NEDAS QG model benchmarks

"truth" directory contains the QG model state from the nature run.

"prior" directory contains the ensemble prior states for the 400 members.

"code" directory contains:

"run.sh" that is submitted to supercomputer job queue to perform the analysis

jupyter notebooks contain steps that created the figures

"default" is a source file for environment variables

"benchmark" directory contains data output from the analysis, ${assim_mode}/n${nens}/dx${grid_dx}_dz${grid_dz}/no${nobs}/roi_${hroi}_${vroi}/np${nproc}/try$i/ are directories containing the output from a case with specified parameters:

assim_mode = batch or serial

nens is the ensemble size

grid_dx and grid_dz determine the model grid size

nobs is the number of observations

hroi and vroi are the horizontal and vertical localization distances

nproc is the number of processors

try$i is the i-th random realization for the same setting (5 tries are stored in this dataset)

The NEDAS main branch (commit 27da8d81656ae7f2fa6cdfba28838e89d20eda90) is the version that performed this benchmarking.