联合Sentinel-1与Sentinel-2数据的青藏高原东缘草地地上生物量反演

Inversion of aboveground biomass of grassland on the eastern margin of the Qinghai-Tibet Plateau combined with Sentinel-1 and Sentinel-2 data

为探究协同主被动遥感在估算草地地上生物量(AGB)方面的潜力,本研究以青藏高原东缘阿坝藏族羌族自治州红原县为研究区,Sentinel-1 SAR数据和Sentinel-2多光谱影像为数据源,采用多元线性回归、逐步回归、半经验物理模型方式进行建模,探究Sentinel-1和Sentinel-2数据协同反演草地AGB的能力。结果表明,协同反演精度优于Sentinel-2单一数据源反演精度(多元线性回归模型的模型精度R2从0.74增加到了0.83)。另外,联合Sentinel-1和Sentinel-2数据,采用逐步回归方法建立AGB模型,模型精度R2达到了0.78;半经验物理模型的模型精度R2为0.77。总体上,Sentinel-1影响因子能在一定程度上提高了反演模型精度,多种建模方式最终反演的AGB结果与实测草地AGB相符,研究结果可为研究区乃至整个青藏高原东缘草地AGB精确估算提供科学依据。

The aim of this study was to explore the application potential of collaborative active and passive remote sensing in aboveground biomass(AGB)estimation of grassland.In this study,Hongyuan County,Aba Tibetan and Qiang Autonomous Prefecture,on the eastern edge of the Qinghai-Tibetan Plateau was used as the study area.We investigated the ability to retrieval grass AGB by combining Sentinel-1 and Sentinel-2 data,using multiple linear regression,stepwise regression,and semi-empirical physical modeling,with Sentinel-1 synthetic aperture radar data and Sentinel-2 multispectral images as the data sources.The results showed that the cooperative inversion accuracy was better than the inversion accuracy of the Sentinel-2 single-data source(the accuracy,R^(2),of the multiple linear regression model increased from 0.74 to 0.83).In addition,combining Sentinel-1 and Sentinel-2 data with stepwise regression,the AGB model accuracy(R2)was 0.78,and the semi-empirical physical model accuracy(R^(2))was 0.77.In general,the Sentinel-1 influence factor improved the accuracy of the inversion model to some extent,and the final inverse AGB results from multiple modeling approaches were consistent with the measured grassland AGB.The results of this study provide a scientific basis for the accurate estimation of AGB in the study area and even the eastern edge grassland of the whole Qinghai-Tibet Plateau.