In order to increase the prediction accuracy of digital mapping model of soil properties,accompanied by the increase in data quantity especially remote sensing environmental variables,the improvement of computing power,and the popularization of deep learning frameworks,the digital mapping model of soil propertiesis transitioning from traditional knowledge-driven models to data-driven artificial intelligent deep learning models.This article takes the key property of soil organic carbon as an example to analyze and summarize the theoretical basis,model structure,integration of relevant environmental variable spatial context information and multimodality data that urgently needs to be solved,and interpretability of deep learning models.The aim is to promote the application of artificial intelligent deep learning models in soil properties digital mapping of the Third National Soil Survey.
digital soil mappingdeep learningneural networksoil propertiessoil organic carbon
国家科技期刊平台
NSTL
万方数据
