!
!##################################################################
!##################################################################
!###### ######
!###### SUBROUTINE MEAN_ABSOULTE_ERROR ######
!###### ######
!###### Developed by ######
!###### Center for Analysis and Prediction of Storms ######
!###### University of Oklahoma ######
!###### ######
!##################################################################
!##################################################################
SUBROUTINE MEAN_ABSOLUTE_ERROR (nx, ny, nz, data, model, ilevel, &
jlevel, klevel)
!-----------------------------------------------------------------------
!
! PURPOSE:
!
! Calculates the mean absolute error between the hypothetical data set
! and model forecast and writes this to the "MAE.txt" file
!
! The file "MAE.txt" will always contain the mean absolute error for
! the entire data set at the top of the file. It will also contain
! the mean absolute error based on the i, j, or k levels depending
! upon which of those options are chosen in the "verif.input" file.
!
! Formula:
!
! MAE = (1/n) * SUM[k=1 to n] ABS(y(k) - o(k))
!
! n is the number of forecast / observation pairs
! y(k) is the kth of n forecasts
! o(k) is the kth of n observations
!
!-----------------------------------------------------------------------
!
! AUTHOR: Geoffrey Stano
! 03/01/2002 Initialized.
!
! MODIFICATIONS:
!
! 04/04/2002 Jason J. Levit
! Cosmetic modifications.
!
!-----------------------------------------------------------------------
!
! Variable Declarations.
!
!-----------------------------------------------------------------------
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios
REAL :: maez(nz)
REAL :: maey(ny)
REAL :: maex(nx)
REAL :: mae_all ! variable for MAE of entire data set
REAL :: num
REAL :: data(nx,ny,nz), model(nx,ny,nz)
!@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
!
! Beginning of executable code...
!
!@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
! ##### Calculates mean absolute error for entire data set mae_all #####
mae_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
num = 0.0
num = model(i,j,k) - data(i,j,k)
IF (num < 0) THEN
num = -1 * num ! Performs Absolute Value
END IF
mae_all = mae_all + num
END DO
END DO
END DO
OPEN (UNIT = 11, FILE = "MAE.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (11, FMT = 200)
mae_all = mae_all / (nx * ny * nz)
WRITE (11,'(f7.2)') mae_all
! ########## End of mae_all calculation ##########
! ##### Calculates i-level mean absolute error, meax #####
IF (ilevel == 1) THEN
DO i=1,nx
maex(i) = 0.0
DO j=1,ny
DO k=1,nz
num = 0.0
num = model(i,j,k) - data(i,j,k)
IF (num < 0) THEN
num = -1 * num ! Performs Absolute Value
END IF
maex(i) = maex(i) + num
END DO
END DO
END DO
WRITE (11, FMT = 201)
DO i=1,nx
maex(i) = maex(i) / (ny * nz)
WRITE (11,'(f7.2)') maex(i)
END DO
END IF
! ########## End of i-level mean absolute error ##########
! ##### Calculates j-level mean absolute error, maey #####
IF (jlevel == 1) THEN
DO j=1,ny
maey(j) = 0.0
DO i=1,nx
DO k=1,nz
num = 0.0
num = model(i,j,k) - data(i,j,k)
IF (num < 0) THEN
num = -1 * num ! Performs Absolute Value
END IF
maey(j) = maey(j) + num
END DO
END DO
END DO
WRITE (11, FMT = 202)
DO j=1,ny
maey(j) = maey(j) / (nx * nz)
WRITE (11,'(f7.2)') maey(j)
END DO
END IF
! ######### End of j-level mean absolute error ##########
! ##### Calculates the k-level mean absolute error #####
IF (klevel == 1) THEN
DO k=1,nz
maez(k) = 0.0
DO j=1,ny
DO i=1,nx
num = 0.0
num = model(i,j,k) - data(i,j,k)
IF (num < 0) THEN
num = -1 * num ! Performs Absolute Value
END IF
maez(k) = maez(k) + num
END DO
END DO
END DO
WRITE (11, FMT = 203)
DO k=1,nz
maez(k) = maez(k) / (nx * ny)
WRITE (11,'(f7.2)') maez(k)
END DO
END IF
! ######### End of k-level mean absolute error #########
CLOSE (UNIT = 11)
200 FORMAT ("Mean Absolute Error for all data"/)
201 FORMAT (/"Mean Absolute Error with i-level dependence"/)
202 FORMAT (/"Mean Absolute Error with j-level dependence"/)
203 FORMAT (/"Mean Absolute Error with k-level dependence"/)
END SUBROUTINE Mean_Absolute_Error
!#######################################################################
SUBROUTINE BRIER_SCORE (noe, fcstprob, obsevent)
! Calculates the brier score between the hypothetical probability
! forecast and the observed event and writes this to the "BrierS.txt"
! file
!
! Formula:
!
! BS = (1/n) * SUM[k=1 to n] (fp(k) - v(k))**2
!
! n is the number of forecast / observation pairs
! fp is the forecasted probability for an event
! v is the verification term
! v = 1 if the event did occur
! v = 0 if the event did not occur
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: brier_s(noe) ! brier score
REAL :: fcstprob(noe), obsevent(noe)
DO g=1,noe
brier_s = brier_s + (fcstprob(g) - obsevent(g))**2
END DO
OPEN (UNIT = 22, FILE = "BrierS.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (22, FMT = 300)
brier_s(noe) = (brier_s(noe) / noe)
WRITE (22, FMT = '(f4.2)') brier_s(noe)
CLOSE(22)
300 FORMAT ("Brier Score"/)
END SUBROUTINE Brier_Score
!#######################################################################
SUBROUTINE ROOT_MEAN_SQUARE (nx, ny, nz, data, model, ilevel, & 5
jlevel, klevel,rms_all)
! Calculates the root mean square error between the hypothetical data set
! and the model forecast and writes this to the "RMS.txt" file
!
! Formula:
!
! RMS = SQRT{ (1/n) * SUM[k=1 to n] (y(k) - o(k))**2}
!
! n is the number of forecast / observation pairs
! y(k) is the kth of n forecasts
! o(k) is the kth of n observations
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
INTEGER :: rmscount
REAL :: rms_all
REAL :: rmsz(nz)
REAL :: rmsy(ny)
REAL :: rmsx(nx)
REAL :: data(nx,ny,nz), model(nx,ny,nz)
! ##### Calulates the root mean square error for all data #####
rms_all = 0.0
rmscount=0
DO k=1,nz
DO j=1,ny
DO i=1,nx
IF (data(i,j,k).ne.-99.9) THEN
rms_all = rms_all + (model(i,j,k)-data(i,j,k))**2
rmscount=rmscount+1
END IF
END DO
END DO
END DO
! OPEN (UNIT = 12, FILE = "RMS.txt", IOSTAT = ios, &
! FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
! WRITE (12, FMT = 300)
! print *, 'rmscount=',rmscount
! rms_all = rms_all / (nx * ny * nz)
rms_all = rms_all / real(rmscount)
rms_all = SQRT(rms_all)
! WRITE (12,'(f7.2)') rms_all
! ########## End all root mean square error ##########
! ##### Calculates the i-level root mean square error #####
IF (ilevel == 1) THEN
DO i=1,nx
rmsx(i) = 0.0
DO j=1,ny
DO k=1,nz
rmsx(i) = rmsx(i) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
! WRITE (12, FMT = 301)
DO i = 1,nx
rmsx(i) = rmsx(i) / (ny * nz)
rmsx(i) = SQRT(rmsx(i))
! WRITE (12,'(f7.2)') rmsx(i)
END DO
END IF
! ########## End i-level root mean square error ##########
! ##### Calculates the j-level root mean square error #####
IF (jlevel == 1) THEN
DO j=1,ny
rmsy(j) = 0.0
DO i=1,nx
DO k=1,nz
rmsy(j) = rmsy(j) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
! WRITE (12, FMT = 302)
DO j = 1,ny
rmsy(j) = rmsy(j) / (nx * nz)
rmsy(j) = SQRT(rmsy(j))
! WRITE (12,'(f7.2)') rmsy(j)
END DO
END IF
! ########## End j-level root mean square error ##########
! ##### Calculates the k-level root mean square error #####
IF (klevel == 1) THEN
DO k=1,nz
rmsz(k) = 0.0
DO j=1,ny
DO i=1,nx
rmsz(k) = rmsz(k) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
! WRITE (12, FMT = 303)
DO k = 1,nz
rmsz(k) = rmsz(k) / (nx * ny)
rmsz(k) = SQRT(rmsz(k))
! WRITE (12,'(f7.2)') rmsz(k)
END DO
END IF
! ########## End k-level root mean square error ##########
! CLOSE (UNIT = 12)
300 FORMAT ("Root Mean Square Error for all data"/)
301 FORMAT (/"Root Mean Square Error with i-level dependence"/)
302 FORMAT (/"Root Mean Square Error with j-level dependence"/)
303 FORMAT (/"Root Mean Square Error with k-level dependence"/)
END SUBROUTINE Root_Mean_Square
!#######################################################################
SUBROUTINE MEAN_SQUARE_ERROR (nx, ny, nz, data, model, ilevel, &
jlevel, klevel)
! Calculates the mean square error between the hypothetical data set
! and the model forecast and writes this to the "MSE.txt" file
!
! Formula:
!
! MSE = (1/n) * SUM[k=1 to n] (y(k) -o(k))**2
!
! n is the number of forecast / observation pairs
! y(k) is the kth of n forecasts
! o(k) is the kth of n observations
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: mse_all
REAL :: msez(nz)
REAL :: msey(ny)
REAL :: msex(nx)
REAL :: data(nx,ny,nz), model(nx,ny,nz)
! ##### Calculates the mean square error for all data #####
mse_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
mse_all = mse_all + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
OPEN (UNIT = 13, FILE = "MSE.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (13, FMT = 400)
mse_all = mse_all / (nx * ny * nz)
WRITE (13,'(f7.2)') mse_all
! ########## End mean square error for all data ##########
! ##### Calculates i-level mean square error #####
IF (ilevel == 1) THEN
DO i=1,nx
msex(i) = 0.0
DO j=1,ny
DO k=1,nz
msex(i) = msex(i) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
WRITE (13, FMT = 401)
DO i = 1,nx
msex(i) = msex(i) / (ny * nz)
WRITE (13,'(f7.2)') msex(i)
END DO
END IF
! ######### End i-level mean square error ##########
! ##### Calculates j-level mean square error #####
IF (jlevel == 1) THEN
DO j=1,ny
msey(j) = 0.0
DO k=1,nz
DO i=1,nx
msey(j) = msey(j) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
WRITE (13, FMT = 402)
DO j = 1,ny
msey(j) = msey(j) / (nx * nz)
WRITE (13,'(f7.2)') msey(j)
END DO
END IF
! ########## End j-level mean square error ##########
! ##### Calculates k-level mean square error #####
IF (klevel == 1) THEN
DO k=1,nz
msez(k) = 0.0
DO j=1,ny
DO i=1,nx
msez(k) = msez(k) + (model(i,j,k)-data(i,j,k))**2
END DO
END DO
END DO
WRITE (13, FMT = 403)
DO k = 1,nz
msez(k) = msez(k) / (nx * ny)
WRITE (13,'(f7.2)') msez(k)
END DO
END IF
! ######### End k-level mean square error ##########
CLOSE (UNIT = 13)
400 FORMAT ("Mean Square Error for all data"/)
401 FORMAT (/"Mean Square Error with i-level dependence"/)
402 FORMAT (/"Mean Square Error with j-level dependence"/)
403 FORMAT (/"Mean Square Error with k-level dependence"/)
END SUBROUTINE Mean_Square_Error
!#######################################################################
SUBROUTINE FORECAST_MEAN (nx, ny, nz, model, ilevel, jlevel, &
klevel)
! Calculates the mean of all of the forecast data and writes this
! to the "FcstMean.txt" file
!
! Formula:
!
! Forecast Mean = (Sum of all forecasts) / (Total number of forecasts)
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: fcstmean_all ! mean of all forecasts
REAL :: fcstmeanz(nz)
REAL :: fcstmeany(ny)
REAL :: fcstmeanx(nx)
REAL :: model(nx,ny,nz) ! hypothetical model data
! ##### Calculates the forecast mean for all data #####
fcstmean_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
fcstmean_all = fcstmean_all + model(i,j,k)
END DO
END DO
END DO
fcstmean_all = fcstmean_all / (nx * ny * nz)
OPEN (UNIT = 14, FILE = "FcstMean.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (14, FMT = 500)
WRITE (14,'(f7.2)') fcstmean_all
! ############ End forecast mean for all data ##########
! ##### Calculates i-level forecast mean #####
IF (ilevel == 1) THEN
DO i=1,nx
fcstmeanx(i) = 0.0
DO j=1,ny
DO k=1,nz
fcstmeanx(i) = fcstmeanx(i) + model(i,j,k)
END DO
END DO
END DO
WRITE (14, FMT = 501)
DO i=1,nx
fcstmeanx(i) = fcstmeanx(i) / (ny * nz)
WRITE (14,'(f7.2)') fcstmeanx(i)
END DO
END IF
! ########## End i-level forecast mean ##########
! ##### Calculates j-level forecast mean #####
IF (jlevel == 1) THEN
DO j=1,ny
fcstmeany(j) = 0.0
DO k=1,nz
DO i=1,nx
fcstmeany(j) = fcstmeany(j) + model(i,j,k)
END DO
END DO
END DO
WRITE (14, FMT = 502)
DO j=1,ny
fcstmeany(j) = fcstmeany(j) / (nx * nz)
WRITE (14,'(f7.2)') fcstmeany(j)
END DO
END IF
! ########## End j-level forecast mean ##########
! ##### Calculates k-level forecast mean #####
IF (klevel == 1) THEN
DO k=1,nz
fcstmeanz(k) = 0.0
DO j=1,ny
DO i=1,nx
fcstmeanz(k) = fcstmeanz(k) + model(i,j,k)
END DO
END DO
END DO
WRITE (14, FMT = 503)
DO k=1,nz
fcstmeanz(k) = fcstmeanz(k) / (nx * ny)
WRITE (14,'(f7.2)') fcstmeanz(k)
END DO
END IF
! ########## End k-level forecast mean ##########
CLOSE (UNIT = 14)
500 FORMAT ("Forecast mean for all data"/)
501 FORMAT (/"Forecast mean with i-level dependence"/)
502 FORMAT (/"Forecast mean with j-level dependence"/)
503 FORMAT (/"Forecast mean with k-level dependence"/)
END SUBROUTINE Forecast_Mean
!#######################################################################
SUBROUTINE OBSERVATION_MEAN (nx, ny, nz, data, ilevel, jlevel, &
klevel)
! Calculates the mean of all of the observation data and writes this
! to the "ObsMean.txt" file
!
! Formula:
!
! Observation Mean = (Sum of all obs) / (Total number of obs)
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: obsmean_all ! mean of all observations
REAL :: obsmeanz(nz)
REAL :: obsmeany(ny)
REAL :: obsmeanx(nx)
REAL :: data(nx,ny,nz) ! hypothetical observation data
! ##### Calculates the mean for all the observations #####
obsmean_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
obsmean_all = obsmean_all + data(i,j,k)
END DO
END DO
END DO
obsmean_all = obsmean_all / (nx * ny * nz)
OPEN (UNIT = 15, FILE = "ObsMean.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (15, FMT = 600)
WRITE (15,'(f7.2)') obsmean_all
! ######### End observation mean for all data ##########
! ##### Calculates i-level observation mean #####
IF (ilevel == 1) THEN
DO i=1,nx
obsmeanx(i) = 0.0
DO j=1,ny
DO k=1,nz
obsmeanx(i) = obsmeanx(i) + data(i,j,k)
END DO
END DO
END DO
WRITE (15, FMT = 601)
DO i=1,nx
obsmeanx(i) = obsmeanx(i) / (ny * nz)
WRITE (15,'(f7.2)') obsmeanx(i)
END DO
END IF
! ########## End i-level observation mean #########
! ##### Calculates j-level observation mean #####
IF (jlevel == 1) THEN
DO j=1,ny
obsmeany(j) = 0.0
DO k=1,nz
DO i=1,nx
obsmeany(j) = obsmeany(j) + data(i,j,k)
END DO
END DO
END DO
WRITE (15, FMT = 602)
DO j=1,ny
obsmeany(j) = obsmeany(j) / (nx * nz)
WRITE (15,'(f7.2)') obsmeany(j)
END DO
END IF
! ########## End j-level observation mean ##########
! ##### Calculates k-level observation mean #####
IF (klevel == 1) THEN
DO k=1,nz
obsmeanz(k) = 0.0
DO j=1,ny
DO i=1,nx
obsmeanz(k) = obsmeanz(k) + data(i,j,k)
END DO
END DO
END DO
WRITE (15, FMT = 603)
DO k=1,nz
obsmeanz(k) = obsmeanz(k) / (nx * ny)
WRITE (15,'(f7.2)') obsmeanz(k)
END DO
END IF
! ########## End k-level observation mean ##########
CLOSE (UNIT = 15)
600 FORMAT ("Observation mean for all data"/)
601 FORMAT (/"Observation mean with i-level dependence"/)
602 FORMAT (/"Observation mean with j-level dependence"/)
603 FORMAT (/"Observation mean with k-level dependence"/)
END SUBROUTINE Observation_Mean
!#######################################################################
SUBROUTINE FORECAST_VARIANCE (nx, ny, nz, model, ilevel, &
jlevel, klevel)
! Calculates the varience for the hypothetical forecast set and writes
! this to the "FcstVar.txt" file.
!
! Formula:
!
! Variance = SUM[ (x - x(bar))**2 / N]
!
! x is the value of each individual forecast
! x(bar) is the mean of the set of forecasts
! N is the number of forecasts
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: avg_all ! holds average value
REAL :: avgz(nz)
REAL :: avgy(ny)
REAL :: avgx(nx)
REAL :: fcstvar_all
REAL :: fcstvarz(nz)
REAL :: fcstvary(ny)
REAL :: fcstvarx(nx)
REAL :: model(nx,ny,nz)
! ##### Calculates forecast variance for all data #####
avg_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
avg_all = avg_all + model(i,j,k)
END DO
END DO
END DO
avg_all = avg_all / (nx * ny * nz)
fcstvar_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
fcstvar_all = fcstvar_all + (model(i,j,k) - avg_all)**2
END DO
END DO
END DO
fcstvar_all = fcstvar_all / (nx*ny*nz)
OPEN (UNIT = 17, FILE = "FcstVar.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (17, FMT = 700)
WRITE (17,'(e13.2)') fcstvar_all
! ########## End forecast variance for all data ##########
! ##### Calculates i-level forecast variance #####
IF (ilevel == 1) THEN
DO i=1,nx
avgx(i) = 0.0
DO j=1,ny
DO k=1,nz
avgx(i) = avgx(i) + model(i,j,k)
END DO
END DO
END DO
DO i=1,nx
avgx(i) = avgx(i) / (ny * nz)
END DO
DO i=1,nx
fcstvarx(i) = 0.0
DO j=1,ny
DO k=1,nz
fcstvarx(i) = fcstvarx(i) + (model(i,j,k) - avgx(i))**2
END DO
END DO
END DO
WRITE (17, FMT = 701)
DO i=1,nx
fcstvarx(i) = fcstvarx(i) / (ny*nz)
WRITE (17,'(e13.2)') fcstvarx(i)
END DO
END IF
! ########## End i-level forecast variance ##########
! ##### Calculates j-level forecast variance #####
IF (jlevel == 1) THEN
DO j=1,ny
avgy(j) = 0.0
DO i=1,nx
DO k=1,nz
avgy(j) = avgy(j) + model(i,j,k)
END DO
END DO
END DO
DO j=1,ny
avgy(j) = avgy(j) / (nx * nz)
END DO
DO j=1,ny
fcstvary(j) = 0.0
DO i=1,nx
DO k=1,nz
fcstvary(j) = fcstvary(j) + (model(i,j,k) - avgy(j))**2
END DO
END DO
END DO
WRITE (17, FMT = 702)
DO j=1,ny
fcstvary(j) = fcstvary(j) / (nx * nz)
WRITE (17,'(e13.2)') fcstvary(j)
END DO
END IF
! ########## End j-level forecast variance ##########
! ##### Calculates k-level forecast variance #####
IF (klevel == 1) THEN
DO k=1,nz
avgz(k) = 0.0
DO j=1,ny
DO i=1,nx
avgz(k) = avgz(k) + model(i,j,k)
END DO
END DO
END DO
DO k=1,nz
avgz(k) = avgz(k) / (nx * ny)
END DO
DO k=1,nz
fcstvarz(k) = 0.0
DO j=1,ny
DO i=1,nx
fcstvarz(k) = fcstvarz(k) + (model(i,j,k) - avgz(k))**2
END DO
END DO
END DO
WRITE (17, FMT = 703)
DO k=1,nz
fcstvarz(k) = fcstvarz(k) / (nx * ny)
WRITE (17,'(e13.2)') fcstvarz(k)
END DO
END IF
! ########## End k-level forecast variance ##########
CLOSE (17)
700 FORMAT ("Forecast Variance with all data"/)
701 FORMAT (/"Forecast Variance with i-level dependence"/)
702 FORMAT (/"Forecast Variance with j-level dependence"/)
703 FORMAT (/"Forecast Variance with k-level dependence"/)
END SUBROUTINE FORECAST_VARIANCE
!#######################################################################
SUBROUTINE OBSERVATION_VARIANCE (nx, ny, nz, obs, ilevel, &
jlevel, klevel)
! Calculates the varience for the hypothetical forecast set and writes
! this to the "FcstVar.txt" file.
!
! Formula:
!
! Variance = SUM[ (x - x(bar))**2 / N]
!
! x is the value of each individual observation
! x(bar) is the mean of the set of observations
! N is the number of observations
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: avg_all ! holds average value
REAL :: avgz(nz)
REAL :: avgy(ny)
REAL :: avgx(nx)
REAL :: obsvar_all
REAL :: obsvarz(nz)
REAL :: obsvary(ny)
REAL :: obsvarx(nx)
REAL :: obs(nx,ny,nz), model(nx,ny,nz)
! ##### Calculates observation variance for all data #####
avg_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
avg_all = avg_all + obs(i,j,k)
END DO
END DO
END DO
avg_all = avg_all / (nx * ny * nz)
obsvar_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
obsvar_all = obsvar_all + (obs(i,j,k) - avg_all)**2
END DO
END DO
END DO
obsvar_all = obsvar_all / (nx*ny*nz)
OPEN (UNIT = 18, FILE = "ObsVar.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (18, FMT = 800)
WRITE (18,'(e13.2)') obsvar_all
! ########## End observation variance for all data ##########
! ##### Calculates i-level observation variance #####
IF (ilevel == 1) THEN
DO i=1,nx
avgx(i) = 0.0
DO j=1,ny
DO k=1,nz
avgx(i) = avgx(i) + obs(i,j,k)
END DO
END DO
END DO
DO i=1,nx
avgx(i) = avgx(i) / (ny * nz)
END DO
DO i=1,nx
obsvarx(i) = 0.0
DO j=1,ny
DO k=1,nz
obsvarx(i) = obsvarx(i) + (obs(i,j,k) - avgx(i))**2
END DO
END DO
END DO
WRITE (18, FMT = 801)
DO i=1,nx
obsvarx(i) = obsvarx(i) / (ny*nz)
WRITE (18,'(e13.2)') obsvarx(i)
END DO
END IF
! ########## End i-level observation variance ##########
! ##### Calculates j-level observation variance #####
IF (jlevel == 1) THEN
DO j=1,ny
avgy(j) = 0.0
DO i=1,nx
DO k=1,nz
avgy(j) = avgy(j) + obs(i,j,k)
END DO
END DO
END DO
DO j=1,ny
avgy(j) = avgy(j) / (nx * nz)
END DO
DO j=1,ny
obsvary(j) = 0.0
DO i=1,nx
DO k=1,nz
obsvary(j) = obsvary(j) + (obs(i,j,k) - avgy(j))**2
END DO
END DO
END DO
WRITE (18, FMT = 802)
DO j=1,ny
obsvary(j) = obsvary(j) / (nx * nz)
WRITE (18,'(e13.2)') obsvary(j)
END DO
END IF
! ########## End j-level observation variance ##########
! ##### Calculates k-level observation variance #####
IF (klevel == 1) THEN
DO k=1,nz
avgz(k) = 0.0
DO j=1,ny
DO i=1,nx
avgz(k) = avgz(k) + obs(i,j,k)
END DO
END DO
END DO
DO k=1,nz
avgz(k) = avgz(k) / (nx * ny)
END DO
DO k=1,nz
obsvarz(k) = 0.0
DO j=1,ny
DO i=1,nx
obsvarz(k) = obsvarz(k) + (obs(i,j,k) - avgz(k))**2
END DO
END DO
END DO
WRITE (18, FMT = 803)
DO k=1,nz
obsvarz(k) = obsvarz(k) / (nx * ny)
WRITE (18,'(e13.2)') obsvarz(k)
END DO
END IF
! ########## End k-level forecast variance ##########
CLOSE (18)
800 FORMAT ("Observation Variance with all data"/)
801 FORMAT (/"Observation Variance with i-level dependence"/)
802 FORMAT (/"Observation Variance with j-level dependence"/)
803 FORMAT (/"Observation Variance with k-level dependence"/)
CLOSE (UNIT = 18)
END SUBROUTINE OBSERVATION_VARIANCE
!#######################################################################
SUBROUTINE FORECAST_STD_DEVIATION (nx, ny, nz, model, ilevel, &
jlevel, klevel)
! Calculates the standard deviation for the hypothetical forecast data
! set and writes this to the "FcstStdDev.txt" file
!
! Formula:
!
! Std Dev = SQRT[ SUM[ (x -x(bar))**2 / N]]
!
! x is the value of each individual forecast
! x(bar) is the mean of the set of forecasts
! N is the number of forecasts
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: stddev_all
REAL :: stddevz(nz)
REAL :: stddevy(ny)
REAL :: stddevx(nx)
REAL :: avg_all
REAL :: avgz(nz)
REAL :: avgy(ny)
REAL :: avgx(nx)
REAL :: model(nx,ny,nz)
! ##### Calculates the standard deviation for all data #####
avg_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
avg_all = avg_all + model(i,j,k)
END DO
END DO
END DO
avg_all = avg_all / (nx * ny * nz)
DO k=1,nz
avgz(k) = avgz(k) / (nx * ny)
END DO
stddev_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
stddev_all = stddev_all + (model(i,j,k) - avg_all)**2
END DO
END DO
END DO
OPEN (UNIT = 20, FILE = "FcstStdDev.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (20, FMT = 100)
stddev_all = stddev_all / (nx * ny * nz)
stddev_all = SQRT(stddev_all)
WRITE (20,'(e13.2)') stddev_all
! ########## End standard deviation for all data ##########
! ##### Calculates i-level standard deviation #####
IF (ilevel == 1) THEN
DO i=1,nx
avgx(i) = 0.0
DO j=1,ny
DO k=1,nz
avgx(i) = avgx(i) + model(i,j,k)
END DO
END DO
END DO
DO i=1,nx
avgx(i) = avgx(i) / (ny * nz)
END DO
DO i=1,nx
stddevx(i) = 0.0
DO j=1,ny
DO k=1,nz
stddevx(i) = stddevx(i) + (model(i,j,k) - avgx(i))**2
END DO
END DO
END DO
WRITE (20, FMT = 101)
DO i=1,nx
stddevx(i) = stddevx(i) / (ny * nz)
stddevx(i) = SQRT(stddevx(i))
WRITE (20,'(e13.2)') stddevx(i)
END DO
END IF
! ########## End i-level standard deviation ##########
! ##### Calculates j-level standard deviation #####
IF (jlevel == 1) THEN
DO j=1,ny
avgy(j) = 0.0
DO k=1,nz
DO i=1,nx
avgy(j) = avgy(j) + model(i,j,k)
END DO
END DO
END DO
DO j=1,ny
avgy(j) = avgy(j) / (nx * nz)
END DO
DO j=1,ny
stddevy(j) = 0.0
DO k=1,nz
DO i=1,nx
stddevy(j) = stddevy(j) + (model(i,j,k) - avgy(j))**2
END DO
END DO
END DO
WRITE (20, FMT = 102)
DO j=1,ny
stddevy(j) = stddevy(j) / (nx * nz)
stddevy(j) = SQRT(stddevy(j))
WRITE (20,'(e13.2)') stddevy(j)
END DO
END IF
! ########## End j-level standard deviation ##########
! ##### Calculates k-level standard deviation #####
IF (klevel == 1) THEN
DO k=1,nz
avgz(k) = 0.0
DO j=1,ny
DO i=1,nx
avgz(k) = avgz(k) + model(i,j,k)
END DO
END DO
END DO
DO k=1,nz
avgz(k) = avgz(k) / (nx * ny)
END DO
DO k=1,nz
stddevz(k) = 0.0
DO j=1,ny
DO i=1,nx
stddevz(k) = stddevz(k) + (model(i,j,k) - avgz(k))**2
END DO
END DO
END DO
WRITE (20, FMT = 103)
DO k=1,nz
stddevz(k) = stddevz(k) / (nx * ny)
stddevz(k) = SQRT(stddevz(k))
WRITE (20,'(e13.2)') stddevz(k)
END DO
END IF
! ########## End k-level standard deviation ##########
CLOSE(20)
100 FORMAT ("Forecast standard deviation for all data"/)
101 FORMAT (/"Forecast std dev with i-level dependence"/)
102 FORMAT (/"Forecast std dev with j-level dependence"/)
103 FORMAT (/"Forecast std dev with k-level dependence"/)
END SUBROUTINE FORECAST_STD_DEVIATION
!#######################################################################
SUBROUTINE OBSERVATION_STD_DEVIATION (nx, ny, nz, obs, ilevel, &
jlevel, klevel)
! Calculates the standard deviation for the hypothetical observation
! set and writes this to the "ObsStdDev.txt" file
!
! Formula:
!
! Std Dev = SQRT[ SUM[ (x -x(bar))**2 / N]]
!
! x is the value of each individual observation
! x(bar) is the mean of the set of observations
! N is the number of observations
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
REAL :: stddev_all
REAL :: stddevz(nz)
REAL :: stddevy(ny)
REAL :: stddevx(nx)
REAL :: avg_all
REAL :: avgz(nz)
REAL :: avgy(ny)
REAL :: avgx(nx)
REAL :: obs(nx,ny,nz)
! ##### Calculates the standard deviation for all data #####
avg_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
avg_all = avg_all + obs(i,j,k)
END DO
END DO
END DO
avg_all = avg_all / (nx * ny * nz)
stddev_all = 0.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
stddev_all = stddev_all + (obs(i,j,k) - avg_all)**2
END DO
END DO
END DO
OPEN (UNIT = 21, FILE = "ObsStdDev.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (21, FMT = 200)
stddev_all = stddev_all / (nx * ny * nz)
stddev_all = SQRT(stddev_all)
WRITE (21,'(e13.2)') stddev_all
! ########## End standard deviation for all data ##########
! ##### Calculates i-level standard deviation #####
IF (ilevel == 1) THEN
DO i=1,nx
avgx(i) = 0.0
DO j=1,ny
DO k=1,nz
avgx(i) = avgx(i) + obs(i,j,k)
END DO
END DO
END DO
DO i=1,nx
avgx(i) = avgx(i) / (ny * nz)
END DO
DO i=1,nx
stddevx(i) = 0.0
DO j=1,ny
DO k=1,nz
stddevx(i) = stddevx(i) + (obs(i,j,k) - avgx(i))**2
END DO
END DO
END DO
WRITE (21, FMT = 201)
DO i=1,nx
stddevx(i) = stddevx(i) / (ny * nz)
stddevx(i) = SQRT(stddevx(i))
WRITE (21,'(e13.2)') stddevx(i)
END DO
END IF
! ########## End i-level standard deviation ##########
! ##### Calculates j-level standard deviation #####
IF (jlevel == 1) THEN
DO j=1,ny
avgy(j) = 0.0
DO k=1,nz
DO i=1,nx
avgy(j) = avgy(j) + obs(i,j,k)
END DO
END DO
END DO
DO j=1,ny
avgy(j) = avgy(j) / (nx * nz)
END DO
DO j=1,ny
stddevy(j) = 0.0
DO k=1,nz
DO i=1,nx
stddevy(j) = stddevy(j) + (obs(i,j,k) - avgy(j))**2
END DO
END DO
END DO
WRITE (21, FMT = 202)
DO j=1,ny
stddevy(j) = stddevy(j) / (nx * nz)
stddevy(j) = SQRT(stddevy(j))
WRITE (21,'(e13.2)') stddevy(j)
END DO
END IF
! ########## End j-level standard deviation ##########
! ##### Calculates k-level standard deviation #####
IF (klevel == 1) THEN
DO k=1,nz
avgz(k) = 0.0
DO j=1,ny
DO i=1,nx
avgz(k) = avgz(k) + obs(i,j,k)
END DO
END DO
END DO
DO k=1,nz
avgz(k) = avgz(k) / (nx * ny)
END DO
DO k=1,nz
stddevz(k) = 0.0
DO j=1,ny
DO i=1,nx
stddevz(k) = stddevz(k) + (obs(i,j,k) - avgz(k))**2
END DO
END DO
END DO
WRITE (21, FMT = 203)
DO k=1,nz
stddevz(k) = stddevz(k) / (nx * ny)
stddevz(k) = SQRT(stddevz(k))
WRITE (21,'(e13.2)') stddevz(k)
END DO
END IF
! ########## End k-level standard deviation ##########
CLOSE(21)
200 FORMAT ("Observation standard deviation for all data"/)
201 FORMAT (/"Observation std dev with i-level dependence"/)
202 FORMAT (/"Observation std dev with j-level dependence"/)
203 FORMAT (/"Observation std dev with k-level dependence"/)
END SUBROUTINE OBSERVATION_STD_DEVIATION
!#######################################################################
SUBROUTINE BIAS_CALC (nx, ny, nz, data, model, ilevel, jlevel, & 5
klevel,bias_all)
! Calculates the bias along the k level of the grid between the
! hypothetical data set and model forecast and writes this
! to the "JBias.txt" file
!
! Formula:
!
! Bias = (1/n) * SUM[k=1 to n] (y(k) - o(k))
!
! n is the number of forecast / observation pairs
! y(k) is the kth of n forecasts
! o(k) is the kth of n observations
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios, num
INTEGER :: biascount
REAL :: bias_all
REAL :: biasz(nz)
REAL :: biasy(ny)
REAL :: biasx(nx)
REAL :: data(nx,ny,nz), model(nx,ny,nz)
! ##### Calclulates the bias for all data #####
bias_all = 0.0
biascount=0
DO k=1,nz
DO j=1,ny
DO i=1,nx
IF (data(i,j,k).ne.-99.9) THEN
bias_all = bias_all + (model(i,j,k)-data(i,j,k))
biascount=biascount+1
END IF
END DO
END DO
END DO
! OPEN (UNIT = 16, FILE = "Bias.txt", IOSTAT = ios, &
! FORM = "FORMATTED")
! IF (ios /= 0) THEN
! PRINT*, "Error opening file"
! END IF
! WRITE (16, FMT = 900)
! bias_all = bias_all / (nx * ny * nz)
bias_all = bias_all / (biascount)
! WRITE (16,'(f7.2)') bias_all
! ########## End bias for all data ##########
! ##### Calculates i-level bias ######
IF (ilevel == 1) THEN
DO i=1,nx
biasx(i) = 0.0
DO j=1,ny
DO k=1,nz
biasx(i) = biasx(i) + (model(i,j,k)-data(i,j,k))
END DO
END DO
END DO
WRITE (16, FMT = 901)
DO i = 1,nx
biasx(i) = biasx(i) / (ny * nz)
WRITE (16,'(f7.2)') biasx(i)
END DO
END IF
! ########## End i-level bias ##########
! ##### Calculates j-level bias #####
IF (jlevel == 1) THEN
DO j=1,ny
biasy(j) = 0.0
DO k=1,nz
DO i=1,nx
biasy(j) = biasy(j) + (model(i,j,k)-data(i,j,k))
END DO
END DO
END DO
WRITE (16, FMT = 902)
DO j = 1,ny
biasy(j) = biasy(j) / (nx * nz)
WRITE (16,'(f7.2)') biasy(j)
END DO
END IF
! ########## End j-level bias ##########
! ##### Calculate k-level bias #####
IF (klevel == 1) THEN
DO k=1,nz
biasz(k) = 0.0
DO j=1,ny
DO i=1,nx
biasz(k) = biasz(k) + (model(i,j,k)-data(i,j,k))
END DO
END DO
END DO
WRITE (16, FMT = 903)
DO k = 1,nz
biasz(k) = biasz(k) / (nx * ny)
WRITE (16,'(f7.2)') biasz(k)
END DO
END IF
! ########## End k-level bias ##########
CLOSE (UNIT = 16)
900 FORMAT ("Bias for all data"/)
901 FORMAT (/"Bias with i-level dependence"/)
902 FORMAT (/"Bias with j-level dependence"/)
903 FORMAT (/"Bias with k-level dependence"/)
END SUBROUTINE BIAS_CALC
!#######################################################################
SUBROUTINE THREAT_SCORE (noe, fcstprob, obsevent)
! Calculates the threat score using hypothetical data from a
! probability forecast and writes this calculation to the
! "TS.txt" file. This calculation also goes by the name
! Critical Success Index (CSI).
!
! Formula:
!
! Threat Score = CFE(y) / [ FE(y) + OE(y) - CFE(y) ]
!
!
! CFE(y) is the number of correctly forecast 'yes' events
! FE(y) is the number of forecast 'yes' events
! OE(y) is the number of observed 'yes' events
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: cfey ! number of correctly forecast 'yes' events
REAL :: fey ! number of forecast 'yes' events
REAL :: oey ! number of observed 'yes' events
REAL :: threat ! threat score
REAL :: fcstprob(noe), obsevent(noe)
cfey = 0
fey = 0
oey = 0
DO g=1,noe
IF ((fcstprob(g).gt.(.50)).AND.(obsevent(g) == 1.0)) THEN
cfey = cfey + 1
END IF
END DO
DO g=1,noe
IF (fcstprob(g).gt.(.50)) THEN
fey = fey + 1
END IF
END DO
DO g=1,noe
IF (obsevent(g) == 1) THEN
oey = oey + 1
END IF
END DO
threat = cfey / ( fey + oey - cfey )
OPEN (UNIT = 23, FILE = "TS.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (23, FMT = 100)
WRITE (23, '(f7.4)') threat
CLOSE (23)
100 FORMAT ("Threat Score"/)
END SUBROUTINE THREAT_SCORE
!#######################################################################
SUBROUTINE EQUITABLE_THREAT_SCORE (noe, fcstprob, obsevent)
! Calculates the equitable threat score using hypothetical data from a
! probability forecast and writes this calculation to the
! "ETS.txt" file.
!
! Formula:
!
! Equitable Threat Score =(CFE(y) - RH) / [FE(y) + OE(y) - CFE(y) - RH]
!
! RH = (FE(y) * OE(y)) / n
!
! RH is the number of hits above the threshold that would be expected
! in a random forecast
! CFE(y) is the number of correctly forecast 'yes' events
! FE(y) is the number of forecast 'yes' events
! OE(y) is the number of observed 'yes' events
! n is the total number of forecasts
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: cfey ! number of correctly forecast 'yes' events
REAL :: fey ! number of forecast 'yes' events
REAL :: oey ! number of observed 'yes' events
REAL :: rh ! random hits
REAL :: ets ! equitable threat score
REAL :: fcstprob(noe), obsevent(noe)
cfey = 0
fey = 0
oey = 0
DO g=1,noe
IF ((fcstprob(g).gt.(.50)).AND.(obsevent(g) == 1.0)) THEN
cfey = cfey + 1
END IF
END DO
DO g=1,noe
IF (fcstprob(g).gt.(.50)) THEN
fey = fey + 1
END IF
END DO
DO g=1,noe
IF (obsevent(g) == 1) THEN
oey = oey + 1
END IF
END DO
rh = (fey * oey) / noe
ets = (cfey - rh) / (fey + oey - cfey -rh)
OPEN (UNIT = 24, FILE = "ETS.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (24, FMT = 100)
WRITE (24, '(f7.4)') ets
CLOSE (24)
100 FORMAT ("Equitable Threat Score"/)
END SUBROUTINE EQUITABLE_THREAT_SCORE
!#######################################################################
SUBROUTINE PROB_OF_DETECTION (noe, fcstprob, obsevent)
! Calculates the probability of detection using hypothetical data
! from a probability forecast and writes this calculation to the
! "POD.txt" file.
!
! Formula:
!
! Probability of Detection = CFE(y) / OE(y)
!
! CFE(y) is the number of correctly forecast 'yes' events
! OE(y) is the number of observed 'yes' events
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: cfey ! number of correctly forecast 'yes' events
REAL :: oey ! number of observed 'yes' events
REAL :: pod ! probability of detection
REAL :: fcstprob(noe), obsevent(noe)
cfey = 0
oey = 0
DO g=1,noe
IF ((fcstprob(g).gt.(.50)).AND.(obsevent(g) == 1.0)) THEN
cfey = cfey + 1
END IF
END DO
DO g=1,noe
IF (obsevent(g) == 1) THEN
oey = oey + 1
END IF
END DO
pod = cfey / oey
OPEN (UNIT = 25, FILE = "POD.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (25, FMT = 100)
WRITE (25, '(f7.4)') pod
CLOSE (25)
100 FORMAT ("Probability of Detection"/)
END SUBROUTINE PROB_OF_DETECTION
!#######################################################################
SUBROUTINE FALSE_ALARM_RATE (noe, fcstprob, obsevent)
! Calculates the false alarm rate using hypothetical data
! from a probability forecast and writes this calculation to the
! "FAR.txt" file.
!
! Formula:
!
! False Alarm Rate = IFE(n) / FE(y)
!
! IFE(n) is the number of incorrectly forecast 'no' events
! (ie: The event was forecast but not observed.)
! FE(y) is the number of forecast 'yes' events
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: ifen ! number of incorrectly forecast 'yes' events
REAL :: fey ! number of forecast 'yes' events
REAL :: far ! false alarm rate
REAL :: fcstprob(noe), obsevent(noe)
ifen = 0
fey = 0
DO g=1,noe
IF ((fcstprob(g).gt.(.50)).AND.(obsevent(g) == 0.0)) THEN
ifen = ifen + 1
END IF
END DO
DO g=1,noe
IF (fcstprob(g).gt.(.50)) THEN
fey = fey + 1
END IF
END DO
far = ifen / fey
OPEN (UNIT = 26, FILE = "FAR.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (26, FMT = 100)
WRITE (26, '(f7.4)') far
CLOSE (26)
100 FORMAT ("False Alarm Rate"/)
END SUBROUTINE FALSE_ALARM_RATE
!######################################################################
SUBROUTINE HIT_RATE (noe, fcstprob, obsevent)
! Calculates the hit rate using hypothetical data from a probability
! forecast and writes this calculation to the "HR.txt" file.
!
! Formula:
!
! Hit Rate = (CFE(y) + CFE(n)) / n
!
! CFE(y) is the number of correctly forecast 'yes' events
! CFE(n) is the number of correctly forecast 'no' events
! n is the total number of forecasts
!#######################################################################
IMPLICIT none
INTEGER :: noe ! number of events
INTEGER :: g, ios
REAL :: cfey ! number of correctly forecast 'yes' events
REAL :: cfen ! number of correctly forecast 'no' events
REAL :: hr ! hit rate
REAL :: fcstprob(noe), obsevent(noe)
cfey = 0.0
cfen = 0.0
DO g=1,noe
IF ((fcstprob(g).gt.(.50)).AND.(obsevent(g) == 1.0)) THEN
cfey = cfey + 1
END IF
END DO
DO g=1,noe
IF ((fcstprob(g).le.(.50)).AND.(obsevent(g) == 0.0)) THEN
cfen = cfen + 1
END IF
END DO
hr = (cfey + cfen) / noe
OPEN (UNIT = 27, FILE = "HR.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (27, FMT = 100)
WRITE (27, '(f7.4)') hr
CLOSE (27)
100 FORMAT ("Hit Rate"/)
END SUBROUTINE HIT_RATE
!#######################################################################
SUBROUTINE SKILL_SCORE (nx, ny, nz, model, ilevel, &
jlevel, klevel)
! Calculates the skill score using the hypothetical model forecast
! and a hypothetical reference value writes this to the "SS.txt" file
!
! The file "SS.txt" will always contain the mean absolute error for
! the entire data set at the top of the file. It will also contain
! the mean absolute error based on the i, j, or k levels depending
! upon which of those options are chosen in the "verif.input" file.
!
!
! Formula:
!
! MAE = (1/n) * SUM (forecast - reference) / (1 - reference)
!
! n is the number of forecasts
! forecast is the forecast value
! reference is the "climotology" or "persistance" value
!#######################################################################
IMPLICIT none
INTEGER :: nx ! number of x grid points
INTEGER :: ny ! number of y grid points
INTEGER :: nz ! number of z grid points
INTEGER :: ilevel ! toggles ilevel calculation
INTEGER :: jlevel ! toggles jlevel calculation
INTEGER :: klevel ! toggles klevel calculation
INTEGER :: i, j ,k, ios
REAL :: ssz(nz)
REAL :: ssy(ny)
REAL :: ssx(nx)
REAL :: ss_all ! variable for SS of entire data set
REAL :: num
REAL :: ref ! reference value
REAL :: model(nx,ny,nz)
! ##### Calculates skill score for entire data set mae_all #####
ss_all = 0.0
ref = 50.0
DO k=1,nz
DO j=1,ny
DO i=1,nx
num = 0.0
num = (model(i,j,k) - ref) / (1 - ref)
ss_all = ss_all + num
END DO
END DO
END DO
OPEN (UNIT = 28, FILE = "SS.txt", IOSTAT = ios, &
FORM = "FORMATTED")
IF (ios /= 0) THEN
PRINT*, "Error opening file"
END IF
WRITE (28, FMT = 200)
ss_all = ss_all / (nx * ny * nz)
WRITE (28,'(f7.2)') ss_all
! ########## End of ss_all calculation ##########
! ##### Calculates i-level skill score, ssx #####
IF (ilevel == 1) THEN
DO i=1,nx
ssx(i) = 0.0
DO j=1,ny
DO k=1,nz
num = 0.0
num = (model(i,j,k) - ref) / (1 - ref)
ssx(i) = ssx(i) + num
END DO
END DO
END DO
WRITE (28, FMT = 201)
DO i=1,nx
ssx(i) = ssx(i) / (ny * nz)
WRITE (28,'(f7.2)') ssx(i)
END DO
END IF
! ########## End of i-level skill score ##########
! ##### Calculates j-level skill score, ssy #####
IF (jlevel == 1) THEN
DO j=1,ny
ssy(j) = 0.0
DO i=1,nx
DO k=1,nz
num = 0.0
num = (model(i,j,k) - ref) / (1 - ref)
ssy(j) = ssy(j) + num
END DO
END DO
END DO
WRITE (28, FMT = 202)
DO j=1,ny
ssy(j) = ssy(j) / (nx * nz)
WRITE (28,'(f7.2)') ssy(j)
END DO
END IF
! ######### End of j-level skill score ##########
! ##### Calculates the k-level skill score #####
IF (klevel == 1) THEN
DO k=1,nz
ssz(k) = 0.0
DO j=1,ny
DO i=1,nx
num = 0.0
num = (model(i,j,k) - ref) / (1 - ref)
ssz(k) = ssz(k) + num
END DO
END DO
END DO
WRITE (28, FMT = 203)
DO k=1,nz
ssz(k) = ssz(k) / (nx * ny)
WRITE (28,'(f7.2)') ssz(k)
END DO
END IF
! ######### End of k-level skill score #########
CLOSE (UNIT = 28)
200 FORMAT ("Skill Score for all data"/)
201 FORMAT (/"Skill Score with i-level dependence"/)
202 FORMAT (/"Skill Score with j-level dependence"/)
203 FORMAT (/"Skill Score with k-level dependence"/)
END SUBROUTINE SKILL_SCORE