从MODIS资料提取土壤湿度信息的主成分分析方法

Principal Component Analysis Method Acquiring Soil Moisture Information from MODIS Data

沿袭了遥感地物分类中K-L变换思想,以归一化处理后的遥感数据,结合地面土壤湿度观测数据,应用主成分分析方法,采用MODIS不同红外波段数据,通过相关关系计算,在监测结果中融合MODIS具有250 m分辨率的第2波段数据,建立了青海省多维特征空间土壤湿度监测模型。模型的建立克服了热惯量法监测干旱需多时相遥感数据的缺陷,经初步检验,此模型可以在一定精度范围之内监测表层土壤湿度,具有业务应用潜力。

Monitoring soil moisture exactly is very important. Soil moisture is one important factor of agricultural meteorology, which can reflect humid condition of soil, and is a main base to forecast drought of farmland. But it is influenced by too many factors, so it is difficult to monitor real-time soil moisture of largescale areas. General method such as soil sampling method, neutron probe method and TDR method take much time and efforts, and can only monitor limited spots. However, the development of remote sensing technology can provide assistance to monitoring real-time soil moisture of large-scale areas dynamically. Thermal inertial is a matured technological method to monitor bare soil moisture applying MODIS data. But it needs remote sensing data of both daytime and nighttime, which is difficult to obtain in practical operation. It inherits the idea of K-L transformation that is applied in the remote sensing system of target classification, uses principal component analysis, regression analysis, residual image, and relative reflection of internal average methods to correct remote sensing radiation data, and establishes soil moisture model of multiple-dimensional feature space based on mono temporal normalized MODIS data. Then the result from the image of monitoring is obtained and checked up. Monitored and model resuits of 35 spots are compared, and the accuracy of the model is 80%. It shows that the model has potential to be applied in operation. More problems are discussed, including representative of the data, the calibration of remote sensing detector and so on.