Rice monitoring and production estimation using multitemporal RADARSAT

Rice monitoring and production estimation has special significance to China, as rice is the staple grain and accounts for 42% of the crop production in this country. Radar remote sensing is appropriate for monitoring rice because the areas where this crop is cultivated are often cloudy and rainy. Synthetic Aperture Radar (SAR) is thus anticipated to be the dominant high-resolution remote sensing data source for agricultural applications in tropical and subtropical regions. IP also provides revisit schedules suitable for agricultural monitoring. This paper presents the results of a study examining the backscatter behavior of rice as a function of time using multitemporal RADARSAT data acquired in 1996 and 1997. A rice-type distribution map was produced, showing four types of rice with different life spans ranging from 80 days to 120 - 125 days. The Life span of a rice crop has significant impact on the yield, as well as on the taste and quality of the rice, with the longer growing varieties having the best taste and the highest productivity. The rice production of three counties and two administrative regions, totaling 5000 km(2), was estimated in this study. The accuracy of the rice classification was found to be 91% (97% after postclassification filtering) providing confidence that multitemporal RADARSAT data is capable of rice mapping. An empirical growth model was then applied to the results of the rice classification, which related radar backscatter values to rice life spans. These life spans could then be used to sum up the production estimates, which were obtained from agronomic models already in use for rice by local agronomists. These models related the yield of rice to their life span based on empirical observations for each type of rice. The resulting productivity estimate could not be compared to any other existing data on yield production for the study-area, but was well received by the local authorities. Based on the studies carried out in the Zhaoqing test site since 1993, it is suggested that rice production estimates require three radar data acquisitions taken at three different stages of crop growth and development. These three growth stages an: at the end of the transplanting and seedling development period, during the ear differentiation period, and at the beginning of the harvest period. Alternatively, if multiparameter radar data is available, only two data acquisitions may be needed. These would be at the end of the transplanting and seedling development period, and at the beginning of the harvest period. This paper also proposes an operational scenario for rice monitoring and production estimation. (C) 2001 Elsevier Science Inc. All rights reserved.