基于特征筛选与随机森林的土壤有机质空间预测

打开文本图片集

中图分类号：S153.621;X144;TP183文献标志码：A文章编号：1672-2043（2025）11-2864-11doi：10.11654/jaes.2025-0557

Abstract：Inodertovestigatetectsofarosenviroetalvarblesotespatialdistributionofsoiloganicmaterprdictedby therandomforestodel，thisstudyhosevariousenvironmentalvaribletypesforcombinationndoptimzation.Thegoalwastoiize thedtrimentalefectsofredundantcharacteristicvariablsontheandoforestmodel.Inordertobuildaandomforestpredictionodel ofsoilorganicmaterwithariousombinationsofviroentalvariables，opograpicfactors，cliatefactoegeationfctosoil properties，andanthropogenicfactorswerechosenandprioritized.Spearmancoelationaalysisndimportancenalyssereusedto choosethebestsetofenvironmentalfactors.Thefindingsdemonstratedthattherandomforestpredictionmodelwithanthropogenic influencesandsoilcharacteristicsasinputvariablesproducedsuperioroutcomes，withrotmeansquareerorandcoeficientof determination of 4.387 g⋅kg-1 and O.802，respectively; With an coefficient of determination of O.747，the prediction accuracy was lowest whenclimatefactorwereemployedasindependent inputs;Therandomforestpredictionmodel'sfindingswereoptimizedwithrootmean square error and coefficient of determination of 2.785 g⋅kg-1 and O.911， respectively，following feature screening to eliminate redundant features;Topogaphastepriarydeternatofsologaicmaterineseacheaccdingtoefiningofeaod significaceanalysesTherandomforestmodeloutperfostheonventioalodinaryrigingregresionrigingandgeoaicaly weightedegreionkrgingmodelsintesofprdictionaccracy.eacracyofteandomforestpredictiomodelabefctiely increasedbyractesticfldiotalbles.Ungoyvatio，oesoilttealndnalii with fewer variables，the spatial predictionaccuracyof soil organic mattercontentinthe studiedarea exceeded 0.8. Keywords：soil organic matter; characterization screening; random forest; spatial prediction

土壤有机质（SoilOrganicMatter，SOM）是土壤有机碳的重要载体，是构成土壤固相物质的关键部分。（剩余19725字）