Research On Soil Moisture Extracted Method Based On Multi-source Data

It is very important to accurately estimate surface soil moisture in large area and real time in the work of “Bohai Sea Granary” scientific and technological demonstration. The traditional soil moisture acquisition is based on ground station monitoring, which has high precision and be continuous in time but in point scale. It needs ground-based observation data verify precision while acquired soil moisture with remote sensing retrieval, which has characteristics of macro, dynamic, real- time and multi- source. In order to comprehensively utilize the advantages of two sides to acquire soil moisture data with high precision and special continuity, the paper chose Wudi country of Binhai city in “Bohai Sea Granary” Shandong province as research region, used ground observation network data, ground synchronous observation data and remote sensing image data MODIS and researched on soil moisture extracted based on studying data interpolation method, up-scale method and remote sensing inversion model with comprehensive utilization of different data advantages and complement each other. The main research content is as follows:(1) Put forward an interpolation method, BMA-I, for ground observation soil moisture with comprehensive consideration of temporal and special informationUsing collaborative kriging method as space forecasting model, time-series recursive forecasting method based on LS-SVM algorithm as time-series forecasting model, averaging model by BMA method, we established a new interpolation method BMA-I. Then it was compared with interpolation only using one method such as ordinary kriging or collaborative kriging. By analysis the results, we can see that the new method can improve the smooth effect brought by the kriging method and better reflect the real soil moisture data to some extent.(2) Proposed an improved method to calculate apparent thermal inertia and up-scale soil moistureFirst, calculated temperature difference using site surface soil maximum temperature and minimum temperature acquired by ground observation network, and calculated ATI in pixel scale with MODIS surface reflectance products. Then, established function relationship between soil moisture interpolated above and ATI, acquired up-scaling surface soil moisture data based on Bayesian Linear Regression in pixel scale.(3) Built soil moisture retrieval method based on improved temperature-vegetation drought index modelExtracted Ts and NDVI data from MO DIS remote sensing image, constructed the feature space, calculated TVDI by dry and wet side equation, built up relationship model between soil moisture data up-scale obtained and TVDI to retrieve area surface soil moisture. It has been verified and the results showed that it was much better than only with using small amount of ground observation data.