摘要
Abstract
To reconstruct the missing remote sensing soil moisture information,multiple sets of data models were simulated and constructed using different types of methods—including statistical spatial interpolation(ordinary Kriging),linear regression models(multiple linear regression),and machine learning models(artificial neural networks,random forest)—based on data such as meteorology,topography,vegetation,and soil.Combined with other soil moisture products,the impact of different methods on reconstruction accuracy was evaluated in terms of statistical characteristics and spatial distribution patterns.The results show that ordinary Kriging has significant advantages when data quality is high and the missing data ratio is low;however,when the missing data ratio is high and the spatial distribution of missing data is uneven,multiple linear regression and machine learning models achieve higher accuracy.关键词
土壤湿度/插值法/神经网络/线性回归Key words
soil moisture/interpolation method/neural network/linear regression分类
水利科学
