source: LMDZ6/branches/cirrus/libf/phylmd/lmdz_lscp_poprecip.F90 @ 5225

MODULE lmdz_lscp_poprecip
!----------------------------------------------------------------
! Module for the process-oriented treament of precipitation
! that are called in LSCP
! Authors: Atelier Nuage (G. Riviere, L. Raillard, M. Wimmer,
! N. Dutrievoz, E. Vignon, A. Borella, et al.)
! Jan. 2024


IMPLICIT NONE

CONTAINS

!----------------------------------------------------------------
! Computes the processes-oriented precipitation formulations for
! evaporation and sublimation
! 
SUBROUTINE poprecip_precld( &
           klon, dtime, iftop, paprsdn, paprsup, pplay, temp, tempupnew, qvap, &
           qprecip, precipfracclr, precipfraccld, qvapclrup, qtotupnew, &
           rain, rainclr, raincld, snow, snowclr, snowcld, dqreva, dqssub &
           )

USE lmdz_lscp_ini, ONLY : prt_level, lunout
USE lmdz_lscp_ini, ONLY : coef_eva, coef_sub, expo_eva, expo_sub, thresh_precip_frac
USE lmdz_lscp_ini, ONLY : RCPD, RLSTT, RLVTT, RLMLT, RVTMP2, RTT, RD, RG
USE lmdz_lscp_ini, ONLY : ok_corr_vap_evasub
USE lmdz_lscp_tools, ONLY : calc_qsat_ecmwf

IMPLICIT NONE


INTEGER, INTENT(IN)                     :: klon           !--number of horizontal grid points [-]
REAL,    INTENT(IN)                     :: dtime          !--time step [s]
LOGICAL, INTENT(IN)                     :: iftop          !--if top of the column


REAL,    INTENT(IN),    DIMENSION(klon) :: paprsdn        !--pressure at the bottom interface of the layer [Pa]
REAL,    INTENT(IN),    DIMENSION(klon) :: paprsup        !--pressure at the top interface of the layer [Pa]
REAL,    INTENT(IN),    DIMENSION(klon) :: pplay          !--pressure in the middle of the layer [Pa]

REAL,    INTENT(INOUT), DIMENSION(klon) :: temp           !--current temperature [K]
REAL,    INTENT(INOUT), DIMENSION(klon) :: tempupnew      !--updated temperature of the overlying layer [K]

REAL,    INTENT(INOUT), DIMENSION(klon) :: qvap           !--current water vapor specific humidity (includes evaporated qi and ql) [kg/kg]
REAL,    INTENT(INOUT), DIMENSION(klon) :: qprecip        !--specific humidity in the precipitation falling from the upper layer [kg/kg]

REAL,    INTENT(INOUT), DIMENSION(klon) :: precipfracclr  !--fraction of precipitation in the clear sky IN THE LAYER ABOVE [-]
REAL,    INTENT(INOUT), DIMENSION(klon) :: precipfraccld  !--fraction of precipitation in the cloudy air IN THE LAYER ABOVE [-]

REAL,    INTENT(IN),    DIMENSION(klon) :: qvapclrup      !--clear-sky specific humidity IN THE LAYER ABOVE [kg/kg]
REAL,    INTENT(IN),    DIMENSION(klon) :: qtotupnew      !--total specific humidity IN THE LAYER ABOVE [kg/kg]

REAL,    INTENT(INOUT), DIMENSION(klon) :: rain           !--flux of rain gridbox-mean coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: rainclr        !--flux of rain gridbox-mean in clear sky coming from the layer above [kg/s/m2]
REAL,    INTENT(IN),    DIMENSION(klon) :: raincld        !--flux of rain gridbox-mean in cloudy air coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: snow           !--flux of snow gridbox-mean coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: snowclr        !--flux of snow gridbox-mean in clear sky coming from the layer above [kg/s/m2]
REAL,    INTENT(IN),    DIMENSION(klon) :: snowcld        !--flux of snow gridbox-mean in cloudy air coming from the layer above [kg/s/m2]

REAL,    INTENT(OUT),   DIMENSION(klon) :: dqreva         !--rain tendency due to evaporation [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqssub         !--snow tendency due to sublimation [kg/kg/s]


!--Integer for interating over klon
INTEGER :: i
!--dhum_to_dflux: coef to convert a specific quantity variation to a flux variation
REAL, DIMENSION(klon) :: dhum_to_dflux
!--
REAL, DIMENSION(klon) :: rho, dz

!--Saturation values
REAL, DIMENSION(klon) :: qzero, qsat, dqsat, qsatl, dqsatl, qsati, dqsati
!--Vapor in the clear sky
REAL :: qvapclr
!--Fluxes tendencies because of evaporation and sublimation
REAL :: dprecip_evasub_max, draineva, dsnowsub, dprecip_evasub_tot
!--Specific humidity tendencies because of evaporation and sublimation
REAL :: dqrevap, dqssubl
!--Specific heat constant
REAL :: cpair, cpw

!--Initialisation
qzero(:)  = 0.
dqreva(:) = 0.
dqssub(:) = 0.
dqrevap   = 0.
dqssubl   = 0.

!-- dhum_to_dflux = rho * dz/dt = 1 / g * dP/dt
dhum_to_dflux(:) = ( paprsdn(:) - paprsup(:) ) / RG / dtime
rho(:) = pplay(:) / temp(:) / RD
dz(:) = ( paprsdn(:) - paprsup(:) ) / RG / rho(:)

!--Calculation of saturation specific humidity
!--depending on temperature:
CALL calc_qsat_ecmwf(klon,temp(:),qzero(:),pplay(:),RTT,0,.false.,qsat(:),dqsat(:))
!--wrt liquid water
CALL calc_qsat_ecmwf(klon,temp(:),qzero(:),pplay(:),RTT,1,.false.,qsatl(:),dqsatl(:))
!--wrt ice
CALL calc_qsat_ecmwf(klon,temp(:),qzero(:),pplay(:),RTT,2,.false.,qsati(:),dqsati(:))



!--First step consists in "thermalizing" the layer:
!--as the flux of precip from layer above "advects" some heat (as the precip is at the temperature 
!--of the overlying layer) we recalculate a mean temperature that both the air and the precip in the 
!--layer have.

IF (iftop) THEN

  DO i = 1, klon
    qprecip(i) = 0.
  ENDDO

ELSE

  DO i = 1, klon
    !--No condensed water so cp=cp(vapor+dry air)
    !-- RVTMP2=rcpv/rcpd-1
    cpair = RCPD * ( 1. + RVTMP2 * qvap(i) )
    cpw = RCPD * RVTMP2
    !--qprecip has to be thermalized with 
    !--layer's air so that precipitation at the ground has the
    !--same temperature as the lowermost layer
    !--we convert the flux into a specific quantity qprecip
    qprecip(i) = ( rain(i) + snow(i) ) / dhum_to_dflux(i)
    !-- t(i,k+1) + d_t(i,k+1): new temperature of the overlying layer
    temp(i) = ( tempupnew(i) * qprecip(i) * cpw + cpair * temp(i) ) &
            / ( cpair + qprecip(i) * cpw )
  ENDDO

ENDIF

! TODO Probleme : on utilise qvap total dans la maille pour l'evap / sub
! alors qu'on n'evap / sub que dans le ciel clair
! deux options pour cette routine :
! - soit on diagnostique le nuage AVANT l'evap / sub et on estime donc
! la fraction precipitante ciel clair dans la maille, ce qui permet de travailler
! avec des fractions, des fluxs et surtout un qvap dans le ciel clair
! - soit on pousse la param de Ludo au bout, et on prend un qvap de k+1
! dans le ciel clair, avec un truc comme :
!   qvapclr(k) = qvapclr(k+1)/qtot(k+1) * qtot(k)
! UPDATE : on code la seconde version. A voir si on veut mettre la premiere version.


DO i = 1, klon

  !--If there is precipitation from the layer above
  ! NOTE TODO here we could replace the condition on precipfracclr(i) by a condition
  ! such as eps or thresh_precip_frac, to remove the senseless barrier in the formulas
  ! of evap / sublim
  IF ( ( ( rain(i) + snow(i) ) .GT. 0. ) .AND. ( precipfracclr(i) .GT. 0. ) ) THEN

    IF ( ok_corr_vap_evasub ) THEN
      !--Corrected version - we use the same water ratio between
      !--the clear and the cloudy sky as in the layer above. This
      !--extends the assumption that the cloud fraction is the same
      !--as the layer above. This is assumed only for the evap / subl
      !--process
      !--Note that qvap(i) is the total water in the gridbox, and
      !--precipfraccld(i) is the cloud fraction in the layer above
      qvapclr = qvapclrup(i) / qtotupnew(i) * qvap(i) / ( 1. - precipfraccld(i) )
    ELSE
      !--Legacy version from Ludo - we use the total specific humidity
      !--for the evap / subl process
      qvapclr = qvap(i)
    ENDIF

    !--Evaporation of liquid precipitation coming from above
    !--in the clear sky only
    !--dprecip/dz = -beta*(1-qvap/qsat)*(precip**expo_eva)
    !--formula from Sundqvist 1988, Klemp & Wilhemson 1978
    !--Exact explicit formulation (rainclr is resolved exactly, qvap explicitly)
    !--which does not need a barrier on rainclr, because included in the formula
    draineva = precipfracclr(i) * ( MAX(0., &
             - coef_eva * ( 1. - expo_eva ) * (1. - qvapclr / qsatl(i)) * dz(i) &
             + ( rainclr(i) / MAX(thresh_precip_frac, precipfracclr(i)) )**( 1. - expo_eva ) &
               ) )**( 1. / ( 1. - expo_eva ) ) - rainclr(i)
             
    !--Evaporation is limited by 0
    draineva = MIN(0., draineva)


    !--Sublimation of the solid precipitation coming from above
    !--(same formula as for liquid precip)
    !--Exact explicit formulation (snowclr is resolved exactly, qvap explicitly)
    !--which does not need a barrier on snowclr, because included in the formula
    dsnowsub = precipfracclr(i) * ( MAX(0., &
             - coef_sub * ( 1. - expo_sub ) * (1. - qvapclr / qsati(i)) * dz(i) &
             + ( snowclr(i) / MAX(thresh_precip_frac, precipfracclr(i)) )**( 1. - expo_sub ) &
             ) )**( 1. / ( 1. - expo_sub ) ) - snowclr(i)

    !--Sublimation is limited by 0
    ! TODO: change max when we will allow for vapor deposition in supersaturated regions
    dsnowsub = MIN(0., dsnowsub)

    !--Evaporation limit: we ensure that the layer's fraction below
    !--the clear sky does not reach saturation. In this case, we 
    !--redistribute the maximum flux dprecip_evasub_max conserving the ratio liquid/ice 
    !--Max evaporation is computed not to saturate the clear sky precip fraction
    !--(i.e., the fraction where evaporation occurs)
    !--It is expressed as a max flux dprecip_evasub_max
    
    dprecip_evasub_max = MIN(0., ( qvapclr - qsat(i) ) * precipfracclr(i)) &
                     * dhum_to_dflux(i)
    dprecip_evasub_tot = draineva + dsnowsub

    !--Barriers
    !--If activates if the total is LOWER than the max because
    !--everything is negative
    IF ( dprecip_evasub_tot .LT. dprecip_evasub_max ) THEN
      draineva = dprecip_evasub_max * draineva / dprecip_evasub_tot
      dsnowsub = dprecip_evasub_max * dsnowsub / dprecip_evasub_tot
    ENDIF


    !--New solid and liquid precipitation fluxes after evap and sublimation
    dqrevap = draineva / dhum_to_dflux(i)
    dqssubl = dsnowsub / dhum_to_dflux(i)


    !--Vapor is updated after evaporation/sublimation (it is increased)
    qvap(i) = qvap(i) - dqrevap - dqssubl
    !--qprecip is the total condensed water in the precip flux (it is decreased)
    qprecip(i) = qprecip(i) + dqrevap + dqssubl
    !--Air and precip temperature (i.e., gridbox temperature)
    !--is updated due to latent heat cooling
    temp(i) = temp(i) &
            + dqrevap * RLVTT / RCPD &
            / ( 1. + RVTMP2 * ( qvap(i) + qprecip(i) ) ) &
            + dqssubl * RLSTT / RCPD &
            / ( 1. + RVTMP2 * ( qvap(i) + qprecip(i) ) )

    !--Add tendencies
    !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
    rainclr(i) = MAX(0., rainclr(i) + draineva)
    snowclr(i) = MAX(0., snowclr(i) + dsnowsub)

    !--If there is no more precip fluxes, the precipitation fraction in clear
    !--sky is set to 0
    IF ( ( rainclr(i) + snowclr(i) ) .LE. 0. ) precipfracclr(i) = 0.

    !--Calculation of the total fluxes
    rain(i) = rainclr(i) + raincld(i)
    snow(i) = snowclr(i) + snowcld(i)

  ELSEIF ( ( rain(i) + snow(i) ) .LE. 0. ) THEN
    !--If no precip, we reinitialize the cloud fraction used for the precip to 0 
    precipfraccld(i) = 0.
    precipfracclr(i) = 0.

  ENDIF ! ( ( rain(i) + snow(i) ) .GT. 0. )

  !--Diagnostic tendencies
  dqssub(i) = dqssubl / dtime
  dqreva(i) = dqrevap / dtime

ENDDO ! loop on klon


END SUBROUTINE poprecip_precld

!----------------------------------------------------------------
! Computes the processes-oriented precipitation formulations for
! - autoconversion (auto) via a deposition process
! - aggregation (agg)
! - riming (rim)
! - collection (col)
! - melting (melt)
! - freezing (freez)
! 
SUBROUTINE poprecip_postcld( &
           klon, dtime, paprsdn, paprsup, pplay, ctot_vol, ptconv, &
           temp, qvap, qliq, qice, icefrac, cldfra, &
           precipfracclr, precipfraccld, &
           rain, rainclr, raincld, snow, snowclr, snowcld, &
           qraindiag, qsnowdiag, dqrauto, dqrcol, dqrmelt, dqrfreez, &
           dqsauto, dqsagg, dqsrim, dqsmelt, dqsfreez)

USE lmdz_lscp_ini, ONLY : prt_level, lunout
USE lmdz_lscp_ini, ONLY : RCPD, RLSTT, RLVTT, RLMLT, RVTMP2, RTT, RD, RG, RPI
USE lmdz_lscp_tools, ONLY : calc_qsat_ecmwf

USE lmdz_lscp_ini, ONLY : cld_lc_con, cld_tau_con, cld_expo_con, seuil_neb,    &
                          cld_lc_lsc, cld_tau_lsc, cld_expo_lsc, rain_int_min, & 
                          thresh_precip_frac, gamma_col, gamma_agg, gamma_rim, &
                          rho_rain, r_rain, r_snow, rho_ice,                   &
                          tau_auto_snow_min, tau_auto_snow_max,                &
                          thresh_precip_frac, eps,                             &
                          gamma_melt, alpha_freez, beta_freez, temp_nowater,   &
                          iflag_cloudth_vert, iflag_rain_incloud_vol,          &
                          cld_lc_lsc_snow, cld_lc_con_snow, gamma_freez,       &
                          rain_fallspeed_clr, rain_fallspeed_cld,              &
                          snow_fallspeed_clr, snow_fallspeed_cld


IMPLICIT NONE

INTEGER, INTENT(IN)                     :: klon           !--number of horizontal grid points [-]
REAL,    INTENT(IN)                     :: dtime          !--time step [s]

REAL,    INTENT(IN),    DIMENSION(klon) :: paprsdn        !--pressure at the bottom interface of the layer [Pa]
REAL,    INTENT(IN),    DIMENSION(klon) :: paprsup        !--pressure at the top interface of the layer [Pa]
REAL,    INTENT(IN),    DIMENSION(klon) :: pplay          !--pressure in the middle of the layer [Pa]

REAL,    INTENT(IN),    DIMENSION(klon) :: ctot_vol       !--volumic cloud fraction [-]
LOGICAL, INTENT(IN),    DIMENSION(klon) :: ptconv         !--true if we are in a convective point

REAL,    INTENT(INOUT), DIMENSION(klon) :: temp           !--current temperature [K]
REAL,    INTENT(INOUT), DIMENSION(klon) :: qvap           !--current water vapor specific humidity [kg/kg]
REAL,    INTENT(INOUT), DIMENSION(klon) :: qliq           !--current liquid water specific humidity [kg/kg]
REAL,    INTENT(INOUT), DIMENSION(klon) :: qice           !--current ice water specific humidity [kg/kg]
REAL,    INTENT(IN),    DIMENSION(klon) :: icefrac        !--ice fraction [-]
REAL,    INTENT(IN),    DIMENSION(klon) :: cldfra         !--cloud fraction [-]

REAL,    INTENT(INOUT), DIMENSION(klon) :: precipfracclr  !--fraction of precipitation in the clear sky IN THE LAYER ABOVE [-]
REAL,    INTENT(INOUT), DIMENSION(klon) :: precipfraccld  !--fraction of precipitation in the cloudy air IN THE LAYER ABOVE [-]
                                                          !--NB. at the end of the routine, becomes the fraction of precip
                                                          !--in the current layer

REAL,    INTENT(INOUT), DIMENSION(klon) :: rain           !--flux of rain gridbox-mean coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: rainclr        !--flux of rain gridbox-mean in clear sky coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: raincld        !--flux of rain gridbox-mean in cloudy air coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: snow           !--flux of snow gridbox-mean coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: snowclr        !--flux of snow gridbox-mean in clear sky coming from the layer above [kg/s/m2]
REAL,    INTENT(INOUT), DIMENSION(klon) :: snowcld        !--flux of snow gridbox-mean in cloudy air coming from the layer above [kg/s/m2]

REAL,    INTENT(OUT),   DIMENSION(klon) :: qraindiag      !--DIAGNOSTIC specific rain content [kg/kg]
REAL,    INTENT(OUT),   DIMENSION(klon) :: qsnowdiag      !--DIAGNOSTIC specific snow content [kg/kg]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqrcol         !--rain tendendy due to collection by rain of liquid cloud droplets [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqsagg         !--snow tendency due to collection of lcoud ice by aggregation [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqrauto        !--rain tendency due to autoconversion of cloud liquid [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqsauto        !--snow tendency due to autoconversion of cloud ice [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqsrim         !--snow tendency due to riming [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqsmelt        !--snow tendency due to melting [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqrmelt        !--rain tendency due to melting [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqsfreez       !--snow tendency due to freezing [kg/kg/s]
REAL,    INTENT(OUT),   DIMENSION(klon) :: dqrfreez       !--rain tendency due to freezing [kg/kg/s]



!--Local variables

INTEGER :: i
!--dhum_to_dflux: coef to convert a specific quantity variation to a flux variation
REAL, DIMENSION(klon) :: dhum_to_dflux
REAL, DIMENSION(klon) :: qtot                             !--includes vap, liq, ice and precip

!--Partition of the fluxes
REAL :: dcldfra
REAL :: precipfractot
REAL :: dprecipfracclr, dprecipfraccld
REAL :: drainclr, dsnowclr
REAL :: draincld, dsnowcld

!--Collection, aggregation and riming
REAL :: eff_cldfra
REAL :: coef_col, coef_agg, coef_rim, coef_tmp, qrain_tmp
REAL :: Eff_rain_liq, Eff_snow_ice, Eff_snow_liq
REAL :: rho_snow
REAL :: dqlcol           !--loss of liquid cloud content due to collection by rain [kg/kg/s]
REAL :: dqiagg           !--loss of ice cloud content due to collection by aggregation [kg/kg/s]
REAL :: dqlrim           !--loss of liquid cloud content due to riming on snow [kg/kg/s]

!--Autoconversion
REAL :: qthresh_auto_rain, tau_auto_rain, expo_auto_rain
REAL :: qthresh_auto_snow, tau_auto_snow, expo_auto_snow
REAL :: dqlauto          !--loss of liquid cloud content due to autoconversion to rain [kg/kg/s]
REAL :: dqiauto          !--loss of ice cloud content due to autoconversion to snow [kg/kg/s]

!--Melting
REAL :: dqsmelt_max, air_thermal_conduct
REAL :: nb_snowflake_clr, nb_snowflake_cld
REAL :: capa_snowflake, temp_wetbulb
REAL :: rho, r_ice
REAL :: dqsclrmelt, dqscldmelt, dqstotmelt
REAL, DIMENSION(klon) :: qzero, qsat, dqsat

!--Freezing
REAL :: dqrfreez_max
REAL :: tau_freez
REAL :: dqrclrfreez, dqrcldfreez, dqrtotfreez, dqrtotfreez_step1, dqrtotfreez_step2
REAL :: coef_freez
REAL :: dqifreez        !--loss of ice cloud content due to collection of ice from rain [kg/kg/s]
REAL :: Eff_rain_ice


!--Initialisation of variables


qzero(:)    = 0.

dqrcol(:)   = 0.
dqsagg(:)   = 0.
dqsauto(:)  = 0.
dqrauto(:)  = 0.
dqsrim(:)   = 0.
dqrmelt(:)  = 0.
dqsmelt(:)  = 0.
dqrfreez(:) = 0.
dqsfreez(:) = 0.


DO i = 1, klon

  !--Variables initialisation
  dqlcol  = 0.
  dqiagg  = 0.
  dqiauto = 0.
  dqlauto = 0.
  dqlrim  = 0.

  !-- dhum_to_dflux = rho * dz/dt = 1 / g * dP/dt
  dhum_to_dflux(i) = ( paprsdn(i) - paprsup(i) ) / RG / dtime
  qtot(i) = qvap(i) + qliq(i) + qice(i) &
          + ( raincld(i) + rainclr(i) + snowcld(i) + snowclr(i) ) / dhum_to_dflux(i)

  !------------------------------------------------------------
  !--     PRECIPITATION FRACTIONS UPDATE
  !------------------------------------------------------------
  !--The goal of this routine is to reattribute precipitation fractions
  !--and fluxes to clear or cloudy air, depending on the variation of
  !--the cloud fraction on the vertical dimension. We assume a
  !--maximum-random overlap of the cloud cover (see Jakob and Klein, 2000,
  !--and LTP thesis, 2021)
  !--NB. in fact, we assume a maximum-random overlap of the total precip. frac

  !--Initialisation
  precipfractot = precipfracclr(i) + precipfraccld(i)

  !--Instead of using the cloud cover which was use in LTP thesis, we use the
  !--total precip. fraction to compute the maximum-random overlap. This is
  !--because all the information of the cloud cover is embedded into
  !--precipfractot, and this allows for taking into account the potential
  !--reduction of the precipitation fraction because either the flux is too
  !--small (see barrier at the end of poprecip_postcld) or the flux is completely
  !--evaporated (see barrier at the end of poprecip_precld)
  !--NB. precipfraccld(i) is here the cloud fraction of the layer above
  !precipfractot = 1. - ( 1. - precipfractot ) * &
  !               ( 1. - MAX( cldfra(i), precipfraccld(i) ) ) &
  !             / ( 1. - MIN( precipfraccld(i), 1. - eps ) )


  IF ( precipfraccld(i) .GT. ( 1. - eps ) ) THEN
    precipfractot = 1.
  ELSE
    precipfractot = 1. - ( 1. - precipfractot ) * &
                 ( 1. - MAX( cldfra(i), precipfraccld(i) ) ) &
               / ( 1. - precipfraccld(i) )
  ENDIF

  !--precipfraccld(i) is here the cloud fraction of the layer above
  dcldfra = cldfra(i) - precipfraccld(i)
  !--Tendency of the clear-sky precipitation fraction. We add a MAX on the
  !--calculation of the current CS precip. frac.
  !dprecipfracclr = MAX( 0., ( precipfractot - cldfra(i) ) ) - precipfracclr(i)
  !--We remove it, because precipfractot is guaranteed to be > cldfra (the MAX is activated
  !--if precipfractot < cldfra)
  dprecipfracclr = ( precipfractot - cldfra(i) ) - precipfracclr(i)
  !--Tendency of the cloudy precipitation fraction. We add a MAX on the
  !--calculation of the current CS precip. frac.
  !dprecipfraccld = MAX( dcldfra , - precipfraccld(i) ) 
  !--We remove it, because cldfra is guaranteed to be > 0 (the MAX is activated
  !--if cldfra < 0)
  dprecipfraccld = dcldfra


  !--If the cloud extends
  IF ( dprecipfraccld .GT. 0. ) THEN
    !--If there is no CS precip, nothing happens.
    !--If there is, we reattribute some of the CS precip flux
    !--to the cloud precip flux, proportionnally to the
    !--decrease of the CS precip fraction
    IF ( precipfracclr(i) .LE. 0. ) THEN
      drainclr = 0.
      dsnowclr = 0.
    ELSE
      drainclr = dprecipfracclr / precipfracclr(i) * rainclr(i) 
      dsnowclr = dprecipfracclr / precipfracclr(i) * snowclr(i) 
    ENDIF
  !--If the cloud narrows
  ELSEIF ( dprecipfraccld .LT. 0. ) THEN
    !--We reattribute some of the cloudy precip flux
    !--to the CS precip flux, proportionnally to the
    !--decrease of the cloud precip fraction
    draincld = dprecipfraccld / precipfraccld(i) * raincld(i) 
    dsnowcld = dprecipfraccld / precipfraccld(i) * snowcld(i) 
    drainclr = - draincld
    dsnowclr = - dsnowcld
  !--If the cloud stays the same or if there is no cloud above and
  !--in the current layer, nothing happens
  ELSE
    drainclr = 0.
    dsnowclr = 0.
  ENDIF

  !--We add the tendencies
  !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
  precipfraccld(i) = precipfraccld(i) + dprecipfraccld
  precipfracclr(i) = precipfracclr(i) + dprecipfracclr
  rainclr(i) = MAX(0., rainclr(i) + drainclr)
  snowclr(i) = MAX(0., snowclr(i) + dsnowclr)
  raincld(i) = MAX(0., raincld(i) - drainclr)
  snowcld(i) = MAX(0., snowcld(i) - dsnowclr)
  
  !--If vertical heterogeneity is taken into account, we use
  !--the "true" volume fraction instead of a modified 
  !--surface fraction (which is larger and artificially
  !--reduces the in-cloud water).
  IF ( ( iflag_cloudth_vert .GE. 3 ) .AND. ( iflag_rain_incloud_vol .EQ. 1 ) ) THEN
    eff_cldfra = ctot_vol(i)
  ELSE
    eff_cldfra = cldfra(i)
  ENDIF 


  !--Start precipitation growth processes

  !--If the cloud is big enough, the precipitation processes activate
  ! TODO met on seuil_neb ici ?
  IF ( cldfra(i) .GE. seuil_neb ) THEN

    !---------------------------------------------------------
    !--           COLLECTION AND AGGREGATION
    !---------------------------------------------------------
    !--Collection: processus through which rain collects small liquid droplets
    !--in suspension, and add it to the rain flux
    !--Aggregation: same for snow (precip flux) and ice crystals (in suspension)
    !--Those processes are treated before autoconversion because we do not
    !--want to collect/aggregate the newly formed fluxes, which already
    !--"saw" the cloud as they come from it
    !--The formulas come from Muench and Lohmann 2020

    !--gamma_col: tuning coefficient [-]
    !--rho_rain: volumic mass of rain [kg/m3]
    !--r_rain: size of the rain droplets [m]
    !--Eff_rain_liq: efficiency of the collection process [-] (between 0 and 1)
    !--dqlcol is a gridbox-mean quantity, as is qliq and raincld. They are
    !--divided by respectively eff_cldfra, eff_cldfra and precipfraccld to
    !--get in-cloud mean quantities. The two divisions by eff_cldfra are
    !--then simplified.

    !--The collection efficiency is perfect.
    Eff_rain_liq = 1.
    coef_col = gamma_col * 3. / 4. / rho_rain / r_rain * Eff_rain_liq
    IF ( raincld(i) .GT. 0. ) THEN
      !--Exact explicit version, which does not need a barrier because of
      !--the exponential decrease
      dqlcol = qliq(i) * ( EXP( - dtime * coef_col * raincld(i) / precipfraccld(i) ) - 1. )

      !--Add tendencies
      qliq(i) = qliq(i) + dqlcol
      raincld(i) = raincld(i) - dqlcol * dhum_to_dflux(i)

      !--Diagnostic tendencies
      dqrcol(i)  = - dqlcol  / dtime
    ENDIF

    !--Same as for aggregation
    !--Eff_snow_liq formula:
    !--it s a product of a collection efficiency and a sticking efficiency
    ! Milbrandt and Yau formula that gives very low values:
    ! Eff_snow_ice = 0.05 * EXP( 0.1 * ( temp(i) - RTT ) )
    ! Lin 1983's formula 
    Eff_snow_ice = EXP( 0.025 * MIN( ( temp(i) - RTT ), 0.) ) 
    !--rho_snow formula follows Brandes et al. 2007 (JAMC)
    rho_snow = 1.e3 * 0.178 * ( r_snow * 2. * 1000. )**(-0.922)
    coef_agg = gamma_agg * 3. / 4. / rho_snow / r_snow * Eff_snow_ice
    IF ( snowcld(i) .GT. 0. ) THEN
      !--Exact explicit version, which does not need a barrier because of
      !--the exponential decrease
      dqiagg = qice(i) * ( EXP( - dtime * coef_agg * snowcld(i) / precipfraccld(i) ) - 1. )

      !--Add tendencies
      qice(i) = qice(i) + dqiagg
      snowcld(i) = snowcld(i) - dqiagg * dhum_to_dflux(i)

      !--Diagnostic tendencies
      dqsagg(i)  = - dqiagg  / dtime
    ENDIF


    !---------------------------------------------------------
    !--                  AUTOCONVERSION
    !---------------------------------------------------------
    !--Autoconversion converts liquid droplets/ice crystals into
    !--rain drops/snowflakes. It relies on the formulations by
    !--Sundqvist 1978.

    !--If we are in a convective point, we have different parameters
    !--for the autoconversion
    IF ( ptconv(i) ) THEN
        qthresh_auto_rain = cld_lc_con
        qthresh_auto_snow = cld_lc_con_snow

        tau_auto_rain = cld_tau_con
        !--tau for snow depends on the ice fraction in mixed-phase clouds     
        tau_auto_snow = tau_auto_snow_max &
                      + ( tau_auto_snow_min - tau_auto_snow_max ) * ( 1. - icefrac(i) )

        expo_auto_rain = cld_expo_con
        expo_auto_snow = cld_expo_con
    ELSE
        qthresh_auto_rain = cld_lc_lsc
        qthresh_auto_snow = cld_lc_lsc_snow

        tau_auto_rain = cld_tau_lsc
        !--tau for snow depends on the ice fraction in mixed-phase clouds     
        tau_auto_snow = tau_auto_snow_max &
                      + ( tau_auto_snow_min - tau_auto_snow_max ) * ( 1. - icefrac(i) )

        expo_auto_rain = cld_expo_lsc
        expo_auto_snow = cld_expo_lsc
    ENDIF


    ! Liquid water quantity to remove according to (Sundqvist, 1978)
    ! dqliq/dt = -qliq/tau * ( 1-exp(-(qliqincld/qthresh)**2) )
    !
    !--And same formula for ice
    !
    !--We first treat the second term (with exponential) in an explicit way
    !--and then treat the first term (-q/tau) in an exact way

    dqlauto = - qliq(i) * ( 1. - exp( - dtime / tau_auto_rain * ( 1. - exp( &
               - ( qliq(i) / eff_cldfra / qthresh_auto_rain ) ** expo_auto_rain ) ) ) )

    dqiauto = - qice(i) * ( 1. - exp( - dtime / tau_auto_snow * ( 1. - exp( &
               - ( qice(i) / eff_cldfra / qthresh_auto_snow ) ** expo_auto_snow ) ) ) )


    !--Barriers so that we don't create more rain/snow
    !--than there is liquid/ice
    dqlauto = MAX( - qliq(i), dqlauto )
    dqiauto = MAX( - qice(i), dqiauto )

    !--Add tendencies
    qliq(i) = qliq(i) + dqlauto
    qice(i) = qice(i) + dqiauto
    raincld(i) = raincld(i) - dqlauto * dhum_to_dflux(i)
    snowcld(i) = snowcld(i) - dqiauto * dhum_to_dflux(i)

    !--Diagnostic tendencies
    dqsauto(i) = - dqiauto / dtime
    dqrauto(i) = - dqlauto / dtime


    !---------------------------------------------------------
    !--                  RIMING
    !---------------------------------------------------------
    !--Process which converts liquid droplets in suspension into
    !--snow because of the collision between
    !--those and falling snowflakes.
    !--The formula comes from Muench and Lohmann 2020
    !--NB.: this process needs a temperature adjustment

    !--Eff_snow_liq formula: following Ferrier 1994,
    !--assuming 1
    Eff_snow_liq = 1.0
    !--rho_snow formula follows Brandes et al. 2007 (JAMC)
    rho_snow = 1.e3 * 0.178 * ( r_snow * 2. * 1000. )**(-0.922)
    coef_rim = gamma_rim * 3. / 4. / rho_snow / r_snow * Eff_snow_liq
    IF ( snowcld(i) .GT. 0. ) THEN
      !--Exact version, which does not need a barrier because of
      !--the exponential decrease
      dqlrim = qliq(i) * ( EXP( - dtime * coef_rim * snowcld(i) / precipfraccld(i) ) - 1. )

      !--Add tendencies
      !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
      qliq(i) = qliq(i) + dqlrim
      snowcld(i) = snowcld(i) - dqlrim * dhum_to_dflux(i)

      !--Temperature adjustment with the release of latent
      !--heat because of solid condensation
      temp(i) = temp(i) - dqlrim * RLMLT / RCPD &
                        / ( 1. + RVTMP2 * qtot(i) )

      !--Diagnostic tendencies
      dqsrim(i)  = - dqlrim  / dtime
    ENDIF

  ENDIF ! cldfra .GE. seuil_neb

ENDDO ! loop on klon


!--Re-calculation of saturation specific humidity
!--because riming changed temperature
CALL calc_qsat_ecmwf(klon, temp, qzero, pplay, RTT, 0, .FALSE., qsat, dqsat)

DO i = 1, klon

  !---------------------------------------------------------
  !--                  MELTING
  !---------------------------------------------------------
  !--Process through which snow melts into rain. 
  !--The formula is homemade.
  !--NB.: this process needs a temperature adjustment

  !--dqsmelt_max         : maximum snow melting so that temperature
  !--                      stays higher than 273 K [kg/kg]
  !--capa_snowflake      : capacitance of a snowflake, equal to
  !--                      the radius if the snowflake is a sphere [m]
  !--temp_wetbulb        : wet-bulb temperature [K]
  !--snow_fallspeed      : snow fall velocity (in clear/cloudy sky) [m/s]
  !--air_thermal_conduct : thermal conductivity of the air [J/m/K/s]
  !--gamma_melt          : tuning parameter for melting [-]
  !--nb_snowflake        : number of snowflakes (in clear/cloudy air) [-]

  IF ( ( snowclr(i) + snowcld(i) ) .GT. 0. ) THEN
    !--Computed according to
    !--Cpdry * Delta T * (1 + (Cpvap/Cpdry - 1) * qtot) = Lfusion * Delta q
    dqsmelt_max = MIN(0., ( RTT - temp(i) ) / RLMLT * RCPD &
                        * ( 1. + RVTMP2 * qtot(i) ))
   
    !--Initialisation 
    dqsclrmelt = 0.
    dqscldmelt = 0.

    !--We assume that the snowflakes are spherical
    capa_snowflake = r_snow
    !--Thermal conductivity of the air, empirical formula from Beard and Pruppacher (1971)
    air_thermal_conduct = ( 5.69 + 0.017 * ( temp(i) - RTT ) ) * 1.e-3 * 4.184    
    !--rho_snow formula follows Brandes et al. 2007 (JAMC)
    rho_snow = 1.e3 * 0.178 * ( r_snow * 2. * 1000. )**(-0.922)

    !--In clear air
    IF ( ( snowclr(i) .GT. 0. ) .AND.  ( precipfracclr(i) .GT. 0. ) ) THEN
      !--Formula for the wet-bulb temperature from ECMWF (IFS)
      !--The vapor used is the vapor in the clear sky
      temp_wetbulb = temp(i) &
                   - ( qsat(i) - ( qvap(i) - cldfra(i) * qsat(i) ) / ( 1. - cldfra(i) ) ) &
                   * ( 1329.31 + 0.0074615 * ( pplay(i) - 0.85e5 ) &
                   - 40.637 * ( temp(i) - 275. ) )
      !--Calculated according to
      !-- flux = velocity_snowflakes * nb_snowflakes * volume_snowflakes * rho_snow
      nb_snowflake_clr = snowclr(i) / precipfracclr(i) / snow_fallspeed_clr &
                       / ( 4. / 3. * RPI * r_snow**3. * rho_snow )
      dqsclrmelt = - nb_snowflake_clr * 4. * RPI * air_thermal_conduct &
                   * capa_snowflake / RLMLT * gamma_melt &
                   * MAX(0., temp_wetbulb - RTT) * dtime
      
      !--Barrier to limit the melting flux to the clr snow flux in the mesh
      dqsclrmelt = MAX( dqsclrmelt , -snowclr(i) / dhum_to_dflux(i))
    ENDIF


    !--In cloudy air
    IF ( ( snowcld(i) .GT. 0. ) .AND. ( precipfraccld(i) .GT. 0. ) ) THEN
      !--As the air is saturated, the wet-bulb temperature is equal to the
      !--temperature
      temp_wetbulb = temp(i)
      !--Calculated according to
      !-- flux = velocity_snowflakes * nb_snowflakes * volume_snowflakes * rho_snow
      nb_snowflake_cld = snowcld(i) / precipfraccld(i) / snow_fallspeed_cld &
                       / ( 4. / 3. * RPI * r_snow**3. * rho_snow )
      dqscldmelt = - nb_snowflake_cld * 4. * RPI * air_thermal_conduct &
                   * capa_snowflake / RLMLT * gamma_melt &
                   * MAX(0., temp_wetbulb - RTT) * dtime

      !--Barrier to limit the melting flux to the cld snow flux in the mesh
      dqscldmelt = MAX(dqscldmelt , - snowcld(i) / dhum_to_dflux(i))
   ENDIF
    

    !--Barrier on temperature. If the total melting flux leads to a
    !--positive temperature, it is limited to keep temperature above 0 degC.
    !--It is activated if the total is LOWER than the max
    !--because everything is negative
    dqstotmelt = dqsclrmelt + dqscldmelt
    IF ( dqstotmelt .LT. dqsmelt_max ) THEN
      !--We redistribute the max melted snow keeping
      !--the clear/cloud partition of the melted snow
      dqsclrmelt = dqsmelt_max * dqsclrmelt / dqstotmelt
      dqscldmelt = dqsmelt_max * dqscldmelt / dqstotmelt
      dqstotmelt = dqsmelt_max

    ENDIF

    !--Add tendencies
    !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
    rainclr(i) = MAX(0., rainclr(i) - dqsclrmelt * dhum_to_dflux(i))
    raincld(i) = MAX(0., raincld(i) - dqscldmelt * dhum_to_dflux(i))
    snowclr(i) = MAX(0., snowclr(i) + dqsclrmelt * dhum_to_dflux(i))
    snowcld(i) = MAX(0., snowcld(i) + dqscldmelt * dhum_to_dflux(i))

    !--Temperature adjustment with the release of latent
    !--heat because of melting
    temp(i) = temp(i) + dqstotmelt * RLMLT / RCPD &
                      / ( 1. + RVTMP2 * qtot(i) )

    !--Diagnostic tendencies
    dqrmelt(i) = - dqstotmelt / dtime
    dqsmelt(i) = dqstotmelt / dtime

  ENDIF


  !---------------------------------------------------------
  !--                  FREEZING
  !---------------------------------------------------------
  !--Process through which rain freezes into snow. 
  !-- We parameterize it as a 2 step process:
  !--first: freezing following collision with ice crystals
  !--second: immersion freezing following (inspired by Bigg 1953)
  !--the latter is parameterized as an exponential decrease of the rain
  !--water content with a homemade formulya
  !--This is based on a caracteritic time of freezing, which
  !--exponentially depends on temperature so that it is
  !--equal to 1 for temp_nowater (see below) and is close to
  !--0 for RTT (=273.15 K).
  !--NB.: this process needs a temperature adjustment
  !--dqrfreez_max : maximum rain freezing so that temperature
  !--              stays lower than 273 K [kg/kg]
  !--tau_freez    : caracteristic time of freezing [s]
  !--gamma_freez  : tuning parameter [s-1]
  !--alpha_freez  : tuning parameter for the shape of the exponential curve [-]
  !--temp_nowater : temperature below which no liquid water exists [K] (about -40 degC)

  IF ( ( rainclr(i) + raincld(i) ) .GT. 0. ) THEN

 
    !--1st step: freezing following collision with ice crystals
    !--Sub-process of freezing which quantifies the collision between
    !--ice crystals in suspension and falling rain droplets.
    !--The rain droplets freeze, becoming graupel, and carrying
    !--the ice crystal (which acted as an ice nucleating particle).
    !--The formula is adapted from the riming formula.
    !--it works only in the cloudy part

    dqifreez = 0.
    dqrtotfreez_step1 = 0.

    IF ( ( qice(i) .GT. 0. ) .AND. ( cldfra(i) .GT. 0. ) .AND. &
         ( raincld(i) .GT. 0. ) .AND. ( precipfraccld(i) .GT. 0. ) ) THEN
      dqrclrfreez = 0.
      dqrcldfreez = 0.

      !--Computed according to
      !--Cpdry * Delta T * (1 + (Cpvap/Cpdry - 1) * qtot) = Lfusion * Delta q
      dqrfreez_max = MIN(0., ( temp(i) - RTT ) / RLMLT * RCPD &
                         * ( 1. + RVTMP2 * qtot(i) ))
 

      !--The collision efficiency is assumed unity 
      Eff_rain_ice = 1.
      coef_freez = gamma_freez * 3. / 4. / rho_rain / r_rain * Eff_rain_ice
      !--Exact version, which does not need a barrier because of
      !--the exponential decrease.
      dqifreez = qice(i) * ( EXP( - dtime * coef_freez * raincld(i) / precipfraccld(i) ) - 1. )

      !--We add the part of rain water that freezes, limited by a temperature barrier
      !--This quantity is calculated assuming that the number of drop that freeze correspond to the number
      !--of crystals collected (and assuming uniform distributions of ice crystals and rain drops)
      !--The ice specific humidity that collide with rain is dqi = dNi 4/3 PI rho_ice r_ice**3
      !--The rain that collide with ice is, similarly, dqr = dNr 4/3 PI rho_rain r_rain**3
      !--The assumption above corresponds to dNi = dNr, i.e.,
      !-- dqr = dqi * (4/3 PI rho_rain * r_rain**3) / (4/3 PI rho_ice * r_ice**3)
      !--Dry density [kg/m3]
      rho = pplay(i) / temp(i) / RD
      !--r_ice formula from Sun and Rikus (1999)
      r_ice = 1.e-6 * ( 45.8966 * ( qice(i) / cldfra(i) * rho * 1e3 )**0.2214 &
            + 0.7957 * ( qice(i) / cldfra(i) * rho * 1e3 )**0.2535 * ( temp(i) - RTT + 190. ) ) / 2.
      dqrcldfreez = dqifreez * rho_rain * r_rain**3. / ( rho_ice * r_ice**3. )
      dqrcldfreez = MAX(dqrcldfreez, - raincld(i) / dhum_to_dflux(i))
      dqrcldfreez = MAX(dqrcldfreez, dqrfreez_max) 
      dqrtotfreez_step1 = dqrcldfreez

      !--Add tendencies
      !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
      qice(i) = qice(i) + dqifreez
      raincld(i) = MAX(0., raincld(i) + dqrcldfreez * dhum_to_dflux(i))
      snowcld(i) = MAX(0., snowcld(i) - dqrcldfreez * dhum_to_dflux(i) - dqifreez * dhum_to_dflux(i))
      temp(i) = temp(i) - dqrtotfreez_step1 * RLMLT / RCPD &
                        / ( 1. + RVTMP2 * qtot(i) )

    ENDIF
    
    !-- Second step immersion freezing of rain drops
    !-- with a homemade timeconstant depending on temperature

    dqrclrfreez = 0.
    dqrcldfreez = 0.
    dqrtotfreez_step2 = 0.
    !--Computed according to
    !--Cpdry * Delta T * (1 + (Cpvap/Cpdry - 1) * qtot) = Lfusion * Delta q

    dqrfreez_max = MIN(0., ( temp(i) - RTT ) / RLMLT * RCPD &
                         * ( 1. + RVTMP2 * qtot(i) ))

    
    tau_freez = 1. / ( beta_freez &
              * EXP( - alpha_freez * ( temp(i) - temp_nowater ) / ( RTT - temp_nowater ) ) )


    !--In clear air
    IF ( rainclr(i) .GT. 0. ) THEN
      !--Exact solution of dqrain/dt = -qrain/tau_freez
      dqrclrfreez = rainclr(i) / dhum_to_dflux(i) * ( EXP( - dtime / tau_freez ) - 1. )
    ENDIF

    !--In cloudy air
    IF ( raincld(i) .GT. 0. ) THEN
      !--Exact solution of dqrain/dt = -qrain/tau_freez
      dqrcldfreez = raincld(i) / dhum_to_dflux(i) * ( EXP( - dtime / tau_freez ) - 1. )
    ENDIF

    !--temperature barrier step 2
    !--It is activated if the total is LOWER than the max
    !--because everything is negative
    dqrtotfreez_step2 = dqrclrfreez + dqrcldfreez
  
    IF ( dqrtotfreez_step2 .LT. dqrfreez_max ) THEN
      !--We redistribute the max freezed rain keeping
      !--the clear/cloud partition of the freezing rain
      dqrclrfreez = dqrfreez_max * dqrclrfreez / dqrtotfreez_step2
      dqrcldfreez = dqrfreez_max * dqrcldfreez / dqrtotfreez_step2
      dqrtotfreez_step2 = dqrfreez_max
    ENDIF


    !--Add tendencies
    !--The MAX is needed because in some cases, the flux can be slightly negative (numerical precision)
    rainclr(i) = MAX(0., rainclr(i) + dqrclrfreez * dhum_to_dflux(i))
    raincld(i) = MAX(0., raincld(i) + dqrcldfreez * dhum_to_dflux(i))
    snowclr(i) = MAX(0., snowclr(i) - dqrclrfreez * dhum_to_dflux(i))
    snowcld(i) = MAX(0., snowcld(i) - dqrcldfreez * dhum_to_dflux(i))


    !--Temperature adjustment with the uptake of latent
    !--heat because of freezing
    temp(i) = temp(i) - dqrtotfreez_step2 * RLMLT / RCPD &
                      / ( 1. + RVTMP2 * qtot(i) )

    !--Diagnostic tendencies
    dqrtotfreez = dqrtotfreez_step1 + dqrtotfreez_step2          
    dqrfreez(i) = dqrtotfreez / dtime
    dqsfreez(i) = -(dqrtotfreez + dqifreez) / dtime

  ENDIF



  !--If the local flux of rain+snow in clear/cloudy air is lower than rain_int_min,
  !--we reduce the precipiration fraction in the clear/cloudy air so that the new
  !--local flux of rain+snow is equal to rain_int_min.
  !--Here, rain+snow is the gridbox-mean flux of precip.
  !--Therefore, (rain+snow)/precipfrac is the local flux of precip.
  !--If the local flux of precip is lower than rain_int_min, i.e.,
  !-- (rain+snow)/precipfrac < rain_int_min , i.e., 
  !-- (rain+snow)/rain_int_min < precipfrac , then we want to reduce
  !--the precip fraction to the equality, i.e., precipfrac = (rain+snow)/rain_int_min.
  !--Note that this is physically different than what is proposed in LTP thesis.
  precipfracclr(i) = MIN( precipfracclr(i), ( rainclr(i) + snowclr(i) ) / rain_int_min )
  precipfraccld(i) = MIN( precipfraccld(i), ( raincld(i) + snowcld(i) ) / rain_int_min )

  !--Calculate outputs
  rain(i) = rainclr(i) + raincld(i)
  snow(i) = snowclr(i) + snowcld(i)

  !--Diagnostics
  !--BEWARE this is indeed a diagnostic: this is an estimation from
  !--the value of the flux at the bottom interface of the mesh and
  !--and assuming an upstream numerical calculation
  !--If ok_radocond_snow is TRUE, then the diagnostic qsnowdiag is
  !--used for computing the total ice water content in the mesh
  !--for radiation only
  qraindiag(i) = ( rainclr(i) / ( pplay(i) / RD / temp(i) ) / rain_fallspeed_clr &
                 + raincld(i) / ( pplay(i) / RD / temp(i) ) / rain_fallspeed_cld )
  qsnowdiag(i) = ( snowclr(i) / ( pplay(i) / RD / temp(i) ) / snow_fallspeed_clr &
                 + snowcld(i) / ( pplay(i) / RD / temp(i) ) / snow_fallspeed_cld )


ENDDO ! loop on klon


END SUBROUTINE poprecip_postcld

END MODULE lmdz_lscp_poprecip