As discussed last week with @jbusecke, I put together a plot that shows a snapshot of the small scale fields (so far just for ql) across all algos and both datasets (CM2.6 and CESM). Below is the plot (note coare3p0 is not included here yet).
Note: the colorbar scale here is -10 to 10
The first thing to notice is how similar these fields look across all algorithms! This is in contrast to what we saw previously with our "first results". @jbusecke and I have discussed this a bit and will need to figure out where the error occurred (in the previous computations or in the ones I show here).
To verify that the fields I've plotted are indeed different, I show differences between a couple algos for the CM2.6 data:
Difference between NCAR and ecmwf
Difference between ecmwf without skin and with skin
So there are difference between these plots up to a bit more than 10%!
For reference, the code I used to generate these plots can be found in this notebook.
As discussed last week with @jbusecke, I put together a plot that shows a snapshot of the small scale fields (so far just for
ql) across all algos and both datasets (CM2.6 and CESM). Below is the plot (notecoare3p0is not included here yet).Note: the colorbar scale here is -10 to 10
The first thing to notice is how similar these fields look across all algorithms! This is in contrast to what we saw previously with our "first results". @jbusecke and I have discussed this a bit and will need to figure out where the error occurred (in the previous computations or in the ones I show here).
To verify that the fields I've plotted are indeed different, I show differences between a couple algos for the CM2.6 data:
Difference between
NCARandecmwfDifference between
ecmwfwithout skin and with skinSo there are difference between these plots up to a bit more than 10%!
For reference, the code I used to generate these plots can be found in this notebook.