| 1 |
program driver |
| 2 |
|
| 3 |
c*** This subroutine calculates analysis errors using greg's |
| 4 |
c*** ensemble square root filter, or the traditional ensemble |
| 5 |
c*** kalman filter. |
| 6 |
c*** |
| 7 |
c*** Written by: Jim Hansen |
| 8 |
c*** Last modified: Nov 27, 2002 |
| 9 |
|
| 10 |
implicit none |
| 11 |
|
| 12 |
integer i, j, k, n, ijim, ijuli, isteps, nens, ics |
| 13 |
integer nx, ny, nz, nf, nnx, nny, nnz, mobs |
| 14 |
#include "parameters.h" |
| 15 |
integer iobsloc(mobs) |
| 16 |
real*8 y(n), dydx(n), yobs(mobs), ytmp(n) |
| 17 |
real*8 pert(n), H(mobs,n), R(mobs), yobsfull(n) |
| 18 |
real*8 xens(n,nens), par(n), var(n), ave(n) |
| 19 |
real*8 var_obs, eps, dt, x |
| 20 |
real*8 obserr, analerr, distsub |
| 21 |
real*8 bcup, bcdown, bcin, bcout |
| 22 |
|
| 23 |
common /boundary/ bcup,bcdown,bcin,bcout |
| 24 |
common /dim/ nnx,nny,nnz |
| 25 |
|
| 26 |
external derivsL953d |
| 27 |
|
| 28 |
c*** open some output files |
| 29 |
open(unit=2,file='test.dat',status='unknown') |
| 30 |
open(unit=3,file='fields.dat',status='unknown') |
| 31 |
|
| 32 |
c*** initialisations |
| 33 |
eps=0.2 ! obs error variance |
| 34 |
isteps=1 ! number of steps per obs cycle |
| 35 |
dt=0.05 ! integration stepsize |
| 36 |
do i=1,n |
| 37 |
par(i)=8. ! forcing magnitude |
| 38 |
enddo |
| 39 |
bcup=5. ! boundary conditions |
| 40 |
bcdown=5. |
| 41 |
bcin=5. |
| 42 |
bcout=5. |
| 43 |
nnx=nx |
| 44 |
nny=ny |
| 45 |
nnz=nz |
| 46 |
|
| 47 |
c*** specify observation locations |
| 48 |
do i=1,mobs |
| 49 |
iobsloc(i)=i |
| 50 |
enddo |
| 51 |
do i=1,mobs |
| 52 |
do j=1,n |
| 53 |
H(i,j)=0. |
| 54 |
enddo |
| 55 |
enddo |
| 56 |
do i=1,mobs |
| 57 |
H(i,iobsloc(i))=1. |
| 58 |
enddo |
| 59 |
|
| 60 |
c*** define obs variance |
| 61 |
var_obs=eps**2 |
| 62 |
do i=1,mobs |
| 63 |
R(i)=var_obs |
| 64 |
enddo |
| 65 |
|
| 66 |
c*** make up some initial conditions |
| 67 |
call random2(pert,n) |
| 68 |
do i=1,n |
| 69 |
y(i)=par(1)*pert(i) |
| 70 |
enddo |
| 71 |
|
| 72 |
c*** remove transient |
| 73 |
do i=1,2**14 |
| 74 |
call stepit(x,y,dydx,par,dt,n,isteps,derivsL953d) |
| 75 |
enddo |
| 76 |
|
| 77 |
c*** define an observation about truth |
| 78 |
call random2(pert,n) |
| 79 |
do i=1,n |
| 80 |
yobsfull(i)=y(i)+eps*pert(i) |
| 81 |
enddo |
| 82 |
|
| 83 |
c*** generate an nens member ensemble of initial conditions. |
| 84 |
do i=1,nens |
| 85 |
call random2(pert,n) |
| 86 |
do j=1,n |
| 87 |
xens(j,i)=yobsfull(j)+eps*pert(j) |
| 88 |
enddo |
| 89 |
enddo |
| 90 |
|
| 91 |
c*** loop over initial conditions |
| 92 |
do ijim=1,ics |
| 93 |
|
| 94 |
c*** step truth forward |
| 95 |
call stepit(x,y,dydx,par,dt,n,isteps,derivsL953d) |
| 96 |
|
| 97 |
c*** define an observation about truth |
| 98 |
call random2(pert,n) |
| 99 |
do k=1,n |
| 100 |
yobsfull(k)=y(k)+eps*pert(k) |
| 101 |
enddo |
| 102 |
do k=1,mobs |
| 103 |
yobs(k)=yobsfull(iobsloc(k)) |
| 104 |
enddo |
| 105 |
|
| 106 |
c*** step ensemble forward |
| 107 |
do ijuli=1,nens |
| 108 |
do k=1,n |
| 109 |
ytmp(k)=xens(k,ijuli) |
| 110 |
enddo |
| 111 |
call stepit(x,ytmp,dydx,par,dt,n,isteps,derivsL953d) |
| 112 |
do k=1,n |
| 113 |
xens(k,ijuli)=ytmp(k) |
| 114 |
enddo |
| 115 |
enddo |
| 116 |
|
| 117 |
c*** call the filter |
| 118 |
c call EnSRF3dO(xens,yobs,H,n,mobs,R,nens,nx,ny,nz) |
| 119 |
call EnKF(xens,yobs,H,n,mobs,R,nens) |
| 120 |
|
| 121 |
c*** get ensemble mean and variance |
| 122 |
call ranmean2(xens,ave,n,nens) |
| 123 |
call ranvar(xens,ave,n,nens,var) |
| 124 |
|
| 125 |
c*** calculate obs and anal err |
| 126 |
obserr=distsub(y,yobsfull,n) |
| 127 |
analerr=distsub(y,ave,n) |
| 128 |
write(6,*) "\tOBSERR\t", "\tANALERR" |
| 129 |
write(6,*) obserr, analerr |
| 130 |
write(2,*) ijim, obserr, analerr |
| 131 |
write(3,'(250f18.10)') (ave(i), i=1,n) |
| 132 |
|
| 133 |
call flush() |
| 134 |
|
| 135 |
enddo |
| 136 |
|
| 137 |
return |
| 138 |
end |
Parent Directory
| 
Revision Log
| 
Revision Graph