This changes our ErrorCovariance class to be a wrapper around BUMP. This way, we can pass it an atlas::Field for the horizontal correlation lengths, instead of using a fixed value for the entire globe. umdsstCovar should be used for staticbinit now. After bump has been initialized umdsstCovar and BUMP will behave the same, so it doesn't matter which one is used for dirac and var. I have left those two using BUMP.
Actually filling in the atlas::Field with a horizontally varying field will be in a different PR, it is using the same 3000km globally that was being used before.
closes #57
a couple other small changes I found along the way
the BUMP files are placed in their own directory and are deleted every time before staticbinit is run. This way we never have to worry about manually deleting them again.
the ErrorCovariance test was not setup correctly, It uses BUMP now, and runs after staticbinit is run.
the reference answers for staticbinit have changed. I have no idea why. The answers for dirac and var have not changed so it apparently doesn't matter.
This changes our
ErrorCovarianceclass to be a wrapper around BUMP. This way, we can pass it anatlas::Fieldfor the horizontal correlation lengths, instead of using a fixed value for the entire globe.umdsstCovarshould be used forstaticbinitnow. After bump has been initializedumdsstCovarandBUMPwill behave the same, so it doesn't matter which one is used fordiracandvar. I have left those two usingBUMP.Actually filling in the
atlas::Fieldwith a horizontally varying field will be in a different PR, it is using the same 3000km globally that was being used before.closes #57
a couple other small changes I found along the way
staticbinitis run. This way we never have to worry about manually deleting them again.ErrorCovariancetest was not setup correctly, It uses BUMP now, and runs afterstaticbinitis run.staticbinithave changed. I have no idea why. The answers fordiracandvarhave not changed so it apparently doesn't matter.