I am developing a model using soil moisture to predict missing precipitation.
I correlated soil moisture storage at location A with precipitation at the same location A.
I also correlated precipitation at location A with precipitation at location B. Surprisingly the soil moisture A - precipitation A correlation has higher R and higher RMSE (0.94 and 18.7 mm) whiles precipitation A -precipitation B correlation has lower R and lower RMSE (0.87, 12 mm).
Please, is it possible to get a high R-value with high RMSE and also low R but low RMSE using the same sample size? What could be the reason?
Yes. It is possible. When calibrating you model, calculate average value of two objective functions (RMSE and R). Try to get the maximum A=average (R+RMSE). Or, you can regard RMSE, R and NSE (add this objective function) as three points. Each calibration procedure refers to a specific three-dimensional objective space. Please refer to this article: Evaluation of the impacts of hydrologic model calibration methods on predictability of ecologically-relevant hydrologic indices
Thanks Yanfeng Wu. If I got you correct, is a high maximum A=average (R+RMSE) an indicator of a good model?
Yes, you can try to calibrate it. I read a paper talk about average value of two indicators. Garcia et al. (2018) showed that the Kling and Gupta efficiency (KGE) applied to a transformation of discharge is inadequate to fulfil all assessment
criteria, whereas the mean of the KGE applied to the discharge and the KGE applied to the inverse of the discharge is sufficient. This is, using average KGE and KGE(1/Q) also can get a good result. Please refer to [Which objective function to calibrate rainfall–runoff models for low-flow index simulations?]
- Badges
- Science topic
- Similar topics
- Correction