Adding Ristic microphysics scheme

[Nikolar97]

TYPE: New feature

KEYWORDS: microphysic, forecast, cloud, precipitation

SOURCE: Ivan Ristic (Weather2 - www.weather2.rs)

DESCRIPTION OF CHANGES:
In order to improve cloud and precipitation forecast we developed new cloud prediction scheme and we implemented it in WRF model. Fractional cloud cover, cloud liquid water, cloud ice and cloud snow are explicitly predicted by adding three
prognostic equations for fractional cloud cover, cloud mixing ratio and snow per cloud fraction to the model. Sedimentation of ice and snow is also included in parameterization. Precipitation of rain and snow are determined from cloud fields. Clouds predicted like this can be used also in radiation parameterization.

Thermodynamic wet bulb temperature will be used for describing clouds because it is constant during water phase changes. By using this temperature moist static energy of model grid box and cloudy part inside the grid box is the same and principle of energy conservation is satisfied.

A complete description is now found in Ristic I., Kordic I., 2018: Cloud parameterization and cloud prediction scheme in the Eta numerical weather model. NWCM - Serbian Academy of Sciences and Arts which can be found at https://www.sanu.ac.rs/wp-content/uploads/2018/10/11_Cloud%20parameterization%20and%20cloud%20prediction%20scheme%20in%20Eta%20numerical%20weather%20model.pdf .

LIST OF MODIFIED FILES:
phys/Makefile
phys/module_microphysics_driver.F
phys/module_mp_ivanr_micro.F
phys/module_radiation_driver.F
phys/module_ra_gfdleta.F
Registry/Registry.EM_COMMON
dyn_em/solve_em.F
dyn_em/module_first_rk_step_part1.F

TESTS CONDUCTED:
All the wrf-coop tests passed and the microphysics is being actively used for running a model with NMM.
Integration of the model for test cases indicate that new cloud prediction scheme improved forecast compared to the original model. New fractional cloud cover formula showed good results in practice, since the fractional cloud cover, predicted in this way, was much closer to the real cloud cover values. Significant progress has been made in stratiform precipitation forecast. Positive impact on convection scheme is also noticed.
One test was run on 17. November 2011 with the Wrf model with ECMWF as boundary data. Model run for 72 hours in horizontal resolution of about 22 km and vertical resolution of 38 layers for the Europe domain. Fog was first to be tested.

The second test was run on 24. June 2015 with Wrf model with ECMWF as boundary data. Model was run for 24 hours in
horizontal resolution of about 22 km and a vertical resolution of 38 layers for the Europe domain. The second test situation tested mid-morning precipitation over northern Serbia.

More details can be found in the aforementioned pdf file.

RELEASE NOTE: Ivan Ristic: Custom implementation of microphysics

[weiwangncar]

@Nikolar97 Thanks for doing this new PR. You're almost there. The regression tests show that the code failed data assimilation compilation. It is possible that all you need is to add the new mp declaration and new variables to the DA registry: registry.var, and registry.wrfplus. You can refer to the following PR and see how they added the new variables to those registry files:
https://github.com/wrf-model/WRF/pull/1364/files

[weiwangncar]

@Nikolar97 It looks like more changes are required. Please see these modified files in the PR for NTU microphysics:
wrftladj/module_bc_ad.F
wrftladj/module_bc_tl.F
wrftladj/module_microphysics_driver_ad.F
wrftladj/module_microphysics_driver_tl.F
wrftladj/solve_em_ad.F
wrftladj/solve_em_tl.F

[Nikolar97]

@weiwangncar It seems the problem was the missing noahmp files that were not added after the fork to the main branch.

[Nikolar97]

@dudhia The last commit should be of the full microphysics scheme. We were doing partial updates to check if everything works.

[weiwangncar]

@weiwangncar It seems the problem was the missing noahmp files that were not added after the fork to the main branch.

These noahmp modules should be checked out at the compile time from another repository, and should not be added to this PR. The regression tests suggest that the code failed when compiling WRF 4DVar, and hence my suggestion to add declaration of the scheme in its respective registries, and like passing the variables through the files similar to those indicated in the NTU scheme PR.

[weiwangncar]

The regression tests have passed - congratulations!

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 23           24
Number of Builds       : 60           58
Number of Simulations  : 158           156        0
Number of Comparisons  : 95           92        0

Failed Simulations are: 
None
Which comparisons are not bit-for-bit: 
None

[Nikolar97]

@weiwangncar Thank you. @dudhia What steps should we take next after every test was successful with the new build.