Rice Mapping And Biomass Inversion With Multi-temporal ASAR Data

Rice crop is one of most important grain crops in the world. The planting area andgrowth status is valuable for grain safety. Rice is always planted in the tropic andsub-tropic area, so it is difficult to gather the rice information by optical remote sensingimage due to cloudy and rainy weather condition during the whole rice season. However,microwave remote sensing data has been a primary source of remote sensing data forrice monitoring because of its capabilities of penetrating through clouds, and all-timeand all-weather monitoring.In this paper, three multi-temporal ENVISAT ASAR data were acquired in ricecultivated season in Nanchang County, Jiangxi Province. Rice parameters (such as freshweight, dry weight, plant height, leaf area index) were measured. With the analysis ofthe data, methods of rice area extracting and biomass estimation were developed. Themain contents and results are:(1) Backscattering characteristics of rice were analyzed and compared with thebackscattering characteristics of other surface objects in the study area based on3temporal Advanced Synthetic Aperture Radar (ASAR) datasets. Two classificationmethods (Maximum Likelihood Classification (MLC) and Support Vector Machine(SVM)) were used for the extraction of rice paddies. The results showed that bothmethods could extract rice area, and the accuracy of rice identification of MLC was73.81%but that of SVM was80.95%. The SVM identification accuracy level washigher and was further used for the mapping of rice biomass.(2) Understood the relationship between measured biomass and inversed biomass.The water cloud model could be used for rice biomass inversion. The R2value betweenmeasured biomass and inversed biomass in HH polarization was0.55and that in VVpolarization was0.48. Thus, HH polarization was slightly more sensitive to biomassvariation and more suitable for rice biomass inversion than VV.(3) To further improve the feasibility of inversion as measured with high R2value, wecombined optical and HH datasets in the rice biomass inversion. The result showed thatthe model based on two types of remote sensing data has met the needs for rice biomass inversion and the R2value became0.81. According to inversion result and imageanalysis, rice biomass distribution on two dates was mapped. Within the extracted ricearea, biomass change between two dates in Nanchang County was analyzed.