基于岭估计的青海省东部农业区ET_0遥感反演研究

论文利用青海东部农业区内的12个气象站2003—2005年气象资料,应用Penman-Monteith公式计算得到各站逐旬的ET_0值,并进一步研究其与高程(DEM)、地表温度(LST)及归一化植被指数(NDVI)3个因子之间的关系,提取遥感数据并耦合到时间分辨率为旬,空间分辨率为1 km,将其与计算所得ET_0建立多元反演模型。由于3个自变量因子之间存在着很强的相关关系,LST与NDVI间判定系数R2平均在0.7左右,不能直接用最小二乘回归方法建立模型。为有效避免自变量间相关性对模型的影响,研究中采用岭估计方法建模。结果表明,通过岭估计建立2003年10~33旬区域二元模型反演最低精度达76.19%,区域三元模型反演最低精度也有83.54%。与传统方法所建模型相比,检查点均方根误差减小约1.1,反演最低精度提高11%左右,能满足实际应用需求。

Remote Sensing Inversion of ET0 in Eastern Agricultural Area of Qinghai Province Based on Ridge Estimation

In this study, ET₀ values were calculated by using the Penman-Monteith equation and the meteorological data of the 12 weather stations in the eastern agricultural area of Qinghai Province from 2003-2005. Remote sensing data was assimilated with the digital elevation models (DEM), land surface temperature (LST), and normalized difference vegetation index (NDVI) to acquire the data set with time resolution of ten days and the spatial resolution of 1 km. Then a multivariate retrieval model between ET₀ and these variables was established. Since there is a strong correlation among these independent variables, for example the correlation index R² of LST and NDVI is about 0.7 on average, a partial least-squares regression cannot be directly established. This study established a model with the ridge estimation method which can reduce the model error and improve the retrieval accuracy. The minimum retrieved accuracy reached 76.19% and 83.54% in 2013. Compared with traditional methods, the checkpoint variance was significantly reduced and the retrieval accuracy was improved. The model met the requirements of practical application.