Digital mapping of soil total nitrogen in Wangyedian experimental forest farm based on random forest model
To explore the spatial distribution characteristics of soil total nitrogen content in forest farm and its response to environmental factors,random forest model and Cubist model were used to establish a quantitative relationship between soil total nitrogen content and environmental covariates including eleva-tion,normalized difference vegetation index,mean annual precipitation,mean annual temperature,y-coordi-nate and aspect at soil depths of 0-10,10-30 and 30-50 cm in Wangyedian experimental forest farm.Soil to-tal nitrogen content of the area was predicted and mapped,and the controlling factors affecting the spatial variation of soil total nitrogen were analyzed.The results showed that the average content of soil total nitro-gen at soil depth of 0-10,10-30 and 30-50 cm was 3.20,2.02 and 1.47 g/kg,respectively.It decreased with the increase of soil depth.The results of cross-validation showed that the R2 of the random forest mod-el for predicting soil total nitrogen at the three soil depths was 0.59,0.42,and 0.39,respectively,better than the R2 of Cubist model with 0.56,0.38,and 0.34,respectively.The prediction accuracy of both models decreased with the increase of soil depth.The influence of various environmental factors on the spatial distri-bution of soil total nitrogen decreased with the increase of soil depth.From the prediction map of soil total nitrogen with the random forest model,the content of soil total nitrogen at different soil depth showed a spa-tial pattern of low in the western,northern,and central regions,and high in the southwestern,southeastern,and eastern regions.The uncertainty map showed that the random forest model had a low standard devia-tion in predicting the distribution of the content of soil total nitrogen.The elevation had the greatest impact on the content of soil total nitrogen,followed by the normalized difference vegetation index>mean annual precipitation>mean annual temperature>y-coordinate>aspect.It is indicated that the random forest model can serve as an effective method for predicting the content of soil total nitrogen at different soil depth in the forest farm.
soil total nitrogenrandom forest modelspatial distributionsoil depthdigital soil map-ping
NETL
NSTL
万方数据
