klein21
commented
6 years ago Dustin,
For these changes it is most important that the LIDAR/MODIS developers approve them. If those developers, Helene Chepfer and Robert Pincus respectively, approve of the modifications, then I approve merging this to the trunk.
One small comment, in your first point below I believe you mean to say “do not impact any of the LIDAR diagnostics”, not “MODIS” diagnostics, right?
Steve
From: dustinswales notifications@github.com Reply-To: "CFMIP/COSPv2.0" reply@reply.github.com Date: Thursday, March 22, 2018 at 2:17 PM To: "CFMIP/COSPv2.0" COSPv2.0@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [CFMIP/COSPv2.0] Changes (minor) to MODIS and LIDAR simulators. (#9)
Recently some small issues were uncovered by modeling centers implementing COSP1.4.2. and I propose we include these changes on the COSP2 master branch.
*) In the LIDAR simulator there was a division-by-zero for scenes for really thick clouds. Including a guard against this results in some very minor changes to the COSP2 reference dataset: atb532_perp: 3.33 % of values differ, relative range: -1.10e+00 to -3.84e-05 These changes are associated with very small values and do not impact any of the MODIS diagnostics. (NOTE: this small change would not have been noticed using the previous offline driver, as atb532_perp is internal and is not part of output from the CMOR offline driver)
*) In the MODIS simulator, some of the gridmean diagnostics were being set to zero for clear scenes, when they should have been set to undefined. This impacted the averaging being performed by the model, resulting in much large values than expected. In the offline tests, we see the following: tautmodis: 18.30 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauwmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauimodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 tautlogmodis: 18.30 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauwlogmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauilogmodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 reffclwmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 reffclimodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 lwpmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 iwpmodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00
For both cases, diagnostics requested for CFMIP3/CMIP6 are not impacted.
You can view, comment on, or merge this pull request online at:
https://github.com/CFMIP/COSPv2.0/pull/9
Commit Summary
- Added changes to lidar simulator to guard against divide by zero. In MODIS simulator, set mean in-cloud fields to R_UNDEF instead of 0.
File Changes
- M driver/data/outputs/UKMO/cosp2_output_um.ref.nchttps://github.com/CFMIP/COSPv2.0/pull/9/files#diff-0 (0)
- M src/simulator/MODIS_simulator/modis_simulator.F90https://github.com/CFMIP/COSPv2.0/pull/9/files#diff-1 (20)
- M src/simulator/actsim/lidar_simulator.F90https://github.com/CFMIP/COSPv2.0/pull/9/files#diff-2 (21)
Patch Links:
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/CFMIP/COSPv2.0/pull/9, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AfO_cHkJG2uzk5hjSFTv5zLgua5JrV3Oks5thBS5gaJpZM4S3zuZ.
dustinswales
RobertPincus
Recently some small issues were uncovered by modeling centers implementing COSP1.4.2. and I propose we include these changes on the COSP2 master branch.
*) In the LIDAR simulator there was a division-by-zero for scenes for really thick clouds. Including a guard against this results in some very minor changes to the COSP2 reference dataset: atb532_perp: 3.33 % of values differ, relative range: -1.10e+00 to -3.84e-05 These changes are associated with very small values and do not impact any of the MODIS diagnostics. (NOTE: this small change would not have been noticed using the previous offline driver, as atb532_perp is internal and is not part of output from the CMOR offline driver)
*) In the MODIS simulator, some of the gridmean diagnostics were being set to zero for clear scenes, when they should have been set to undefined. This impacted the averaging being performed by the model, resulting in much large values than expected. In the offline tests, we see the following: tautmodis: 18.30 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauwmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauimodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 tautlogmodis: 18.30 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauwlogmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 tauilogmodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 reffclwmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 reffclimodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00 lwpmodis: 30.72 % of values differ, relative range: -1.00e+00 to -1.00e+00 iwpmodis: 65.36 % of values differ, relative range: -1.00e+00 to -1.00e+00
For both cases, diagnostics requested for CFMIP3/CMIP6 are not impacted.