采樣密度對土壤全氮隨機森林模擬精度的響應

打開文本圖片集

中圖分類號：S158 文獻標志碼：A 文章編號：1672-1098（2025）02-0042-09

引文格式：，等.采樣密度對土壤全氮隨機森林模擬精度的響應[J].安徽理工大學學報（自然科學版），2025，45（2） ：42-50.

Impact of Sampling Density on the Spatial Prediction Accuracy of Soil Total Nitrogen by Using Random Forest LI Xiaopeng1 ，ZHANG Shiwen' ，LIU Xiaoxue2，YAN Fang2 KONG Chenchen1 ，JIAO Yangqing' ，ZHAO Baoyu

（1.SchoolofEarthandEnvironmnt，AnhuiUniversityofSienceandTechnology，HuainanAnhui220o1，China;2.FengtaiAgriculturalProductsQualityandSafetyIspectioStation，BeijingOoo，China;3.DepartmentofFarmlandInformationMaagee， Beijing Cultivated Land Construction and Protection Center，Beijing 1OOo2O，China） Abstract：Objective A reasonable sampling densityfor regional soil total nitrogen（STN） investigation enables accurate assessment of STN content dynamics whileoptimizing resource eficiency. Understandingsampling density on spatial simulation accuracy is therefore crucial.Methods Byutilizing soil sampledata from Fangshan District，Beijing，optimal sampling numbers were calculated and diferent density gradients were established.Topographicand vegetation variables were incorporatedas auxiliary factors in random forest modeling to predictSTNspatial distribution and evaluate sampling density impacts on simulation accuracy.Results Sampling density determination solely by Cochran's formula （neglecting spatial heterogeneity）resulted in low STN predictionaccuracy.Semivariogram analysis revealed moderate spatial autocorrelation of STN. Spatial distribution exhibited a west-high-eastlow patern，aligning with elevation trends. Increased sampling density significantly enhanced random forest accuracy until reaching 37O samples，beyond which accuracy stabilized. Maximum R2（0.82） and minimum RMSE （204號 （0.15g/kg） ） occurred at 497 samples.Conclusion The optimal sampling density for STN prediction in Fangshan District ranges between 222 and 37Osamples，balancing accuracy and cost-effctiveness.This finding providesguidance for regional soil surveys.

Key Words ： soil total nitrogen ;sampling density ;spatial interpolation;random forest ;interpolation accuracy

土壤全氮（Soil total nitrogen，STN）是決定土壤質量的主要因素，也是衡量土壤肥力的重要指標，由于人為活動[1]、成土因素[2]和地形因子[3]等環(huán)境因子的影響，STN通常表現出顯著的變異，準確估計STN的空間分布可為農業(yè)管理提供理論支持和指導[4]。（剩余7927字）