Snow depth variable name - missing from constants

[zakwatts]

When training PVNet, the snow depth variable is specified in the config. But PVNet fails to train. This is due to the snow depth variable name sd not being in the constants.py file.

For the old ECMWF and Met Office datasets (no extending past 48 hours) the snow depth variable we pulled was called sde.

From the new extended NWP datasets, ECMWF snow depth variable is sd and Met Offices is sde.

Need to research the difference between sd and sde.

Will require a new constant to be added to the constants.py file to be able to train with snow depth.

[zakwatts]

From a previous issues, appears we receive sd in live from the met office. (as apposed to sde in the backtest)

From @devsjc comment

sd is from the met office live data, it's the Snow Depth Water Equivalent in and it's measured in kgm-2
sde is from ceda, also snow depth water equivalent, but this one is measured in m

They differ by a factor of ~1000

openclimatefix/MetOfficeDataHub#71

[zakwatts]

In production both ECMWF and Met Office are sde. Based off internal dashboard for ECMWF and Met Office.

[zakwatts]

sde in the old ECMWF and sd in the new extended ECMWF appear to be the same.

Also worth noting that the extended ECMWF has half the spatial resolution of the old ECMWF.