I propose that we take advantage of GSI's Single_Observation utility for GSI vs JEDI assimilation validation experiments. In addition, we can build a framework around this tool to better facilitate assimilation validation experiment between all observation platforms that can be shared among developers and reduce the effort required to perform such tests.
I just do not know where such framework should be housed if at all in the RDASApp. Suggestions are most welcome!
Basic Idea:
Using GSI's Single_Observation utility, create a realistic single bufr observation of any ob platform (not just ADPUPA). This is intentionally different from using GSI's single observation test namelist capability since that doesn't actually use an observation, but uses user provided innovation etc. The point is to have a single bufr observation that can be used by GSI and then also run through a bufr2ioda convert to have an equivalent single ioda observation that can be used in JEDI. Then we can assimilate the observations and compare between GSI, FV3-JEDI, and MPAS-JEDI once those test cases are in place. Finally, this capability should include a plotting utility to easily make comparisons for validation.
I propose that we take advantage of GSI's Single_Observation utility for GSI vs JEDI assimilation validation experiments. In addition, we can build a framework around this tool to better facilitate assimilation validation experiment between all observation platforms that can be shared among developers and reduce the effort required to perform such tests.
I just do not know where such framework should be housed if at all in the RDASApp. Suggestions are most welcome!
Basic Idea:
Using GSI's Single_Observation utility, create a realistic single bufr observation of any ob platform (not just ADPUPA). This is intentionally different from using GSI's single observation test namelist capability since that doesn't actually use an observation, but uses user provided innovation etc. The point is to have a single bufr observation that can be used by GSI and then also run through a bufr2ioda convert to have an equivalent single ioda observation that can be used in JEDI. Then we can assimilate the observations and compare between GSI, FV3-JEDI, and MPAS-JEDI once those test cases are in place. Finally, this capability should include a plotting utility to easily make comparisons for validation.