Oceanic influence on precipitation in Venezuela, under

Transcription

Oceanic influence on precipitation in
Venezuela, under current and future
climate
Master’s Thesis in Meteorology
Nele Tim
Mentors:
Prof. Dr. Martin Claußen
Prof. Dr. Lelys Bravo de Guenni
Fachbereich Geowissenschaften
Universität Hamburg
Bundesstraße 55
20146 Hamburg
Hamburg, December 2011
Abstract
This is a study on the statistical relation between rainfall in Venezuela and sea surface
temperature.
The oceanic influence on rainfall in Venezuela in general is analysed with Canonical Correlation Analysis (CCA) using the Climate Predictability Tool (CPT), a program developed
for seasonal forecast in the tropics. Station data of precipitation are correlated with the
Extended Reconstructed Sea Surface Temperature (ERSST), v3b data set (1951 - 2010).
In addition, the correlations are analysed in historical (1946 - 2005) and future (2041 2100) simulations performed with the Max Planck Institute for Meteorology Earth System Model MPI-ESM (CMIP5 historical and RCP85 scenario experiments). Four oceanic
regions (north tropical Atlantic (NTA), Niño3 and Niño3.4 and an area which includes
all previous three (Pac-Atl)) are used for the CCAs while precipitation of two regions in
Venezuela is considered: a part of the central coast and one area in the inland. The two
seasons in Venezuela (dry and wet), separated into an early and a late period, and the
months November, December and February, are analysed with the CCAs.
For the observations, the oceanic impact on the precipitation of the station data is, in the
majority of the cases, higher in the inland than at the coast. The Pacific’s influence is
stronger in the early dry season than in the wet seasons, whereas the Atlantic’s influence is
stronger in the wet seasons (inland). In contrast, CCAs applied to the model data provide
highest coefficients in the late wet season with all oceanic regions. In most cases the NTA
has a stronger influence than the Niño regions. The differences between coast and inland
are minimal. In addition to the zero-lag correlations, CCAs have been applied to lagged
sea surface temperatures, with sea surface temperature (SST) leading precipitation by up
to six months.
The statistical connection of extreme precipitation and SST shows that an anomalous
warm NTA and an anomalously cold eastern equatorial Pacific alone are not sufficient to
provoke extreme rainfall in Venezuela. Other meteorological circumstances are necessary
at the same time.
Four extreme events in precipitation in Venezuela are analysed using the NCEP/NCAR
Reanalysis I data. The ’Situación Norte’, the key factor for extreme events during the
dry season, is a meteorological situation when there is an upper-level trough just north of
Venezuela and a cold front located at the northern coast of Venezuela. Additionally, these
events are favoured and intensified by La Niña conditions in the eastern equatorial Pacific,
an NTA SST which is warmer than usual, northeast trades with a strong northern component, low level convection and high level divergence, a high Quasi-Biennial Oscillation
(QBO) and a strong moisture transport towards the coast of Venezuela.
I
Contents
Abstract
I
1 Introduction
1
1.1
Introductory remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2
Settings of the problem
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2 Data and methodology
2.1
2.2
7
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
2.1.1
Station data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
2.1.2
Sea surface temperature . . . . . . . . . . . . . . . . . . . . . . . . .
9
2.1.3
Model data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.4
Data of extreme events . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1
Canonical Correlation Analysis . . . . . . . . . . . . . . . . . . . . . 13
2.2.2
Climate Predictability Tool . . . . . . . . . . . . . . . . . . . . . . . 14
3 Precipitation in Venezuela and the sea surface temperature of the north tropical
Atlantic and the east equatorial Pacific
17
3.1
Characteristics of the precipitation in Venezuela . . . . . . . . . . . . . . . . 17
3.2
Characteristics of the sea surface temperature of the north tropical Atlantic
and the east equatorial Pacific . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3
Extreme precipitation in Venezuela . . . . . . . . . . . . . . . . . . . . . . . 29
III
Contents
4 Canonical Correlation Analysis of the relationship between precipitation in Venezuela and sea surface temperature
4.1
4.2
41
Canonical Correlation Analysis of the precipitation during the seasons . . . 42
4.1.1
Station data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1.2
Model data of the past . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.3
Model data of the future . . . . . . . . . . . . . . . . . . . . . . . . . 59
Canonical Correlation Analysis of the precipitation during November, December and February . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3
4.2.1
Station data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.2
Model data of the past . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.2.3
Model data of the future . . . . . . . . . . . . . . . . . . . . . . . . . 79
Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5 Extreme events in precipitation in Venezuela
87
5.1
Statistic of extreme events in precipitation and sea surface temperature . . . 88
5.2
Meteorological conditions during four extreme events in precipitation . . . . 91
5.3
5.2.1
15th - 17th of February 1951 . . . . . . . . . . . . . . . . . . . . . . 92
5.2.2
14th - 16th of December 1999 . . . . . . . . . . . . . . . . . . . . . . 94
5.2.3
7th - 10th of February 2005 . . . . . . . . . . . . . . . . . . . . . . . 97
5.2.4
24th - 26th of November 2010 . . . . . . . . . . . . . . . . . . . . . . 100
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6 Conclusions
105
Appendix
111
Bibliography
159
Acknowledgements
165
IV
1
Introduction
1.1 Introductory remarks
Venezuela is located in the north of South America, in the tropics between 0.7 ◦ N - 12.2 ◦ N
and 59.8 ◦ W - 73.4 ◦ W. It is a third-world and socialist country. The economy is based on
oil sources in the west of the country and southern part of the country along the Orinoco
river (Faja petrolífera del Orinoco). Most of the necessary products are imported. The
capital Caracas, in the central north of Venezuela, is the financial and industrial centre.
Thirty percent of the population lives there.
The suburbanisation is strong, because of the lack of jobs and the bad living conditions
in the countryside. The huge barrios are vulnerable to damages by rainfall. Extreme precipitation can cause landslides which in turn can cause huge destruction because people
have to live on the hillsides surrounding the city. The biggest catastrophe took place in
December 1999. The effects were huge: 10 000 dead people were reported and the estimated price for reconstruction adds up to 1.8 billion US Dollars (Lyon, 2003). The whole
coast was affected and important transport ways were destroyed. The strong influence
of precipitation, especially of extreme events, on the life and infrastructure of Venezuela
reveals the importance of further investigation.
1
1 Introduction
INAMEH, the Instituto Nacional de Meteorología e Hidrología de Venezuela, forecasts the
weather and conducts some research on parameters which influence precipitation. Furthermore, the Universidad Simón Bolívar (USB) in Caracas is developing an early warning
system in the Vargas State at the north central coast of Venezuela. At CESMa, Centro
de Estadística y Software Matemático, of the USB, a group of researchers and students is
investigating the impact of precipitation, especially in Vargas. This thesis has partly been
developed at CESMa.
1.2 Settings of the problem
The characteristics of the seasons
The geographical position of Venezuela leads to a strong influence of the seasons by the
Intertropical Convergence Zone (ITCZ). Two seasons exist during the year, the dry season
in the boreal summer and the wet season in the boreal winter. The precipitation defines
the seasons with its presence or absence. The duration, start and ending of each season
vary with the years. On average, the wet season spans the months May to October, the
dry season lasts from November until April (Mundaray, 2005).
Hastenrath (1966) described the beginning of the rainy season which is characterised by
the ITCZ lying north of the equator and over Venezuela. This is followed by a convergence
of winds and a moisture transport from the east and the west towards the Caribbean. At
the same time, the anticyclone over the north Atlantic moves northward. This reduces the
pressure gradient between the southern flank of the high and the trough at the equator
and weakens the northeasterly trades.
The easterly waves bring warm moist air to the Caribbean and the north coast of Venezuela. The low level convergence of the trade winds is followed by a strong upward motion,
which favours deep convection (Mundaray, 2005).
As the rainy season proceeds (described by Ashby et al., 2005), the trades from the east
near the surface get weaker and the convergence expands to the mid-troposphere. The
weaker trades decrease the heat loss of the Atlantic. The vertical wind shear, which is
strong at the beginning of the rainy season, is then reduced. Knaff (1997) added that
these three factors, the low surface pressure, the low vertical wind shear and the weaker
east surface trades, forward the convection over the tropical Atlantic. The same study
reveals, that the southward movement of the north Atlantic high, the reinforcement of the
trades and the increment of vertical wind shear define the end of the rainy season.
The weather in Venezuela during the dry season is described by FAV (2000). During the
dry season (November - April) the weather in Venezuela is influenced by the northeasterly
trade winds and an anticyclonic system in the high levels of the troposphere. The ITCZ is
2
located in the south of Venezuela. The weather is characterised by clear skies, sometimes
partly cloudy and with little rain.
Influence of the oceans on precipitation
Different studies about the influence of the oceans on precipitation have been conducted. There is no doubt about a connection between the sea surface temperature (SST) of
the east equatorial Pacific and the north tropical Atlantic (NTA) and precipitation in the
Caribbean and Venezuela.
Impact of the north tropical Atlantic on precipitation in Venezuela:
A positive sea surface temperature anomaly in the north tropical Atlantic causes warmer
and more humid air over the NTA itself, upwind of Venezuela. This leads to decreased
stability and the development of convection, especially with temperatures over 26.5 ◦ C
(Gray, 1968). The warm and humid air reaches the coast of Venezuela with the northeast
trade winds. Previous studies agree that a positive anomaly over the north tropical Atlantic favours strong precipitation at the northern coast of Venezuela and in the Caribbean
(Enfield, 1996; Guenni et al., unpublished manuscript; Chen et al., 2002; Taylor et al.,
2002; Martelo, 2003,I).
The effect of El Niño-Southern Oscillation (ENSO) on Venezuelan rainfall:
Sequera (2009) found out that El Niño is one of the most important mechanisms for the
climate variability in tropical South America. Previous studies came to a consistent result:
El Niño produces a deficit of rain and La Niña increases the amount of rain. During El
Niño, warm ENSO, the ITCZ is located more in the southwest, followed by a southward
motion of the high pressure systems. An anomalous Hadley cell, located over the north of
South America, blocks the upward motion and convection over northern South America
and leads to less precipitation (Poveda et al., 1997). The flow from the south to the
Caribbean is strengthened and the moisture transport from the Atlantic to the Pacific is
weakened. This is associated with decreased precipitation over northern South America
(Mestas-Nuñez et al., 2007). There is not only less precipitation during El Niño, but also
tropical disturbance, hurricanes and storms appear less frequently (Cárdenas et al., 2002).
Cárdenas et al. (2002, 2003) determined that La Niña has the same or even a stronger
influence on precipitation in Venezuela than El Niño and that it favours heavy rain at the
coast of Venezuela. Additionally, Lyon (2002) pointed out that a change from a warm to
a cold ENSO phase causes stronger seasonal rainfall (June - August).
The influence of ENSO varies during the seasons. A developing La Niña produces heavy
late wet season rainfall (August - October) in the Caribbean (Taylor et al., 2002). Cárdenas
et al. (2002) pointed out that the influence of ENSO is stronger during the dry than during
the wet season. Martelo (2003,II) added the reason for it: the gradients in temperature,
3
1 Introduction
energy and wind are stronger. La Niña is linked with a wetter dry season and El Niño
with a drier wet season (Martelo, 2003,II).
Signs of the sea surface temperature anomalies and time until the response of the precipitation:
There is broad agreement about the constellation of temperatures in the north tropical
Atlantic and east equatorial Pacific that favours wet and dry conditions in the Caribbean
and at northern coast of Venezuela. Positive rainfall anomalies are associated with a cold
east equatorial Pacific and a warm north tropical Atlantic. Enfield et al. (1999) added that
the strongest rainfall response occurs with a meridional dipole of sea surface temperature
anomalies in the tropical Atlantic and when the eastern tropical Pacific and NTA SST
anomalies are of opposite sign. An interruption of the common sign of the SST anomalies
and its precipitation response was mentioned by Martelo (2003,I): some dryer Mays occur
during La Niña conditions. Taylor et al. (2002) agreed with that and add that a dry
early rainy season in the Caribbean is connected with La Niña conditions in the early dry
season. Also, Chen et al. (2002) suggested that a wet early wet season is linked to a warm
equatorial Pacific the winter before but with a warm NTA the spring before. This points
out that the time until the precipitation responds to the SST differs with the studies.
Martelo (2003,I) revealed that NTA SSTs influence precipitation in Venezuela with a lead
time of zero to two months, Niño3 SST with up to four, and Niño3.4 SST with up to three
months.
Atmospheric bridge:
The zonal SST gradient between the Pacific and the Atlantic influences the strength of the
trade winds over the Atlantic and in the Caribbean Low Level Jet (a maximum of easterly
winds in 950 hPa over the Caribbean (Wang, 2007)). The sea surface temperature gradient
over both oceans and the resulting trade winds are the bridge connecting the oceans and
have a positive feedback with the Walker Circulation over the Pacific basin (Wang et al.,
2009). The signal of ENSO reaches the north tropical Atlantic. 50-80% of the anomalous
SST variability of the tropical Atlantic is associated with the Pacific ENSO (Enfield et al.,
1997). The Pacific is ahead of the Atlantic. Enfield et al. (1997) concluded that a warming
in the Atlantic occurs four to five months after a heating in the Pacific.
Seasonal difference between the impact of the Pacific and the Atlantic:
The seasonal difference in the influence of the Pacific and the Atlantic has been analysed
as well. Taylor et al. (2002) pointed out, that the influence of the tropical Pacific on
Caribbean rainfall is stronger in the late than in the early wet season, while the impact
of the tropical Atlantic is more intense in the early wet season but has an effect on the
whole rainy season. The same investigation detects that the early wet season is modulated
directly by the SST, while the late rainy season is effected mainly by vertical wind shear
4
modification of convective conditions. Enfield et al. (1999) figured out, that the wet season
is more strongly influenced by the NTA than by the tropical eastern Pacific.
Martelo (2003,I) and Sequera (2009) additionally investigated the difference between the
wet season and the dry season and came to the results that the north tropical Atlantic
influences more the wet season while the regions Niño3 and Niño3.4 have a stronger impact
on the late wet and dry season.
Geographical differences in Venezuela:
Precipitation in Venezuela is highly influenced by the topography. The country has a high
diversity of vegetation zones and climates. There is the tropical coast at the Caribbean
Sea in the north, the Andes in the west, the plains in the centre and the rainforest in
the south. Cárdenas et al. (2003) determined almost always higher correlations between
SST in the Pacific regions and rainfall in Bolívar (located in the south east of Venezuela)
than with precipitation in Vargas (at the north central coast). Martelo (2003,I) discovered
that northern Venezuela is more influenced by the Atlantic, eastern Venezuela more by the
Pacific.
This thesis will investigate the influence of the north tropical Atlantic and the eastern
equatorial Pacific on precipitation in Venezuela. Earlier studies show that there is definitely a connection between the SST of both oceans and the precipitation in the Caribbean
and northern South America. This investigation will analyse which ocean influences more
strongly temporally, during the different seasons (early/late wet and early/late dry), and
spatially, at the coast and in the inland. It is interesting to find out if there is a geographical difference in oceanic influence as well. Previous studies do not have concurrent results
about the impact on precipitation in the four seasons. Additionally, from this point of view,
the dry season (early and late) and the spatial variations have rarely been investigated.
Most of the mentioned articles are studies about the Caribbean rainfall and not about the
precipitation in Venezuela.
The data and methods used for this study will be presented in the following chapter 2. The
characteristics of precipitation in Venezuela, the sea surface temperatures of the north tropical Atlantic and the east equatorial Pacific and extreme precipitation will be displayed
in chapter 3. The influence of the SST in the NTA and the Niño regions 3 and 3.4 on
the precipitation in Venezuela will be investigated with Canonical Correlation Analyses in
chapter 4. In chapter 5 the connection between extreme precipitation and extreme SST
and four extreme events in precipitation in Venezuela will be analysed.
5
2
Data and methodology
2.1 Data
The data used for the analyses are presented in this chapter. The precipitation data described in section 2.1.1 are measured at stations in Venezuela and are used for the Canonical
Correlation Analyses in chapter 4. The sea surface temperature data of paragraph 2.1.2 are
the counterpart to the station data for the correlations. Canonical Correlation Analyses
are additionally applied to model data, which are presented in 2.1.3. The description of the
characteristics of extreme precipitation in Venezuela (section 3.3) and a statistical analysis
on the connection between extreme events in precipitation and sea surface temperature
(section 5.1) are based on the data described in sections 2.1.1, 2.1.2 and 2.1.3. The statistical measures of the months of extreme precipitation (section 3.3) are calculated with the
station data described in section 2.1.4. The extreme events in precipitation in section 5.2
are analysed with the model data described in section 2.1.4.
7
2.1.1 Station data
The station data include 127 stations, 36 in the state of Vargas and 91 in the state of
Bolívar (see figure 2.1). The measurements at the stations in Vargas are conducted by
the Dirección de Hidrologia y Meteorología, Ministerio del Ambiente y Recursos Naturales
(MARN), the data in Bolívar are collected by the Base de datos Estaciones Cuenca del
Rio CARONI, Electrificación del Caroní (EDELCA), with the exception of three stations
(MARN). Further information is listed in tables A.1 - A.5 (appendix).
Vargas is a state on the central coast of Venezuela, north and northeast of Caracas, the
capital of Venezuela. The precipitation in this state is strongly influenced by its topography.
To the north is the Caribbean Sea, to the south the Avila, a cordillera. The state of Bolívar
is located in southeastern Venezuela. The Rio Orinoco forms the northern, Guyana the
eastern, Brazil the southern and the Amazons and the Llanos the western border of Bolivar.
These two regions were selected to represent the situation at the coast and in the inland.
These data are used for the Canonical Correlation Analyses in chapter 4. Seasonal means
and single months are correlated. In both cases the monthly values are accumulated values
of the particular month. The four seasons are as following: early dry season (November
- January), late dry season (February - April), early wet season (May - July) and late
dry season (August - October). Before calculating the seasonal means, missing values
are replaced by long term means of the corresponding month. The stations all measured
during different time periods. That makes it very difficult to specify the period for the
investigation. To have the same period in both states, to compare the regions with each
other, and include as many as possible of the four events discussed in section 5.2, all sixty
years between 1951 - 2010 are chosen.
Figure 2.1: Map of the stations (black dots), the precipitation regions of the model data (black
boxes: coast and inland) and the oceanic regions, pink box: north tropical Atlantic, turquoise box:
Pac-Atl, red: Niño3.4, blue area: Niño3
8
2.1 Data
2.1.2 Sea surface temperature
The Canonical Correlation Analyses in section 4.1.1 and 4.2.1 are done with the following
sea surface temperature (SST) data together with the station data, described in the previous paragraph.
The most recent version of the Extended Reconstructed Sea Surface Temperature (ERSST),
v3b, is used as SST data for the correlations. The analysis is based on the International
Comprehensive Ocean-Atmosphere Data Set (ICOADS) release 2.4. At the end of every
month, the ERSST analysis is updated with the available Global Telecommunication System (GTS) ship and buoy data for that month. The anomalies are computed with respect
to the 1971 - 2000 climatology (Xue et al., 2003). The data have a 2
◦
x2
◦
resolution.
For the Atlantic, a region called north tropical Atlantic (NTA) is chosen. It spans the area
6.0 ◦ N - 18.0 ◦ N, 60.0 ◦ W - 10.0◦ W (see figure 2.1). This area is defined after Penland et
al. (1998) and is used by NOAA (National Oceanic and Atmospheric Administration) for
calculating the NTA index. In addition, an unpublished study by Bravo de Guenni, one of
the supervisors of this thesis, was based on this region too. All previous studies split the
area in two parts: 6.0 ◦ N - 18.0 ◦ N, 60.0 ◦ W - 20.0 ◦ W and 6.0 ◦ N - 10.0 ◦ N, 20.0 ◦ W 10.0 ◦ W, probably to avoid having parts of the African continent in the chosen window.
Here a rectangular region is used and values over land are masked.
In the equatorial Pacific, the investigated area is limited to the western hemisphere. To
find out which region has the strongest influence on precipitation in Venezuela, Canonical
Correlation Analyses are performed for the months November, December and February.
All Niño regions in the east equatorial Pacific are correlated with the precipitation in Vargas and Bolívar. The regions are: Niño1.2 (0
◦
- 10 ◦ S, 90 ◦ W - 80 ◦ W), Niño3 (6 ◦ S -
6 ◦ N, 150 ◦ W - 90 ◦ W) and Niño3.4 (6 ◦ S - 6 ◦ N, 170 ◦ W - 120 ◦ W) The regions Niño3 and
Niño3.4 are officially defined as the region between 5 ◦ S - 5 ◦ N. In this study, the areas are
chosen to be one degree larger because of the spatial 2
◦
resolution of the data.
Table 2.1: Correlations between sea surface temperature (Niño1.2, Niño3, Niño3.4) and precipitation (Vargas, Bolívar) to choose the best region in the western equatorial Pacific
Region
Niño1.2
Niño3
Niño3.4
Niño1.2
Niño3
Niño3.4
Vargas, Nov.
0.40
0.43
0.49
Bolívar, Nov.
0.38
0.45
0.52
Vargas, Dec.
0.29
0.35
0.33
Bolívar, Dec.
0.35
0.57
0.56
Average of Vargas and Bolívar
Vargas, Feb.
0.52
0.39
0.43
Bolívar, Feb.
0.63
0.68
0.38
Niño1.2 = 0.43
Vargas, average of Nov., Dec., Feb.
0.40
0.39
0.42
Bolívar, average of Nov., Dec., Feb.
0.45
0.57
0.49
Niño3 = 0.48
Niño3.4 = 0.45
9
In Vargas, the correlation coefficient between precipitation and SST (averaged over these
three months) is very similar for all three western equatorial Pacific regions. Table 2.1
shows that Niño3.4 has the highest influence on precipitation in Vargas with a coefficient of
0.42. Niño1.2 has nearly the same averaged correlation coefficient over these three months
(0.40). Even the difference to Niño3 is quite small with 0.39. In Bolívar, the coefficient
of Niño1.2 is smaller, with 0.45, compared to the other two regions. Niño3 has by far the
highest correlation averaged over these three months with 0.57 followed by Niño3.4 with
0.49. The averaged correlation coefficients over both regions in Venezuela have the same
order like in Bolívar. Niño3 leads with 0.48, followed by Niño3.4 with 0.45 and by Niño1.2
with 0.43. The results of Niño3 and Niño3.4 seem to be very similar as mentioned before
by Cárdenas et al (2003) and Martelo (2003,I). They have a higher influence than Niño1.2,
in Bolívar and in total. According to this table the regions Niño3 and Niño3.4 are chosen
to be investigated in more detail in this study. Furthermore, a fourth area will be used
which includes all three regions (NTA, Niño3 and Niño3.4) called Pac-Atl (6 ◦ S - 18 ◦ N,
150 ◦ W - 10 ◦ W).
2.1.3 Model data
In addition to the station data, Canonical Correlation Analyses (CCAs) are applied to
data from climate model simulations performed with MPI-ESM, the Earth System Model
(ESM) of the Max Planck Institute for Meteorology (MPIM). MPI-ESM is a coupled model consisting of four main components: the atmospheric model ECHAM6 (Röckner et al.,
2003), the land surface and vegetation model JSBACH (Raddatz et al., 2007), the ocean
and sea ice model MPIOM (Marsland et al., 2003), and the ocean biogeochemistry model
HAMOCC (Maier-Reimer et al., 2005). The sixth-generation atmospheric general circulation model ECHAM6 is the most recent version in a series of ECHAM models evolving
originally from the spectral weather prediction model of the European Centre for Medium
Range Weather Forecasts (ECMWF) (Simmons et al., 1989). In this study, data from two
simulations are used that haven been performed with MPI-ESM within the Coupled Model
Intercomparison Project, Phase5 (CMIP5, http://cmip-pcmdi.llnl.gov/cmip5): a historical simulation (experiment id MPI-ESM-LR_historical_r1i1p1) from 1850 to 2005 using
observed forcings for atmospheric greenhouse gas concentrations (among others), and a
future simulation (experiment id MPI-ESM-LR_rcp85_r1i1p1) from 2006 to 2100 using
a scenario of relatively high greenhouse gas concentrations that would lead to a radiative
forcing of about 85W/m2 by the year 2100. The sea surface temperature and precipitation
form these simulations are monthly values in T63 horizontal (approx. 1.9
◦
x 1.9◦ ) and
L47 vertical resolution (top level at 10 hPa), where the numbers stand for the wavelength
at which it is truncated triangularly. Hagemann et al. (2006) found out, that the validation of the modelled precipitation (here ECHAM5) provides much better results for
10
2.1 Data
the precipitation over the oceans. The same investigation mentions that higher horizontal
resolution (T42 to T159) does not have a strong effect on the precipitation as increased
vertical resolution (L19 to L31). The relatively coarse horizontal resolution should not lead
to large errors in the precipitation amount. Hagemann et al. (2006) analysed as well the
precipitation response to ENSO. The model is able to reproduce most of the precipitation
which is connected with ENSO, even though the precipitation might be higher than observed.
The precipitation areas of the model data are chosen to represent Vargas and Bolívar as
well as possible. Vargas is represented by the coastal area of three grid boxes at 10.3 ◦ N
and between 67.5 ◦ W - 63.8 ◦ W (see figure 2.1). Table 2.2 shows the orography of the
model in the boxes and the corresponding latitude and longitude. In the western grid box
the elevation is highest with 200 m, decreasing to the box in the centre with 120 m and
further down to 100 m at the eastern boundary of the chosen area.
Table 2.2: ECHAM6 orography of area at the coast
Latitude/Longitude
10.3
-67.5
200 m
-65.6
120 m
-63.8
100 m
The region which represents the inland and the state Bolívar of the station data is a larger
area with 3x3 grid boxes, between 4.7 ◦ N - 8.4 ◦ N and 63.8 ◦ W - 60.0 ◦ W. Its orography
(table 2.3) varies more. The altitude is lowest in the northeast grid box is near the delta of
the Orinoco (100 m) and highest in the southwest with 780 m which stands for the table
mountains in this part of the country.
Table 2.3: ECHAM6 orography of the area in the inland
Latitude/Longitude
8.394
6.528
4.663
-63.750
280 m
620 m
780 m
-61.875
200 m
490 m
630 m
-60.000
100 m
310 m
420 m
The oceanic regions of the model data span the following areas. NTA: 6.5 ◦ N - 17.7 ◦ N,
60.0 ◦ W - 11.3 ◦ W, Niño3: 4.7 ◦ S - 4.7 ◦ N, 150.0 ◦ W - 90.0 ◦ W, Niño3.4: 4.7 ◦ S - 4.7 ◦ N,
168.8 ◦ W - 120.0 ◦ W, Pac-Atl: 4.7 ◦ S - 17.7 ◦ N, 168.8 ◦ W - 11.3 ◦ W.
2.1.4 Data of extreme events
Data of two meteorological stations are used to calculate the mean, maximum, minimum
and standard deviation during the extreme events and the monthly and annual mean to
reveal the extremeness of the events in section 3.3. The station Maiquetia is operated by
the Fuerza Aérea Venezolana (FAV) and is located at 10.6 ◦ N and 67.0 ◦ W (Cárdenas et
11
al., 2003). La Armada Bolivariana controls the station Cagigal which has the coordinates
10.5 ◦ N and 66.9 ◦ W and is located at an absolute altitude of 1035 m.
The four extreme events in section 5.2 are studied with the NCEP/NCAR Reanalysis I
data. These data are released by the National Oceanic and Atmospheric Administration of
the United States (NOAA), by the department NCEP (National Center for Environmental
Prediction) and NCAR (National Center of Atmospheric Research). The NCEP/NCAR
Reanalysis I data are produced by using a state-of-the-art analysis/forecast system which
assimilates the data by using past data from 1948 to present. The data are available since
1949, globally, monthly, daily and 4 times daily (00 Z, 06 Z, 12 Z, and 18 Z). The 6-hourly
data is used here. The anomalies are the mean of the chosen time period of the event minus
the long term mean, the climatology, of the years 1981 - 2010. In addition to this, synoptic
maps were downloaded from the Service Records Retention System (SRRS) Analysis and
Forecast Charts National of the Climatic Data Center, taken from FAV (2000) and from the
TROPICAL PREDICTION CENTER MIAMI FLORIDA, after Mundaray (2005). The
indices of El Niño-Southern Oscillation (ENSO), Southern Oscillation Index (SOI) and the
Quasi-Biennial Oscillation (QBO) were downloaded from NOAA. For ENSO three-month
running means of the Niño3.4 region are used. The season is called warm or cold when a
minimum of five consecutive over-lapping seasons have an anomaly of at least +/-0.5 ◦ C.
Monthly values are used for the calculations of the SOI (sea level pressure anomalies) and
of the QBO (averaged 30 hPa zonal wind at the equator).
2.2 Methodology
The characteristics of precipitation of the four seasons in Venezuela are analysed with the
station data and the model data (section 3.1). The time periods of the model data are
chosen as 1946 - 2005 and 2046 - 2100. Thereby, both periods span sixty years like the
station data,the model data of the past have a similar time period as the station data and
the data of the future contain the furthest possible future.
The sea surface temperature anomalies are displayed in histograms in section 3.2. The anomalies of the model data are calculated with respect to the climatology of 1966 - 1995 and
2061 - 2090. The trend is subtracted from the SST anomalies of ERSST and ECHAM6.
Extreme precipitation is analysed in section (3.3). The major extreme precipitation events
took place during November, December and February. These three months are analysed by
histograms and statistical summary measures. Furthermore, the extremeness of the events
is shown with the aid of two stations which measured continuously the last sixty years.
In chapter 4, the relationship between precipitation and sea surface temperature is investigated with Canonical Correlation Analysis (CCA, for further details see the following
section). Canonical correlations are chosen because it is a good method to analyse the
12
2.2 Methodology
relationship of multivariate time series. CCA is a linear statistical method. The precipitation of the four seasons in the states Vargas (coast) and Bolívar (inland) will be canonically
correlated with the sea surface temperature of the regions Niño3 and Niño3.4 in the equatorial Pacific, the NTA and Pac-Atl. The dry and the wet season are split into an early and
a late part because the precipitation even varies during these seasons. Each of these four
seasons (early wet, late wet, early dry and late dry) contains three months. The monthly
values are averaged over these three months. The CCAs are performed with zero-lagged
SST and precipitation and with lags up to six months, with SST leading precipitation.
Each lag is shifted one month backwards with respect to the previous one. That means
that the first two lags still contain at least one month of the corresponding season. The
spatial loadings are shown as correlations of the original data with the CCA time series.
The spatial loadings present the typical patterns of the anomalies of SST and precipitation which develop together. The CCA coefficients and spatial loading of the first CCA
mode which describes a physically meaningful pattern are described. In section 4.2 CCAs
are performed with the three months during which the four major extreme events took
place (November, December and February). The precipitation of the four extreme events
(section5.2) does account for only up to 4.6% of the precipitation of one season or one
month within the whole time period of the station data. It is therefore reasonable to use
a linear method like CCA to analyse these months.
2.2.1 Canonical Correlation Analysis
Canonical correlations are a multiple regression analysis to characterise the relationship
between two sets of variables, each set has more than one member (Everitt, 2005). It was
originally developed by Hotelling in 1936. The description is based on Everitt (2005) and
Zorita et al. (1992). The Canonical Correlation Analysis is about finding linear functions
of the sets of variables to maximise the correlation among them.
There are two sets of variables: x = (x1 , ..., xn ) and y = (y1 , ..., ym )
The CCA finds two new sets of variables Ui and Vj . Ui and Vj , the canonical correlation
time series, are linear combinations of x and y, Ui = αi x and Vj = βj y and uncorrelated.
The problem is reduced by an eigenproblem. The coupled eigenproblem is defined as:
Cxx −1 Cxy Cyy −1 Cxy T α = λ2 α
Cyy −1 Cxy T Cxx −1 Cxy β = λ2 β
The eigenvalues λ2 are the same in both cases, α and β are the eigenvectors or weights.
Cxx and Cyy are the autocovariance matrices and Cxy and Cxy the cross-covariance matrices. The eigenvectors are chosen in such a way that the correlation between U1 =α1 x
and V1 =β 1 y is maximised.
The size of the covariance matrix can be reduced via an empirical orthogonal function
13
(EOF) analysis, previous to CCA. Limiting the number of EOFs (modes) from n, m to i, j,
the size of the covariance matrix is reduced from (n x m) to (i x j). The number of modes
are chosen in a way that the remaining modes explain most of the total variance. This
filtering is done to eliminate noise but can also ignore useful data.
The new variables which enter the Canonical Correlation are the Principal Components
(PCs), the projected eigenvectors. Therefore, the calculation of the autocorrelation matrices Cxx and Cyy is more facile because they are diagonal to their inverse. The canonical
correlation time series Ui and Vj are calculated as linear combinations of the PCs. The
patterns of the canonical correlation can be obtained from the original variables by averaging in time:
gi = Cxx αi = hUi xi
hj = Cyy β j = hUj yi
g and h, the spatial loadings, can be seen as the local covariance between the variables x
and y and the canonical correlation time series and can be considered as the indicator of
the strength of the signal.
2.2.2 Climate Predictability Tool
Climate Predictability Tool (CPT) is a program for performing Canonical Correlation Analysis and used in this study. The following description is based on the CPT user guide.
CPT was developed by the International Research Institute (IRI) for Climate and Society
of the University of Columbia, USA. CPT is a powerful tool to forecast seasonal climate
in tropical and sub-tropical areas.
The CPT is prefiltering the input data with an EOF analysis. The maximum number of
modes can be specified. The Principal Components are calculated by using the correlation matrix (the analysis is based on the standardised anomalies). By constructing the
model, a cross-validated forecast is made. The cross-validated forecast is a prediction of
the hindcasts. The best cross-validated result is defined by the highest goodness index,
calculated with the Pearson correlation coefficient. The goodness index can be seen as an
average correlation between the transformed cross-validated forecasts and observations for
all series. The correlations are first transformed to the Fisher z-scale, to use a transform
which is more normally distributed, averaged, and then transformed back to the correlation
scale. The number of used modes is chosen according to this goodness index.
The predictand, the rainfall data, is transformed to a normal distribution before the calculations because, as can be seen in the statistics of the station data 3.1, the distribution
is skew. CCA generally works best with normally distributed data. The empirical distribution is the basis for the transformation. The precipitation data are transformed to
an uniform distribution on the unit interval. Standard normal distribution deviates are
14
2.2 Methodology
calculated using these percentiles. The process is done backwards by converting to the
percentiles and linearly interpolating them on the original data. The goodness index and
the spatial loadings apply to the transformed data. Furthermore, the precipitation data
are zero-bounded so that negative values are not predicted.
Settings of CPT
The length of the cross validation window is chosen as five. It defines how many years are
left out by computing the cross-validation. The number of years has to be odd, because
just the year in the middle of the cross-validation window is predicted at each step.
The time periods are 1951 - 2010 (station data), 1946 - 2005 (model data of the past) and
2041 - 2100 (model data of the future). The length of training period defines the number
of years used for constructing the model and is chosen as the first fifty years of the time
periods. The remaining ten years are the forecast period. Starting years of both input files
(predictor and predictand) are the same. When computing CCAs with a lag of SST leading
the precipitation and the lag crosses the calendar year, the time period of the predictand
starts one year later and the training period infolds only forty-nine years. Three-month
means of November - January and December - February cross the calendar year and refer
to the year of November and December, thus the data contains just fifty-nine years.
Missing values are replaced by the median of each station. The maximum percentage of
missing values per stations and time step is 50%. It has to be such a high limit because of
the problematic station data. With a lower value, the number of used stations would be
too small to make a reasonable analysis. The replacement is done using the untransformed
data.
The connection between extreme events in precipitation and SST is proved in 5.1 by plotting the time series of both variables. The anomalies of precipitation and SST are used,
calculated by subtracting the climatologies (station data: 1971 - 2000, model data: 1966 1995 and 2061 - 2090) from the time periods (the same as used for the CCAs). The SST
anomalies are detrended. The threshold of extremeness is the standard deviation of the
anomalies of the corresponding variable over time period.
The meteorological conditions during four extreme events in the past are analysed in section 5.2. Plots of different parameters are used to find similarities in the conditions which
cause the events. The parameters are studied in the three or four days of the peaks of the
events.
15
3
Precipitation in Venezuela and the sea surface
temperature of the north tropical Atlantic
and the east equatorial Pacific
In this chapter properties of the precipitation during the four seasons are displayed and
analysed. Beside that, the sea surface temperature anomalies of the four oceanic regions
are shown and discussed. In addition to the precipitation in general, this study investigates
extreme precipitation. For this reason, the three months of the year when the last major
extreme events took place are studied by looking at the characteristics and statistical
measures of the precipitation during these months. The values, mention in context with
the histograms are approximate values.
3.1 Characteristics of the precipitation in Venezuela
The histograms of the precipitation of the station data during the four seasons in Vargas
are displayed in figure 3.1. As expected, there are higher values of precipitation in the wet
seasons and the most cases have between 35 mm - 100 mm (early wet) and 50 mm 150 mm (late wet). More events of strong rain and also more light precipitation occur in
the late wet season than at their beginning. During the early dry season it rains more
strongly than in the late dry season.
17
The figure 3.2 shows the statistical measures of the four seasons in Vargas with the station
data. Most of the stations show the expected order after looking at the histograms. Highest
precipitation in the late wet season, followed by the early wet season, early dry and late
dry. In contrast, some stations show the highest mean and median in the early dry season.
The minima of the late dry season are again the lowest compared to all other seasons.
(a)
(b)
(c)
(d)
Figure 3.1: Histograms of the precipitation of the four seasons in Vargas (station data), early
wet (a), late wet (b), early dry (c), late dry (d)
At most stations the standard deviation is highest during the dry season which can be
explained by normally few precipitation combined with heavy rain events. The stations
are ordered from the west to the east of Vargas. It can be seen, that the precipitation
increases in the eastern part of the state. The minima show it only in the early dry season.
When ordering the stations according to the latitude (not shown) it becomes clear that in
the north of Vargas, closer to the sea, the precipitation amounts are higher in both dry
seasons (except for the minima).
18
(a)
(b)
(c)
(d)
(e)
Figure 3.2: Mean (a), median (b), maximum (c), minimum (d), standard deviation (e) of the
precipitation of the four seasons in Vargas (station data)
The figure 3.3 displays the histograms of the mean precipitation per season of the station
data in Bolívar. It can be seen, that there is more precipitation in Bolívar than in Vargas.
The difference in precipitation between the two parts of the rainy season is minimal. Events
with 120 mm - 480 mm and 108 mm - 402 mm of precipitation occur most often. The
19
(a)
(b)
(c)
(d)
Figure 3.3: Histograms of the precipitation of the four seasons in Bolívar (station data), early
wet (a), late wet (b), early dry (c), late dry (d)
spectra are much wider than in Vargas, values up to 1180 mm were measured. This
again shows that it rains much more in this region than in Vargas. During the early dry
season, the bar with the highest frequency (700) is surprisingly the bar which represents
the precipitation amounts between 80 mm - 160 mm. In contrast to this, in the late dry
season light rain events up to 60 mm occur most often (in 800 cases).
The statistical parameters of the precipitation in Bolívar (figure 3.4) have a more clear
structure than in Vargas. Means, medians, maxima and minima present the same order
of the season when compared to each other. The early wet season always has the highest
values, closely followed by the late rainy season. In Vargas, it was the other way round.
The order of the dry seasons is the same as at the coast. More rain at the beginning of
the dry season than during its second part. The standard deviation has similar values in
all four seasons but here the values are higher during the wet than during the dry seasons.
This result is contrary to the one in Vargas. An explanation can be the lack of heavy
rain events affecting the inland of Venezuela. The stations in Bolívar are ordered from the
20
(a)
(b)
(c)
(d)
(e)
precipitation of the four seasons in Bolívar (station data)
south to the north. It can be seen, that the statistical measures have their lowest values
in the north of the state and the highest values during both wet seasons in the middle of
Bolívar. When ordering the stations from the west to the east, the mean and the median
of the wet seasons are higher in the west.
21
The figure 3.5 shows the histograms of the precipitation of the four seasons at the coast
with the model data of the past. In general, the modelled precipitations are less than
the measured ones. The highest values in precipitation take place during the first half of
the rainy season with maximum values of 320 mm. Amounts between 0 mm and 20 mm
never occur, values between 50 mm and 80 mm the most often. In the late wet season
the precipitation amounts are lower and the distribution is wider with a high number of
values between 3 mm and 87 mm. The precipitation in the dry seasons is mostly near
zero. In the early dry season, values up to 45 mm occur more than 30 times in this period.
Surprisingly, values of zero were never modelled. In the second half of the dry season,
more than 120 of the 180 values are between 0 mm and 10 mm. The maximum values are
smallest during this season. The histograms of the four seasons at the coast of the model
data are quite different than the ones of the station data where highest values of more than
450 mm occur during the late wet and early dry season.
(a)
(b)
(c)
(d)
Figure 3.5: Histograms of the precipitation of the four seasons at the coast (model data of the
past), early wet (a), late wet (b), early dry (c), late dry (d)
22
(a)
(b)
(c)
(d)
Figure 3.6: Histograms of the precipitation of the four seasons in the inland (model data of the
past), early wet (a), late wet (b), early dry (c), late dry (d)
In the inland, the spectrum of precipitation amounts does not differ within the four seasons
(figure 3.6). Only the early wet season shows a higher maximum with precipitation up to
250 mm/months. During the three other seasons the maximum is around 150 mm. Values
of zero or near zero occur most frequently and the occurrence get stepwise less with higher
precipitation amounts. But in the early wet season rainfall amounts between 115 mm
185 mm take place most frequently. The values of the model are much smaller than the
ones of the station data. The distribution differs strongly for the late wet season, where the
station data show a similar pattern like in the early wet season. The differences between
coast and inland are much smaller with the modelled precipitation than with the measured
one.
23
Also the histograms of the four seasons of the future at the coast show a trend to less
rain in the model data. The figure 3.7 displays the development, especially for the late
wet season. The historical model data have no values of zero and the most frequent values
are up to 87 mm while the histogram for the future data has its highest occurrence in the
first two bars with maximum amounts of 42 mm. The late dry season represents a drastic
reduction of the rainfall amount from the past to the future and compared to the early dry
season.
(a)
(b)
(c)
(d)
Figure 3.7: Histograms of the four seasons at the coast (model data of the future), early wet (a),
late wet (b), early dry (c), late dry (d)
The inland precipitation of the model data of the future is shown in figure 3.8. The precipitation during the early wet season has nearly the same frequency for amounts between
0 mm and 200 mm while the data for the past were roughly normally distributed. The
spectrum of the late wet and early dry season is very similar to the one of the past but
the concentration on small precipitation amounts is again stronger. The second half of the
dry season has much lower rain in the future than in the past with a maximum of around
65 mm.
24
(a)
(b)
(c)
(d)
Figure 3.8: Histograms of the four seasons in the inland (model data of the future), early wet
(a), late wet (b), early dry (c) late dry (d)
To sum up, the precipitation data of the model does not show the differences between
coast and inland as can be seen with the station data. One reason for this seems to be
the orography. The model orography do not change strongly between these areas. At
the coast, the orography is from 100 m up to 200 m. These low elevations can even be
found in the inland where the height differs between 100 m and 780 m. Furthermore, the
modelled precipitation is less than the measured one at the stations. Both regions show,
as expected, more rain in the wet than in the dry season. Always the late dry season is
the driest and gets drier in the future, the early wet season is always the wettest in the
inland. At the coast, it rains less in the early wet season than the late wet season. The
statistical measures of Vargas and Bolívar show a strong contrast. In Vargas, they have
higher values in the east, in Bolívar the mean and median of the wet seasons have their
maximum in the west. Bolívar show a stronger north-south contrast (with minima in the
north) than east-west and compared to the north-south contrast in Vargas. Unlike Vargas,
where the highest values occur in the north.
25
3.2 Characteristics of the sea surface temperature of the
north tropical Atlantic and the east equatorial Pacific
(a)
(b)
(c)
(d)
Figure 3.9: Histograms of the sea surface temperature anomalies (ERSST), NTA (a), Niño3 (b),
Niño3.4 (c), Pac-Atl (d)
The figure 3.9 shows the histograms of the SST anomalies of the two regions in the Pacific
Ocean, Niño3 and Niño3.4, the north tropical Atlantic and Pac-Atl of the ERSST data.
NTA has more positive anomalies than negative. Niño3 has slightly more negative values,
Niño3.4 as well but nearly balanced. The spectrum of values is in both of them wider than
in the NTA. Pac-Atl has more positive anomalies than negative ones. The Niño regions
and Pac-Atl have higher positive anomalies than negative with the strongest contrast in
graph (d) with a range of values between -4 K and +6 K.
26
3.2 Characteristics of the sea surface temperature of the north tropical Atlantic and the
east equatorial Pacific
The histograms of the modelled sea surface temperature anomalies of the past are shown in
figure 3.10. All four regions show a distribution concentrated around zero and a spectrum
from around -6 K to 5 K. The maximum negative values are always higher than the positive
ones. Additionally, slightly more negative anomalies occur. Compared to the ERSST data,
the spectrums are wider, especially for the north tropical Atlantic.
(a)
(b)
(c)
(d)
Figure 3.10: Histograms of the sea surface temperature anomalies (ECHAM6, past), NTA (a)
Niño3 (b) Niño3.4 (c) Pac-Atl (d)
27
The figure 3.11 shows the histograms of the four regions of SST anomalies for the future.
The NTA has more negative values than positive ones. The Niño3 region as well have
more negative anomalies than positive ones, contrary to the historical data. This indicates
more La Niña conditions in the future than in the past. The Niño3.4 has quite a mirrored
distribution, as well as Pac-Atl. There are higher maxima than in the past, except for the
NTA.
(a)
(b)
(c)
(d)
Figure 3.11: Histograms of the sea surface temperature anomalies (ECHAM6, future), NTA (a),
Niño3 (b), Niño3.4 (c), Pac-Atl (d)
In summary, the anomalies of the SST seem to be more skew with the ERSST and the
spectrums are more narrow. The anomalies are stronger in the future than in the past
(ECHAM6 and ERSST) for the Niño regions and Pac-Atl. The extreme events La Niña
and El Niño become more extreme in the future.
28
3.3 Extreme precipitation in Venezuela
Extreme events in precipitation in Venezuela occurred mainly in the months November,
December and February. For this reason, the characteristics of the precipitation during
these months are displayed and analysed.
(a)
(b)
(c)
Figure 3.12: Histograms of the precipitation in Vargas (station data) in November (a), December
(b), February (c)
The figure 3.12 shows more rain in November and December than in February. The definition of dry and wet season is based on the average, so that in some years the rainy season
can last also during November. But it could be an indicator for more heavy rain events
in the beginning of the dry season as well. Moreover, there is a similar distribution of
the amount of the different strength of precipitation in November and December, while in
February low values (0 - 70 mm) were measured much more frequently than higher ones.
29
(a)
(b)
(c)
(d)
(e)
precipitation of November, December and February in Vargas (station data)
In figure 3.13 statistical parameters of the precipitation in Vargas are shown. Again
November and December show higher precipitation than February. In December there is
a higher variability in the minimal measured value. The stations are again ordered from
the west to the east. It can be seen that there is an increment of precipitation in the east,
30
especially in November and December. In these two months, the statistical measures are
as well higher in the north than in the south (not shown).
(a)
(b)
(c)
Figure 3.14: Histograms of the precipitation in Bolívar (station data) in November (a), December
(b), February (c)
The histograms of the precipitation in Bolívar (figure 3.14) make again clear how different
the two regions in Venezuela are. Bolívar has much more rain during these three months.
In November values between 0 mm and 300 mm occur nearly with the same frequency
while in December most of the measured rainfall amounts are between 0 mm and 120 mm.
In February the spectrum is even more narrow.
31
(a)
(b)
(c)
(d)
(e)
precipitation of November, December and February in Bolívar (station data)
The figure 3.15 shows the statistical parameters of the precipitation in Bolívar. The order
of the months after the precipitation is as in Vargas. In February, the precipitation is
lowest, followed by December and in November the highest values occur. The stations are
ordered from the south to the north. The means, medians and the maxima show lowest
32
values in the north and highest ones in the middle of the state.
(a)
(b)
(c)
Figure 3.16: Histograms of the precipitation at the coast (model data of the past) in November
(a), December (b), February (c)
The graph (figure 3.16) shows the monthly precipitation of the model data (past) at the
coast. In November, precipitation amounts up to 320 mm occur. The distribution is
concentrated at values near zero but nevertheless precipitation up to 90 mm takes place
more than 30 times. In contrast, in December the highest modelled precipitation amount
is only 190 mm. More than 120 times the precipitation is only between 0 and 20 mm. This
narrow distribution is even stronger during February at the coast of Venezuela. Nearly all
the precipitation has values under 5 mm. Compared to the station data, the precipitation
of the model data has lower values during all three months. Even so, the tendency to less
rain as the dry season progresses is the same in both data sets.
33
The histograms of the precipitation of the model data of the past in the inland (figure 3.17)
are quite similar to the ones at the coast. Again, the values get smaller and less widespread
from November to February. The model data differ quite a lot from the station data in
the inland (Bolívar). The model does not make a difference in the amount of precipitation
between the two regions while the measured precipitation is much higher in the inland.
(a)
(b)
(c)
Figure 3.17: Histograms of the precipitation in the inland (model data of the past) in November
34
The histograms (figure 3.18) show the monthly precipitation of November, December and
February of the model data of the future at the coast. For November, the spectrum is
similar compared to the historical data. For the other two months the spectra are more
narrow and in all three cases the concentration is highest for values of zero or near zero.
This tendency to small precipitation amounts is more drastic than in the past.
(a)
(b)
(c)
Figure 3.18: Histograms of the precipitation at the coast (model data of the future) in November
35
In the inland, this development from the past to the future is even a bit stronger (see
figure 3.19). As in the past, the precipitation gets lower with the development of the dry
season but the concentration of small values is much stronger in the future. In February,
over 500 of the 540 cases show precipitation of less than 6 mm/month. The model and the
station data of both regions show more precipitation in November and December than in
February (November always with the highest values).
(a)
(b)
(c)
Figure 3.19: Histograms of the precipitation in the inland (model data of the future) in November
In summary, the precipitation in strongest in November, followed by December and the
fewest rainfall occur in February. This can be seen in all data set and for both precipitation
regions. The station data provide a stronger contrast between coast and inland. This agrees
with the results of the four seasons. The model data show a tendency to less rain in the
future for all three months.
36
The following table (3.1) displays the mean, maximum, minimum and the standard deviation at the stations Maiquetia (Vargas) and Cagigal (Caracas) during the three months
(November, December and February) and for the whole year.
Table 3.1: Statistical measures (mean, maximum, minimum, standard deviation) of the precipitation in November, December, February and annual of the period 1951 - 2010 at the stations
Maiquetia and Cagigal
Month
November
December
February
annual
Measurement
mean
max
min
std
mean
max
min
std
mean
max
min
std
mean
max
min
std
Maiquetia (Vargas)
70.0 mm
428.5 mm
2.0 mm
79.9 mm
60.7 mm
388.5 mm
0.0 mm
70.0 mm
31.5 mm
468.0 mm
0.0 mm
78.1 mm
552.3 mm
985.6 mm
205.0 mm
185.4 mm
Cagigal (Caracas)
84.3 mm
229.1 mm
3.2 mm
43.2 mm
42.5 mm
181.3 mm
1.8 mm
35.3 mm
16.4 mm
224.0 mm
0.0 mm
37.0 mm
862.2 mm
1657.5 mm
508.8 mm
196.3 mm
Table 3.1 shows that the means get smaller during the progression of the dry season. November has the highest values with 70.0 mm (Maiquetia) and 84.3 mm (Cagigal), followed
by December with 60.7 mm/42.5 mm and February with 31.5 mm/16.4 mm. The annual
mean, maximum and minimum are again much higher in Caracas than at the coast, in
Maiquetia. The standard deviation is higher in Maiquetia than in Caracas but for the
entire year the statistic shows the opposite. These results agree with the previous ones:
Maiquetia well represents the coast/Vargas and the precipitation in Caracas has already
different characteristics which show the tendency of the inland/Bolívar. This is not valid
for the mean precipitation of the three months, when Maiquetia provides higher values
than Cagigal.
In the following, the means in precipitation are calculated to derive the extreme events of
the past. Previous studies revealed that the major extreme events took place in February
1951, December 1999, February 2005 and November 2010, and all in the dry season. The
two stations used here are located in Vargas (Maiquetia) and Caracas (Cagigal). Vargas
was more affected than Caracas (altitude: 700 m - 900 m) because of its location between
the sea and the Cordillera de la Costa (altitude: 1500 m - 2500 m) which extends the
region between 10 ◦ N - 11 ◦ N and 65 ◦ W - 68.5 ◦ W (Hidalgo, 2001). The damage was huge
37
especially in 1999 when thousands of people died and lost their houses because of floods
and landslides in Vargas (López et al, 2005).
In table 3.2 highest 10% precipitation events (the six years with the highest precipitation),
at the station Maiquetia are listed.
Table 3.3 shows the same for the station in Caracas.
Table 3.2: Ranking of the highest 10% precipitation events [mm] in November, December, February and annual with the corresponding year of the period 1951 - 2010 at the station Maiquetia
in Vargas
November
428.5 | 2010
293.0 | 2000
274.0 | 1991
221.0 | 1973
212.9 | 2004
184.4 | 2008
December
388.5 | 1999
221.0 | 1985
207.0 | 1954
193.0 | 1975
186.0 | 1966
161.0 | 1960
February
468.0 | 1951
381.8 | 2005
147.8 | 2006
75.0 | 1969
61.0 | 1981
50.0 | 1976
Annual
985.6 | 2010
961.0 | 1951
940.0 | 2005
892.0 | 1956
883.5 | 1999
820.0 | 2008
Table 3.3: Ranking of the highest 10% precipitation events [mm] in November, December, February and annual with the corresponding year of the period 1951 - 2010 at the station Cagigal in
Caracas
November
229.1 | 2005
198.8 | 2008
161.4 | 1989
159.2 | 2010
149.2 | 1958
148.2 | 1966
December
181.3 | 1999
135.3 | 1985
114.8 | 1952
103.0 | 2007
98.2 | 1954
89.7 | 1990
February
224.0 | 1951
172.5 | 2005
69.8 | 1981
51.7 | 1997
41.3 | 1968
33.3 | 1969
Annual
1657.5 | 2010
1316.2 | 2005
1213.0 | 1996
1196.4 | 1954
1161.0 | 1990
1039.5 | 1999
In Maiquetia, the highest accumulated monthly values are always registered at one of the
four events. In November 2010 the highest value was measured with 428.5 mm, where the
second highest value has 135.5 mm less precipitation recorded. December 1999 leads the
December statistic with 388.5 mm, followed by 1985 with 221.0 mm (a difference of 167.5
mm). The monthly value in February was the highest in 1951 with 468.0 mm, followed
by the year 2005 with 381.8 mm per month. The third highest precipitation took place in
2006 with only 147.8 mm. There is a big difference (134 mm) between the two first ones
and the third. Therefore it is possible to say, that these two events were the strongest in
the last sixty years in February in Maiquetia, representing the central coast of Venezuela.
The annual statistic is not so clear. There are three years with a precipitation of more
than 900 mm: 2010 has 985.6 mm, 1951 961.0 mm and 2005 940.0 mm. 1999 has only the
fifth highest value with 883.5 mm. The annual values show that the extreme events during
one month do not have to produce a record of precipitation in the mean of the whole year.
In Caracas (Cagigal) November 2010 has only the fourth highest measured rainfall and
38
December 1999 leads the statistic but only with a difference of 43.3 mm to the second
position. A similar tendency as in Vargas can be seen in the values of February. 1951 and
2005 have the highest values and there is a big gap to the following, 102.7 mm. The annual
value at Cagigal is the highest in 2010, followed by 2005. 1999 is the last in the list of the
10% and 1951 does not occur.
Outstanding is that during all three months the registered mean precipitation in Cagigal
is lower than in Maiquetia which can be a consequence of the different topography. In
contrast, the annual values are much higher in Caracas than at the coast. That can be
explained by the more northward surface winds during the wet season which reverse the
effect of the topography.
39
4
Canonical Correlation Analysis of the
relationship between precipitation in
Venezuela and sea surface temperature
In this chapter, the influence of oceanic regions in the Atlantic and the Pacific on the
precipitation in Venezuela is investigated linearly with the Canonical Correlation Analyses.
The impact on the precipitation of station data and model data is analysed with this
method. Furthermore, the different effect on precipitation at the coast and in the inland
and variability of the oceanic influence during the seasons is studied in order to investigate
how the impacts change spatially and temporally. In the first part, the CCAs are performed
for the four seasons, in the second part for the three months during which the extreme
events occurred.
41
4 Canonical Correlation Analysis of the relationship between precipitation in Venezuela
and sea surface temperature
4.1 Canonical Correlation Analysis of the precipitation during
the seasons
4.1.1 Station data
In the following, the results of Canonical Correlation Analyses with the station data are
presented and discussed.
Before computing the CCA, an empirical orthogonal function (EOF) analyses is performed.
Two EOFs are relevant for the CCAs of the precipitation in Vargas. In the dry seasons,
the EOFs explain 67% (early) and 64% (late) of the variance. The explained variance of
the two EOFs of the wet season is 46% for the early wet season and 44% for the late rainy
season. The CCAs, applied to the precipitation in Bolívar, are the first two EOFs as well,
only for the early wet season the first four EOFs seem to explain the relevant variance of
65%. The EOFs explain 59% of the early dry season, 66% of the variance of the late dry
season and 53% of the late wet season. For the correlations with the NTA and Vargas,
three (two in the early dry season in Bolívar) SST EOFs were used which explain between
83% and 95% of the variance. Two EOFs of the Niño fields enter the CCA. They stand
for a variance between 93% and 98%. For Pac-Atl, between two and four EOFs seem to
explain the important variance of 63% to 85%. The remaining EOFs were truncated as
noise. The number of used EOFs for each CCA, chosen according to the goodness index
so that the best possible forecast is provided, can be seen in the corresponding table in the
appendix.
Table 4.1 shows the order of the zero-lag canonical correlation coefficients for the different
seasons and SST regions.
Table 4.1: Ranking of the correlation coefficients from the Canonical Correlation Analyses of the
sea surface temperature (ERSST) and the precipitation in Vargas (station data) of the four seasons
NTA
late dry (0.38)
early wet (0.28)
early dry (0.22)
late wet (0.18)
early dry
Pac-Atl (0.49)
Niño3 (0.49)
Niño3.4 (0.47)
NTA (0.22)
Niño3
early dry (0.49)
late wet (0.35)
early wet (0.31)
late dry (0.20)
late dry
NTA (0.38)
Niño3.4 (0.29)
Pac-Atl (0.26)
Niño3 (0.20)
Niño3.4
early dry (0.47)
early wet (0.38)
late wet (0.36)
late dry (0.29)
early wet
Niño3.4 (0.38)
Pac-Atl (0.35)
Niño3 (0.31)
NTA (0.28)
Pac-Atl
early dry (0.50)
late wet( 0.45)
early wet (0.35)
late dry (0.26)
late wet
Pac-Atl (0.44)
Niño3.4 (0.36)
Niño3 (0.35)
NTA (0.18)
The early dry season spans the months November, December and January. The influence
of the NTA is smaller, with 0.22, than the coefficients of the CCAs of the regions in the
42
4.1 Canonical Correlation Analysis of the precipitation during the seasons
Pacific. The precipitation pattern of the spatial loadings of the first CCA mode cannot
be explained physically when correlating with the NTA. There seems to be no plausible
explanation for inhomogeneous anomalies in this small state. Thus the second CCA mode
is discussed here. The spatial loadings show a large area with positive SST anomalies east
of Africa and a positive rainfall anomaly in Vargas. The correlation coefficients of Niño3
and Niño3.4 are quite similar with 0.49 and 0.47. The CCA patterns look very similar.
The spatial loadings of the CCA with Niño3.4 are displayed in figure 4.1. The spatial
loadings of the precipitation in (b) show the anomalies at each station. The station data
contain many missing values, thus the number of stations is reduced by the 50% threshold
of missing values. Nevertheless, a positive SST anomaly in the Niño3.4 region is associated
with a negative rainfall anomaly.
(a)
(b)
Figure 4.1: Spatial loadings of SST in the Niño3.4 region (a) and of precipitation in Vargas (b)
from the Canonical Correlation Analysis in the early dry season (ERSST and station data)
The region Pac-Atl also has a strong influence on the early season rainfall with a correlation
coefficient of 0.49. The spatial loadings show positive anomalies over the whole domain,
with smaller ones in the NTA and negative anomalies in Vargas.
In the late dry season, the NTA has the strongest influence compared to the other regions
and seasons with 0.38. The spatial loadings of the SST show a dipole structure with negative anomalies in the east and positive ones in the west. The loadings of the precipitation
have negative values. The correlation coefficient is quite small with 0.20. The coefficient
of Niño3.4 is higher with 0.29 and the spatial loadings are of the opposite sign. Niño3 has
homogeneous positive SST anomalies and negative precipitation anomalies while Niño3.4
has negative SST anomalies with stronger values at the western boundary of the domain
and positive anomalies in Vargas. The correlation coefficient of the CCA with Pac-Atl is
0.26 and the spatial loadings are shown in figure 4.2. Again the SST anomalies in the
Pacific and Atlantic are positive. That is contrary to the results of the correlation with
the NTA and the order of the coefficients suggests a higher influence of the NTA on late
dry season rainfall.
43
(a)
(b)
Figure 4.2: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation in Vargas (b)
from a Canonical Correlation Analysis in the late dry season (ERSST and station data)
The CCA of the early wet season with the NTA has a correlation coefficient of 0.28, again
lower than of the Pacific regions. Just as the spatial loadings of SST and precipitation
which have again the same sign, positive. The correlation coefficient of the CCA with Niño3
is only slightly higher compared to NTA with 0.31. Niño3.4 has a higher influence with
a coefficient of 0.38. The spatial loadings of these two regions look contrary (figure 4.3).
The corresponding precipitation pattern is positive with Niño3 and negative with Niño3.4.
(a)
(b)
Figure 4.3: Spatial Loadings of SST in the Niño3 region (a) and of SST in the Niño3.4 region
(b) from a Canonical Correlation Analysis in the early wet season (ERSST)
The CCA with Pac-Atl has a correlation of 0.35. The spatial loadings show a dipole
between Pacific and Atlantic, positive anomalies in the equatorial Pacific and negative
anomalies in the NTA. This pattern is combined with negative rainfall anomalies in Vargas. This result corresponds with the patterns of all three regions separately.
The late wet season is more influenced by the SST of the Pacific. Like in the early dry
season only the precipitation pattern of the second CCA mode when correlating NTA can
be explained physically. The correlation coefficient is small with 0.18. The spatial loadings
show negative SST anomalies in most of the domain, only at its southern boundary do the
anomalies have the same sign like the ones of the precipitation, positive. The coefficients
and patterns of the spatial loadings of the CCAs with the Niño3 and Niño3.4 region are
44
very similar. The coefficients are 0.35 and 0.36 and the spatial loadings show negative precipitation anomalies and positive SST anomalies in both cases. The correlation coefficient
of the CCA with Pac-Atl is quite high with 0.44. The patterns of the spatial loadings for
SST and precipitation agree perfectly with the results of every single region. Negative anomalies in the Pacific and northern equatorial Atlantic, positive anomalies in the southern
Atlantic domain and in Vargas (figure 4.4).
(a)
(b)
Figure 4.4: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation in Vargas (b)
from a Canonical Correlation Analysis in the late wet season (ERSST and station data)
To sum up, the influence of the NTA is highest during the late dry season when the lowest
precipitation is measured. The regions Niño3 and Niño3.4 show similar results with slightly
higher correlation coefficients with Niño3.4 in most cases (except for the early dry season).
The spatial loadings only differ among both Pacific regions in the early wet season. The
region Pac-Atl has correlations of the same order or higher than the Niño regions. Its
pattern reflects the pattern of the NTA only in the wet season, the ones of the Pacific
during all four seasons.
Table 4.2 shows the results of zero-lagged CCAs of the four SST regions and the precipitation of the four seasons in Bolívar.
The canonical correlation coefficient is very small with the NTA (0.17) in the early dry
season. The correlation with Niño3 and Niño3.4 are much higher and very similar with
0.77 and 0.76. The spatial loadings of the CCAs with both Niño regions show oppositely
signed anomalies of SST and precipitation. The coefficient of the Pac-Atl region is high
as well with 0.70 and its spatial loadings underline the order of coefficients. There are
positive SST anomalies over the whole domain with stronger ones in the Pacific than in
the Atlantic and negative precipitation anomalies in Bolívar.
In the late dry season all zero-lag correlations of the four regions have a first spatial loading of the precipitation that cannot be explained physically. Therefore, the second is
mentioned in this discussion. The correlation coefficients do not differ much between the
45
sea surface temperature (ERSST) and the precipitation in Bolívar (station data) of the four seasons
NTA
early wet (0.85)
late wet (0.67)
late dry (0.21)
early dry (0.17)
early dry
Niño3 (0.77)
Niño3.4 (0.76)
Pac-Atl (0.70)
NTA (0.17)
Niño3
early dry (0.77)
late wet (0.64)
early wet (0.53)
late dry (0.34)
late dry
Niño3 (0.34)
Niño3.4 (0.21)
NTA (0.21)
Pac-Atl (0.20)
Niño3.4
early dry (0.76)
early wet (0.75)
late wet (0.64)
late dry (0.21)
early wet
NTA (0.85)
Pac-Atl (0.82)
Niño3.4 (0.75)
Niño3 (0.53)
Pac-Atl
late wet (0.82)
early wet (0.82)
early dry (0.70)
late dry (0.20)
late wet
Pac-Atl (0.82)
NTA (0.67)
Niño3.4 (0.64)
Niño3 (0.64)
regions. The correlations with the NTA and Niño3.4 have a coefficient of 0.21, Pac-Atl a
slightly lower one with 0.20 and the Niño3 has the highest with 0.34. The spatial loadings
of the CCA with the NTA show negative anomalies in the eastern NTA and in Bolívar
(figure 4.5).
(a)
(b)
Figure 4.5: Spatial loadings of SST in the NTA region (a) and of precipitation in Bolívar (b)
from a Canonical Correlation Analysis in the late dry season (ERSST and station data)
The spatial loadings of Niño3 have negative values in the east and positive ones in the
west of the oceanic region and in the precipitation. That corresponds well with the results
of Niño3.4 where the anomalies of SST are negative over the whole domain and of preci-
46
pitation positive in Bolívar. The spatial loadings of Pac-Atl show negative values in the
Pacific and opposite ones in precipitation. In the Atlantic there is a north-south dipole
with positive anomalies in the south.
The CCAs of the early wet season have higher correlation coefficients than the previous
season. Previous studies agree with that only for the NTA. The coefficient of the CCA with
the NTA is very high and highest compared to the other regions with 0.85. The spatial
loadings of the precipitation do not have anomalies of one sign over whole Bolívar. The
SST has negative anomalies, as precipitation at most stations. Some stations in the southeast have slightly positive values. For this season the results differ within the two Niño
regions. The CCA with Niño3 has a correlation coefficient of 0.53. The spatial loadings
show positive values in the SST and in the main part of Bolívar. Anomalies with contrary
signs occur in the south of the state. The CCA with Niño3.4 has a higher coefficient (0.75).
The spatial loadings have the same pattern but with opposite signs compared to the ones
of the CCA with Niño3. The coefficient of the CCA with Pac-Atl is high as well with a
value 0.82. The spatial loadings of the CCA with Pac-Atl provide a good summary of the
results of each region. They are shown in figure 4.6.
(a)
(b)
Figure 4.6: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation in Bolívar (b)
from a Canonical Correlation Analysis in the early wet season (ERSST and station data)
The CCA with the NTA and the precipitation in the late wet season provides a correlation
coefficient which is slightly higher than in the Niño regions with 0.67 compared to 0.64
47
(Niño3 and Niño3.4). The spatial loadings of the CCA with the NTA show negative anomalies in precipitation associated with a dipole in the SST anomalies with positive values
in the north of the domain. The patterns of the spatial loadings of the CCA with Niño3
have positive anomalies in the SST and negative ones in the precipitation. The spatial loadings of the CCA with Niño3.4 show similar patterns: Negative anomalies in precipitation
and positive ones in the SST. The SST anomalies are stronger in the western part of the
region. The CCA with Pac-Atl shows that the impact of both oceans is important in this
season. The coefficient is higher than of any single region with 0.82. The spatial loadings
figure 4.7 summarise that negative SST anomalies in the equatorial Pacific and positive
ones in the NTA provide positive rainfall anomalies in the late wet season.
(a)
(b)
Figure 4.7: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation in Bolívar (b)
from a Canonical Correlation Analysis in the late wet season (ERSST and station data)
In summary, the influence of the NTA is low during the dry and high during the wet seasons. The Niño regions and Pac-Atl have similar and high coefficients for both wet and
the early dry season.
For Vargas the goodness indices of the CCA with the NTA are all low, between -0.063
(late dry) and 0.114 (early wet). The indices of the correlations with the Niño regions are
lower than 0.1 in the late dry and early wet season, still low with 0.150 and 0.157 in the
late wet and highest in the early dry season. Good forecasts for the precipitation in Vargas
48
can only be made with the Niño regions and Pac-Atl in the early dry season. The indices
differ between 0.309 (Niño3.4) and 0.345 (Pac-Atl).
The zero-lag goodness indices of the correlations with the NTA and precipitation in Bolívar
are low during both dry seasons (0.074 and -0.039). The CCA of the early wet season has a
goodness index of 0.177 and of the late wet season the highest with the NTA (0.222). The
goodness indices of the Niño regions are mostly similar. The best forecast can be made of
the early dry season with indices of 0.515 and 0.519. Followed by the late wet season with
0.365 and 0.354, the late dry season with 0.241 and 0.133, and the lowest indices occur in
the early wet season with 0.121 and 0.004. The Pac-Atl region has indices of the same size
like the Niño regions. The highest index appears as well with the early dry season (0.468),
second is the index of the late wet with 0.460, third of the early wet (0.146) and goodness
index of the late dry season provides the worst forecast with 0.064.
The goodness index is higher with the Pacific regions than with the NTA. Furthermore,
the precipitation in Vargas and Bolívar can possibly be predicted in the early dry season
by the Niño regions.
In the table 4.3 the lags of the highest canonical correlation coefficients and the highest
goodness indices are listed.
Table 4.3: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature (ERSST) and the precipitation (station data) of the
four seasons, SST leading the precipitation up to six months
Season and region
early dry Vargas
late dry Vargas
early wet Vargas
late wet Vargas
early dry Bolívar
late dry Bolívar
early wet Bolívar
late wet Bolívar
NTA
5|3
0|6
5|3
4|4
6|4
6|6
0|2
1|2
Niño3
1;3| 1
5|0
0|1
3|3
1|0
2|0
1|1
4|0
Niño3.4
5|1
2|1
0|3
3|3
0|0
6|6
0|3
2|0
Pac-Atl
5|3
2|2
2|2
4|3
3|4
5|6
2|2
3 | 0;1
The detailed results of all lags are shown in the appendix in tables A.6 - A.13.
The precipitation of the early dry season in Vargas, is highest correlated with the SST of
NTA, Niño3.4 and Pac-Atl during JJA, while Niño3 influences strongest with one or three
lags. As well in the late dry season Niño3 has a different result than the other three with
the highest correlation with five lags. But in the early dry season the correlations of the
Niño regions and Pac-Atl are all very similar. In the early wet season the Niño regions and
Pac-Atl have highest coefficients at small lags (0 - 3), NTA influences earlier (December February). The results of the late wet season correspond well for all regions. The highest
correlations are found at lag 4 (April - June) or 3 (May - July). The highest goodness
49
index coincides well with the highest correlation coefficient for all regions in this season.
A good agreement is defined as a difference of maximal 2 lags. For the NTA and Niño3
this is not fulfilled in the late dry season. The lags of the CCA with Niño3.4 agree among
coefficient and index only in both late seasons, for Pac-Atl in all four seasons.
In Bolívar the NTA has its highest correlation coefficients at a large lag (six) when correlating with the precipitation of the dry seasons and at a small lag (zero or one) with the
precipitation of the wet seasons. The lags with the highest goodness index correspond with
these results. The strongest influence of Niño3 is short time in advance (between lag one
and two). Only for the precipitation of the late wet season is the coefficient highest at lag
four. The goodness indices underline this with a good conformity of the lags in these three
seasons. The lags of the highest correlation coefficient and goodness index differ with four
in the CCAs of the precipitation of the late wet season. Niño3.4 influences strongest at the
same time or with a small lag of two (precipitation of the late wet season) but the rain in
the late dry season is highest correlated with a lag of six. The goodness index is highest
at the same or at similar lags. Also for Pac-Atl both lags correspond well. The lags of the
highest correlation coefficients seem to be temporally in the middle between NTA and the
Niño regions.
The spatial loadings fulfil the pattern of anomalies which the literature suggests (sign of
SST anomalies of NTA and precipitation are the same and opposite to the SST anomalies
in the Pacific). Exception of this patterns are the rainfall in Vargas in the early dry season
correlated with Pac-Atl at the lag of the highest goodness index, in the early wet season
at the lag of the goodness index when correlating Niño3 and during the late wet season
in both lags of NTA and Pac-Atl. For the precipitation in Bolívar, the spatial loadings
fulfil the pattern with exception of lags when correlating the precipitation of the early wet
season with the Niño regions and Pac-Atl. That agrees with the results of previous studies.
In summary, for precipitation in Vargas there is a high variability of the results for different
lags, where only the late wet season provides homogenous results among the SST regions.
In Bolivar the results of the NTA suggest that the SST of the wet season influences the
whole year. The Niño regions influence only short time in advance with the exception of
Niño3.4 in the late dry season. The anomalies of the spatial loadings differ in the NTA
and Pac-Atl for the precipitation in Vargas in the late wet season and in the Niño regions
and Pac-Atl for the early wet season rainfall in Bolívar.
4.1.2 Model data of the past
In this part, the results of the CCAs of the historical model data of sea surface temperature
and the precipitation are displayed and discussed.
All EOFs which explain at least 1% enter the CCAs. None of the EOFs of the precipitation
50
at the coast were truncated, so that 100% of the variance remains for the CCAs. For the
data of the inland between five and seven of the nine EOFs correspond with this criterion.
They represent between 97% and 99% of the total variance. For the NTA data six or seven
EOFs remain, a 95% variance. Between three and five EOFs of the SST of Niño3 fulfil
the criterion so that 96% or 97% variance is covered. With the Niño3.4 region there are
only three or four EOFs which enter the CCAs but explain between 98% and 99% of the
variance. The remaining variance of Pac-Atl is lower with values between 91% and 93%
even if the number of EOFs is much higher with eight to twelve. The explained variance
of the leading mode of the Niño regions is much higher than the explained variance of the
NTA and even more compared to Pac-Atl. This induces the different number of remaining
modes.
Table 4.4 shows the ranking of the canonical correlation coefficients with the model data
of the past.
sea surface temperature and the precipitation at the coast of the four seasons (model data of the
past)
NTA
late wet (0.85)
early dry (0.74)
early wet (0.45)
late dry (0.34)
early dry
Pac-Atl (0.81)
NTA (0.74)
Niño3.4 (0.49)
Niño3 (0.38)
Niño3
late wet (0.75)
late dry (0.42)
early wet (0.40)
early dry (0.38)
late dry
Pac-Atl (0.58)
Niño3 (0.42)
Niño3.4 (0.39)
NTA (0.34)
Niño3.4
late wet (0.75)
early dry (0.49)
early wet (0.41)
late dry (0.39)
early wet
Pac-Atl (0.68)
NTA (0.45)
Niño3.4 (0.41)
Niño3 (0.40)
Pac-Atl
late wet (0.91)
early dry (0.81)
early wet (0.68)
late dry (0.58)
late wet
Pac-Atl (0.91)
NTA (0.85)
Niño3 (0.75)
Niño3.4 (0.75)
In the early dry season, the highest correlation is the one with Pac-Atl (0.81), followed
by NTA (0.74), Niño3.4 (0.49) and Niño3 (0.38). The spatial loadings of the CCA with
the NTA show negative SST anomalies in the centre of the region and positive anomalies
at its southeast and northwest corner. The precipitation anomalies are negative with the
strongest anomalies in the centre. Niño3 and Niño3.4 have similar results, as expected,
with anomalies of opposite sign in SST and precipitation. The SST anomalies are less
strong in the Niño3.4 region, the precipitation anomalies look exactly like in the CCA
with the NTA. The spatial loadings of the Pac-Atl show the same patterns of positive SST
anomalies in the Niño regions and negative anomalies in the NTA and the precipitation at
the coast (figure 4.8).
In the late dry season the influence of the Niño regions becomes stronger compared to the
previous season and stronger than the one of the NTA. The coefficient of the correlation
with Pac-Atl is still the highest with 0.58, but this time it is followed by Niño3 with 0.42,
51
(a)
(b)
Figure 4.8: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation at the coast (b)
from a Canonical Correlation Analysis in the early dry season (model data of the past)
the third strongest influence has Niño3.4 with 0.39 and the coefficient of NTA is only of
0.34. The spatial loadings of the NTA show negative rainfall anomalies, slightly stronger in
the west and positive SST anomalies with highest values in the southeast and a little area
with negative anomalies in the northwest. The spatial loadings of the Niño regions show
positive SST and negative rainfall anomalies. In the Niño3 region the anomalies are higher
in the west. The spatial loadings of the CCA with Pac-Atl are shown in figure 4.9. The
SST anomalies are positive with the exception of a small area in the NTA and north of the
Caribbean coast of Venezuela. The precipitation anomalies are negative. These patterns
show, additionally to the coefficients, that the equatorial Pacific has a stronger influence
than the NTA in the late dry season on the precipitation at the coast of Venezuela.
(a)
(b)
Figure 4.9: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation at the coast (b)
from a Canonical Correlation Analysis in the late dry season (model data of the past)
In the early wet season, which spans the months May to July, the correlation coefficients
are a bit higher than in the season before and have the same order as at the beginning of
the dry season. The Pac-Atl has again the highest correlation with 0.68. With 0.45, the
NTA has the second strongest influence in this season. The correlations of the Niño regions
are lower and similar to each other with coefficients of 0.40 (Niño3) and 0.41 (Niño3.4).
The spatial loadings of the CCA with the NTA are shown in figure 4.10. There is a warm
pool between the coast of Africa and 50 ◦ W. Negative anomalies occur in the north and
the west of the region. The precipitation anomalies are negative with higher values in the
east.
The spatial loadings of the CCAs with the Niño regions show strong positive SST anoma-
52
(a)
(b)
Figure 4.10: Spatial loadings of SST in the NTA region (a) and of precipitation at the coast (b)
from a Canonical Correlation Analysis in the early wet season (model data of the past)
lies and strong negative anomalies in the precipitation with the maximum in the central
grid box. The anomalies of the SST of Pac-Atl summarise the results of each region separately. The SST anomalies of the Pacific and of the southern part of the Atlantic area
are negative and positive only in the north of the NTA and in the Caribbean Sea; the
precipitation anomalies are positive too.
The oceanic influence on the precipitation at the coast of Venezuela seems to be strongest
in the late wet season. Both Niño regions have a correlation coefficient of 0.75. The NTA
has an even higher coefficient of 0.85. The Pac-Atl region has again the highest correlation
coefficient with 0.91. In this season the area of the NTA where the SST anomalies have the
same sign as the precipitation is even smaller than in the season before. Just a small part
in the northwest has negative anomalies, the rest of the spatial loadings is positive. The
corresponding rainfall anomalies are homogeneous and strong. The spatial loadings of the
Niño regions have the same patterns and signs. The figure 4.11 shows those of Niño3.4.
The negative SST anomalies are strongest in the west, the precipitation anomalies are very
high with the same strength in all three grid boxes.
(a)
(b)
Figure 4.11: Spatial loadings of SST in the Niño3.4 region (a) and of precipitation at the coast
(b) from a Canonical Correlation Analysis in the late wet season (model data of the past)
The spatial loadings of the CCA with Pac-Atl show negative anomalies in the Pacific with
highest values in the west and at the west coast of Central America. Positive SST anomalies occur on the north coast of South America and in the Atlantic south of the Equator.
The positive rainfall anomalies are strong and homogeneous.
To sum up, the coefficients of Pac-Atl are always the highest, followed by NTA with the
53
exception of the late dry season where both Niño regions have a stronger influence. The
oceanic influence is highest in the late wet season. The anomalies of the Niño regions are
always of the opposite sign than the ones of the precipitation. The SST anomalies of the
NTA have the same sign like the rainfall at the beginning of the dry and the wet season, the
patterns of the Pac-Atl combine the results of the correlations with the regions separately.
In addition to the precipitation at the coast, the correlations were performed with the
area in the inland of Venezuela. In table 4.5 the correlation coefficients and the order of
the SST regions with respect to the coefficients are shown.
sea surface temperature and the precipitation in the inland of the e seasons (model data of the past)
NTA
late wet (0.85)
early dry (0.74)
late dry (0.67)
early wet (0.49)
early dry
Pac-Atl (0.84)
NTA (0.74)
Niño3 (0.63)
Niño3.4 (0.53)
Niño3
late wet (0.73)
early dry (0.63)
early wet (0.52)
late dry (0.44)
late dry
Pac-Atl (0.73)
NTA (0.67)
Niño3.4 (0.53)
Niño3 (0.44)
Niño3.4
late wet (0.72)
late dry (0.53)
early dry (0.53)
early wet (0.48)
early wet
Pac-Atl (0.57)
Niño3 (0.52)
NTA (0.49)
Niño3.4 (0.48)
Pac-Atl
late wet (0.91)
early dry (0.84)
late dry (0.73)
early wet (0.57)
late wet
Pac-Atl (0.91)
NTA (0.85)
Niño3 (0.73)
Niño3.4 (0.72)
In the early dry season the order of influence of the oceanic regions on inland rainfall is
similar as on coastal precipitation, only the Niño regions change their position in the ranking. The coefficient of Pac-Atl is of 0.84, slightly higher than with the precipitation at
the coast. That of the CCA with the NTA is 0.74 exactly as with the coastal precipitation.
Niño3 has a correlation coefficient of 0.63 much higher than with the rainfall at the coast
and also 0.1 higher than with Niño3.4. The spatial loadings of the correlation with the
Pac-Atl are shown in figure 4.12. The SST anomalies are positive in the eastern equatorial Pacific, as with the correlation of Niño3. The Atlantic is mostly colder with positive
anomalies only in small parts in the east and at the east coast of Central America. The
spatial loadings of SST of the CCAs with the three regions separately show the same. The
precipitation anomalies are strongest in the north and east where the altitude is low when
correlated with the NTA and strongest in the grid boxes with intermediate elevation in
the spatial loadings of the CCA with Niño3. The spatial loadings of precipitation of the
CCA with Pac-Atl are strongly negative in all grid boxes except the one with the highest
altitude in the southwest.
The results of the CCAs with late dry season rainfall are different for the coast and the
inland. In the inland, the highest correlation coefficient is of the CCA with the Pac-Atl
with a coefficient of 0.73, followed by the CCA with NTA with a coefficient of 0.67. The
54
(a)
(b)
Figure 4.12: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation in the inland
(b) from a Canonical Correlation Analysis in the early dry season (model data of the past)
Niño regions have lower influence with coefficients of 0.53 (Niño3.4) and 0.44 (Niño3). The
spatial loadings of the correlation with the NTA show positive anomalies in the south and
in the centre of the NTA and precipitation anomalies of the same sign, strongest where
the orography is lowest. Niño3 and Niño3.4 have again the same pattern in loadings in
the spatial loadings although the coefficients differ with 0.09. The SST anomalies are positive and the rainfall anomalies negative. The spatial loadings of the Pac-Atl correlation
are shown in figure 4.13. SST anomalies are positive in the Pacific and in the north and
south of the tropical Atlantic. In its centre the anomalies are negative like the ones of the
precipitation in the inland, with highest values in the southwest and northeast. Again,
the spatial loadings of the CCA with Pac-Atl are a good summary of the patterns of the
correlation with one of the other three oceanic region.
In the early wet season the canonical correlation coefficients are more similar between the
regions than in the previous seasons. Again, the highest coefficient provides the CCA with
Pac-Atl, 0.57. The three small regions have similar coefficients. The CCA with Niño3 has
a coefficient of 0.52, NTA of 0.49 and Niño3.4 of 0.48. It seems that there is no notable
difference between the influence of the three regions on the inland precipitation in this
season. The spatial loadings of the CCA with the NTA can be seen in figure 4.14. The
SST anomalies show a dipole with positive values in the northwest and negative ones in
the southeast. The precipitation anomalies are positive, strongly in the north and in the
east.
55
(a)
(b)
(b) from a Canonical Correlation Analysis in the late dry season (model data of the past)
(a)
(b)
Figure 4.14: Spatial loadings of SST in the NTA region (a) and of precipitation in the inland (b)
from a Canonical Correlation Analysis in the early wet season (model data of the past)
The spatial loadings of precipitation of the CCAs with the NTA and Pac-Atl are positive,
with the Niño regions negative. The loadings of the SST in the Niño regions are positive.
The SST of the CCA with Pac-Atl has negative anomalies in the Pacific with a strong cold
56
tongue just at the Equator. In the Atlantic the anomalies are positive in the north, in the
south and especially in the Caribbean Sea and negative in the centre of the area.
In the second part of the rainy season, the coefficients are the highest compared to the
other seasons and similar compared to the coast. The Pac-Atl again provides the strongest
influence with a coefficient of 0.91, followed by the NTA with 0.85. The Niño regions
have coefficients of 0.73 (Niño3) and 0.72 (Niño3.4). The SST anomalies of the NTA are
negative in the major part and positive only in the southwest and in the precipitation.
The Niño region has negative SST anomalies with maxima at the north boundary of the
areas and positive rainfall anomalies. The spatial loadings of the Pac-Atl (figure 4.15)
show positive anomalies in the Pacific with highest values just north of the Equator and
at the west coast of Central America. In the Atlantic, there are positive anomalies in the
major part north of the Equator and negative ones south of the Equator and in a thin line
at the coast of South America.
(a)
(b)
(b) from a Canonical Correlation Analysis in the late wet season (model data of the past)
In summary, the influence of Pac-Atl is always the strongest, followed by the NTA with the
exception only of the early wet season. In this season, all regions have similar coefficients.
The signs of the anomalies are always contrary between Niño regions and precipitation and
the same between NTA and precipitation, with the exception of the late wet season. The
patterns of Pac-Atl always give a summary of the three regions separately. The oceanic
influence is highest in the late wet season.
57
The goodness index of the zero-lag correlations with the NTA and the coastal precipitation is highest in the late wet season with 0.780, followed by the early dry (0.579), early
wet (0.304) and late dry season (-0.075). The order of the seasons is the same for the
indices and for the coefficients. The Niño regions make a good forecast in the late wet
season with 0.670 (Niño3) and 0.679 (Niño3.4). In the other seasons the goodness index is
always just around 0.3 or 0.4. Pac-Atls goodness indices have the same order as the ones of
the NTA and are between 0.815 and 0.274. The correlations with the inland do not differ
greatly in the goodness indices compared to the coast. The only difference is that Pac-Atl
has a stronger influence in the late dry (0.475) than in the early wet season (0.370).
Summing up, Pac-Atl and NTA provide a better zero-lag forecast in the second half of the
year, while the Niño regions have a higher index than the CCAs with the NTA in the early
wet and the late dry season. Highest of all is always the goodness index of Pac-Atl. This
conclusion is valid for the precipitation in both areas in Venezuela.
In table 4.6 the lags with the highest correlation coefficients and goodness indices are
listed. In tables A.14 - A.21 (appendix) the correlation coefficients and goodness indices
of each lag are shown.
Table 4.6: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature and the precipitation of the four seasons (model
data of the past), SST leading the precipitation up to six months
Season and region
early dry coast
late dry coast
early wet coast
late wet coast
early dry inland
late dry inland
early wet inland
late wet inland
NTA
1|0
2|6
0|0
1|0
0|0
0|5
2|0
1|0
Niño3
4|0
0|0
3|1
3|3
0|0
3|1
2|2
3|3
Niño3.4
0|0
0|0
6|0
2|2
0|0
0|0
1|0
2|3
Pac-Atl
0|0
0|0
0|0
0|0
0|0
0|0
1|1
0|3
The NTA and Pac-Atl have their strongest CCA coefficients always at small lags which
include at least one of the three months of the corresponding season. Furthermore, the
lags of the highest goodness indices differ only up to two lags compared to the highest
correlation coefficient except of the CCAs with NTA and the precipitation in both regions
in the late dry seasons and the CCA of Pac-Atl with the inland precipitation of the late wet
season. The strongest influence of Niño3 is at lags between zero and four, simultaneously
in the early dry (inland) and late dry season (coast). The results of the goodness indices
are similar but in the early dry season at the coast they differ with four lags. The results of
the Niño3.4 region correspond within coast and inland. The highest correlation coefficient
is at the zero-lag in both dry seasons, at lag two in the late wet and at lag six (coast) and
58
one (inland) in the early wet season. This big lag of six seems to be an exception and the
correlation coefficients between lag zero and six only differ in 0.03.
The spatial loadings at the lags with the highest coefficients and indices of the Niño regions always have anomalies of opposite signs in SST and precipitation. NTA differs from
anomalies of the same sign in SST and rainfall at the lags with the highest coefficients
and indices in the late wet seasons and the late dry season at the coast. Furthermore, the
common pattern of the spatial loadings is not fulfilled at the lag of the highest correlation
coefficient in the early wet season and at the lag of the highest index in the late wet season,
both with precipitation in the inland. In agreement with this, the spatial loadings at the
lags of the highest coefficients and indices of the CCAs with Pac-Atl vary in the late wet
seasons and at the lag of the highest goodness index in the late dry season at the coast.
In conclusion, the influence is highest and forecast is as good as possible nearly always at
small lags when at least one of the months of the particular precipitation season is included
in the three-month average. At these lags, the anomalies in the east equatorial Pacific are
always of opposite sign to the ones of the precipitation and the NTA, which have the same
sign with the exception of both late seasons in the NTA and the Atlantic part of Pac-Atl.
4.1.3 Model data of the future
In this part, the results of CCAs of the model data of the future with sea surface temperature and precipitation are displayed and discussed.
The EOFs of the precipitation at the coast all explain more than 1% so that all of them
enter the CCAs and 100% variance remains (three EOFs). For the precipitation in the
inland between five and eight EOFs enter the CCAs which explain between 97% and 99%.
The truncated variance of the oceanic regions is small as well. The three, four or five
EOFs of the NTA explain at least 96%, of the Niño3 it is even more with 98% during
all four seasons (three or four EOFs) and between 98% and 99% of the total variance are
represented in the three or two EOFs of the Niño3.4 region. The variance which enters
the CCAs with the Pac-Atl, is a little less (93% to 95%) with between five and seven used
EOFs. Again, the leading mode of the Niño regions explain more variance compared to
the NTA and Pac-Atl.
The canonical correlation coefficients of the seasonal analysis of the future data are listed
in table 4.7.
The rainfall in the early dry season at the coast is most strongly influenced by the SST of
the Pac-Atl with 0.53, followed by the two Niño regions with 0.45. The impact of the NTA
is quite low with 0.30 (second CCA mode). The spatial loadings of the CCA with Pac-Atl
summarise the results of the three regions. There are positive SST anomalies, stronger
ones in the Pacific than in the Atlantic and negative precipitation anomalies.
59
sea surface temperature and the precipitation at the coast of the four seasons (model data of the
future)
NTA
late wet (0.76)
early wet (0.64)
early dry (0.30)
late dry (0.22)
early dry
Pac-Atl (0.53)
Niño3 (0.45)
Niño3.4 (0.45)
NTA (0.30)
Niño3
late wet (0.75)
early wet (0.46)
early dry (0.45)
late dry (0.19)
late dry
Pac-Atl (0.26)
NTA (0.22)
Niño3 (0.19)
Niño3.4 (0.12)
Niño3.4
late wet (0.74)
early dry (0.45)
early wet (0.42)
late dry (0.12)
early wet
Pac-Atl (0.68)
NTA (0.64)
Niño3 (0.46)
Niño3.4 (0.42)
Pac-Atl
late wet (0.82)
early wet (0.68)
early dry (0.53)
late dry (0.26)
late wet
Pac-Atl (0.82)
NTA (0.76)
Niño3 (0.75)
Niño3.4 (0.74)
The late dry season rainfall is not strongly correlated with the oceanic surface temperatures. Strongest is the Pac-Atl with only 0.26, followed by the NTA with 0.22, Niño3
with 0.19 and Niño3.4 with 0.12. Except for Niño3, the second CCA modes were used to
get a physically meaningful precipitation pattern of the spatial loadings. In the spatial
loadings of the NTA there is a north-south dipole with positive values in the north. The
corresponding precipitation anomalies are positive. The patterns of the correlations with
the Niño regions do not agree. The correlation with Niño3 shows negative anomalies in
both variables while Niño3.4 has contrary anomalies between SST and precipitation. The
spatial loadings of Pac-Atl show an El Niño situation with negative values in the NTA
and positive rainfall anomalies. This does not agree with the literature (Mestas-Nuñez
et al. (2007), Cárdenas et al. (2002, 2003), Enfield (1996), Guenni et al. (unpublished
manuscript) Chen et al. (2002), Taylor et al. (2002), Martelo (2003,I)).
The rainfall of the early wet season is highly influenced by the NTA (0.64) and Pac-Atl
(0.68). The Niño regions play an important role too with 0.46 (Niño3) and 0.42 (Niño3.4).
The anomalies of SST and precipitation are of contrary sign in all four correlations. Just
one small area in the northeast of the NTA has negative anomalies like the precipitation.
Even stronger coefficients are provided by the correlations of the late wet season. The
SST of all areas in the oceans have high correlations, strongest again Pac-Atl with 0.82,
NTA has the second highest with 0.76 and the Niños just slightly lower with 0.75 (Niño3)
and 0.74 (Niño3.4). All spatial loadings show positive SST anomalies and negative precipitation anomalies. In figure 4.16 the spatial loadings of the correlation with Pac-Atl are
displayed. The SST anomalies of the Pacific are stronger than the ones of the Atlantic but
with a line of lower anomalies in the east near the equator. This pattern could indicate
a La Niña condition. The Caribbean SST anomalies are negative like the precipitation
anomalies.
In conclusion, the Pac-Atl always provides the highest correlation, the Niños the lowest
60
(a)
(b)
Figure 4.16: Spatial loadings of SST in the Pac-Atl region (a) and of precipitation at the coast
(b) from a Canonical Correlation Analysis in the late wet season (model data of the future)
except for the early dry season. The strongest oceanic influence is in the late wet season,
followed by the early wet, early dry and late dry (Niño3 has a higher correlation in the
early dry than in the early wet). The spatial loadings show anomalies of opposite sign for
all correlations with the Niño regions, except Niño3 in the late dry season. Even the NTA
has only one exception with anomalies of the same sign in SST and precipitation, in the
late dry. The correlations with Pac-Atl have contrary anomalies in SST and precipitation
in both early seasons, in the late dry an El Niño and in the late wet a condition that could
be La Niña, with stronger rain during the El Niño event and dry conditions during La
Niña. This stands in contrast to previous studies which suggest that an El Niño favours
positive rainfall anomalies and a La Niña negative ones.
The table 4.8 show the results of the correlations with the precipitation in the inland
during the four seasons.
sea surface temperature and the precipitation in the inland of the four seasons (model data of the
future)
NTA
late wet (0.83)
early wet (0.71)
late dry (0.41)
early dry (0.41)
early dry
Niño3 (0.71)
Pac-Atl (0.68)
Niño3.4 (0.61)
NTA (0.41)
Niño3
early wet (0.74)
late wet (0.73)
early dry (0.71)
late dry (0.46)
late dry
Pac-Atl (0.68)
Niño3 (0.46)
NTA (0.41)
Niño3.4 (0.38)
Niño3.4
late wet (0.73)
early wet (0.71)
early dry (0.61)
late dry (0.38)
early wet
Pac-Atl (0.74)
Niño3 (0.74)
NTA (0.71)
Niño3.4 (0.71)
Pac-Atl
late wet (0.75)
early wet (0.74)
early dry (0.68)
late dry (0.68)
late wet
NTA (0.83)
Pac-Atl (0.75)
Niño3.4 (0.73)
Niño3 (0.73)
The influence of the oceans on the rainfall in the inland in the early dry season is higher
than on the coastal precipitation. Highest is the correlation with Niño3 (0.71), followed by
the Pac-Atl (0.68), the Niño3.4 (0.61) and the NTA with 0.41. The spatial loadings show
contrary anomalies of SST and precipitation in all four correlations.
61
In the late dry season the coefficients are again higher compared to the results of the rainfall at the coast. Pac-Atl has the highest with 0.68, the other three regions have similar
coefficients. Niño3 has the strongest influence of them all with 0.46, NTA has a coefficient
of 0.41 and Niño3.4 of 0.38. The spatial loadings of the CCA with NTA provide anomalies
of opposite sign between SST and precipitation. Both Niño regions show mainly anomalies
with contrary sign but with a pool of anomalies like the precipitation between 110 ◦ W
and 130 ◦ W. This pattern can also be seen in the spatial loadings of the Pac-Atl where
the Atlantic has negative anomalies and the precipitation positive ones. This is shown in
figure 4.17.
(a)
(b)
(b) from a Canonical Correlation Analysis in the late dry season (model data of the future)
The precipitation in the early wet season is highly influenced by the SST of all regions. The
canonical correlation coefficients differ between 0.74 (Pac-Atl and Niño3) and 0.71 (NTA
and Niño3.4). The Niño regions and Pac-Atl show positive SST anomalies and negative
precipitation anomalies in the spatial loadings. The CCA with NTA has contrary signs
in the anomalies too, except for a small area in the northeast of the NTA with negative
anomalies like the precipitation.
The correlation coefficients of the CCAs of the late wet season are similarly high. The NTA
has a coefficient of 0.83. The second strongest influence has Pac-Atl with 0.75, closely followed by the Niño regions with 0.73. The spatial loadings of the correlation with the NTA
show mainly positive SSTs and only in the very southwest negative ones like the sign of the
precipitation anomalies. Both Niño regions have anomalies of opposite sign, also Pac-Atl
62
with the exception of the Caribbean Sea at the coast of Venezuela and Columbia.
In short, like at the coast the oceanic influence is higher in the wet than in the dry seasons.
The impact of Pac-Atl is always strong, strongest during the late dry and early wet. The
NTA has a higher coefficient than both Niños only in the late wet season. The NTA has
the smallest influence compared to the other regions in the early dry and similar to the
Niños in the late dry and early wet season. The spatial loadings have anomalies of contrary
sign with all regions in the early dry and both wet seasons. The NTA has it additionally in
the late dry season but in the spatial loadings of the Niños and Pac-Atl the SST anomalies
are contrary to the precipitation anomalies only in a pool in the middle of this Pacific area
between 110 ◦ W and 130 ◦ W.
All four regions provide the best zero-lag forecast in the late wet season (between 0.544
and 0.588). The Niño regions and Pac-Atl have similar high indices in the early wet and
the Niños even in the early dry season. Least predictable is the late dry season with all
oceanic regions.
Table 4.9 lists the lags of the highest correlation coefficients and of the highest goodness indices.
Table 4.9: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature and the precipitation of the four seasons (model
data of the future), SST leading the precipitation up to six months
Season and region
early dry coast
late dry coast
early wet coast
late wet coast
early dry inland
late dry inland
early wet inland
late wet inland
NTA
1|2
6|6
0|6
0|0
3|2
4|6
0|0
1|0
Niño3
5|0
2|0
2|2
0|0
0|2
2|2
0|0
0|0
Niño3.4
1|1
6|1
1|0
0|0
0|0
2|4
0|0
0|0
Pac-Atl
2|3
6|6
0|0
0|0
2|3
0|5
0|0
4|3
All lags with their coefficients and indices can be seen in tables A.22 - A.29 in the appendix.
The lags with the highest correlation coefficients are small (between zero and two) in the
early dry and the wet seasons so that at least one month of the three-month average of the
corresponding season is still included in the mean. Exceptions are the CCA with Niño3
and the precipitation in the early dry season at the coast, NTA with the rainfall of both
regions in the same season and Pac-Atl with precipitation in the late wet season in the
inland. The late dry season rainfall at the coast is influenced by large lags of six (except
Niño3), the inland precipitation with large lags only with the NTA. The lags with the
highest goodness indices differ from the ones with the highest correlation coefficients by
63
maximally two lags. Every oceanic region has one exception to this pattern.
The spatial loadings of the Niño regions show contrary anomalies between SST and precipitation, except for late dry season at the coast. The patterns with the NTA are contrary
too in nearly all cases and the Pac-Atl mirrors that with homogeneous SST anomalies in
Pacific and Atlantic, oppositely signed to the rainfall anomalies.
In summary, the highest oceanic influence and the best predictability with the SSTs are
fulfilled with lags between zero and two. Only the late dry season at the coast seems
to be influenced already by the SST of the previous autumn. The spatial loadings show
anomalies which are mainly of contrary signs between the SSTs and the precipitation.
4.2 Canonical Correlation Analysis of the precipitation during
November, December and February
In this section the relationship between the sea surface temperature and the precipitation
of November, December and February is investigated. As shown in Section 3.3, there are
strong differences in the characteristics of the precipitation within these three months.
Additionally, the CCAs of the four seasons show that the oceanic influence varies within
the dry season.
4.2.1 Station data
The first two EOFs of the precipitation enters the CCAs. The first two EOFs of the November precipitation in Vargas explain 57%, and of December and February 66%. The
EOFs of the precipitation in Bolívar represent for November 53%, for December a variance
of 55% and for February 71%. The first two or three EOFs of the are used as input for
the CCA. They represent between 84% and 92% of the total variance. Two EOFs of the
Pacific regions enter the CCA which explain between 95% and 97%. Of Pac-Atl, between
two and four EOFs are chosen to represent the important variance of 61% to 84%. The
remaining EOFs were truncated as noise.
In table 4.10 the results of the CCA coefficients are displayed.
The NTA has a correlation coefficient of 0.07, when correlating with the precipitation of
November in Vargas. The spatial loadings of the CCA time series are shown in figure 4.18.
Positive SST anomalies in the north tropical Atlantic are associated with precipitation
anomalies in Vargas of the same sign.
The correlation of the precipitation of this month with the Niño3 region has a much
higher coefficient of 0.43. The spatial loadings show that positive Niño3 SST anomalies are
associated with negative precipitation anomalies the central coast of Venezuela. Niño3.4
64
4.2 Canonical Correlation Analysis of the precipitation during November, December and
February
Table 4.10: Ranking of the correlation coefficients from the Canonical Correlation Analyses of
the sea surface temperature (ERSST) and the precipitation in Vargas (station data) of November,
December and February
Niño3
Niño3.4
Pac-Atl
NTA
Feb (0.23) Nov (0.43) Nov (0.49) Feb (0.45)
Dec (0.21) Feb (0.39) Feb (0.43) Nov (0.45)
Nov (0.07) Dec (0.35) Dec (0.32) Dec (0.40)
November
December
February
Niño3.4 (0.49) Pac-Atl (0.40) Pac-Atl (0.45)
Pac-Atl (0.45)
Niño3 (0.35) Niño3.4 (0.43)
Niño3 (0.43) Niño3.4 (0.32) Niño3 (0.39)
NTA (0.21)
NTA (0.23)
NTA (0.07)
(a)
(b)
Figure 4.18: Spatial loadings of SST in the NTA region (a) and of precipitation in Vargas (b)
from the Canonical Correlation Analysis in November (station data)
influences even a bit more strongly with 0.49. The spatial loadings show the same results as
for the CCA with Niño3, opposite signed SST and rainfall anomalies. Strongest anomalies
are found in the western part of Niño3.4. The correlation coefficient of Pac-Atl is of the
same scale as the Pacific regions with 0.45. The SST pattern, shown in figure 4.19, has
negative SST anomalies in the Pacific, positive anomalies at the coasts of the NTA and
negative anomalies again in its centre. That underlines that the impact of the SST of the
equatorial Pacific is much stronger than the one of the NTA.
Figure 4.19: Spatial loadings of SST in the Pac-Atl region from the Canonical Correlation Analysis in November (ERSST)
65
In December the order of the coefficients is different. The highest correlation is now found
with the Pac-Atl, 0.40, followed by the Niño regions with 0.35 and 0.32. Correlating the
NTA, the first spatial pattern of the precipitation cannot be explained physically. The
second CCA mode has a correlation coefficient of only 0.21. The loadings (figure 4.20) are
different than the ones of November. The anomalies of precipitation are again positive but
the SST shows a dipole. Positive anomalies in the west and negative ones in the east.
(a)
(b)
Figure 4.20: Spatial loadings of SST in the NTA region (a) and of precipitation in Vargas (b)
from the Canonical Correlation Analysis in December (ERSST and station data)
Niño3 (0.35) and Niño3.4 (0.32) show the same patterns, with homogeneous SST anomalies
of positive sign and negative precipitation anomalies at all stations. The spatial loadings
of Pac-Atl (0.40) are positive in the Pacific too. The NTA of CCA of Pac-Atl has slightly
positive values but a small region of negative SSTs in front of Africa and at the delta of
the Amazon.
The correlation of the precipitation of February with the NTA provides a correlation coefficient of 0.23 (second CCA mode). Negative anomalies are located in the eastern NTA
and in the precipitation. Again, the SST of the Pacific have a stronger influence. Niño3,
with a coefficient of 0.39, and Niño3.4 (0.43) show anomalies of positive sign in SST and
negative ones in precipitation. At the western boundary of the Niño3 domain, and the
eastern boundary of the Niño3.4 domain the anomalies are less strong than in the rest of
the areas. Thus, the highest anomalies are located in the central western equatorial Pacific.
Pac-Atl has the highest correlation coefficient with 0.45.
Figure 4.21: Spatial loadings of SST in the Pac-Atl region from the Canonical Correlation Analysis in February (ERSST)
66
February
The figure 4.21 shows the spatial loadings of the SST. Strong positive anomalies are located in the equatorial Pacific. The Atlantic part of Pac-Atl has opposite signed anomalies
in its centre and a region with positive anomalies in its southwest. This corresponds with
the pattern of the NTA correlated separately. The precipitation and SST anomalies in
the centre of the Atlantic are again of the same sign and oppositely to that of equatorial
Pacific. Comparing with the spatial loadings of the CCA with November rainfall, it is a
very similar pattern.
In summary, the influence of the Pacific is strongest in December and February. The Niño
regions have always higher correlation coefficients than the NTA. The influence for every
single region do not change strongly within these months. Only the influence of the NTA
in November can be neglected.
Table 4.11 shows the results of CCAs with the precipitation of this three months in Bolívar.
the sea surface temperature (ERSST) and the precipitation in Bolívar (station data) of November,
NTA
Niño3
Niño3.4
Pac-Atl
Feb (0.26) Feb (0.68) Dec (0.56) Dec (0.53)
Nov (0.18) Dec (0.57) Nov (0.52) Nov (0.47)
Dec (0.07) Nov (0.45) Feb (0.38) Feb (0.39)
November
December
February
Niño3.4 (0.52) Niño3 (0.57)
Niño3 (0.68)
Pac-Atl (0.47) Niño3.4 (0.56) Pac-Atl (0.39)
Niño3 (0.45)
Pac-Atl (0.53) Niño3.4 (0.38)
NTA (0.07)
NTA (0.26)
NTA (0.18)
The results of the CCAs of November in Bolívar are quite similar to the ones in Vargas.
The CCA with the NTA provides a correlation coefficient of only 0.18 (second CCA mode).
The anomalies of the SST are negative with strongest ones in the western half. The anomalies of precipitation are positive. This stands in contrast to the result of Vargas where both
patterns show anomalies of the same sign. Higher canonical correlation coefficients can be
found by correlating the Pacific with the November rainfall in Bolívar. Niño3.4 provides
the highest one with 0.52. The SST anomalies of this region are negative, more negative
at the eastern boundary of its domain and the precipitation anomalies are positive. In the
Niño3 region, where the correlation coefficient is 0.45, the spatial loadings show strong and
homogenous positive anomalies in the SSTs and negative anomalies in the rainfall. Both
areas in the Pacific have anomalies of SST and rainfall with contrary signs as mentioned
before in Vargas. The correlation coefficient of the CCA with Pac-Atl is similar to the
ones of Niño3 and Niño3.4 with 0.47. The pattern corresponds with the ones of the region
67
in the Pacific. The SST anomalies of the Pacific is negative, while the loadings of the
precipitation are positive. The Atlantic SST anomalies of Pac-Atl are divided in a part
with positive anomalies in the western half and a part with negative ones in the eastern
side.
In December, again the NTA has the lowest influence on the precipitation in the state Bolívar. It has a correlation coefficient of only 0.07 (second CCA mode). The spatial loadings
are of the same sign in SST and precipitation, negative. Niño3 has the highest influence
on the December precipitation has Niño3 with 0.57. The pattern of the spatial loadings
is inhomogeneous. The anomalies are negative with strongest values in the eastern part.
The pattern of the precipitation shows positive anomalies. Both can be seen in figure 4.22.
(a)
(b)
Figure 4.22: Spatial loadings the SST in the Niño3 region (a) and of precipitation in Bolívar (b)
from the Canonical Correlation Analysis in December (ERSST and station data)
The correlation with Niño3.4 has nearly the same coefficient (0.56). The anomalies of SST
and precipitation are of opposite sign as well. Here the SST has positive anomalies and the
precipitation negative ones. The region Pac-Atl has a slightly lower correlation coefficient
of 0.53. The loadings show negative anomalies of precipitation and positive anomalies in
SST. The anomalies are stronger in the Pacific than in the Atlantic where even a small
area at the Amazon delta is negative. This underlines that the influence of the Pacific is
much stronger than the impact of the Atlantic on December precipitation in Bolívar.
68
February
The CCA of SST of the NTA and precipitation in February provides a higher correlation
coefficient compared to the coefficients of November and December with 0.26 (second CCA
mode). In figure 4.23 the spatial loadings can be seen. In the western NTA part the anomalies are negative and in the eastern positive. The loadings of the precipitation have the
same sign like the eastern area of the NTA.
Figure 4.23: Spatial loadings of SST in the NTA region from the Canonical Correlation Analysis
in February (ERSST)
The coefficient of the correlation with the Niño3 region is quite high with 0.68. The SST
has strong positive anomalies and the precipitation has slightly negative anomalies. The
canonical correlation coefficients of Niño3.4 and Pac-Atl are very similar to 0.38 and 0.39.
The spatial loadings of the analysis with Niño3.4 have the same signs like with Niño3 but
here the precipitation anomalies are stronger. The loadings of the CCA with Pac-Atl show
a dipole and correspond perfectly to the patterns of the three regions separately. There
are negative anomalies in the Pacific, the main part of the NTA has positive anomalies
and the precipitation has homogeneous positive anomalies.
In Bolívar the influence of the regions in the Pacific is in all three months higher than the
ones of the NTA. In most of the cases the anomalies of the Pacific have an opposite sign
to the ones in the NTA and the anomalies of the precipitation.
The zero-lag goodness indices of the north tropical Atlantic are always very low. The
index is between -0.069 and 0.071. The goodness indices of Niño3 are between 0.182 for
the precipitation in Vargas during February and 0.320 for the rainfall in Bolívar in December. Always very similar to this are the indices of the Niño3.4 region and Pac-Atl. They
vary from 0.160 to 0.330 and 0.189 to 0.331 in the same order of the months like with
Niño3.
Not only the higher correlation coefficients of the CCAs with the regions of the equatorial
Pacific, but also the goodness indices underline the greater influence of the Niño regions
and furthermore the better predictability of Venezuelan rain during these three months in
the dry season.
The results of the lags of the CCAs can be seen in table 4.12.
The correlation coefficients and goodness indices of each lag are listed in tables A.30 - A.35
(appendix).
69
Table 4.12: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature (ERSST) and the precipitation (station data) of
November, December and February, SST leading the precipitation up to six months
Month and region
November Vargas
December Vargas
February Vargas
November Bolívar
December Bolívar
February Bolívar
NTA
5|5
5|2
1|2
3|6
3|6
5|6
Niño3
3|1
5|5
2|0
5|1
3|2
0|4
Niño3.4
0|0
6|6
0|1
0|0
0|2
1|1
Pac-Atl
2|2
5 | 4;5
1|1
2|2
5|5
6|6
In Vargas the highest correlation coefficient with the November precipitation and the SST
of the NTA occurs in June while the regions in the Pacific and Pac-Atl influence simultaneously or with a lag of maximal three months. For the precipitation of December the
lags of all regions are high. NTA has its strongest influence during July like Niño3 and
Pac-Atl, Niño3.4 in June. These big differences between the influence of the Pacific regions
and Pac-Atl on the precipitation of November and December cannot be explained. The
rainfall in February is influenced strongest by all regions at similar lags. The highest influence occurs between December (Niño3) and February (Niño3.4). The lags of the highest
goodness indices are at the same lag or differ up to two months compared to the ones of
the highest coefficients with three exception for NTA and two for Niño3.
In Bolívar the influence of the NTA is for all three months of precipitation quite long in
advance (between three and five months) and the best goodness index always appears at
the sixth lag. In all three months of precipitation the lag of the highest coefficient with
Niño3 and Niño3.4 does not differ much and is close to the month of precipitation, except
for the influence of Niño3 on the precipitation in November (lag five). In contrast, the
strongest influence of Pac-Atl varies between lag two and lag six (July - September) which
is during the wet season.
The spatial loadings at the lags with the highest correlation coefficients and goodness indices of the CCAs with the NTA all have anomalies of the same sign between SST and
precipitation. When correlating the Niño regions, the anomalies of SST and precipitation
have opposite signs. The patterns of the correlations with Pac-Atl summarise the ones of
the three regions.
In short, the lags of the highest correlation coefficients of the CCAs with Niño3 and Niño3.4
and the precipitation in Vargas are always similar but quite different compared to the NTA
region. The Pacific influences strongest with small or zero lags the precipitation of November and February while the rainfall of December is associated most strongly with the
SSTs five or six months in advance. In contrast, the NTA influences the precipitation of
November and December with a large lag of five months and the precipitation of February
with one lag. Correlating the precipitation of Bolívar, always at least one of the Niño
70
February
regions has its strongest influence after the NTA. The lags with the highest coefficients
differ strongly between the Niño regions and between the months. The lag of NTA and
Pac-Atl are never simultaneous. The spatial loadings always fulfil the expected scheme.
4.2.2 Model data of the past
The monthly precipitation of the months November, December and February at the coast
and in the inland is correlated with the SST of the regions Niño3 and Niño3.4 in the equatorial Pacific, the NTA and Pac-Atl.
EOFs which explain at least 1% variance were used for the CCAs. For all three months
of the data at the coast all EOFs (three) enter the CCAs. So that 100% of the variance
remains and no data are rejected. For the inland, the EOFs represent 98% for November
(six EOFs), 99% of the variance of December (six EOFs) and 99% for February (eight
EOFs). For the CCAs with the NTA between seven and eight EOFs are used as input.
They represent between 93% and 94% of the total variance. Between four and five EOFs
enter the CCAs of the Niño3 region, with an explained variance of 96% or 97%. The three
or four first EOFs explain between 97% and 98% variance of the Niño3.4 region. For PacAtl between eleven and twelve EOFs represent a variance of at least 1%, which makes a
total variance of 90%. The remaining EOFs were truncated as noise. The different amount
of EOFs with importance show the difference between the SST regions. The Niño regions
always have a strong leading EOF of at least 70%, the NTA has a less strong one and
the Pac-Atl region only around 44% or 46%. This shows that the variability is higher in
the NTA and especially in the region which combines both oceans. This agrees with the
analyses of the four seasons.
Table 4.13 shows the canonical correlation coefficients of the rainfall of November, December and February in the coastal region.
the sea surface temperature and the precipitation at the coast of November, December and February
(model data of the past)
NTA
Niño3
Niño3.4
Pac-Atl
Nov (0.69) Feb (0.35) Feb (0.36) Nov (0.78)
Dec (0.43) Dec (0.31) Nov (0.34) Dec (0.54)
Feb (0.36) Nov (0.30) Dec (0.26) Feb (0.38)
November
December
February
Pac-Atl (0.78) Pac-Atl (0.54) Pac-Atl (0.38)
NTA (0.69)
NTA (0.43)
Niño3.4 (0.36)
Niño3.4 (0.34) Niño3 (0.31)
NTA (0.36)
Niño3 (0.30) Niño3.4 (0.26) Niño3 (0.35)
In November the correlation is highest with the Pac-Atl, followed by the NTA and then by
71
the Niño regions. The NTA has a correlation coefficient of 0.69. The spatial loadings of
the SST show slightly negative anomalies in nearly the whole area with just some positive
anomalies in the southeast and northwest corners of the region. The loadings of the precipitation anomalies are negative with stronger anomalies in the two western grid boxes.
The correlation with Niño3 has a coefficient of 0.30. The leading modes have homogenous
patterns with opposite signs. The spatial loadings of the precipitation look exactly like
the ones correlated with the NTA while the ones of the Niño3 are positive. The correlation results of Niño3.4 are very similar to the ones of the Niño3 region. The coefficient is
slightly higher with 0.34, the SST spatial loadings are positive in the whole SST region and
the precipitation anomalies are negative, with higher values in the east, but they are less
strong. The Pac-Atl, with the highest correlation of 0.78, combines the spatial loadings of
the regions. In figure 4.24 it can be seen that the Pacific has negative anomalies while the
major part of the NTA has positive SST anomalies and the precipitation anomalies are
positive too. Again the precipitation anomalies are stronger in the two grid boxes in the
west. The spatial loadings correspond with the ones of the station data.
(a)
(b)
(b) from the Canonical Correlation Analysis in November (model data of the past)
In December the correlation coefficients are lower, except Niño3 with 0.31. The order stays
in general the same. The combined region Pac-Atl has the highest correlation coefficient
with 0.54, followed by the NTA with 0.43 and the by the Niño regions. Now the Niño3.4
has the lowest coefficient with only 0.26. The spatial loadings of the CCA with the NTA
(figure 4.25) show negative anomalies in SST and precipitation.
(a)
(b)
from the Canonical Correlation Analysis in December (model data of the past)
72
February
The Niño regions have a very similar result. The spatial loadings are homogeneous with
opposite sign in the two variables. In addition, the spatial loadings of the Pac-Atl are
similarly compared to the previous month. There are negative anomalies in the Pacific
and positive ones in the NTA and precipitation at the central coast of Venezuela. The
rainfall anomalies are higher in the east than in the west.
The correlation coefficients are similar of all four regions in February. Again Pac-Atl has
the highest with 0.38. The NTA and Niño3.4 have the same with 0.36 and Niño3 has a
coefficient of only 0.01 less. The first CCA mode of the correlation with the NTA provides
patterns of the spatial loading with opposite signs in the three grid boxes of the precipitation. This cannot be explained physically so that the second CCA mode is discussed here.
The spatial loadings are shown in figure 4.26. The SST has positive anomalies in the east,
in front of the coast of Africa, and negative ones in the western half. The corresponding
precipitation anomalies are positive with stronger values in the west.
(a)
(b)
from the Canonical Correlation Analysis in February (model data of the past)
Again, the spatial loadings of the Niño regions are the same. Positive SST anomalies and
negative precipitation anomalies. The anomalies of the correlation with Pac-Atl support
the results of the Niño regions. The SST anomalies are negative in the whole area, Pacific
and Atlantic, and positive in all three grid boxes of the precipitation.
In short, the correlation coefficients of NTA are higher than the ones of the Niño regions in
November and December and the similar in February. The Pac-Atl always has the highest
coefficient. The spatial loadings of November and December have anomalies of the same
sign with the NTA and of the opposite sign with the Pacific regions, Pac-Atl shows the
same. In February, the NTA has a dipole in the spatial loadings, the sign of the anomalies
agrees between the precipitation and the SST in the eastern half of the area. The spatial
loadings of the Pac-Atl show overall negative anomalies with corresponding positive precipitation anomalies. This supposes that the influence of the NTA diminishes in February.
The same correlations are performed with the precipitation data of the inland. The correlation coefficients and their order can be seen in table 4.14.
The results of the correlations with November precipitation in the inland are similar to the
73
sea surface temperature and the precipitation in the inland of November, December and February
(model data of the past)
NTA
Niño3
Niño3.4
Pac-Atl
Nov (0.67) Nov (0.57) Feb (0.34) Nov (0.79)
Dec (0.56) Feb (0.38) Nov (0.31) Dec (0.66)
Feb (0.17) Dec (0.16) Dec (0.27) Feb (0.54)
November
December
February
NTA (0.67)
NTA (0.56)
Niño3 (0.38)
Niño3 (0.57) Niño3.4 (0.27) Niño3.4 (0.34)
Niño3.4 (0.31) Niño3 (0.16)
NTA (0.17)
ones at the coast. Again, the Pac-Atl region has the highest correlation with 0.79, followed
by the NTA with 0.67 and the Niño regions. However, the coefficients of the Niño regions
differ quite strongly. Niño3 has a much higher influence with a coefficient of 0.57 while
Niño3.4 has one of only 0.31. The spatial loadings of the CCA of the NTA are negative
in both fields. Only a little area at the coast of Africa has positive SST anomalies. The
precipitation anomalies are strong with the exception of the grid box with the highest
orography. The CCA with Niño3 shows positive SST anomalies with highest values in the
centre of the area. The anomalies of the precipitation have the opposite sign with lowest
values in the northwest, increasing to the southeast. In general, the spatial loadings do
not differ strongly between the two Niño regions. The signs are the same but the SST
anomalies are more homogeneous with the Niño3.4 region. The spatial loadings of the
rainfall have stronger values with highest values in the southeast and centre and lowest in
the northeast. The loadings of the Pac-Atl (figure 4.27) show a summary of the previous
results. The anomalies of the Atlantic part and the rainfall have anomalies of the same
sign and the eastern equatorial Pacific anomalies of the opposite sign. The precipitation
anomalies are highest in the grid boxes of intermediate altitude (420 m and 490 m), in the
centre and in the southeast.
In December, the order of the correlation coefficients starts again with the Pac-Atl with a
coefficient of 0.66, followed by the NTA with 0.56 and the Niño regions. For this month the
correlation with Niño3.4 is higher with 0.27 (second CCA mode) compared to Niño3 with
only 0.16. The spatial loadings of the CCA with the NTA can be seen in figure 4.28. The
major part of the NTA shows negative anomalies as well as the precipitation in the whole
chosen region in the inland with lowest values in the grid box with the highest elevation.
The spatial loadings of Niño3.4 are shown in figure 4.29. The first CCA mode was eliminated because of its meaningless pattern of precipitation. The loadings show negative
anomalies in the SST and positive ones in the precipitation with strongest values in the
south.
74
February
(a)
(b)
(b) from the Canonical Correlation Analysis in November (model data of the past)
(a)
(b)
Figure 4.28: Spatial loadings of SST in the NTA region (a) and of precipitation in the inland (b)
from the Canonical Correlation Analysis in December (model data of the past)
The spatial loadings of the CCA with Pac-Atl correspond with the results of the regions
separately and with the results of November. The anomalies of NTA and rainfall have the
same sign and are opposite to the ones in eastern equatorial Pacific.
75
(a)
(b)
Figure 4.29: Spatial loadings of SST in the Niño3.4 region (a) and of precipitation in the inland
(b) from the Canonical Correlation Analysis in December (model data of the past)
The correlations for February exhibit a different order in the strength of the oceanic influence compared with November and December. However, Pac-Atl has the highest coefficient with 0.54 but for this month it is followed by the Niño regions (Niño3: 0.38 and
Niño 3.4: 0.34) and NTA has the smallest influence with 0.17. The spatial loadings of
the correlation with the NTA show anomalies which interrupt the common patterns. The
anomalies of SST and precipitation have an opposite sign: Positive anomalies in the SST
with highest values in the centre and negative anomalies in the inland with highest values
in the southwest. The same patterns appear in the spatial loadings of Niño3 and Niño3.4
and are summarised in the patterns of Pac-Atl (figure 4.30). The Pacific and the Atlantic
show positive anomalies while the rainfall anomalies are negative.
To sum up, the influence of the north tropical Atlantic on the precipitation in November
and December is higher than of the Niño regions and the coefficient of Pac-Atl is always the
highest. Furthermore, the spatial loadings of these two months show anomalies in rainfall
and the SST in the NTA region of the same sign and in the Niño regions of opposite sign
to them. In February, all regions have positive SST anomalies corresponding to negative
precipitation anomalies.
The goodness indices of the zero-lag correlation of November have a similar order like
the correlation coefficients. Pac-Atl has the highest with 0.567, followed by the NTA with
0.523, Niño3 with 0.215 and Niño3.4 with 0.163. In December the indices are lower and
76
February
(a)
(b)
(b) from the Canonical Correlation Analysis in February (model data of the past)
closer together. NTA has a goodness index of 0.226 similar to Niño3 with 0.240. Again,
the Niño3.4 region has the lowest with 0.191 and Pac-Atl the highest with 0.357. NTA
and Pac-Atl have their lowest indices in February with -0.105 (NTA) and 0.252 (Pac-Atl).
The Niño regions have their highest indices with 0.292 (Niño3) and 0.309 (Niño3.4). This
corresponds with the results of the correlation coefficient.In the inland the goodness index
of NTA decreases with the development of the dry season. In November the index is 0.506,
in December 0.127 and in February only 0.034. The other three regions have the highest
index in February and lowest in December.
In summary, the rainfall at the coast in November and December, can be predicted the
best with both oceans together and also good with the NTA. In February the Niño regions
make a better forecast. In the inland the precipitation in November can be predicted quite
well with Pac-Atl or NTA, December is difficult with all regions and in February the Niño
regions and Pac-Atl give a hint of how the precipitation will be.
Table 4.15 lists the lags which have the highest canonical correlation coefficient and provide
the best forecast (highest goodness index).
All lags with their correlation coefficients and goodness indices can be seen in the appendix
in tables A.36 - A.41.
At the coast, the NTA and Pac-Atl have their highest coefficient and index simultaneously
with the November precipitation, while the Niño regions influence the strongest in June
77
Table 4.15: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature and the precipitation of November, December and
February (model data of the past), SST leading the precipitation up to six months
Month and region
November coast
December coast
February coast
November inland
December inland
February inland
NTA
0|0
1|6
4|6
0|0
1|6
5|6
Niño3
4|5
2|0
6|4
6|6
2|2
5|6
Niño3.4
5|2
1|1
2|5
4|6
2|2
2|2
Pac-Atl
0|0
1|0
3|6
0|0
6|0
6|6
and July. The rainfall of December is strongly influenced by the SST of November and
October. The goodness indices agrees with that, only ones of the NTAs differs strongly
with the highest index at lag six. The precipitation in February is first influenced by Niño3
(August), then by the NTA (October), followed by the Pac-Atl in November and finally
Niño3.4 in December. The goodness indices vary with two or three months.
In the inland the results are similar. The precipitation in November is strongly influenced
simultaneously by the SST of Pac-Atl and NTA and with lags of four or six respectively by
the Niño regions. The goodness indices are highest in the same months, only with Niño3.4
two months later. The December rainfall is influenced mainly by the SSTs of October
and November, but of the Pac-Atl already in June. For this month of precipitation the
goodness indices do not match with the correlation coefficients (except the Niño3 region).
Unlike the rainfall of February when the lags of both, correlation coefficient and goodness
index, are highest in the same month or only with one month difference.
The CCAs of the NTA and the precipitation in November and December with the highest
coefficients and goodness indices provide SST and precipitation anomalies of the same sign.
With Niño3 the anomalies have opposite signs when index and coefficient are highest except for the correlation coefficient with the precipitation in November at the coast. For the
region Niño3.4, the rainfall in November in the inland is the only exception. The CCAs of
Pac-Atl and November and December precipitation show spatial loadings with anomalies
of the same sign in the NTA and the precipitation and opposite signs in the Niño regions.
For the CCAs with the precipitation of February it matches only with the eastern equatorial Pacific.
In conclusion, the differences between coast and inland are small. The NTA influences
most strongly between September and November. Niño3 has its strongest influence on the
precipitation of December with small lags and on the rainfall of November and February
with high lags. The region Niño3.4 has high lags only with the rainfall in November. The
SST of Pac-Atl shows different results in the lags between coast and inland. The highest
goodness indices of the NTA are at the same lags while correlating the November precipitation, one/two months later with February rain and five months later with December rain.
78
February
The goodness indices of Niño3 always differ with only zero, one or two months from the
correlation coefficient, while Niño3.4 has its highest lags of both simultaneously in some
cases. Furthermore, the lags of goodness and correlation coefficient correspond well when
correlating the SST of Pac-Atl, with the exception of the December rainfall in the inland
when it varies from six (correlation coefficient) to zero (goodness index). The spatial loadings show the common pattern of same signed anomalies in the NTA and precipitation
and anomalies of contrary to this in the Niño regions. In general, the pattern is interrupted
for NTA and November rainfall, for the Niño regions and February precipitation at the
coast and for Pac-Atl and November and February precipitation.
4.2.3 Model data of the future
To subtract the noise, only EOFs which explain at least 1% variance enter the CCAs. For
the precipitation at the coast this criterion allows all EOFs to enter. That means 100%
of the variability is represented. The EOFs which stand for at least 1% in the inland explain 98% of November and December (six EOFs) precipitation and 100% of the rainfall
in February (all nine EOFs). The EOFs (five or six) of the NTA represent 96% variance in
November and December and 97% in February. The explained variance of the Niño regions
is slightly higher with 97% to 98% (three or five EOFs), while of Pac-Atl the seven or eight
remaining EOFs explain only 91% in November and 93% in December and February. Like
in the model data of the past, the leading EOF of the Niño regions explains a higher part
of the total variance than in the NTA and especially in the Pac-Atl. The variability is
stronger in the Atlantic and in the Pac-Atl than in the Niño regions. Suggesting that the
leading mode is the warming of the oceans, there seems to be a stronger trend in the SST
of the Pacific.
Table 4.16 shows the correlation coefficients of the CCAs with the coastal precipitation in
November, December and February and its order within the oceanic regions.
the sea surface temperature and the precipitation at the coast of November, December and February
(model data of the future)
NTA
Niño3
Niño3.4
Pac-Atl
Nov (0.58) Dec (0.48) Nov (0.42) Dec (0.71)
Dec (0.58) Nov (0.42) Dec (0.41) Nov (0.65)
Feb (0.28) Feb (0.35) Feb (0.29) Feb (0.49)
November
December
February
NTA (0.58)
NTA (0.58)
Niño3 (0.35)
Niño3 (0.42)
Niño3 (0.48) Niño3.4 (0.29)
Niño3.4 (0.42) Niño3.4 (0.41)
NTA (0.28)
79
The November precipitation is influenced most strongly by the Pac-Atl region with a coefficient of 0.65. The second strongest influencer is the NTA with 0.5. Both Niño regions
have lower correlation coefficients with 0.42. The patterns of the spatial loadings look the
same for all SST regions. The anomalies of the SST are positive and the precipitation
anomalies negative. The SST anomalies in the Pacific are stronger than in the Atlantic,
the precipitation anomalies are strong in the middle grid box and additionally in the east
when correlating with the NTA.
The correlation with the precipitation in December have coefficients of the same order
within the oceans as in November. The highest coefficient has again the CA with Pac-Alt
with 0.71. The coefficient of the CCA with the NTA is 0.58, while the Niño regions again
have lower ones with 0.50 (Niño3) and 0.41 (Niño3.4). The spatial loadings of the NTA
show a dipole with positive anomalies in the east, negative ones in the west and in the
precipitation at the coast. The spatial loadings of the CCAs with the Niño regions have
opposite anomalies in SST and rainfall. The spatial loadings of the Pac-Atl can be seen in
figure 4.31. The anomalies are positive with the exception of a small area in the Caribbean
Sea at the coast of Venezuela. The precipitation anomalies are homogeneously negative.
(a)
(b)
(b) from the Canonical Correlation Analysis in December (model data of the future)
The correlation coefficients of the CCAs with the precipitation in February are lower than
in the two months discussed before. Again Pac-Atl has the highest influence with a coefficient of 0.49, followed by Niño3 with 0.35 and Niño3.4 with 0.29. The NTA has the
lowest correlation coefficient with 0.28. The spatial loadings of the CCA with the NTA
show negative anomalies in both fields, with the Niño regions anomalies of opposite signs
between SST and precipitation. The spatial loadings of Pac-Atl (figure 4.32) show SST
anomalies of the Pacific like during El Niño. A warm tongue reaches from the coast of
Ecuador and Peru westward into the equatorial Pacific, surrounded by negative anomalies.
The SST of the Atlantic is colder with the exception of the Caribbean Sea at the east coast
of Central America. The spatial loadings of the precipitation show rainfall which is less
than normal. This agrees with previous studies.
In short, the influence of the oceanic regions on the precipitation at the coast of Venezuela
is stronger in November and December and weaker in February. In November and Decem-
80
February
(a)
(b)
(b) from the Canonical Correlation Analysis in February (model data of the future)
ber, the influence of the NTA is higher than the one of the Niño regions. These results do
not reflect those of the early dry season discussed before. The Pac-Atl has the strongest
impact on the precipitation during all the months. Contrary signs of the anomalies between SST and precipitation occur in all three months with the Niño3, in November and
December with the Niño3.4 and in November with the NTA.
In table 4.17 the coefficients of the CCAs with the precipitation in the inland are listed.
sea surface temperature and the precipitation in the inland of November, December and February
(model data of the future)
NTA
Niño3
Niño3.4
Pac-Atl
Dec (0.62) Nov (0.59) Dec (0.60) Dec (0.79)
Nov (0.41) Dec (0.51) Nov (0.55) Nov (0.71)
Feb (0.41) Feb (0.48) Feb (0.40) Feb (0.61)
November
December
February
Niño3 (0.59)
NTA (0.62)
Niño3 (0.48)
Niño3.4 (0.55) Niño3.4 (0.60)
NTA (0.41)
NTA (0.41)
Niño3 (0.51) Niño3.4 (0.40)
The November precipitation is strongly influenced by the Pac-Atl with a coefficient of 0.71.
In contrast to the coastal rainfall, the correlations with the Niño regions have higher coefficients than the NTA with 0.59 (Niño3) and 0.55 (Niño3.4) compared to 0.41 (NTA).
The patterns of the spatial loadings correspond within all regions. The SST anomalies are
positive and the rainfall anomalies negative, exactly as at the coast.
The correlations with the precipitation in December have higher coefficients than in November except for Niño3. Pac-Atl has the highest values (0.79), followed by the NTA with
0.62. Nearly the same strength of influence has the Niño3.4 region with 0.60. Niño3 has a
lower one with 0.51. The NTA show a dipole in the spatial loadings with positive values
in the west and negative ones in the east. The precipitation anomalies are positive with
81
highest values in the southwest. Also the results of the anomalies of the Niño regions agree
with the ones of the correlations with the rainfall at the coast, contrary signed anomalies
between SST and precipitation. The spatial loadings of the Pac-Atl summarise these, negative anomalies in the whole SST area with stronger ones in the Pacific and positive ones
in the Caribbean Sea.
The February precipitation is most strongly influenced by the Pac-Atl (second CCA mode).
Like at the coast, the impact of Niño3 is the strongest of all the small regions, with 0.48.
The NTA provides a correlation coefficient of 0.41, similar to Niño3.4 with 0.40. The
patterns of the spatial loadings show the same as for November, except the anomalies of
Pac-Atl and its precipitation (shown in figure 4.33).
(a)
(b)
(b) from the Canonical Correlation Analysis in February (model data of the future)
To sum up, the influence of NTA, Niño3 and Pac-Atl is strongest in December, of Niño3.4
in November and lowest for all regions in February. The impact of Pac-Atl is always the
highest, followed by NTA in December and Niño3 in November and February. The spatial
loadings of the correlations with the Niño regions and with the NTA in November and
February always show contrary signs in the anomalies of SST and precipitation.
The goodness indices for the zero-lag correlations are always highest with Pac-Atl, followed by the Niño regions except for the coastal February rain when Niño3.4 provides the
lowest predictability. These results stand in contrast to the ones of the model data of the
past. In the past, the goodness indices are highest for the regions which influence most
82
February
strongly (highest coefficients).
The lags with the highest canonical correlation coefficients and goodness indices are displayed in table 4.18.
Table 4.18: Lags of the highest correlation coefficients|goodness indices from the Canonical Correlation Analyses of the sea surface temperature and the precipitation of November, December and
February (model data of the future), SST leading the precipitation up to six months
Month and region
November coast
December coast
February coast
November inland
December inland
February inland
NTA
0|1
0|0
6|6
2|1
1|0
5|4
Niño3
5|1
6|6
6|6
1|1
6|6
0|4
Niño3.4
4|4
6|2
2|4
1|1
0|1
0|6
Pac-Atl
0|1
0 | 0;3
5|0
0|1
0|4
1|0
The detailed tables of coefficients and indices are shown in the appendix in tables A.42 A.47.
NTA and Pac-Atl influence the rainfall in November and December at the coast simultaneously and that of February with lags of six and five. The lags with the highest goodness
indices correspond with those results, except for Pac-Atl correlated with the February precipitation. The Niño regions always influence with a big lag of four, five or six while the
lag is small with Niño3.4 and the precipitation in February. The lags of the highest indices
do not match well, only in December (Niño3) and November and February (Niño3.4).
For the inland precipitation, the results of NTA and Pac-Atl are similar compared to the
coast. Niño3 influences with small lags in November and December and Niño3.4 only in
November. The lags of the highest goodness index vary from the ones of the highest coefficient for the CCAs of the Niño regions and rainfall in February and of Pac-Atl and the
precipitation in December.
The spatial loadings at the highest coefficient and index show anomalies of contrary signs
for the Niño regions except for the correlations with rainfall in February at the coast.
The anomalies of the NTA and the precipitation correspond in their signs for December
and February rainfall except for the loadings of the lag with the highest coefficient in the
inland. The loadings of the CCA with Pac-Atl only match with the expected pattern of
anomalies with the same sign in the NTA and the precipitation and contrary ones in the
Pacific when correlating the December rainfall at the coast and inland (only coefficient)
and for the February rainfall the lag with the highest coefficient at the coast.
In summary, the differences between coast and inland are small. Mainly, the precipitation
of November and December is influenced simultaneously or with small lags. February rainfall is influenced with higher lags at the coast by all SST regions, in the inland only by
the NTA. In the inland the lags are small for the region with strong influence. The lags of
83
the index match with the ones of the coefficient in all cases for the NTA and in four cases
for the Niño regions and for Pac-Atl. The spatial loadings correspond with the expected
pattern for the Niños (not for February at the coast) and for the NTA and Pac-Atl mainly
only in December.
4.3 Summary and Conclusion
Comparison of model data, past and future:
For the four seasons, there is a clear change in the order of seasons looking at the zero
lag coefficient. In the past, the order is mainly: late wet season, early dry season, early
wet season and late dry. This changes in the future, then the early wet season has higher
connection to the rainfall than the early dry season. At the coast, Pac-Atl is, in the past
and the future, the strongest influencer and NTA the second with two exceptions, late
dry season in the past and early dry season in the future. In the inland, the influence of
Pac-Atl decreases in the early dry and late wet season, of the NTA in the dry seasons and
gains impact on the precipitation of the late wet season.
The common pattern in the spatial loadings is always fulfilled for the Niño regions. In
the past, it does not match for the spatial loadings of the CCA with NTA and late dry
season rainfall at the coast and correlated with the precipitation of early wet season in
both regions of Venezuela. In the future, the patterns agree with the expected one of the
NTA only in the late dry season at the coast.
The oceanic influence of Pac-Atl is stronger in November and December in both time periods, of the Niño regions only in the future. The influence of Pac-Atl is in all these three
months highest in the past and the future. The comparison of NTA and the Niños does
not make a difference at the coast between past and future (NTA coefficients are higher in
November and December). In the inland NTA remains stronger than the Niños in December, the influence diminishes in November and increases in February from the past to the
future.
The spatial loadings of the zero-lag correlations show anomalies of the same sign in the
NTA and the precipitation and contrary signs in the Pacific regions and the precipitation.
In the past, this pattern has only two exceptions, the correlation with Pac-Atl and rainfall in February at the coast and NTA and Pac-Atl with precipitation in February in the
inland. In the future this pattern is not fulfilled in November (Pac-Atl and NTA) in both
precipitation regions and for the inland rainfall in the February (Pac-Atl and NTA).
The CCAs with the highest coefficients and goodness indices are mainly at small lags. The
late dry season in the future at the coast seems to be an exception. The lags with the
highest correlation coefficients agree between past and future for the NTA and for Pac-Atl
at the coast. The other regions differ without system. The patterns of the spatial loadings
84
4.3 Summary and Conclusion
at the lags with the highest influence do not match with the pattern suggested by previous
studies in the CCAs with NTA and Pac-Atl and the precipitation of February and the late
seasons in the past. In the future, the anomalies of the Atlantic are only of the same sign
like the precipitation of December when correlating Pac-Atl and NTA.
Comparison of station and model data:
In most cases, the correlations with the station data show a stronger influence of the Niño
regions than of the NTA while the model data reveal exactly the opposite. The station
data reveal strong differences between the two regions in Venezuela. There is more precipitation in Bolívar, the correlation coefficients are higher too. The model data results
do not show these differences. In general, the correlation coefficients are higher with the
model data in the wet seasons. Maybe the variance is smaller so that the scheme is clearer
and the predictability is higher too.
The goodness index of the zero-lag CCAs with the NTA of the model data (past) is in
most cases higher than of the station data, especially in November, December, the early
dry season and the wet seasons, comparing the station data with model data of the future,
additionally in the late dry season. For the Niño regions and Pac-Atl, the indices with
the model data are higher than the ones of the station data for the CCAs of the late dry
season and the wet seasons and, additionally, compared to the future, of November and
December.
The results of the lags with the highest coefficients and indices are more clear and structured with the model data and seem to be more reliable. The results do not correspond
between CCAs of the model data and the station data. In most cases, CCAs of the model
data of the past and of the future show both highest coefficients and indices at small lags,
the exception differ between both data sets.
The spatial loadings of the station data show, except for the rain in February and the
early dry season, that positive SST anomalies in the NTA and negative anomalies in the
equatorial Pacific provide rains which are stronger than usual in Venezuela. The model
data does not show this so clearly in the NTA and thereby in the Atlantic part of Pac-Atl.
The data of the past have a anomalies which differ from that pattern in late wet season, the
future data do not fulfil it at all. The exceptions to this pattern increase from station data
to model data past and future, also for the spatial loadings at the lags with the highest
correlation coefficient and goodness index.
Comparison of the two Niño regions:
The comparison of the two Niño regions provides different results for model and station
data. With the station data Niño3 has higher zero-lag correlation coefficients in the early
dry and in December with rainfall in Vargas and Bolívar and additionally in February
85
correlated with precipitation in Bolívar. In Vargas, the difference between the two Pacific
regions is small, in the second decimal place, while with rainfall in Bolívar the distinctions
are bigger in the early wet season and in February when Niño3.4 has a stronger influence.
Correlating the model data of the past and the coast rainfall, the eastern Pacific region has
a higher impact in the late seasons and December, with small differences. The inland rain
is more strongly influenced by the Niño3 region in the early seasons, the late wet season,
November and February, with differences in the first decimal place in the early dry season
and November.
In the future, the influence of the Niño3 region is stronger in most cases, only in December
and the late wet season in the inland is the coefficient of Niño3.4 higher. The differences
are all very small. To sum up, with five exceptions the differences between the two Niño
regions can be neglected.
The explanation for the patterns of the spatial loadings which occur most of the time
in the station data and in the Pacific regions of the model data is given in the following.
La Niña provides a positive rainfall anomaly in Venezuela. This can be explained by the
zonal and the meridional atmospheric circulation, the Walker Circulation and the Hadley
Cell. The Walker Circulation (January) during La Niña leads to the ascent of air over
South America, stronger and faster than during the normal phase of ENSO (Lau et al.,
2002). The winter Hadley Cell (December-February) is weakened and has its branch of
ascending air anomalously north of the equator (Quan et al. 2004). The upward motion,
in addition supported with positive NTA SST anomalies, leads to positive precipitation
anomalies in northern South America, at least in the early dry season.
Furthermore, the Subtropical Jetstream influences the intensity of the precipitation. The
Subtropical Jetstream generates a subsidence which causes the subtropical desert regions
(Martelo, 2003,I). During the La Niña phase, this high wind speed air current is located
more in the north than normally. This diminishes the subsidence over Venezuela and enables convection (Arévalo, personal communication).
The coherence of La Niña and the precipitation in Venezuela are being studied at the
moment and not fully investigated yet.
86
5
Extreme events in precipitation in Venezuela
The previous chapter showed that there is a connection between the sea surface temperatures of the north tropical Atlantic and the east equatorial Pacific and the precipitation
in Venezuela. Furthermore, it becomes clear that negative anomalies in this part of the
Pacific and positive anomalies in the NTA favour positive rainfall anomalies in Venezuela.
In the following section, the relationship between extreme events in SST and precipitation
is analysed statistically.
87
5.1 Statistic of extreme events in precipitation and sea
surface temperature
Statistics are collected to see how the presence of negative SST anomalies in the Niño
regions (La Niña) or/and positive SST anomalies in the NTA are connected with extreme
events in precipitation in Vargas at the coast of Venezuela.
Table 5.1 shows the number of extreme events in SST, precipitation and of both simultaneously. The station data detect between four and seven extreme precipitation events
in the sixty years. The same time periods as for the CCAs are used. In November 2004,
there was one heavy rain event, three months before the extreme event in February 2005.
In February itself the data show the extreme events of 1951 and 2005. In December, the
event in 2010 does not occur because of the lack of data. Only during one of these events,
namely November 2004, is the NTA extremely positive,three months before the extreme
precipitation occur. Both Niño regions have extreme negative anomalies in November
2010, December 2010 and 1999 and in February 2000. Extreme events in the NTA and
the precipitation occur only twice at the same time, once in November and once in December. With the Pacific regions it happens twice in November and December, and once
in February that an event is detected simultaneously in SST and precipitation. It never
happens simultaneously that the NTA is extremely warm, the equatorial Pacific extremely
cold and that the precipitation over Vargas is extremely strong.
The model data do not show a stronger connection between extreme SST and extreme
precipitation. In general, there are more heavy rain events than in the station data. NTA
and rainfall have three events in November simultaneously, Niño3 only one in December
and Niño3.4 four distributed over the three months. Again, no event occurs in the Pacific,
the Atlantic and precipitation simultaneously. The data of the future show that extreme
precipitation events occur less frequently with the progressing of the dry season. In November eleven heavy rains occurred during these sixty years, only two in February. The
point of time does of extreme events in precipitation not correspond well with the events
in the SST, there is only one event in December which takes place in both oceans and in
the rainfall.
The figure 5.1 and figure 5.2 shows that, even when the extreme events do not correspond
well in the data, the pattern, that the NTA and the precipitation have the same signed
anomalies and that the precipitation and the SST of the Niño regions have opposite signed
anomalies, is mostly fulfilled in all data sets.
The figure 5.2 shows in an example the change of the SST from the past to the future.
In the past there are two major El Niño events of four or five years. Each one followed
by a La Niña event in 1975 - 1978 and 1994 - 1997. The minima and maxima are around
+/- 1.2 K. In the future, the time series looks quite different. There occur only two long
88
5.1 Statistic of extreme events in precipitation and sea surface temperature
(a)
(b)
Figure 5.1: Time series of sea surface temperature in the NTA region and of precipitation of the
Novembers 1951 - 2010 (a) of sea surface temperature in the Niño3.4 region and of precipitation
of the Decembers 1951 - 2010 (b)
(a)
(b)
Figure 5.2: Time series of sea surface temperature in the Niño3 region and of precipitation of
the Novembers 1946 - 2005 (a) and 2041 - 2100 (b)
period, one of strong positive anomalies and one of negative anomalies of nine years in the
first half of these sixty years (interrupted by a two-year non-extreme period). Later, there
are only values near zero and positive anomalies again at the end of the whole period.
Thus, the fluctuation decreases and the anomalies get stronger and more concentrated on
one time period.
To sum up, the extreme events do not agree well in SST and precipitation but the figures
show the tendency to anomalies with the same sign in the NTA and the precipitation in
Vargas/at the coast and contrary signs between Pacific SST and rainfall. These results
suggest that La Niña conditions and positive anomalies in the NTA are sufficient for positive rainfall anomalies but not for extreme events. Even though extreme events also occur
without an anomalous cold east equatorial Pacific or an anomalous warm NTA.
89
Table 5.1: Number of extreme events in precipitation, in sea surface temperature and in both
simultaneously of the ERSST/station data and the model data (past and future) in November,
Data
NTA-Nov.
Station data:
Model data past:
Model data future:
NTA-Dec.
Station data:
Model data past:
Model data future:
NTA-Feb.
Station data:
Model data past:
Model data future:
Niño3-Nov.
Station data:
Model data past:
Model data future:
Niño3-Dec.
Station data:
Model data past:
Model data future:
Niño3-Feb.
Station data:
Model data past:
Model data future:
Niño3.4-Nov.
Station data:
Model data past:
Model data future:
Niño3.4-Dec.
Station data:
Model data past:
Model data future:
Niño3.4-Feb.
Station data:
Model data past:
Model data future:
90
Number of extreme
precipitation
Number of extreme
sea surface temperature
Number of both simultaneously
6
11
11
8
9
8
1
3
2
7
7
7
9
10
7
1
0
1
4
7
2
8
9
9
0
0
0
6
11
11
9
8
7
2
0
1
7
7
7
11
11
8
2
1
1
4
7
2
11
9
7
2
0
0
6
11
11
11
11
6
2
1
0
7
7
7
11
9
9
2
1
1
4
7
2
6
13
7
1
2
0
5.2 Meteorological conditions during four extreme events in precipitation
5.2 Meteorological conditions during four extreme events in
precipitation
The previous section reveals, that negative SST anomalies in the Niño regions and/or
positive anomalies in the NTA lead to positive precipitation anomalies but that the SST
is not a sufficient precondition for extreme events. Therefore, four specific extreme events
in precipitation are researched regarding their meteorological conditions. The aim is to
reveal similarities between the events. These four events were extreme in their amount of
precipitation and as well in the consequences they caused. The four events are:
• 15th - 17th of February 1951
• 14th - 16th of December 1999
• 7th - 10th of February 2005
• 24th - 26th of November 2010
The ’Situación Norte’ is a meteorological situation which causes extreme events in precipitation in northern Venezuela. This name stands for the anomalous occurrence of midlatitude meteorological systems more in the south, in the tropics, north of the Caribbean coast
of Venezuela (ARMADA et al., 2000). Mundaray (2005) explained that the northeastward
replacement of the high pressure over the North Atlantic allows the midlatitude cyclones
and troughs to pass farther south than normal and to influence the weather in Venezuela.
Furthermore, cold fronts can pass just north of the Venezuelan coast. ARMADA et al.
(2000) added that the fronts are characterised by a line of clouds of hundreds of kilometres
and cause weak or moderate precipitation which lasts for two to five days. They inform
further that if these systems are very stationary the precipitation lasts longer and that the
number of days with persistent rain rises too, when a second frontal system occurs.
The trade winds favour precipitation by bringing the humidity from the sea to the coast
and consequently to the cordillera. The air is forced to rise very fast because of the narrow
zone between the sea and the mountains. This causes orographic clouds and precipitation
of medial intensity. The winds also favour precipitation positively with convergence in
low levels and divergence in high levels which leads to deep convection (ARMADA et al.,
2000).
Cárdenas et al. (2003) pointed out another modulator of rainfall in Venezuela, the QuasiBiennial Oscillation (QBO) and describe it as followed. The QBO is a circulation in the
lower stratosphere between 10 - 100 hPa. Like the name indicates, the circulation has a
cycle of 24 - 30 months. After one cycle the zonal circulation changes its direction and
velocity. The circulation takes place in the tropics, the average position is between 12.0 ◦ S
91
and 12.0 ◦ N. When there are higher velocities than 4 m/s the precipitation in Venezuela
is higher, especially with west winds. A lower value diminishes the rainfall.
In the following the four extreme events in precipitation in Venezuela in the past are analysed. The synoptic situation is shown by displaying meteorological parameters. Each event
is studied separately and the chapter is completed by a conclusion.
5.2.1 15th - 17th of February 1951
In February 1951 a very strong rain event occurred at the central coast of Venezuela.
Precipitation values as high as during this February were never measured before.
468.0 mm per month, 193 mm on the 16th of February 1951 (Maiquetia). These heavy
rains had very extensive consequences for the people living at the central coast, especially
in Vargas (Linares et al., 1999). 700 people died and hundreds of people lost their houses
(Mundaray, 2005). The poor neighbourhoods and slums were the most affected because
their houses were built very simply and on the hillsides of the cordillera where the landslides
took place.
The pressure field during these three days is shown in figure 5.3 (a). There are two high
pressure fields. One over the southeast edge of the United States and one over the North
Atlantic at the coast of Africa. In the middle of these two highs a slightly low pressure
field can be seen. In figure 5.3 (b) the sea level pressure anomaly of these three days is
presented. All the pressure fields are stronger and more in the south than usual.
(a)
(b)
Figure 5.3: Sea level pressure (a) sea level pressure anomaly (b) of the 15th - 17th of February
1951
This surface pressure formation and the cold front can be seen as well in figure 5.4 of the
17th of February. ’A’ stands for Alta (high) and ’B’ for Baja (low). This cold front was
lying exactly over the coast of Venezuela and is one factor causing strong rain as explained
in the previous paragraph of this chapter.
The figure 5.5 (a), (b) and (c) show the three wind fields in different altitudes: at the
surface, in 850 hPa and in the upper troposphere in 250 hPa. At the level of 850 hPa
92
Figure 5.4: Synoptic map of the surface of the 17th of February 1951 12 UTC (FAV, 2000)
the anticyclonic flow over the Azores and the cyclonic circulation in the east of it can be
observed. The surface wind at the coast of Venezuela comes from the northeast. In
250 hPa the subtropical Jetstream is clearly shown. The wind field at 250 hPa is divergent
at the coast of Venezuela and convergent in low latitudes. This causes an ascent of air
which favours deep convection.
(a)
(b)
(c)
Figure 5.5: Wind field at the surface (a) in 850 hPa (b) in 250 hPa (c) of the 15th - 17th of
February 1951
The wind field anomalies at the surface (figure 5.6) show that the flow comes more from
the north. This forces the rising of air at the cordillera and leads to precipitation.
In addition to this, there was a QBO in February 1951 of -6.01 m/s, according to Cárdenas
et al. (2003). This means slightly forced precipitation.
93
Figure 5.6: Wind field anomalies at the surface of the 15th - 17th of February 1951
In the graph of the air temperature anomalies (figure 5.7 (a)) the cold front can be seen
in the strong temperature differences at the central coast of Venezuela and northwest of
it. The slightly higher air temperature over the central north tropical Atlantic can be
explained by the warmer sea surface temperatures shown in figure 5.7 (b).
(a)
(b)
Figure 5.7: Air temperature anomaly in 850 hPa (a) sea surface temperature anomaly (b) of the
15th - 17th of February 1951
As shown in chapter 4 and in the first part of this chapter, a cold ENSO phase which means
a La Niña event (cold equatorial Pacific) favours strong rain at the coast of Venezuela.
The three-month mean SST of the equatorial Pacific shows negative anomalies in January
- March 1951 of -0.9 ◦ C and a positive Southern Oscillation Index (SOI) in February 1951
of +1.5. This cannot be seen in the SST graph because the window reaches only a very
small part of the Pacific ocean at the coast of South America.
5.2.2 14th - 16th of December 1999
The extreme event in the dry season 1999/2000 took place between the 14th and the 16th
of December 1999. The registered precipitation at the station of Maiquetia was
388.5 mm during this December. This heavy rain had even worse consequences than in
1951 (ARMADA et al., 2000). Alone in Vargas around 15 000 people died (López et al,
94
2005) and more than 5 000 houses were destroyed because of the floods and landslides
(CEPAL, 2000).
In figure 5.8 (a) a similar pattern like in February 1951 can be seen. A low pressure field
over the central North Atlantic and two highs at its sides. The anomalies of the sea level
pressure are shown in figure 5.8 (b). The Azores high pressure system is higher than normally and the low is much stronger and more in the south.
(a)
(b)
Figure 5.8: Sea level pressure (a) sea level pressure anomaly (b) of the 14th - 16th of December
1999
The synoptic map of the surface (figure 5.9 (a)) displays the strong cyclone surrounded
from the high pressure fields. The trough in the high levels of these pressure constellation
can be seen in figure 5.9 (b).
(a)
(b)
Figure 5.9: Synoptic map of the surface (a) synoptic map at 250 hPa (b) of the 15th of December
1999 00 UTC (SRRS, NOAA)
In the wind field in figure 5.10 (a), (b) and (c) it can be seen, that there is a strong
convergence in the low levels with high wind speed and a surface wind direction from the
northeast. In 250 hPa the flow shows an upper-level trough northeast of Venezuela.
The wind field anomalies in figure 5.11 (a) and (b) show a more southward surface flow
with higher than normal wind speeds. In 250 hPa the much stronger flow around the
trough gets visible. This makes clear how uncommon this flow in the upper-levels is and
95
(a)
(b)
(c)
Figure 5.10: wind field at the surface(a) in 850 hPa(b) in 250 hPa(c)
(a)
(b)
Figure 5.11: Wind field anomalies at the surface(a) in 250 hPa(b) of the 14th - 16th of December
1999
that the Jetstream normally appears more in the south and without this bulge.
The QBO has a similar high value like in February 1951 but with the opposite sign. A
westerly flow took place with a speed of 6.43 m/s.
The air temperature anomalies (figure 5.12 (a)) were partly warmer over the north tropical
Atlantic. This warm air is transported to the coast of Venezuela by the flow of the trade
winds. As in 1951, there is a temperature contrast which displays the location of the cold
front. The sea surface temperature anomalies in figure 5.12 (b) show perfectly the negative anomalies in the equatorial Pacific. This was also observed in the ENSO parameter
96
in November - January 1999/2000 with -1.6 ◦ C and the SOI in December of +2.4. Both
values are very high. Furthermore, there was a change from an El Niño to a La Niña
phase. Lyon (2002) revealed that this leads to stronger rainfall (in the wet season). The
equatorial Pacific had an anomaly of +2.3 ◦ C and +2.5 ◦ C in 1997 - 1998. In addition, the
north tropical Atlantic and the Caribbean Sea at the coast of Venezuela showed positive
anomalies.
(a)
(b)
Figure 5.12: Air temperature anomaly in 850 hPa (a) sea surface temperature anomaly (b) of
the 14th - 16th of December 1999
The moisture flux in figure 5.13 (a) explains where the necessary moisture for this strong
precipitation came from. There was a strong moisture flux from the north tropical Atlantic towards the coast of South America, especially to northeast Brazil, Guyana, French
Guyana, Surinam and Venezuela. In the anomalies in figure 5.13 (b) it can be seen how
much stronger this moisture transport was.
(a)
(b)
Figure 5.13: Moisture flux in 850 hPa (a) moisture flux anomaly in 850 hPa (b) of the 14th 16th of December 1999
5.2.3 7th - 10th of February 2005
The extreme event at the 7th to the 10th of February 2005 had the second highest registered
amount of monthly rainfall during a February in the last sixty years. In Maiquetia
97
381.8 mm were measured, the mean of Februaries is 31.49 mm. Nevertheless, after 1999
the country was a bit more prepared for an event like this so that less destruction was
caused. Even so tens of people lost their lives, houses were destroyed as well as bridges
and streets (Lopéz et al., 2005).
The surface pressure fields during the event are displayed in figure 5.14 (a). The Azores
high is strongly distinct with a core pressure of 1030 hPa. East of it a lightly low pressure
field is located and over Florida higher pressures take place again. In the anomalies in
figure 5.14 (b) it gets clear that the Azores high is much stronger than usual and that the
low pressure at its east flank is even stronger and with a anomalous southern position.
(a)
(b)
Figure 5.14: Sea level pressure (a) sea level pressure anomaly (b) of the 7th - 10th of February
2005
The analysed synoptic surface map (figure 5.15) shows the cold front lying directly over
the coast of Venezuela.
Figure 5.15: Analysed synoptic map of the surface of the 7th of February 2005 18 UTC (TROPICAL PREDICTION CENTER MIAMI FLORIDA, after Mundaray 2005)
The surface wind (figure 5.16 (a)) and the wind field in 850 hPa (figure 5.16 (b)) show the
expected circulations following the pressure fields, the anticyclonic flow where the Azores
high is located and a narrow cyclonic flow over the central North Atlantic. In 250 hPa
(figure 5.16 (c)) a divergence over the coast of Venezuela takes place. In addition, there is
98
a high level flow towards Venezuela coming from the equatorial Pacific.
(a)
(b)
(c)
February 2005
The anomalies in figure 5.17 (a) and (b) show that there is a surface wind more from the
north at the coast, like in the events discussed before. In high levels, the flow is much
stronger than normal. The flow from the Pacific is anomalous and of very high wind
speeds.
(a)
(b)
Figure 5.17: Wind field anomalies at the surface (a) in 250 hPa (b) of the 7th - 10th of February
2005
The QBO seems to have no influence during this event. The value for February 2005 is
only of -0.96 m/s, which supposedly to diminishes the precipitation.
99
The air temperature anomalies in 850 hPa (figure 5.18 (a)) show, as during the other
events, positive temperature anomalies over the north tropical Atlantic. In connection
with the trades, this warm and humid air reaches the coast of Venezuela. The temperature
difference, which is characteristic for the cold front, can be seen as well. The sea surface
temperature anomalies in figure 5.18 (b) shows a strong, large warm pool in the north
tropical Atlantic. The Pacific is only colder near the equator, surrounded by positive
anomalies. These do not correspond with the ENSO value of +0.4 ◦ C (January - March
2005) and the negative SOI of -5.2 (February 2005). The positive SST anomalies and the
strong high level winds from the eastern equatorial Pacific could explain the high moisture
content necessary for rain events like this.
(a)
(b)
the 7th - 10th of February 2005
5.2.4 24th - 26th of November 2010
The extreme event in November 2010 was the strongest registered during Novembers of
the last sixty years. A monthly precipitation of 428.5 mm was measured. Again many
people died and thousands lost their houses. A high percentage of these homeless people
still have to live in emergency shelters, ten months after the disaster (Garcia, personal
communication).
During this event, the pressure field (figure 5.19 (a)) looks quite different. Southern North
America and the North Atlantic are dominated by a widespread low pressure field. In the
graph of the anomalies (figure 5.19 (b)) it can be seen that this anomaly is very strong,
with up to 10 hPa.
In figure 5.20 the upper-level trough can be observed. It has a wider zonal elongation than
the ones observed in the previous events.
In the surface windfield (figure 5.21 (a)) and in the one at 850 hPa (figure 5.21 (b)) it
becomes clear, that the low pressure over the North Atlantic are two cyclones. This can
explain the persistent rain over several days. Both figures show two cyclonic flows, one in
100
(a)
(b)
Figure 5.19: Sea level pressure (a) sea level pressure anomaly (b) of the 24th - 26th of November
2010
Figure 5.20: Synoptic map at 250 hPa of the 25th of November 2010 00 UTC (SRRS, NOAA)
the central Atlantic and one at the west coast of Africa. In the surface plot, high wind
speeds of northerly winds at the coast of Venezuela and a convergent flow can be observed.
In 250 hPa (figure 5.21 (c)) the high wind speeds of the Subtropical Jetstream just north
of Venezuela took place, producing a divergence in high levels. In addition, the trough
appears in the flow as well.
The surface flow anomalies (figure 5.22 (a)) show a stronger wind at the coast with a more
northern component. The figure of the wind field anomalies in 250 hPa (figure 5.22 (b))
explains only one wide upper-level trough.
101
(a)
(b)
(c)
November 2010
(a)
(b)
Figure 5.22: Wind field anomalies at the surface (a) in 250 hPa (b) of the 24th - 26th of November
2010
102
Furthermore, the QBO has a very high value with 12.16 m/s.
The air temperature anomalies in 850 hPa (figure 5.23 (a)) have a high and widespread
positive anomaly over the north tropical Atlantic just upstream of the trade winds. Moreover, a temperature contrast, typical for a cold front, can be seen.
The sea surface temperature anomalies (figure 5.23 (b)) show a positive anomaly in nearly
all the north tropical Atlantic and a negative anomaly in the equatorial Pacific. This
matches with the negative ENSO of -1.4 ◦ C (October - December 2010) and the positive
Southern Oscillation Index of +2.1 in November 2010. Additionally, there was an El Niño
phase in 2009 - 2010 with values between +1.5 ◦ C and +1.8 ◦ C. A strong moisture flux at
(a)
(b)
the 24th - 26th of November 2010
850 hPa from the north tropical Atlantic toward Venezuela can be seen in figure 5.24 (a).
The figure 5.24 (b) shows that this flow is slightly anomalous during the dry season.
(a)
(b)
Figure 5.24: Moisture flux in 850 hPa (a) moisture flux anomaly in 850 hPa (b) of the 24th 26th of November 2010
103
5.3 Conclusions
Section 5.1 reveals that only the tendency of precipitation anomalies can be seen from
the SST anomalies. In general, the precipitation at the coast of Venezuela has positive
anomalies with anomalies in the NTA of the same and anomalies of the opposite sign in
the Niño regions.
Furthermore, these results show that La Niña conditions in the Pacific and positive anomalies in the NTA lead to positive rainfall anomalies but this it is not sufficient for an
extreme.
The ’Situación Norte’ seems to be the key factor that extreme events in precipitation occur.
All events have this situation in common but the additional factors differ in their influence.
In 1999 and 2010 the SST anomalies of both oceans are contrary and strong and the ENSO
phase changed from positive to negative. This seems to be not only a support for strong
rain in the wet season. In addition to this, a strong moisture transport towards the coast
of Venezuela took place. This explains the huge amount of moisture measured in precipitation.
In 1951, the comparatively strong temperature contrast at the cold front seems to be a
strong favourable factor.
In 2005, the connection of an even more southward displacement of the trough, its closed
cyclonic flow, the high level wind coming from the warm Atlantic and a Pacific with mostly
positive anomalies seem to be the conditions leading to the event.
104
6
Conclusions
The aim of this Master Thesis was to investigate the influence of sea surface temperature
on precipitation in Venezuela. Two regions in Venezuela were chosen, one at the coast and
one in the inland to detect spatial differences. Not only precipitation in general was of
interest, but also extreme precipitation. Therefore the characteristics of the precipitation
of the seasons and of the months during which the extreme precipitation mostly took place
were studied. Station data as well as model data were used.
The study of precipitation of the four seasons shows at the coast more precipitation in the
late wet than in the early wet season. The inland precipitation is higher during the early
wet season. Comparing both parts of the dry season, more precipitation occur in the early
dry season in both precipitation regions. The analysis of the precipitation of November,
December and February detects a decrement of rainfall with progression of the dry season.
Strong differences in precipitation between the two regions were detected with the station
data. The precipitation in the inland is higher than at the coast. The model has difficulties to differentiate between the precipitation between coast and inland. In addition,
the measured precipitation amounts are much larger than the modelled ones. A reason
for this could be the lack of variations in the orography (because of the coarse resolution)
which leads to a lack of orography uplift so that the rainfall caused by the cordillera at
the coast and the table mountains in the inland are not accounted for. At the coast, the
105
6 Conclusion
precipitation increases from the west to the east and from the south to the north. In the
inland the highest precipitation occurs in the west in the centre of the region.
The oceanic regions north tropical Atlantic, the Niño regions 3 and 3.4 and Pac-Atl which
covers all three previous ones are used in this study. Their characteristics show stronger
anomalies in the Niño regions than in the NTA which get even stronger in the future.
The connection between SST and precipitation was analysed by Canonical Correlation
Analyses. The CCA results of the four seasons are as following:
• The precipitation of the station data in Vargas in the late dry season and in Bolívar
in the wet seasons is more strongly influenced by the NTA than by the Niño regions.
The results for Bolívar correspond with previous studies. The Niño regions and PacAtl have similar and high canonical correlation coefficients for both parts of the wet
season and the early dry season. The model data provide contrary results. In most of
the cases the NTA has a stronger influence than the Niño regions. The literature does
not suggest such a dominating influence of the NTA. The Pac-Atl always provides
the highest correlation, except for the CCA of inland precipitation in the early dry
and the late wet season in the future. The oceanic influence is highest in the late
wet season where most precipitation occurs. Low correlations take place in the late
dry season in the future in the correlations with both precipitation regions and in
the past in the inland, when the rainfall is minimal.
• With the station data, the zero-lag goodness index is higher with the Pacific than with
the Atlantic. Furthermore, the precipitation in Vargas and in Bolívar can possibly
be predicted in the early dry season by the Niño regions. With the model data, all
four regions provide the best zero-lag forecast in the late wet season. Predictability
is lowest in the late dry season (except for the coastal precipitation of the past with
Niño3 and the inland precipitation of the past with Niño3.4 and Pac-Atl). In the
past, the early dry and late wet season precipitation in both region can be better
predicted by the NTA than by the Niño regions. In the future, the Niño regions
provide the better forecast than the NTA on the precipitation of all seasons (except
the late dry season in the inland).
• The results of the CCAs with the station data of the lags with the highest impact
suggest that the SST in the NTA region of the wet season influences the precipitation
of the whole year in Bolívar. In most cases the Niño regions influence most strongly a
short time in advance of the precipitation. With the model data, the highest influence
is almost always at small lags where at least one of the months of the particular season
is included in the three-month average. In general, the lags with the highest goodness
106
6 Conclusion
indices do not vary strongly from the lags of the highest coefficients. Conclusively,
the best forecast with the highest correlation coefficient can be made with the SST
a short time in advance of the precipitation.
The CCAs of November, December and February has the following results:
• The canonical correlations with the station data provide stronger zero-lag correlation
coefficients with the Pacific regions than with the NTA. This corresponds with the
results of Martelo (2003,I). The results of the model data stand in contrast to this.
In general, the correlation coefficients of the CCAs with the NTA are higher than
with the Niño regions, with the model in November and December. Pac-Atl has the
highest coefficients, except for the coastal precipitation in February in the past. In
the future, the influence of the oceanic regions on the precipitation in Venezuela is
strong in November and December and weak in February.
• With the station data, the zero-lag goodness index supports the stronger influence
of the Niño regions during these three months of the dry season with a better predictability. This corresponds with the analysis of the dry season. In the past, a good
prediction with the model data can be made with Pac-Atl and a good one even with
the NTA on the coastal rain in November and December. The Niño regions and PacAtl give a hint at how the precipitation will be in February. In the future, the Niño
regions provide better forecasts than the NTA, February rainfall can be predicted, if
at all, by the SST of Pac-Atl.
• With the station data, the lags of the SSTs in the NTA region which provide the
highest influence are large with the precipitation in Vargas in November and December and in Bolívar in February. The lags of the SSTs in the NTA and in the Pac-Atl
region are never simultaneous with the precipitation in both regions. In general, the
lags of the highest goodness index agree with only few months difference with the one
with the highest coefficient. The lags of highest coefficient and highest index of the
NTA differ stronger between the two precipitation regions. A reason for this could
be the low influence of the NTA. With the model data of the past, Niño3 affects most
strongly with small lags the precipitation in December, Niño3.4 the rainfall of December and February, the NTA the one of November and December. The lags with
the highest coefficient with Pac-Atl differ with the precipitation regions. Pac-Atl
influences highest at small lags the precipitation of November (coast and inland) and
the precipitation of December in the coast. In the future, the results of the lags are
quite different. At the coast, November and December are influenced most strongly
with no or small lags. In the inland, the regions with a strong influence have their
strongest impact a short time in advance of the precipitation. The lags of the highest
goodness indices correspond with the lags of the highest correlation coefficients in
three-fourths of the model data, past and future together.
107
6 Conclusion
Comparison of the two Niño regions show that there is no important difference in the influence of the precipitation in Venezuela between the two areas in the equatorial Pacific. In
most of the cases, the station data provide a higher influence with Niño3.4, also the model
data of the past with rainfall at the coast. In the future, Niño3 has a stronger influence in
nearly all correlations.
The patterns of the spatial loadings of the canonical correlations with the station data
show, in most of the cases, anomalies of the same sign in the NTA and the precipitation.
However, the anomalies of the Niño regions have opposite signs to the ones of the NTA and
the precipitation. These results correspond with previous studies (Enfield, 1996; Guenni
et al., unpublished manuscript; Chen et al., 2002; Taylor et al., 2002; Martelo, 2003,I;
Martelo, 2003,II; Cárdenas et al., 2002, Cárdenas et al., 2003). With the model data of
the past this changes to anomalies of the opposite sign in NTA and precipitation in the
correlations with the NTA and Pac-Atl in February, in the late dry season and in the late
wet season. The spatial loadings of the CCAs with the model data of the future do not
show this pattern in most of the correlations with the NTA and Pac-Atl. Thus, the model
seems to be able to capture the precipitation response of ENSO but not the impact of the
NTA in the correct way.
Looking at the statistical connection between the SST of NTA, Niño3 and Niño3.4 and the
precipitation in Vargas during November, December and February, it seems that extreme
events do not agree well in SST and precipitation. Nevertheless, the tendency to anomalies
with the same sign in the north tropical Atlantic and the precipitation in Vargas/coast and
contrary signs between Pacific SST and rainfall can be seen.
These results show, that La Niña conditions are not sufficient to cause an extreme event
but seem to favour in positive rainfall anomalies. Even so, extreme events also occur without an anomalous cold equatorial Pacific. The SSTs cannot provide a forecast on extreme
events alone, but show the tendency on how the precipitation behaves.
Four major extreme events in precipitation at the coast of Venezuela were analysed on
the basis of their meteorological conditions. They took place in February 1951, December
1999, February 2005 and November 2010. All four occurred in the dry season and caused
immense destruction.
• The ’Situación Norte’ can be seen in all four events. There was a stronger subtropical
high pressure system over the eastern North Tropical Atlantic and at its west a field
of anomalous low pressure. There was a second high pressure field located over the
southeast of North America. This constellation of pressure fields took place during
108
6 Conclusion
the events in 1951, 1999, 2005 and is associated with a trough. Even in December
1999 there was a trough in high levels. This becomes clear in the wind field in 250 hPa
and the synoptic maps where patterns and flows typical for this synoptic situation
appear. In addition to the trough, there was a cold front lying over or close to the
coast of Venezuela. These cold fronts do not move fast and cause around five days
of continuous rain (ARMADA et al, 2000).
• In all events, the surface winds had a stronger than usual northern component at the
coast of Venezuela during all four events which favours strong precipitation because
of the coastal topography. Winds from the north which reach the coast of Venezuela
lead to upward movements because of the cordillera located parallel to the northern
coast.
• Additionally, in the events of 1951, 2005 and 2010 low level convergence and high
level divergence occur, which favour rising air and convection.
• The QBO is another factor which seems to influence the rain in Venezuela. High
values are connected with strong rain, even stronger precipitation with positive values
(west wind) (Cárdenas et al., 2003). A high QBO was found in 1951, 1999 and 2010.
In 1999 and 2010 the values are not only high but also positive. The exact physical
mechanism of the connection between QBO and precipitation is still unknown.
• The air over the north tropical Atlantic had a positive anomaly in all four events,
which favours more instability and the trade winds transport this warm and humid
air to the coast of Venezuela.
• The SST anomalies of the NTA were warmer than usual during all four events but
negligible in 1951. The ENSO values were negative in 1951, 1999 and 2010. Furthermore, the equatorial Pacific had a negative anomaly in the graphs of the events in
1999 and 2010. During these two events where the oceanic influence was strongest,
there was additionally a strong moisture transport toward the coast of Venezuela
which can explain the huge amount of humidity necessary for strong precipitation as
in those two events.
109
Appendix
111
Appendix
Table A.1: List of stations in Vargas (red: remaining stations after the 50% threshold of CPT)
part I
Serial
422
502
503
Latitude [◦ N ]
10.37
10.36
10.36
Longitude [◦ W ]
67.02
66.57
66.59
Altitude [m]
7
75
43
Time period
1970 - 1983
1948 - 1983
1970 - 1979
508
1401
10.36
10.35
66.53
67.03
53
16
1951 - 1999
1956 - 1983
1404
10.32
67.21
5
1951 - 2006
1412
1413
1414
1439
1501
1503
1504
1560
10.28
10.28
10.29
10.24
10.37
10.37
10.36
10.36
67.16
67.13
67.11
67.01
66.27
66.21
66.18
66.36
644
1537
1021
34
6
-
5001
10.36
66.5
74
1958 - 1983
5004
10.37
66.47
30
1953 - 1976
5005
10.37
66.44
49
1951 - 2006
112
1948
1953
1949
1949
1967
1954
1954
2001
-
2006
1984
1983
1983
1983
1958
1996
2002
Name
Catia la mar
Maiquetia
Maiquetia
Aeropuerto
Macuto
Mamo
(only all
seasons, Feb.)
Puerto
la Cruz
La Guitarra
La Penita
Las Mercedes
El Carite
Todasana
Caruao
Chuspa
Quebrada
San Julian
Caraballeda
(only
all seasons)
Uria (only
both wet,
late dry)
Naiguata
Appendix
Table A.2: List of stations in Vargas (red: remaining stations after the 50% threshold of CPT)
part I
Serial
5006
5011
5016
Latitude [◦ N ]
10.37
10.37
10.36
Longitude [◦ W ]
66.36
66.34
66.35
Altitude [m]
74
15
145
Time period
1967 - 1979
1967 - 2006
1967 - 1981
5017
10.36
66.37
575
1967 - 1983
5020
5040
10.36
10.36
66.41
66.44
-
1969 - 1983
1972 - 1983
5041
5042
5043
5044
5045
5046
5051
5058
10.36
10.37
10.37
10.34
10.34
10.43
10.37
10.37
66.21
66.29
66.31
66.57
66.58
66.58
66.51
66.43
10
80
30
1350
1050
640
15
52
5075
10.35
66.56
-
2001 - 2002
9304
10.33
67.1
12
1967 - 1999
9311
9312
10.32
10.33
67.08
67.14
972
22
1967 - 1998
1974 - 1983
9313
10.29
67.14
408
1967 - 1980
1969
1969
1969
1970
1970
1970
1974
2000
-
1999
1983
1999
1983
1980
1979
1974
2001
Name
Anare
Los Caracas
Los Caracas
El Limon
Los Caracas
Fila de Indio
Maria Isabel
Naiguata
Estangue
Caruao
Oripato
Osma
Piedra Azul 1
Piedra Azul 2
Piedra Azul 3
Los Corales
USB
Camburi
Grande
Quebrada
Osorio
Puerto
Oricao
Carayaca
Puerto
Chichiriviche
Hacienda
Naranjal
113
Appendix
Table A.3: List of stations in Bolívar (red: remaining stations after the 50% threshold of CPT)
part I
Serial
8101
8106
8108
8109
8111
8117
Latitude [◦ N ]
8.2
8.1729
7.4541
7.3936
8.19
7.4613
Longitude [◦ W ]
62.41
62.3921
63.0259
63.0208
62.42
62.5833
Altitude [m]
42
79
293
274
164
260
8126
8138
8145
8146
8155
8174
8177
6.5537
8.08
7.3539
6.4902
7.392
7.55
8.0922
62.3606
62.53
62.5024
63.0019
62.6314
63.07
62.4745
320
87
346
316
267
103
97
1979
1991
1997
1983
1998
1998
1998
-
2010
2000
2010
2007
2004
1999
2007
8202
8204
8206
8207
8218
8228
6.3027
6.1445
4.5214
6.2554
6.2233
6.0125
62.5307
62.5052
62.2515
63.3512
62.4615
62.534
392
405
411
280
786
474
1965
1958
1989
1974
1995
1974
-
2010
1970
1992
2007
2006
2005
8231
8232
8246
8251
8253
8255
8258
8259
8262
8272
5.2132
5.3606
5.5429
5.5741
5.2244
5.2132
5.5444
5.5643
5.4258
6.1445
62.4013
62.5457
62.4442
61.4719
62.054
62.4604
62.3449
62.181
62.2046
62.5144
420
558
431
928
2516
420
1749
460
487
412
1958
1997
1997
1997
1997
1979
1979
1980
1964
1998
-
2010
2010
2010
2010
2006
2010
2010
2009
2010
2007
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8298
6.1314
6.1431
6.0243
6.0596
5.5311
5.4118
5.323
5.5245
6.0111
6.16
6.12
62.4723
62.4412
62.3815
62.2172
62.2017
62.2822
62.2449
62.1347
62.0837
62.53
62.52
392
392
428
743
490
532
505
449
473
386
412
1984
1984
1984
1984
1984
1984
1984
1984
1984
1984
1987
-
2007
2007
2010
2007
1998
2007
2005
2007
2007
1995
2010
114
Time period
1998 - 2010 Club
1949 - 2010
1957 - 2010
1991 - 2004
1974 - 1981
1995 - 2007
Name
Nautico
Macagua
Las Babas
Guri Dique C
Alta Vist
Guri Presa
Derecha
San Pedro
Caruachi
Guri Medio
Periquera
Guri Dique J
El Merey
Caruachi
Ataguia A
Arekuna
Canaima
Pigmaken
El Cazabe
Antavaria
San Salvador
de Paul
Uriman
Takupay
Cucurital
Carrao Alto
Auacapa
Capaura
Auyantepuy
Guarimba
Kamarata
Carrao en
Canaima
Mayupa
Tepochi
Ahonda
Kapín
Peipa
Uruyen
Quibaty
Waiquimba
Tuna
Ara
Canaima
Appendix
part II
Serial
8302
8303
8304
8305
8307
8308
8313
8314
8315
8319
8323
8325
8333
8334
8335
8336
8337
8338
8339
8341
8342
Latitude [◦ N ]
5.4341
4.4452
4.2626
5.0745
4.5007
5.0856
4.5352
4.2808
4.2001
5.3503
5.4102
5.0105
4.5857
4.5951
4.5711
4.5446
5.1946
5.1205
4.4441
4.3435
4.1429
Longitude [◦ W ]
61.3236
62.2015
61.474
61.0019
60.5216
61.4843
61.5057
61.3546
61.4314
61.4506
61.338
61.0906
61.4408
62.0508
60.3626
61.4526
61.3816
60.495
61.1508
62.1245
62.0612
Altitude [m]
820
434
639
902
893
818
1014
847
475
1235
1207
874
817
780
1176
807
1119
2598
1587
415
425
Time period
1992 - 2002
1977 - 2010
1992 - 1997
1982 - 2010
1990 - 2007
1990 - 2001
1992 - 2003
1994 - 2009
1974 - 2007
1959 - 2010
1974 - 2010
1974 - 2010
1959 - 2010
1997 - 2010
1997 - 2007
2001 - 2010
1997 - 2010
1997 - 2010
1997 - 1997
1997 - 2004
1997 - 2009
8345
8347
8349
8355
4.4936
4.1911
5.1423
4.5904
61.0455
62.1919
61.1908
61.2015
910
458
860
840
1979
1997
1997
1997
8360
8380
4.3019
5.22
61.0838
61.13
907
911
1953 - 2010
1971 - 1983
8412
8413
5.0409
5.0352
63.0358
63.4114
354
435
1975 - 2010
1976 - 2009
8416
8417
8420
4.4706
5.2311
4.2622
63.2151
61.1214
63.0004
346
1130
360
1975 - 2004
1974 - 1999
1976 - 2000
8430
8431
8448
8450
3.3545
4.2354
4.1344
4.551
62.5721
62.4001
63.1701
62.4532
671
409
480
519
1997
1997
1997
1997
-
-
2002
2010
1998
2000
2010
2010
2010
2010
Name
Lorikeima
Aripichí
Peika
Compuiba
Morocmeru
Mauroken
Eutobarima
El Pauji
Icabaru
Kavanayen
Parupa
Yuruani
Wonken
Techime
Arabopo
Caruaiken
Ueitepuy
Kukenantepuy
Uadaparu
Mariva
Icabaru
Medio
Aqua Fria
Parcupi
Vist Alegra
Nueva
Esperanza
Santa Elena
Aponguao
(Marn)
Guiguata
Guaina (only
all seasons)
Ichum
Kamá
Mahigia (only
both wet,
late dry,
Nov., Dec.,
Feb.)
Paragua Alto
Karun Alto
Ichúm Alto
Aureme
115
Appendix
part III
Serial
8505
Latitude [◦ N ]
5.414
Longitude [◦ W ]
63.3241
Altitude [m]
311
Time period
1976 - 2007
8514
5.4818
63.5251
307
1974 - 2005
8518
8522
8529
5.1855
6.5
6.0806
63.2408
63.2
63.4254
322
283
306
1975 - 2010
1957 - 2010
1976 - 2004
8533
6.1614
63.1614
369
1997 - 2008
8534
8543
6.2204
5.5343
63.3401
63.3402
280
1843
1979 - 2010
1979 - 2010
8703
5.58
61.25
1334
1988 - 2010
8721
8727
7.21
5.343
62.32
61.2055
242
1064
1958 - 2005
1974 - 2002
8730
8750
8778
7.5854
7.3035
6.022
62.1943
63.1636
61.02405
376
383
155
1962 - 2007
1969 - 2010
1971 - 1992
8781
7.5205
62.0425
350
1975 - 1992
116
Name
Carapo (only
both wet,
late dry,
Nov., Dec.)
Gauiquinimita
Cauce
Karum
La Paragua
Tonoro (only
both wet,
late dry,
Nov., Dec.)
Chiguao
Medio
Auraima
Guaiquinimita
Tepuy
Sierra
de Lema
Manteco
San Rafael
de Kamoiran
Upata
Cuidad Piar
Kilómetro 88
(Marn)
Villa Lola
(Marn)
Appendix
Table A.6: Results of the Canonical Correlation Analyses of sea surface temperature (ERSST)
and of precipitation (station data) in the early dry season in Vargas, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3480; 0.2243
0.3937; 0.0821
0.3020; 0.1727
0.2702
0.2703
0.2805
0.2589
0.4923
0.4997
0.4967
0.4997
0.4926
0.4937
0.4696
0.4654
0.4719
0.4700
0.4620
0.4442
0.4881
0.4876
0.4936
0.4972
0.4896
0.5008
0.4889
0.5173
0.4949
-0.053; 0.077
0.007; -0.058
0.063; 0.056
0.119
0.115
0.095
-0.022
0.341
0.349
0.346
0.345
0.337
0.337
0.317
0.309
0.316
0.314
0.305
0.289
0.296
0.306
0.348
0.351
0.347
0.353
0.344
0.320
0.297
Number
of used EOFs,
SST|precipitation
3|2
3|2
3|2
3|1
3|1
3|1
3|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
2|1
2|1
1|1
1|1
1|1
1|1
1|1
2|1
2|1
117
Appendix
and of precipitation (station data) in the late dry season in Vargas, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
118
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3841;
0.1948
0.2195
0.2398
0.1633
0.2681
0.3782
0.1966
0.1806
0.1316
0.0928
0.2835
0.2943
0.1010
0.2853
0.2740
0.3150;
0.3040
0.3089
0.3044
0.2903;
0.2615;
0.3966;
0.5267;
0.5245;
0.3989;
0.3800;
0.1086
0.041; -0.063
-0.077
-0.056
0.046
0.004
0.002
0.153
0.041
0.026
-0.013
-0.053
-0.056
-0.057
-0.039
0.029
0.036
-0.052; -0.037
-0.067
-0.074
-0.081
-0.081; -0.144
-0.069; -0.017
0.001; -0.063
0.034; -0.121
0.008; 0.009
-0.036; 0.000
-0.073; -0.027
-0.015
0.1369
0.1611
0.1058
0.1984
0.0678
0.1264
0.0903
0.0368
0.0196
Number
of used EOFs,
SST|precipitation
3|2
2|1
1|2
1|2
1|2
3|1
3|1
1|1
1|1
1|1
1|1
1|2
1|2
1|1
2|1
2|1
2|2
1|2
1|2
1|2
2|2
2|2
2|2
4|2
3|2
2|2
2|2
1|1
Appendix
and of precipitation (station data) in the early wet season in Vargas, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.2787
0.3327
0.3581
0.4409
0.3683
0.4563;
0.4425;
0.3070
0.2833
0.2146
0.3996;
0.2735;
0.0649
0.0643
0.3791;
0.3350;
0.3157;
0.2788;
0.2707;
0.3244;
0.3539;
0.3518
0.3160
0.3605
0.1212
0.2693
0.1924
0.3182;
0.114
0.160
0.177
0.186
0.174
0.083; 0.136
0.081; 0.102
0.006
0.026
0.009
0.013; -0.012
-0.027; -0.042
-0.074
-0.079
-0.087; -0.127
-0.074; -0.120
-0.020; -0.023
-0.001; 0.007
0.002; -0.005
-0.014; -0.013
-0.042; -0.045
0.053
0.090
0.157
-0.033
0.054
-0.023
-0.042; -0.127
0.2575
0.2204
0.0649
0.0634
0.0190
0.0307
0.0887
0.1044
0.0867
0.0760
0.0693
0.0096
Number
of used EOFs,
SST|precipitation
1|1
1|1
1|1
3|1
1|1
3|2
3|2
2|1
2|1
2|1
2|2
2|2
1|1
1|1
2|2
2|2
2|2
2|2
2|2
2|2
2|2
4|1
3|1
3|1
1|1
2|1
2|1
2|2
119
Appendix
and of precipitation (station data) in the late wet season in Vargas, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
120
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.4033;
0.3745;
0.2800;
0.2718
0.3618
0.3246
0.2033
0.3487
0.3458
0.3804
0.4256
0.3543;
0.2231
0.2831;
0.3563
0.3561
0.3968
0.4031
0.3365
0.2538
0.2167
0.4438
0.3836
0.4136
0.5366
0.4585;
0.3360
0.3867
-0.030; -0.039
-0.013; -0.068
-0.165; -0.045
0.020
0.106
0.065
0.013
0.157
0.159
0.184
0.196
0.153; 0.144
0.031
-0.049; -0.098
0.150
0.149
0.182
0.193
0.150
0.074
0.040
0.164
0.161
0.186
0.197
0.079; 0.153
0.090
0.094
0.1764
0.1564
0.1986
0.2380
0.0705
0.3970
Number
of used EOFs,
SST|precipitation
3|2
3|2
2|2
2|1
2|1
2|1
1|1
1|1
1|1
1|1
2|1
2|2
2|1
2|2
1|1
1|1
1|1
1|1
1|1
2|1
2|1
3|1
1|1
1|1
4|1
3|2
2|1
2|1
Appendix
and of precipitation (station data) in the early dry season in Bolívar, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.1659
0.1512;
0.1877;
0.1733;
0.2848
0.2346
0.5265;
0.7660
0.7669
0.7647
0.7530;
0.7288;
0.6943
0.6597
0.7612
0.7585
0.7481
0.7258
0.6936
0.6500
0.5371
0.6998
0.6750
0.7282;
0.7395;
0.7386;
0.6977;
0.6212;
0.074
-0.042; -0.048
-0.055; -0.101
-0.082; -0.057
0.111
0.066
0.066; -0.102
0.515
0.510
0.499
0.487; 0.483
0.465; 0.461
0.434
0.396
0.519
0.518
0.510
0.488
0.458
0.417
0.322
0.468
0.446
0.456; 0.464
0.475; 0.483
0.481; 0.489
0.421; 0.434
0.336; 0.359
0.0050
0.0046
0.1572
0.3795
0.2980
0.2955
0.1510
0.1750
0.1883
0.2388
0.2860
Number
of used EOFs,
SST|precipitation
1|1
2|2
2|2
2|2
1|1
1|1
3|2
2|1
2|1
2|1
2|2
2|2
2|1
3|1
1|1
2|1
2|1
1|1
1|1
1|1
1|1
1|1
1|1
2|2
2|2
2|2
2|2
2|2
121
Appendix
and of precipitation (station data) in the late dry season in Bolívar, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
122
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.2868;
0.1314
0.0934
0.3258
0.3558;
0.3955
0.5812
0.5979;
0.6087;
0.6227;
0.2375
0.2437
0.2785
0.3295
0.6469;
0.2892
0.6581;
0.6717;
0.6793;
0.6649;
0.4377
0.6122;
0.2026
0.7117;
0.2905
0.3088
0.7282;
0.7802;
-0.093; -0.039
-0.071
-0.019
-0.105
-0.012; 0.006
0.288
0.315
0.062; 0.241
-0.012; 0.184
-0.017; 0.142
0.015
0.042
0.112
0.194
0.054; 0.133
0.170
-0.010; 0.182
-0.064; 0.205
-0.104; 0.245
-0.075; 0.273
0.303
0.035; 0.064
0.056
-0.011; 0.158
0.169
0.194
0.054; 0.226
0.213; 0.279
0.2106
0.1807
0.3446
0.2809
0.2490
0.2135
0.2641
0.3177
0.3742
0.3921
0.1979
0.2065
0.3664
0.3395
Number
of used EOFs,
SST|precipitation
2|2
1|1
1|1
1|2
2|2
2|1
3|1
2|2
2|2
2|2
1|1
1|1
1|1
1|1
2|2
2|1
2|2
2|2
2|2
2|2
2|1
3|2
1|1
3|2
1|1
1|1
3|2
4|2
Appendix
and of precipitation (station data) in the early wet season in Bolívar, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.8487;
0.8175;
0.8229
0.8146;
0.7833
0.7659;
0.7367;
0.5299;
0.5888
0.5837
0.5332
0.4816
0.4692
0.3735
0.7456;
0.5081
0.5332
0.5107
0.4783
0.4604
0.4613
0.8161;
0.6523;
0.8346;
0.8004;
0.6049;
0.7410;
0.6560;
0.177; 0.258;
0.241;0.330
0.250
0.189; 0.221
0.203
0.239; 0.230;
0.167;0.211
0.121
0.163
0.156
0.122
0.101
0.102
0.112
0.004; 0.069
0.120
0.161
0.160
0.151
0.144
0.145
0.146; 0.314;
0.233; 0.249
0.254; 0.397;
0.234; 0.274;
0.137; 0.199
0.158; 0.230
0.033; 0.211
0.5200; 0.2723
0.4909
0.3982;
0.5082; 0.2280
0.4320
0.3970
0.7996; 0.1653
0.2941
0.7864; 0.1624
0.5294; 0.0969
0.3389
0.4148
0.4462
0.275
0.245
0.295
0.328
0.259
Number
of used EOFs,
SST|precipitation
3|4
3|2
1|2
3|2
2|1
3|4
2|4
1|2
1|2
1|2
1|2
1|2
1|2
1|1
2|4
1|2
1|2
1|2
1|2
1|2
1|2
4|3
2|2
4|3
3|3
2|4
2|4
2|3
123
Appendix
and of precipitation (station data) in the late wet season in Bolívar, SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
124
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.6719;
0.7210;
0.6731;
0.6157;
0.5349;
0.5145;
0.5448;
0.6369
0.6337
0.6271
0.5658
0.6508
0.2459
0.1029
0.6389;
0.6479;
0.6824;
0.6497
0.4327
0.2940
0.1703
0.8238
0.8150
0.8236
0.8241
0.4941
0.8000
0.5571
0.222; 0.242
0.319; 0.364
0.326; 0.351
0.300; 0.289
0.021; 0.064
0.027; 0.058
0.107; 0.126
0.365
0.356
0.345
0.298
0.280
0.020
-0.142
0.354; 0.345
0.332; 0.314
0.336; 0.310
0.285
0.164
0.080
-0.026
0.460
0.460
0.450
0.430
0.136
0.390
0.044
0.3547
0.4588
0.3909
0.3407
0.2692
0.2030
0.1805
0.1365
0.1035
0.0804
Number
of used EOFs,
SST|precipitation
3|2
3|2
3|2
3|2
3|2
3|2
3|2
1|1
1|1
1|1
1|1
3|1
1|1
1|1
2|2
2|2
2|2
2|1
1|1
1|1
1|1
4|1
3|1
3|1
3|1
2|1
4|1
3|1
Appendix
Table A.14: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the early dry season at the coast (model data of the past), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.7402
0.7562;
0.6348;
0.3458
0.3670;
0.3527;
0.4486;
0.3779
0.3268
0.2954
0.2701
0.4423;
0.3684;
0.3661;
0.4912;
0.2949
0.2776
0.2624
0.2566
0.2479
0.2363
0.8136;
0.8111
0.6535;
0.3506
0.3985;
0.2629
0.2423
0.579
0.573; 0.593
0.210; 0.239
0.227
0.153; 0.117
0.125; 0.066
-0.221; 0.053; -0.004
0.315
0.259
0.228
0.204
0.025; 0.210; 0.181
-0.024; 0.201; 0.172
0.096; 0.210
0.397; 0.418
0.222
0.208
0.198
0.193
0.180
0.164
0.659; 0.648
0.638
0.488; 0.482
0.265
0.213; 0.181
0.198
0.166
0.4947
0.4685
0.0800
0.0342
0.3811; 0.1633
0.3636; 0.0884
0.3332; 0.0741
0.2258
0.1642
0.4449
0.3530
0.0409
Number
of used EOFs,
SST|precipitation
7|1
7|2
7|2
2|1
2|2
2|2
5|3
1|1
1|1
1|1
1|1
4|3
3|3
2|2
2|2
2|1
1|1
1|1
1|1
1|1
2|1
9|2
8|1
7|2
2|1
2|2
1|1
1|1
125
Appendix
Table A.15: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the late dry season at the coast (model data of the past), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
126
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3383;
0.2779;
0.5316;
0.3079
0.2417
0.2918
0.2798
0.4239
0.3908
0.3551
0.3482
0.3339
0.3458
0.3617
0.3906
0.3697
0.3854;
0.3760
0.3517
0.3363
0.3395
0.5767;
0.4172
0.4562
0.3396
0.3409
0.3423
0.3311
-0.075; -0.149
-0.018; -0.267
0.032; -0.085; -0.076
0.103
0.164
0.167
0.170
0.357
0.335
0.292
0.264
0.258
0.272
0.291
0.334
0.318
0.261; 0.318
0.311
0.284
0.261
0.263
0.274; 0.347; 0.322
0.312
0.301
0.267
0.265
0.268
0.255
0.0228
0.0822
0.1559; 0.1058
0.2617
0.3886; 0.2404
Number
of used EOFs,
SST|precipitation
2|2
2|2
4|3
4|1
1|1
2|1
2|1
1|1
1|1
1|1
3|1
1|1
1|1
1|1
1|1
1|1
2|2
2|1
1|1
1|1
1|1
6|3
2|1
5|1
1|1
1|1
1|1
1|1
Appendix
Table A.16: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the early wet season at the coast (model data of the past), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5698;
0.3169
0.3131
0.2762
0.3068;
0.1524
0.4783;
0.3980
0.3873
0.3506
0.4001
0.1666
0.2223;
0.3576;
0.4107
0.3372
0.4187
0.2127
0.3649;
0.4033
0.4371;
0.6818
0.6263
0.4803
0.4081
0.6238;
0.0626
0.0616
0.304;
0.191
0.157
0.081
0.006;
0.027
0.205;
0.297
0.319
0.303
0.229
0.137
0.093;
0.159;
0.344
0.293
0.237
0.181
0.210;
0.328
0.318;
0.430
0.417
0.336
0.203
0.136;
0.142
0.153
0.3561
0.2208; 0.1639
0.1255
0.1608
0.1414
0.2552; 0.1021
0.2616
0.4100; 0.2999
0.272
-0.090; -0.003
0.149
0.116
0.194
0.208; 0.195
0.374
0.069; 0.043
Number
of used EOFs,
SST|precipitation
7|2
2|1
5|1
2|1
3|3
1|1
5|2
1|1
1|1
1|1
4|1
1|1
2|2
2|2
1|1
1|1
3|1
1|1
3|3
3|1
3|2
10 | 1
6|1
3|1
4|1
9|3
1|1
1|1
127
Appendix
Table A.17: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the late wet season at the coast (model data of the past), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
128
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.8486;
0.8741;
0.7677;
0.6396
0.4832
0.3936;
0.3718;
0.7489
0.7876
0.7827
0.8184
0.7460
0.7436
0.6617
0.7484
0.7760
0.8386
0.8264
0.8144
0.7452
0.6400
0.9068
0.8479;
0.8488
0.8742
0.8825
0.8557
0.7717
0.780;
0.763;
0.621;
0.505
0.344
0.180;
0.246;
0.670
0.723
0.719
0.750
0.709
0.690
0.578
0.679
0.710
0.769
0.760
0.741
0.669
0.510
0.817
0.750;
0.761
0.793
0.789
0.757
0.644
0.0714
0.3822; 0.2504
0.3200; 0.2319
0.2129
0.0213
0.2830
0.770
0.745; 0.744
0.606; 0.610
0.122
0.046
0.740
Number
of used EOFs,
SST|precipitation
4|2
7|3
5|3
7|1
4|1
5|2
2|2
3|1
4|1
3|1
4|1
1|1
4|1
4|1
3|1
3|1
4|1
3|1
4|1
4|1
4|1
9|1
7|2
6|1
7|1
9|1
8|1
8|1
Appendix
Table A.18: Results of the Canonical Correlation Analyses of sea surface temperature and of
precipitation in the early dry season in the inland (model data of the past), SST leading the precipitation up to six months
Region
Lag
CCA coefficient
Goodness index
NTA
0
1
2
3
4
5
6
0.7447; 0.4108
0.7392; 0.4164; 0.3643
0.5116
0.4222
0.3062
0.2468
0.6499; 0.5342; 0.3715;
0.3450; 0.2556; 0.0796
0.6297;
0.4873;
0.1104;0.0734
0.3744
0.3286
0.2969
0.5033; 0.1667
0.2738
0.3506
0.5341
0.4076
0.3156
0.2853
0.2585
0.2473
0.2260
0.8391; 0.6743; 0.3806
0.7917; 0.5442; 0.2779
0.6536; 0.3792
0.5867
0.5680; 0.4675; 0.3699;
0.2670; 0.2303; 0.0561
0.5680; 0.4564
0.5824; 0.4328; 0.3754;
0.1038
0.528; 0.511
0.493; 0.446; 0.456
0.325
0.266
0.162
0.085
-0.035; 0.066; 0.070;
0.164; 0.139; 0.130
0.347; 0.444; 0.439;
0.433
0.296
0.250
0.205
0.251; 0.209
0.160
0.261
0.373
0.259
0.220
0.179
0.140
0.139
0.127
0.599; 0.652; 0.643
0.597; 0.595; 0.582
0.460;0.467
0.379
0.077; 0.250; 0.172;
0.181; 0.206; 0.204
-0.096; 0.187
-0.068; -0.075; 0.174;
0.162
0
Niño3
Niño3.4
Pac-Atl
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
Number
of used EOFs,
SST|precipitation
7|2
6|3
4|1
3|1
2|1
2|1
7|6
4|6
1
1
1
4
1
3
3
3
1
1
1
1
1
10
9
6
8
6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1
1
1
2
1
1
1
1
1
1
1
1
1
3
3
2
1
6
7|2
6|4
129
Appendix
Table A.19: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the late dry season in the inland (model data of the past), SST leading the precipitation
up to six months
Region
Lag
CCA coefficient
Goodness index
0
0.6692;
0.2400;
0.6495;
0.2149;
0.4854;
0.1409
0.2432
0.7438;
0.3288;
0.3365
0.2401
0.4380
0.4354
0.4221
0.4630
0.4095
0.4298
0.4415
0.5349;
0.5326;
0.1104
0.4233
0.4538;
0.4193
0.4128
0.4219
0.7340;
0.4618
0.4420
0.4119
0.5360;
0.5719;
0.6949;
0.2819;
0.076; 0.133; 0.079;
0.004; -0.017; -0.019;
0.039; -0.026; -0.095;
-0.094; -0.082
0.065; -0.013; 0.107;
-0.075
0.148
0.220; 0.159; 0.130;
0.141; 0.135; 0.131
0.233
0.123
0.359
0.362
0.344
0.348
0.329
0.348
0.359
0.363; 0.347
-0.001; 0.338; 0.327;
0.325
0.326
0.329; 0.323
0.324
0.319
0.330
0.475; 0.433; 0.435
0.382
0.354
0.335
0.350; 0.334
0.358; 0.371
0.434; 0.398; 0.394;
0.388; 0.399
1
NTA
2
3
4
Niño3
Niño3.4
Pac-Atl
130
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5591; 0.4587;
0.0857; 0.0295
0.5295; 0.4166;
0.1588
0.3274; 0.1599;
0.4033; 0.3480;
0.1929; 0.0552
0.3120
0.4901; 0.2433;
0.1595
0.4221; 0.3616
0.1168
0.3670
0.4879; 0.3747;
0.1828
Number
of used EOFs,
SST|precipitation
6|6
7|5
4|4
1|1
7|6
2
2
1
1
1
3
1
1
1
4
4
|
|
|
|
|
|
|
|
|
|
|
1
1
1
1
1
1
1
1
1
2
6
1
2
1
1
1
7
2
3
1
4
6
9
|
|
|
|
|
|
|
|
|
|
|
|
1
3
1
1
1
3
1
1
1
2
2
5
Appendix
precipitation in the early wet season in the inland (model data of the past), SST leading the precipitation up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
0.4885
0.3656
0.5915;
0.1067;
0.6576;
0.2653;
0.4596;
0.2974
0.3294
0.5178
0.4988
0.5442
0.5439
0.6238;
0.1856;
0.3210
0.3898
0.4781
0.5345
0.5128
0.4465
0.3388
0.3459
0.5075;
0.5706
0.7103
0.6244
0.5581
0.3102
0.3867
0.5098
0.313
0.216
-0.054; 0.131; 0.152;
0.119; 0.111
0.007; 0.045; 0.096;
0.067; 0.064; 0.052
0.065; -0.003; -0.014
0.139
0.220
0.361
0.348
0.369
0.329
0.145; 0.051; -0.040;
-0.046; -0.037
0.179
0.299
0.339
0.305
0.296
0.233
0.292
0.216
0.053; 0.173; 0.194
0.370
0.410
0.383
0.341
0.189
0.221
0.330
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5176; 0.4194;
0.0707
0.4628; 0.4493;
0.2305; 0.0365
0.2726; 0.1883
0.4497; 0.3144;
0.1100
0.4640; 0.2303
Number
of used EOFs,
SST|precipitation
2|1
2|1
5|6
7|6
7
4
2
1
1
2
4
5
|
|
|
|
|
|
|
|
3
1
1
1
1
1
1
6
3
2
1
3
3
4
3
3
3
3
10
5
6
2
4
4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1
1
1
1
1
1
1
1
5
1
1
1
1
1
1
1
131
Appendix
Table A.21: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the late wet season in the inland (model data of the past), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
0.8530;
0.8758;
0.2726
0.8075;
0.7372
0.2987;
0.3620;
0.3970;
0.3935;
0.7326
0.7818
0.7959
0.8149
0.7815;
0.7693
0.7197;
0.7174
0.7493
0.8281
0.8236
0.8186
0.7731
0.7166;
0.9089;
0.3279
0.8273
0.8569
0.8879;
0.2006;
0.8515;
0.8309
0.7732
0.693;
0.643;
0.619
0.556;
0.394;
0.359;
0.192;
0.108;
0.075;
0.598
0.658
0.672
0.691
0.650;
0.641
0.522;
0.576
0.611
0.695
0.707
0.695
0.640
0.506;
0.693;
0.670
0.693
0.712
0.731;
0.736;
0.720;
0.691
0.606
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
132
0.1907; 0.1143
0.4603; 0.4334;
0.2840
0.5108; 0.4184;
0.0543
0.0614
0.1875; 0.0892
0.1937; 0.0700
0.2982
0.1652
0.1169; 0.0733
0.5461; 0.4856;
0.5428; 0.4636;
0.0426
0.4181
0.676; 0.676
0.605; 0.625;
0.503
0.357; 0.351;
0.359
0.168
0.048; 0.072
-0.072; -0.075
0.657
0.504
0.469; 0.466
0.665; 0.667;
0.730;
0.736
0.729
0.741;
Number
of used EOFs,
SST|precipitation
4|3
7|4
6|2
7|5
3
3
5
3
4
3
4
3
5
5
3
3
4
3
4
4
4
9
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2
3
3
1
1
1
1
2
1
2
1
1
1
1
1
1
3
4
4|1
6|1
7|5
5|2
5|1
8|1
Appendix
precipitation in the early dry season at the coast (model data of the future), SST leading the
precipitation up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.4194;
0.5550;
0.5198
0.3908
0.3327
0.2728
0.3668;
0.4507
0.3474
0.4447
0.4239
0.4432;
0.4849
0.4747
0.4499
0.4559
0.4503
0.4356
0.4243
0.3879
0.3094
0.5287;
0.5515;
0.6664;
0.5860;
0.5477;
0.4898;
0.5012;
-0.037; 0.143; 0.143
0.296; 0.279; 0.277
0.381
0.319
0.226
0.102
0.105; 0.089
0.472
0.405
0.416
0.400
0.394; 0.387
0.382
0.363
0.415
0.422
0.416
0.404
0.394
0.339
0.229
0.393; 0.371
0.439; 0.412; 0.402
0.454; 0.467; 0.447
0.447; 0.458; 0.451
0.371; 0.400; 0.384
0.341; 0.329
0.242; 0.292; 0.275
0.3010; 0.0588
0.2473; 0.0316
0.1053
0.1122
0.0546
0.4363;
0.4021;
0.3118;
0.3098;
0.1475
0.2292;
0.0282
0.2535
0.0486
0.0790
0.0793
Number
of used EOFs,
SST|precipitation
3|3
4|3
4|1
3|1
3|1
3|1
3|2
3|1
1|1
1|1
1|1
2|3
3|1
3|1
1|1
1|1
1|1
1|1
1|1
2|1
2|1
2|3
3|3
5|3
3|3
4|3
3|2
4|3
133
Appendix
Table A.23: Results of the Canonical Correlation Analyses of sea surface temperature and of precipitation in the late dry season at the coast (model data of the future), SST leading the precipitation
up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
134
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3250;
0.3409;
0.2749;
0.2963
0.2913
0.3669;
0.4547;
0.1901;
0.4244;
0.4309;
0.3086
0.2997;
0.1167
0.1426
0.4451;
0.4323;
0.3760;
0.1656;
0.2889
0.1248;
0.3341;
0.4405;
0.3291
0.2978
0.2791
0.2651
0.2510
0.3659;
-0.208;
-0.133;
-0.199;
0.031
-0.021
-0.130;
-0.066;
-0.088;
-0.143;
-0.128;
-0.377
-0.440;
-0.372
-0.282
-0.097;
-0.131;
-0.353;
-0.374;
-0.431
-0.340;
-0.302;
-0.189;
-0.186
-0.209
-0.194
-0.180
-0.171
-0.059;
0.2160; 0.0751
0.0741
0.1026
0.3434;
0.3600;
0.0262
0.1410;
0.1075;
0.1279
0.1185
0.1245
0.0254
0.0931
0.1232
0.1148
0.1116; 0.0267
0.0335
0.0458
0.1124; 0.0365
0.2640; 0.2332
0.1726
-0.175; -0.169
-0.199
-0.296
0.086; 0.061
0.105; 0.064
-0.142
-0.262; -0.209
-0.301; -0.292
-0.417
-0.291
-0.268
-0.430; -0.446
-0.448
-0.369
-0.425; -0.420
-0.311; -0.254
-0.117
Number
of used EOFs,
SST|precipitation
3|3
2|3
4|2
5|1
5|1
4|3
5|3
2|2
3|3
3|3
1|3
2|3
2|1
2|1
2|3
2|3
3|3
3|2
1|3
2|2
3|3
7|3
1|3
1|3
1|3
1|3
1|3
6|2
Appendix
precipitation in the early wet season at the coast (model data of the future), SST leading the
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.6409;
0.6007;
0.4201;
0.4126;
0.4633;
0.3247
0.4962
0.4611
0.4453
0.5669
0.4977
0.3369
0.3819
0.3833
0.4183
0.5237
0.4870
0.3230
0.3060
0.4061
0.4488
0.6827;
0.6441;
0.5999;
0.5370;
0.3906
0.4752
0.5078
0.342;
0.351;
0.133;
0.120;
0.180;
0.200
0.354
0.372
0.363
0.416
0.300
0.244
0.269
0.251
0.346
0.343
0.316
0.228
0.204
0.207
0.246
0.483;
0.453;
0.396;
0.295;
0.292
0.346
0.379
0.0899;
0.1108;
0.0984;
0.1491;
0.1578;
0.0176
0.0402
0.0600
0.0394
0.0389
0.0922;
0.1124;
0.0697;
0.0390;
0.0543
0.0481
0.0332
0.0082
0.269;
0.288;
0.060;
0.043;
0.123;
0.268
0.282
0.051
0.029
0.121
0.455;
0.422;
0.354;
0.237;
0.453
0.424
0.356
0.238
Number
of used EOFs,
SST|precipitation
3|3
3|3
3|3
3|3
3|3
1|1
3|1
2|1
2|1
4|1
4|1
1|1
2|1
2|1
1|1
3|1
3|1
1|1
1|1
3|1
3|1
3|3
3|3
3|3
3|3
1|1
3|1
3|1
135
Appendix
precipitation in the late wet season at the coast (model data of the future), SST leading the precipitation up to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
136
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.7587;
0.7416
0.7160
0.5057
0.3481
0.2683
0.1953
0.7514;
0.7055
0.7026
0.6562
0.5635
0.6894
0.6023
0.7421
0.7417
0.7091
0.6410
0.7031
0.6643
0.5825;
0.8169;
0.8023;
0.7811;
0.8130;
0.8105;
0.7659;
0.6876;
0.632;
0.630
0.597
0.404
0.240
0.124
0.046
0.663;
0.652
0.627
0.573
0.508
0.584
0.488
0.689
0.686
0.651
0.584
0.571
0.552
0.446;
0.688;
0.671;
0.629;
0.606;
0.615;
0.577;
0.448;
0.4245
0.2146; 0.1298
0.2938
0.4287;
0.4063;
0.3564;
0.4786;
0.4923
0.4754;
0.4544;
0.1563
0.1872
0.1957
0.2401
0.2687
0.3524
0.649
0.647; 0.651
0.473
0.706;
0.685;
0.640;
0.627;
0.642
0.608;
0.474;
0.703
0.684
0.641
0.625
0.612
0.492
Number
of used EOFs,
SST|precipitation
5|2
4|1
4|1
3|1
3|1
2|1
1|1
3|3
1|1
3|1
3|1
1|1
4|1
4|1
1|1
1|1
1|1
1|1
3|1
3|1
3|2
3|3
3|3
3|3
7|3
7|2
5|3
6|3
Appendix
precipitation in the early dry season in the inland (model data of the future), SST leading the
Region
NTA
Lag
CCA coefficient
Goodness index
0
1
0.4052;
0.5787;
0.1166
0.5846
0.6788;
0.0359
0.6465;
0.0638
0.5804;
0.5639;
0.7129;
0.5795
0.6564;
0.6132;
0.6580;
0.6364;
0.6490;
0.6054
0.6039
0.5983
0.5825
0.5679
0.5069
0.4214
0.6767;
0.6898;
0.7305;
0.6924;
0.6949;
0.7046;
0.6868;
0.222;
0.287;
0.307
0.432
0.327;
0.385
0.292;
0.271
0.199;
0.074;
0.519;
0.515
0.532;
0.488;
0.481;
0.468;
0.456;
0.540
0.536
0.530
0.516
0.503
0.454
0.381
0.474;
0.456;
0.519;
0.551;
0.504;
0.431;
0.385;
2
3
4
Niño3
Niño3.4
Pac-Atl
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.0402
0.3685; 0.1767;
0.5073; 0.2463;
0.4900; 0.2595;
0.3047
0.4683; 0.1397
0.4649; 0.2927
0,4691
0.2146
0.2340; 0.0371
0.2720; 0.0109
0.2798; 0.0170
0.5321
0.5394
0.5604; 0.1819
0.0674
0.1496
0.4670; 0.0509
0.4498; 0.1215
Number
of used EOFs,
SST|precipitation
2|3
4|4
0.229
0.334;
0.300;
0.375;
0.396;
4|1
4|5
0.276;
0.287;
4|5
0.134
0.145; 0.126
0.534; 0.544
0.513
0.473
0.466; 0.467
0.447; 0.443
0.433; 0.425
0.519
0.532
0.539; 0.545
0.548
0.496
0.460; 0.443
0.423 ;0.400
2
3
3
1
2
2
3
3
3
1
1
1
1
1
1
1
2
2
3
3
5
3
3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5
5
5
1
5
3
3
3
3
1
1
1
1
1
1
1
5
5
5
2
2
5
5
137
Appendix
precipitation in the late dry season in the inland (model data of the future), SST leading the
Region
NTA
Niño3
Niño3.4
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
0.4062; 0.2996
0.3165
0.3277
0.3923
0.5323; 0.3529; 0.0673
0.3673
0.3741
0.4619; 0.1969
0.4513
0.5221; 0.0905; 0.0497
0.3968
0.4194; 0.0583
0.4134
0.4240;0.0608
0.3798; 0.1981
0.3799; 0.1958
0.4140
0.3869
0.3961
0.3854
0.3759
0.6753; 0.5028; 0.4049;
0.3256; 0.1597; 0.1040
0.6516; 0.5580; 0.3869;
0.2291; 0.1775; 0.0907
0.3398
0.3653
0.3813
0.3926
0.3883
0.072;
0.234
0.245
0.267
0.261;
0.280
0.287
0.294;
0.264
0.310;
0.236
0.252;
0.248
0.258;
0.169;
0.179;
0.228
0.245
0.260
0.249
0.171
0.231;
0.168;
0.260;
0.165;
0.245
0.272
0.290
0.300
0.297
1
Pac-Atl
138
2
3
4
5
6
0.207
0.271; 0.235
0.309
0.275; 0.275
0.224
0.222
0.220
0.228
0.163; 0.190;
0.157; 0.161
0.212; 0.205;
0.152; 0.161
Number
of used EOFs,
SST|precipitation
3|2
1|1
1|1
2|1
4|3
1|1
1|1
2|2
3|1
3|3
3|1
3|2
3|1
2|3
2|2
2|2
3|1
2|1
2|1
2|1
3|1
7|6
6|6
1
1
1
1
1
|
|
|
|
|
1
1
1
1
1
Appendix
precipitation in the early wet season in the inland (model data of the future), SST leading the
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
0.7142;
0.0020
0.6348;
0.6024;
0.5167
0.5799;
0.5510;
0.5913;
0.7431;
0.6204
0.7140;
0.6450;
0.4843
0.4374
0.3994
0.7061;
0.6331;
0.5536
0.5186
0.5157
0.5914;
0.3974
0.7431;
0.7275;
0.7226;
0.6287;
0.6174;
0.6060;
0.6176;
0.397; 0.369; 0.472;
0.467
0.284; 0.351; 0.385
0.174; 0.144
0.064
0.050; 0.069
-0.064; 0.126
0.086; 0.257
0.560; 0.603
0.510
0.460; 0.493; 0.491
0.413; 0.445
0.383
0.336
0.291
0.543; 0.586
0.458; 0.507
0.453
0.419
0.400
0.153; 0.355
0.294
0.563; 0.608
0.536; 0.587
0.496; 0.529; 0.518
0.348; 0.470
0.274; 0.426; 0.404
0.140; 0.404
0.254; 0.422
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.4473; 0.3939;
0.4525; 0.2636
0.3558
0.2489
0.3319
0.4067
0.5008
0.3748; 0.1487
0.3183
0.4233
0.3660
0.3929
0.4763
0.5140
0.4914; 0.0968
0.5108
0.5290; 0.0686
0.5355
0.4990
Number
of used EOFs,
SST|precipitation
4|4
4
2
1
4
4
5
4
2
4
4
1
1
1
2
2
1
1
2
2
1
3
4
3
4
3
6
5
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3
6
3
2
2
2
2
1
3
2
1
1
1
2
2
1
1
1
3
1
2
2
4
2
4
2
2
139
Appendix
precipitation in the late wet season in the inland (model data of the future), SST leading the
Region
Lag
CCA coefficient
Goodness index
0
0.8254;
0.2321;
0.8363;
0.1885
0.7976;
0.1381
0.5351;
0.3835
0.3548
0.2396
0.7257
0.7149
0.6872
0.6711;
0.6266
0.7574;
0.0914
0.6908;
0.7299
0.7257
0.6872
0.6367
0.6579;
0.7174;
0.6641;
0.7459
0.7673;
0.7617;
0.8352;
0.8395;
0.2969;
0.8171;
0.2452;
0.7427;
0.544;
0.529;
0.542;
0.536
0.498;
0.485
0.345;
0.255
0.225
0.150
0.579
0.563
0.554
0.520;
0.512
0.545;
0.539
0.519;
0.581
0.580
0.571
0.532
0.423;
0.489;
0.370;
0.588
0.584;
0.573;
0.641;
0.584;
0.596;
0.583;
0.554;
0.456;
1
NTA
Niño3
Niño3.4
Pac-Atl
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
140
0.6007; 0.3597;
0.0839
0.6044; 0.2809;
0.5626; 0.2872;
0.3331; 0.2852
0.2725
0.6109; 0.3727;
0.3070
0.5534; 0.0738
0.5584; 0.1047
0.5617; 0.1430
0.5274; 0.4085
0.5297; 0.4877
0.4354
0.6676; 0.4890;
0.2218; 0.0842
0.6789; 0.3573;
0.1339
0.5904; 0.2670
Number
of used EOFs,
SST|precipitation
5|6
0.561;
0.526
0.569;
0.541;
0.538;
4|6
0.517;
0.497;
4|6
0.315; 0.304
0.514
0.554;
0.544;
0.505
0.495; 0.491
0.497; 0.489
0.444; 0.431
0.595; 0.598
0.590; 0.595
0.624
0.611; 0.607;
0.590; 0.590
0.579; 0.553;
0.554
0.510; 0.507
3
3
2
1
3
3
2
3
2
4
|
|
|
|
|
|
|
|
|
|
5
1
1
1
1
1
1
2
1
6
4
2
2
1
1
3
3
3
3
3
3
7
7
|
|
|
|
|
|
|
|
|
|
|
|
|
2
1
1
1
1
4
6
6
1
6
6
2
6
7|5
7|3
Appendix
and of precipitation (station data) in November in Vargas, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.0742
0.2226
0.1471
0.1819
0.3396
0.4154
0.2593; 0.2087
0.4290
0.4437
0.4281
0.4959; 0.0639
0.4111
0.3826
0.3065
0.4888
0.4785
0.3562
0.4361; 0.0546
0.4116
0.4183
0.3867
0.4515
0.4040
0.4790
0.3874
0.4139
0.4383
0.3826
0.002
0.058
0.029
0.050
0.110
0.144
-0.079; -0.050
0.269
0.277
0.257
0.264; 0.248
0.223
0.187
0.114
0.281
0.273
0.209
0.208; 0.195
0.188
0.189
0.164
0.253
0.252
0.254
0.226
0.209
0.203
0.103
Number
of used EOFs,
SST|precipitation
1|1
3|1
1|1
1|1
3|1
3|1
3|2
1|1
1|1
1|1
2|2
1|1
1|1
1|1
2|1
2|1
1|1
2|2
2|1
2|1
2|1
3|1
1|1
2|1
1|1
2|1
2|1
2|1
141
Appendix
and of precipitation (station data) in December in Vargas, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niñ03.4
Pac-Atl
142
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5334;
0.4156
0.4286;
0.2362;
0.2581;
0.2438
0.2426
0.3508
0.3461
0.3780
0.3490
0.3908
0.4537
0.4269
0.3220
0.3268
0.3662
0.3419
0.3398
0.3842
0.4669
0.3992
0.3829
0.3991
0.3939
0.4242
0.4297
0.4212
-0.034; 0.057
0.041
0.036; 0.117
-0.068; 0.013
0.045; -0.080
-0.063
-0.033
0.214
0.215
0.239
0.218
0.255
0.289
0.259
0.191
0.196
0.222
0.203
0.204
0.226
0.261
0.262
0.252
0.260
0.254
0.276
0.276
0.206
0.2116
0.2281
0.2275
0.0850
Number
of used EOFs,
SST|precipitation
3|2
1|2
3|2
3|2
3|2
3|1
3|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
2|1
1|1
1|1
1|1
1|1
1|1
1|1
2|1
Appendix
and of precipitation (station data) in February in Vargas, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5045;
0.4640;
0.2519
0.0241
0.2166
0.3314
0.2838
0.3880
0.3513
0.3908
0.2993
0.3038
0.3570
0.2518
0.4268;
0.3730
0.3597
0.3423
0.3492
0.3098
0.3026
0.4540;
0.5175;
0.4473
0.4221
0.3702;
0.3833
0.4476
0.113; -0.055
0.113; 0.046
0.150
-0.232
-0.158
0.002
0.054
0.182
0.176
0.173
0.135
0.128
0.080
0.079
0.160; 0.135
0.184
0.181
0.162
0.161
0.119
0.122
0.189; 0.182
0.217; 0.125
0.215
0.146
0.108; 0.134
0.086
0.214
0.2266
0.1678
0.0891
0.1981
0.0330
0.1709
Number
of used EOFs,
SST|precipitation
3|2
3|2
1|1
1|2
3|1
3|1
3|1
1|2
1|1
2|1
1|1
1|1
2|1
1|1
2|2
1|1
1|1
1|1
1|1
1|1
1|1
2|2
2|2
2|1
3|1
2|2
2|1
3|1
143
Appendix
and of precipitation (station data) in November in Bolívar, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niño3.4
Pac-Atl
144
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.4316;
0.5710;
0.3087;
0.4591;
0.4555;
0.3179
0.3584
0.4456
0.4600
0.4418
0.4394
0.4744
0.5409
0.3869
0.5238
0.4820
0.5131
0.4365
0.4490
0.4973
0.4287
0.4682
0.6431
0.7613;
0.5929;
0.6424
0.5366
0.4327
-0.069; -0.013
-0.002; -0.030
0.043; 0.053
0.172; 0.099
0.113; 0.140
0.169
0.197
0.264
0.266
0.247
0.244
0.262
0.263
0.083
0.303
0.278
0.291
0.224
0.271
0.201
0.129
0.213
0.325
0.385; 0.357
0.297; 0.268
0.361
0.180
0.059
0.1801
0.3245
0.2334
0.1922
0.1560
0.2438
0.1288
Number
of used EOFs,
SST|precipitation
2|2
3|2
2|2
3|2
3|2
2|1
2|1
1|1
1|1
1|1
1|1
1|1
3|1
3|1
2|1
2|1
2|1
2|1
2|1
2|1
3|1
3|1
3|1
4|2
3|2
3|1
3|1
2|1
Appendix
and of precipitation (station data) in December in Bolívar, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3962;
0.0917
0.1599;
0.3889;
0.2430;
0.3860
0.3574
0.5731
0.5389
0.5932
0.6192
0.5817
0.6063
0.6070
0.5591
0.5440
0.5568
0.5493
0.5093
0.5082
0.4117
0.5333
0.5163
0.4968
0.5213
0.5580
0.5858
0.5588
-0.009; -0.069
0.098
0.159; -0.028
0.012; 0.121
-0.079; -0.261
0.176
0.210
0.320
0.315
0.345
0.322
0.303
0.336
0.303
0.330
0.333
0.341
0.337
0.295
0.291
0.230
0.331
0.329
0.312
0.317
0.243
0.335
0.315
0.0692
0.0376;
0.3175
0.0038
Number
of used EOFs,
SST|precipitation
2|2
1|1
2|2
2|2
2|2
2|1
2|1
2|1
2|1
2|1
2|1
2|1
2|1
3|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
1|1
2|1
3|1
3|1
3|1
145
Appendix
and of precipitation (station data) in February in Bolívar, SST leading the precipitation up to six
months
Region
NTA
Niño3
Niño3.4
Pac-Atl
146
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3827;
0.3151;
0.2637
0.2317;
0.1878
0.4688
0.4413
0.6782;
0.4043
0.3533
0.3537
0.3773
0.3345
0.3654
0.3846
0.4116
0.3722
0.3497
0.3721
0.3658
0.3763
0.3867
0.4657
0.5082
0.4076
0.4892
0.5544
0.5576
-0.039; 0.071
0.019; -0.184
0.214
0.242; 0.064
0.196
0.307
0.335
0.248; 0.190
0.216
0.173
0.209
0.257
0.223
0.239
0.248
0.273
0.218
0.211
0.259
0.259
0.254
0.180
0.312
0.285
0.314
0.399
0.438
0.441
0.2614
0.0953
0.1611
0.3053
Number
of used EOFs,
SST|precipitation
2|2
2|2
1|1
2|2
1|1
2|1
1|1
2|2
2|1
1|1
1|1
1|1
1|1
2|1
1|1
1|1
1|1
1|1
1|1
1|1
2|1
2|1
2|1
4|1
2|1
2|1
2|1
2|1
Appendix
precipitation in November at the coast (model data of the past), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.6918
0.6130;
0.2796
0.4879;
0.3875;
0.1252
0.3756
0.3026
0.2949
0.3199
0.3713;
0.4275;
0.3777
0.3061
0.3405;
0.2634
0.2950
0.2097
0.2115
0.3427;
0.3523;
0.7792
0.4814
0.3537
0.5906;
0.2628
0.4675;
0.5483;
0.523
0.328; 0.298
0.063
0.310; 0.277
0.059; -0.044; -0.099
-0.015
-0.017
0.215
0.206
0.237
0.092; 0.218
0.072; 0.283; 0.280
0.284
0.182
0.163; 0.156
0.169
0.230
0.135
0.127
0.172; 0.227
-0.149; 0.209
0.567
0.301
0.246
0.277; 0.289
0.191
0.159; 0.217
-0.015; 0.196
0.3446
0.1537
0.2335; 0.0669
0.2909
0.3674; 0.0902
0.0178
0.2062
0.3313
0.4977
0.3270
0.5089
Number
of used EOFs,
SST|precipitation
7|1
9|2
3|1
4|2
3|3
1|1
7|1
1|1
1|1
1|1
3|2
3|3
2|1
2|1
2|2
1|1
1|1
1|1
1|1
2|2
2|2
1|1
5|1
2|1
10 | 2
1|1
5|2
10 | 2
147
Appendix
precipitation in December at the coast (model data of the past), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
148
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.4322;
0.6204;
0.4199
0.4079;
0.3072
0.4525;
0.4642
0.3066
0.2354
0.5006;
0.3294;
0.3140;
0.3732;
0.3189;
0.2620
0.4366;
0.3145
0.3555;
0.1547
0.4099;
0.1657
0.5371
0.7388;
0.5332
0.2800
0.4810;
0.5901;
0.3511
0.226; 0.126
0.063; 0.276
0.248
-0.161; -0.041
0.179
-0.072; -0.099; 0.101
0.318
0.240
0.151
0.101; 0.052; 0.069
0.028; -0.003
.0.064; -0.103
0.002; -0.059
-0.031; -0.099
0.191
0.196; 0.006
0.132
0.086; 0.009
0.012
0.126; 0.011; -0.017
0.068
0.357
0.197; 0.304; 0.329
0.323
0.175
-0.008; 0.171; 0.088
0.013; 0.273
0.190
0.1702
0.4919
0.2328
0.3671; 0.3196
0.2954; 0.1576
0.0807
0.0629
0.1827
0.1430
0.1498
0.0636
0.2085; 0.0548
05410; 0.4029
0.4032; 0.2069
0.5092
Number
of used EOFs,
SST|precipitation
5|2
8|2
4|1
3|2
3|1
5|3
4|1
1|1
1|1
4|3
4|2
4|2
5|2
2|3
1|1
4|2
1|1
2|3
1|1
3|3
1|1
7|1
10 | 3
10 | 1
2|1
7|3
10 | 2
4|1
Appendix
precipitation in February at the coast (model data of the past), SST leading the precipitation up to
six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5069;
0.3358;
0.4115;
0.4144
0.7460;
0.3862
0.5711;
0.3460
0.4009;
0.4803;
0.3674
0.4167
0.3557
0.5112;
0.3631
0.3498
0.3821
0.3753
0.3711
0.3088
0.3425
0.3785
0.4359;
0.4450
0.6477
0.5026;
0.4876;
0.6463
-0.087; -0.105; -0.169
-0.137; -0.372; -0.398
-0.113; -0.359; -0.429
0.155
0.179; 0.262; 0.234
0.248
0.332; 0.347; 0.327
0.292
0.191; 0.274
0.116; 0.275
0.295
0.345
0.284
0.282; 0.315
0.309
0.285
0.304
0.301
0.299
0.321
0.276
0.252
0.300; 0.241
0.306
0.326
0.288; 0.217
0.307; 0.345
0.354
0.3623; 0.0902
0.1156; 0.0132
0.2655; 0.1187
0.4671; 0.2423
0.2188; 0.1327
0.2270
0.2585
0.2650
0.1409
0.0678
0.3263
Number
of used EOFs,
SST|precipitation
8|3
3|3
6|3
5|1
9|3
2|1
5|3
1|1
2|2
3|2
1|1
1|1
2|1
4|2
1|1
1|1
2|1
1|1
1|1
1|1
1|1
2|1
2|3
4|1
1|1
2|2
3|2
10 | 1
149
Appendix
precipitation in November in the inland (model data of the past), SST leading the precipitation up
to six months
Region
Lag
CCA coefficient
Goodness index
NTA
0
1
2
3
0.6862;
0.6111;
0.4041
0.6290;
0.1568;
0.5896;
0.2022,
0.5927;
0.1691;
0.2813
0.5702;
0.3310
0.3144
0.5157;
0.0216
0.5920;
0.5561;
0.6491;
0.0357
0.3068
0.3158
0.3077
0.2224
0.5629;
0.2007
0.4953;
0.5440;
0.7938;
0.3223
0.6648
0.6485;
0.2021
0.7393;
0.7926;
0.5004;
0.6227;
0.5715;
0.506; 0.481
0.350; 0.272
0.228
0.275; 0.342; 0.353;
0.324; 0.326
-0.149; 0.091; 0.068;
0.033; 0.035
0.006; 0.086; 0.144;
0.097; 0.076
0.151
0.226; 0.283; 0.252
0.254
0.234
0.240; 0.193; 0.188;
0.191
0.345; 0.354; 0.321
0.212; 0.299
0.393; 0.345; 0.318;
0.315
0.213
0.221
0.223
0.104
-0.054; -0.013; 0.194;
0.167
0.292; 0.267
0.293; 0.274
0.522; 0.569; 0.582;
0.578
0.510
0.430; 0.360; 0.364;
0.370
0.481; 0.431; 0.461
0.239; 0.084; 0.071;
0.117; 0.171
0.048; 0.312; 0.298
0.185; 0.349
4
5
Niño3
6
0
1
2
3
4
5
6
Niño3.4
0
1
2
3
4
5
6
0
1
2
Pac-Atl
3
4
5
6
150
0.2513
0.2045
0.5363; 0.4447;
0.1295
0.5047; 0.2656;
0.0571
0.5371; 0.3520;
0.1076
0.4599; 0.1317
0.2640; 0.2144;
0.4004; 0.0303
0.3167
0.3426; 0.1556;
0.4263; 0.3231;
0.3164
0.2234
0.6294; 0.4956;
0.3259; 0.2721;
0.5439; 0.4337
0.6243; 0.5840;
0.3733
0.5848; 0.2823
0.4896
Number
of used EOFs,
SST|precipitation
6|2
6|2
3|1
5|6
5|5
6|5
2
3
1
1
4
|
|
|
|
|
1
6
1
1
4
4|3
3|2
4|4
1
1
1
1
4
|
|
|
|
|
1
1
1
1
5
2|4
2|3
9|4
6|1
4|5
10 | 3
1|5
7|3
4|2
Appendix
precipitation in December in the inland (model data of the past), SST leading the precipitation up
to six months
Region
Lag
CCA coefficient
Goodness index
0
1
0.5599;
0.6503;
0.2893;
0.5027;
0.1158
0.1832
0.2419
0.1996
0.5323;
0.1989
0.1622
0.3485;
0.3791;
0.4256;
0.4140;
0.4572
0.4074
0.4273;
0.2548
0.3519
0.1096
0.3288
0.3388
0.3563
0.6639;
0.7124;
0.2285
0.5348;
0.6277;
0.3767;
0.5079;
0.7156;
0.3043
0.127; 0.313
0.264; 0.216; 0.205;
0.196; 0.154
0.191; 0.234; 0.224;
0.197
0.107
0.107
0.104
0.137; 0.234; 0.266;
0.268
0.104
0.013; 0.130
0.011; 0.146
-0.030; 0.017; -0.071
-0.131; -0.265
-0.061
-0.102
0.030; 0.103; 0.082
0.096
0.231
-0.014
-0.095
-0.038
-0.027
0.064; 0.436
0.430; 0.324; 0.377;
0.371
0.268; 0.304; 0.289
0.147; 0.096; 0.104
0.005; 0.081
0.098; 0.062; 0.239
0.163; 0.133; 0.267;
0.242
2
NTA
Niño3
Niño3.4
Pac-Atl
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.3870
0.5314; 0.3428;
0.1094
0.3903; 0.2559;
0.4429; 0.3272;
0.2392
0.3055
0.3186; 0.0609
0.0517
0.2683; 0.0428
0.5847
0.4576; 0.4189;
0.3578; 0.1092
0.3898; 0.3217
0.1830
0.4712; 0.4197
0.6243; 0.5356;
Number
of used EOFs,
SST|precipitation
5|2
8|5
4|6
1
1
1
4
|
|
|
|
1
2
1
4
1
2
4
4
3
1
1
3
2
4
1
1
1
1
9
9
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1
2
2
3
2
5
3
3
1
1
1
3
3
3
2
4
5
1
2
7
2
|
|
|
|
|
3
3
2
3
4
151
Appendix
precipitation in February in the inland (model data of the past), SST leading the precipitation up
to six months
Region
NTA
Lag
CCA coefficient
Goodness index
0
1
2
3
4
0.1681
0.2226
0.2934
0.2605
0.6178;
0.0229
0.7489;
0.3186;
0.0461
0.6747;
0.3767
0.3579
0.3245
0.3869
0.3648
0.6023;
0.1912
0.3848
0.3414
0.3431
0.6289;
0.4495;
0.4065;
0.3572
0.4125
0.5442;
0.3865
0.3778
0.3885
0.5352;
0.4084
0.5603
0.034
0.104
0.187
0.160
0.174;
0.091
0.295;
0.140;
0.177
0.310;
0.279
0.245
0.227
0.283
0.273
0.254;
0.211
0.294
0.240
0.243
0.303;
0.290;
0.275;
0.258
0.305
0.284;
0.275
0.270
0.273
0.268;
0.289
0.309
5
Niño3
Niño3.4
Pac-Atl
152
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5043; 0.1716;
0.4741; 0.3608;
0.2598; 0.2194;
0.3310
0.4759; 0.3899;
4245; 0.3556
0.2150
0.2686
0.3902
0.1507
0.152;
0.104;
0.183;
0.155;
0.157;
0.183;
0.267
0.189;
0.196;
0.335; 0.312
0.318
0.290
0.248
0.230
Number
of used EOFs,
SST|precipitation
1|1
1|1
1|1
1|1
4|8
8|7
7
1
1
1
1
1
4
|
|
|
|
|
|
|
2
1
1
1
1
1
6
1
1
1
3
2
2
1
1
2
1
1
1
2
2
7
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1
1
1
7
2
2
1
1
8
1
1
1
4
1
1
Appendix
precipitation in November at the coast (model data of the future), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5822;
0.5373
0.4113
0.2501;
0.2605;
0.4613;
0.2009
0.4194
0.4313
0.3938
0.4327
0.5171
0.5974;
0.4829
0.4165
0.4160
0.4108
0.3890
0.5319
0.4843;
0.4976;
0.6456;
0.5627;
0.5150;
0.5417
0.5814;
0.6049
0.4830
0.287; 0.250; 0.228
0.399
0.308
0.164; 0.106
0.189; 0.119
-0.009; 0.153
-0.074
0.385
0.396
0.354
0.398
0.430
0.292; 0.267; 0.275
0.281
0.379
0.381
0.372
0.355
0.407
0.302; 0.293
0.286; 0.254; 0.266
0.448; 0.423; 0.418
0.454; 0.449
0.374; 0.417; 0.414
0.411
0.405; 0.395
0.425
0.322
0.2660; 0.1068
0.0360
0.0119
0.2262
0.1230; 0.0321
0.0357
0.1625; 0.0836
0.2759; 0.1266
0.2828
0.3217; 0.0140
0.1630
Number
of used EOFs,
SST|precipitation
6|3
4|1
4|1
2|2
2|2
3|2
3|1
1|1
1|1
1|1
2|1
2|1
3|3
5|1
1|1
1|1
1|1
1|1
3|1
3|2
4|3
5|3
3|2
3|3
5|1
5|2
8|1
5|1
153
Appendix
precipitation in December at the coast (model data of the future), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
154
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.5757;
0.4713;
0.5379
0.5351;
0.3759;
0.4031;
0.3404;
0.4979
0.4942
0.3851
0.3658
0.3718
0.4340
0.5945;
0.4061
0.4216
0.4316
0.3873
0.3516
0.3806
0.5726;
0.7127;
0.4324
0.4537
0.5080
0.4845
0.4451
0.5427
0.309; 0.337; 0.316
-0.212; -0.115; -0.009
0.223
0.152; 0.105; 0.109
-0.167; 0.107
0.120; 0.010; -0.042
0.012; -0.078; -0.154
0.390
0.363
0.335
0.326
0.296
0.319
0.406; 0.381; 0.381
0.351
0.358
0.381
0.321
0.277
0.341
0.354; 0.327; 0.330
0.394; 0.373; 0.347
0.332
0.375
0.394
0.350
0.287
0.258
0.4598; 0.1199
0.4263; 0.1822
0.3195; 0.1645
0.2709
0.2376; 0.0635
0.1687; 0.0367
0.1356; 0.1143
0.1285; 0.0490
0.4222; 0.2404
Number
of used EOFs,
SST|precipitation
5|3
6|3
5|1
4|3
3|2
3|3
3|3
3|1
3|1
1|1
1|1
2|1
2|1
4|3
1|1
1|1
1|1
1|1
1|1
1|1
3|3
7|3
2|1
2|1
3|1
3|1
3|1
8|1
Appendix
precipitation in February at the coast (model data of the future), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.2802
0.2662
0.3483;
0.3706;
0.4791;
0.4745;
0.4947;
0.3514;
0.1430
0.2477;
0.1546
0.2487
0.2288
0.3735;
0.2908;
0.1267
0.4283;
0.1291
0.3768
0.3069;
0.3319
0.4859;
0.1364
0.4257
0.3411;
0.5195;
0.5308;
0.3951;
0.039
0.017
-0.110; -0.273
-0.076; -0.227
-0.128; -0.227; -0.280
-0.056; -0.104; -0.126
0.026; 0.078; 0.021
0.050; -0.035
-0.082
-0.023; -0.071
-0.045
0.029
0.000
0.075; 0.017
-0.081; -0.131; -0.148
-0.127
-0.049; -0.105; -0.131
-0.096
0.101
0.022; -0.008
0.066
0.105; 0.051; 0.046
-0.085
-0.040
-0.114; -0.308
0.002; -0.105; -0.177
-0.075; -0.162; -0.096
-0.067; -0.117
0.1956
0.2021
0.1770; 0.0905
0.2609; 0.1551
0.3185; 0.2511
0.1329
0.1554
0.1013
0.1330; 0.0310
0.1289; 0.0187
0.0596
0.3232; 0.1540
0.1777
0.3109; 0.1562
0.3502; 0.2935
0.1821
Number
of used EOFs,
SST|precipitation
2|1
2|1
5|2
6|2
5|3
5|3
5|3
2|3
1|1
3|2
1|1
2|1
2|1
4|2
3|3
1|1
3|3
1|1
3|1
3|2
3|1
6|3
1|1
6|1
6|2
7|3
7|3
6|2
155
Appendix
precipitation in November in the inland (model data of the future), SST leading the precipitation
up to six months
Region
NTA
Niño3
Lag
CCA coefficient
Goodness index
0
1
2
0.4080;
0.6135
0.6649;
0.1183
0.4905;
0.4133;
0.3266;
0.2274
0.5870
0.6762
0.5395
0.5203
0.4953
0.3393
0.6325;
0.1008
0.5543
0.6757;
0.1628
0.5523
0.5056
0.4669
0.3414
0.6010;
0.0097
0.7098;
0.7072;
0.6698;
0.6262;
0.6731;
0.1320;
0.6630
0.5544
0.373; 0.367
0.506
0.100; 0.382; 0.414;
0.410
0.208; 0.367; 0.343
0.341; 0.307
0.262; 0.175
0.186
0.507
0.549
0.454
0.421
0.394
0.272
0.242; 0.322; 0.271;
0.263
0.474
0.479; 0.471; 0.472;
0.465
0.468
0.421
0.396
0.281
0.330; 0.325; 0.309;
0.306
0.489; 0.534; 0.527
0.547; 0.536
0.348; 0.481
0.291; 0.438; 0.426
-0.013; 0.360; 0.410;
0.394; 0.389
0.470
0.377
3
4
5
6
0
1
2
3
4
5
6
0
1
Niño3.4
Pac-Atl
2
3
4
5
6
0
1
2
3
4
5
6
156
0.0318
0.6044; 0.4584;
0.3948; 0.0730
0.3135
0.1323
0.5183; 0.2554;
0.4364; 0.3479;
0.4107; 0.1614;
0.4419; 0.2578
0.2929
0.5996
0.5571; 0.0353
0.5890; 0.4781;
0.0398
Number
of used EOFs,
SST|precipitation
2|2
4|1
4|6
3
2
2
1
1
4
1
2
2
1
5
|
|
|
|
|
|
|
|
|
|
|
3
3
3
1
1
1
1
1
1
1
4
1|1
4|6
1
1
1
1
4
|
|
|
|
|
1
1
1
1
4
5
5
2
3
5
|
|
|
|
|
3
2
6
5
6
8|1
6|1
Appendix
precipitation in December in the inland (model data of the future), SST leading the precipitation
up to six months
Region
Lag
CCA coefficient
Goodness index
0
0.6146;
0.0501
0.6502;
0.2055;
0.5757;
0.4054;
0.0281
0.4009;
0.4901;
0.0137
0.4807;
0.5100
0.5161
0.5098;
0.5831;
0.0612
0.3396
0.3350
0.6223;
0.0649
0.6012;
0.5358
0.4134
0.5160;
0.3616
0.3734
0.5757;
0.7899;
0.2677;
0.6649;
0.1176
0.6987;
0.2274;
0.6739;
0.0469
0.6466;
0.6307;
0.6460;
0.315; 0.281; 0.264;
0.255
0.094; 0.114; 0.008;
0.026; 0.056; 0.043
0.159; 0.307; 0.291
-0.057; 0.067; -0.003;
0.003;
0.071; -0.014; -0.014
-0.061; 0.091; 0.023;
0.022
-0.219; -0.057; -0.101
0.352
0.334
0.270; 0.242
0.266; 0.253; 0.215;
0.220
0.274
0.278
0.377; 0.387; 0.358;
0.343
0.363; 0.315
0.393
0.347
0.328; 0.307
0.279
0.310
0.243; 0.233; 0.215
0.415; 0.429; 0.464;
0.439; 0.420; 0.424
0.355; 0.317; 0.329;
0.313
0.347; 0.363; 0.382;
0.367; 0.359; 0.351
0.400; 0.446; 0.427;
0.416
0.429; 0.402
0.377; 0.335
0.263; 0.388; 0.358
1
NTA
2
3
4
5
Niño3
6
0
1
2
3
4
5
6
Niño3.4
0
1
2
3
4
5
6
0
1
Pac-Atl
2
3
4
5
6
0.3873; 0.1986;
0.4873; 0.3114;
0.1624; 0.0400
0.4439; 0.1217
0.3507; 0.0950;
0.1061; 0.0157
0.2852; 0.1083;
0.3779; 0.1136
0.0499
0.3354; 0.1628;
0.3986; 0.2614;
0.0412
0.0799
0.2759; 0.0655
0.5887; 0.4602;
0.1969; 0.1054
0.3633; 0.3427;
0.4947; 0.4631;
0.1452; 0.0549
0.4311; 0.2662;
0.0980
0.0902
0.5324; 0.2744
Number
of used EOFs,
SST|precipitation
5|4
6|6
5|3
4|4
3|3
4|4
3
3
3
3
4
|
|
|
|
|
6
1
1
2
5
1|1
1|1
4|6
2
3
2
3
1
1
3
7
|
|
|
|
|
|
|
|
3
1
1
2
1
1
4
6
5|4
6|6
4|6
3|2
3|2
8|3
157
Appendix
precipitation in February in the inland (model data of the future), SST leading the precipitation up
to six months
Region
NTA
Niño3
Niño3.4
Pac-Atl
Lag
CCA coefficient
Goodness index
0
1
2
0.4061;
0.5174;
0.5773;
0.1434
0.5104;
0.5076;
0.5970;
0.5770;
0.4814;
0.3336
0.3306
0.2880
0.3858
0.2394
0.4721;
0.3952;
0.3151
0.3285
0.3171
0.4503;
0.3106
0.4568;
0.6131;
0.4518;
0.5992;
0.5446;
0.0402
0.6171;
0.6017;
0.4534;
-0.002; 0.088
0.073; 0.091; 0.103
-0.029; 0.110; 0.101;
0.087
0.060; 0.120; 0.179
-0.036; 0.189
0.223; 0.302
0.035; 0.302
0.068; 0.241
0.165
0.179
0.157
0.212
0.128
0.151; 0.191
0.206; 0.179
0.160
0.183
0.181
0.138; 0.188
0.163
0.135; 0.188
0.085; 0.311
0.002; 0.248
0.036; 0.290
0.048; 0.085; 0.249;
0.245
-0.005; 0.277
-0.002; 0.214
-0.040; 0.240
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
158
0.2204
0.2733; 0.0636
0.5153; 0.3433;
0.4407; 0.3197
0.3314
0.3961
0.4518
0.3276
0.2187
0.0789
0.1973
0.2238
0.5466
0.4175
0.5132
0.4607; 0.4016;
0.4109
0.4200
0.3674
Number
of used EOFs,
SST|precipitation
3|2
3|4
4|8
6
5
5
5
3
1
1
1
2
1
3
2
1
1
1
2
1
3
6
3
6
4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3
2
2
2
2
1
2
1
1
2
2
2
1
2
2
2
2
2
2
2
2
4
5|2
5|2
3|2
Bibliography
La Armada Bolivariana (ARMADA), Consejo Nacional de Investigaciones Científicas y
Tecnológicas (CONICIT), Corporacion Venezolana de Guayana–Electrificación del Caroní
(CVG–EDELCA), Fondo Nacional de Investigaciones Agropecuarias (FONAIAP),
Ministerio del Ambiente y Recursos Naturales (MARN), Servicio de Meteorología de la
Fuerza Aérea Venezolana (SEMETFAV), Universidad Central de Venezuela (UCV), Programa para la Modernización del Sistema de Medición y Pronóstico Hidrometerológico
Nacional (VENEHMET) 2000. Situación meteorologica generadora de las inundaciones y
flujo de lodo en el norecentro de Venezuela, con especial énfasis en la costa norte (16 de
diciembre de 1999). Caracas, Mayo de 2000, made available by Juan Arévalo.
Ashby, S, Taylor, M, Chen, A 2005. Statistical models for predicting rainfall in the Caribbean. Theoretical and Applied Climatology 82, 65–80, DOI 10.1007/s00704–004–0118–8.
Cárdenas, P, Garcia, L F, Gil, A 2002. Impacto de los eventos El Niño–Oscilacion del
Sur en Venezuela. Corporación Andina de Fomento.
Cárdenas, P, Martelo, M T, Garcia, L F, Gil, A 2003. Impacto de los eventos El Niño–Oscilacion
del Sur en Venezuela. Parte II. Corporación Andina de Fomento.
159
Bibliography
Comisión Económica para América Latina y el Caribe (CEPAL) 2000. Los efectos
socioeconómicos de las inundaciones y deslizamientos en Venezuela en 1999. Sede Subregional en México Programa Naciones Unidas para el Desarrollo, Caracas, Febrero 2000.
Chen, A, Taylor, M 2002. Investigating the link between early season caribbean rainfall and the El Niño+1 year. International Journal of Climatology, Vol.22, 87–106 (2002),
DOI: 10.1002/joc.711.
Climate Predictability Tool (CPT) User Guide. International Research Institute for Climate and Society, The Earth Institute of Columbia University, made available by Ashley
Curtis.
Enfield, D and Alfaro, E 1999. The Dependence of Caribbean Rainfall on the Interaction of the Tropical Atlantic and Pacific Oceans. Journal of Climate, Vol 12, 2093–2103.
Enfield, D, Mayer, D 1997. Tropical Atlantic sea surface temperature variability and its
relation to El Niño-Southern Oscillation. Journal of Geophysical Research, Vol 12, 929–945.
Enfield, D 1996. Relationships of inter-American rainfall to tropical Atlantic and Pacific SST variability. Geophysical Research Letters, Vol.23, No.23, 3305–3308, November 15,
1996.
Everitt, B 2005.
An R and S–Plus Companion to Multivariate Analysis, chapter 8.
Springer-Verlag, pp 221.
FAV 2000. Análisis de las situaciones meteorológicas que afectaron al área norte
costera del país durante el mes de diciembre del año 1999. Informe preparado por el Servicio de Meteorología de la Fuerza Aérea de Venezuela.
Gray, W M 1968. Global view of the origin of tropical disturbance and storms. Monthly
Weather Review, Vol. 96, No. 10, 669 700, October 1968.
Guenni, L, Nobre, C, Marengo J, Huerta, G, Sanso, B. Oceanic Influence on Extreme
Rainfall Trends in the North Central Coast of Venezuela: Present and future climate assessments. unpublished manuscript.
Hagemann, S, Arpe, K, Roeckner, E 2006. Evaluation of the Hydrological Cycle in the
ECHAM5 Model. Journal of Climate–Special Section, Vol 19, 3801–3827.
160
Bibliography
Hastenrath, S L 1966. On general circulation and energy budget in the area of the Central
American seas. Journal of Atmospheric Sciences, Vol. 23, Issue 6, 694–711.
Hidalgo, L G 2001. Servicio de enlace para la Cooperación Internacional (Actuación de
las ONGS frente a la emergencia), pp 180. Lluvias torrenciales e inundaciones en el sector
norte-centro-costero de Venezuela (14-16 Dic 1999) pp 150-180. Red Venezolana de Organizaciones para el Desarrollo Social (REDSOC). Caracas.
Knaff, J A 1997. Implications of summertime sea level pressure anomalies in the tropical Atlantic region. Journal of Climate, Vol. 10, 789–804, April 1997.
Lau, K–M, Yang, S 2002. Walker Circulation. Encyclopedia of Atmospheric Sciences,
Academic Press, doi:10.1006/rwas.2002.0450.
Linares , Y, Davies, V, Duque, J R, Hernández, T 1999. Y el Ávila bajó al mar. Los
Libros De EL NACIONAL, 122 pp.
López, J L, Courtel, F, Bello, M T 2005. Las inundaciones, los aludes torrenciales y
sus medidas de prevención y mitigación. Instituto de Mecánica de Fluidos, Facultad de
Ingenería, Universidad Central de Venezuela, Caracas, Septiempre del 2005.
Lyon, B 2002. Modulation of seasonal precipitation in northern Venezuela and antecedent
rainfall to the floods and landslides of December 1999. Proceedings, 13th Symposium on
Global Change and Climate Variations, Orlando, 155–156.
Lyon, B 2003. Notes and Correspondence, Enhanced seasonal rainfall in northern Venezuela and the extreme events of December 1999. Journal of Climate, Vol 16, 2302–2306.
Maier–Reimer, E, Kriest, I, Segschneider, J, Wetzel, P 2005. The HAMburg Ocean Carbon
Cycle Model HAMOCC5.1 –Technical description release 1.1. Reports on Earth System
Science 14, 57 pp, Max–Planck Institut für Meteorologie, Hamburg.
Marsland, S J, Haak, H, Jungclaus, H J, Latif, M, Röske, F 2003. The Max–Planck–Institute
global ocean/sea ice model with orthogonal curvilinear coordinates. Ocean Modelling (5),
91–127.
161
Bibliography
Martelo, M T 2003 (I). Influencia de las variables macroclimáticas en el clima de Venezuela. Dirección de Hidrología, Meteorología y Oceanologia–Dirección General de Cuencas
Hidrográficas–MARN.
Martelo, M T 2003 (II). La precipitación en Venezuela y su relacíon con el sistema
climático. Dirección de Hidrología, Meteorología y Oceanologia–Dirección General de
Cuencas Hidrográficas–MARN.
Mestas-Nuñez, A, Enfield, D, Zhang, C 2007. Water Vapor Fluxes over the Intra–Americas
Sea: Seasonal and Interannual Variability and Associations with Rainfall. Journal of Climate, Vol 20, 1910–1922.
Mundaray M, R E 2005. Condiciones meteorologicas, que provocaron las inundaciones
en febrero del 2005 en las costad venezolanas. Trabajo especial de grado (Master’s Thesis), Universidad Central de Venezuela .
Penland, C, Matrosova, L 1998. Prediction of tropical Atlantic sea surface temperatures
using Linear Inverse Modeling. Journal of Climate, Vol 11, 483–496.
Poveda, G, Mesa, O J 1997. NOTES AND CORRESPONDENCE, Feedbacks between
Hydrological Processes in Tropical South America and Large–Scale Ocean–Atmospheric
Phenomena. Journal of Climate, Vol. 10, 2690–2702, October 1997.
Quan, X-W, Diaz, H F, Hoerling, M P 2004. Change of the Tropical Hadley Cell Since
1950. Advances in global change research, The Hadley Circulation: Past, Present, and
Future edited by Diaz, H F, Bradley R S, Kluwer Academic Publishers, 85–120.
Raddatz, T J, Reick, C J, Knorr, W, Kattge, J, Roeckner, E, Schnur, R, Schnitzler, K-G,
Wetzel P, Jungclaus, J 2007. Will the tropical land biosphere dominate the climate–carbon
cycle feedback during the twenty–first century?. Climate Dynamics, Vol. 29, 565–574,
doi: 10.1007/s00382–007–0247–8.
Roeckner, E, Bäuml, G, Bonaventura, L, Brokopf, R, Esch, M, Giorgetta, M, Hagemann,
S, Kirchner, I, Kornblueh L, Manzini, E, Rhodin, A, Schlese, U, Schulzweida, U, Tompkins,
A 2003. The atmospheric general circulation model ECHAM5, Part I. Model description,
Report No. 349, 127 pp, Max–Planck Institut für Meteorologie, Hamburg, November 2003.
162
Bibliography
Sequera P, P E 2009. Análisis de la dependencia de variables climáticas a gran escala
sobre la precipitación en Venezuela. Trabajo de Grado (Master’s Thesis), Universidad
Simón Bolívar.
Simmons, A J, Burridge, D M, Jarraud, M, Girard, C, Wergen, W 1989. The ECMWF
medium-range prediction models development of the numerical formulations and the impact of increased resolution. Meteorology and Atmospheric Physics, Vol. 40, 28–60,
Springer-Verlag.
Taylor, M A, Enfield, D, Chen, A 2002. Influence of tropical Atlantic versus the tropical Pacific on Caribbean rainfall. Journal of Geophysical Research, Vol. 107, No. C9,
3127, doi:10.1029/2001JC001097.
Wang, C, Kucharski, F, Barimalala, R, Bracco, A 2009. Teleconnections of the tropical
Atlantic to the tropical Indian and Pacific Oceans: A review of recent findings. Meteorologische Zeitschrift, Vol. 18, No. 4, 445–454, August 2009.
Wang, Ch 2007. Variability of the Caribbean Low-Level Jet and its relations to climate.
Climate Dynamics, Vol 29, 411–422, DOI 10.1007/s00382–007–0243–z.
Xue, Y, Smith, T M, Reynolds, R W 2003. NOTES AND CORRESPONDENCE Interdecadal Changes of 30–Yr SST Normals during 1871–2000. Journal of Climate, Vol.13,
1601–1612, 15 May 2003.
Zorita, E, Kharin, V, von Storch, H 1992. The Atmospheric Circulation and Sea Surface Temperature in the North Atlantic Area in Winter: Their Interaction and Relevance
for Iberian Precipitation. Journal Of Climate, Vol 5, 1097–1108, October 1992.
NCEP/NCAR Reanalysis I 4-times daily values downloaded from (27/08/2011, 31/08/2011,
01/09/2011):
http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html
NCEP/NCAR Reanalysis I anomalies and climatologies downloaded from (27/08/2011,
31/08/2011, 01/09/2011):
http://www.esrl.noaa.gov/psd/data/composites/day/
ENSO values taken from (February 2011):
http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml
163
Bibliography
SOI indices taken from (February 2011):
http://www.esrl.noaa.gov/psd/data/correlation/soi.data
QBO values taken from (19/08/2011):
http://www.esrl.noaa.gov/psd/data/correlation/qbo.data
Synoptic maps downloaded from (surface: 16/08/2011, 250 hPa: 21/08/2011):
http://nomads.ncdc.noaa.gov/ncep/dates
ERSST data downloaded from (17/03/2011, 08/09/2011):
http://iridl.ldeo.columbia.edu/SOURCES/.NOAA/.NCDC/.ERSST/.version3b/.anom/
CPT downloaded from (28/06/2011):
http://portal.iri.columbia.edu/portal/server.pt?open=512&objID=697&PageID=7264&cached=true&mode=2
Synoptic maps of the surface of the 7th of February 2005 18 UTC:
TROPICAL PREDICTION CENTER MIAMI FLORIDA, NOAA
164
Acknowledgements
My heartful thanks go to my mentors Prof. Dr. Martin Claußen from Hamburg and Prof.
Dr. Lelys Bravo de Guenni from Caracas who extensively supported and helped me with
advice and constructive comments throughout the last year.
Furthermore, special thanks go to Dr. Reiner Schnur from the University of Hamburg who
invested a lot of time discussing, explaining and correcting my work.
Thanks to Cynthia Eller and Jochen Hettchen who helped me with the English language.
I would like to thank my parents Birgit Dose and Klaus Tim and my brother Ole Tim for
their support during the whole time of my studies. They helped me with advice and their
faith in me at all times.
My friends Nele von Pein, Nele Cornils, Sarah Gabriel, Maja Voigt, Laura Roschewitz,
Moritz Geisthövel, Jonas Rauh, Lena Tepe, Ann-Kristin Naumann, Amelie Tetzlaff and
Stefanie Grünwald supported me during my time at the university.
During my time in Venezuela, there were many people around who extensively helped
me. I would like to mention Marco Polo, Juan Arévalo and Dr. Pedro Cárdenas from
INAMEH and Luis Felipe Garcia from Corpoelec, as well as Carlos Contrerar, Desiree
Villalta, Yannixia Abril and Raúl Ramirez from the Simón Bolívar University of Caracas.
All of them helped me to find my way around the University, with language issues and
with my work.
Moreover, a big thank you goes to my Venezuelan friends Aquiles Lacruz, Aimara Argotty
and Rodolfo Alvizu, Helian and Salvador. Muchas gracias to Mercedes Rojas Almeida for
her mental support, her help with language barriers, organisation of meetings and for her
general generosity.
My very special thanks go to my novio Demetrio José Benito Rojas Almeida. He supported
me during the whole time of my studies, he advised me, he believed in me and found the
right words for every situation. He was my backup in stressful times and made my time
in Venezuela an experience I will never forget.
165
This Master’s Thesis was written independently and without the aid of unfair or
unauthorised resources. Indications of sources are given whenever content was taken
directly or indirectly from other sources. This Master’s Thesis was not presented before
in an examination procedure. The written version is identical with the electronic one.
I agree to the publication of this Master’s Thesis.
Hamburg, den 22.12.2011
Nele Tim

Oceanic influence on precipitation in Venezuela, under

Transcription

Similar documents

Drivhuseffekt og global opvarmning

Free-air gravity

Drivhuseffekt og global opvarmning

Recent heavy rainfalls in Denmark Fun and facts Media

Direct Access Lab Test Order

Biomes Slides

BEACONSFIELD ASBESTOS PROSPECT

Exame Médico da Floresta Amazônica Brasileira: Situação Atual e