Hi,
I am currently working on calibrating CTSM using multiple observation sites. I have around 12 observation sites, and I would like to run these sites simultaneously within a single CTSM case rather than creating and compiling separate cases for each site. But there is no method for generating multiple mesh files with non-spatially contiguous grid points.
My understanding is that CTSM does not require the simulated grid cells to be spatially continuous, and a possible approach is to create a user-defined regional grid containing all observation sites as independent grid cells.
The workflow I am considering is:
I am currently working on calibrating CTSM using multiple observation sites. I have around 12 observation sites, and I would like to run these sites simultaneously within a single CTSM case rather than creating and compiling separate cases for each site. But there is no method for generating multiple mesh files with non-spatially contiguous grid points.
My understanding is that CTSM does not require the simulated grid cells to be spatially continuous, and a possible approach is to create a user-defined regional grid containing all observation sites as independent grid cells.
The workflow I am considering is:
- Create an ESMF grid file with 12 grid points, where each observation site corresponds to one grid cell, and assign the latitude and longitude of each observation site to the respective grid cell.
- subset 12 single point surface dataset from my regional surfdata by subset tool, combine these surfdata contain 12 grids at each observation.
- Create and run a simulation case using the combined 12-site mesh file and surfdata file. The site data can then be validated against the corresponding simulation output along the matching grid dimension.
- Is it possible and recommended to run multiple independent single-point sites as multiple grid cells within a single CTSM case?
- Is there an existing tool or script to generate an ESMF mesh/domain file for a collection of irregularly distributed sites?
- For this application, is it better to manually construct the mesh/domain files, or should I create multiple single-point cases and run them separately?
