中山大学学报（自然科学版）（中英文）2025，Vol.64Issue(2)：33-41,9.DOI:10.13471/j.cnki.acta.snus.ZR20240201
基于机器学习的环境因子与土壤孔隙度模拟
Simulation of environmental factors and soil porosity based on Machine Learning:A case study of tobacco-growing soils in Baise,Guangxi
摘要
Abstract
Soil porosity is a crucial parameter in the study of soil physical quality,agriculture,and environmental protection.This research focuses on the 0-20 cm surface soil of tobacco planting in Baise,Guangxi Province.Four machine learning models were employed to simulate the predictive potential of six climatic factors,three topographic factors,and one soil attribute factor on soil porosity.The study also analyzed the magnitude and spatial distribution characteristics of porosity.The findings reveal that the Random Forest model is the most effective,achieving a mean porosity prediction value of 41.257%,the lowest root mean square error of 5.738,and the highest coefficient of determination of 0.648.The predicted results closely align with the measured values,indicating that the Random Forest model demonstrates strong generalization performance and effective predictive capabilities for simulating environmental factors and soil porosity.Meanwhile,results from Kriging interpolation indicate that the porosity values in Debao County and Jingxi City areas are generally low.This suggests potential land degradation issues,such as land slumping,compaction,and a reduction of soil organic carbon storage.These problems could be mitigated through restoration measures such as selective operation,the reasonable application of organic fertilizer,and deep plowing,which would help promote tobacco productivity in the study area.Overall,this study provides an effective method for predicting regional soil porosity and offers a valuable reference for understanding the characteristics of soil porosity in tobacco-growing regions across the country,as well as for developing land degradation management strategies.关键词
土壤孔隙度/气候因子/地形因子/土壤属性因子/随机森林模型Key words
soil porosity/climate factors/topographic factors/soil property factors/random forest model分类
农业科技引用本文复制引用
胡炎凤,邹天祥,梁志鹏,涂俊喜,周萌,沈文杰,张介棠,范东升,卢燕回..基于机器学习的环境因子与土壤孔隙度模拟[J].中山大学学报（自然科学版）（中英文）,2025,64(2):33-41,9.基金项目
国家自然科学基金-广东联合基金(U1911202) （U1911202）
国家重点研发计划(2022YFF0801201) （2022YFF0801201）
广东省科技计划项目(2020B1111370001) （2020B1111370001）
