Hi Sander,
I am trying to bring multiple datasets into a similar format so that I can later compare partial and total columns of CO. The datasets I am working with are reported on different pressure levels and horizontal resolutions. For example, I would like to harmonize CO mixing ratios across pressure levels before comparison.
Methodologically, is this approach correct? For this purpose, I am rebinning, regridding, and collocating the datasets. So far, rebinning and regridding seem to work well. However, some datasets, such as those from EAC4, are not supported directly by HARP.
Could you please suggest a possible solution for handling EAC4 data with HARP? In addition, I would appreciate an example code that demonstrates how to use harpcollocate on two datasets (e.g., EAC4 and TROPOMI CO).
Thanks in advance for your guidance.
Best regards,
There is some basic support for reading CAMS model data in HARP, but there is no direct support for EAC4. The general approach to bringing external data into HARP format when this is not already supported in the HARP library is to create your own converter. Some sample code for this for satellite data is provided in Add additional support for more product files by producing similar product files with already-supported product files - #2 by sander.niemeijer, but a similar approach can be used to turn gridded data into HARP format.
However, the important point is that harpcollocate was written with ground based data in mind. It is more tailored towards comparing satellite data vs. ground based data and finding the collocation pairs between both datasets that match in space and time (i.e. overpasses).
When comparing global data such as model vs. satellite, satellite vs. satellite, or model vs. model, you generally won’t be using a measurement by measurement matchup.
One approach is to create regular grids at the same resolution where the data matches in time and space. If a quantity is long-lived and doesn’t move that much you can often simply compare daily grids (e.g. the daily L3 from satellite vs. a daily average of the model data). And you don’t need any harp collocation for this. But when you have to deal with diurnal variations you will have to (for polar orbiting satellites) align each satellite L3 grid cell with a model grid cell of the same ‘local time’, which means that you have to slice your different model steps during the day based on the satellite local overpass time. But that is currently not a type of operation that is supported in the harp toolset.
Another approach is to regrid your L3 data to the grid of the satellite. See Using HARP and CAMS AQ forecast profile as an apriori to derive TROPOMI NO2 tropospheric column - #2 by sander.niemeijer for some information on that.
Thank you for wonderful explanation and the great help !!! I will go through the links you indicated.