Soil Moisture Estimation Under Different Soil-vegetation Types In High Elevation And Cold Mountainous Area

Soil moisture is an important variable in hydrological analysis and watershed simulations.The upstream of the Heihe River Watershed is located in Qilian Mountain,Northwest China.In this area,soil moisture shows strong variations because of complex climatic,hydrological and topographical conditions.Constrained by accessibility,data availability,and financial resources,it is difficult to monitor soil moisture over the entire upstream of the Heihe River Watershed by in situ observations.While model simulations and satellite products can provide soil moisture data over large scales,their uncertainties are difficult to assess in the mountain areas such as the study watershed.Thus,it is necessary to determining the best scaling scheme of soil moisture in mountainous area based on limited in situ observations.Soil and vegetation are two of the most important factors influencing hydrological processes in the upstream of the Heihe River Watershed.This study aims to: 1)determine the best expansion methods of soil texture based on the field sampling data at 178 random sampling points,and generate the soil texture map;2)identify the main soil-vegetation types by combining the generated soil texture map,the GSCC(Genetic Soil Classification of China)and vegetation types;3)select the best soil moisture estimation methods in horizontal direction based on the sampling data at hill slope scale;4)choose the best soil moisture estimation methods in vertical direction based on the observations at the 32 in situ points;5)analyze the impacts of environmental factors on the estimation methods,and determine the best estimation schemes of the soil moisture distribution in the study area.The main conclusions are as follows.1.The predicted soil texture values by the OK(Ordinary Kriging)method is better performance in identifying the major soil type: silt loam.The MCRF(Markov Chain Random Field)method is better performance in identifying the minor soil types: loam and sandy loam.2.At the watershed scale,the OK method can only estimate the main soil texture type(silt loam),but not the minor types(loam and sandy loam).However,the MCRF method can accurately estimate 75% of the points from the minor types when all the sampling data have been applied in the calculations.Thus,as a classical linear method,the OK method is more suitable for mapping soil texture with little variation.The MCRF method shows advantage in mapping soil texture with strong spatial heterogeneity,which is more suitable in the study area.3.Expanding soil moisture in horizontal direction,both the IDW(Inverse distance weighting)and OK methods perform better than the RBF(Radial Basis Functions)method.Either in the different grids or at the different resolutions,all the three methods perform better when the soil moisture variations are smaller.4.The accuracy performances of Exp(Exponential filter),RBF(Radial Basis Functions)and LR(Liner Regression)are better from daily scale to hourly scale,while the tendencies of the three methods are different at different temporal scales.The performance of Exp at hourly scale is worse than its performance at the daily scale.The RBF and LR method,show little influence of the temporal scales.All the three methods perform best in sparse grassland,then dense grassland and shrub,with different performances in farmland and barren land.5.Estimaing soil moisture in vertical direction,linear regression(LR),Radial Basis Functions(RBF)and Exp methods show a declining performance order in estimation accuracy,and an increasing performance order in trend fitting.All the three methods,LR,RBF and Exp methods,show declining performance with increasing soil depths.For the estimation of soil moisture,the methods show same regular under different soilvegetation types.In horizontal direction,the IDW method is the best one and then is OK and RBF method at hill slope scale.Estimating soil moisture in vertical direction,the LR method is the best method in estimation accuracy with worst performance in trend fitting.The Exp method is the worst method estimation accuracy with best performance in trend fitting.The performance of RBF method is between them.