Soil Moisture Monitoring System Based On Remotely Sensed Data Reconstruction

Soil Moisture is the most important water source for plant growth, and is the key variable for the land surface parameterization in the regional and global scale. Satellite remote sensing is an important way for monitoring soil moisture. Because satellite sensors can easily be affected by clouds, aerosol and so on, the information obtained often doesn’t accurately reflect the surface information, especially in the southern area of our country. So we proposed a data reconstructed method, named background method, based on background database and mTSF principle, together with quality control information (QA flag) from2003-2011, to rebuild MODIS LST (Land Surface Temperature) and NDVI (Normolized Difference Vegetation Index) in Jiangsu Province. The method fills the vacant data, improves the accuracy of the inversed soil moisture and meets the work needs. Finally, the study applied C#and IDL to build the drought monitor system of cropland. The main conclusions are as follows:(1) The pixels ratio of good quality of pre-reconstructed MOD09A18days land surface reflectance products reaches the maximum from2003-2011. The ratio of cloud contamination is the highest in summer, followed by winter and autumn, spring is relatively lower. The MODIS LST daily products show the result that the pixels’ ratios of cloudy effects reach the highest from2003-2011. The cloud contamination mainly occurs in summer, followed by autumn and winter, spring is the lowest. In order to improve the precision and effiency of remotely sensed soil moisture monitoring, rebuilding the time-series LST and NDVI is a way to restore the data quality.(2) The data reconstructed method based on the background database and mTSF principle can fill the NDVI invalied pixels, and the results achieve relatively good precision. Compared with different methods, the average correlation coefficient between the method we proposed and the original good quality NDVI is0.84, which is the highest. But the average correlation of Savitzky-Golay (S-G), Asymmetric Gaussian (A-G) and Doubled Logistic (D-L) method is0.80,0.73and0.72respectively. In the supposed cloudy area experiment, the reconstructed NDVI based on background database method has good accuracy, the average absolute error with the original NDVI is0.049, the average relative error is10.14%, the average RMSE is0.059and the average correlation efficient is0.82.(3) The data reconstructed method based on the background database and mTSF principle can keep the original real LST and reflect the trends of LST in time series. The rebuilt LST in the cloudy area has fine response relationship with the in situ land surface temperature from2009-2011. The average correlation between LST pixels affected by cloudy contamination and the in situ values is0.87. After the reconstruction, the average absolute error decrease. In the supposed cloudy area experiment, the reconstructed LST based on background database method reach good precision, the average absolute error with the original LST is0.57℃, the average RMSE is0.71and the average correlation efficient is0.78.(4) The rebuilt data provide remote sensing soil moisture monitoring effective service, and the result can accurately reflect the soil moisture change in the winter wheat area in Huaibei. In no-rain times, the deviation between the inversed data and the in situ data is0.22%-2.89%, the average relative error is20.28%-26.69%. On the contrary, before the reconstruction, only the original image on the March26th has LST values in the in situ stations, and the average relative error reaches35.19%.(5) The integrated application of IDL and C#build the soil moisture retrieving system of wheat area in Huaibei. The system allows users to monitor near real-time drought condition dynamically and to improve the intention and efficiency of anti-drought work. The platform will provide information on drought warning and drought decision timely and effectively.