add blending and associated pre-processing tools. (#38)

CMakeLists.txt

@@ -76,4 +76,5 @@ add_subdirectory(FV3_ensmean_recenter)
 add_subdirectory(bufrsnd)
 add_subdirectory(rtma_minmaxtrh)
 add_subdirectory(pm)
+add_subdirectory(blending)
 

blending/CMakeLists.txt

@@ -0,0 +1,7 @@
+cmake_minimum_required(VERSION 3.0)
+project(blending)
+
+add_subdirectory(chgres_winds)
+add_subdirectory(raymond)
+add_subdirectory(remap_dwinds)
+add_subdirectory(remap_scalar)

blending/chgres_winds/CMakeLists.txt

@@ -0,0 +1,17 @@
+cmake_minimum_required(VERSION 3.0)
+project(chgres_winds)
+
+# Find the path to libiomp5.so
+find_library(LIBIOMP5_PATH NAMES libiomp5.so)
+set(FCODE ${CMAKE_CURRENT_SOURCE_DIR}/chgres_winds.f90)
+
+# Compile the Fortran code using f2py
+execute_process(
+  COMMAND $ENV{BLENDINGPYTHON} -m numpy.f2py -DF2PY_REPORT_ATEXIT --fcompiler=intelem --build-dir ${CMAKE_CURRENT_BINARY_DIR} -c --f90flags='-fopenmp' -L${LIBIOMP5_PATH} -liomp5 -m chgres_winds ${FCODE}
+    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+    RESULT_VARIABLE F2PY_RESULT
+)
+
+#install chgres_winds.so
+file(GLOB_RECURSE SOFILE "${CMAKE_CURRENT_BINARY_DIR}/*chgres_winds*.so")
+install(FILES ${SOFILE} DESTINATION ${CMAKE_INSTALL_LIBDIR} RENAME chgres_winds.so)

blending/chgres_winds/chgres_winds.f90

@@ -0,0 +1,293 @@
+ subroutine main(gridx,gridy,u_s,v_s,u_w,v_w,ud,vd)
+ use ISO_FORTRAN_ENV
+ use omp_lib
+! use, intrinsic :: ieee_arithmetic
+ implicit none
+ integer, parameter     :: f_p = selected_real_kind(20)
+ integer, parameter     :: f_d  = 8
+ real(kind=8), dimension(:,:,:), intent(IN)    :: u_s,v_s
+ real(kind=8), dimension(:,:,:), intent(INOUT) :: ud
+ real(kind=8), dimension(:,:,:), intent(IN)    :: u_w,v_w
+ real(kind=8), dimension(:,:,:), intent(INOUT) :: vd
+ real(kind=8), dimension(:,:),   intent(IN)    :: gridx,gridy !(nlon,nlat)
+ real(kind=8) :: inner_prod
+ real(kind=8), dimension(2):: p1,p2,p3
+ real(kind=8), dimension(3):: e1,e2,ex,ey
+ integer :: i,j,k
+ integer :: is,ie,js,je,ks,ke
+ real(kind=8), parameter:: pi=4.D0*DATAN(1.D0)
+ real(kind=8), parameter:: deg2rad=pi/180.D0
+ real(kind=8), parameter:: rad2deg=180.D0/pi
+ real(kind=8), parameter:: r360=360.D0
+
+ is = lbound(ud, dim=1)
+ ie = ubound(ud, dim=1)
+ js = lbound(vd, dim=2)
+ je = ubound(vd, dim=2)
+ ks = lbound(ud, dim=3)
+ ke = ubound(ud, dim=3)
+
+!$OMP parallel do default(none) &
+!$OMP          shared(is,ie,js,je,ks,ke,gridx,gridy,ud,vd,u_s,v_s,u_w,v_w) &
+!$OMP          private(p1,p2,p3,e1,e2,ex,ey)
+ do k=ks,ke
+   do j=js,je+1 !lats
+     do i=is,ie !lons
+       if(i .lt. ie-1 .and. j .le. je) then
+       p1(1) = gridx(i,  j)*deg2rad
+       p1(2) = gridy(i,  j)*deg2rad
+       p2(1) = gridx(i+1,j)*deg2rad
+       p2(2) = gridy(i+1,j)*deg2rad
+       endif
+       call mid_pt_sphere(p1, p2, p3)
+       call get_unit_vect2(p1, p2, e1)
+       call get_latlon_vector(p3, ex, ey)
+       ud(i,j,k) = u_s(i,j,k)*inner_prod(e1,ex) + v_s(i,j,k)*inner_prod(e1,ey)
+       if ( isnan(ud(i,j,k))) then
+         write(*,*) "in ud loop: NaN at i,j,k",i,j,k
+         stop
+       end if
+     enddo
+   enddo
+   do j=js,je
+     do i=is,ie+1
+       if(i .lt. ie .and. j .le. je-1) then
+       p1(1) = gridx(i,  j)*deg2rad
+       p1(2) = gridy(i,  j)*deg2rad
+       p2(1) = gridx(i,j+1)*deg2rad
+       p2(2) = gridy(i,j+1)*deg2rad
+       endif
+       call mid_pt_sphere(p1, p2, p3)
+       call get_unit_vect2(p1, p2, e2)
+       call get_latlon_vector(p3, ex, ey)
+       vd(i,j,k) = u_w(i,j,k)*inner_prod(e2,ex) + v_w(i,j,k)*inner_prod(e2,ey)
+       if ( isnan(vd(i,j,k))) then
+         write(*,*) "in vd loop: NaN at i,j,k",i,j,k
+         stop
+       end if
+     enddo
+   enddo
+ enddo
+
+ return
+ end subroutine main
+
+   real*8 function inner_prod(v1, v2)
+       integer, parameter     :: f_p = selected_real_kind(20)
+       integer, parameter     :: f_d  = 8
+       real(kind=8),intent(in):: v1(3), v2(3)
+       real(kind=f_p) :: vp1(3), vp2(3), prod16
+       integer k
+
+         do k=1,3
+            vp1(k) = real(v1(k),kind=f_p)
+            vp2(k) = real(v2(k),kind=f_p)
+         enddo
+         prod16 = vp1(1)*vp2(1) + vp1(2)*vp2(2) + vp1(3)*vp2(3)
+         inner_prod = prod16
+
+  end function inner_prod
+
+ subroutine mid_pt_sphere(p1, p2, pm)
+ use ISO_FORTRAN_ENV
+ implicit none
+      integer, parameter     :: f_d  = 8
+      real(kind=8) , intent(IN)  :: p1(2), p2(2)
+      real(kind=8) , intent(OUT) :: pm(2)
+!------------------------------------------
+      real(kind=8) e1(3), e2(3), e3(3)
+
+      call latlon2xyz(p1, e1)
+      call latlon2xyz(p2, e2)
+      call mid_pt3_cart(e1, e2, e3)
+      call cart_to_latlon(1, e3, pm(1), pm(2))
+
+ end subroutine mid_pt_sphere
+
+ subroutine get_unit_vect2( e1, e2, uc )
+ use ISO_FORTRAN_ENV
+ implicit none
+   integer, parameter     :: f_d  = 8
+   real(kind=8), intent(in) :: e1(2), e2(2)
+   real(kind=8), intent(out):: uc(3) !< unit vector e1--->e2
+! Local:
+   real(kind=8), dimension(3):: pc, p1, p2, p3
+
+! RIGHT_HAND system:
+   call latlon2xyz(e1, p1)
+   call latlon2xyz(e2, p2)
+
+   call mid_pt3_cart(p1, p2,  pc)
+   call vect_cross(p3, p2, p1)
+   call vect_cross(uc, pc, p3)
+   call normalize_vect( uc )
+
+ end subroutine get_unit_vect2
+
+ subroutine get_latlon_vector(pp, elon, elat)
+ use ISO_FORTRAN_ENV
+ implicit none
+ integer, parameter     :: f_d  = 8
+ real(kind=8), intent(IN)  :: pp(2)
+ real(kind=8), intent(OUT) :: elon(3), elat(3)
+
+         elon(1) = -SIN(pp(1))
+         elon(2) =  COS(pp(1))
+         elon(3) =  0.0
+         elat(1) = -SIN(pp(2))*COS(pp(1))
+         elat(2) = -SIN(pp(2))*SIN(pp(1))
+!!! RIGHT_HAND
+         elat(3) =  COS(pp(2))
+! Left-hand system needed to be consistent with rest of the codes
+!        elat(3) = -COS(pp(2))
+
+ end subroutine get_latlon_vector
+
+!-----------------------------------------------------------------------
+
+!>@brief The subroutine 'normalize_vect' makes 'e' a unit vector.
+ subroutine normalize_vect(e)
+ use ISO_FORTRAN_ENV
+ implicit none
+
+ integer, parameter     :: f_p = selected_real_kind(20)
+ integer, parameter     :: f_d  = 8
+ real(kind=8), intent(inout):: e(3)
+ real(kind=f_p):: pdot
+ integer k
+
+    pdot = e(1)**2 + e(2)**2 + e(3)**2
+    pdot = sqrt( pdot )
+
+    do k=1,3
+       e(k) = e(k) / pdot
+    enddo
+
+ end subroutine normalize_vect
+
+!>@brief The subroutine 'vect_cross performs cross products
+!! of 3D vectors: e = P1 X P2
+ subroutine vect_cross(e, p1, p2)
+ use ISO_FORTRAN_ENV
+ implicit none
+ integer, parameter     :: f_d  = 8
+ real(kind=8), intent(in) :: p1(3), p2(3)
+ real(kind=8), intent(out):: e(3)
+
+      e(1) = p1(2)*p2(3) - p1(3)*p2(2)
+      e(2) = p1(3)*p2(1) - p1(1)*p2(3)
+      e(3) = p1(1)*p2(2) - p1(2)*p2(1)
+
+ end subroutine vect_cross
+
+!>@brief The subroutine 'latlon2xyz' maps (lon, lat) to (x,y,z)
+ subroutine latlon2xyz(p, e)
+ use ISO_FORTRAN_ENV
+ implicit none
+
+ integer, parameter     :: f_p = selected_real_kind(20)
+ integer, parameter     :: f_d  = 8
+ real(kind=8), intent(in) :: p(2)
+ real(kind=8), intent(out):: e(3)
+
+ integer n
+ real(kind=f_p):: q(2)
+ real(kind=f_p):: e1, e2, e3
+
+ logical, save       :: first_time = .true.
+ integer, save       :: id_latlon
+
+    do n=1,2
+       q(n) = p(n)
+    enddo
+
+    e1 = cos(q(2)) * cos(q(1))
+    e2 = cos(q(2)) * sin(q(1))
+    e3 = sin(q(2))
+!-----------------------------------
+! Truncate to the desired precision:
+!-----------------------------------
+    e(1) = e1
+    e(2) = e2
+    e(3) = e3
+
+ end subroutine latlon2xyz
+
+
+
+ subroutine mid_pt3_cart(p1, p2, e)
+ use ISO_FORTRAN_ENV
+ implicit none
+       integer, parameter     :: f_p = selected_real_kind(20)
+       integer, parameter     :: f_d  = 8
+       real(kind=8), intent(IN)  :: p1(3), p2(3)
+       real(kind=8), intent(OUT) :: e(3)
+!
+       real(kind=f_p):: q1(3), q2(3)
+       real(kind=f_p):: dd, e1, e2, e3
+       integer k
+
+       do k=1,3
+          q1(k) = p1(k)
+          q2(k) = p2(k)
+       enddo
+
+       e1 = q1(1) + q2(1)
+       e2 = q1(2) + q2(2)
+       e3 = q1(3) + q2(3)
+
+       dd = sqrt( e1**2 + e2**2 + e3**2 )
+       e1 = e1 / dd
+       e2 = e2 / dd
+       e3 = e3 / dd
+
+       e(1) = e1
+       e(2) = e2
+       e(3) = e3
+
+ end subroutine mid_pt3_cart
+
+ subroutine cart_to_latlon(np, q, xs, ys)
+ use ISO_FORTRAN_ENV
+ implicit none
+! vector version of cart_to_latlon1
+  integer, intent(in):: np
+  integer, parameter     :: f_p = selected_real_kind(20)
+  integer, parameter     :: f_d  = 8
+  real(kind=8), intent(inout):: q(3,np)
+  real(kind=8), intent(inout):: xs(np), ys(np)
+! local
+  real(kind=8), parameter:: esl=1.d-10
+  real(kind=8), parameter:: pi=4.D0*DATAN(1.D0)
+  real(kind=f_p):: p(3)
+  real(kind=f_p):: dist, lat, lon
+  integer i,k
+
+  do i=1,np
+     do k=1,3
+        p(k) = q(k,i)
+     enddo
+     dist = sqrt(p(1)**2 + p(2)**2 + p(3)**2)
+     do k=1,3
+        p(k) = p(k) / dist
+     enddo
+
+     if ( (abs(p(1))+abs(p(2)))  < esl ) then
+          lon = real(0.,kind=f_p)
+     else
+          lon = atan2( p(2), p(1) )   ! range [-pi,pi]
+     endif
+
+     if ( lon < 0.) lon = real(2.,kind=f_p)*pi + lon
+! RIGHT_HAND system:
+     lat = asin(p(3))
+
+     xs(i) = lon
+     ys(i) = lat
+! q Normalized:
+     do k=1,3
+        q(k,i) = p(k)
+     enddo
+  enddo
+
+ end  subroutine cart_to_latlon

blending/raymond/CMakeLists.txt

@@ -0,0 +1,15 @@
+cmake_minimum_required(VERSION 3.0)
+project(raymond)
+
+set(FCODE ${CMAKE_CURRENT_SOURCE_DIR}/raymond.f)
+
+# Compile the Fortran code using f2py
+execute_process(
+    COMMAND $ENV{BLENDINGPYTHON} -m numpy.f2py  -DF2PY_REPORT_ATEXIT --fcompiler=intelem --build-dir ${CMAKE_CURRENT_BINARY_DIR} -c -m raymond ${FCODE}
+    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+    RESULT_VARIABLE F2PY_RESULT
+)
+
+#install raymond.so
+file(GLOB_RECURSE SOFILE "${CMAKE_CURRENT_BINARY_DIR}/*raymond*.so")
+install(FILES ${SOFILE} DESTINATION ${CMAKE_INSTALL_LIBDIR} RENAME raymond.so)