Geoinformation, monitoring and assessment Environment

Posted June 2000

Incorporating the use of VEGETATION data in FAO programmes

Paper presented at the symposium VEGETATION 2000, Belgirate, Italy, 3 to 6 April 2000

by
F.L. Snijders
Remote Sensing Officer
Environment and Natural Resource Service
Research, Extension and Training Division
FAO Sustainable Development Department


Summary

Data from low-resolution earth observation satellites have been used operationally by FAO in the fields of early warning for food security and desert locust plague prevention for many years. Originally, this focused on data at 4-10 km resolution, which proved very suitable for the monitoring of crop growing conditions over large areas, in particular when a long-year archive of historic data was available. In the field of desert locust control, this was less successful and efforts were undertaken to obtain data at 1 km from local HRPT stations. However, operational use of data from a variety of sources proved very difficult and the quality and timeliness was highly variable among sources. The launch of the VEGETATION instrument on board SPOT offered an excellent opportunity to change this situation, as it offered global coverage processed at a single centre.

In order to explore the possible uses of VEGETATION data, an agreement was established between FAO and the EU Joint Research Centre. Within the framework of this agreement, a number of activities were undertaken. First, some tools had to be developed to facilitate access to the data, which is in HDF format. This resulted, first, in a simple information extractor and a conversion utility that has been put in the public domain and allows the use of the data on simple PCs. Second, the use of 10-daily composits has been incorporated in the operations of the FAO Global Information and Early Warning System (GIEWS) and the locust control group, through the ARTEMIS system. Third, new applications of the data are being explored.

The experiences obtained so far clearly indicate that VEGETATION is a unique and very valuable instrument that has a very wide field of applications, although some improvements are still needed to both the processing and the delivery systems. While in the field of locust control the benefits are more direct, in the field of early warning for food security the importance will grow more steadily as this application relies heavily on the availability of historic data. Other applications, such as in the field of irrigation monitoring, are still under investigation.

Introduction

Data from low-resolution earth observation satellites have been used operationally by the Food and Agriculture Organization (FAO) of the United Nations for many years for the monitoring and assessment of environmental conditions related to food security. Programmes using this data include the Global Information and Early Warning System (GIEWS), the Emergency Centre for Locust Applications (ECLO), the Emergency Prevention System against transboundary animal and plant pests and diseases (EMPRES) and the Programme Against African Trypanosomiasis (PAAT). Furthermore, assistance in the use of such data was, and continues to be, provided to many national and regional activities, including early warning food information systems in the SADC and IGAD regions, desert locust control organizations, the Nile Forecast Centre and others.

In particular the GIEWS, which monitors global food demand and supply, and the locust control group, which monitors desert locust and other migratory pests and coordinates national, regional and international action, have a need for timely and reliable information on agro-meteorological and crop/vegetation conditions covering large areas. This information can often not be obtained from ground reporting alone. In order to assist these programmes, FAO established in 1988 the Africa Real Time Environmental Monitoring System (ARTEMIS), with the objective of providing a routine flow of satellite imagery in near real-time indicating the status of the growing season and vegetation development over Africa. Through cooperation with the GIMMS group of NASA Goddard Space Flight Centre and the TAMSAT group of the University of Reading, an operational processing of METEOSAT TIR and NOAA-AVHRR GAC data was implemented. METEOSAT data was used to monitor rainfall and NDVI imagery derived from the AVHRR instrument to monitor the vegetation development. All products were made available at a common resolution of 7.6 km.

Over the years, as the use of satellite-derived information became more and more integrated in the operations of the above programmes, there was a growing demand for more and better data, which could only be partially met. The area coverage was extended first to South and Central America with NOAA GAC-derived NDVI and, later, through cooperation with the Japanese Meteorological Agency, monsoon monitoring over Asia was also explored, based on GMS data. Although this was certainly an improvement, many areas of special interest to GIEWS, such as the Commonwealth of Independent States (CIS) and North Korea, were still not covered.

For locust control, there was a strong demand to increase the spatial and spectral resolution of the coverage as the available data did not allow for the reliable identification of vegetated areas in the desert suitable for locust breeding and development. However, this was not easily implemented. Meteorological satellites do not (yet) produce at higher resolution, and NOAA data at LAC (1 km) resolution cannot, routinely, be obtained centrally. Data from local HRPT stations were obtained, but only with the Centre Agrhymet in Niamey, Niger, did this result in an operational and timely data flow. This certainly improved the situation, but it covered only a part of the locust recession area, which stretches from Senegal/Mauritania up to India. Also, when data from other HRPT stations were obtained, the combined use of the various sources was complicated by the heterogeneity in processing, image quality and timeliness of the data.

Exploring the use of VEGETATION

The launch of the VEGETATION instrument on board the SPOT satellite offered an excellent opportunity to solve a number of the problems mentioned in the previous section and to improve the services FAO provides in early warning for food security as:

Tool development

The first hurdle faced when trying to use data from VEGETATION is its data format: a "science data set" (SDS) of the Hierarchical Data Format (HDF). This is a very powerful and well-structured format, but it is not directly compatible with the main format used within FAO for low-resolution satellite images. This format was originally defined by the "Image Display and Analysis" (IDA) software that was developed as an international effort by USAID-FEWS, USGS and FAO-ARTEMIS.

IDA has been instrumental in expanding the use of satellite data to a wide range of users, in particular in developing countries, as it runs on very simple DOS-based microcomputers instead of highly priced digital image processing systems. IDA itself has now been superseded by WinDisp, which was originally developed as part of a dedicated Workstation for GIEWS with funding from the European Commission. Though originally developed as an IDA image viewer, many of the analytical capabilities of IDA were added to WinDisp, making it the Windows-based successor. Several versions of WinDisp have since been released with contributions coming from many sources in addition to the GIEWS, including FAO-ARTEMIS, USAID FEWS, SADC RRSP, USFS and the USGS.

The IDA image format remains a de facto standard within the world of early warning and is supported by the whole suite of software developed either by, or in collaboration with, FAO for the analysis of low-resolution satellite data. The first task towards operational use of VEGETATION data, therefore, was to develop appropriate conversion and intake routines. This resulted in two programmes: VegInfo and AICON.

VegInfo can be seen as a spin-off of the AICON development and is a small, PC-based programme that runs in a DOS window and extracts all the basic information from the HDF file, such as the length of the file header and the height and width of the image. While this is not required for the more popular high-end image processing systems that have HDF support built in, this information is important for those who work in low-cost application environments found in many developing countries. Figure 1 gives an example of the information extracted by VegInfo.


Figure 1

The second tool developed is AICON, the Artemis Image CONverter. AICON is a general, PC-based tool that can convert geographic projection and data coding of 8- and 16-bit images. It was already used routinely within ARTEMIS for the conversion of 16-bit imagery obtained from NASA/GSFC to IDA. The programme has now been expanded to support the HDF format as used by VEGETATION. For 8-bit images, such as the NDVI, it simply extracts the image data and adds the appropriate header, while for the 16-bit data, as used for single band imagery, it also provides a recoding to 8-bit. Although AICON is fully interactive, it can be highly automated. Within ARTEMIS it is used operationally for the intake of VEGETATION images in batch-mode. Figure 2 shows the dialogue window to define an image format in AICON.


Figure 2


Both VegInfo and AICON were developed by Silvio Griguolo of the University of Venice and can be obtained through the software section of the FAO ARTEMIS/AgroMet website. This site also contains pointers to the other members of the software suite mentioned, including WinDisp, which is now multilingual, with all menus, on-line help and reference documents available in English, French and Spanish.

Integration of VEGETATION into routine operations

The exploration of routine use of VEGETATION data focused on dekadal NDVI imagery, as the analysts of the GIEWS and the locust group were already very familiar with the characteristics of this index, based on the use of NOAA-AVHRR data.

Dekadal NDVI composites are converted by ARTEMIS to IDA format and cut into a number of continental or subcontinental windows in such a way that almost all regional groupings, such as IGAD, CILSS or ASEAN, fall fully within one window. Although this introduced considerable overlap between the windows, it did facilitate the use of the data. Each window is between 30 and 50 MB and stored on the ARTEMIS server, which is available internally to all FAO staff through a local area network. Limiting the size to 50 MB ensures reasonable response time of both the network and the client software.

The GIEWS workstation, which facilitates access to the information available to country analysts, such as country reports, news agency reports, cereal balance sheets, reference information and satellite data, was modified to incorporate the VEGETATION imagery. See Figure 3.


Figure 3


With the above two actions, all relevant analysts have easy access to the dekadal NDVI imagery and can explore their usefulness. The main problem associated with the integration of the data concerns its size. Moving from GAC-type resolutions to LAC resolutions results in a dramatic increase in storage requirements. Also, with the arrival of VEGETATION data, not only did the resolution increase, but also the area coverage and the number of available bands. For routine use, the expansion was around 100 times, excluding the storage need for application development. While ARTEMIS can still accommodate the data, new storage solutions will have to be found in the near future, when the historic archive, which should be kept on line for most applications, starts to grow.

Application development

A variety of activities were undertaken to explore possible new applications using the VEGETATION data or to see where it could strengthen existing applications. These included:

In due course, other applications will also be evaluated, including some in the field of animal health and tsetse control. Furthermore, as more and more science groups produce results, FAO will watch closely for those applications that fall within its mandate and which, in close collaboration with the researchers, can be applied operationally.

First results and conclusions

As VEGETATION data has only recently become available on line, only tentative results are available at this time. In the field of locust control the advantages of the new data are already clear and the use of NDVI images has demonstrated an improved identification and monitoring of potential locust breeding grounds. The usefulness of, and interpretation techniques to be applied to, SWIR data is not yet fully known and will need some more time to be determined. Field checks of NDVI images have generally been positive, in particular in Mauritania, although in other areas some "false positives" have been found which have not yet been explained. More details of this work will be presented by Michiel Cherlet.

Only partial results are currently available for the use of VEGETATION data in the field of early warning for food security. Much of the analysis applied routinely by the country officers of GIEWS is based on the availability of historic data and this is not (yet) available. This makes it difficult, for instance, to assess if indeed a season is earlier or later than usual. Here NOAA GAC will still have a use for a number of years to come. But VEGETATION does allow the ability to "zoom in" on identified problem areas and it is the only routine source of satellite information currently available to GIEWS for important regions, such as North Korea and the CIS. Even with the very short time-series available, the data have already proved a number of times to be very useful in depicting adverse crop growing conditions.

In general, it can be concluded that VEGETATION is a unique and very valuable instrument that will allow many programmes of FAO to move from 4-10 km resolution to 1 km resolution. Its importance will grow steadily as the historic archive increases in size. Not only will the existing application be improved, but many new applications will also develop. In this respect VEGETATION should be seen as the first of a whole new range of earth observation systems that will shortly become available. Within the fields traditionally covered by the ARTEMIS system of FAO, it will team up very nicely with the new METEOSAT series of satellites (METEOSAT Second Generation) to be launched in the near future and certainly contribute to increased food security in developing countries.

However, despite the many good characteristics of the VEGETATION data, there is, as always, room for improvement.

The first area for improvement concerns the timeliness of image delivery. For many of FAO's programmes, this is an essential characteristic of any routine data flow, but, at present, this is hampering full operational use of VEGETATION data. As an example, in November 1999, reports were received of vegetation development in a number of potential locust breeding areas in Mauritania. Ground teams were ready to verify this information and had hoped to base their findings on VEGETATION. Unfortunately, the data arrived too late, and an opportunity was missed.

The second field where improvements would increase the value of VEGETATION concerns the quality of cloud masking and the validity of the Status Maps. For a proper assessment of the confidence that can be assigned to changes in NDVI the availability of reliable cloud masking information is essential.

The third field relates to the image speckle that results from the compositing procedure. Dekadal composits of single-band imagery can be rather noisy, which forces the analysts to use monthly images instead of dekadal images, thereby reducing the temporal resolution.

Any improvements that can be made in the above three fields will further increase the value of this very successful mission. From the side of FAO, it is sincerely hoped that VEGETATION will be able to realize a long-term data flow, which will benefit many users, both in developed and in developing countries.

Additional information about the VEGETATION initiative can be obtained from the VEGETATION website or e-mail: fred.snijders@fao.org



Back to Top FAO Homepage