| GISdevelopment.net ---> AARS ---> ACRS 1989 ---> Agriculture & Soil |
The use of NOAA AVHRR data as
a tool for operational agroclimatic assessment in Asia*
Gary E. Johnson and V. Rao Achutuni
NOAA/NESDIS Climate Applications Branch and University of Missouri,
Co-operative Institute for Applied Meteorology,
Columbia, Missouri, USA
Jan B. Orsini
U.N. Economic and Social Commission for Asia and the Pacific,
Agriculture and Rural Development Division Bangkok, Thailand
Gary E. Johnson and V. Rao Achutuni
NOAA/NESDIS Climate Applications Branch and University of Missouri,
Co-operative Institute for Applied Meteorology,
Columbia, Missouri, USA
Jan B. Orsini
U.N. Economic and Social Commission for Asia and the Pacific,
Agriculture and Rural Development Division Bangkok, Thailand
Abstract
National oceanic and atmospheric Administration (NOAA) and University of Missouri Scientists began operational drought early warning activities in 1979. Data from the Advanced very High Resolution Radiometer (AVHRR) sensor abroad the NOAA polar-orbiting satellites were introduced into the process in 1982. The integration of AVHRR data in the agrocilmatic assessment process in Asia began with a training course for professionals from Malaysia and Thailand in 1985. Current activities in Asia focus in enhancing existing agroclimatic assessment programs with composited imagery and NDVI products derived from NOAA satellite data.
Introduction
The National Oceanic and Atmospheric Administration (NOAA), with the logistical support of the United Nations / Economic and Social Commission for Asia and the Pacific (UN/ESCAP), is conducting Phase-II of a drought early warning program for South and South East Asia. This Agro-Climatic Assessment Project, begun with phase-I in 984 has the long-term objective of establishing a stand - alone crop monitoring tool to be used on a real time basis by food security decision makers in each country.
Technology transfer during Phase-I focused on the utilization of cumulative rainfall indices and statistical climate crop yield models to (1) monitor drought, (2) Assess the impact on crop production, and (3) produce and disseminated in operational crop assessment early warning bulletin within each participating country. The agencies in the program generally included the National Meteorological Department, the Ministry of Agriculture, and the agency responsible for national food security. Bangladesh, India, Indonesia Malaysia, Nepal Pakistan, the Philippines, Sri Lank, and Thailand participated. Many drought early warning units within their governmental framework which still continue to produce and disseminate crop assessment bulletins. Users report that this rainfall analysis technology proved especially effective during the severe drought of 1987.
The Phase- II project utilizes NOAA polar orbiting satellite imagery and vegetative index analysis to enhance the Phase-I analysis. Although relatively coarse in resolution.
The data enable a continuous real time analysis of the technology of vegetation provided that clouds can be removed by selection of the cloud free areas of a sequence of superimposed images. This low cost technology, developed and refined by the climate Applications Branch (CB) of BOAA and the co-operative institute for Applied Meteorological (CIAM) for use on personal computers, was tested extensively in the Sahelian countries of Africa. The data can be received twice daily by any country with the appropriate antenna and data reception and analysis system.
Two Phase II activities are completed. A national workshop mission was conducted in each country at the outset to : introduce the new technology; liaise directly with each participating agency. Help clarify the inter agency relationships; facilitate the selection of trainees; and assess computer software and hardware needs. A regional training workshop for sixteen trainees from four countries (Indonesia, Malaysia, the Philippines, and Thailand) was held at the Asian Institute of Technology (AIT) in Bangkok for four weeks in June and July. A follow ups in country technical training mission is currently being conducted to assure that the technology is in place and operational. A regional evaluation seminar will follow at the conclusion of the project to: assess the usefulness of the technology; identify countries wishing to join the Phase- III replication program; and assess possible sources and future funding support.
NOAA satellite and sensor characteristics
The remotely sensed data used in the operational agro-climatic assessment process are acquired from the NOAA polar orbiting satellites. Selected characteristics of the NOAA 11 satellite and the Advanced very high resolution radiometer (AVHRR) sensor appear in table 1.
and AVHRR sensor system
| Characteristic | NOAA/AVHRR |
| Inclination of orbit | 98.922° |
| Height above surface | 851 km |
| Number of orbits/day | 14.1 |
| Times of coverage at equator | 01:41 descending 13:41 ascending |
| Orbital period | 102 min. |
| Latitudinal coverage | 90°N - 90°s |
| Cycle duration | 1 day |
| Ground coverage | 2842 km |
| Field of view (FOV) | ± 55.4° |
| Instantaneous field of view (IFOV) | 1.39 - 1.51 mrad |
| Ground resolution (nadir) | 1.1 km |
| Number of channels | One visible One near infrared One middle infrared Two thermal infrared |
Date are routinely acquired from the satellite and sensor system at two spatial resolutions. The finest resolution data, known as High Resolution Picture Transmission (HRPT) data, are transmitted to ground receiving stations in real-time. Selected data sets of the same spatial resolution area recorded aboard the satellite for alter playback to ground receiving stations.
Recorder limitations dictate that a maximum of ten minutes of this data may be recoded on any single orbit. These data are subsequently processed as Local Area Coverage (LAC) data. A reduced spatial resolution data set, known as Global area coverage GAC data are created by the processor aboard the satellite by sampling the full resolution data and recording it at GAC resolution. Full resolution data are sampled by averaging four of every five adjacent pixels along a scan line and processing every third line. LAC and HRPT data are normally 1.1 kilometer spatial resolution data at the satellite subpoint; GAC data are generally resolution data (Kidwell, 1988). A third spatial resolution AVHRR product, known as global vegetation Index (GVI) data, by selecting the last value in an array of four by four GAC pixels and assigning its radio metric value to the resulting GVI pixel.
Satellite data products
The spatial resolution of data for use in operational agro climatic assessments must be appropriate in terms of data availability, cost, project objectives, available hardware and software, and the timeliness expected of an operational system. All three types of NOAA satellite data GVI, GAC and LAC, were considered for assessment products. GVI resolution data represents the mot economical data source with least quantity of data to handle. For this reason, a three year data set of GVI data covering the Asian continent was created for use in the satellite crop monitoring workshop component for this project. Because of the agricultural conditions in South East Asia, however, mot countries wish to work with the highest spatial resolution data available from the NOAA satellite. Consequently, software for analysis the 1.1 km resolution HRPt.LAC data has also been made available.
Remotely sensed data products take two distinct forms :
- composited imagery which is interpreted for vegetative condition; and
- the Normalized difference vegetation index (NDVI) which is analyzed for vegetative trend through time an for change from a reference period.
Image compositing software written by NOAA and CIAM personnel extracts desired :windows: from the image data, maps these data to plate carree projection, and composites the data over any selected time period. In some instances, compositing periods are chosen to match available ancillary data. In working with African data for example, ten day (decadal) data were composite to match the time frame of available precipitation data. In the case of the GVI data used for the satellite crop monitoring workshop, however, only preprocessed weekly data were available for analysis. Image data are composited over time to minimize the clouds present in the image and to reduce the atmospheric and scan-angle effects which act to reduce the atmospheric and can angle effects which act to reduce the vegetation index. It has been demonstrated that when several images of an area are composited to obtain a clear view, the NDVI values from days with haze, sub-pixel clouds, and high scan angles are eliminated (Tarpley, 19840. The compositing algorithm was designed to retain the greatest NDVI value for each pixel over the compositing period. Therefore, areas which appear cloudy on composite images have clouds presents for all days at the satellite overpass time.
Derived composite images are transferred to floppy diskettes for analysis on a personal computer. The addition of a DRAGON, a commercial software package, enhanced the software written by NOAA and CIAM. The personal computer system permits an analyst to display, enlarge and "roam" around the image. Color components of the image may be altered, images may be annotated with boundaries or text information, and the image may be saved to a floppy or hard disk for subsequent comparison.
Images are displayed according to a color coordinate system (Ambroziak 1984). Colors display a continuum of hue, intensity, and saturation designed to match both the data and mind's perception system of color. Different hues separate vegetation and water from clouds and bare soil. Intensity distinguishes variations within a hue; for example,. Light green may represent crops and dark green deciduous forest. A thermal channel of AVHRR data generally cheannel4, is used to equate cloud temperature to saturation by reducing saturation so that colder clouds are represented as white.
Using the personal computer system, analysts can qualitatively photo interpret the images for changing vegetative conditions and trends. Employing a split screen option, analysts can also compare images from multiple dates side-by-side.
The Normalized difference vegetation index (NDVI)
The normalized difference vegetation index (NDVI), a ratio of near infrared (NIR) and visible (VIS) radiance, has proved to be indicative of the quantity and quality of photo synthetically active biomass. The general formula for NDVI is (NIR - VIS) / (NIR + VIS). In the case of the AVHRR sensor, the NDVI is expressed as :
Where:
CH1 = Channel 1 VIS reflectance
CH2 = Channel 2 NIR reflectance
The NDVI has found numerous applications in monitoring worldwide crop conditions and in rangeland assessment (Perry and Lautenschlager, 1984; justice and Hiernaux, 1986; Malingrea, 1986; Tucker and Selelrs, 1986; Johnson et al., 1987 van Dijk et al. 1987).
NDVI time series analysis
Time series of NDVI permit the monitoring of the dynamic nature of vegetation phonology. The assessor can make qualitative judgements about the current growing season by comparing it with some benchmark year or years.
The NOAA/CIAM software uses the GVI resolution data for computing the NDVI time series. The raw NDVI data are smoothed using curve fitting techniques discussed by van Dijk et. al (1987). The use can overlay and display upto three years of DVI time series simultaneously. The software also permits the user the compute the total area under the curve between any two time periods. Figure 1 shows an example for 0.50 lat. By 1.00 long. Polygon in Northeast Thailand.
One of the main concerns in dealing with NDVI data in South East Asia is cloud contamination. Malingreu (1986) used a 3-week compositing technique to minimize cloud contamination in Indonesia. There will probably always be some areas under persistent clouds in south East Asia. Software to compute the NDVI for various compositing periods is being developed by NOAA/CIAM.
Figure 1. Smoothed NDVI times series (1985, 1986, 1987) for a
(1/2)° X 1° polygon including Khon Kaen in Northeast Thailand.
NDVI Versus yield relationships
There is a strong relationship between the NDVI and seasonal biomass production over a given region. Interest in obtaining quantitative yield estimates on the basis of spectral reflectance is increasing rapidly as this is seen as a viable alternative to area survey sampling. The technique is still in its infancy and far fro being operational. The lack of a large historical spectral reflectance data base precludes the development o conventional models. Consequently, one has to take the covariance modeling approach where selecting more locations compensates for the fewer number of years of data. Several studies relating crop yield to spectral reflectance from hand held radiometers, landsat and NOASS AVHRR data are available (Barnett and Thompson, 1982; Hat filed, 1983; Ayyanagr, 1984). Sakamoto 91988) examined the relationship between millet and sorghum yield and NDVI (raw, area under the curve, and rate of change) in the African Sahel as part of USAID Famine Early Warning System (FEWS) project. Similar being tested by NOAA in South East Asia.
Integrating satellite data in operational Agroclimatic assessments in South East Asia
The goal of the Phase II effort is to integrate the NOAA AVHRR satellite data products (color images and NDVI time series), at least on a quasi real-time basis, into the agroclimatic assessment bulletins established in phase-I. Real-time capability cannot be achieved until and unless each country develops the ability to receive and process level 1b data. At present, as per of the test and evaluation phase, NOAA is sending GVI resolution weekly NDVI data to Indonesia, Malaysia, the Philippines, and Thailand.
The flow of information and data into the assessment process is :
- Data analysis,
- products,
- Ancillary information,
- Interpretation and preparation and of the impact assessment in the form of a bulletin and
- Dissemination of bulletins t user groups in government agencies.
Interpretation and assessment preparation involves analyzing all of the information available to the assessor coming to conclusions regarding present and prospective crop conditions. Finally, it is important to make sure that each individual country disseminates these bulletins to the appropriate individuals within the key government agencies concerned with national food security.
Present Status of NOAA data Reception
Software provided in the context of this project is designed to accept standard level 1b NOAA data. At the present time, only Thailand both receives and processes AVHRR data in a level 1b format. Indonesia and Malaysia both receive HRPT data from the NOAA polar orbiting satellite, nut do not produce level 1b data. The Philippines does no currently receive does not currently receive digital AVHRR data but plans to construct receiving station within the next year with the assistance f the Australian government.
Conclusions
Data from the AVHRR sensor aboard NOAA's polar orbiting satellites have great potential for monitoring the vegetation resources of the plant. Numerous studies are currently being undertaken to exploit this valuable data source. While spatial regulation is often considered a limiting factor in the use of AVHRR data, the elements of daily global coverage, near real time data availability, a data set which is manageable by personal computers, and economical cost have made AVHRR data attractive to many users of remotely sensed data.
The NOAA and CIAM approach to agroclimatic assessment has focused on the integration of several data sources. We believe that the convergence of evidence concept in assessing agricultural conditions is vital. The use of rainfall data and field reports in support of remotely sensed data provides a better analytical tool than remotely sensed data used alone.
The four nations with which we are working in South East Asia are eager to evaluate the usefulness of AVHRR data for themselves. Working together we hope to advance the use of this toll in the context of operational acgroclimatic assessment.
Acknowledgements
The authors express appreciation to Rita Terry for hereditorial work on the manuscript and to Lutine Hatley for typing the manuscript. Tom Philips and Sandy Weisman are acknowledged for their programming work in support of the project. This work is funded by the United Nations Development programme and the United States Agency for International Development Office of Foreign Disaster Assistance. The authors support is provided by their respective organizations, the national Oceanic and Atmospehric Administration of the United States Department of Commerce and the Economic and Social Commission for Asia and the Pacific of the United Nations.
References
- Ambroziak, R.A., 1984: A new method for incorporating meteorological satellite data into global crop monitoring. Eighteenth International Symposium on Remote Sensing of the Environment, Paris.
- Ayyanagr, R.S. 1984: Applications of Remote Sensing for rice : Indian experience. In Deepak and Rao (eds), Applications of Remote Sensing for Rice production. A Deepak Publishing, Hampton, Virginia, pp. 157-172.
- Barnett, T.L. And D.R. Thompson, 1982: The use of large-area spectral data in wheat yield estimation Remote Sensing of Environment, 12: 509-518.
- Hatfield, J.L. 1983: Remote Sensing estimator of potential and actual crop yield. Remote Sensing of Environment, 13: 301-311.
- Johnson, G.E. A. Van Dijk, and C.M. Sakamoto, 1987: The use of AVHRR data in operational agricultural assessment in Africa. Geocarto International, 1:41-60.
- Justice, C.O. and P. Hiernaux, 1986: Monitoring the grasslands of the Sahel using NOAA/AVHRR data: Niger 1983. International Journal of Remote Sensing, 7:1475-1498.
- Kidwell, D.B. 1988: NOAA Polar orbiter data (TIROS-N, NOAA-6, NOAA-7, NOAA-8, NOAA-9 and NOAA-10) users guide. National Oceanic and Atmospheric administration, Washington, D.C.
- Malingreau, J.P. 1986: global Vegetation dynamics: Satellite observations over Asia. International Journal of Remote Sensing, 7:1121-1146.
- Perry, C.r., Jr. and L.F. Lautenschlager, 1984: Functional equivalence of spectral vegetation indices. Remote sensing of the Environment, 14:169-182.
- Sakamoto, C.M. 1988: Feasibility of using NOAA AVHRR data for estimating millet and sorghum yield in the Sahel. University of Missouri Agricultural Experiment Station Journal Series No. 10, 673.
- Tarpley, J.D., S.R. Schneider, and R.L. Money, 1984: Global Vegetation indices from the NOAA-7 meteorological satellite. Journal of Climate and Applied Meteorology, 23:491-494.
- Tucker, C.J. and P. Sellers, 1986: Satellite Remote Sensing Primary production. International Journal of Remote Sensing, 7: 1395-1416.
- Van Dijk, A., S.L. Callis, C.M. Sakamoto, and W.L. Decker, 1987: Smoothing vegetation index profiles: an alternative method for reducing radiometric disturbances in NOAA/AVHRR data. Phtogrammetric Engineering and Remote Sensing, 53: 1059-1067.
*presented at the 10th Asian Conference on Remote Sensing, Kuala Lumpur, Malaysia, November 23-29, 1989