+ diff --exclude=.git --exclude=version_clubb_core.txt --exclude=version_silhs.txt -r clubb clubb_release
diff '--exclude=.git' '--exclude=version_clubb_core.txt' '--exclude=version_silhs.txt' -r clubb/src/CLUBB_core/setup_clubb_pdf_params.F90 clubb_release/src/CLUBB_core/setup_clubb_pdf_params.F90
371a372,374
>
> ice_supersat_frac_1(j,:) = pdf_params(j)%ice_supersat_frac_1
> ice_supersat_frac_2(j,:) = pdf_params(j)%ice_supersat_frac_2
405,410c408
<
< do j = 1, ngrdcol
< ice_supersat_frac_1(j,:) = pdf_params(j)%ice_supersat_frac_1
< ice_supersat_frac_2(j,:) = pdf_params(j)%ice_supersat_frac_2
< end do
<
---
>
496c494
< ! call stat_update_var( ihmp2_zt(i), hydrometp2_zt(:,i), stats_zt )
---
> ! call stat_update_var( ihmp2_zt(i), hydrometp2_zt(:,i), stats_zt )
584c582
< ! call stat_update_var( iprecip_frac, precip_frac, stats_zt )
---
> ! call stat_update_var( iprecip_frac, precip_frac, stats_zt )
593c591
< ! call stat_update_var( iprecip_frac_1, precip_frac_1, stats_zt )
---
> ! call stat_update_var( iprecip_frac_1, precip_frac_1, stats_zt )
602c600
< ! call stat_update_var( iprecip_frac_2, precip_frac_2, stats_zt )
---
> ! call stat_update_var( iprecip_frac_2, precip_frac_2, stats_zt )
611c609
< ! call stat_update_var( iNcnm, Ncnm, stats_zt )
---
> ! call stat_update_var( iNcnm, Ncnm, stats_zt )
697,712d694
<
< do k = 2, nz, 1
< do j = 1, ngrdcol
< call calc_cholesky_corr_mtx_approx &
< ( pdf_dim, corr_array_1_n(j,:,:,k), & ! intent(in)
< corr_cholesky_mtx_1(j,:,:,k), corr_array_1_n(j,:,:,k) ) ! intent(out)
< end do
< end do
<
< do k = 2, nz, 1
< do j = 1, ngrdcol
< call calc_cholesky_corr_mtx_approx &
< ( pdf_dim, corr_array_2_n(j,:,:,k), & ! intent(in)
< corr_cholesky_mtx_2(j,:,:,k), corr_array_2_n(j,:,:,k) ) ! intent(out)
< end do
< end do
715,772c697,710
<
< if ( .not. l_interp_prescribed_params .and. l_fix_w_chi_eta_correlations &
< .and. .not. l_calc_w_corr ) then
<
< ! When the flags are set this way, the correlation matrices do not vary with any vertical
< ! values, and instead are determined entirely by prescribed values. This results in there
< ! being only two unique correlation matrices, one for when the grid box is in cloud
< ! and one for when it is not. So instead of setting up correlation matrices for all
< ! grid boxes then calculating their Cholesky decomps, we can simply set up two correlation
< ! matrices, one for in cloud and one for out cloud, calculate the corresponding
< ! Cholesky decompositions, then use the value of rc at each grid box to determine whether
< ! we assign the in cloud or out of cloud matrices to that grid box.
< call calc_corr_norm_and_cholesky_factor( nz, ngrdcol, pdf_dim, iiPDF_type, &
< rc_1, rc_2, &
< corr_array_n_cloud, corr_array_n_below, &
< corr_array_1_n, corr_array_2_n, &
< corr_cholesky_mtx_1, corr_cholesky_mtx_2 )
<
< else
<
< ! The correlation matrices can vary with vertical values. So we need to set the
< ! correlation matrices up for each grid box, then find the Cholesky decomp for each
< ! grid box individually. This is very computationally expensive.
<
< call comp_corr_norm( nz, pdf_dim, ngrdcol, wm_zt(:,:), rc_1(:,:), rc_2(:,:), &
< cloud_frac_1(:,:), cloud_frac_2(:,:), mixt_frac(:,:), &
< precip_frac_1(:,:), precip_frac_2(:,:), &
< wpNcnp_zt(:,:), wphydrometp_zt(:,:,:), &
< mu_x_1(:,:,:), mu_x_2(:,:,:), &
< sigma_x_1(:,:,:), sigma_x_2(:,:,:), &
< sigma_x_1_n(:,:,:), sigma_x_2_n(:,:,:), &
< corr_array_n_cloud, corr_array_n_below, &
< pdf_params, &
< iiPDF_type, &
< l_calc_w_corr, &
< l_fix_w_chi_eta_correlations, &
< corr_array_1_n(:,:,:,:), corr_array_2_n(:,:,:,:) )
<
< ! Compute choleksy factorization for the correlation matrix of 1st PDF component
< do k = 2, nz, 1
< do j = 1, ngrdcol
< call Cholesky_factor( pdf_dim, corr_array_1_n(j,:,:,k), & ! In
< corr_array_scaling, corr_cholesky_mtx_1(j,:,:,k), & ! Out
< l_corr_array_scaling ) ! Out
< end do
< end do
<
< ! Compute choleksy factorization for the correlation matrix of 2nd PDF component
< do k = 2, nz, 1
< do j = 1, ngrdcol
< call Cholesky_factor( pdf_dim, corr_array_2_n(j,:,:,k), & ! In
< corr_array_scaling, corr_cholesky_mtx_2(j,:,:,k), & ! Out
< l_corr_array_scaling ) ! Out
< end do
< end do
<
< end if
<
---
>
> call comp_corr_norm( nz, pdf_dim, ngrdcol, wm_zt(:,:), rc_1(:,:), rc_2(:,:), &
> cloud_frac_1(:,:), cloud_frac_2(:,:), mixt_frac(:,:), &
> precip_frac_1(:,:), precip_frac_2(:,:), &
> wpNcnp_zt(:,:), wphydrometp_zt(:,:,:), &
> mu_x_1(:,:,:), mu_x_2(:,:,:), &
> sigma_x_1(:,:,:), sigma_x_2(:,:,:), &
> sigma_x_1_n(:,:,:), sigma_x_2_n(:,:,:), &
> corr_array_n_cloud, corr_array_n_below, &
> pdf_params, &
> iiPDF_type, &
> l_calc_w_corr, &
> l_fix_w_chi_eta_correlations, &
> corr_array_1_n(:,:,:,:), corr_array_2_n(:,:,:,:) )
777,781c715,719
< sigma_x_1_n(:,:,:), sigma_x_2_n(:,:,:), &
< sigma2_on_mu2_ip_1(:,:,:), sigma2_on_mu2_ip_2(:,:,:), &
< corr_array_1_n(:,:,:,:), &
< corr_array_2_n(:,:,:,:), &
< corr_array_1(:,:,:,:), corr_array_2(:,:,:,:) )
---
> sigma_x_1_n(:,:,:), sigma_x_2_n(:,:,:), &
> sigma2_on_mu2_ip_1(:,:,:), sigma2_on_mu2_ip_2(:,:,:), &
> corr_array_1_n(:,:,:,:), &
> corr_array_2_n(:,:,:,:), &
> corr_array_1(:,:,:,:), corr_array_2(:,:,:,:) )
815a754,793
> if ( l_diagnose_correlations ) then
>
> do k = 2, nz, 1
> do j = 1, ngrdcol
> call calc_cholesky_corr_mtx_approx &
> ( pdf_dim, corr_array_1_n(j,:,:,k), & ! intent(in)
> corr_cholesky_mtx_1(j,:,:,k), corr_array_1_n(j,:,:,k) ) ! intent(out)
> end do
> end do
>
> do k = 2, nz, 1
> do j = 1, ngrdcol
> call calc_cholesky_corr_mtx_approx &
> ( pdf_dim, corr_array_2_n(j,:,:,k), & ! intent(in)
> corr_cholesky_mtx_2(j,:,:,k), corr_array_2_n(j,:,:,k) ) ! intent(out)
> end do
> end do
>
> else
>
> ! Compute choleksy factorization for the correlation matrix (out of
> ! cloud)
> do k = 2, nz, 1
> do j = 1, ngrdcol
> call Cholesky_factor( pdf_dim, corr_array_1_n(j,:,:,k), & ! In
> corr_array_scaling, corr_cholesky_mtx_1(j,:,:,k), & ! Out
> l_corr_array_scaling ) ! Out
> end do
> end do
>
> do k = 2, nz, 1
> do j = 1, ngrdcol
> call Cholesky_factor( pdf_dim, corr_array_2_n(j,:,:,k), & ! In
> corr_array_scaling, corr_cholesky_mtx_2(j,:,:,k), & ! Out
> l_corr_array_scaling ) ! Out
> end do
> end do
>
> end if
>
1145,1189d1122
<
< !=============================================================================
< subroutine calc_corr_norm_and_cholesky_factor( nz, ngrdcol, pdf_dim, iiPDF_type, &
< rc_1, rc_2, &
< corr_array_n_cloud, corr_array_n_below, &
< corr_array_1_n, corr_array_2_n, &
< corr_cholesky_mtx_1, corr_cholesky_mtx_2 )
<
< ! Description: This subroutine computes the correlation arrays and correlation
< ! Cholesky matrices of PDF vars for both components. Here, we assume that
< ! there are only two unique correlation arrays, which allows us to compute
< ! these two unique arrays and their corresponding Cholesky decompositions,
< ! then use rc to determine which one to assign to each grid column and
< ! vertical level. If the correlation arrays vary based on vertically varying
< ! values, then this subroutine is not appropriate.
< !
< ! References:
< ! https://github.com/larson-group/cam/issues/129#issuecomment-816205563
< !-----------------------------------------------------------------------
<
< use constants_clubb, only: &
< rc_tol, &
< zero
<
< use clubb_precision, only: &
< core_rknd ! Variable(s)
<
< use array_index, only: &
< iiPDF_chi, & ! Variable(s)
< iiPDF_eta, &
< iiPDF_w, &
< iiPDF_Ncn
<
< use model_flags, only: &
< iiPDF_ADG1, & ! Variable(s)
< iiPDF_ADG2, &
< iiPDF_new_hybrid
<
< use pdf_parameter_module, only: &
< pdf_parameter ! Variable(s)
<
< use matrix_operations, only: &
< Cholesky_factor ! Procedure(s)
<
< implicit none
1191,1324d1123
< ! Input Variables
< integer, intent(in) :: &
< nz, & ! Number of vertical levels
< pdf_dim, & ! Number of variables in the corr/mean/stdev arrays
< ngrdcol ! Number of grid columns
<
< real( kind = core_rknd ), dimension(ngrdcol,nz), intent(in) :: &
< rc_1, & ! Mean of r_c (1st PDF component) [kg/kg]
< rc_2 ! Mean of r_c (2nd PDF component) [kg/kg]
<
< real( kind = core_rknd ), dimension(pdf_dim,pdf_dim), intent(in) :: &
< corr_array_n_cloud, & ! Prescribed correlation array in cloud [-]
< corr_array_n_below ! Prescribed correlation array below cloud [-]
<
< integer, intent(in) :: &
< iiPDF_type ! Selected option for the two-component normal (double
< ! Gaussian) PDF type to use for the w, rt, and theta-l (or
< ! w, chi, and eta) portion of CLUBB's multivariate,
< ! two-component PDF.
<
< ! Output Variables
< real( kind = core_rknd ), dimension(ngrdcol,pdf_dim,pdf_dim,nz), intent(out) :: &
< corr_array_1_n, & ! Corr. array (normal space) of PDF vars. (comp. 1) [-]
< corr_array_2_n, & ! Corr. array (normal space) of PDF vars. (comp. 2) [-]
< corr_cholesky_mtx_1, & ! Transposed corr. cholesky matrix, 1st comp. [-]
< corr_cholesky_mtx_2 ! Transposed corr. cholesky matrix, 2nd comp. [-]
<
< ! Local Variables
< real( kind = core_rknd ), dimension(pdf_dim,pdf_dim) :: &
< corr_array_cloud, & ! General in cloud corr. matrix
< corr_array_below, & ! General out of cloud corr. matrix
< corr_cholesky_mtx_cloud, & ! General in cloud Cholesky matrix
< corr_cholesky_mtx_below ! General out of cloud Cholesky matrix
<
< logical :: &
< l_corr_array_scaling ! Dummy variable that we need for calling Cholesky_factor
<
< real( kind = core_rknd ), dimension(pdf_dim) :: &
< corr_array_scaling ! Dummy variable that we need for calling Cholesky_factor
<
< integer :: jvar, j, k ! Indices
<
< logical, parameter :: &
< l_follow_ADG1_PDF_standards = .true.
<
< !-------------------- Begin Code --------------------
<
< ! Initialize correlation arrays with prescribed values
< corr_array_cloud(:,:) = corr_array_n_cloud(:,:)
< corr_array_below(:,:) = corr_array_n_below(:,:)
<
< ! The ADG1 PDF fixes the correlation of w and rt and the correlation of
< ! w and theta_l to be 0, which means the correlation of w and chi and the
< ! correlation of w and eta must also be 0.
< if ( ( iiPDF_type == iiPDF_ADG1 .or. iiPDF_type == iiPDF_ADG2 &
< .or. iiPDF_type == iiPDF_new_hybrid ) &
< .and. l_follow_ADG1_PDF_standards ) then
<
< corr_array_cloud(iiPDF_w,iiPDF_chi) = zero
< corr_array_below(iiPDF_w,iiPDF_chi) = zero
<
< corr_array_cloud(iiPDF_w,iiPDF_eta) = zero
< corr_array_below(iiPDF_w,iiPDF_eta) = zero
<
< end if
<
< ! Ncn is an inherently in-cloud property, so replace out of cloud correlation values
< ! with in cloud ones.
< corr_array_below(iiPDF_Ncn,iiPDF_chi) = corr_array_cloud(iiPDF_Ncn,iiPDF_chi)
< corr_array_below(iiPDF_Ncn,iiPDF_eta) = corr_array_cloud(iiPDF_Ncn,iiPDF_eta)
<
<
< ! Estimates the correlation of the natural logarithm of a
< ! hydrometeor species and eta using the correlation of chi and eta and the
< ! correlation of chi and the natural logarithm of the hydrometeor. This
< ! facilitates the Cholesky decomposability of the correlation array that will
< ! inevitably be decomposed for SILHS purposes. Without this estimation, we
< ! have found that the resulting correlation matrix cannot be decomposed.
< do jvar = iiPDF_Ncn+1, pdf_dim
<
< corr_array_cloud(jvar,iiPDF_eta) = corr_array_cloud(iiPDF_eta,iiPDF_chi) &
< * corr_array_cloud(jvar,iiPDF_chi)
<
< corr_array_below(jvar,iiPDF_eta) = corr_array_below(iiPDF_eta,iiPDF_chi) &
< * corr_array_below(jvar,iiPDF_chi)
<
< end do
<
< ! Calc in cloud Cholesky
< call Cholesky_factor( pdf_dim, corr_array_cloud(:,:), & ! In
< corr_array_scaling, corr_cholesky_mtx_cloud(:,:), & ! Out
< l_corr_array_scaling ) ! Out
<
< ! Calc out of cloud Cholesky
< call Cholesky_factor( pdf_dim, corr_array_below(:,:), & ! In
< corr_array_scaling, corr_cholesky_mtx_below(:,:), & ! Out
< l_corr_array_scaling ) ! Out
<
< ! Use rc_1 to determine which correlation and Cholesky matrices to assign to 1st PDF
< do k = 1, nz
< do j = 1, ngrdcol
<
< if ( rc_1(j,k) > rc_tol ) then
< ! Assign in cloud matrices to 1st PDF component
< corr_array_1_n(j,:,:,k) = corr_array_cloud
< corr_cholesky_mtx_1(j,:,:,k) = corr_cholesky_mtx_cloud
< else
< ! Assign out of cloud matrices to 1st PDF component
< corr_array_1_n(j,:,:,k) = corr_array_below
< corr_cholesky_mtx_1(j,:,:,k) = corr_cholesky_mtx_below
< end if
<
< end do
< end do
<
< ! Use rc_1 to determine which correlation and Cholesky matrices to assign to 2nd PDF
< do k = 1, nz
< do j = 1, ngrdcol
<
< if ( rc_2(j,k) > rc_tol ) then
< ! Assign in cloud matrices to 2nd PDF component
< corr_array_2_n(j,:,:,k) = corr_array_cloud
< corr_cholesky_mtx_2(j,:,:,k) = corr_cholesky_mtx_cloud
< else
< ! Assign out of cloud matrices to 2nd PDF component
< corr_array_2_n(j,:,:,k) = corr_array_below
< corr_cholesky_mtx_2(j,:,:,k) = corr_cholesky_mtx_below
< end if
<
< end do
< end do
<
< end subroutine calc_corr_norm_and_cholesky_factor
<
![[Jenkins]](/jenkins/static/57c8a776/images/svgs/logo.svg){kind=logo}
