Hi Amin Mohebbi Tafreshi

In the following article, you will find that the authors use the Partial Least-squares Regression (PLR) method to downscale the GRACE total water storage change. The regression is carried out at a residual signal level obtained by removing the dominant seasonal signal and linear trend and by considering time lag/lead between various water budget components:

Article Downscaling GRACE total water storage change using partial l...

The Random Forest Model is also another useful method:

Article Downscaling of GRACE-Derived Groundwater Storage Based on th...

Hope this helps!

Best regards,

Ebrahim

Dear Dr. Ghaderpour

Thank you very much for your valuable answer.

Best regards

Hi Amin

It depends on the data you are working with. If the data volume is large for training, the artificial neural network will do better

Dear Dr. Memarian Sorkhabi

Thank you very much for your valuable answer.

Best regards