I am working with temperature, precipitation data (0.1 degree from CDS store) and population density data (2.5 min from SEDAC) and often have to change the resolution of either data to achieve a common resolution. I do this using cdo but often get confused on which interpolation method would be more appropriate or give accurate results for a particular variable. And does the choice of method also depends on whether we are interpolating from coarser to finer grid or vice-versa? What method should be used for population data? Can you guide me on this or provide some good references?
cdo remapbil,target_file_name temp.nc outfile.nc cdo remapcon,target_file_name prec.nc outfile.nc
Hello,
In my opinion, if you intend to continuously change the resolution of the main dataset from which you obtain precipitation and temperature data, it is better to first create a fishnet in ArcGIS for each scale that corresponds to the boundary of the area of interest. For example, you can create 1 km, 4 km, and other fishnets. After that, based on the points created for each spatial scale, you can extract your data from the Google Earth Engine platform. Then, you can perform the interpolation operations.
Regarding the second issue, if your goal is to compare interpolation methods, it is better to evaluate errors such as Nugget, RMSE, ME, R², and correlation for each method to understand which method has the least error during interpolation.
In general, reducing the spatial scale inherently involves errors, and this is a fact.
You can uses remapdis o remapbil for regular grids. The interpolation for both methods are very same. You can analized performance compared reanalysis with local meteorological data. The metrics are used common RSME, Bias, Taylor Diagram.
Best regards Axel
- Badges
- Science topic