Following the aim of assessing the ecohydrological impacts of climate change in tropical subhumid watershed of Venezuela, we need to select some climate global models (CGMs). So, we will obtain several future daily climate data bases for running SWAT simulations under different "feasible and expected" conditions.
We have some basic questions before go ahead:
How to select the "better" GCMs?
One way appears to be comparing the CGM outputs with the records of historical reference time period (selecting those models showing the better evaluation stats).
But:
1) Considering the CGM outputs (raw, not downscaled) spatial resolution:
Comparing against historical local record data would be an admitted procedure?
2) If GCM "historic" outputs are downscaled ( i.e: bias correction by quantile mapping) using a historical data record, then these corrected CGM outputs will be compared with the same data record. Doesn't it imply a redundancy that leads to a obvious improvement?
3) There is no way of validating the future CGM simulation. Then: it really worth to select the "best" CGM according actual historic data records?. In other words: you may have a very good simulation for an actual historical record, but: will model perform a good simulation of the future?
Different points of view will be very valuable on our research. Thank you very much in advance.
Dear Oscar,
The way you proposed 'One way appears to be comparing the CGM outputs with the records of historical reference time period (selecting those models showing the better evaluation stats)' is recommended. As you know we can use several statistics including R^2, RMSE, Standard deviation for that.
I have some GCM's with me. If you send your email ID I am happy to share them with you.
Hello Miyuru Bandara Gunathilake . Thank you very much.
Then, do you think the better way is using raw data in the reference period? I mean, raw model outputs (with no downscaling).
For evaluation purposes,we apply a "ranking" method based in three model performance qualities (for monthly values): a) Accuracy (through MAE, RMSE and Willmott's Concordance Index), b) data distribution (mean, standard deviation, skewness) and c) consistency (R^2, slope and intercept).
I will mail you asking for the model you kindly offer to me.
Thanks again.
Yeah Professor,
I would prefer that.
I would say that we check the efficiency of our bias correction method through indices RMSE, mean and R^2.
That means we would evaluate the statistics of (1)raw historical GCM with historical observed and (2) bias corrected raw historical with historical observed.
through these indices.
If RMSE has decreased when we move from (1) to (2)
If R^2 has increased when we move from (1) to (2)
If the mean of historical observed and bias corrected raw historical has closer values We can say the selected GCM is a good representative for our study area.
My email is
I more than happy to share the GCMs which I extracted
Best
Miyuru
Thanks again Miyuru!
We will consider your very valuable suggestions and move forward over that approach.
This paper aims: to select representative GCMs for the river basin at short‐term (2016‐2045), mid‐century (2036‐2065) and end‐of‐century (2071‐2100) periods, and developed range of possible future precipitation and temperature scenarios
Article Representative General Circulation Models (GCMs) selection a...
- Badges
- Science topic
- Similar topics
- Correction