Changes between Initial Version and Version 1 of PDAF_diag_crps

Dec 9, 2021, 5:15:10 PM (2 years ago)



  • PDAF_diag_crps

    v1 v1  
     1= PDAF_diag_CRPS =
     4This page documents the routine `PDAF_diag_CRPS` of PDAF, which was introduced with PDAF V2.0.
     6This routine computes the Continuous Ranked Probability Score (CRPS) and its decomposition into resolution and reliability. The CRPS provide information about the statistical consistency of the ensemble with the observations. In toy models, the CRPS can also be computed with raegard to the true state.
     8Inputs are an array holding the observed ensemble and a corresponding vector of observations.
     10The routine can be called in the pre/poststep routine of PDAF both before and after the analysis step to compute the CRPS.
     12The interface is the following:
     14  SUBROUTINE PDAF_diag_CRPS(dim, dim_ens, element, oens, obs, &
     15             CRPS, reli, resol, uncert, status)
     17with the following arguments:
     19  INTEGER, INTENT(in) :: dim                ! PE-local state dimension
     20  INTEGER, INTENT(in) :: dim_ens            ! Ensemble size
     21  INTEGER, INTENT(in) :: element            ! ID of element to be used
     22       !< If element=0, mean values over all elements are computed
     23  REAL, INTENT(in)    :: oens(dim, dim_ens) ! State ensemble
     24  REAL, INTENT(in)    :: obs(dim)           ! State ensemble
     25  REAL, INTENT(out)   :: CRPS               ! CRPS
     26  REAL, INTENT(out)   :: reli               ! Reliability
     27  REAL, INTENT(out)   :: resol              ! resolution
     28  REAL, INTENT(out)   :: uncert             ! uncertainty
     29  INTEGER, INTENT(out) :: status            ! Status flag (0=success)
     33 * using `element` one can select a since element of the observation vector for which the CRPS is computed (by multiple computations, it allows to computed a CRPS individually for each entry of the state vector). For `element=0` the CRPS over all elements is computed
     34 * A perfectly reliable system gives `reli=0`. An informative system gives `resol << uncert`.
     35 * Compared to Hersbach (2000), `resol` here is equivalent to `CPRS_pot`.
     36 * The routine is not parallelized. In addition, it uses a rather simple sorting algorithm. Accordingly, the performance is likely suboptimal for high-dimensional cases.