TEMPO RAD netCDF dimensions for wavecal_params

[thomasd]

Hi, I get strange results when looking up the dimensions for the "wavecal_params" variable in the TEMPO L1B RAD product files using the netCDF library. The dimensions should be "(mirror_step, xtrack, wavecal_par)", e.g. 123 * 2048 * 1 in the V3 RAD files I have. However, when I look up these dimensions for the wavecal_params in band_540_740_nm using the netCDF library, the final dimension is wrongly reported as "ephemeris_time", with a size of 35. This happens when I use Python (tested in using netCDF versions 1.6.5 and 1.7.1) and when I use the C API directly.

Are you aware of this and/or do you have a solution?

Example session in python:

ds = nc.Dataset("./Downloads/TEMPO_RAD_L1_V03_20241001T185431Z_S007G05.nc")
ds["/band_540_740_nm/wavecal_params"]

Out[9]:
<class 'netCDF4._netCDF4.Variable'>
float32 wavecal_params(mirror_step, xtrack, ephemeris_time)
num_coefficients: 1
start_spectral_channel: 470
num_spectral_channels: 100
adjust_nominal_wavelength: 1
path = /band_540_740_nm
unlimited dimensions:
current shape = (131, 2048, 37)
filling on, default _FillValue of 9.969209968386869e+36 used

thanks!

Thomas

[ASDC - ghayescrepps]

Hi Thomas,
Thank you for your post. We have verified that this issue occurs with the netCDF4 library and have informed the TEMPO Science Team. At this time, our suggested workaround is to use the h5py library, in which this error has not been seen to occur. For example, "h5py.File(filename, 'r')" works in almost the same way as "nc.Dataset(filename, 'r')".

[thomasd]

Hi, thanks for forwarding the issue!

Note that a similar issue exists in band_290_490_nm: there the final dimension of wavecal_params is reported to be 'xtrack' instead of 'wavecal_par'. The quickest way to see it, is using the ncdump command line utility ("ncdump -h <L1B filename>").

[ASDC - ghayescrepps]

Thank you for letting us know of this second error; we will make note of it as well.