Feature 2003 briercl

met/data/config/GridStatConfig_default

@@ -107,6 +107,7 @@ climo_cdf = {
    cdf_bins    = 1;
    center_bins = FALSE;
    write_bins  = TRUE;
+   direct_prob = FALSE;
 }
 
 ////////////////////////////////////////////////////////////////////////////////

met/data/config/PointStatConfig_default

@@ -149,6 +149,7 @@ climo_cdf = {
    cdf_bins    = 1;
    center_bins = FALSE;
    write_bins  = TRUE;
+   direct_prob = FALSE;
 }
 
 ////////////////////////////////////////////////////////////////////////////////

met/data/config/SeriesAnalysisConfig_default

@@ -85,6 +85,12 @@ climo_stdev = {
    file_name = [];
 }
 
+climo_cdf = {
+   cdf_bins    = 1;
+   center_bins = FALSE;
+   direct_prob = FALSE;
+}
+
 ////////////////////////////////////////////////////////////////////////////////
 
 //
@@ -111,8 +117,9 @@ mask = {
 }
 
 //
-// Number of grid points to be processed concurrently.  Set smaller to use
-// less memory but increase the number of passes through the data.
+// Number of grid points to be processed concurrently. Set smaller to use less
+// memory but increase the number of passes through the data. If set <= 0, all
+// grid points are processed concurrently.
 //
 block_size = 1024;
 

met/docs/Users_Guide/config_options.rst

@@ -1307,8 +1307,8 @@ over the "climo_mean" setting and then updating the "file_name" entry.
 The "climo_cdf" dictionary specifies how the the climatological mean
 ("climo_mean") and standard deviation ("climo_stdev") data are used to
 evaluate model performance relative to where the observation value falls
-within the climatological distribution. This dictionary consists of 3
-entries:
+within the climatological distribution. This dictionary consists of the
+following entries:
 
 (1)
     The "cdf_bins" entry defines the climatological bins either as an integer
@@ -1320,6 +1320,8 @@ entries:
 (3)
     The "write_bins" entry may be set to TRUE or FALSE.
 
+(4) The "direct_prob" entry may be set to TRUE or FALSE.
+
 MET uses the climatological mean and standard deviation to construct a normal
 PDF at each observation location. The total area under the PDF is 1, and the
 climatological CDF value is computed as the area of the PDF to the left of
@@ -1348,11 +1350,11 @@ an even number of bins can only be  uncentered. For example:
   5   centered bins (cdf_bins = 5; center_bins = TRUE;) yields:
     0.0, 0.125, 0.375, 0.625, 0.875, 1.0
 
-When multiple climatological bins are used, statistics are computed
-separately for each bin, and the average of the statistics across those bins
-is written to the output. When "write_bins" is true, the statistics for each
-bin are also written to the output. The bin number is appended to the
-contents of the VX_MASK output column.
+When multiple climatological bins are used for Point-Stat and Grid-Stat,
+statistics are computed separately for each bin, and the average of the
+statistics across those bins is written to the output. When "write_bins"
+is true, the statistics for each bin are also written to the output.
+The bin number is appended to the contents of the VX_MASK output column.
 
 Setting the number of bins to 1 effectively disables this logic by grouping
 all pairs into a single climatological bin.
@@ -1363,6 +1365,7 @@ all pairs into a single climatological bin.
      cdf_bins    = 11;    or an array of floats
      center_bins = TRUE;  or FALSE
      write_bins  = FALSE; or TRUE
+     direct_prob = FALSE; or TRUE
   }
 
 .. _climato_data:
@@ -1381,7 +1384,18 @@ directly to compute Brier Skill Score (BSS).
 When "climo_mean" and "climo_stdev" are both set to non-probability fields,
 the MET tools use the mean, standard deviation, and observation event
 threshold to derive a normal approximation of the climatological
-probabilities. Those derived probability values are used to compute BSS.
+probabilities.
+
+The "direct_prob" option controls the derivation logic. When "direct_prob" is
+true, the climatological probability is computed directly from the
+climatological distribution at each point as the area to the left of
+the event threshold value. For greater-than or greater-than-or-equal-to
+thresholds, 1.0 minus the area is used. When "direct_prob" is false, the
+"cdf_bins" values are sampled from climatological distribution. The probability
+is computed as the proportion of those samples which meet the threshold criteria.
+In this way, the number of bins impacts the resolution of the climatological
+probabilities. These derived probability values are used to compute the
+climatological Brier Score and Brier Skill Score.
 
 .. _mask_missing_flag:
 
@@ -3593,6 +3607,8 @@ Computation may be memory intensive, especially for large grids.
 The "block_size" entry sets the number of grid points to be processed
 concurrently (i.e. in one pass through a time series). Smaller values
 require less memory but increase the number of passes through the data.
+If set less than or equal to 0, it is automatically reset to the number
+of grid points, and they are all processed concurrently.
 
 .. code-block:: none
 

met/docs/Users_Guide/series-analysis.rst

@@ -123,8 +123,7 @@ ____________________
 
   block_size = 1024;
 
-Number of grid points to be processed concurrently. Set smaller to use less memory but increase the number of passes through the data. The amount of memory the Series-Analysis tool consumes is determined by the size of the grid, the length of the series, and the block_size entry defined above. The larger this entry is set the faster the tool will run, subject to the amount of memory available on the machine.
-
+Number of grid points to be processed concurrently. Set smaller to use less memory but increase the number of passes through the data. The amount of memory the Series-Analysis tool consumes is determined by the size of the grid, the length of the series, and the block_size entry defined above. The larger this entry is set the faster the tool will run, subject to the amount of memory available on the machine. If set less than or equal to 0, it is automatically reset to the number of grid points, and they are all processed concurrently.
 
 ____________________
 

met/src/basic/vx_config/config_constants.h

@@ -300,10 +300,11 @@ struct RegridInfo {
 //
 
 struct ClimoCDFInfo {
-   bool        flag;       // Flag to turn on/off climo CDF logic
-   int         n_bin;      // Number of climo CDF cdf bins
-   ThreshArray cdf_ta;     // Array of CDF thresholds
-   bool        write_bins; // Flag for writing the individual bins
+   bool        flag;        // Flag to turn on/off climo CDF logic
+   int         n_bin;       // Number of climo CDF cdf bins
+   ThreshArray cdf_ta;      // Array of CDF thresholds
+   bool        write_bins;  // Flag for writing the individual bins
+   bool        direct_prob; // Flag for the direct computation of probs
 
    ClimoCDFInfo();
    void clear();
@@ -665,6 +666,7 @@ static const char conf_key_climo_cdf[]          = "climo_cdf";
 static const char conf_key_cdf_bins[]           = "cdf_bins";
 static const char conf_key_center_bins[]        = "center_bins";
 static const char conf_key_write_bins[]         = "write_bins";
+static const char conf_key_direct_prob[]        = "direct_prob";
 static const char conf_key_time_interp_method[] = "time_interp_method";
 static const char conf_key_day_interval[]       = "day_interval";
 static const char conf_key_hour_interval[]      = "hour_interval";

met/src/basic/vx_config/config_util.cc

@@ -1498,6 +1498,7 @@ void ClimoCDFInfo::clear() {
    n_bin = 0;
    cdf_ta.clear();
    write_bins = false;
+   direct_prob = false;
 }
 
 ///////////////////////////////////////////////////////////////////////////////
@@ -1589,9 +1590,13 @@ ClimoCDFInfo parse_conf_climo_cdf(Dictionary *dict) {
    center = cdf_dict->lookup_bool(conf_key_center_bins);
 
    // Conf: write_bins
-   // Used by Grid-Stat and Point-Stat but not by Ensemble-Stat
+   // Used by Grid-Stat and Point-Stat
+   // Not used by Ensemble-Stat or Series-Analysis
    info.write_bins = cdf_dict->lookup_bool(conf_key_write_bins, false, false);
 
+   // Conf: direct_prob
+   info.direct_prob = cdf_dict->lookup_bool(conf_key_direct_prob, false, false);
+
    // Check that at least one value is provided
    if(bins.n() == 0) {
       mlog << Error << "\nparse_conf_climo_cdf() -> "

met/src/libcode/vx_statistics/compute_stats.cc

@@ -762,7 +762,7 @@ void compute_pctinfo(const PairDataPoint &pd, bool pstd_flag,
    // Use input climatological probabilities or derive them
    if(cmn_flag) {
       if(cprob_in) climo_prob = *cprob_in;
-      else         climo_prob = derive_climo_prob(pd.cdf_info,
+      else         climo_prob = derive_climo_prob(pd.cdf_info_ptr,
                                                   pd.cmn_na, pd.csd_na,
                                                   pct_info.othresh);
    }

met/src/libcode/vx_statistics/ens_stats.cc

@@ -522,7 +522,7 @@ void RPSInfo::set(const PairDataEnsemble &pd) {
       climo_pct.zero_out();
 
       // Derive climatological probabilities
-      if(cmn_flag) climo_prob = derive_climo_prob(pd.cdf_info,
+      if(cmn_flag) climo_prob = derive_climo_prob(pd.cdf_info_ptr,
                                                   pd.cmn_na, pd.csd_na,
                                                   fthresh[i]);