The hydrological and soil nature is a significant parameter in the study of hydrology, ecology, environment and agriculture. Its spatial distribution is of great significance to the revelation and simulation of the migration and transformation mechanism of substances and energies in the earth critical zone. As the large-scale and concentrated sampling, analysis, supervision and inspection of hydrological and soil nature consumes too much time and energy, the prediction of its spatial distribution is often carried out by the traditional spatial interpolation method, such as Ordinary Kriging, Regression Kriging and Random Forest. However, during the interpolation, the uncertainty of the model structure, target variable and auxiliary variable greatly impacts the accuracy of the interpolation. Therefore, the contribution rate of three uncertain sources of quantitative analysis to the spatial interpolation deviation of the hydrological and soil nature, and its removal are keys to improving the interpolation accuracy.

Zhu Qing Research Group systematically and deeply analyzes different uncertainty sources of the spatial interpolation of the hydrological and soil nature and their contribution rates by selecting key hydrological and soil natures such as the soil water contents, temporal stability of soil moisture and cation exchange capacity and auxiliary variables such as the detection of earth conductivity meter, remote sensing image, soil texture and topographic index. It is found in the research that the uncertainties of auxiliary variables (such as the topographic index of different spatial resolution rates and the accuracy of soil texture) have greatest impacts on the interpolation of key hydrological and soil nature parameters, while the uncertainty of model structure (such as different models and model parameters) has a less impact, and the uncertainty of target variable (such as the sampling density and sample sizes) has the least impact. Meanwhile, the impact of three uncertainty sources on the accuracy of interpolation is affected by land-using and the alternation of drying and wetting. For example, the impact of model structures on the accuracy of the interpolation is greater in dry seasons while it decreases greatly in wetting seasons. The impact of auxiliary variables on the accuracy of interpolation is greater in forests while it is smaller in tea gardens which have been affected greatly by human activities. Therefore, to improve the accuracy of the spatial interpolation of hydrological and soil natures, the first step is to reduce the uncertainty of auxiliary variables. For example, obtain topographical parameters by using the DEM with higher resolution ratio and increase the detection density of earth conductivity meter; additionally, optimize selected models and other parameter sets, for example, adopt an advanced machine-based learning model and the method of collective forecast based on multiple model parameters.