C $Header: /u/gcmpack/MITgcm/pkg/sphere/sphere.F,v 1.3 2007/10/09 00:11:39 jmc Exp $ C $Name: $ #include "CPP_OPTIONS.h" c ================================================================== c c shpere.F: Routines that handle the projection onto sherical c harmonics. c c Routines: c c o adfsc4dat - Adjoint routine of fsc4dat. c o shc2grid - Evaluate a spherical harmonics model on a c regular grid. c o shc4grid - Evaluate a spherical harmonics model on a c regular grid. c c o shcrotate - s/r used for the sph analysis ... c o shc2zone - s/r used for the sph analysis ... c o shc4zone - s/r used for the sph analysis ... c o fsc2dat - s/r used for the sph analysis ... c o frsbase - s/r used for the sph analysis ... c o helmholtz - s/r used for the sph analysis ... c o recur_z - s/r used for the sph analysis ... c c o shcError - Print error message and stop program (see below). c c IMSL Routines: c c o fftrb - Compute the real periodic sequence from its Fourier c coefficients. c --> replace by NAGLIB routine. C06... c c Platform-specific: c c M abort - FORTRAN intrinsic on SUN and, presumably, on CRAY to c exit the program if an error condition is met. c c --> Replaced ABORT by shcError ( Christian Eckert MIT 22-Mar-2000 ) c c Note: c ===== c c Where code is written in lower case letters, changes were c introduced. The changes are mainly related to F90 syntax. c Additionally, I replaced the call to the intrinsic *AMOD* c by its generic name *MOD*. ( Ch.E. 25-May-1999 ) c c c Documentation: c c c ================================================================== SUBROUTINE FSC4DAT(N,FSC) IMPLICIT NONE INTEGER N REAL FSC(N) REAL SCALE integer i #ifdef USE_SPH_PROJECTION CALL FFTRF(N,FSC,FSC) #else do i=1,n fsc(i) = 0.0 enddo #endif SCALE = 2.0/FLOAT(N) FSC(1) = FSC(1)/FLOAT(N) CE FSC(2:N-1:2) = FSC(2:N-1:2)*SCALE CE FSC(3:N :2) =-FSC(3:N: 2)*SCALE ! change sign of SINE coeffs. do i = 2,n-1,2 fsc( i ) = fsc( i )*scale enddo do i = 3,n,2 fsc( i ) = -fsc( i )*scale ! change sign of sine coeffs. enddo IF (MOD(N,2).EQ.0) THEN FSC(N) = FSC(N)/FLOAT(N) ENDIF RETURN END c ================================================================== subroutine adfsc4dat(n,adfsc) implicit none integer n real adfsc(n) integer i #ifdef USE_SPH_PROJECTION call fftrb(n,adfsc,adfsc) #else do i=1,n adfsc(i) = 0.0 enddo #endif do i=1,n adfsc(i) = adfsc(i)/float(n) enddo end c ================================================================== SUBROUTINE SHC2GRID( LMAX, SHC, NLON, NLAT, GRID, P ) C C --- Evaluate a spherical harmonics model on a regular grid. C C SHC ! spherical harmonic coefficients in the order of C ! C ! C00 : SHC(1) C ! S11 C10 C11 : SHC(2) SHC(3) SHC(4) C ! S22 S21 C20 C21 C22 : SHC(5) SHC(6) SHC(7) SHC(8) SHC(9) C ! C ! i.e., C(L,M) = SHC(L*L+L+1+M) C ! S(L,M) = SHC(L*L+L+1-M) C NLAT ! number of latitude (zonal) lines. C NLON ! number of longitude (meridional) lines. C ! C GRID ! grid values in the order of C ! ((west to east),south to north) C ! ((LON=1,NLON),LAT=1,NLAT) C ! grid values are defined on the "centers" C ! of geographically regular equi-angular blocks. C C --- Coded by Dr. Myung-Chan Kim, Center for Space Research, 1997 C University of Texas at Austin, Austin, Texas, 78712 C IMPLICIT NONE INTEGER LMAX ! INPUT max. degree REAL SHC((1+LMAX)*(1+LMAX)) ! INPUT spherical harmonics INTEGER NLAT ! INPUT number of latitude lines. INTEGER NLON ! INPUT number of longitude lines. REAL GRID(NLON,NLAT) ! OUTPUT grid values REAL P((LMAX+1)*(LMAX+2)/2) ! work space INTEGER NLONLMT PARAMETER (NLONLMT=10000) REAL HS (NLONLMT) REAL HN (NLONLMT) REAL DLAT, ANGLE, XLAT1, XLAT2 INTEGER LATS, LATN ce integer js, jn integer i ce IF(NLON.GT.NLONLMT) CALL ABORT('NLON.GT.NLONLMT') ce IF(NLON.LT.LMAX*2 ) CALL ABORT('NLON.LT.LMAX*2 ') IF(NLON.GT.NLONLMT) CALL shcError('NLON.GT.NLONLMT',1) IF(NLON.LT.LMAX*2 ) CALL shcError('NLON.LT.LMAX*2 ',1) DLAT = 180.0/FLOAT(NLAT) ANGLE = 180.0/FLOAT(NLON) CALL SHCROTATE(LMAX,SHC,ANGLE) DO LATS = 1,(NLAT+1)/2 LATN = NLAT-(LATS-1) XLAT2 =-90.0 + FLOAT(LATS)*DLAT XLAT1 = XLAT2-DLAT do i=1,nlon hs(i) = 0.0 hn(i) = 0.0 enddo CALL SHC2ZONE( LMAX, SHC, XLAT1, XLAT2, HS, HN, P ) CALL FSC2DAT( NLON, HS ) CALL FSC2DAT( NLON, HN ) do i=1,nlon grid(i,lats) = hs(i) grid(i,latn) = hn(i) enddo ENDDO CALL SHCROTATE(LMAX,SHC,-ANGLE) RETURN END c ================================================================== SUBROUTINE SHC4GRID( LMAX, SHC, NLON, NLAT, GRID, P ) IMPLICIT NONE INTEGER LMAX ! INPUT max. degree REAL SHC((1+LMAX)*(1+LMAX)) ! OUTPUT spherical harmonics INTEGER NLAT ! INPUT number of latitude lines. INTEGER NLON ! INPUT number of longitude lines. REAL GRID(NLON,NLAT) ! INPUT grid values REAL P((LMAX+1)*(LMAX+2)/2) ! work space INTEGER NLONLMT PARAMETER (NLONLMT=10000) REAL HS (NLONLMT) REAL HN (NLONLMT) REAL DLAT, ANGLE, XLAT1, XLAT2 INTEGER LATS, LATN ce integer js, jn integer i IF(NLON.GT.NLONLMT) THEN PRINT *, 'NLON = ', NLON PRINT *, 'NLONLMT = ', NLONLMT ce CALL ABORT('NLON.GT.NLONLMT') CALL shcError('NLON.GT.NLONLMT',1) END IF IF(NLON.LT.LMAX*2 ) THEN PRINT *, 'NLON = ', NLON PRINT *, 'LMAX = ', LMAX ce CALL ABORT('NLON.LT.LMAX*2') CALL shcError('NLON.LT.LMAX*2',1) END IF DLAT = 180.0/FLOAT(NLAT) ANGLE = 180.0/FLOAT(NLON) DO LATS = 1,(NLAT+1)/2 LATN = NLAT-(LATS-1) do i = 1,nlon hs(i) = grid(i,lats) hn(i) = grid(i,latn) enddo CALL FSC4DAT(NLON,HS) CALL FSC4DAT(NLON,HN) XLAT2 =-90.0 + FLOAT(LATS)*DLAT XLAT1 = XLAT2-DLAT CALL SHC4ZONE(LMAX,SHC,XLAT1,XLAT2,HS,HN,P) ENDDO CALL SHCROTATE(LMAX,SHC,-ANGLE) RETURN END c ================================================================== SUBROUTINE SHCROTATE(LMAX,SHC,ANGLE) IMPLICIT NONE INTEGER LMAX ! max. degree of spherical harmonics. REAL SHC((1+LMAX)*(1+LMAX)) ! spherical harmonic coeffs. REAL ANGLE ! in degree. INTEGER LMAXLMT PARAMETER (LMAXLMT=10000) REAL H(LMAXLMT*2+1) INTEGER K, L, M REAL SINA, COSA, C, S if (mod(angle,360.0) .ne. 0.0) then CALL FRSBASE(ANGLE,H,1,1+LMAX+LMAX) DO M = 0,LMAX IF(M.EQ.0) THEN SINA = 0.0 COSA = 1.0 ELSE COSA = H(M+M) SINA = H(M+M+1) ENDIF DO L = M,LMAX K = L*L+L+1 C = SHC(K+M) S = SHC(K-M) SHC(K+M) = COSA*C + SINA*S SHC(K-M) =-SINA*C + COSA*S ENDDO ENDDO ENDIF RETURN END c ================================================================== SUBROUTINE SHC2ZONE(LMAX,SHC,XLAT1,XLAT2,HS,HN,P) IMPLICIT NONE INTEGER LMAX ! INPUT max. degree of spher. harmonics. REAL SHC((1+LMAX)*(1+LMAX)) ! INPUT spherical harmonic coeffs. REAL XLAT1 ! INPUT latitude. REAL XLAT2 ! INPUT latitude. REAL HS(1+LMAX+LMAX) ! OUTPUT REAL HN(1+LMAX+LMAX) ! OUTPUT REAL P((LMAX+1)*(LMAX+2)/2) ! INPUT INTEGER LMAXLMT PARAMETER (LMAXLMT=5000) REAL FACT(0:LMAXLMT) ! INTEGER LMAX1, J, K, L, M REAL DEG2RAD, SLAT REAL A, B, E, F, Q, DA, DB, XLAT ce IF (LMAX.GT.LMAXLMT) CALL ABORT('LMAX.GT.LMAXLMT') cph IF (LMAX.GT.LMAXLMT) CALL shcError('LMAX.GT.LMAXLMT',1) DEG2RAD = ACOS(-1.0)/180.0 XLAT = 0.5*(XLAT1+XLAT2) do k = 1,lmax fact(k) = 1.0 enddo SLAT = SIN(XLAT*DEG2RAD) CALL HELMHOLTZ(LMAX,SLAT,P) LMAX1 = LMAX+1 K = 0 DO M = 0,LMAX A = 0.0 B = 0.0 E = 0.0 F = 0.0 DO L = M,LMAX-1,2 K = K+1 J = L*L+L+1 Q = FACT(L)*P(K) DA = Q*SHC(J+M) DB = Q*SHC(J-M) A = A+DA B = B+DB E = E+DA F = F+DB K = K+1 J = J+L+L+2 Q = FACT(L+1)*P(K) DA = Q*SHC(J+M) DB = Q*SHC(J-M) A = A+DA B = B+DB E = E-DA F = F-DB ENDDO IF(MOD(LMAX-L,2).EQ.0) THEN K = K+1 J = LMAX*LMAX+LMAX+1 Q = FACT(LMAX)*P(K) DA = Q*SHC(J+M) DB = Q*SHC(J-M) A = A+DA B = B+DB E = E+DA F = F+DB ENDIF IF(M.EQ.0) THEN HS(1) = A HN(1) = E ELSE HS(M+M ) = A HS(M+M+1) = B HN(M+M ) = E HN(M+M+1) = F ENDIF ENDDO RETURN END c ================================================================== SUBROUTINE SHC4ZONE(LMAX,SHC,XLAT1,XLAT2,HS,HN,P) IMPLICIT NONE INTEGER LMAX REAL SHC((1+LMAX)*(1+LMAX)) ! spherical harmonic coeffs. REAL XLAT1 ! latitude. REAL XLAT2 ! latitude. REAL P((LMAX+1)*(LMAX+2)/2) C C - output - C REAL HS(1+LMAX+LMAX) REAL HN(1+LMAX+LMAX) C C - work space - C INTEGER LMAXLMT PARAMETER (LMAXLMT=5000) REAL XLAT, SIN0, SIN1, SIN2, SCALE REAL DEG2RAD, CE, CO, SE, SO INTEGER I, J, K, L, M INTEGER JP, I1, I2 ce IF(LMAX.GT.LMAXLMT) CALL ABORT('LMAX.GT.LMAXLMT') cph IF(LMAX.GT.LMAXLMT) CALL shcError('LMAX.GT.LMAXLMT',1) IF(XLAT1 .LT. -90.0+1.E-10) THEN do i = 1,(1+lmax)*(1+lmax) shc(i) = 0.0 enddo ENDIF XLAT = 0.5*(XLAT1+XLAT2) DEG2RAD = ACOS(-1.0)/180.0 SIN0 = SIN(XLAT *DEG2RAD) SIN1 = SIN(AMAX1(-90.0,XLAT1)*DEG2RAD) SIN2 = SIN(AMIN1( 0.0,XLAT2)*DEG2RAD) SCALE = (SIN2 - SIN1)*0.25 do i = 1,lmax+lmax+1 hs(i) = hs(i)*scale hn(i) = hn(i)*scale enddo CALL HELMHOLTZ(LMAX,SIN0,P) I = 0 cadj loop = parallel DO M = 0,LMAX IF (M .EQ. 0) THEN J = 1 JP = 1 ELSE J = 2*M JP = J+1 ENDIF CE = HS(J)+HN(J) CO = HS(J)-HN(J) SE = HS(J)+HN(JP) SO = HS(J)-HN(JP) I1 = I+1 I2 = I+2 cadj loop = parallel DO L = M,LMAX-1,2 K = L*L+L+1 SHC(K+M) = SHC(K+M) + P(I1) * CE SHC(K-M) = SHC(K-M) + P(I1) * SE K = K+L+L+2 SHC(K+M) = SHC(K+M) + P(I2) * CO SHC(K-M) = SHC(K-M) + P(I2) * SO ENDDO I = I + 2 IF(MOD(LMAX-M,2).NE.1) THEN I = I+1 K = LMAX*LMAX+LMAX+1 SHC(K+M) = SHC(K+M) + P(I) * CE SHC(K-M) = SHC(K-M) + P(I) * SE ENDIF ENDDO RETURN END c ================================================================== subroutine fsc2dat(n,fsc) c c --- this routine are coded to clarify the imsl's fftrf/fftrb. c implicit none integer n real fsc(n) integer i do i = 2,n-1,2 fsc(i ) = fsc(i )*0.5 fsc(i+1) = -fsc(i+1)*0.5 ! change sign of sine coeffs. enddo #ifdef USE_SPH_PROJECTION call fftrb(n,fsc,fsc) #else do i = 1,n fsc( i ) = 0.0 enddo #endif return end c ================================================================== SUBROUTINE FRSBASE(A,H,I,J) C C --- Given A (in degree), return H(I:J) = 1,COS(A),SIN(A)..... C IMPLICIT NONE REAL A REAL H(1) INTEGER I, J REAL DEG2RAD, T, ARG INTEGER N, L DEG2RAD = ACOS(-1.0)/180.0 N = J-I+1 IF(N.LE.0) RETURN H(I) = 1.0 IF(N.EQ.1) RETURN ARG = A * DEG2RAD H(I+1) = COS(ARG) IF(N.EQ.2) RETURN H(I+2) = SIN(ARG) IF(N.EQ.3) RETURN T = H(I+1)+H(I+1) H(I+3) = T*H(I+1) - 1.0 IF(N.EQ.4) RETURN H(I+4) = T*H(I+2) IF(N.EQ.5) RETURN DO L = I+5,J H(L) = T*H(L-2)-H(L-4) END DO RETURN END c ================================================================== SUBROUTINE HELMHOLTZ(LMAX,S,P) C C --- compute the fully normalized associated legendre polynomials. C IMPLICIT NONE INTEGER LMAX ! INPUT max. degree. REAL S ! INPUT sin(latitude). REAL P(1) ! OUTPUT assoc. legendre polynomials. INTEGER LMAXLMT PARAMETER(LMAXLMT=10800) REAL A(0:LMAXLMT) REAL ISECT(0:LMAXLMT) REAL X(LMAXLMT) REAL Y(LMAXLMT+LMAXLMT) REAL Z(LMAXLMT) INTEGER LMAXOLD DATA LMAXOLD/-1/ SAVE LMAXOLD, ISECT, X, Y, Z REAL C, CM, AK INTEGER K, L, N, M IF(LMAX.NE.LMAXOLD) THEN ISECT(0) = 1 DO L = 1,LMAX X(L) = SNGL(1.0D0/DSQRT(DBLE(FLOAT(4*L*L-1)))) Y(L) = SNGL(DSQRT(DBLE(FLOAT(L)))) Y(L+LMAX) = SNGL(DSQRT(DBLE(FLOAT(L+LMAX)))) ISECT(L) = L*LMAX-L*(L-3)/2+1 ENDDO CALL RECUR_Z(Z,LMAX) LMAXOLD = LMAX ENDIF C = SNGL(DSQRT(1.0D0-DBLE(S)*DBLE(S))) CM = 1.0 P(1) = 1.0 DO M = 1,LMAX K = ISECT(M) CM = C * CM P(K) = Z(M) * CM ENDDO N = 1 DO M = 0,LMAX-N K = ISECT(M)+N L = N+M AK = X(L)*Y(L+M)*Y(N) P(K) = S*P(K-1)/AK A(M) = AK ENDDO DO N = 2,LMAX DO M = 0,LMAX-N K = ISECT(M)+N L = N+M AK = X(L)*Y(L+M)*Y(N) P(K) =(S*P(K-1)-P(K-2)*A(M))/AK A(M) = AK ENDDO ENDDO RETURN END c ================================================================== SUBROUTINE RECUR_Z(Z,LMAX) C C --- Coefficients for fully normalized sectorial Legendre polynomials. C IMPLICIT NONE INTEGER LMAX REAL Z(LMAX) DOUBLE PRECISION ZZ INTEGER L ZZ = 2.0D0 DO L = 1,LMAX ZZ = ZZ * DBLE(FLOAT(L+L+1))/DBLE(FLOAT(L+L)) Z(L) = SNGL(DSQRT(ZZ)) ENDDO RETURN END c ================================================================== subroutine shcError( errstring, ierr ) c c --- Print an error message and exit. Written in order to replace c the machine specific routine ABORT. c ( Christian Eckert MIT 22-Mar-2000 ) c implicit none integer ierr character*(*) errstring if ( ierr .ne. 0 ) then print* print*,' sphere: ',errstring print* stop ' ... program stopped.' endif return end c ==================================================================