Inversion and distribution of soil moisture in belly of Maowusu sandy land based on comprehensive drought index

Abstract: In order to accurately estimate soil moisture content based on index of drought monitored in the complex type cover, this paper studied the feasibility of integrating the single index to estimate soil moisture. Wushen Banner in the abdomen of Maowusu sandy land was the study area. The annual wind velocity of this area was 3.4 m/s. The land use of this area included grassland, sandy land, forest, farmland, water body and construction land, which accounted for 55.66%, 29.3%, 10.9%, 2.75%, 0.7% and 0.69% of the total area, respectively. The remote sensing data were from Landsat 8 OLI images on August 26 and September 28 of 2016. The spatial resolution was 30 m. The data were carefully preteated. The sampling was carried out in 24 plots of 12 sandy land and 12 grassland. Traditional 3 single drought monitoring indexes were selected. The weights of each index was determined by analytic hierarchy process. Together with the measured soil moisture data from different cover types, regression analysis was applied to establish the multi-index comprehensive drought monitoring model (CDI). The accuracy of the model was validated by determination coefficient, average relative error and root mean square error (RMSE). Moreover, the spatial distribution of surface soil moisture in the studied area was inverted and analyzed based on multi-index model. The results showed that the 3 single drought indexes could objectively reflect the drought characteristics, displaying a negative correlation with surface soil moisture content. The best temperature vegetation dryness index (TVDI) correlation was 0.604. The comprehensive drought monitoring index model and multiple indexes were introduced. The determination coefficient in surface soil moisture index models for grassland and sandy land in August and September were 0.751 and 0.749, respectively. On the whole, the determination coefficient of the exponential model reached 0.744 and 0.727, respectively, which were higher than the fitting effect of a single index. With the increase of soil moisture content, the CDI value showed an obvious decreasing trend, which was ranked as sandy land higher in grassland in size. It indicated that the comprehensive drought monitoring index could better reflect soil moisture distribution in the studied area. The surface soil moisture in this area based on the model inversion was generally low, and the area with the moisture content less than 0.15 cm3/cm3 accounted for 96.47%. And most of these areas were far away from lakes and reservoirs. According to different land types and soil moisture distribution, 2 different land types such as grassland and sandy land were statistically analyzed. The results showed that the different cover types had a quite different moisture content, with an average of 0.087 cm3/cm3 for grassland in August, 0.026 cm3/cm3 for sandy land in August, 0.102 cm3/cm3for grassland in September, 0.029 cm3/cm3 for a sandy land in September. The surface characteristics of soil moisture spatial distribution for grassland was consistent with the sandy land. The accuracy verification of grassland, sandy land and the whole inversion results showed that the average relative error was less than 10%. Generally, the soil moisture content in the studied area decreased gradually from east to west and from north to south.