Reproducing Global Mean Error for tas NRMSE in Random Forest Model

[ShihengDuan]

Hi,

I appreciate the effort and resources you've

made available through the open-sourced scripts and data. Currently, I'm reproducing the tas NRMSE values as reported in the manuscript. However, I've encountered difficulty in achieving similar results specifically for the random forest model.

While the spatial error metrics appear reasonable with values similar to those reported, the global mean error shows large differences compared to the second column of Table 2 in the JAMES manuscript (I'm getting 0.16). I have attached the results for your reference, where you'll observe that the time series of global mean tas exhibits a substantial error.

Could you kindly assist in investigating this matter? Thanks!

randomforest.pdf