[Dart-dev] DART/branches Revision: 12746

[dart at ucar.edu]

thoar at ucar.edu
2018-07-13 10:07:06 -0600 (Fri, 13 Jul 2018)
45
Making consistent with current git version.




Added: DART/branches/rma_wrfHydro/R/PlotEnsTimeSeries.R
===================================================================
--- DART/branches/rma_wrfHydro/R/PlotEnsTimeSeries.R	                        (rev 0)
+++ DART/branches/rma_wrfHydro/R/PlotEnsTimeSeries.R	2018-07-13 16:07:06 UTC (rev 12746)
@@ -0,0 +1,149 @@
+PlotEnsTimeSeries <- function(pinfo) {
+## PlotEnsTimeSeries: Plots time series of ensemble members, mean and truth.
+#
+#
+# PlotEnsTimeSeries is intended to be called by 'plot_ens_time_series'
+# The only input argument is a structure with model-dependent
+# components.
+#
+# USAGE: PlotEnsTimeSeries( pinfo );
+#
+# STRUCTURE COMPONENTS FOR low-order models
+# truth_file      name of netCDF DART file with copy tagged 'true state'
+# diagn_file      name of netCDF DART file with copies tagged 'ensemble mean'
+#                 and 'ensemble spread'
+# var             name of netCDF variable of interest
+# var_inds        indices of variables of interest
+
+# Example 1
+  if(FALSE){
+    library(rwrfhydro)
+    dartPath <-
+      strsplit(grep('wrf_hydro_dart', readLines('~/.wrf_hydro_tools'), value=TRUE),'=')[[1]][2]
+    source(paste0(dartPath,'/R/PlotEnsTimeSeries.R'))
+
+    pinfo <- list()
+    pinfo$truth_file  = 'True_State.nc'
+    pinfo$diagn_file  = 'Prior_Diag.nc'
+    pinfo$model       = 'wrfHydro'      ## not really necessary currently
+
+    pinfo$var         = 'qlink1'
+    ##pinfo$var_inds  = which(trimws(ncdump("../DOMAIN/RouteLink.nc",'gages', q=TRUE))!='')
+    pinfo$var_inds    = c(130,136,157)
+    pinfo$var_ind_name= trimws(ncdump("../DOMAIN/RouteLink.nc",'gages', q=TRUE)[pinfo$var_inds])
+    PlotEnsTimeSeries(pinfo)
+       
+    pinfo$var         = 'qBucketMult'
+    pinfo$var_inds    = NA
+    pinfo$var_ind_name= NA
+    PlotEnsTimeSeries(pinfo)
+
+    pinfo$var         = 'qSfcLatRunoffMult'
+    PlotEnsTimeSeries(pinfo)
+
+  }
+  ## Code starts here -----------------------------------------------------------------
+
+  if(!file.exists(pinfo$diagn_file)) stop(paste0("No such file: ", pinfo$diagn_file))
+  pinfo$diagn_file_abs <- if(substr(pinfo$diagn_file,1,1)=='/') pinfo$diagn_file else paste0(getwd(),'/',pinfo$diagn_file)
+  
+  ## Query the metadata to determine which "copy" is the ensemble mean.
+  ncMeta <- ncdump(pinfo$diagn_file, q=TRUE)
+  varunits <- ncMeta$var[[pinfo$var]]$units
+  nVarDims <- ncMeta$var[[pinfo$var]]$ndims
+  varDims <- unlist(lapply(ncMeta$var[[pinfo$var]]$dim, function(vv) vv$name))
+  nMembers <- ncMeta$dim$copy$len
+  copyNames <- ncdump(pinfo$diagn_file,'CopyMetaData', q=TRUE)
+  ens_mean_index <- which( trimws(copyNames) ==  'ensemble mean')
+  whEnsMems <- grep('ensemble member',copyNames)
+  gregorianOrigin <- "1601-01-01 00:00:00 UTC"
+  time <- as.POSIXct( ncdump(pinfo$diagn_file,'time', q=TRUE) * 24*3600,
+                      origin=gregorianOrigin, tz='UTC')
+
+  ## JLM <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
+  ## WOW, going to need to check that this is correct via performing the calculation...
+  # because of issues with ncdf4
+  ## JLM <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
+
+  ## If the truth is known, great.
+  ## JLM, deal with truth when we have/fake it
+  #have_truth <- FALSE
+  #if(file.exists(pinfo$truth_file)) {
+  #  have_truth  = TRUE
+  #  truth_index = get_copy_index(pinfo$truth_file, 'true state')
+  #}
+
+  ## This works for scalar multipliers (e.g. qBucketMult, qSfcLatRunoffMult)
+  if(all(varDims==c("scalar", "copy", "time"))) {
+    ens_mean <-
+      ncdump(pinfo$diagn_file, pinfo$var, q=TRUE)[ens_mean_index, ,drop=FALSE]
+    ens_members <-
+      ncdump(pinfo$diagn_file, pinfo$var, q=TRUE)[whEnsMems, ,drop=FALSE]
+    if(!(abs(mean(ens_members[,1])-ens_mean[1,1])) < 1e-7)
+      warning("The ensemble mean does not match a spot check of its value.\n",
+              "Please investigate, may be a dimension issue with ncdf4 package.")
+
+    dimnames(ens_mean) <- list()
+    dimnames(ens_mean)[[2]] <- format(time,'%Y-%m-%d %H:%M:%S')
+    dimnames(ens_mean)[[1]] <- trimws(copyNames[ens_mean_index])
+    meanDf <- plyr::adply(ens_mean, c(1,2))
+    meanDf <- plyr::rename(meanDf, c(X1='member', X2='time', V1=pinfo$var))
+
+    dimnames(ens_mean) <- list()
+    dimnames(ens_members)[[1]] <- trimws(copyNames[whEnsMems])
+    dimnames(ens_members)[[2]] <- format(time,'%Y-%m-%d %H:%M:%S')
+    ensDf <- plyr::adply(ens_members, c(1,2))
+    ensDf <- plyr::rename(ensDf, c(X1='member', X2='time', V1=pinfo$var))