What problem does your feature request solve?
To compare model rainfall fields with Nimrod radar observations, the model data should be on the same grid as the observations. Then a filtering operation is required to ensure that only model data points corresponding to good radar observations are considered for comparison.
Describe the solution you'd like
As shown in the images below, model rainfall fields (left) have different domains and gridding to the Nimrod radar observations (right).
A further complication is that there are multiple Nimrod radar grids e.g. Nimrod_xkm (as shown above), Nimrod_1km and Nimrod 2km. Each if these products has a corresponding weights field that records how many 5 minute instantaneous value of radar rainfall rate were used to integrate up an hourly rainfall estimate. 13 of these 5 minute data are required for a complete integration, but hourly values constructed from 12 or 11 values are accepted. The weights fields can be plotted to show where hourly data are reliable:
The solution proposed is to first regrid the mode data to fit the Nimrod domain. Then a filtering operation should be employed to ensure that only the model data coincident with weights>10 are used to compare against radar observations.
Describe alternatives you've considered
An alternative would be to regrid the Nimrod observations onto the model grid. But best practice is to transform model output to compare against observations. Manipulating observations to resemble model data is discouraged.
What problem does your feature request solve?
To compare model rainfall fields with Nimrod radar observations, the model data should be on the same grid as the observations. Then a filtering operation is required to ensure that only model data points corresponding to good radar observations are considered for comparison.
Describe the solution you'd like
As shown in the images below, model rainfall fields (left) have different domains and gridding to the Nimrod radar observations (right).
A further complication is that there are multiple Nimrod radar grids e.g. Nimrod_xkm (as shown above), Nimrod_1km and Nimrod 2km. Each if these products has a corresponding weights field that records how many 5 minute instantaneous value of radar rainfall rate were used to integrate up an hourly rainfall estimate. 13 of these 5 minute data are required for a complete integration, but hourly values constructed from 12 or 11 values are accepted. The weights fields can be plotted to show where hourly data are reliable:
The solution proposed is to first regrid the mode data to fit the Nimrod domain. Then a filtering operation should be employed to ensure that only the model data coincident with weights>10 are used to compare against radar observations.
Describe alternatives you've considered
An alternative would be to regrid the Nimrod observations onto the model grid. But best practice is to transform model output to compare against observations. Manipulating observations to resemble model data is discouraged.