Daniel Senftleben (daniel.senftleben[@]dlr.de) and Nikolay Koldunov (nikolay.koldunov[@]awi.de)
The issue:
Interpolating gridded data from one spatial grid to another (from here on termed “regridding”) is necessary when two datasets on different grids are to be compared grid-point-wise. There are different regridding methods available, each involving different interpolation algorithms. Regridding sea ice concentration can introduce an error in the integrated sea ice area of up to 10% when compared to the sea ice area obtained without regridding. The error strongly depends on the selected algorithm and the grid type. Furthermore, regridding requires much computing time and can easily be the most expensive step in the analysis.
The scientific questions that will be studied:
-Where do the interpolation errors stem from?
-Can we find a “best practice” for regridding, as in developing a solution that works for all grid types?
-What is the most efficient way to perform regridding, to save as much cpu time as possible?
The type of analyses that will be conducted:
-Testing regridding with different grid types from CMIP5/CMIP6 models with different algorithms and estimating the respective error in each case
-Based on the above results, developing recommendations for the community on which method to use in different situations.
-Implementing several regridding methods into the ESMValTool to make it available to the scientific community
Update:
• Project is no longer ongoing.
• Based on limited analysis, the following recommendation is provided by the project leads:
– No best interpolation method was found across several models for September Arctic sea ice area
– For the sea ice fields, the approach that is taken by OMIP, to compute most of the characteristics on the original grid, is the best way to do sea ice analysis.