subroutine intw(ru,rv,nlath,nlon,nsig,nwrecs,uvfile)
c$$$ subprogram documentation block
c . . . .
c subprogram: intw apply observation operator for winds
c prgmmr: derber org: w/nmc23 date: 91-02-26
c
c abstract: apply observation operator for winds
c
c program history log:
c 91-02-26 derber
c
c input argument list:
c ru - search direction for u-components
c rv - search direction for v-components
c nlath - half the number of latitudes on gaussian grid
c nlon - number of longitudes on gaussian grid
c nsig - number of sigma levels
c nwrecs - number of wind records
c nblk - blocking factor for iunit
c iunit - data scratch file
c
c output argument list:
c ru - results from observation operator (0 for no data)
c rv - results from observation operator (0for no data)
c
c attributes:
c language: cft77
c machine: cray ymp
c
c$$$
c
C-CRA dimension ru(2*nlath+1,nlon+2,nsig)
C-CRA dimension rv(2*nlath+1,nlon+2,nsig)
C-CRA dimension uvfile(*)
C-CRA dimension su(2*nlath+1,nlon+2,nsig)
C-CRA dimension sv(2*nlath+1,nlon+2,nsig)
dimension ru(2*48+1,192+2,28)
dimension rv(2*48+1,192+2,28)
dimension uvfile(*)
dimension su(2*48+1,192+2,28)
dimension sv(2*48+1,192+2,28)
c--------
C-CRA su=ru
c dimension su(2*nlath+1,nlon+2,nsig)
DO ITMP=1,(2*nlath+1)*(nlon+2)*nsig
su(ITMP,1,1)=ru(ITMP,1,1)
ENDDO
C-CRA sv=rv
c dimension sv(2*nlath+1,nlon+2,nsig)
DO ITMP=1,(2*nlath+1)*(nlon+2)*nsig
sv(ITMP,1,1)=rv(ITMP,1,1)
ENDDO
C-CRA ru=0.
c dimension ru(2*nlath+1,nlon+2,nsig)
DO ITMP=1,(2*nlath+1)*(nlon+2)*nsig
ru(ITMP,1,1)=0.
ENDDO
C-CRA rv=0.
c dimension rv(2*nlath+1,nlon+2,nsig)
DO ITMP=1,(2*nlath+1)*(nlon+2)*nsig
rv(ITMP,1,1)=0.
ENDDO
if(nwrecs .eq. 0)return
npp=17
c--------
c
c--------
is=1
do 100 i=1,nwrecs
ngrp=uvfile(is)
is=is+1
cdir$ ivdep
do 101 k=1,ngrp
jlat=uvfile((k-1)*npp+is)
jlon=uvfile((k-1)*npp+is+1)
jsig=uvfile((k-1)*npp+is+2)
jlatp=uvfile((k-1)*npp+is+3)
jlonp=uvfile((k-1)*npp+is+4)
jsigp=uvfile((k-1)*npp+is+5)
wgt000=uvfile((k-1)*npp+is+6)
wgt100=uvfile((k-1)*npp+is+7)
wgt010=uvfile((k-1)*npp+is+8)
wgt110=uvfile((k-1)*npp+is+9)
c udat=uvfile((k-1)*npp+is+14)
c vdat=uvfile((k-1)*npp+is+15)
c aerr=uvfile((k-1)*npp+is+16)
c uges=uvfile((k-1)*npp+is+17)
c vges=uvfile((k-1)*npp+is+18)
c ityp=uvfile((k-1)*npp+is+19)
c if(ityp .ne. 283)then
wgt001=uvfile((k-1)*npp+is+10)
wgt101=uvfile((k-1)*npp+is+11)
wgt011=uvfile((k-1)*npp+is+12)
wgt111=uvfile((k-1)*npp+is+13)
valu=wgt000*su(jlat,jlon,jsig)+wgt100*su(jlatp,jlon,jsig)
* +wgt010*su(jlat,jlonp,jsig)+wgt110*su(jlatp,jlonp,jsig)
* +wgt001*su(jlat,jlon,jsigp)+wgt101*su(jlatp,jlon,jsigp)
* +wgt011*su(jlat,jlonp,jsigp)+wgt111*su(jlatp,jlonp,jsigp)
valv=wgt000*sv(jlat,jlon,jsig)+wgt100*sv(jlatp,jlon,jsig)
* +wgt010*sv(jlat,jlonp,jsig)+wgt110*sv(jlatp,jlonp,jsig)
* +wgt001*sv(jlat,jlon,jsigp)+wgt101*sv(jlatp,jlon,jsigp)
* +wgt011*sv(jlat,jlonp,jsigp)+wgt111*sv(jlatp,jlonp,jsigp)
ru(jlat,jlon,jsig)=ru(jlat,jlon,jsig)+wgt000*valu
ru(jlatp,jlon,jsig)=ru(jlatp,jlon,jsig)+wgt100*valu
ru(jlat,jlonp,jsig)=ru(jlat,jlonp,jsig)+wgt010*valu
ru(jlatp,jlonp,jsig)=ru(jlatp,jlonp,jsig)+wgt110*valu
ru(jlat,jlon,jsigp)=ru(jlat,jlon,jsigp)+wgt001*valu
ru(jlatp,jlon,jsigp)=ru(jlatp,jlon,jsigp)+wgt101*valu
ru(jlat,jlonp,jsigp)=ru(jlat,jlonp,jsigp)+wgt011*valu
ru(jlatp,jlonp,jsigp)=ru(jlatp,jlonp,jsigp)+wgt111*valu
rv(jlat,jlon,jsig)=rv(jlat,jlon,jsig)+wgt000*valv
rv(jlatp,jlon,jsig)=rv(jlatp,jlon,jsig)+wgt100*valv
rv(jlat,jlonp,jsig)=rv(jlat,jlonp,jsig)+wgt010*valv
rv(jlatp,jlonp,jsig)=rv(jlatp,jlonp,jsig)+wgt110*valv
rv(jlat,jlon,jsigp)=rv(jlat,jlon,jsigp)+wgt001*valv
rv(jlatp,jlon,jsigp)=rv(jlatp,jlon,jsigp)+wgt101*valv
rv(jlat,jlonp,jsigp)=rv(jlat,jlonp,jsigp)+wgt011*valv
rv(jlatp,jlonp,jsigp)=rv(jlatp,jlonp,jsigp)+wgt111*valv
c else
c valu=wgt000*su(jlat,jlon,jsig)+wgt100*su(jlatp,jlon,jsig)
c * +wgt010*su(jlat,jlonp,jsig)+wgt110*su(jlatp,jlonp,jsig)
c valv=wgt000*sv(jlat,jlon,jsig)+wgt100*sv(jlatp,jlon,jsig)
c * +wgt010*sv(jlat,jlonp,jsig)+wgt110*sv(jlatp,jlonp,jsig)
c uanl=uges*aerr+valu
c vanl=vges*aerr+valv
c spdanl=sqrt(uanl*uanl+vanl*vanl)
c spdn=(spdanl-aerr*udat)/spdanl
c valu=aerr*uanl*spdn
c valv=aerr*vanl*spdn
c ru(jlat,jlon,jsig)=ru(jlat,jlon,jsig)+wgt000*valu
c ru(jlatp,jlon,jsig)=ru(jlatp,jlon,jsig)+wgt100*valu
c ru(jlat,jlonp,jsig)=ru(jlat,jlonp,jsig)+wgt010*valu
c ru(jlatp,jlonp,jsig)=ru(jlatp,jlonp,jsig)+wgt110*valu
c rv(jlat,jlon,jsig)=rv(jlat,jlon,jsig)+wgt000*valv
c rv(jlatp,jlon,jsig)=rv(jlatp,jlon,jsig)+wgt100*valv
c rv(jlat,jlonp,jsig)=rv(jlat,jlonp,jsig)+wgt010*valv
c rv(jlatp,jlonp,jsig)=rv(jlatp,jlonp,jsig)+wgt110*valv
c end if
101 continue
is=is+npp*ngrp
100 continue
return
end