典型红壤区农田土壤酸化空间分异主导因子识别模型

According to the Status of the World's Soil Resources Main Report in 2015, soil acidity is a serious constraint to food production worldwide, soil degradation caused by soil acidification has become a global consensus. Guangdong province where is a typical red earths region of Southern China: agricultural soil acidification was significantly, with complex topography and geomorphology, the level of economic development in different regions is extremely unbalanced, so it is necessary to identify the dominant factors in the spatial heterogeneity of agricultural soil acidification. This study selected the 30 impact factors including soil, meteorology, vegetation, topography, hydrology, acidic precipitation, fertilization, land use, road and social economy. An optimal model for the impact factors of agricultural soil acidification building based on machine learning data mining and through cross validation to validate this model; Using model of importance analysis to select the important impact factors and to analyze the overall impact factors of agricultural soil acidification in the perspective of spatial pattern and process; Simulation and prediction of spatial distribution in agricultural soil acidification based on the model, analysis of spatial heterogeneity law and its dominant impact factors from different river basins and the regions in different economic development level base on further optimal modeling for spatial heterogeneity and redundancy analysis method, and distinguish between artificial and natural sources. The research can provide a new model building method for the study of soil acidification and provide an important reference for the prevention and control of soil acidification.

2015年世界土壤资源状况报告指出土壤酸化是世界粮食产量的重要制约因素, 土壤酸化是土壤退化的重要因素已在全球达成共识。南方典型红壤区——广东省农田土壤酸化显著，且地形、地貌复杂，各区域经济发展水平极不平衡，识别农田土壤酸化空间分异主导因子显得非常重要。本研究综合考虑土壤、气象、植被、地形、水文、酸沉降、施肥、土地利用、道路和社会经济10类要素的30个影响因子，基于机器学习数据挖掘方法构建农田土壤酸化影响因子的最优模型，并使用交叉验证方法加以验证与优化；使用模型重要性分析模块筛选主导影响因子，从格局与过程的角度分析农田土壤酸化的整体性主导影响因子；基于最优模型模拟预测农田土壤酸化空间分布，对空间分异部分进一步优化建模，并采用冗余分析方法，从不同流域与不同经济发展水平的角度分析其空间分异规律并识别其主导影响因子，通过主导因子区分人为源与自然源。本项目能为土壤酸化综合治理与分区防治提供重要的科学参考依据。