search

ecmwf / ecpoint-calibrate

Interactive GUI (developed in Python) for calibration and conditional verification of numerical weather prediction model outputs.
GNU General Public License v3.0
21 stars 8 forks source link

Bias correction factor may be wrong #150

Closed EstiGascon closed 3 years ago

EstiGascon commented 3 years ago

ERA5_FE_2mminT In the example, it looks like the value of the bias correction factor on the top of the mapping functions could be wrong. For example, I think that the value on the left plot should be lower. Could you double-check the computation of the correction factor? Maybe it is something related to the computation depending on if it is accumulated (FER) or instantaneous variable (FE).

FatimaPillosu commented 3 years ago

We have a meeting today with Anirudha, we'll discuss this later. Thanks for pointing this out @EstiGascon

EstiGascon commented 3 years ago

Hi Anirudha,

We have been testing the version 0.23.0 with the correction of this issue, but it is still something wrong. We think that it could be due to some outliers in the FE which makes the bias correction value look "wrong". Just to be sure that it does not happen, we think that the calculation of the mean bias correction (so the black number on the top right of the mapping function) should be calculated only for the range of values plotted on the mapping function. For example, in the plots above, the mean bias correction would be calculated just with the FE between -30 and 30, but if we have any other value below or above that range (i.e. 52 or -40), then they are not used in the calculation of the bias correction factor.

Also, be sure that the calculation of the bias correction factor corresponds to the mean of the FE, being FE = OBS - FC.

This is quite urgent, so I would appreciate if you could have a look as soon as possible. Many thanks in advance!