Early detection system of drought in East Asia using NDVI from NOAA/AVHRR data

Transcription

1 INT. J. REMOTE SENSING, 20 AUGUST, 2004, VOL. 25, NO. 16, Cover Early detection system of drought in East Asia using NDVI from NOAA/AVHRR data X. SONG* Institute of Geographical Sciences & Natural Resources Research, Chinese Academy of Sciences, Building 917, Datun Road, Anwai, Beijing , China; G. SAITO National Institute for Agro-Environmental Sciences, Kannondai, Tsukuba, Ibaraki , Japan M. KODAMA Computer Center for Agriculture, Forestry & Fisheries Research, Kasumigaseki, Chiyodaku, Tokyo , Japan and H. SAWADA Forestry and Forest Products Research Institute, 1 Matunosato, Tsukuba, Ibaraki , Japan (Received 24 April 2003 ) 1. Introduction Drought causes damage to agricultural production and affects the balance of food supply and demand. As Obasi (1994) mentioned, droughts had affected 50% of 2.8 billion people who suffered from weather-related disasters. Moreover, 1.3 million people have died due to the direct and indirect cause of drought during East and South East Asia has the largest population in the world, and it is also one of the largest agriculture-producing regions in the world. If drought occurs in this area, it will affect the balance of food supply and demand all over world. On the other hand, according to Gol dsberg (1972) using information of vulnerability of major agricultural areas to drought based on the difference between annual precipitation and annual potential evaporation, the negative difference (210 to 2500 mm annually) areas are listed as the frequent intensive drought area. East Asia, especially China, is also one of the largest regions where the drought occurs easily in the world. It is very important to mitigate the damage caused by the drought. Since timely information of drought about its extent and intensity can help *Corresponding author. International Journal of Remote Sensing ISSN print/issn online # 2004 Taylor & Francis Ltd DOI: /

2 3106 X. Song et al. to reduce the food supply shortage by the transport of necessary food, it is required to develop early detection system of drought effects on agriculture. In recent years, the Advanced Very High Resolution Radiometer (AVHRR)- based Normalized Difference Vegetation Index (NDVI) has been used for drought monitoring (Kogan 1997). However, the different characteristics of vegetation in different areas, make it difficult to detect the effects of drought on agriculture. Therefore, the objective of this research is to develop an early detection system of drought effects on agriculture in East and South East Asia using the NDVI from National Oceanic and Atmospheric Administration (NOAA)/AVHRR data. With this system, the drought effects on agriculture in China in 2000 are detected and monitored successfully. 2. Satellite data We have used the NOAA/AVHRR data from the Satellite Image Database System in Agriculture, Forestry and Fisheries (SIDaB) (Kodama and Song 2000). In this system the user can search and download the satellite images freely using world wide web (WWW) and file transfer protocol (FTP) ( agropedia/). The NOAA/AVHRR data of this system are received at three stations in Japan, Shiogama, Yokohama and Ishigaki, and one station in Thailand, Bangkok belonging to the Asian Institute of Technology (AIT). These data are transmitted using the high-speed network system, and processed at the Computer Center for Agriculture, Forestry and Fisheries Research, Japan. The geometric correction and mosaic are done in short time using TeraScan software and EWS. The daily, weekly, 10-day and monthly NDVI, MCSST and other datasets are supplied automatically. 3. Methodology We use the changes of up-to-date and normal NDVI from NOAA/AVHRR data for detecting and monitoring drought effects on agriculture. Ideally, for detecting and monitoring drought, cloud-free satellite data are required. Therefore, to minimize the cloud effect, the largest NDVI value of every pixel is used for the daily, weekly and 10-day images. However, even using this method, we found that there are too many cloudy cover areas in the daily and weekly images, which makes it difficult to detect drought effects on agriculture using these data. On the other hand, monthly data are averaged over a period that is too long to describe the development of vegetation because morphological changes and leaf appearances occur in a shorter period. Therefore, in this work, the detection and monitoring of drought effects on each 10-day interval are discussed. The algorithm of this research is that (figure 1): the standard NDVI images are created first. Then, the up-to-date NDVI is calculated just after the 10 days and cloud pixels are masked for elimination. And then, the difference NDVI images between the standard and up-to-date NDVI images are calculated. Lastly, using the difference NDVI image, it is possible to detect the area and intensity of drought damage on agriculture. The bigger negative difference pixels are listed as drought areas, and the values of difference NDVI show the intensities of the drought damages.

3 Early detection of drought in East Asia 3107 Figure 1. The flow chart of the research Creating the standard NDVI image Firstly, we select a suitable period NOAA data for creating the standard NDVI images. On thinking of climatic periodicity, it seems that the longer time series data are suitable for creating the standard NDVI image. The NOAA/AVHRR data in SIDaB are useful from However, it is known that clouds and other atmosphere constituents obscure the land surface, reducing NDVI considerably (Goward et al. 1991, Gutman 1991, Townshend 1994). Moreover, due to a volcanic eruption, the NDVI was depressed for a long time. On the other hand, on thinking of agriculture, since there are changes in the vegetation from one year to the next, the too old data are of little use for the explanation of vegetation changes. Therefore, considering both these factors, because the atmospheric state has been almost stable since 1997, the NOAA/AVHRR data in were selected for creating the standard NDVI images. In order to minimize the cloud effects, the 10-day NDVI images are created using the maximum value of every pixel. Then, the 10-day NDVI images are masked because there are also cloudy pixels even in this method. The cloudy pixels (NDVIv0) are substituted as the special value areas so that these pixels are eliminated when the averages of NDVI data are calculated. The average values of NDVI data in are those of standard NDVI images. Finally, the standard NDVI images are created. One example of the standard NDVI images is shown in

4 3108 X. Song et al. Figure 2. One sample of the standard NDVI image. figure 2. There are seven classes of NDVI values. This shows the vegetation condition in the last 10 days of May Creating the up-to-date NDVI image The up-to-date NDVI is calculated just after the 10 days and cloud pixels are masked for elimination. As described above, there are also cloudy pixels in the upto-date 10-day NDVI image, even using the maximum value of each pixel for the daily and 10-day images. For these areas, it is impossible to monitor the change of

5 Early detection of drought in East Asia 3109 Figure 3. One sample of the up-to-date NDVI image. vegetation. Therefore, these pixels (NDVIv0) are substituted as the special value area. In the NDVI image, these areas are shown in black (figure 3). When the NDVI image is used, these areas will be eliminated. We do not discuss the changes of vegetation at-these-areas. Comparing figure 2 with figure 3, the difference between May 2000 and the standard May is clearly found. The values of NDVI in May 2000 in the areas of N, E and N, E are less than that of the standard May.

6 3110 X. Song et al Creating the drought risk map The difference NDVI images between the standard and up-to-date 10-day NDVI images are calculated, and the drought risk maps are created using this difference NDVI. The bigger difference NDVI pixels are listed as the drought risk areas, and the values of this difference NDVI show the intensities of drought effects. The bigger negative difference pixels (difference NDVIv20.1) are listed as the drought risk areas (yellow and red areas are shown in figure 4 and cover), and the values of the difference NDVI show the intensities of drought damages. 4. Detecting and monitoring drought in China China is essentially an agricultural country with about 80% of its total population engaged in agriculture. It is the biggest developing country, with onefifth of the world s population. Therefore, it is important to detect and monitor Figure 4. One sample of the drought risk map by the difference NDVI between standard and up-to-date NDVI.

7 Early detection of drought in East Asia 3111 the changes of the agriculture in China. In 2000, there were heavy droughts in China. In this research, using the early detection system described above, the drought area and intensity have been successfully detected and monitored. Using the early detection system described above, the drought risk maps for each 10-day interval in 2000 were created and the NDVI difference and intensity in the last 10 days of May is shown in figure 4. From this figure we can see that there were drought areas in East China, especially, it seems, the heavy drought occurred at areas of N, E and N, E. The drought caused much damage to agriculture in Conclusions In this research, using the changes of up-to-date and normal NDVI from NOAA/AVHRR data, an early detection system for drought is developed. The standard NDVI images are created using NOAA/AVHRR data in The drought risk maps are created. The drought risk maps in China in 2000 are discussed as the example. Compiling the analyses of characteristics of NDVI graph and meteorological precipitation data, the drought effects on agriculture in 2000 are detected and monitored successful. It is shown that this system is possible and useful to detect the drought area and intensity on agriculture. Acknowledgments This research was supported by the following research grants: the Knowledge Innovation Key Project (KZCX-SW-317/CX10G-E01-08), Chinese Academy of Sciences and the director s grant of Institute of Geographical Sciences & Natural Resources Research, Chinese Academy of Sciences. It also was financially supported by ACT-JST: Research and Development for Applying Advanced Computational Science and Technology of Japan Science and Technology Corporation. References GOL DSBERG, I., 1972, Agroclimatic Atlas of the World (Leningrad: Gidrometeoizdat). GOWARD, S. N., MARKHAM, B., DYE, D. G., DULANEY, W., and YANG, J., 1991, Normalized difference vegetation index measurements from the Advanced Very High Resolution Radiometer. Remote Sensing of Environment, 35, GUTMAN, G. G., 1991, Vegetation indices from AVHRR data: an update and future prospects. Remote Sensing of Environment, 35, KODAMA, M., and SONG, X., 2000, A new remote sensing database system in Ministry of Agriculture, Forestry and Fisheries, Japan. 51st International Astronautical Congress, 2 6 October 2000, Rio de Janeiro, Brazil. KOGAN, F. N., 1997, Global drought watch from space. Bulletin of the American Meteorological Society, 78, OBASI, O. P., 1994, WMO s role in the international decade for natural disaster reduction. Bulletin of the American Meteorological Society, 75, TOWNSHEND, J. R. G., 1994, Global data sets for land applications from the Advanced Very High Resolution Radiometer: an introduction. International Journal of Remote Sensing, 15,