Based on a dataset of δ15N in soil from 812 sampling sites worldwide and environmental factors such as meteorology,soil,terrain and vegetation,a total model and sub model of the relationship betweenδ15N in soil and environmental factors were constructed using random forest regression algorithm,and the opti-mal model was identified through accuracy comparison.The spatial distribution of δ15N in soil from major terres-trial natural ecosystems worldwide with a spatial resolution of 0.25° was mapped based on the optimal model and environmental factor data.The results indicated that the accuracy of the total model was RMSE=1.63‰ and R2=0.61.The sub model based on normalized difference vegetation index(NDVI)classification had the high-est accuracy and was the optimal model,with an average accuracy of RMSE=1.19‰ and R2=0.69.The aver-age of δ15N in soil from major terrestrial natural ecosystems worldwide was 4.88‰,with a variation range of-0.49‰～12.29‰ and a standard deviation of 1.74‰.
维普
万方数据