Ten-Daily Global Composites of METOP-AVHRR

Ten-Daily Global Composites of METOP-AVHRR
Herman Eerens*a, Bettina Baruthb, Lieven Bydekerkea, Bart Derondea, Jan Driesa, Erwin Goora,
Walter Heynsa, Tim Jacobsa, Bart Oomsa, Isabelle Piccarda, Antoine Royerb, Else Swinnena,
Adri Timmermansa, Tom Van Roeya, Johan Vereeckena, Yves Verheijena
a
Vlaamse Instelling voor Technologisch Onderzoek (VITO-TAP), Boeretang, 2400-Mol, Belgium
Joint Research Centre (JRC-IPSC), Monitoring Agricultural Resources Unit (MARS), Ispra, Italy
b
ABSTRACT
Systematic scanning of the earth surface could be achieved for the first time in 1978, with the launch of the earth
observation system NOAA-AVHRR. Some twenty years later, the SPOT-VEGETATION instrument introduced
significant improvements at the levels of image quality, timeliness and availability. Since the start in April 1998, VITO
is responsible for the central processing, archiving and distribution of the VEGETATION data. This paper briefly
announces how a similar service is being established at VITO to provide the same kind of image data from the recently
launched METOP-AVHRR.
Keywords: NOAA-AVHRR, SPOT-VEGETATION, EUMETCAST, JRC-MARS.
1. INTRODUCTION & BACKGROUNDS
The history and applications of the AVHRR-sensor (Advanced Very High Resolution Radiometer) on board of the series
of the US NOAA-satellites has been thoroughly documented by Cracknell [1]. Originally designed for meteorological
purposes and to complement the equatorial view of the geostationary GOES-platforms, the NOAA-satellites follow a
standard, near-polar orbit (roughly at 830 km, 100° inclination, 14.1 cycles/day). With its wide swath of 2920 km, the
AVHRR sensor is capable to provide synoptic imagery with daily global coverage, be it only at a nadir-resolution of 1.1
km [2]. Since the start in 1978 many different NOAA-platforms were launched, and the last one so far was NOAA19
which became operational around July 2009. The objective has always been to maintain two instruments in space: the
one in the "morning orbit" scans the earth surfaces at 08h|20h, the other in the "noon orbit" passes by at around 14h|02h
(local solar times, ascending|descending paths, somewhat generalised and omitting the platform drift which occurred in
reality). Beyond the meteorological context, the synoptic AVHRR imagery soon proved very useful as well for the
monitoring of the land surfaces at continental to global scales. This gave rise to a number of innovative data sets with
global images composited at fixed intervals (dekad to month), such as AVHRR-Pathfinder [3] (8 km resolution, years
1981-2001), GIMMS [4] (8 km, 1981-present) and EROS [5] (1 km, some dekads over the years 1992-1995). Stimulated
by this new information, the research on global vegetation dynamics bloomed and numerous studies brought new
insights on the distribution of ecosystems and their susceptibility to changes caused by climatic and human factors (see
for instance [6] and [7]).
The NOAA-AVHRR system also had some technical shortcomings to name just a few: the poor stabilisation of the
platform, the lack of measures against orbital drift (retarding overpass times) and the absence of an on-board shortwave
calibration standard. But especially the limited on-board data storage facilities hampered the production of global
composite images at 1 km resolution, because most of the 1 km information was definitely lost if not registered in real
time (and stored) by a local ground receiving station. The VEGETATION sensor (VGT), carried since 1998 by the
European satellites SPOT-4/5, continued with the lessons learned from twenty years of operation of NOAA-AVHRR. It
tried to keep the positive elements, especially the daily global coverage at 1 km resolution, but without the mentioned
drawbacks [8, 9]. All the land data registered during a single orbit are stored on board and transmitted to the antenna of
Kiruna in Northern Sweden.
* Corresponding Author: E-mail [email protected]; phone 00 32 14 336872; fax 00 32 14 322795.
The further processing, archiving and data dissemination is performed by VITO's data centre (Centre de Traitement
d’Images Végétation, CTIV). The global ten-daily syntheses (VGT-S10) are by far the main product demanded by the
user community. The VGT-S10 are composite images, which comprise calibrated and atmospherically corrected reflectances (incl. NDVI), projected to a geographical Longitude/Latitude grid with a resolution of 1°/112 (roughly 1 km along
a great circle). As with AVHRR, the new VGT data boosted research on global vegetation mapping and monitoring (see
for instance [10] and [11]).
The European Commission's Joint Research Centre (JRC) in Ispra-Italy has always been a convinced user of remote
sensing data. Since 1989, the JRC-MARS unit (Monitoring Agriculture with Remote Sensing) co-ordinates all efforts to
improve, modernize and unify the collection of agro-statistics throughout the member states of the European Union.
Over the years, a generic Crop Growth Monitoring System (CGMS) was developed which combines different techniques
(GIS, meteo-processing, crop growth modelling and statistical calibration) to predict the growth state and yield of the
main crops throughout Europe [12]. Since the early days, the imagery of NOAA-AVHRR was used as an auxiliary
source of information. Later on, the global S10-composites of SPOT-VGT were also included in the system, as JRCMARS had to cope more and more with global issues concerning food security, humanitarian aid and the volatile supply
and demand of competing producers and markets [13].
Over the years, JRC-MARS systematically collected an archive of 1 km-resolution NOAA-AVHRR scenes over Europe.
It comprises all (or most) raw scenes registered since 1981 by different ground stations: Dundee (1981-1988), DLROberpfaffenhofen (1989-1995), JRC-Ispra (1996-2000), FU-Berlin (2001-2008) and EUMETSAT-EARS (from 2009
onwards). On behalf of the MARS-programme, VITO developed a processing chain for the automated treatment of the
AVHRR data. Individual tracks/orbits are ingested, corrected in different steps (calibration, geometric correction,
atmospheric correction, cloud/snow labelling) and then composited to ten-daily "Synthesis" images (S10), similar to the
ones of SPOT-VGT but restricted to the pan-European continent. Recently, the full European archive has been
reprocessed in a standardized way with this software [14].
On 19 October 2006, the European Space Agency (ESA) and the European Organisation for the Exploitation of
Meteorological Satellites (EUMETSAT) launched the METOP-satellite, which can be considered as the European
counterpart of the NOAA-platforms [15]. METOP carries a wide range of different sensors, amongst which also the
same AVHRR-instrument as installed on NOAA. In a joint European-US initiative, METOP will assume the "morning
orbit" (overpasses at around 08h|20h LST), while NOAA remains responsible for the "noon orbit" (14h|02h). Contrary to
NOAA, METOP provides advanced on-board data storage capacities, so all the registered 1 km imagery of the last orbit
is systematically transmitted (via an antenna in Svalbard) to EUMETSAT. After some first but crucial adaptations,
EUMETSAT distributes the "raw" data freely and in near-real time via its EUMETCAST broadcasting service [16]. The
entire system became operational from 15 May 2007 onwards.
On demand of JRC-MARS, VITO adapted its AVHRR processing chain so it could also treat the new METOP data.
Although the action initially focused on the European continent, the METOP imagery soon revealed its excellent
potentials for global land monitoring. Therefore VITO decided to start the routine processing of the METOP-AVHRR
data on a global scale. The resulting ten-daily composites can be considered as a complement to the global VGT-S10,
derived from SPOT-VEGETATION. This paper briefly presents the methodology and data distribution plans.
2. METHODOLOGY FOR THE PRODUCTION OF GLOBAL S10-COMPOSITES
The AVHRR-3 instrument on board of METOP and the recent NOAA's (since NOAA-15) registers in five spectral bands:
bands 1/2 in the shortwave range (RED and NIR=near infrared), bands 4/5 in the thermal infrared (TIR, 10-12µm) and
band 3 which can be switched between the shortwave infrared (B3A=SWIR, around 1.6 µm, mostly during daytime) and
the middle infrared (B3B=MIR, around 3.7 µm, during the night).
At EUMETSAT, the incoming stream of 1 km METOP-AVHRR imagery is submitted to a number of pre-processing
operations [17]. Compared to the situation of NOAA-AVHRR, the resulting "Level1b" files have four crucial merits:
Thanks to METOP’s advanced on-board storage facilities, 1 km resolution data are available for the entire earth
surface.
 The Level1b files always provide the longitude and latitude on the WGS84 geodetical datum for a subsample of pixels.
These "Lon/Lat-planes" are needed for the mapping of the raw images towards a geographical projection system.

In the Level1b files of NOAA-AVHRR, the planes often lack precision due to errors of the on-board clock, unknown
platform attitude (yaw, pitch, roll) and other reasons. To reduce mapping errors, the planes thus must be adjusted by
means of complicated and computer-intensive chip matching procedures, without guaranteed success [13]. However,
for METOP these problems no longer exist because the Lon/Lat-values in the Level1b appear to be highly accurate.
 In the Level1b-files, distributed by EUMETSAT, all the measurements are already converted into well-calibrated
radiances. For the MIR/TIR bands the conversion is based on the classical inverse Planck approach and the
measurements of the on-board calibration standards [18]. The calibration of the shortwave bands has always been
problematic due to the absence of an on-board reference lamp, but EUMETSAT uses the same "vicarious" calibration
methods for METOP as NASA does for NOAA. This approach guarantees an optimal agreement between the
measurements of the AVHRR's on both platforms [19].
 EUMETSAT directly applies an advanced cloud/snow detection algorithm on the calibrated data [20]. The Level1bfiles of METOP-AVHRR thus already contain an indication of each pixel’s observation state
(clear cloud snow/ice).
EUMETSAT cuts the pre-processed Level1b data stream into segments of three minutes (1080 scan lines, or 180 seconds
x 6 lines per second) and broadcasts them via the EUMETCAST system in the so-called EPS-format [17]. VITO
systematically ingests all the global information via its two EUMETCAST satellite receiving stations (primary &
backup). Each individual segment is further treated as follows:






Segment selection: The never-ending data stream comprises all kinds of METOP-AVHRR imagery, registered over
land and sea, during the day and the night. But as the focus is on global vegetation monitoring, only the daytime
segments with at least some land pixels are retained for further processing. In line with the VGT-S10, no further
attempts are made to process the sea pixels. And the elimination of the night time registrations implies that band 3
always corresponds with the SWIR (B3A).
Spatial - Remap: Using the mentioned Lon/Lat-planes, included by EUMETSAT in the Level1b, and a "nearest
neighbour" resampling scheme, the five spectral bands of AVHRR are converted to the WGS84 Geographical Lon/Lat
system with the same framing and resolution as used for SPOT-VGT (pixel size of 1°/112 1 km along a great
circle). All further images follow this system.
Spatial - Angles: Similar images are computed providing for each pixel the angular position (zenith/azimuth) of the
sun and the sensor at the moment of the registration.
Spectral - Shortwave: The on-board registered radiances are converted into surface reflectance factors by means of the
SMAC algorithm [21] which removes (at least partially) the unwanted atmospheric and angular impacts on the ground
signal. The SMAC-coefficients for METOP's three shortwave channels (RED, NIR, SWIR) were computed on behalf
of the MARS-project [22]. They are also made available via CESBIO's SMAC-website [23]. In addition to these
band-specific coefficients, SMAC also needs the input of the atmospherical state at registration time, in terms of water
vapour, aerosol load and ozone content. For the assessment of these three atmospheric state variables, the same
approach is followed as for SPOT-VGT: for aerosols and ozone only climatologic values are used (long term monthly
averages), but for water vapour more precise data are acquired from ECMWF, i.e. global water vapour maps with a
spatial resolution of 0.5° and a renewal frequency of six hours (four water vapour images per day). After the
atmospheric correction, the Normalised Difference Vegetation Index (NDVI) is computed from the surface
reflectances: NDVI = (NIR-RED)/(NIR+RED).
Spectral - Longwave: Land surface temperatures (LST) are derived, separately for land and sea pixels, from the two
TIR brightness temperatures using the split window technique proposed by Coll & Caselles [24], which also requires
the input of water vapour and TIR emissivities. For water vapour, the same six-hourly ECMWF data are used as for
SMAC. The TIR emissivities of the land pixels are assessed via a simple linear equation from their NDVI.
Quality - Masking: Each pixel's observational state is expressed via subsequent 0/1-switches in a bitmap image. This
"status mask" classifies each pixel according to criteria such as: land sea, and clear cloud snow/ice. While the
GLC2000 map [10] is used to separate land from sea pixels, the distinction between “clear cloud snow/ice” is
fully based on the results of the cloud/snow detection added by EUMETSAT in the Level1b-files [20].
The final ten-daily composites (S10) are then created in a similar way as for SPOT-VGT:

Spatial aspects: The composite images follow the same map system as the corrected segments, i.e. WGS84
Geographical Lon/Lat with a resolution of 1°/112. But while the segments only cover limited zones, the
S10-composites always extend over the same near-global area, ranging from -180° to +180° in longitude and from
-56° to +75° in latitude (40 320 columns x 14 673 lines). As for SPOT-VGT, the composites only contain information
for the land pixels. All the water pixels are flagged with unique missing values codes.
 Temporal aspects: Every month is divided in three "dekads". The first two always comprise ten days (1-10, 11-20),
the third one has variable length as it runs from day 21 until the end of the month. The procedure starts with the
selection of all segments registered within the concerned dekad and overlapping at least partially with the mentioned
target zone.
 Spectral aspects: In general, for each land pixel in the composite, different observations are available, from different
segments or registration dates. The compositing selects the "best available" observation and transfers all its
components (reflectances, temperatures, angles, status, ...) to the corresponding layers in the S10 synthesis. The
selection is realised per pixel as follows. First, all the available observations are classified in a number of groups
based on their status (clear, snow/ice, cloud) and geometry (sun/view zenith angles). These groups are ranked in a
logical way (clear with good geometry > clear with acceptable geometry > ... > cloudy with bad geometry). After the
classification, the best available group is found. If it contains only one observation, this will be the "best" one. If there
are more, the one with the highest NDVI is selected. This method favours the near-nadir views and suppresses the
observations which are still partly affected by clouds, snow and water (which all have low NDVI).
 Quality Control: Scenes of NOAA-AVHRR are often affected by radiometric errors (stripes, waves) or geometrical
shifts, especially when a platform reaches its nominal lifetime. Hence, after the pre-processing each individual
segment is visually checked by an operator who identifies and rejects the bad scenes. Without this measure, the bad
segments can spoil the quality of the final composites. But after one year of similar checks on the AVHRR data of
METOP, no such errors could be detected. So as an alternative, we now only check the daily global composites,
which are produced as well in the background. This requires less time and is as effective as checking hundreds of
individual segments (480 per day).
The compositing is a crucial step. Whereas the individual segments contain a lot of clouds and occupy different and
scattered areas, the final ten-daily composites are better "filled", less contaminated by clouds and they always cover the
full area of the “target zone”, in this case the major part of the globe. The global S10-composites derived from METOPAVHRR have the same spatial characteristics as the S10 of SPOT-VGT both can even directly be superimposed. As
mentioned, all sea pixels are flagged and information is only provided for the land pixels. Each S10 comprises twelve
image layers, which are stored in separate binary images with data type “unsigned 8-bit” and with the annotation in
accessory text-files:

Surface reflectances in RED, NIR and SWIR.
 Derived scenes with surface NDVI and (land) surface temperature .
 The four angles of the viewing geometry (zenith/azimuth angles of sun and sensor, as seen from the pixel centres).

Status BitMask: land sea, clear cloud snow/ice, etc.
Time grid: Registration day of the selected observation.
 The number of "clear" (cloud/snowfree) observations available for the compositing.

3. CONCLUSIONS & PROSPECTS
The systematic ingestion and processing of METOP-AVHRR started at VITO in January 2009, but all the Level1b data
registered in the course of 2008 could be acquired from EUMETSATS's UMARF archiving service [15]. As a result, the
available time series starts in January 2008 and new ten-daily composites are systematically added in near-real time (at
last three days after the end of each dekad).
In parallel, a data distribution service has been established (http://www.metopS10.vito.be), which allows users to freely
download specific products, in the same style as the free VGT-portal (http://free.vgt.vito.be). The service is expected to
become operational in April 2010.
In terms of geometry and contents, the ten-daily global composites of METOP-AVHRR are very comparable to the ones
of SPOT-VEGETATION. Although in-depth validations still must be carried out, the geometric and radiometric quality
of the data seems excellent.
Today's global land monitoring activities strongly rely on sensors such as SPOT-VEGETATION, TERRA/AQUAMODIS and ENVISAT-MERIS/AATSR, which all reach the end of their nominal lifetimes. If one of these systems fails,
METOP-AVHRR can serve as an immediately available alternative. Even today, METOP-AVHRR can play an
important complementary role thanks to the early-morning cross-over and presence of thermal bands.
Nowadays, the European METOP-S10 composites are operationally used in the crop monitoring system of JRC-MARS.
In the future, new developments might be considered such as the inclusion of more advanced biophysical products, the
regular delivery of similar composites for the polar regions, etc.
ACKNOWLEDGEMENTS
In addition to our own institutes (VITO-TAP, JRC-IPSC), we want to thank EUMETSAT for the provision of the basic
METOP-data in real-time, for the delivery of the complete global imagery of 2008 by the UMARF-archive, and also for
the authorisation to distribute the derived S10-products presented in this text. The Belgian Science Policy Office
(BelSPO) is rendered thanks for the sponsoring of our activities - processing and publishing. We also greet our
colleagues from Alterra-NL, MeteoConsult-NL and MC-Wetter-GE, with whom we collaborate for JRC-MARS.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Cracknell A., “The Advanced Very High Resolution Radiometer”, Taylor & Francis, ISBN 0-7484-0209-8
(1997).
Kidwell K. (Editor), “NOAA polar orbiter data user’s guide”, US Department of Commerce, National Oceanic
and Atmospheric Administration, National Environmental Satellite, Data and Information Service, Washington,
DC 20233 (1997). See: http://www2.ncdc.noaa.gov/docs/podug/.
El Saleous N., Vermote E., Justice C., Townshend J., Tucker C. and Goward S., “Improvements in the global
biospheric record from the Advanced Very High Resolution Radiometer (AVHRR)”, Int. J. Remote Sensing, vol.
21, no. 6&7, 1251-1277 (2000).
Tucker C., Pinzon J., Brown M., Slayback D., Pak E., Mahoney R., Vermote E. and El Saleous N., “An extended
AVHRR 8-km NDVI dataset compatible with MODIS and SPOT vegetation NDVI data”, Int. J. Remote
Sensing, vol. 26, no. 20, 4485-4498 (2005).
Eidenshink J. and Faundeen J., “The 1 km AVHRR global land data set: first stages in implementation”,
International Journal of Remote Sensing, vol.15, no. 17, 3443-3462 (1994).
Loveland T. and Belward A., “The IGBP-DIS global 1km land cover data set, DISCover: first results”,
International Journal of Remote Sensing, vol. 18, no. 15, 3289-3295 (1997).
Myneni R., Keeling C., Tucker C., Asrar G. and Nemani R., “Increased plant growth in the northern high
latitudes from 1981 to 1991”, Nature, vol. 386, no. 6626, 698-702 (1997).
Passot X., “VEGETATION image processing methods in the CTIV”, Proceedings of the first VEGETATION
User's Conference, ed. G. Saint, 3-6 April 2000, 15-31 (2000).
SPOT VEGETATION website: http://www.vgt.vito.be.
Bartholomé E. and Belward A., “GLC2000: a new approach to global land cover mapping from Earth
observation data”, International Journal of Remote Sensing, vol. 26, no. 9, 1959-1977 (2005).
Tansey K., Grégoire J-M., Stroppiana D., Sousa A., Silva J., Pereira J., Boschetti L., Maggi M., Brivio P., Fraser
R., Flasse S., Ershov D., Binaghi E., Graetz D. and Peduzzi P., “Vegetation burning in the year 2000: Global
burned area estimates from SPOT VEGETATION data”, Journal of Geophysical Research - Atmospheres, 109,
D14S03 (2004).
van Diepen K., Boogaard H., Supit I., Lazar C., Orlandi S., Van der Goot E. and Schapendonk A., “Methodology
of the MARS crop yield forecasting system. Vol. 2: Agrometeorological modelling, processing and analysis”.
Eds. Lazar C. and Genovese G., EUR 21291 EN/2, 98 p. (2004).
Eerens H., Piccard I., Royer A. and Orlandi S., 2004, “Methodology of the MARS crop yield forecasting system.
Vol. 3: Remote sensing information, data processing and analysis”, Eds. Royer A. and Genovese G., EUR 21291
EN/3, 76 p. (2004).
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
Swinnen E., Eerens H., Heyns W., Piccard I., Viaene P. and Claes P., “An integrated long time series of 1 km
resolution NDVI for Europe from the NOAA-AVHRR and SPOT-VEGETATION sensors”, In: “Proceedings of
MultiTemp 2007: 4th International workshop on the analysis of multi-temporal remote sensing images”, 18-20
July, 2007, Leuven, Belgium. Eds. De Lannoy G. et al., Leuven, Gent: Katholieke Universiteit Leuven,
Universiteit Gent, ISBN 1-4244-0846-6 (2007).
EUMETSAT: http://www.eumetsat.int/HOME/index.htm.
EUMETCAST: http://www.eumetsat.int/HOME/Main/What_We_Do/EUMETCast/index.htm.
EUMETSAT, “AVHRR Level1b Products Guide”, EUMETSAT, Darmstadt, Germany, Document EUM/OPSEPS/MAN/04/0029, 123 p. (2008).
See: http://oiswww.eumetsat.org/WEBOPS/eps-pg/AVHRR/AVHRR-PG-0TOC.htm.
Planet W. (Editor), “Data extraction and calibration of TIROS-N/NOAA radiometers”, NOAA Technical
Memorandum NESS 107, Revision 1, Oct. 1988, 130 p. (1988).
Rao C. and Chen J., “Revised post-launch calibration of the visible and near-infrared channels of the Advanced
Very High Resolution Radiometer (AVHRR) on the NOAA-14 spacecraft”, Int. J. Remote Sensing, vol. 20, no.
18, 3485-3491 (1999).
EUMETSAT, “EPS ground segment – AVHRR L1 product generation specification”, EUMETSAT, Darmstadt,
Germany, Document EUM.EPS.SYS.SPE.990004, 158 p. (2004).
Rahman H. and Dedieu G., “SMAC: a Simplified Method for the Atmospheric Correction of Satellite
Measurements in the Solar Spectrum”, International Journal of Remote Sensing, 15(1), 123-143 (1994).
Berthelot B., “SMAC coefficients for METOP AVHRR/3”, VEGA Technologies SAS, Toulouse, Internal report
SMAC01-TN-AVHRR3-VEGA, 63 p. (2008).
SMAC coefficients for METOP-AVHRR: http://www.cesbio.ups-tlse.fr/fr/smac.htm.
Coll C. and Caselles V., “A split-window algorithm for land surface temperature from Advanced Very High
Resolution Radiometer data: validation and algorithm comparison”, Journal of Geophysics Research, 102(D14),
16697-16713 (1997).