|
| 基于多源环境协变量的热带县域农田土壤pH空间制图及驱动因子识别 |
| Spatial Mapping of Farmland Soil pH and Identification of Driving Factors in a Tropical County Based on Multi-Source Environmental Covariates |
| 投稿时间：2025-08-18 修订日期：2025-09-12 |
| DOI： |
| 中文关键词: 热带农田 土壤pH 机器学习 递归特征消除 驱动因子 空间制图 |
| 英文关键词: tropical farmland soil pH machine learning recursive feature elimination driving factors spatial mapping |
| 基金项目:国家重点研发计划项目(2023YFD1900105) |
|
| 摘要点击次数: 646 |
| 全文下载次数: 0 |
| 中文摘要: |
| 【目的】为明确热带农区土壤pH的空间格局及驱动因子，【方法】基于海南省澄迈县的130个农田表层（0~20 cm）土壤样本，在融合地形、气候、遥感等45个环境协变量的基础上，集成特征选择策略和机器学习模型，开展土壤pH空间制图研究。【结果】基于递归特征消除方法选择的13个环境协变量构建的随机森林模型制图效果最优，相关系数为0.57。澄迈县酸性土壤主要分布在澄迈县西部和南部地区，其中酸性土壤（4.5 |
| 英文摘要: |
| In order to clarify the driving factors of soil pH in tropical agricultural areas. Based on 130 farmland surface (0-20 cm) soil samples from Chengmai County, Hainan Province, this study integrated 45 environmental covariates—including topography, climate, and remote sensing indices—with feature selection strategies and machine learning models to conduct spatial mapping of soil pH. The results indicate that the random forest model constructed using 13 environmental covariates selected through the recursive feature elimination (RFE) method achieved the best mapping performance, with a correlation coefficient of 0.57. In Chengmai County, acidic soils are predominantly distributed in the western and southern regions, with acidic soils (4.5 < pH ≤ 5.5) and slightly acidic soils (5.5 < pH ≤ 6.5) accounting for 28.24% and 71.76%, respectively. Topography (30.86%) and soil properties (30.09%) were identified as the most influential factors for soil pH spatial variability, followed by vegetation dynamics (15.09%), climate (14.55%), and soil management (9.40%). Our study indicates that farmland soils in Chengmai County generally exhibit low pH, with spatial patterns jointly determined by soil-forming factors and soil-forming conditions. The integration of recursive feature elimination with the random forest model provides a feasible approach for county-scale spatial mapping of tropical farmland soil pH, effectively enhancing variable selection efficiency and model predictive performance. The results provide a scientific reference for managing farmland acidification, optimizing agricultural resource use, and mitigating heavy metal pollution risks in tropical agricultural regions. |
| HTML
View Fulltext
查看/发表评论 下载PDF阅读器 |
