Simcloud: A simple cloud scheme for Isca #246

daw538 · 2023-03-04T19:30:27Z

SimCloud is a simple cloud diagnostic scheme for general circulation models, implemented previously by @lqxyz for Isca and described in Liu et al. 2021. Currently the model sits as a private branch -- this PR brings the scheme inline with the current master in order to merge into the master, giving Isca a full cloud scheme that supersedes the simpler SPOOKIE-2 derived scheme currently in the master.

I propose adding an additional test case to the current suite to provide an example of SimCloud in situ (trip test file would need updating which I have not yet done). Qun has provided one in his branch which can be used. I also propose either moving or removing the two SPOOKIE-2 cloud test cases that are currently present in the SOCRATES test case directory to avoid confusion.

There is one change to the file mixed_layer.F90 that will need resolving. The present master has ice-dependent albedo calculated in the subroutine albedo_calc via the following method:

where(ice_concentration.gt.ice_concentration_threshold)
    albedo_inout=ice_albedo_value
end where

In this situation all the trip tests pass as shown below.

Results for all of the test cases ran comparing 9560521 and 193e7f0 are as follows...
axisymmetric : pass
bucket_model : pass
frierson : pass
giant_planet : pass
held_suarez : pass
MiMA : pass
realistic_continents_fixed_sst : pass
realistic_continents_variable_qflux : pass
socrates_aquaplanet : pass
top_down_test : pass
variable_co2_grey : pass
variable_co2_rrtm : pass
ape_aquaplanet : pass
Congratulations, all tests have passed

In the PR branch however, this calculation has been modified from a step function (as above) to a ramp function, entirely replacing the where statement thus

albedo_inout = albedo_inout*(1.0-ice_concentration) + ice_albedo_value*ice_concentration

and the following section was additionally commented out (in the current state of the PR I have reverted this behaviour)

! if(update_albedo_from_ice) then
!   call read_ice_conc(Time_next)
!   land_ice_mask=.false.
!   where(land_mask.or.(ice_concentration.gt.ice_concentration_threshold))
!     land_ice_mask=.true.
!   end where
! else
land_ice_mask=land_mask
! endif

Using the ramp function rather than the step function causes the trip tests for realistic_continents_fixed_sst and realistic_continents_variable_qflux to fail, which is understandable. We hence need to decide how to resolve the potential conflict between the two methods: do we bin one or do we introduce some form of choice with a namelist parameter? What would be the justification for using either?

Once this PR has been accepted I will follow up with a secondary PR that updates the docs to include the cloud module and a list of the available test cases (which is currently present but hidden in the Isca/exp folder.

…near relationship with RH

…tion diagnosis and critic RH profiles

…itted critical RH profiles

Conflicts: src/atmos_spectral/driver/solo/idealized_moist_phys.F90

…into simple_clouds_diag_cf

…ion)

…into simple_clouds_diag_cf

…s of CRE

…here it changes with temperature

…) in AMIP simulations

…ssure)

…n random-overlap overlap assumption

…ofile

daw538 · 2023-03-15T10:20:51Z

Hi Daniel, Sorry I did not see this review request until now. Happy to discuss in person at the Friday meeting if that is of interest?

Hi Penny, yes I think that might be worth it. Dan

…d controls for each scheme so both cannot be used.

daw538 · 2023-03-17T16:12:34Z

I have re-run the trip tests for the latest commit c68238b, which pass on all cases except socrates_aquaplanet_cloud.

Results for all of the test cases ran comparing 9560521 and c68238b are as follows...
axisymmetric : pass
bucket_model : pass
frierson : pass
giant_planet : pass
held_suarez : pass
MiMA : pass
realistic_continents_fixed_sst : pass
realistic_continents_variable_qflux : pass
socrates_aquaplanet : pass
socrates_aquaplanet_cloud : fail
top_down_test : pass
variable_co2_grey : pass
variable_co2_rrtm : pass
ape_aquaplanet : pass
Nightmare, some tests have failed

The failure behaviour is expected since I have changed the namelist parameter associated with using the available cloud schemes. The master commit does not have this option in the source so the associated run for this experiment fails. I have independently tested the equivalent runs from 9560521 and c68238b using the methodology from the trip test function which reports a pass.

The changes to the namelist include changing the default behaviour of do_cloud_simple in idealised_moist_physics_nml so that it turns on the SimCloud cloud scheme when true, whilst the SPOOKIE scheme is activated with a new control do_cloud_spookie. The default state of both of these options is false to preserve previous behaviour.

The current failure in the github test run is a result of change separate to this PR and is being addressed in #247 . An additional trip test for the SimCloud scheme will be provided in a future PR.

penmaher

Hi Dan, great job on getting this into review and making both cloud schemes available. As we discussed in the group meeting, my review has been focused on the technical and readability of the code. In time it would have good to do a more rather focused science review but that is far from a small thing to do. In general the quality of the code is good so most of the suggestions I have made are things you can choose to do or not. Let me know if you don't agree with anything or don't understand my comments.

exp/test_cases/simple_clouds/socrates_simcloud.py

penmaher · 2023-03-30T10:15:40Z

exp/test_cases/simple_clouds/socrates_simcloud.py

+    },
+
+    'diffusivity_nml': {
+        'do_entrain': True, #False,


I ran my test cases with this as True but don't recall why. Noted here in case helpful.

I'm not sure either. Looking at the source the default is set to True but it is not used in a number of the other test cases.

penmaher · 2023-03-30T10:17:05Z

exp/test_cases/simple_clouds/socrates_simcloud.py

+        'depth': 20.0,                          # Depth of mixed layer used
+        'land_option': 'input',                 # Tell mixed layer to get land mask from input file
+        'land_h_capacity_prefactor': 0.1,       # What factor to multiply mixed-layer depth by over land.
+        'albedo_value': 0.12,                   # Ocean albedo value


albedo looks low to me. Is this what Qun used?

Agreed. It looks low to me also and I'm wondering if it's a typo—the value is referenced in the namelist as coming from the control case of Jucker & Gerber, 2017, which I think is this paper. Looking in the paper however states a value of 20.5% (0.205), which seems more reasonable.

I wonder if this is because the normal mixed-layer albedo is so high precisely because we don't have clouds. But once we add clouds we don't need the surface albedo to be quite so high.

penmaher · 2023-03-30T14:03:44Z

src/atmos_param/cloud_simple/large_scale_cloud.F90

+
+  end subroutine large_scale_cloud_init
+
+  subroutine large_scale_cloud_diag(pfull, ps, rh, q_hum, qsat, qcl_rad, wg_full, cf, Time)


I found the '_diag' routine name a bit misleading. In this context it means diagnosing clouds in the column, but it reads as though we are initialising diagnostics or similar. If it is not too much of a pain, could we change this to 'large_scale_cloud'

penmaher · 2023-03-30T14:07:34Z

src/atmos_param/cloud_simple/large_scale_cloud.F90

Looks fine to me. This part of the code was easier to follow as the comments are much better. As with the cloud_simple.F90 I can't comment too much on the science parts without a lot more time to investigate, but it all looks superficially fine to me (i.e. nothing jumps out).

penmaher · 2023-03-30T14:08:38Z

src/atmos_param/cloud_simple/lcl.F90

@@ -0,0 +1,920 @@
+! The original code is from http://romps.berkeley.edu/papers/pubdata/2016/lcl/lcl.f90


Can you check that we can reproduce this code in here? I.e. take a look at the license Romps uses. I also see that there are other functions with different licenses in here. Can you make sure they are all open enough for us to use too please?

Given it was taken directly from Romps (which is a very trustworthy source), I propose we don't make any changes)

penmaher · 2023-03-30T14:12:57Z

src/atmos_param/cloud_simple/marine_strat_cloud.F90

+
+  end subroutine marine_strat_cloud_init
+
+  subroutine marine_strat_cloud_diag(temp, p_full, p_half, z_full, rh, q_hum, temp_2m, &


Same criticism here about diag being misleading. Propose 'marine_strat_cloud'

penmaher · 2023-03-30T14:15:00Z

src/atmos_param/cloud_simple/marine_strat_cloud.F90

There is quite a bit of science in this one, which probably should be reviewed in more detail than I have, but I don't have a week to spend on this PR so I am giving it the benefit of the doubt.

penmaher · 2023-05-05T14:21:19Z

I am accepting this merge, in advance of going on mat leave, so that I do not hold the process up. I pass the review over to Neil and Stephen who can continue discussion with Dan. The issues with the PR I have raised have been discussed with Dan, and I am satisfied there is nothing preventing this from going onto the master (majority of comments are nice to have suggestions but not needed per say). Penny

penmaher

Sorry I closed the PR rather than approved it. Have reopened and now accepting the PR.

AndrewILWilliams · 2023-07-13T12:50:25Z

Hi there! Can I ask if SimCloud passes its properties to RRTM? Or does RRTM currently just do clear-sky radiation?

sit23 · 2023-07-13T13:01:43Z

At the moment Simcloud just works in Socrates. We did try to get some cloud stuff working for RRTM, but we didn't finish it as we just switched to Socrates instead.

sit23

This all looks good to me. I haven't checked all of it massively closely, but I think it's important we get this in soon. If it's not breaking the trip tests then that's fine. Given that the master hasn't changed since they were run, I think it's fine to put this in. Thanks for all your work on this Dan. I guess one thing that should be done at some point would be to add a bit to the docs about the clouds. That could be left as a to-do!

sit23 · 2023-10-18T12:57:05Z

exp/test_cases/simple_clouds/socrates_simcloud.py

+        'depth': 20.0,                          # Depth of mixed layer used
+        'land_option': 'input',                 # Tell mixed layer to get land mask from input file
+        'land_h_capacity_prefactor': 0.1,       # What factor to multiply mixed-layer depth by over land.
+        'albedo_value': 0.12,                   # Ocean albedo value


I wonder if this is because the normal mixed-layer albedo is so high precisely because we don't have clouds. But once we add clouds we don't need the surface albedo to be quite so high.

sit23 · 2023-10-18T12:59:47Z

exp/test_cases/trip_test/trip_test_functions.py

@@ -74,12 +74,18 @@ def get_nml_diag(test_case_name):
        input_files = exp_temp.inputfiles
        nml_out = exp_temp.namelist        

+    #if 'soc_realistic_continents_fixed_sst_with_linear_cld_scheme' in test_case_name:


Given that the trip tests seem to have passed for the relevant cases, let's add these tests back.

sit23 · 2023-10-18T13:04:07Z

@daw538 - I think that we're done here now. I reckon we should just merge it in. I see @ntlewis is still requested for a review. Do you have time to do this @ntlewis or are you happy for us to go ahead? I'll try and fix the automated testing in #255

ntlewis · 2023-10-18T13:32:17Z

@sit23 I'm happy for this to go in. Also, regarding the albedo issue above, I think it's low because clouds have been introduced. For example, a typical albedo for open ocean would be something like 0.08, so it's still higher than that. Given that the land prefactor is also > 1, I think this'll end up with a sensible albedo once the effect of clouds is included.

ntlewis

Happy for this to go in. See comment above about albedo.

sit23 and others added 30 commits December 14, 2018 10:49

Experiment setup changes.

3bf10f2

Checkout cloud_simple from column_clouds_diag_cf branch

d63ce1e

Test: run socrates with clouds; diagnose cloud fraction based on a li…

c95129f

…near relationship with RH

add experiment to test socrates with clouds with different cloud frac…

7e4ef20

…tion diagnosis and critic RH profiles

Add experiment to test the cloud diagnosis method: spookie with its f…

8460918

…itted critical RH profiles

Ignore file or directory begin with tmp

56dd444

Ignore files or directories begin with tmp

f1000ac

Test: changing albedos for the aquaplanet simulations with clouds

8afffbb

Test: changing albedos for the aquaplanet simulations with clouds

95878a7

Run simple clouds model with Socrates varying albedos

e3cc2ef

Add diagnostics to calculate the cloud radiative effect (CRE) at TOA

3bdc291

Conflicts: src/atmos_spectral/driver/solo/idealized_moist_phys.F90

Add two variables, soc_flux_lw_clr and soc_flux_sw, for clear sky

31dfdf0

pull from origin for simple_clouds_diag_cf branch

f634226

Merge branch 'simple_clouds_diag_cf' of https://github.com/lqxyz/Isca …

c348bf6

…into simple_clouds_diag_cf

Test scripts for cloud radiation effect (outputs for clear sky condit…

3cfcc3f

…ion)

Merge branch 'simple_clouds_diag_cf' of https://github.com/lqxyz/Isca …

e6c7e42

…into simple_clouds_diag_cf

Test scripts: varying critical RH at 200 and 700hPa to see the change…

5ce9823

…s of CRE

Test script for CRE calculation with linear diagnostic method

333d947

Test script: change albedo

670d97f

simple clouds: prescribe the liquid water in cloud based on temperature

43ced0b

Prescribe qcl in new method: fitting the observation

859eea5

Test scripts update

797bb8d

change the variable and function name for the q_in_cld calculation, w…

db9e7ca

…here it changes with temperature

Test scripts: Run simple clouds with new qcl (change with temperature…

6bad944

…) in AMIP simulations

Test scripts: run experiments with prescribed in-cloud q (func of pre…

069ef48

…ssure)

test: diagnose cloud low, middle, high and total cloud amount based o…

050de2e

…n random-overlap overlap assumption

Add diagnosis of marine stratocumulus using method from CAM

b1b740f

Diagnose stratcumulus based on estimated inversion strength (EIS)

1a4efff

change the simple_cloud code structure and add linear coefficients pr…

66875fa

…ofile

Check the iostat in reading namelist

65fc12f

daw538 force-pushed the simcloud branch from 2b4d7b9 to a556a6b Compare March 15, 2023 14:10

daw538 added 2 commits March 15, 2023 14:29

Added diagnostic messages to output to show cloud scheme in use. Adde…

386ec8a

…d controls for each scheme so both cannot be used.

Allowing both SimCloud and SPOOKIE to be passed to SOCRATES

3d506c8

daw538 force-pushed the simcloud branch from 090f383 to 3d506c8 Compare March 16, 2023 12:57

Update existing socrates test cases for consistency

c68238b

daw538 requested review from penmaher and ntlewis March 17, 2023 15:12

penmaher requested changes Mar 30, 2023

View reviewed changes

daw538 added 3 commits May 4, 2023 15:10

Removing files that should not be in PR

d4ddecb

Removing personalised comments and tidying terminology

9603436

Further comment tidying

abc8fcb

sit23 self-requested a review May 5, 2023 14:12

penmaher closed this May 5, 2023

penmaher reopened this May 5, 2023

penmaher self-requested a review May 5, 2023 14:23

penmaher approved these changes May 5, 2023

View reviewed changes

sit23 approved these changes Oct 18, 2023

View reviewed changes

ntlewis approved these changes Oct 18, 2023

View reviewed changes

daw538 added 4 commits October 26, 2023 12:55

Reinstate spookie cloud test case

c067351

Merge remote-tracking branch 'upstream/master' into simcloud

5f9daa3

Fixes for gfortran compilation

ea22d13

Further fixes for gfortran compilation

1feab10

daw538 merged commit 7acc5d2 into ExeClim:master Oct 26, 2023
1 check passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Simcloud: A simple cloud scheme for Isca #246

Simcloud: A simple cloud scheme for Isca #246

daw538 commented Mar 4, 2023

daw538 commented Mar 15, 2023

daw538 commented Mar 17, 2023 •

edited

Loading

penmaher left a comment

penmaher Mar 30, 2023

daw538 Mar 31, 2023

penmaher Mar 30, 2023

daw538 Mar 31, 2023

sit23 Oct 18, 2023

penmaher Mar 30, 2023

penmaher Mar 30, 2023

penmaher Mar 30, 2023

penmaher Mar 30, 2023

penmaher Mar 30, 2023

penmaher commented May 5, 2023

penmaher left a comment

AndrewILWilliams commented Jul 13, 2023

sit23 commented Jul 13, 2023

sit23 left a comment

sit23 Oct 18, 2023

sit23 Oct 18, 2023

sit23 commented Oct 18, 2023

ntlewis commented Oct 18, 2023

ntlewis left a comment


		end subroutine large_scale_cloud_init

		subroutine large_scale_cloud_diag(pfull, ps, rh, q_hum, qsat, qcl_rad, wg_full, cf, Time)

		@@ -0,0 +1,920 @@
		! The original code is from http://romps.berkeley.edu/papers/pubdata/2016/lcl/lcl.f90


		end subroutine marine_strat_cloud_init

		subroutine marine_strat_cloud_diag(temp, p_full, p_half, z_full, rh, q_hum, temp_2m, &

Simcloud: A simple cloud scheme for Isca #246

Simcloud: A simple cloud scheme for Isca #246

Conversation

daw538 commented Mar 4, 2023

daw538 commented Mar 15, 2023

daw538 commented Mar 17, 2023 • edited Loading

penmaher left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

penmaher commented May 5, 2023

penmaher left a comment

Choose a reason for hiding this comment

AndrewILWilliams commented Jul 13, 2023

sit23 commented Jul 13, 2023

sit23 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sit23 commented Oct 18, 2023

ntlewis commented Oct 18, 2023

ntlewis left a comment

Choose a reason for hiding this comment

daw538 commented Mar 17, 2023 •

edited

Loading