Hello everyone,
I'm planning to show the differences between various statistical downscaling methods in determining the results of my hydrological model. And, I decided to construct another downscaling model employing Support Vector Regression (SVR) and carry out the downscaling operation by SVR.
However, I have no idea where to start realizing this purpose of mine.
I'd be so glad if you shared with me the info on how and where I can start learning about this process.
Best
Dear Ibrahim
I don't know if you exactly intend to use SVR or you want to include some machine learning in your downscaling analysis. If you have enough time, there is a brilliant course in coursera about machine learning, taught by Andrew Ng that includes the support vector machines as well. You can find loads of immensely useful material there and will definitely gain valuable insight on the subject. If you are just focused on the results, ok with using a pre-written code and don't care about the concept behind it, I have written a MatLab code that you can use for your purpose.
I really recommend using various machine learning techniques, it always ends up with exciting results.
Sina Masoumzadeh Sayyar Thanks a lot for your positive and constructive response. I'll definitely follow the machine learning courses. Plus, I'd be so glad if you could share your MATLAB codes with me.
Zülküf İbrahim Erkol There are some problems that I could not solve, although I have used many sources for statistical downscaling. I would like to obtain information about the flows of the next period within the scope of CMIP6 by making use of the meteorological observation data through the stations. To put it simply;
Is it enough for me to enter CMIP6 data as input to a model like svr, take station data as output, create a separate model for precipitation and temperatures, and then predict meteorological data by entering future GCM data? If I then examine the flow with these estimated values, will I be downscaling (correctly)? Or am I misunderstanding the issue?
Sefa Nur Yesilyurt that's not how SVR works out, unfortunately. Your observation data will be your target dataset and you'll have your own climate data from different climate circulation models. Then, you'll determine certain parameters in SVR model depending on your datasets and process the data statistically using SVR
Dear Zülküf İbrahim Erkol , is it possible for me to get more detailed information from you in our native language via e-mail? I am adding my own mail, and if you add your e-mail, there are some issues I would like to consult with you. [email protected]
Also thank you for your reply
- Badges
- Science method