HRRR producing drastically different values than PyAPS/GACOS about raider

[jlmaurer]

That specific fix is not but I can add it.

[dbekaert]

For HRRR suggest you walk through the model downloaded and make a vertical profile of PTe and compare it to profile of PTE of the other models. That migth also show if its in the model itself.

[bbuzz31]

Likely this is due to a problem in HRRR itself, as this date looks relatively okay in ERA5 from RAiDER

@jlmaurer @dbekaert any thoughts?

[jlmaurer]

@bbuzz31 interesting. Just to clarify, this is happening only for a select few datetimes or consistently at many datetimes?

[piyushrpt]

#501 (comment)

I would try and make sure that lla -> model via lla on a sphere vs EPSG:4326 produces same results. If they don't it is possible that coordinates are being mapped incorrectly. Should test just the coordinate transforms with sample points around the US

[bbuzz31]

@jlmaurer just about 3 or 4 of the 150 i checked

Thanks @piyushrpt. @jlmaurer i belive this is part of #507 ?

[dbekaert]

as this date looks relatively okay in ERA5 from RAiDER

What do you quantify as "relative ok". The plots for WRA5 shows dont look the same for ERA5 in PyAPS verus Raider. Is one a residual plot?

[bbuzz31]

as this date looks relatively okay in ERA5 from RAiDER

What do you quantify as "relative ok". The plots for WRA5 shows dont look the same for ERA5 in PyAPS verus Raider. Is one a residual plot?

The difference between PyAPS (ERA5) and RAiDER (ERA5) is okay relative to HRRR (RAiDER) on that 20181103, suggesting that there is not something fundamentally wrong happening in RAiDER. The plots in the original comment vs the second comment have different colorbars, sorry for the confusion.

[dbekaert]

Im not in sync here. "Okay" is not a good quantification its very subjective. ERA 5 is the same model input and you show one to have a range 0.4 cm and the other 0.1 range. Independent of comparing those difference between ERA5 and HRRR. Why are the difference between ERA5 itself so large...

[bbuzz31]

Closed with #566