Documentation of OCND2 modification of ICE3/ICE4 microphysics in AROME

Introduction

This option was implemented in ICE3/ICE4 microphysics in 2014 in order to improve the performance of the HARMONIE-AROME model configuration in Winter over the Arctic/Subarctic region. Errors corrected were mainly for missing low clouds in moderately cold conditions, an excess of ice clouds in the case of severe cold weather and an excess of cirrus clouds.

Implementation in CY46 - switching on the parameterisation

To use the parameterisation go to nam/harmonie_namelists.pm and set LOCND2 = .TRUE. in the namparar namelist.

About the Code

In CY46, there are two coding versions of ICE3/ICE4 - rain_ice_old.F90 and rain_ice.F90. The variable CMICRO determines which is used, OLD3 for rain_ice_old.F90 and ICE3 for rain_ice.F90. The structure of the code differs between these versions, and since the time stepping procedure is different, the result differs too. But the content of the modification for OCND2 is the same in both versions. The modifications can be be found by searching for

IF(OCND2) THEN \
    --- new code ---\
ENDIF

The main OCDN2 modifications are

Tuning factors for reducing the rate of deposition/evaporation of snow and graupel. See code block “1.2 COMPUTE SOME CONSTANT PARAMETERS” in rain_ice_old.F90 or in ice4_slow.F90. The tuning factors are then used later in rain_ice_old.F90, see code block “3.4.3 compute the deposition on rs: RVDEPS" for snow and in “3.4.6 compute the deposition on rg: RVDEPG” for graupel. In the rain_ice.F90 framework it is all done in the routine ice4_slow.F90. More information about the tuning parameters is included later in this documentation.
Mask to limit computation: Set by tuning parameters in the code block “1.2 COMPUTE SOME CONSTANT PARAMETERS“ in rain_ice_old.F90 or in aro_rain_ice.F90 within the rain_ice.F90 framework. For OCND2=FALSE the limits are hard-coded.
The cloud ice crystal concentration: Modified with OCND2, see code block “3.1.1 compute the cloud ice concentration” in rain_ice_old or ice4_nucleation.F90 within the rain_ice.F90 framework.
Turn large cloud ice crystals into snow: See code block “3.4.5 B:” in rain_ice_old.F90 or ice4_fast_si.F90 within the rain_ice.F90 framework.
Omit collision between snow and graupel since the effect in nature is very small and thus better to omit and speed up the computation a little. See code block “6.2.5” in rain_ice_old.F90 or ice4_fast_rg.F90 respectively
Sub grid-scale calculation of deposition/evaporation of cloud ice. See code block “7.2 Bergeron-Findeisen effect: RCBERI” in rain_ice_old.F90 or ice4_fast_si.F90 for the rain_ice.F90 set up.

There is also an important difference in condensation.F90: With OCND2, only liquid cloud water is handled within the statistical cloud scheme, not both ice and water as is the case with OCND2=F. With OCND2=F, the total cloud cover is calculated directly from the statistical cloud scheme. With OCND2=T, the total cloud cover is calculated as a sum of a liquid part, which is basically just the cloud cover from the statistical cloud scheme and an ice part which is based on the relative humidity with respect to ice and on the content of solid water species.

There are two new routines for OCND2:

icecloud.F90 is used for the sub grid-scale handling of relative humidity with respect to ice and thus for ice clouds. It is called from condensation.F90.
ice4_fast_si.F90 is only used by the newer rain_ice.F90 routine. As already mentioned, it deals with deposition/evaporation of cloud ice.

Tuning parameters

The tuning parameters used specifically for OCND2 can be divided into three categories:

Only having an effect if OCND2 is set to TRUE and used for SPP (April 2023).

Variable	Description
`RFMIN(21)`	Tuning factor for ice clouds, such as cirrus. A larger value means a larger effect of the presence of solid water and thus more ice clouds. (The value is somewhat dependent on what kind of measurement one compares with, and how thin a cirrus cloud should be to be counted as a cloud. A range of 0.5 to 3 should be enough.)

Only having effect if OCND2 is set to TRUE but currently (April 2023) not used in SPP.

Variable	Description
`RFMIN(12)`	Threshold supersaturation with respect to ice in the supersaturated part of the grid-box for treatment in the microphysics computation. A larger value gives more supersaturation and a somewhat faster computation. Values that are too large are physically unrealistic, but there seems to be no consensus about the best value.
`RFMIN(13)`	Threshold mixing ratio for different non-vapor water species treated in the microphysics computation. Larger values result in faster computation, but possible important processes, when only small mixing ratios of water species are present, may be missed.
`RFMIN(15)`	Ice crystal diameter(m) for conversion from cloud ice to snow. Larger values lead to more ice and less snow.
`RFMIN(27)`	Experimental! Minimum temperature (K) used for Meyers ice number concentration. Larger values give less ice for temperatures below `RFRMIN(27)`.
`RFMIN(39)`	Speed factor for deposition/evaporation rate of graupel. Larger values give faster deposition /evaporation.
`RFMIN(40)`	Speed factor for deposition/evaporation rate of snow. Larger values give faster deposition /evaporation.

Have an effect even when OCND2 is not used, but designed for OCND2

Variable	Description
`RFRMIN(1),RFRMIN(2),RFRMIN(3) and RFRMIN(4)`	Different thresholds for snow, ice, graupel and graupel again, respectively, leading to conversion of super-cooled rain into graupel. A higher value gives more super-cooled rain, but may be less physically realistic.
`RFRMIN(7)`	Tuning factor for the collisions between rain and snow. Higher values give less super-cooled rain and more snow. Zero means that those collisions are disregarded (probably OK).

Variable	Value	Description
`RFMIN(1)`	`1.0E−5`	Higher value means more supercooled rain and somewhat less graupel.
`RFMIN(2)`	`1.0E−8`	""
`RFMIN(3)`	`3.0E-7`	""
`RFMIN(4)`	`3.0E-7`	""
`RFMIN(5)`	`1.0E-7`	Higher value means less graupel and more snow. Experimental.
`RFMIN(6)`	`0.15`	Higher value means more graupel and less snow. Experimental.
`RFMIN(7)`	`0.`	Higher value means less supercooled rain and somewhat more snow.
`RFMIN(8)`	`1.`	> 1. Increase melt of graupel, < 1 decrease it. Experimental.
`RFMIN(9)`	`1.`	> 1 means increase IN-concentration and <1 decrease.
`RFMIN(10)`	`10.`	>10 means faster Kogan autoconversion <10 slower, only active for `LKOGAN=T`. This originates from the fact that the formula was based on an LES model with a higher horizontal resolution. It is easy to show that with a coarser resolution and an inhomogeneous cloud liquid field one has to add a compensating factor in order to retain the original mean autoconversion. Tests shows that a lower value e.g 3 would be better, and more in line with what ECMWF is using. The value 10 is, to some extent, a way of decreasing fog, but now we have a lot of other ways to reduce fog.
`RFMIN(11)`	`1.`	Setting e.g. 0.01 means that subgrid-scale fraction of cloud water is used. Minimum cloud fraction=0.01. Only active for `LKOGAN=T`.
`RFMIN(12)`	`0.`	The level of supersaturation in the ice-supersaturated part of grid-box needed to be treated in ice microphysics. (Greg Thompson recommend a higher value 0.05-0.25, in MetCoOp 0.05 is used) Higher value means faster computations, but also that any ice deposition in clear sky is neglected for ice-supersaturated between zero and `RFRMIN(12)`. Only used with OCND2.
`RFMIN(13)`	`1.0E-15`	The mixing-ratio of any water species needed to be treated in ice microphysics. The value 1.0E-15 is taken from old Hirlam. Only used with OCND2.
`RFMIN(14)`	`120.`	Time scale for conversion of large ice crystals to snow. Only used with LMODICEDEP (Experimental).
`RFMIN(15)`	`1.0E-4`	Diameter for conversion ice crystals into snow. Larger value gives more ice and less snow.
`RFMIN(16)`	`0.`	“C” parameter for size distribution of snow. (constant for number concentration, N=Cλ^x) Only active if non-zero. Experimental
`RFMIN(17)`	`0.`	“x” parameter for size distribution of snow. (slope for number concentration, N=Cλ^x) Only active if `RFRMIN(16)` is non-zero. Experimental.
`RFMIN(18)`	`0.`	With `RFRMIN(18)=1`, snow and graupel melt are based on wet bulb temperature, instead of temperature and leads to slower melting. Experimental.
`RFMIN(19)`	`0.`	Threshold cloud thickness for StCu/Cu transition [m] Only active for EDMF scheme and if non-zero, but very small effect.
`RFMIN(20)`	`0.`	Threshold cloud thickness used in shallow/deep decision [m]. Only active for EDMF scheme and if non-zero, higher value gives more shallow convection and less deep model resolved convection.
`RFMIN(21)`	`1.`	Tuning parameter for ice clouds. Larger value gives more cirrus and other ice clouds.
`RFMIN(22)`	`1.`	Tuning parameter for CDNC at lowest model level . Lower value give lower CDNC. `RFRMIN(22)=0.5` means CDNC= old CDNC x 0.5.
`RFMIN(23)`	`0.5`	Tuning parameter only active with `LHGT_QS`. The lower limit for reduction of VSIGQSAT.
`RFMIN(24)`	`1.5`	Tuning parameter only active with `LHGT_QS`. The upper limit for increase of VSIGQSAT.
`RFMIN(25)`	`30.`	Tuning parameter only active with `LHGT_QS`. The level thickness for which VSIGQSAT is unchanged with `LHGT_QS`.
`RFMIN(26)`	`0.`	If > 0.01, it replaces default CDNC everywhere. So `RFRMIN(26)=50E6` (Beware of that it is in m-3!) gives CDNC = 50 cm-3 at reference level (1000 hPa) and RFRMIN(26) x pressure/ ref-pressure elsewhere.
`RFMIN(27)`	`0.`	Minimum assumed temperature with respect to Meyers IN - concentration (K). Gives less IN concentration for temperatures below the value set. Experimental!
`RFMIN(28)`	`0.`	Currently not used.
`RFMIN(29)`	`0.`	If >0. and `RFRMIN(22)>0` it gives the upper limit in metres for which the reduction of CDNC has an effect. A linear decrease from the lowest level to `RFRMIN(29)` meters is assumed.
`RFMIN(30)`	`1.`	If not unity, CDNC is reduced/increased over sea with a factor `RFRMIN(30)` for the lowest model level and linearly reaching "no change" at `RFRMIN(29)` m height. If `RFRMIN(29)` is unset, `RFRMIN(30)` only affects the lowest model level.
`RFRMIN(31:38)`	`0.`	Currently not used.
`RFRMIN(39)`	`0.25.`	Reduction factor for deposition/evaporation of graupel. Only used when `OCND2=T` and `LMODICEDEP=F`.
`RFRMIN(40)`	`0.15`	Reduction factor for deposition/evaporation of snow. Only used when `OCND2=T` and `LMODICEDEP=F`.