Studies On Monitoring Soil Drought By Remote Sensing Based On The TVDI And Their Validation

The Temperature-Vegetation-Drought-Index method (TVDI method) was used for dynamic remote sensing monitoring of drought in Anhui province between April to October 2013. The results of remote sensing monitoring was verified by using the data of soil moisture automatic station in this paper.During the study, EVI and NDVI data were extracted from the MODIS data products of vegetation index data (MOD13A2),by combining with the surface temperature data which was extracted from the land surface temperature data products (MOD11A1), the LST-VI feature space based on two kinds of vegetation index was generated. The MOD13A2 and MOD11A1 were released by the NASA. On this basis, similarities and differences of the various period LST-NDVI space and LST-EVI space were compared, then the merits of the two kinds feature space by fitting the Equation of dry-edge and wet-edge were analyzed. This paper also used FY-3A/VIRR data and selected pixels in all the periods in study area to draw LST-VI scatterplots, then fitted the equation of wet-edge and dry-edge on the basis of the scatterplots.Basing on the above results, results of spatial distribution of three kinds of TVDI index (TVDIa,TVDIb,TVDIc) at different periods of time in the area of the research were calculated, and respectively pointed out the time and the place of drought occurrence reflected by the three kinds of TVDI. By using the spatial information extraction tool of the ArcGIS, TVDIa, TVDIb and TVDIc of pixels at all sites at each period of time were extracted, for the correlation analysis of TVDI index and measured values of soil moisture.Principal components were analyzed by the measured data of soil moisture from automatic station, and three principal components which can reflect most information of the measured data were obtained. Then taking the three principal components as independent variables, proceeded stepwise regression for the three kinds of TVDI respectively, explored the correlation between the three TVDI indices and the measured moisture value of different soil layers.The main conclusions of this paper are as follows:(1) The equation of dry-edge and wet-edge were fitted by using MODIS data, the shape of the generated LST-Ⅵ scatterplot presented triangular or trapezoidal distribution, and the scatterplot of taking EVI as vegetation index presented the characteristics of triangle more obviously; the fitting results of the equation of dry-edge and wet-edge showed that the generated equation of dry-edge and wet-edge using two kinds of vegetation indices mostly had significant correlation with dry-edge and wet-edge presented from the scatterplots (when selecting EVI as vegetation index, the result was all highly significant correlated); the fitting results with taking EVI as vegetation index was more ideal than that with NDVI.(2) The full time LST-EVI scatterplot generated by using the FY-3/VIRR data obviously presented the characteristics of trapezoidal distribution, which match the premise of TVDI model application, and showed that selecting full time LST-EVI scatterplot to fit the equation of dry-edge and wet-edge is feasible.(3) By calculating three kinds of TVDI values of each pixels in the study area in different periods of time, three kinds of TVDI reflecting the drought distribution were obtained, distributions of the three kinds of TVDI had certain consistency in time and space, but the detailed differences were obvious.(4) Through principal component analysis for the automatic station soil moisture data, three principal components (FAC1 FAC2, FAC3) that can reflect all the soil water information were obtained; wherein, FAC1 was comprehensive soil moisture index, reflecting the comprehensive condition of soil moisture; FAC2 was deep soil moisture index, mainly reflecting the status of soil moisture below 50cm or deeper, FAC3 was shallow soil moisture index, mainly indicating soil moisture conditions of shallow layer within 1 Ocm.(5) Three stepwise linear regression modes of three TVDI indices (TVDIa、 TVDIb and TVDIc) and three principal components of soil water content were obtained; the results of correlation analysis for three kinds of indices of three principal components showed that the TVDIa can reflect the shallow soil moisture information very well, TVDIb can excellently reflect the shallow soil moisture information of less than 10 cm and soil moisture content of deeper than 50 cm, but the soil moisture information of the middle layer cannot reflected, and TVDIc can be used to reflect overall information of soil water content to a certain extent.