What to do when margin is colder than cloud top?

[gerritholl]

In fogpy.algorithms.LowCloudHeightAlgorithm.procedure:

https://github.com/gerritholl/fogpy/blob/2c9aa9436e4924f55879d52a9de14817d370df5a/fogpy/algorithms.py#L648-L651

When zcenter==0 and (zneigh==0).all(), this causes zmargin==[0.0, 0.0, 0.0], delta_z==0, which means that apply_lapse_rate returns nan. Not sure what to do about it.

This causes the RuntimeWarning: Mean of empty slice warning referred to in #9.

[gerritholl]

This is covered by:

https://github.com/gerritholl/fogpy/blob/2c9aa9436e4924f55879d52a9de14817d370df5a/fogpy/algorithms.py#L689-L691

So the implementation does appear to consider this, which means the warning must have been there before.

[gerritholl]

Actually, it rather seems to happen when the cloud top temperature is higher (warmer) than the centre pixel.

[gerritholl]

Example of an extreme case:

ipdb> p tcenter, tmargin
(256.846, [249.6363525390625, 253.1262664794922])

This would place the cloud top height at -1335 metre and -689 metre, which is why it gets set to nan:

https://github.com/gerritholl/fogpy/blob/2c9aa9436e4924f55879d52a9de14817d370df5a/fogpy/algorithms.py#L909-L917

And then np.nanmean([nan, nan]) gives the RuntimeWarning.

Now why is the cloud top temperature here so warm?

[gerritholl]

So, in summary:

1. Channel 108 measures higher temperature at cloud top than at first pixel outside cloud
2. Fogpy calculates negative cloud top height (in the case where topography does not allow cloud top height calculation)
3. Fogpy then sets cloud top height to nan
4. Those nans later cause a RuntimeWarning in nanmean