干旱区生态环境知识资源中心

Knowledge of spatial distributions of soil moisture, particularly at large spatial scales, is critical for many practical reasons. Unlike point-scale measurements, remote sensing provides an efficient way to estimate soil moisture over large areas. By resampling MODIS images with different scaling techniques, the impacts of those scaling methods on accuracy of soil moisture retrieval from MODIS data was first investigated using in situ soil moisture measurements. A soil moisture retrieval model was then constructed to obtain and analyse the spatial pattern of soil moisture in the Xinjiang Uygur Autonomous region of China in 2007. The following results were obtained. (1) With the process of spatial scale increases, the value of goodness-of-fit of the model change is relatively large, showing a strong randomness. (2) With resampling scales for 2 and 4 km when compared with measured data, correlation coefficient showed apparent fluctuation. With increased scale sampling, random changes in the model appeared to fluctuate less. Comparison of six different scaling methods, the results indicated that soil moisture retrieval model showed better correlation and higher accuracy of fitting under the scaling method of 2 km, respectively, followed by 1 and 16 km. (3) In order to test the accuracy of the retrieval model, the distribution of soil moisture was analysed by using of satellite image of the study area and retrieval of in situ soil moisture data, the data demonstrated high consistency with fieldwork. Evidence thus indicates that 2 km x 2 km is a significant level for retrieval distribution of soil moisture in the study area. The results also provide some reference for land-use planning and policy-making of sustainable utilisation of land resources.