Changes to enable dust and aerosols in sea ice will be tested using the seaice/develop testsuite.  The current aerosol test suite uses only a single snow layer and snicar-ad is not enabled.  Both these features will be activated.  A short test run should bases parallelism and restartability tests.  We expect that regression tests for the shortwave test suite will fail.  All other tests should pass.  This ensures that the changes are BFB with the aerosol scheme off.

Summary

The seaice/develop testsuites consists of five namelist and streams configurations:  aerosol_shortwave, prescribed_ice, snow_tracer_physics, standard, and standard_bgc.  Each configuration for our development branch was run as a regression test against master, a parallelism test, and a restartability test.  The standard configuration uses the current defaults for water cycle, and so all tests for this configuration are required to pass in a BFB development feature.  The standard_bgc also has aerosols, but does not include shortwave feedbacks.  The aerosol_shortwave test did not include multiple layers or snicar-ad.  We modified the test so that both features are exercised. All tests, except the regression tests for aerosol_shortwave and the standard_bgc, passed.  This latter tests failed because of bug fixes to the code discussed below.

In order to enable the aerosols and dust in a multi-layer snow model and with snicar-AD active, several code changes were necessary.  The first was a correction to the snow surface scattering layer which has a maximum value defined as a function of the number of grid layers.  There are several places in the code where this dependency was not included properly and in one instance the aerosol snow grid upper layer was defined incorrectly.  Another problem was with the implementation of atmosphere-snow-on-sea-ice fluxes.  The aerosols in snow scheme uses the same grid as the shortwave delta-Eddington routine.  Thus the upper-most layer is the surface scattering layer which can get very small for thin snow accumulations over thick sea ice.  If all deposition fluxes are deposited into the upper layer regardless of thickness, the aerosol/dust concentration can get very large. This is not a problem for delta-Eddington, but it is for snicar-ad.  The problem is further exacerbated with the melt-water high scavenging of BC/dust suggested by observations and controlled in the model by the kscavz parameter.  We corrected this problem by using the minimum snow thickness as a condition for deposition into the snow rather than the sea ice, and by introducing a length scale for the spreading of atmospheric fluxes into the snow interior. 

Unit Testing 

Unit Testing: standard testsuite

Date last modified:02/23/2021

Contributors:Nicole Jeffery

Provenance: njeffery:seaice/njeffery_aerosolConcFix

Results:

Testing MPAS-Seaice
===================
Test suite: /lcrc/project/ACME/jeffery/MPAS-seaice/CASES/20210219_aerosolFix/MPAS-Model/testing_and_setup/seaice/testing/testsuites/testsuite.all.xml

Domains directory: /lcrc/project/ACME/jeffery/DATA/E3SM/RegionFiles

Using configuration standard_physics and domain domain_QU120km...

Test: Regression, Configuration: standard_physics, Domain: domain_QU120km
=========================================================================
No non-contents errors
PASS: Test regression passed

Test: Parallelism, Configuration: standard_physics, Domain: domain_QU120km
==========================================================================
multipleBlocks: False
No non-contents errors
PASS: Test parallelism passed

Test: Restartability, Configuration: standard_physics, Domain: domain_QU120km
=============================================================================
No non-contents errors
PASS: Test restartability passed

Verification Test

Verification Test : aerosol_shortwave test

Date last modified: 02/23/2021

Contributors: Nicole Jeffery

Provenance:njeffery:seaice/njeffery_aerosolConcFix

Results: