Using Google Earth to evaluate GCOS weather station sites

Weather – January 2009, Vol. 64, No. 1
Using Google Earth to evaluate
GCOS weather station sites
4
Ian Strangeways
TerraData, Wallingford, Oxfordshire
The Global Climate Observing System (GCOS)
was set up in 1992, co-sponsored by the
World Meteorological Organization (WMO),
the Intergovernmental Oceanographic
Commission (IOC) of UNESCO, the United
Nations
Environmental
Programme
(UNEP) and the International Council for
Science (ICSU) with the aim of improving
climate measurements. This was not a
new network of modern, high-precision,
automatic instruments but a selection from
amongst the best of the long-established
existing network of manual stations,
administered mostly by the world’s National
Meteorological Services (NMSs). These
stations use the traditional instruments
that have been in operation for all of the
twentieth century, and earlier, and which are
still widely used today (Strangeways, 2003).
In addition to instrument networks measuring land-surface conditions, GCOS also
includes networks measuring marine-air and
sea-surface temperature along with salinity
and barometric pressure from across the
oceans, subsurface ocean conditions using
the Argo floats, the radiosonde upper air
network as well as outflow measurements
of the world’s major rivers and the surface
radiation network. GCOS also concerns itself
with satellite measurements. In this article I
look at land surface meteorological stations,
in particular as regards their temperature
measurements.
The GCOS land-surface network is made
up of stations drawn from four sources
(Peterson et al., 1997): the Global Historical
Climatology Network (7283 stations); the
Climatic Research Unit’s dataset (2525 stations); WMO 1961–1990 ‘Normals Stations’
(stations that produce means for that 30year period) (3342 stations); and the list
of Reference Climatological Stations (RCS)
– specially selected, high-quality stations
with a long history of observations (2283
stations). According to WMO recommendations, each member country should
operate at least one RCS. Stations from
these four sources, totalling 15 433, were
sifted, to avoid duplication, and then ranked
according to length of record; data quality
(homogeneity); whether the station was an
RCS; whether the station reported in real
time over the GTS; the size of the surrounding population – to minimize urban heat
island (UHI) effects; good prospects of the
site continuing to operate into the future;
and the probability of the site remaining
substantially unchanged over time. Having
rated the stations according to these criteria, they were further selected depending
on their geographical position and how this
affected the calculation of gridded means.
The final selection is made up of just over
1000 stations (WMO 1, WMO 2).
The aim of GCOS was first to identify existing networks of stations suitable for generating datasets and secondly to improve
and upgrade these stations to meet the
more demanding requirements of climatemonitoring – requirements which are more
rigorous than those traditionally required for
simple meteorology and weather forecasting. For the task of climate monitoring, WMO
compiled a list of ten GCOS principles (WMO
3). This does not necessarily mean that the
stations are now all to the new standard and
meeting the required principles; it is up to
the individual countries whether they wish
to, can afford to, or are able to modify existing stations and procedures to meet the
new standards. Nevertheless, these stations
provide a substantial part of the data used
for the production of global climate datasets
and the more that can be done, therefore, to
enhance them for the future, the better. This
article will show how Google Earth can be
used as such an aid.
Google Earth
Google Earth allows users to look at any point
on Earth in considerable detail; anyone with
a PC can download the software and access
the images free of charge (Google). The PC
must have certain minimum requirements
and a broadband internet connection is
advisable for speed, but most users will have
an adequate set-up. The system uses images
collected from satellites and aircraft joined
together seamlessly. It has endless uses
from simply looking at one’s own house, to
its regular use in TV news reports. Here, its
use for looking at GCOS weather stations
sites on land is explored.
Using Google Earth to evaluate
GCOS sites
The advantage of being able to look closely
at any of the 1000+ GCOS land stations is
that much can be gleaned about the reliability, quality and representativeness of the
sites simply by looking at them in this way.
It is almost as good as flying over them in
a helicopter. While a snapshot picture of a
station as it is today (or as it was recently
– Google Earth images are not real-time) will
tell us little about its long-term history, there
will be many clues regarding its present
state as well as hints about its past, such
as the presence of trees around the site,
which can be assumed to have grown with
time. How much the site is embedded in a
town, city or airport, and its exact whereabouts within them, are all valuable pieces
of additional information to anyone using
the data from the sites for climate research,
particularly in allowing for any biases that
may have been introduced by the site itself,
and which could change slowly over time
as the sites changed, introducing a false
trend. The recent availability of Google Earth
enables, for the first time, useful detective
work of this kind to be done, and to be
done at no cost other than the time taken.
Such details are lacking in the WMO GCOS
list of stations and most researchers have so
far been working largely in the dark about
the physical nature of the sites and their
surroundings.
But there is a problem. The positions of the
GCOS sites are only specified in the WMO literature to one minute-of-arc resolution. This
resolution locates stations to within about a
km of their actual position. While this is generally adequate for identifying which grid
box to place them in when calculating areal
means, it is not precise enough to enable
the exact location of the site to be seen, just
the general surrounding area. To see the site
itself, to within a few metres, the locations
need to be specified to the nearest secondof-arc (or to the fourth decimal place if given
in decimal degrees).
However, the UK Met Office has kindly provided me with the National Grid References
of the six UK GCOS stations and these are
to higher resolution than the WMO coordinates, enabling the actual met. enclosures
The six UK GCOS weather station sites.
WMO
Latitude
Longitude
Elevation
Index Place name Degrees minutes seconds Degrees minutes seconds
(m)
Figure
3005
Lerwick
3026
3162
60
8 21.57082 N
1 11 10.45602 W
Stornoway
58 12 51.56459 N
Eskdalemuir
55 18 41.45749 N
3302
Valley
53 15
3377
Waddington
53 10 32.40429 N
0 31 23.67795 W
69
5
3808
Camborne
50 13
5 19 40.17160 W
84
6
7.84727 N
3.68642 N
78
2
6 19 10.48552 W
6
3
3 12 24.47368 W
217
4 32 15.52381 W
6
4
A wide view of the six stations
The Google Earth view of the UK from around
1500 km altitude (Figure 1) takes in all six
stations. An obvious feature is how the sites
are dominated by the sea, Eskdalemuir and
Waddington being the only inland sites. Nor
are the Midlands or the south-east of England
represented, although these regions are
represented in the (unconnected) Central
England Temperature record and in many
other UK datasets; we are looking here just
at GCOS sites.
Weather – January 2009, Vol. 64, No. 1
to be located to within about 50 m or less.
The six UK GCOS sites will, therefore, be used
to illustrate the potential of Google Earth in
evaluating the GCOS sites.
It is important to remember that GCOS
uses existing stations, generally National
Weather Service stations (NWSs), which have
at least a 30-year-long record. Such stations
in the UK (and globally) were not installed
for climate studies but for the immediate
and practical purposes of weather forecasting and more recently for aviation. The six
UK GCOS sites met these original requirements very well indeed, and still do, but this
does not mean that they also necessarily suit
GCOS requirements, which are somewhat
different. The same will apply to most stations globally. Table 1 gives details of the
UK sites.
Evaluating weather stations with Google Earth
Table 1
Lerwick
Figure 1. This Google Earth view of the UK, from an altitude of about 1470 km, shows the location
of all six GCOS sites. The strong influence of the sea is apparent. (© TerraMetrics 2008 http://www.
truearth.com; GeoPerspectives 2008; and GeoContent 2008.)
The Lerwick site in the Shetland Islands
(Figure 2) is clearly free from the influence
of any town or airport and will represent
the island’s weather and climate extremely
well. The station will, however, be strongly
influenced by its close proximity to the open
North Atlantic, except when wind speeds are
very low; a rarity in these latitudes and open
exposures. Readings will often, therefore, be
more representative of the temperature of
the air over the surrounding seas (Marine Air
Temperature – MAT) than that over the land.
Stornoway
Figure 2. The Lerwick site in the Shetland Islands. In common with the other UK sites, this figure shows
a wide-angle, three-dimensional view and (inset) a close-up view looking vertically downwards at
the location of the met. enclosure; in this case probably the small square enclosure to the bottom
right. The location of the station is shown in the wide view by a black square. See text for discussions.
(© TerraMetrics 2008 http://www.truearth.com; DigitalGlobe 2008.)
The precise location of the Stornoway site
is obscured by a small, thin cloud (Figure 3),
but we know where it is to within a few
metres and it is clear that the measurements will be dominated by the sea’s influence, even more so than at Lerwick, since
it is almost on the beach. But being on the
east side of the Isle of Lewis, the prevailing southwesterly winds will have travelled
across the island before reaching the site
and will have been modified during their
transit, to an extent dependent on conditions. Temperature readings will, therefore,
often be influenced by the sea. Being at an
airfield, complications may result, depending on the exact positioning of the instruments amongst the buildings, concrete and
5
Evaluating weather stations with Google Earth
Valley
Weather – January 2009, Vol. 64, No. 1
Valley, on Anglesey (Figure 4), is one of
those sites presented by Google Earth with
less pixel resolution than usual, making it
impossible to see the finer details of the site.
However, sufficient can be seen to be certain
that the site is within the airfield, and even
where it is located amongst the runways.
The fact that the site is again adjacent to the
coast, and on this occasion on the west side
of an island exposed to the prevailing wind,
tells us that, except under rare calm conditions, measured temperatures will normally
be closer to those that would be observed
over the sea rather than the land.
Waddington
Figure 3. The Stornoway site on the Isle of Lewis in the Outer Hebrides. (©TerraMetrics 2008 http://
www.truearth.com and DigitalGlobe 2008.)
tarmac, which we cannot quite see in this
image; but this will only occur in rare calm
conditions and so will be minor.
Eskdalemuir
Being obscured by cloud, the site is not
visible, making it impossible to judge the
fine details of the site, such as the number
and proximity of surrounding buildings and
trees, which can turn a seemingly rural site
into one with a semi-urban nature. In assessing a site, two main features are important
– the close-up details that allow a judgement to be made about its microclimate
(and whether this might be changing) and
the wide-angle view that tells us about how
well the site represents the area. The general
location of the site, just into Scotland, represents a typical inland and upland region, but
because it is not possible to see any local
detail, no figure is included.
This inland, lowland, site (Figure 5), is
probably the most representative of the
UK of all six GCOS stations. The clarity of
the image is also one of the highest of the
six sites, the Stevenson screen and other
instruments being just about recognizable.
The surrounding airfield may affect readings on calm days (and nights), the buildings and concrete affecting windspeed and
temperature, but this will probably be minor
since the open country is very close by. The
reason that sites are at airports in amongst
the buildings, or in towns, is because the
old manual instruments, still in wide use at
GCOS stations, need operators.
Yet a word of caution is necessary here. A
hurried, uncoordinated rush to change to
automatic instruments would be a disaster
for future measurements if it results in a host
of different new designs being deployed
globally. Great care is required in organizing
any change to automatic weather stations
(AWSs), not only as regards the need for an
overlap period for comparison of new with
old, but more importantly regarding the
actual design of the AWSs. A further article
will be prepared on the important matter
of appropriate station designs for future
climate monitoring.
Camborne
6
Figure 4. The Valley site on Anglesey. (© TerraMetrics 2008 http://www.truearth.com)
The site at Camborne (Figure 6) is clearly
visible and is at a good location, being
away from a town or airport, as at Lerwick.
However, it is just a few kilometres from
the west coast of Cornwall on this narrow stretch of land at the tip of England,
exposed to the prevailing westerly winds;
this coast is renowned for surfing. This
will make the influence of the sea strong
again, as at Lerwick, Stornoway and Valley.
Operating land stations near to discontinuities, like the coast, makes their day-to-day
measurements more sensitive to changes
in wind direction than sites surrounded by
more homogeneous landscapes, such as
at Waddington and Eskdalemuir. On some
days, the measurements will represent
the sea; on others, the land. It might be a
Other UK sites – Thurleigh
Google Earth is, of course, applicable to
all met. sites everywhere. An example of
a non-GCOS site in the UK is at the Royal
Aeronautical Establishment (RAE), Thurleigh,
in Bedfordshire (previously an RAF and USAF
airbase) (53° 13’ 36.9”N, 00° 27’ 55.68”W).
This site is also the WMO synoptic station
BEDFORD 03560. (Its location is given as 52°
13’N, 00° 29’W but, as with all WMO coordinates, it is only given to one minute-of-arc
and in consequence locates the station 1.8 km
Figure 6. The GCOS site near Camborne in Cornwall (not to be confused with Cambourne in
Cambridgeshire). The met. enclosure is probably the square enclosure seen at the bottom left.
(© GeoPerspectives 2008.)
Weather – January 2009, Vol. 64, No. 1
useful refinement, therefore, to differentiate
between the two states rather than to lump
all the measurements together.
Exactly where stations are, however, is
less important than their stability over time,
stability being vital for the detection of
small, climate-induced changes over long
periods. Where stations are sited, however,
might influence how stable they have been.
Temperature anomalies (changes relative to
a 30-year reference period) are less affected,
over a long period, by station location than
by station and site changes.
Evaluating weather stations with Google Earth
Figure 5. The Waddington GCOS site in Lincolnshire. (© GeoPerspectives 2008.)
away from its real location – see earlier
about WMO’s limited resolutions.)
This particular site has been chosen as a
practical example of the use of Google Earth
since it is relevant to the issue of changing
local land-use. Decommissioned in 1994, the
runways at Thurleigh are now used for the
mass storage of new cars (Figure 7). The presence of around 10 000 cars will inevitably
cause some change in temperature of the
air as it passes around them and this will be
reflected in the measured temperature at the
met. site just 200 m away. The dark-coloured
crops surrounding the site to the east and
south will also warm the air passing over
them more than a lighter surface would. The
longer-term changes in land-use represented
by the storage of cars, combined with the
year-by-year short-term changes in crops,
will combine and be reflected in the recorded temperature (pers. comm. Ledger and
Cox, 2008). Most ‘rural’ areas in the UK are not
stable ‘natural-rural’, but managed and fluid,
and so are not immune to changes that can
affect the local temperature. Google Earth
brings to light details of local influences such
as these in a way not previously possible.
It is important to note, however, that while
UHIs are a reality, what matters is not the
temperature bias introduced by the town or
airport but whether the bias changes over
time, which can happen if the sites change
slowly.
Global sites
A comparison between the accurate Met
Office coordinates and those of the WMO’s
at the six UK sites show that the average difference in fixing the locations is
about 700 m, ranging from 340 to 1160 m.
Assuming that this level of uncertainty is
reflected around the globe, it means that
even though it will not be possible to identify the actual met. enclosures when using
the lower-resolution WMO coordinates (and
thereby judge the microclimate they are
operating in), the WMO locations are good
enough to allow a judgement to be made
of how representative the sites are of the
surrounding country and whether they are
associated with an airfield or town. There is
not space here to investigate sites beyond
the UK, but suffice to say that from a preliminary look there are a high proportion
of airport locations worldwide, reflecting
the situation in the UK. Very often towns
are also associated with the stations, raising the possibility, if not the certainty, of a
UHI trend.
It is sometimes possible to locate sites
more precisely if the name of the station
in the WMO list is particularly descriptive,
such as ‘Jakarta Observatory’ (GCOS site
number 96745 at 06° 11’S, 106° 50’E). By
keying in the name, instead of the coordinates, the actual observatory is displayed by
Google Earth (Figure 8). (It should be noted
that Google Earth allows town names, post
7
Weather – January 2009, Vol. 64, No. 1
Evaluating weather stations with Google Earth
Conclusions
Google Earth is an invaluable tool for climate
research, allowing meteorological stations
to be seen and assessed both in close-up
(to evaluate a site’s current microclimate
and local influencing factors) and in the
wider view (to make a judgement about
the site’s representativeness and the likelihood of it having changed over time). This
gives those researchers assembling climate
datasets the ability to look at most of the
GCOS (and other) stations as if from a lowflying aircraft, and thereby make additional
judgements about them. WMO is, therefore,
urged to upgrade its station list to give all
locations to the enhanced resolution used
here. Today this is a simple task given the
wide availability of GPS navigation systems,
although whether it will be done is another
matter.
Figure 7. RAE Thurleigh, Bedfordshire. The met. site is shown in the lower left box (with a close-up
inset). The lower-right image is a wide-angle view of the whole area. The top section shows the
thousands of cars stored on the runways and the dark crops to the south. Inset is a section of the
runway showing the cars in more detail. (© GeoPerspectives 2008. Aerial image published with kind
permission from the GeoInformation Group. Cities Revealed aerial imagery © The GeoInformation
Group, 1998–2008.)
Acknowledgements
I would like to thank John Prior and David
Parker at the Met Office for their help in
reviewing an early draft of this paper and
for making many useful suggestions. Thanks
are due also to the several organizations
and companies holding copyrights on the
images for their kind permission to use
them without charge to illustrate the stations; detailed credits are given against each
figure. The images are a key part of this
article.
References
Figure 8. The Jakarta Observatory site, Java, Indonesia. (© DigitalGlobe 2008 and Europa
Technologies 2008.)
codes, sports centres and other forms of
site-identification to be used to search for a
site as well as latitude and longitude.) Such
inner-city locations, as here in Jakarta, chosen no doubt many decades ago for simple
meteorology, do not well suit the requirements of today’s climate studies because of
the unrepresentative nature of large cities
and because of the danger of temperature
8
biases slowly changing with time. The other
two GCOS sites in Java (at Cilacap at 07° 44’S,
109° 01’E and Sangkapura at 05° 51’S, 112°
38’E) appear equally unrepresentative of the
island. Whether they have remained stable
over time is, however, of greater importance
than their actual location; stations in or
near cities, however, are more likely to have
changed over time.
Peterson T, Daan H, Jones PD. 1997.
Initial selection of a GCOS surface network. Bull. Amer. Met. Soc. 78: 2145–2152.
Strangeways IC. 2003. Measuring the
natural environment (2nd edn). Cambridge
University Press: Cambridge, UK.
WMO 1 http://www.gosic.org/gcos/GCOSdev.htm (Click on ‘list of GSN Stations’)
[Accessed March 2008].
WMO 2 http://www.wmo.int/pages/prog/
gcos/documents/GSN_Station_Map.pdf
[Accessed March 2008].
WMO 3 http://www.wmo.int/pages/
prog/gcos/documents/GCOS_Climate_
Monitoring_Principles.pdf [Accessed
March 2008].
Google http://earth.google.com/intl/en/
userguide/v4/ [Accessed March 2008].
Correspondence to: Ian Strangeways,
7 Cherwell Close, Thames Street, Wallingford,
Oxfordshire, OX10 0HF, UK
Email: [email protected]
© Royal Meteorological Society, 2008
DOI: 10.1002/wea.334