search

SantanderMetGroup / downscaleR

An R package for climate data bias correction and downscaling (part of the climate4R bundle)
https://github.com/SantanderMetGroup/climate4R
GNU General Public License v3.0
104 stars 59 forks source link

NA in all days of the observed record/Dimensions mismatch in predictions #59

Open bettolli opened 5 years ago

bettolli commented 5 years ago

Hi everyone! I’m performing a daily precipitation downscaling using different predictands sources: station data and gridded data. When I performed the downscaling with station data, everything went ok. But when I changed to the gridded data (MSWEP), two errors came out. 1) First error Error in glm.fit(x = numeric(0), y = numeric(0), weights = NULL, start = NULL, : object 'fit' not found In addition: There were 13 warnings (use warnings() to see them)

This error appears when downscale.cv is executed and it stopped at the first fold. I thought it could be a problem with the NA data in the ocean part of the domain (where all days in the record are NA). I tried then with a smaller continental domain and the error did not appear. However, a second error/warning came out

2) Warning message: In (function (..., deparse.level = 1) : number of columns of result is not a multiple of vector length (arg 2)

This warning appears when I tried to use both analogs and GLM with downscale.cv. And effectively, dimensions did not match between predictions and observations. For instance, observations are 5468x5x65 (time x lat x lon) and predictions are 6x 296400.

Thanks a lot! Maria Laura

jorgebanomedina commented 5 years ago

Hi Maria Laura!

Concerning to your first question, downscaleR admits regular grids as predictand but that was a recent modification so maybe you just need to update your downscaleR package from the devel branch.

I have checked your second question and you are right. Downscale.cv failed when regular grids as predictands were fed. I made a modification and now it should work. Again to implement this modification you will need to update from the devel branch.

Please tell me if it works when you try it and thank you for your useful feedback!

Regards,

Jorge

El 11/3/19 a las 15:24, bettolli escribió:

Hi everyone! I’m performing a daily precipitation downscaling using different predictands sources: station data and gridded data. When I performed the downscaling with station data, everything went ok. But when I changed to the gridded data (MSWEP), two errors came out.

  1. First error Error in glm.fit(x = numeric(0), y = numeric(0), weights = NULL, start = NULL, : object 'fit' not found In addition: There were 13 warnings (use warnings() to see them)

This error appears when downscale.cv is executed and it stopped at the first fold. I thought it could be a problem with the NA data in the ocean part of the domain (where all days in the record are NA). I tried then with a smaller continental domain and the error did not appear. However, a second error/warning came out

2.

Warning message: In (function (..., deparse.level = 1) : number of columns of result is not a multiple of vector length (arg 2)

This warning appears when I tried to use both analogs and GLM with downscale.cv. And effectively, dimensions did not match between predictions and observations. For instance, observations are 5468x5x65 (time x lat x lon) and predictions are 6x 296400.

Thanks a lot! Maria Laura

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/SantanderMetGroup/downscaleR/issues/59, or mute the thread https://github.com/notifications/unsubscribe-auth/ATg44QgCX3JECIO1QeMOvLXaWcHp8786ks5vVmcbgaJpZM4boudI.