Rainfed Rice Mapping using RADARSAT

Rainfed Rice Mapping using RADARSAT

Surassawadee Phoompanich, Thanwarat Anan, Wimon Pathong,
Ekkarat Pricharchon, Siam Lawawirojwong, Supapis Polngam
Geo-Informatics and Space Technology Development Agency (Public Organization)
196 Phahonyothin Road, Chatuchak, Bangkok, Thailand 10900
Tel. +66(0)-2940-6420-9 Fax: +66(0)-2579-0116
Email: surassa@hotmail.com, twarat@gistda.or.th, wimon@gistda.or.th, ekkarat@gistda.or.th, siam@gistda.or.th, supapis@gistda.or.th

ABSTRACT
RADARSAT imagery were used with relevant geo-informatic data to study the suitability of radar for rice mapping in the rainy season. A typical rice growing area in Thailand was selected as a case study project area. The area is located in Nakhon Nayok province in the central of Thailand. Multi-temporal RADARSAT data C-HH band acquired on October 15, November 8, and December 19, 2003 were processed to compare with the ground truth data of first tilling period, vegetative period and harvesting period. The composite satellite imagery were interpreted according to the tone, color, texture, pattern, shape and association. The results were also compared with the existing land use map. The accuracy of the classification was also checked against the ground information.

INTRODUCTION
Rice is not only a food staple, but also a key export commodity in Thailand. Its monitoring and yield estimation has special significance. For several time, using optical remote sensing is particularly difficult due to cloud cover since the rice growing seasons often coincide with the rainy seasons. Synthetic Aperture Radar (SAR), like the RADARSAT satellite can provide information which cloud-penetrating is useful alternative for rice crop. Therefore, it is important to study an efficient way to face this disadvantage. Beside this, Geo-Informatics technology can be used to address rainfed rice area with the usefulness of multi-temporal RADARSAT data for rice area acreage throughout the rice growing season as well as to select the optimal combination of imagery for the discrimination of rice growth stages.

OBJECTIVE
To apply RADARSAT imagery and related geo-informatics data for a study of rainfed rice area, and to extend the developed methodology to other areas.

APPROACH
The process of determining rainfed rice area is based on rice crop in 2003 which can be performed using RADARSAT data. The study site is in Nakhon Nayok province (see Figure 1). It is located in the central of Thailand. The main physical areas are flat, except the northern part is a mountainous area which a part of Khao Yai National Park. Soil characteristic is clay and sandy clay. This study using RADARSAT data were acquired during different growth stages shown in Table 1. Then, the RADARSAT imagery was processed in combination imagery were used to characterize a rice backscattering. Finally, the best band combination which can provide an excellent seperability between rice crop and other crops was chosen or interpretation rice crop area. The process of rainfed rice area determination for Nakhon Nayok province is shown in Figure 2.

Figure 1: Nakhon Nayok case study

Table 1: RADARSAT data used (C-HH band)

Figure 2: The process of rainfed rice mapping.

RESULTS AND DISCUSSION
The multi-temporal RADARSAT imagery were to generate color composite imagery. The imagery of different rice growth stage (see Figure 3) showed the radar backscatter of each stage (see Figure 4), and the different color combination from different dates of imagery. The optimal image combination can reveal an excellent seperation between rice and other crops. This study found that the best color composite was derived using September 28 image in red, October 29 in green and December 16 in blue (see Figure 5). The various colors of yellow, greenish blue, magenta and blue (see Figure 6) in the color composite image represent the different growth stages of rice such as red color representing first tilling period, green color a vegetative period, and blue color a harvesting period.

Figure 3: The radar backscatter from different rice growth stages.

Figure 4: The radar backscatter signatures from different rice growth stages in terms
of decibel with ground truth.

Figure 5: The best band combination of RADARSAT Black & White imagery.

Blue color; Ban Khlong Yi Sip Sam, Ongkarak district

Greenish yellow color; Ban Nong Bua, Ongkarak district

Magenta color; Ban Chik Sung, Pak Phli district

Yellow and yellowish color; Ban Sabok Khieo, Mueang Nakhon Nayok district

Figure 6: Comparison RADARSAT characteristics of Black&White and color
composite imagery during different rice growth stages.

The rainfed rice area could be derived from visual interpretation of the best band combination (see Figure 7), the results of which are as follows.

1) The ploughing or soil preparing stage with water being released into rice paddy results in the radar backscatter from water surface. As results, RADARSAT imagery appears as a black color.
2) A rice growth stage is in a first tilling and vegetative period. The high radar backscatter is likely to occur due to moisture content and multiple corner reflectors of rice trees, and water surface. RADARSAT imagery appears as a white color.
3) A ripe rice and harvesting period is the stage with thick leaves and declining moisture content; therefore, resulting in the low radar backscatter from the top of rice trees. RADARSAT imagery show gray color.

The total rained rice area in Nakhon Nayok province (see Figure 8) was 100,688 hectares. Most of the rainfed rice area was found in Ongkarak district, with 35,672 hactares, Mueang Nakhon Nayok district, Banna district, and Pak Phli district, with 28,371 , 19,424 and 17,221 hectares, respectively. Then, to check the accuracy of image interpretation, ground observation was made at 16 sampling points of the study area. The accuracy was found to be 82 percent.

Figure 7: Color composite imagery of Nakhon Nayok province.

Figure 8: The rainfed rice area from RADARSAT imagery.

CONCLUSION
The results indicate that the multi-temporal RADARSAT imagery have been successfully used to delineate and map the different rice crop in the test site. This study based on the time series of radar backscatter from the rice crop. Moreover, the SAR data can used to determine rainfed rice area could be addressed the rice growth stage development and the environmental conditions at the acquisition time in order to improve interpretation accuracy. In this study the best band combination of multi-temporal RADARSAT imagery acquired on 3 different dates (September 28, October 29 and December 16, 2003 in red, green, and blue) revealed the best identification accuracy.

This study demonstrates the potential of RADARSAT imagery in rice crop monitoring which can be applied or performed to other plants. Beside this, the users should have the basic knowledge of SAR data. So this basic data could help to analyze RADARSAT data, and can improve interpretation and accuracy.

REFERENCES

Staples, G.C. and J. Hurley., 1996. Rice crop monitoring in Zhaoqing, China using RADARSAT SAR intial results. Available Source: http://www.gisdevelopment.net/aars/acrs/1996/ts1/ts1001.shtml, May 16, 2002.
Liew, S.C., 1997. Rice crop monitoring in the Mekong River Delta using combined ERS and RADARSAT Synthetic Aperture Radar. Available Source: http://www.crisp.nus.edu.sg/~research/research/forest/
Toan, T.L., F. Ribbes, L.F. Wang, N. Fluory, K-H. Ding and J.A. Kung and M. Fujita. 1997. Rice crop mapping and monitoring using ERS-1 data based on experiment and modeling results. IEEE Trans Geoscience and Remote Sensing.
Vibulsresth et al. 1993. Preliminary Study of SAR data for Coastal area in Thailand. Proceeding of the UN/CHINA/ESA Workshop on microwave Remote Sensing applications. Remote Sensing Division National Research Council of Thailand, Bangkok.
Mard, P.V.C., 1999. Monitoring rice crops in Vietnam. Available Source: http://www.rsi.ca/resources/agriculture/agi.htm, May 16, 2002.
Shao, Y. and R.N. Treuhaft., 1999. Optimum Systhetic Aperture Radar system parameters for rice and tropic vegetation monitoring. 19th Asian Conference on Remote Sensing, Hong Kong.