基于全极化Radarsat-2数据的水稻生物量估算模型

依据苏州东桥试验区水稻生长期观测数据,采用多时相水稻拔节到抽穗期全极化Radarsat-2数据,分别分析了不同极化HH、VV、CROSS、比值HH/VV的雷达后向散射系数时域变化特征与生物量的相关关系,构建水云模型、二次多项式模型和指数模型反演水稻生物量。反演结果表明:HH、CROSS水云模型都有不错的反演效果,相关系数分别为0.910、0.902,而HH水云模型反演生物量尤佳,均方根误差为0.190。指数模型普遍优于二次多项式模型,HH/VV指数模型效果出众,相关系数为0.929,均方根误差为0.164。通过比较分析不同极化的水云模型、二次多项式模型和指数模型,HH水云模型与HH/VV指数模型反演水稻生物量精度相对较高。

Rice Biomass Estimation Based on Full-polarization Radarsat-2 Data

The objective of this study was to investigate the interaction between microwave backscatter signatures and rice biomass.Our unique data consisted of Radarsat-2 microwave backscattering coefficients at all polarizations(HH,VV,CROSS and HH/VV Polarization) for two rice crop stages.Fresh biomass of the whole plant was measured at the same time in Dongqiao of Suzhou.The study analyzed microwave backscatter signatures in time domain,then built a simple backscatter process model(the water cloud model) and regression models.Analyses based on the water cloud model showed that fresh biomass was better correlated with HH and cross-polarization,correlation coefficient was 0.910,0.902 respectively,while fresh biomass was best correlated with HH,RMSE was 0.190.The exponent models were better than polynomial models,while fresh biomass was best correlated with HH/VV using exponent model.In a word,HH water cloud model and HH/VV exponent model were optimal to simulate rice fresh biomass.