engine.cc

// *=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*
// ** Copyright UCAR (c) 1992 - 2022
// ** University Corporation for Atmospheric Research (UCAR)
// ** National Center for Atmospheric Research (NCAR)
// ** Research Applications Lab (RAL)
// ** P.O.Box 3000, Boulder, Colorado, 80307-3000, USA
// *=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*


///////////////////////////////////////////////////////////////////////

using namespace std;

#include <cstdio>
#include <iostream>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <cmath>

#include "engine.h"
#include "mode_columns.h"
#include "vx_util.h"

///////////////////////////////////////////////////////////////////////

static const int print_interest_log_level = 5;

///////////////////////////////////////////////////////////////////////

static inline double area_ratio_conf(double t) { return(t); }

///////////////////////////////////////////////////////////////////////
//
// Code for class ModeFuzzyEngine
//
///////////////////////////////////////////////////////////////////////

ModeFuzzyEngine::ModeFuzzyEngine() {

   init_from_scratch();
}

///////////////////////////////////////////////////////////////////////

ModeFuzzyEngine::~ModeFuzzyEngine() {

   clear_features();

   //
   // Clear fcst ShapeData objects
   //
   if(fcst_raw) {
      delete fcst_raw;
      fcst_raw = (ShapeData *) 0;
   }
   if(fcst_thresh) {
      delete fcst_thresh;
      fcst_thresh = (ShapeData *) 0;
   }
   if(fcst_conv) {
      delete fcst_conv;
      fcst_conv = (ShapeData *) 0;
   }
   if(fcst_mask) {
      delete fcst_mask;
      fcst_mask = (ShapeData *) 0;
   }
   if(fcst_split) {
      delete fcst_split;
      fcst_split = (ShapeData *) 0;
   }
   if(fcst_clus_split) {
      delete fcst_clus_split;
      fcst_clus_split = (ShapeData *) 0;
   }

   //
   // Clear obs ShapeData objects
   //
   if(obs_raw) {
      delete obs_raw;
      obs_raw = (ShapeData *) 0;
   }
   if(obs_thresh) {
      delete obs_thresh;
      obs_thresh = (ShapeData *) 0;
   }
   if(obs_conv) {
      delete obs_conv;
      obs_conv = (ShapeData *) 0;
   }
   if(obs_mask) {
      delete obs_mask;
      obs_mask = (ShapeData *) 0;
   }
   if(obs_split) {
      delete obs_split;
      obs_split = (ShapeData *) 0;
   }
   if(obs_clus_split) {
      delete obs_clus_split;
      obs_clus_split = (ShapeData *) 0;
   }

   //
   // Clear fcst and obs engines
   //
   if(fcst_engine) {
      delete fcst_engine;
      fcst_engine = (ModeFuzzyEngine *) 0;
   }
   if(obs_engine) {
      delete obs_engine;
      obs_engine = (ModeFuzzyEngine *) 0;
   }
}

///////////////////////////////////////////////////////////////////////

ModeFuzzyEngine::ModeFuzzyEngine(const ModeFuzzyEngine &eng) {

   mlog << Error << "\nModeFuzzyEngine::ModeFuzzyEngine(const ModeFuzzyEngine &) -> "
        << "should never be called!\n\n";
   exit(1);
}

///////////////////////////////////////////////////////////////////////

ModeFuzzyEngine & ModeFuzzyEngine::operator=(const ModeFuzzyEngine & eng) {

   mlog << Error << "\nModeFuzzyEngine::operator=(const ModeFuzzyEngine &) -> "
        << "should never be called\n\n";
   exit(1);
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::init_from_scratch() {

   grid = (Grid *) 0;

   //
   // Reset all fcst and obs processing flags to initial state
   //
   need_fcst_conv       = true;
   need_fcst_thresh     = true;
   need_fcst_filter     = true;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;

   need_obs_conv        = true;
   need_obs_thresh      = true;
   need_obs_filter      = true;
   need_obs_split       = true;
   need_obs_merge       = true;
   need_obs_clus_split  = true;

   need_match           = true;

   fcst_raw         = new ShapeData;
   fcst_thresh      = new ShapeData;
   fcst_conv        = new ShapeData;
   fcst_mask        = new ShapeData;
   fcst_split       = new ShapeData;
   fcst_clus_split  = new ShapeData;

   obs_raw          = new ShapeData;
   obs_thresh       = new ShapeData;
   obs_conv         = new ShapeData;
   obs_mask         = new ShapeData;
   obs_split        = new ShapeData;
   obs_clus_split   = new ShapeData;

   fcst_engine      = (ModeFuzzyEngine *) 0;
   obs_engine       = (ModeFuzzyEngine *) 0;

   n_fcst           = 0;
   n_obs            = 0;
   n_clus           = 0;

   n_valid          = 0;

   collection.clear();

   clear_features();

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::clear_features() {

   fcst_single.clear();
    obs_single.clear();

   fcst_cluster.clear();
    obs_cluster.clear();

    pair_single.clear();
   pair_cluster.clear();

   n_fcst = 0;
   n_obs  = 0;
   n_clus = 0;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::set_grid(const Grid *g) {

   grid = g;

   return;

}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::clear_colors() {
   int j;

   for(j=0; j<(fcst_color.n()); j++) {
      fcst_color[j] = unmatched_color;
   }

   for(j=0; j<(obs_color.n()); j++) {
      obs_color[j]  = unmatched_color;
   }

   return;
}


///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::set(const ShapeData &fcst_wd, const ShapeData &obs_wd)

{

   ConcatString path;

   clear_features();
   clear_colors();
   ctable.clear();

   collection.clear();

   set_fcst(fcst_wd);
   set_obs(obs_wd);

   path = replace_path(conf_info.object_pi.color_table.c_str());

   ctable.read(path.c_str());

   return;

}

///////////////////////////////////////////////////////////////////////


void ModeFuzzyEngine::set_no_conv(const ShapeData &fcst_wd, const ShapeData &obs_wd)

{

   ConcatString path;

   clear_features();
   clear_colors();
   ctable.clear();

   collection.clear();

   set_fcst_no_conv (fcst_wd);
   set_obs_no_conv  ( obs_wd);

   path = replace_path(conf_info.object_pi.color_table.c_str());

   ctable.read(path.c_str());

   return;

}


///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::set_fcst(const ShapeData &fcst_wd) {

   *fcst_raw = fcst_wd;

   need_fcst_conv       = true;
   need_fcst_thresh     = true;
   need_fcst_filter     = true;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   do_fcst_convolution();
   do_fcst_thresholding();
   do_fcst_filtering();
   do_fcst_splitting();

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::set_obs(const ShapeData &obs_wd) {

   *obs_raw = obs_wd;

   need_obs_conv       = true;
   need_obs_thresh     = true;
   need_obs_filter     = true;
   need_obs_split      = true;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   do_obs_convolution();
   do_obs_thresholding();
   do_obs_filtering();
   do_obs_splitting();

   return;
}


///////////////////////////////////////////////////////////////////////


void ModeFuzzyEngine::set_fcst_no_conv(const ShapeData &fcst_wd)

{

   // *fcst_raw = fcst_wd;

   need_fcst_conv       = false;
   need_fcst_thresh     = true;
   need_fcst_filter     = true;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   // do_fcst_convolution();
   do_fcst_thresholding();
   do_fcst_filtering();
   do_fcst_splitting();

   return;
}


///////////////////////////////////////////////////////////////////////


void ModeFuzzyEngine::set_obs_no_conv(const ShapeData &obs_wd) {

   // *obs_raw = obs_wd;

   need_obs_conv       = false;
   need_obs_thresh     = true;
   need_obs_filter     = true;
   need_obs_split      = true;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   // do_obs_convolution();
   do_obs_thresholding();
   do_obs_filtering();
   do_obs_splitting();

   return;
}


///////////////////////////////////////////////////////////////////////


int ModeFuzzyEngine::two_to_one(int n_f, int n_o) const {
   int n;

   n = n_o*n_fcst + n_f;

   return(n);
}


///////////////////////////////////////////////////////////////////////


void ModeFuzzyEngine::do_fcst_convolution() {
   int r;

   if(!need_fcst_conv) return;

   r = conf_info.fcst_conv_radius;

   *fcst_conv = *fcst_raw;

   mlog << Debug(3) << "Applying circular convolution of radius "
        << r << " to the forecast field.\n";

   //
   // Apply a circular convolution to the raw field
   //
   if(r > 0) fcst_conv->conv_filter_circ(2*r + 1, conf_info.fcst_vld_thresh);

   need_fcst_conv       = false;
   need_fcst_thresh     = true;
   need_fcst_filter     = true;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   return;
}


///////////////////////////////////////////////////////////////////////


void ModeFuzzyEngine::do_obs_convolution() {
   int r;

   if(!need_obs_conv) return;

   r = conf_info.obs_conv_radius;

   *obs_conv = *obs_raw;

   mlog << Debug(3) << "Applying circular convolution of radius "
        << r << " to the observation field.\n";

   //
   // Apply a circular convolution to the raw field
   //
   if(r > 0) obs_conv->conv_filter_circ(2*r + 1, conf_info.obs_vld_thresh);

   need_obs_conv       = false;
   need_obs_thresh     = true;
   need_obs_filter     = true;
   need_obs_split      = true;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_thresholding() {

   if(need_fcst_conv) do_fcst_convolution();

   if(!need_fcst_thresh) return;

   *fcst_mask = *fcst_conv;

   *fcst_thresh = *fcst_raw;
   fcst_thresh->threshold(conf_info.fcst_conv_thresh);

   //
   // Threshold the convolved field
   //
   fcst_mask->threshold(conf_info.fcst_conv_thresh);

   mlog << Debug(3) << "Applying convolution threshold "
        << conf_info.fcst_conv_thresh.get_str()
        << " resulted in " << fcst_mask->n_objects()
        << " simple forecast objects.\n";

   need_fcst_thresh     = false;
   need_fcst_filter     = true;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_thresholding() {

   if(need_obs_conv) do_obs_convolution();

   if(!need_obs_thresh) return;

   *obs_mask = *obs_conv;

   *obs_thresh = *obs_raw;
   obs_thresh->threshold(conf_info.obs_conv_thresh);

   //
   // Threshold the convolved field
   //
   obs_mask->threshold(conf_info.obs_conv_thresh);

   mlog << Debug(3) << "Applying convolution threshold "
        << conf_info.obs_conv_thresh.get_str()
        << " resulted in " << obs_mask->n_objects()
        << " simple observation objects.\n";


   need_obs_thresh     = false;
   need_obs_filter     = true;
   need_obs_split      = true;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_filtering() {

   if(need_fcst_thresh) do_fcst_thresholding();

   if(!need_fcst_filter) return;

   //
   // Apply object attribute filtering logic
   //
   if(conf_info.fcst_filter_attr_map.size() > 0) {

      fcst_mask->threshold_attr(conf_info.fcst_filter_attr_map,
                                fcst_raw, conf_info.fcst_conv_thresh, grid,
                                conf_info.fcst_info->is_precipitation());

      mlog << Debug(3) << "Applying object attribute filtering"
           << " resulted in " <<  fcst_mask->n_objects()
           << " simple forecast objects.\n";
   }
   else {
      mlog << Debug(3)
           << "No forecast object attribute filtering logic defined.\n";
   }

   need_fcst_filter     = false;
   need_fcst_split      = true;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_filtering() {

   if(need_obs_thresh) do_obs_thresholding();

   if(!need_obs_filter) return;

   //
   // Apply object attribute filtering logic
   //
   if(conf_info.obs_filter_attr_map.size() > 0) {

      obs_mask->threshold_attr(conf_info.obs_filter_attr_map,
                               obs_raw, conf_info.obs_conv_thresh, grid,
                               conf_info.obs_info->is_precipitation());

      mlog << Debug(3) << "Applying object attribute filtering"
           << " resulted in " <<  obs_mask->n_objects()
           << " simple observation objects.\n";
   }
   else {
      mlog << Debug(3)
           << "No observation object attribute filtering logic defined.\n";
   }

   need_obs_filter     = false;
   need_obs_split      = true;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_splitting() {

   if(need_fcst_filter) do_fcst_filtering();

   if(!need_fcst_split) return;

   *fcst_split = split(*fcst_mask, n_fcst);

   need_fcst_split      = false;
   need_fcst_merge      = true;
   need_fcst_clus_split = true;
   need_match           = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_splitting() {

   if(need_obs_filter) do_obs_filtering();

   if(!need_obs_split) return;

   *obs_split = split(*obs_mask, n_obs);

   need_obs_split      = false;
   need_obs_merge      = true;
   need_obs_clus_split = true;
   need_match          = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_merging() {

   do_fcst_merging("", "");

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_merging() {

   do_obs_merging("", "");

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_merging(const char *default_config,
                                      const char *merge_config) {

   if(need_fcst_thresh) do_fcst_thresholding();

   if(!need_fcst_merge) return;

   if(conf_info.fcst_merge_flag == MergeType_Both ||
      conf_info.fcst_merge_flag == MergeType_Thresh)
      do_fcst_merge_thresh();

   if(conf_info.fcst_merge_flag == MergeType_Both ||
      conf_info.fcst_merge_flag == MergeType_Engine)
      do_fcst_merge_engine(default_config, merge_config);

   //
   // Done
   //

   need_fcst_merge      = false;
   need_fcst_clus_split = true;
   need_match           = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_merging(const char *default_config,
                                     const char *merge_config) {

   if(need_obs_thresh) do_obs_thresholding();

   if(!need_obs_merge) return;

   if(conf_info.obs_merge_flag == MergeType_Both ||
      conf_info.obs_merge_flag == MergeType_Thresh)
      do_obs_merge_thresh();

   if(conf_info.obs_merge_flag == MergeType_Both ||
      conf_info.obs_merge_flag == MergeType_Engine)
      do_obs_merge_engine(default_config, merge_config);

   //
   // Done
   //

   need_obs_merge      = false;
   need_obs_clus_split = true;
   need_match          = true;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_matching() {

   if(!need_match) return;

   if(conf_info.match_flag == MatchType_None) {
      mlog << Warning << "\nModeFuzzyEngine::do_matching() -> "
           << "no matching requested in configuration file\n\n";
      do_no_match();
   }
   else if(conf_info.match_flag == MatchType_MergeBoth) {
      do_match_merge();
   }
   else if(conf_info.match_flag == MatchType_MergeFcst) {
      do_match_fcst_merge();
   }
   else if(conf_info.match_flag == MatchType_NoMerge) {
      do_match_only();
   }
   else {
      mlog << Error << "\nModeFuzzyEngine::do_matching() -> "
           << "invalid match_flag value of " << conf_info.match_flag
           << " specified.\n\n";
      exit(1);
   }

   //
   // Generate the fcst and obs composite split fields
   //
   do_fcst_clus_splitting();
   do_obs_clus_splitting();

   //
   // Compute the cluster single and pair features
   //
   do_cluster_features();

   mlog << Debug(3) << "Applying matching logic resulted in "
        << n_clus << " matched cluster objects.\n";

   //
   // Done
   //

   need_match           = false;
   need_fcst_clus_split = false;
   need_obs_clus_split  = false;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform no matching, but still define the single features.
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_no_match() {
   int j, k, n;
   ShapeData * fcst_shape = (ShapeData *) 0;
   ShapeData * obs_shape = (ShapeData *) 0;

   do_fcst_splitting();
   do_obs_splitting();

   clear_colors();

   fcst_shape = new ShapeData [n_fcst];
   obs_shape  = new ShapeData [n_obs];

   if(!fcst_shape || !obs_shape) {

      mlog << Error << "\nModeFuzzyEngine::do_no_match() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Do the single features
   //
   fcst_single.set_size(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_shape[j] = select(*fcst_split, j+1);
      fcst_single[j].set(*fcst_raw, *fcst_thresh, fcst_shape[j],
                         conf_info.inten_perc_value,
                         conf_info.fcst_info->is_precipitation());
      fcst_single[j].object_number = j+1;
   }

   obs_single.set_size(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_shape[j] = select(*obs_split, j+1);
      obs_single[j].set(*obs_raw, *obs_thresh, obs_shape[j],
                        conf_info.inten_perc_value,
                        conf_info.obs_info->is_precipitation());
      obs_single[j].object_number = j+1;
   }

   //
   // Set all interest values to zero
   //
   info_singles.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         info_singles[n].fcst_number    = (j+1);
         info_singles[n].obs_number     = (k+1);
         info_singles[n].pair_number    = n;
         info_singles[n].interest_value = 0.0;

      }
   }

   //
   // Clear out any empty sets
   //
   collection.clear_empty_sets();

   //
   // Assign the (unmatched) colors
   //
   fcst_color.extend(n_fcst);

   for(j=0; j<n_fcst; j++) fcst_color.add(unmatched_color);

   obs_color.extend(n_obs);

   for(j=0; j<n_obs; j++) obs_color.add(unmatched_color);

   //
   // Done
   //

   delete [] fcst_shape; fcst_shape = (ShapeData *) 0;
   delete [] obs_shape;  obs_shape  = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_match_merge() {
   int j, k, n;
   InterestInfo junkinfo;
   ShapeData * fcst_shape = (ShapeData *) 0;
   ShapeData * obs_shape = (ShapeData *) 0;

   do_fcst_splitting();
   do_obs_splitting();

   clear_colors();

   fcst_shape = new ShapeData [n_fcst];
   obs_shape = new ShapeData [n_obs];

   if(!fcst_shape || !obs_shape) {

      mlog << Error << "\nModeFuzzyEngine::do_match_merge() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Do the single features
   //
   fcst_single.set_size(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_shape[j] = select(*fcst_split, j+1);
      fcst_single[j].set(*fcst_raw, *fcst_thresh, fcst_shape[j],
                         conf_info.inten_perc_value,
                         conf_info.fcst_info->is_precipitation());
      fcst_single[j].object_number = j+1;
   }

   obs_single.set_size(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_shape[j] = select(*obs_split, j+1);
      obs_single[j].set(*obs_raw, *obs_thresh, obs_shape[j],
                        conf_info.inten_perc_value,
                        conf_info.obs_info->is_precipitation());
      obs_single[j].object_number = j+1;
   }

   //
   // Do the pair features
   //
   pair_single.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         pair_single[n].set(fcst_single[j], obs_single[k],
                     conf_info.max_centroid_dist);
         pair_single[n].pair_number = n;
      }
   }

   //
   // Calculate the interest values
   //
   info_singles.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         info_singles[n].fcst_number    = (j+1);
         info_singles[n].obs_number     = (k+1);
         info_singles[n].pair_number    = n;
         info_singles[n].interest_value = total_interest(conf_info,
                                             pair_single[n], j+1, k+1,
                                             true);

      }
   }

   //
   // Sort the interest values in decreasing order
   //
   n = n_fcst*n_obs;

   for(j=0; j<(n-1); j++) {
      for(k=(j+1); k<n; k++) {

         if(info_singles[j].interest_value < info_singles[k].interest_value) {

            junkinfo = info_singles[j];
            info_singles[j] = info_singles[k];
            info_singles[k] = junkinfo;
         }
      }
   }

   //
   // Form the sets
   //
   n = n_fcst*n_obs;

   for(j=0; j<n; j++) {

      if(info_singles[j].interest_value < conf_info.total_interest_thresh)
         continue;

      collection.add_pair(info_singles[j].fcst_number, info_singles[j].obs_number);
   }

   //
   // Clear out any empty sets
   //
   collection.clear_empty_sets();

   //
   // Assign the colors
   //
   if(collection.n_sets > ctable.n_entries()) {
      mlog << Warning << "\nModeFuzzyEngine::do_match_merge() -> "
           << "more object matches (" << collection.n_sets
           << ") than defined object colors (" << ctable.n_entries()
           << ") ... reusing some colors!\n\n";
   }

   fcst_color.extend(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {

         if(collection.set[k].has_fcst(j+1)) {
            fcst_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }
   }

   obs_color.extend(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {

         if(collection.set[k].has_obs(j+1)) {
            obs_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }
   }

   //
   // Done
   //

   delete [] fcst_shape; fcst_shape = (ShapeData *) 0;
   delete [] obs_shape;  obs_shape = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform merging of the forecast field based on a different convolution
// threshold.  It is the user's responsibility to choose a threshold type and
// value to define objects which are suitable for merging.
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_merge_thresh() {
   int j, k, x, y;
   int n_fcst_merge, intersection;
   int count, first_k;
   ShapeData fcst_merge_mask, fcst_merge_split;
   ShapeData * fcst_shape = (ShapeData *) 0;
   ShapeData * fcst_merge_shape = (ShapeData *) 0;

   do_fcst_splitting();

   //
   // Define the forecast merge field by applying the specified threshold
   // to the convolved field
   //
   fcst_merge_mask = *fcst_conv;

   //
   // Threshold the forecast merge field
   //
   fcst_merge_mask.threshold(conf_info.fcst_merge_thresh);

   //
   // Split up the forecast merge field
   //
   fcst_merge_split = split(fcst_merge_mask, n_fcst_merge);

   //
   // Allocate space for all of the simple forecast shapes and
   // forecast merge shapes
   //
   fcst_shape       = new ShapeData [n_fcst];
   fcst_merge_shape = new ShapeData [n_fcst_merge];

   if(!fcst_shape || !fcst_merge_shape) {

      mlog << Error << "\nModeFuzzyEngine::do_fcst_merge_thresh() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Select all of the simple forecast shapes and
   // forecast merge shapes
   //
   for(j=0; j<n_fcst; j++) {
      fcst_shape[j] = select(*fcst_split, j+1);
   }

   for(j=0; j<n_fcst_merge; j++) {
      fcst_merge_shape[j] = select(fcst_merge_split, j+1);
   }

   //
   // Calculate the composite object sets
   //
   for(j=0; j<n_fcst_merge; j++) {

      count = first_k = 0;

      for(k=0; k<n_fcst; k++) {

         //
         // Calculate intersection area
         //
         intersection = 0;

         for(x=0; x<fcst_merge_mask.data.nx(); x++) {
            for(y=0; y<fcst_merge_mask.data.ny(); y++) {

               if(fcst_merge_shape[j].s_is_on(x, y) &&
                  fcst_shape[k].s_is_on(x, y)) intersection++;
            }
         }

         //
         // Add to set collection only if the fcst object is
         // completely contained in the merge object.  Meaning,
         // intersection area >= fcst area
         //

         if(intersection >= fcst_shape[k].area()) {

            collection.make_room();

            //
            // Keep track of the first embedded shape.  You only want to
            // create a composite if there are more than 1 shapes in it.
            //
            if(count == 0) first_k = k+1;

            else if(count == 1) {
               collection.set[collection.n_sets].add_pair(first_k, -1);
               collection.set[collection.n_sets].add_pair(k+1, -1);
            }
            else {
               collection.set[collection.n_sets].add_pair(k+1, -1);
            }

            count++;
         }
      }  // end for k

      if(count > 0) collection.n_sets++;
   }  // end for j

   //
   // Done
   //

   delete [] fcst_shape;       fcst_shape = (ShapeData *) 0;
   delete [] fcst_merge_shape; fcst_merge_shape = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform merging of the observation field based on a different convolution
// threshold.  It is the user's responsibility to choose a threshold type and
// value to define objects which are suitable for merging.
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_merge_thresh() {
   int j, k, x, y;
   int n_obs_merge, intersection;
   int count, first_k;
   ShapeData obs_merge_mask, obs_merge_split;
   ShapeData * obs_shape = (ShapeData *) 0;
   ShapeData * obs_merge_shape = (ShapeData *) 0;

   do_obs_splitting();

   //
   // Define the forecast merge field by applying the specified threshold
   // to the convolved field
   //
   obs_merge_mask = *obs_conv;

   //
   // Threshold the forecast merge field
   //
   obs_merge_mask.threshold(conf_info.obs_merge_thresh);

   //
   // Split up the forecast merge field
   //
   obs_merge_split = split(obs_merge_mask, n_obs_merge);

   //
   // Allocate space for all of the simple observation shapes and
   // observation merge shapes
   //
   obs_shape       = new ShapeData [n_obs];
   obs_merge_shape = new ShapeData [n_obs_merge];

   if(!obs_shape || !obs_merge_shape) {

      mlog << Error << "\nModeFuzzyEngine::do_obs_merge_thresh() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Select all of the simple observation shapes and
   // simple merge shapes
   //
   for(j=0; j<n_obs; j++) {
      obs_shape[j] = select(*obs_split, j+1);
   }

   for(j=0; j<n_obs_merge; j++) {
      obs_merge_shape[j] = select(obs_merge_split, j+1);
   }

   //
   // Calculate the composite object sets
   //
   for(j=0; j<n_obs_merge; j++) {

      count = first_k = 0;

      for(k=0; k<n_obs; k++) {

         //
         // Calculate intersection area
         //
         intersection = 0;

         for(x=0; x<obs_merge_mask.data.nx(); x++) {
            for(y=0; y<obs_merge_mask.data.ny(); y++) {

               if(obs_merge_shape[j].s_is_on(x, y) &&
                  obs_shape[k].s_is_on(x, y)) intersection++;
            }
         }

         //
         // Add to set collection only if the obs object is
         // completely contained in the merge object.  Meaning,
         // intersection area >= obs area
         //
         if(intersection >= obs_shape[k].area()) {

            collection.make_room();

            //
            // Keep track of the first embedded shape.  You only want to
            // create a composite if there are more than 1 shapes in it.
            //
            if(count == 0) first_k = k+1;

            else if(count == 1) {
               collection.set[collection.n_sets].add_pair(-1, first_k);
               collection.set[collection.n_sets].add_pair(-1, k+1);
            }
            else {
               collection.set[collection.n_sets].add_pair(-1, k+1);
            }

            count++;
         }
      }  // end for k

      if(count > 0) collection.n_sets++;
   }  // end for j

   //
   // Done
   //

   delete [] obs_shape;       obs_shape = (ShapeData *) 0;
   delete [] obs_merge_shape; obs_merge_shape = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform merging of the forecast field using a fuzzy engine matching
// approach
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_merge_engine(const char *default_config,
                                           const char *merge_config)

{

   int i, j;
   ShapeData fcst_merge_split;
   ConcatString path;

   do_fcst_splitting();

   //
   // Will be deleted by destructor if allocated
   //
   fcst_engine = new ModeFuzzyEngine;

   if(!fcst_engine) {
      mlog << Error << "\nModeFuzzyEngine::do_fcst_merge_engine() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Store the grid
   //
   fcst_engine->grid = grid;

   //
   // Specify the configuration for the forecast merging fuzzy engine
   //
   fcst_engine->ctable = ctable;
   if(default_config && merge_config) {
      fcst_engine->conf_info.read_config(default_config, merge_config);
      fcst_engine->conf_info.process_config(conf_info.fcst_info->file_type(),
                                            conf_info.obs_info->file_type());
      path = replace_path(fcst_engine->conf_info.object_pi.color_table.c_str());
      fcst_engine->ctable.read(path.c_str());
   }

   //
   // Copy over fcst_info and obs_info
   //
   *(fcst_engine->conf_info.fcst_info) = *(conf_info.fcst_info);
   *(fcst_engine->conf_info.obs_info)  = *(conf_info.fcst_info);

   //
   // Copy over the forecast threshold values
   //
   fcst_engine->conf_info.fcst_conv_thresh     = conf_info.fcst_conv_thresh;
   fcst_engine->conf_info.obs_conv_thresh      = conf_info.fcst_conv_thresh;

   fcst_engine->conf_info.fcst_filter_attr_map = conf_info.fcst_filter_attr_map;
   fcst_engine->conf_info.obs_filter_attr_map  = conf_info.fcst_filter_attr_map;

   fcst_engine->conf_info.fcst_merge_thresh    = conf_info.fcst_merge_thresh;
   fcst_engine->conf_info.obs_merge_thresh     = conf_info.fcst_merge_thresh;

   //
   // Copy the previously defined fuzzy engine fields
   // setting the current forecast field to both the
   // forecast and forecast fields in the fcst_engine.
   // this will capture any previous merging performed
   // in the current forecast field.
   //
   *(fcst_engine->fcst_raw)    = *(fcst_raw);
   *(fcst_engine->obs_raw)     = *(fcst_raw);

   *(fcst_engine->fcst_conv)   = *(fcst_conv);
   *(fcst_engine->obs_conv)    = *(fcst_conv);

   *(fcst_engine->fcst_thresh) = *(fcst_thresh);
   *(fcst_engine->obs_thresh)  = *(fcst_thresh);

   *(fcst_engine->fcst_mask)   = *(fcst_mask);
   *(fcst_engine->obs_mask)    = *(fcst_mask);

   *(fcst_engine->fcst_split)  = *(fcst_split);
   *(fcst_engine->obs_split)   = *(fcst_split);

   fcst_engine->n_fcst = n_fcst;
   fcst_engine->n_obs  = n_fcst;

   fcst_engine->need_fcst_conv       = false;
   fcst_engine->need_fcst_thresh     = false;
   fcst_engine->need_fcst_split      = false;
   fcst_engine->need_fcst_merge      = true;
   fcst_engine->need_fcst_clus_split = true;

   fcst_engine->need_obs_conv        = false;
   fcst_engine->need_obs_thresh      = false;
   fcst_engine->need_obs_split       = false;
   fcst_engine->need_obs_merge       = true;
   fcst_engine->need_obs_clus_split  = true;

   fcst_engine->need_match           = true;

   //
   // Copy the top level set collection down to the fcst engine
   // to capture any merging that has already occurred, via
   // the double threshold method
   //
   for(i=0; i<collection.n_sets; i++) {
      if(collection.set[i].n_fcst > 0) {
         fcst_engine->collection.make_room();
         for(j=0; j<collection.set[i].n_fcst; j++) {

            fcst_engine->collection.set[fcst_engine->collection.n_sets].add_pair(
               -1, collection.set[i].fcst_number[j]);
         }
         fcst_engine->collection.n_sets++;
      }
   }

   //
   // Perform merging and matching for the fcst_engine using the
   // simple merging and matching approach
   //
   fcst_engine->do_match_merge();

   //
   // Generate the fcst and obs composite split fields
   //
   fcst_engine->do_fcst_clus_splitting();
   fcst_engine->do_obs_clus_splitting();

   //
   // Based on the results of the merging and matching
   // of the fcst_engine, update the set collection
   // to indicate merging
   //
   for(i=0; i<fcst_engine->collection.n_sets; i++) {

      //
      // All objects will at least match themselves.
      // only define sets when they are composed of two or
      // more simple objects.
      //
      if(fcst_engine->collection.set[i].n_fcst >= 2) {


         for(j=0; j<fcst_engine->collection.set[i].n_fcst; j++) {

            collection.make_room();

            collection.set[collection.n_sets].add_pair(
               fcst_engine->collection.set[i].fcst_number[j], -1);
         }
         collection.n_sets++;
      }
   } // end for i

   //
   // Compute the cluster single and pair features
   //
   fcst_engine->do_cluster_features();

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform merging of the observation field using a fuzzy engine
// matching approach
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_merge_engine(const char *default_config,
                                          const char *merge_config) {
   int i, j;
   ShapeData obs_merge_split;
   ConcatString path;

   do_obs_splitting();

   //
   // Will be deleted by destructor if allocated
   //
   obs_engine = new ModeFuzzyEngine;

   if(!obs_engine) {
      mlog << Error << "\nModeFuzzyEngine::do_obs_merge_engine() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Store the grid
   //
   obs_engine->grid = grid;

   //
   // Specify the configuration for the observation merging fuzzy engine
   //
   obs_engine->ctable = ctable;
   if(default_config && merge_config) {
      obs_engine->conf_info.read_config(default_config, merge_config);
      obs_engine->conf_info.process_config(conf_info.fcst_info->file_type(),
                                           conf_info.obs_info->file_type());
      path = replace_path(obs_engine->conf_info.object_pi.color_table.c_str());
      obs_engine->ctable.read(path.c_str());
   }

   //
   // Copy over fcst_info and obs_info
   //
   *(obs_engine->conf_info.fcst_info) = *(conf_info.obs_info);
   *(obs_engine->conf_info.obs_info)  = *(conf_info.obs_info);

   //
   // Copy over the observation threshold values
   //
   obs_engine->conf_info.fcst_conv_thresh     = conf_info.obs_conv_thresh;
   obs_engine->conf_info.obs_conv_thresh      = conf_info.obs_conv_thresh;

   obs_engine->conf_info.fcst_filter_attr_map = conf_info.obs_filter_attr_map;
   obs_engine->conf_info.obs_filter_attr_map  = conf_info.obs_filter_attr_map;

   obs_engine->conf_info.fcst_merge_thresh    = conf_info.obs_merge_thresh;
   obs_engine->conf_info.obs_merge_thresh     = conf_info.obs_merge_thresh;

   //
   // Copy the previously defined fuzzy engine fields
   // setting the current observation field to both the
   // forecast and observation fields in the obs_engine.
   // this will capture any previous merging performed
   // in the current observation field.
   //
   *(obs_engine->fcst_raw)    = *(obs_raw);
   *(obs_engine->obs_raw)     = *(obs_raw);

   *(obs_engine->fcst_conv)   = *(obs_conv);
   *(obs_engine->obs_conv)    = *(obs_conv);

   *(obs_engine->fcst_thresh) = *(obs_thresh);
   *(obs_engine->obs_thresh)  = *(obs_thresh);

   *(obs_engine->fcst_mask)   = *(obs_mask);
   *(obs_engine->obs_mask)    = *(obs_mask);

   *(obs_engine->fcst_split)  = *(obs_split);
   *(obs_engine->obs_split)   = *(obs_split);

   obs_engine->n_fcst = n_obs;
   obs_engine->n_obs  = n_obs;

   obs_engine->need_fcst_conv       = false;
   obs_engine->need_fcst_thresh     = false;
   obs_engine->need_fcst_split      = false;
   obs_engine->need_fcst_merge      = true;
   obs_engine->need_fcst_clus_split = true;

   obs_engine->need_obs_conv        = false;
   obs_engine->need_obs_thresh      = false;
   obs_engine->need_obs_split       = false;
   obs_engine->need_obs_merge       = true;
   obs_engine->need_obs_clus_split  = true;

   obs_engine->need_match           = true;

   //
   // Copy the top level set collection down to the obs engine
   // to capture any merging that has already occurred, via
   // the double threshold method.  But only copy down those
   // sets containing merged observation objects.
   //
   for(i=0; i<collection.n_sets; i++) {
      if(collection.set[i].n_obs > 0) {
         obs_engine->collection.make_room();
         for(j=0; j<collection.set[i].n_obs; j++) {

            obs_engine->collection.make_room();

            obs_engine->collection.set[obs_engine->collection.n_sets].add_pair(
               -1, collection.set[i].obs_number[j]);
         }
         obs_engine->collection.n_sets++;
      }
   }

   //
   // Perform merging and matching for the obs_engine using the
   // simple merging and matching approach
   //
   obs_engine->do_match_merge();

   //
   // Generate the fcst and obs composite split fields
   //
   obs_engine->do_fcst_clus_splitting();
   obs_engine->do_obs_clus_splitting();

   //
   // Based on the results of the merging and matching
   // of the obs_engine, update the set collection
   // to indicate merging
   //
   for(i=0; i<obs_engine->collection.n_sets; i++) {

      //
      // All objects will at least match themselves.
      // only define sets when they are composed of two or
      // more simple objects.
      //
      if(obs_engine->collection.set[i].n_obs >= 2) {
         collection.make_room();
         for(j=0; j<obs_engine->collection.set[i].n_obs; j++) {

            collection.set[collection.n_sets].add_pair(
               -1, obs_engine->collection.set[i].obs_number[j]);
         }
         collection.n_sets++;
      }
   }

   //
   // Compute the cluster single and pair features
   //
   obs_engine->do_cluster_features();

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform matching between the already merged forecast and
// observation fields allowing no additional merging of objects in
// the observation field
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_match_fcst_merge() {
   int j, k, n;
   InterestInfo junkinfo;
   ShapeData * fcst_shape = (ShapeData *) 0;
   ShapeData * obs_shape = (ShapeData *) 0;

   do_fcst_splitting();
   do_obs_splitting();

   clear_colors();

   fcst_shape = new ShapeData [n_fcst];
   obs_shape = new ShapeData [n_obs];

   if(!fcst_shape || !obs_shape) {
      mlog << Error << "\nModeFuzzyEngine::do_match_fcst_merge() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Do the single features
   //
   fcst_single.set_size(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_shape[j] = select(*fcst_split, j+1);
      fcst_single[j].set(*fcst_raw, *fcst_thresh, fcst_shape[j],
                         conf_info.inten_perc_value,
                         conf_info.fcst_info->is_precipitation());
      fcst_single[j].object_number = j+1;
   }

   obs_single.set_size(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_shape[j] = select(*obs_split, j+1);
      obs_single[j].set(*obs_raw, *obs_thresh, obs_shape[j],
                        conf_info.inten_perc_value,
                        conf_info.obs_info->is_precipitation());
      obs_single[j].object_number = j+1;
   }

   //
   // Do the pair features
   //
   pair_single.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         pair_single[n].set(fcst_single[j], obs_single[k],
                     conf_info.max_centroid_dist);
         pair_single[n].pair_number = n;
      }
   }

   //
   // Calculate the interest values
   //
   info_singles.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         info_singles[n].fcst_number    = (j+1);
         info_singles[n].obs_number     = (k+1);
         info_singles[n].pair_number    = n;
         info_singles[n].interest_value = total_interest(conf_info,
                                             pair_single[n], j+1, k+1,
                                             true);

      }
   }

   //
   // Sort the interest values in decreasing order
   //
   n = n_fcst*n_obs;

   for(j=0; j<(n-1); j++) {
      for(k=(j+1); k<n; k++) {

         if(info_singles[j].interest_value < info_singles[k].interest_value) {

            junkinfo = info_singles[j];
            info_singles[j] = info_singles[k];
            info_singles[k] = junkinfo;

         }
      }
   }

   //
   // Form the sets
   //
   n = n_fcst*n_obs;

   for(j=0; j<n; j++) {

      if(info_singles[j].interest_value < conf_info.total_interest_thresh) {
         continue;
      }

      //
      // Only add this pair if the forecast object is not already matched
      //
      if(!collection.is_fcst_matched(info_singles[j].fcst_number)) {
         collection.add_pair(info_singles[j].fcst_number, info_singles[j].obs_number);
      }
   }

   //
   // Clear out any empty sets
   //
   collection.clear_empty_sets();

   //
   // Assign the colors
   //
   if(collection.n_sets > ctable.n_entries()) {
      mlog << Warning << "\nModeFuzzyEngine::do_match_fcst_merge() -> "
           << "more object matches (" << collection.n_sets
           << ") than defined object colors (" << ctable.n_entries()
           << ") ... reusing some colors!\n\n";
   }

   fcst_color.extend(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {

         if(collection.set[k].has_fcst(j+1)) {
            fcst_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }

   }

   obs_color.extend(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {

         if(collection.set[k].has_obs(j+1)) {
            obs_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }
   }

   //
   // Done
   //

   delete [] fcst_shape;  fcst_shape = (ShapeData *) 0;
   delete [] obs_shape;   obs_shape = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Perform matching between the already merged forecast and observation
// fields allowing no additional merging of objects in the either field
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_match_only() {
   int j, k, n;
   InterestInfo junkinfo;
   ShapeData * fcst_shape = (ShapeData *) 0;
   ShapeData * obs_shape = (ShapeData *) 0;

   do_fcst_splitting();
   do_obs_splitting();

   clear_colors();

   fcst_shape = new ShapeData [n_fcst];
   obs_shape = new ShapeData [n_obs];

   if(!fcst_shape || !obs_shape) {
      mlog << Error << "\nModeFuzzyEngine::do_match_only() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Do the single features
   //
   fcst_single.set_size(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_shape[j] = select(*fcst_split, j+1);
      fcst_single[j].set(*fcst_raw, *fcst_thresh, fcst_shape[j],
                         conf_info.inten_perc_value,
                         conf_info.fcst_info->is_precipitation());
      fcst_single[j].object_number = j+1;
   }

   obs_single.set_size(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_shape[j] = select(*obs_split, j+1);
      obs_single[j].set(*obs_raw, *obs_thresh, obs_shape[j],
                        conf_info.inten_perc_value,
                        conf_info.obs_info->is_precipitation());
      obs_single[j].object_number = j+1;
   }

   //
   // Do the pair features
   //
   pair_single.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         pair_single[n].set(fcst_single[j], obs_single[k],
                     conf_info.max_centroid_dist);
         pair_single[n].pair_number = n;
      }
   }

   //
   // Calculate the interest values
   //
   info_singles.set_size(n_fcst*n_obs);

   for(j=0; j<n_fcst; j++) {
      for(k=0; k<n_obs; k++) {

         n = two_to_one(j, k);

         info_singles[n].fcst_number    = (j+1);
         info_singles[n].obs_number     = (k+1);
         info_singles[n].pair_number    = n;
         info_singles[n].interest_value = total_interest(conf_info,
                                             pair_single[n], j+1, k+1,
                                             true);
      }
   }

   //
   // Sort the interest values in decreasing order
   //
   n = n_fcst*n_obs;
   for(j=0; j<(n-1); j++) {
      for(k=(j+1); k<n; k++) {

         if(info_singles[j].interest_value < info_singles[k].interest_value) {

            junkinfo = info_singles[j];
            info_singles[j] = info_singles[k];
            info_singles[k] = junkinfo;
         }
      }
   }

   //
   // Form the sets
   //
   n = n_fcst*n_obs;
   for(j=0; j<n; j++) {

      if(info_singles[j].interest_value < conf_info.total_interest_thresh)
         continue;

      //
      // Only add this pair if both the forecast and observation objects
      // are not already matched
      //
      if(!collection.is_fcst_matched(info_singles[j].fcst_number) &&
         !collection.is_obs_matched(info_singles[j].obs_number)) {
         collection.add_pair(info_singles[j].fcst_number, info_singles[j].obs_number);
      }
   }

   //
   // Clear out any empty sets
   //
   collection.clear_empty_sets();

   //
   // Assign the colors
   //
   if(collection.n_sets > ctable.n_entries()) {
      mlog << Warning << "\nModeFuzzyEngine::do_match_only() -> "
           << "more object matches (" << collection.n_sets
           << ") than defined object colors (" << ctable.n_entries()
           << ") ... reusing some colors!\n\n";
   }

   fcst_color.extend(n_fcst);

   for(j=0; j<n_fcst; j++) {
      fcst_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {
         if(collection.set[k].has_fcst(j+1)) {
            fcst_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }
   }

   obs_color.extend(n_obs);

   for(j=0; j<n_obs; j++) {
      obs_color.add(unmatched_color);

      for(k=0; k<(collection.n_sets); k++) {
         if(collection.set[k].has_obs(j+1)) {
            obs_color[j] = ctable.nearest((k%ctable.n_entries())+1);
            break;
         }
      }
   }

   //
   // Done
   //

   delete [] fcst_shape;   fcst_shape = (ShapeData *) 0;
   delete [] obs_shape;    obs_shape = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Create a split composite field based on the merging of the single objects
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_fcst_clus_splitting() {
   int x, y, set_number;
   double v;
   ShapeData junk;

   //
   // Initialize fcst_clus_split
   //
   *fcst_clus_split = *fcst_split;

   for(x=0; x<fcst_clus_split->data.nx(); x++) {
      for(y=0; y<fcst_clus_split->data.ny(); y++) {

         v = fcst_split->data(x, y);

         if( is_eq(v, 0.0) ) continue;

         set_number = collection.fcst_set_number(nint(v));

         fcst_clus_split->data.set(set_number + 1, x, y);
      }  // end for y
   }  // end for x

   fcst_clus_split->calc_moments();

   //
   // Done
   //

   need_fcst_clus_split = false;

   return;
}

///////////////////////////////////////////////////////////////////////
//
// Create a split composite field based on the merging of the single
// objects
//
///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_obs_clus_splitting() {
   int x, y, set_number;
   double v;
   ShapeData junk;

   //
   // Initialize obs_clus_split
   //
   *obs_clus_split = *obs_split;

   for(x=0; x<obs_clus_split->data.nx(); x++) {
      for(y=0; y<obs_clus_split->data.ny(); y++) {

         v = obs_split->data(x, y);

         if(is_eq(v, 0.0)) continue;

         set_number = collection.obs_set_number(nint(v));

         obs_clus_split->data.set(set_number + 1, x, y);
      }  // end for y
   }  // end for x

   obs_clus_split->calc_moments();

   //
   // Done
   //

   need_obs_clus_split = false;

   return;
}

///////////////////////////////////////////////////////////////////////

void ModeFuzzyEngine::do_cluster_features() {
   int j;
   ShapeData * fcst_clus_shape = (ShapeData *) 0;
   ShapeData * obs_clus_shape  = (ShapeData *) 0;

   if(need_fcst_clus_split) do_fcst_clus_splitting();
   if(need_obs_clus_split)  do_obs_clus_splitting();

   //
   // Store the number of clusters
   //
   n_clus = collection.n_sets;

   fcst_clus_shape = new ShapeData [n_clus];
   obs_clus_shape  = new ShapeData [n_clus];

   if(!fcst_clus_shape || !obs_clus_shape) {

      mlog << Error << "\nModeFuzzyEngine::do_cluster_features() -> "
           << "memory allocation error\n\n";
      exit(1);
   }

   //
   // Do the single features for clusters
   //
   fcst_cluster.set_size(n_clus);
    obs_cluster.set_size(n_clus);

   for(j=0; j<n_clus; j++) {
      fcst_clus_shape[j] = select(*fcst_clus_split, j+1);
      fcst_cluster[j].set(*fcst_raw, *fcst_thresh, fcst_clus_shape[j],
                       conf_info.inten_perc_value,
                       conf_info.fcst_info->is_precipitation());
      fcst_cluster[j].object_number = j+1;

      obs_clus_shape[j] = select(*obs_clus_split, j+1);
      obs_cluster[j].set(*obs_raw, *obs_thresh, obs_clus_shape[j],
                      conf_info.inten_perc_value,
                      conf_info.obs_info->is_precipitation());
      obs_cluster[j].object_number = j+1;
   }

   //
   // Do the pair features
   //
   pair_cluster.set_size(n_clus);

   for(j=0; j<n_clus; j++) {
      pair_cluster[j].set(fcst_cluster[j], obs_cluster[j],
                       conf_info.max_centroid_dist);
      pair_cluster[j].pair_number = j+1;
   }

   //
   // Calculate the interest values
   //
   info_clus.set_size(n_clus);

   for(j=0; j<n_clus; j++) {
      info_clus[j].fcst_number    = (j+1);
      info_clus[j].obs_number     = (j+1);
      info_clus[j].pair_number    = j;
      info_clus[j].interest_value = total_interest(conf_info,
                                       pair_cluster[j], j+1, j+1,
                                       false);
   }

   //
   // Done
   //

   delete [] fcst_clus_shape; fcst_clus_shape = (ShapeData *) 0;
   delete [] obs_clus_shape;  obs_clus_shape  = (ShapeData *) 0;

   return;
}

///////////////////////////////////////////////////////////////////////

int ModeFuzzyEngine::get_info_index(int pair_n) const {
   int i;

   for(i=0; i<(info_singles.n()); i++) {

      if(info_singles[i].pair_number == pair_n) return(i);

   }

   return(-1);
}

///////////////////////////////////////////////////////////////////////

int ModeFuzzyEngine::get_matched_fcst(int area) const {
   int i, count;

   count = 0;

   for(i=0; i<n_fcst; i++) {

      if(collection.fcst_set_number(i+1) != -1) {
         if(area) count += nint(fcst_single[i].area);
         else     count += 1;
      }
   }

   return(count);
}

///////////////////////////////////////////////////////////////////////

int ModeFuzzyEngine::get_unmatched_fcst(int area) const {
   int i, count;

   count = 0;

   for(i=0; i<n_fcst; i++) {

      if(collection.fcst_set_number(i+1) == -1) {
         if(area) count += nint(fcst_single[i].area);
         else     count += 1;
      }
   }

   return(count);
}

///////////////////////////////////////////////////////////////////////

int ModeFuzzyEngine::get_matched_obs(int area) const {
   int i, count;

   count = 0;

   for(i=0; i<n_obs; i++) {

      if(collection.obs_set_number(i+1) != -1) {
         if(area) count += nint(obs_single[i].area);
         else     count += 1;
      }
   }

   return(count);
}

///////////////////////////////////////////////////////////////////////

int ModeFuzzyEngine::get_unmatched_obs(int area) const {
   int i, count;

   count = 0;

   for(i=0; i<n_obs; i++) {

      if(collection.obs_set_number(i+1) == -1) {
         if(area) count += nint(obs_single[i].area);
         else     count += 1;
      }
   }

   return(count);
}

///////////////////////////////////////////////////////////////////////

double total_interest(ModeConfInfo &mc, const PairFeature &p,
                      int fcst_num, int obs_num, bool is_single) {
   double attribute;
   double interest, weight;
   double confidence;
   double aspect_obs, aspect_fcst;
   double conf_obs, conf_fcst;
   double term, sum, weight_sum;
   double total;

   if(mlog.verbosity_level() >= print_interest_log_level) {

      ConcatString cs;
      if(is_single) {
         cs.format("single pair F%03d_O%03d", fcst_num, obs_num);
      }
      else {
         cs.format("cluster pair CF%03d_CO%03d", fcst_num, obs_num);
      }

      mlog << Debug(print_interest_log_level)
           << "\n" << sep_str << "\n\n"
           << "Computing total interest for " << cs << ".\n\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // If the distance flag is set and the centroid distance is too
   // large, don't compute the interest for this pair.
   //
   ////////////////////////////////////////////////////////////////////

   if(is_single && p.centroid_dist > mc.max_centroid_dist) {
      total = bad_data_double;

      if(mlog.verbosity_level() >= print_interest_log_level) {
         mlog << Debug(print_interest_log_level)
              << "Total Interest = " << total << "\n"
              << "Centroid Distance (" << p.centroid_dist
              << ") > Max Centroid Distance ("
              << mc.max_centroid_dist << ")\n";
      }
      return(total);
   }

   sum = 0.0;
   weight_sum = 0.0;

   ////////////////////////////////////////////////////////////////////
   //
   // Centroid distance
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.centroid_dist;
   interest    = (*mc.centroid_dist_if)(attribute);
   confidence  = area_ratio_conf(p.area_ratio);
   weight      = mc.centroid_dist_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Centroid Distance:\n"
           << "------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Boundary distance
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.boundary_dist;
   interest    = (*mc.boundary_dist_if)(attribute);
   confidence  = 1.0;
   weight      = mc.boundary_dist_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Boundary Distance:\n"
           << "------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Convex Hull distance
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.convex_hull_dist;
   interest    = (*mc.convex_hull_dist_if)(attribute);
   confidence  = 1.0;
   weight      = mc.convex_hull_dist_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Convex Hull Distance:\n"
           << "---------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Angle difference
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.angle_diff;
   interest    = (*mc.angle_diff_if)(attribute);
   aspect_obs  = p.Obs->aspect_ratio;
   aspect_fcst = p.Fcst->aspect_ratio;
   conf_obs    = aspect_ratio_conf(aspect_obs);
   conf_fcst   = aspect_ratio_conf(aspect_fcst);
   confidence  = sqrt(conf_obs*conf_fcst);
   weight      = mc.angle_diff_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Angle Difference:\n"
           << "-----------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Aspect ratio difference
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.aspect_diff;
   interest    = (*mc.aspect_diff_if)(attribute);
   confidence  = 1.0;
   weight      = mc.aspect_diff_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Aspect Ratio Difference:\n"
           << "------------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Area ratio
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.area_ratio;
   interest    = (*mc.area_ratio_if)(attribute);
   confidence  = 1.0;
   weight      = mc.area_ratio_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Area Ratio:\n"
           << "-----------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Intersection/area ratio
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = (p.intersection_area)/(min(p.Obs->area, p.Fcst->area));
   interest    = (*mc.int_area_ratio_if)(attribute);
   confidence  = 1.0;
   weight      = mc.int_area_ratio_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Intersection/Area Ratio:\n"
           << "------------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Curvature ratio
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.curvature_ratio;
   interest    = (*mc.curvature_ratio_if)(attribute);
   confidence  = 1.0;
   weight      = mc.curvature_ratio_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Curvature Ratio:\n"
           << "----------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Complexity ratio
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.complexity_ratio;
   interest    = (*mc.complexity_ratio_if)(attribute);
   confidence  = 1.0;
   weight      = mc.complexity_ratio_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Complexity Ratio:\n"
           << "-----------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Nth Percentile intensity ratio
   //
   ////////////////////////////////////////////////////////////////////

   attribute   = p.percentile_intensity_ratio;
   interest    = (*mc.inten_perc_ratio_if)(attribute);
   confidence  = 1.0;
   weight      = mc.inten_perc_ratio_wt;
   term        = weight*interest*confidence;
   sum        += term;
   weight_sum += weight*confidence;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Percentile (" << mc.inten_perc_value
           << "th) Intensity Ratio:\n"
           << "--------------------\n"
           << "   Value      = " << attribute  << "\n"
           << "   Interest   = " << interest   << "\n"
           << "   Confidence = " << confidence << "\n"
           << "   Weight     = " << weight     << "\n"
           << "   Term       = " << term       << "\n"
           << "\n";
   }

   ////////////////////////////////////////////////////////////////////
   //
   // Done
   //
   ////////////////////////////////////////////////////////////////////

   total = sum/weight_sum;

   if(mlog.verbosity_level() >= print_interest_log_level) {
      mlog << Debug(print_interest_log_level)
           << "Total Interest = (sum of terms)/(sum of weights*confidence)\n\n"
           << "               = " << sum << "/" << weight_sum << "\n\n"
           << "               = " << total << "\n\n";
   }

   return(total);
}

///////////////////////////////////////////////////////////////////////

double interest_percentile(ModeFuzzyEngine &eng, const double p, const int flag) {
   int i, fcst_i, obs_i, n_values;
   double interest, ptile;
   double *v = (double *) 0;
   NumArray fcst_na, obs_na;

   if(eng.conf_info.match_flag == 0 ||
      eng.n_fcst                    == 0 ||
      eng.n_obs                     == 0) return(0.0);

   //
   // Initialize the maximum interest value for each object to zero.
   //
   for(i=0; i<eng.n_fcst; i++) fcst_na.add(0.0);
   for(i=0; i<eng.n_obs;  i++) obs_na.add(0.0);

   //
   // Loop through all the pairs and keep track of the maximum interest values.
   //
   for(i=0; i<(eng.n_fcst*eng.n_obs); i++) {

      //
      // Skip this interest info if it's not assigned
      //
      if(eng.info_singles[i].fcst_number == 0 ||
         eng.info_singles[i].obs_number  == 0) continue;

      interest = eng.info_singles[i].interest_value;
      fcst_i   = eng.info_singles[i].fcst_number - 1;
      obs_i    = eng.info_singles[i].obs_number - 1;

      if(interest >= fcst_na[fcst_i]) fcst_na.set(fcst_i, interest);
      if(interest >= obs_na[obs_i])   obs_na.set(obs_i,   interest);
   }

   //
   // Allocate memory
   //
   v = new double [eng.n_fcst + eng.n_obs];

   //
   // Add the maximum interest values for the forecast and/or observation
   //
   n_values = 0;
   if(flag == 1 || flag == 3) {
      for(i=0; i<eng.n_fcst; i++) {
         v[n_values] = fcst_na[i];
         n_values++;
      }
   }
   if(flag == 2 || flag == 3) {
      for(i=0; i<eng.n_obs; i++) {
         v[n_values] = obs_na[i];
         n_values++;
      }
   }

   //
   // Sort the maximum interest values
   //
   sort(v, n_values);

   //
   // Get the requested percentile
   //
   ptile = percentile(v, n_values, p/100.0);

   //
   // Done
   //

   if(v) { delete [] v; v = (double *) 0; }

   return(ptile);
}

///////////////////////////////////////////////////////////////////////

void write_engine_stats(ModeFuzzyEngine & eng, const Grid & grid, AsciiTable & at)

{

   int i, j, row;

   //
   // Compute the maximum number of rows possible
   //
   i = 1;                        // Header row
   i += eng.n_fcst;              // Simple forecast objects
   i += eng.n_obs;               // Simple observation objects
   i += eng.n_fcst * eng.n_obs;  // Pairs of simple objects
   i += 3*eng.collection.n_sets; // Clusters: fcst, obs, pairs

   //
   // Setup the AsciiTable to be used
   //
   at.clear();
   j = n_mode_hdr_columns + n_mode_obj_columns;
   at.set_size(i, j);                         // Set table size
   justify_mode_cols(at);                     // Justify columns
   at.set_precision(                          // Set the precision
      eng.conf_info.conf.output_precision());
   at.set_bad_data_value(bad_data_double);    // Set the bad data value
   at.set_bad_data_str(na_str);               // Set the bad data string
   at.set_delete_trailing_blank_rows(1);      // No trailing blank rows

   //
   //  calculate n_valid
   //

int x, y;

eng.n_valid = 0;

for (x=0; x<(grid.nx()); ++x)  {

   for (y=0; y<(grid.ny()); ++y)  {

      if ( eng.fcst_raw->is_bad_data(x, y) || eng.obs_raw->is_bad_data(x, y) )  continue;

      ++(eng.n_valid);

   }

}

   //
   // Initialize row count
   //
   row = 0;

   //
   // Column headers
   //
   write_header_row(eng, at, row);
   row++;

   //
   // Single objects
   //
   for(i=0; i<eng.n_fcst; i++) {
      write_fcst_single(eng, i, grid, at, row);
      row++;
   }

   for(i=0; i<eng.n_obs; i++) {
      write_obs_single(eng, i, grid, at, row);
      row++;
   }

   //
   // If no matching was requested, don't write any more
   //
   if(eng.conf_info.match_flag == 0) return;

   //
   // Object pairs, increment the counter within the subroutine
   //
   for(i=0; i<eng.n_fcst; i++) {
      for(j=0; j<eng.n_obs; j++) {
         write_pair(eng, grid, i, j, at, row);
      }
   }

   //
   // Single composites
   //
   for(i=0; i<(eng.collection.n_sets); i++) {
      write_fcst_cluster(eng, i, grid, at, row);
      row++;
   }

   for(i=0; i<(eng.collection.n_sets); i++) {
      write_obs_cluster(eng, i, grid, at, row);
      row++;
   }

   //
   // Composite Pairs
   //
   for(i=0; i<(eng.collection.n_sets); i++) {
      write_cluster_pair(eng, grid, i, at, row);
      row++;
   }

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////

void write_header_row(ModeFuzzyEngine &eng, AsciiTable &at, const int row)

{

   int i, c;
   ConcatString tmp_str;

   //
   // Write out the MODE header columns
   //
   for(i=0; i<n_mode_hdr_columns; i++) {
      at.set_entry(row, i, mode_hdr_columns[i]);
   }

   //
   // Write out the MODE objects columns
   //
   for(i=0; i<n_mode_obj_columns; i++) {
      at.set_entry(row, i + n_mode_hdr_columns, mode_obj_columns[i]);
   }

   //
   // Over-ride the name of the INTENSITY_USER column
   //
   c = METHdrTable.header(met_version, "MODE", "OBJ")->col_offset("INTENSITY_USER");
   if(nint(eng.conf_info.inten_perc_value) == 101) {
      tmp_str = "INTENSITY_MEAN";
   }
   else if(nint(eng.conf_info.inten_perc_value) == 102) {
      tmp_str = "INTENSITY_SUM";
   }
   else {
     tmp_str.format("INTENSITY_%d",
               nint(eng.conf_info.inten_perc_value));
   }
   at.set_entry(row, c, tmp_str);

   return;
}

///////////////////////////////////////////////////////////////////////

void write_header_columns(ModeFuzzyEngine & eng, const Grid & grid, AsciiTable & at, const int row)

{
   int c = 0;
   ConcatString s;


   // Version
   at.set_entry(row, c++,
                (string) met_version);

   // Model Name
   s = check_hdr_str(conf_key_model,
                     eng.conf_info.model);
   at.set_entry(row, c++, s.text());

     //  N_valid
   at.set_entry(row, c++, eng.n_valid);

     //  grid res
   at.set_entry(row, c++, eng.conf_info.grid_res);


   // Description
   s = check_hdr_str(conf_key_desc, eng.conf_info.desc);
   at.set_entry(row, c++, s.text());

   // Forecast lead time
   at.set_entry(row, c++,
                sec_to_hhmmss(eng.fcst_raw->data.lead()));

   // Forecast valid time
   at.set_entry(row, c++,
                unix_to_yyyymmdd_hhmmss(eng.fcst_raw->data.valid()));

   // Forecast accumulation time
   at.set_entry(row, c++,
                sec_to_hhmmss(eng.fcst_raw->data.accum()));

   // Observation lead time
   at.set_entry(row, c++,
                sec_to_hhmmss(eng.obs_raw->data.lead()));

   // Observation valid time
   at.set_entry(row, c++,
                unix_to_yyyymmdd_hhmmss(eng.obs_raw->data.valid()));

   // Observation accumulation time
   at.set_entry(row, c++,
                sec_to_hhmmss(eng.obs_raw->data.accum()));

   // Forecast convolution radius
   at.set_entry(row, c++,
                eng.conf_info.fcst_conv_radius);

   // Forecast convolution threshold
   at.set_entry(row, c++,
                eng.conf_info.fcst_conv_thresh.get_str());

   // Observation convolution radius
   at.set_entry(row, c++,
                eng.conf_info.obs_conv_radius);

   // Observation convolution threshold
   at.set_entry(row, c++,
                eng.conf_info.obs_conv_thresh.get_str());

   // Forecast Variable Name
   s = check_hdr_str(conf_key_fcst_var,
                     eng.conf_info.fcst_info->name_attr());
   at.set_entry(row, c++, s.text());

   // Forecast Variable Units
   s = check_hdr_str(conf_key_fcst_units,
                     eng.conf_info.fcst_info->units_attr(), true);
   at.set_entry(row, c++, s.text());

   // Forecast Variable Level
   s = check_hdr_str(conf_key_fcst_lev,
                     eng.conf_info.fcst_info->level_attr(), true);
   at.set_entry(row, c++, s.text());

   // Observation Variable Name
   s = check_hdr_str(conf_key_obs_var,
                     eng.conf_info.obs_info->name_attr());
   at.set_entry(row, c++, s.text());

   // Observation Variable Units
   s = check_hdr_str(conf_key_obs_units,
                     eng.conf_info.obs_info->units_attr(), true);
   at.set_entry(row, c++, s.text());

   // Observation Variable Level
   s = check_hdr_str(conf_key_obs_lev,
                     eng.conf_info.obs_info->level_attr(), true);
   at.set_entry(row, c++, s.text());

   // Observation type
   s = check_hdr_str(conf_key_obtype,
                     eng.conf_info.obtype);
   at.set_entry(row, c++, s.text());

   return;
}

///////////////////////////////////////////////////////////////////////

void write_fcst_single(ModeFuzzyEngine &eng, const int n, const Grid &grid,
                       AsciiTable &at, const int row) {
   int i, c;
   double lat, lon;
   ConcatString tmp_str;

   if(n >= eng.n_fcst) {
      mlog << Error << "\nwrite_fcst_single(const ModeFuzzyEngine &, int, "
           << "AsciiTable, int) -> "
           << n << " >= number of fcst, " << eng.n_fcst << "\n\n";
      exit(1);
   }

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("F%03d", (n+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   i = eng.collection.fcst_set_number(n + 1);
   tmp_str.format("CF%03d", (i + 1));
   at.set_entry(row, c++, tmp_str);

   // Convert x,y to lat,lon
   grid.xy_to_latlon(eng.fcst_single[n].centroid_x,
                     eng.fcst_single[n].centroid_y,
                     lat, lon);

   // Object centroid, x-coordinate
   at.set_entry(row, c++, eng.fcst_single[n].centroid_x);

   // Object centroid, y-coordinate
   at.set_entry(row, c++, eng.fcst_single[n].centroid_y);

   // Object centroid, latitude
   at.set_entry(row, c++, lat);

   // Object centroid, longitude
   at.set_entry(row, c++, -1.0*lon);

   // Axis angle
   at.set_entry(row, c++, eng.fcst_single[n].axis_ang);

   // Object length
   at.set_entry(row, c++, eng.fcst_single[n].length);

   // Object width
   at.set_entry(row, c++, eng.fcst_single[n].width);

   // Area of object
   at.set_entry(row, c++,
                nint(eng.fcst_single[n].area));

   // Area in the raw field that meets the threshold criteria
   at.set_entry(row, c++,
                nint(eng.fcst_single[n].area_thresh));

   // Object curvature
   at.set_entry(row, c++, eng.fcst_single[n].curvature);

   // Center of curvature, x-coordinate
   at.set_entry(row, c++, eng.fcst_single[n].curvature_x);

   // Center of curvature, y-coordiante
   at.set_entry(row, c++, eng.fcst_single[n].curvature_y);

   // Object complexity
   at.set_entry(row, c++, eng.fcst_single[n].complexity);

   // 10th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.p10);

   // 25th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.p25);

   // 50th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.p50);

   // 75th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.p75);

   // 90th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.p90);

   // Specified percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.pth);

   // Sum of the object intensity values
   at.set_entry(row, c++,
                eng.fcst_single[n].intensity_ptile.sum);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_pair_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////

void write_obs_single(ModeFuzzyEngine &eng, const int n, const Grid &grid,
                      AsciiTable &at, const int row) {
   int i, c;
   double lat, lon;
   ConcatString tmp_str;

   if(n >= eng.n_obs) {
      mlog << Error << "\nwrite_obs_single(const ModeFuzzyEngine &, int, "
           << "AsciiTable, int) -> "
           << n << " >= number of obs, " << eng.n_obs << "\n\n";
      exit(1);
   }

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("O%03d", (n+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   i = eng.collection.obs_set_number(n + 1);
   tmp_str.format("CO%03d", (i + 1));
   at.set_entry(row, c++, tmp_str);

   // Convert x,y to lat,lon
   grid.xy_to_latlon(eng.obs_single[n].centroid_x,
                     eng.obs_single[n].centroid_y,
                     lat, lon);

   // Object centroid, x-coordinate
   at.set_entry(row, c++, eng.obs_single[n].centroid_x);

   // Object centroid, y-coordinate
   at.set_entry(row, c++, eng.obs_single[n].centroid_y);

   // Object centroid, latitude
   at.set_entry(row, c++, lat);

   // Object centroid, longitude
   at.set_entry(row, c++, -1.0*lon);

   // Axis angle
   at.set_entry(row, c++, eng.obs_single[n].axis_ang);

   // Object length
   at.set_entry(row, c++, eng.obs_single[n].length);

   // Object width
   at.set_entry(row, c++, eng.obs_single[n].width);

   // Area of object
   at.set_entry(row, c++,
                nint(eng.obs_single[n].area));

   // Area in the raw field that meets the threshold criteria
   at.set_entry(row, c++,
                nint(eng.obs_single[n].area_thresh));

   // Object curvature
   at.set_entry(row, c++, eng.obs_single[n].curvature);

   // Center of curvature, x-coordinate
   at.set_entry(row, c++, eng.obs_single[n].curvature_x);

   // Center of curvature, y-coordiante
   at.set_entry(row, c++, eng.obs_single[n].curvature_y);

   // Object complexity
   at.set_entry(row, c++, eng.obs_single[n].complexity);

   // 10th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.p10);

   // 25th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.p25);

   // 50th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.p50);

   // 75th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.p75);

   // 90th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.p90);

   // Specified percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.pth);

   // Sum of the object intensity values
   at.set_entry(row, c++,
                eng.obs_single[n].intensity_ptile.sum);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_pair_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////

void write_pair(ModeFuzzyEngine &eng, const Grid & grid, const int n_f, const int n_o,
                AsciiTable &at, int &row) {
   int n, i, c, fcst_i, obs_i;
   ConcatString tmp_str;

   if(n_f >= eng.n_fcst || n_o >= eng.n_obs) {
      mlog << Error << "\nwrite_pair(const ModeFuzzyEngine &, int, int, "
           << "AsciiTable &, const int) -> "
           << n_f << " >= number of fcst, " << eng.n_fcst << " or "
           << n_o << " >= number of obs, " << eng.n_obs << "\n\n";
      exit(1);
   }

   n = eng.two_to_one(n_f, n_o);

   //
   // Only dump out pair features if the total interest value for the pair
   // is greater than the print interest threshold
   //
   if( eng.info_singles[eng.get_info_index(n)].interest_value <
       eng.conf_info.print_interest_thresh)  return;

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("F%03d_O%03d", (n_f+1), (n_o+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   fcst_i = eng.collection.fcst_set_number(n_f+1);
   obs_i  = eng.collection.obs_set_number(n_o+1);
   tmp_str.format("CF%03d_CO%03d", (fcst_i+1), (obs_i+1));
   at.set_entry(row, c++, tmp_str);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_single_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   // Distance between centroids
   at.set_entry(row, c++, eng.pair_single[n].centroid_dist);

   // Distance between boundaries
   at.set_entry(row, c++, eng.pair_single[n].boundary_dist);

   // Distance between convex hulls
   at.set_entry(row, c++,
                eng.pair_single[n].convex_hull_dist);

   // Difference in angles in degrees
   at.set_entry(row, c++, eng.pair_single[n].angle_diff);

   // Difference in aspect ratio
   at.set_entry(row, c++, eng.pair_single[n].aspect_diff);

   // Fcst/Obs area ratio
   at.set_entry(row, c++, eng.pair_single[n].fo_area_ratio);

   // Intersection area
   at.set_entry(row, c++,
                nint(eng.pair_single[n].intersection_area));

   // Union area
   at.set_entry(row, c++,
                nint(eng.pair_single[n].union_area));

   // Symmetric difference area
   at.set_entry(row, c++,
                nint(eng.pair_single[n].symmetric_diff));

   // Intersection over area
   at.set_entry(row, c++,
                eng.pair_single[n].intersection_over_area);

   // Curvature ratio
   at.set_entry(row, c++,
                eng.pair_single[n].curvature_ratio);

   // Complexity ratio
   at.set_entry(row, c++,
                eng.pair_single[n].complexity_ratio);

   // Percentile intensity ratio
   at.set_entry(row, c++,
                eng.pair_single[n].percentile_intensity_ratio);

   // Total interest value
   at.set_entry(row, c++,
                eng.info_singles[eng.get_info_index(n)].interest_value);

   //
   // Increment row counter
   //
   row++;

   return;
}

///////////////////////////////////////////////////////////////////////

void write_fcst_cluster(ModeFuzzyEngine &eng, const int n, const Grid &grid,
                          AsciiTable &at, const int row) {
   int i, c;
   double lat, lon;
   ConcatString tmp_str;

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("CF%03d", (n+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   at.set_entry(row, c++, tmp_str);

   // Convert x,y to lat,lon
   grid.xy_to_latlon(eng.fcst_cluster[n].centroid_x,
                     eng.fcst_cluster[n].centroid_y,
                     lat, lon);

   // Object centroid, x-coordinate
   at.set_entry(row, c++, eng.fcst_cluster[n].centroid_x);

   // Object centroid, y-coordinate
   at.set_entry(row, c++, eng.fcst_cluster[n].centroid_y);

   // Object centroid, latitude
   at.set_entry(row, c++, lat);

   // Object centroid, longitude
   at.set_entry(row, c++, -1.0*lon);

   // Axis angle
   at.set_entry(row, c++, eng.fcst_cluster[n].axis_ang);

   // Object length
   at.set_entry(row, c++, eng.fcst_cluster[n].length);

   // Object width
   at.set_entry(row, c++, eng.fcst_cluster[n].width);

   // Area of object
   at.set_entry(row, c++,
                nint(eng.fcst_cluster[n].area));

   // Area in the raw field that meets the threshold criteria
   at.set_entry(row, c++,
                nint(eng.fcst_cluster[n].area_thresh));

   // Object curvature
   at.set_entry(row, c++, eng.fcst_cluster[n].curvature);

   // Center of curvature, x-coordinate
   at.set_entry(row, c++, eng.fcst_cluster[n].curvature_x);

   // Center of curvature, y-coordiante
   at.set_entry(row, c++, eng.fcst_cluster[n].curvature_y);

   // Object complexity
   at.set_entry(row, c++, eng.fcst_cluster[n].complexity);

   // 10th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.p10);

   // 25th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.p25);

   // 50th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.p50);

   // 75th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.p75);

   // 90th percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.p90);

   // Specified percentile of object intensity
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.pth);

   // Sum of the object intensity values
   at.set_entry(row, c++,
                eng.fcst_cluster[n].intensity_ptile.sum);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_pair_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////

void write_obs_cluster(ModeFuzzyEngine &eng, const int n, const Grid &grid,
                         AsciiTable &at, const int row) {
   int i, c;
   double lat, lon;
   ConcatString tmp_str;

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("CO%03d", (n+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   at.set_entry(row, c++, tmp_str);

   // Convert x,y to lat,lon
   grid.xy_to_latlon(eng.obs_cluster[n].centroid_x,
                     eng.obs_cluster[n].centroid_y,
                     lat, lon);

   // Object centroid, x-coordinate
   at.set_entry(row, c++, eng.obs_cluster[n].centroid_x);

   // Object centroid, y-coordinate
   at.set_entry(row, c++, eng.obs_cluster[n].centroid_y);

   // Object centroid, latitude
   at.set_entry(row, c++, lat);

   // Object centroid, longitude
   at.set_entry(row, c++, -1.0*lon);

   // Axis angle
   at.set_entry(row, c++, eng.obs_cluster[n].axis_ang);

   // Object length
   at.set_entry(row, c++, eng.obs_cluster[n].length);

   // Object width
   at.set_entry(row, c++, eng.obs_cluster[n].width);

   // Area of object
   at.set_entry(row, c++,
                nint(eng.obs_cluster[n].area));

   // Area in the raw field that meets the threshold criteria
   at.set_entry(row, c++,
                nint(eng.obs_cluster[n].area_thresh));

   // Object curvature
   at.set_entry(row, c++, eng.obs_cluster[n].curvature);

   // Center of curvature, x-coordinate
   at.set_entry(row, c++, eng.obs_cluster[n].curvature_x);

   // Center of curvature, y-coordiante
   at.set_entry(row, c++, eng.obs_cluster[n].curvature_y);

   // Object complexity
   at.set_entry(row, c++, eng.obs_cluster[n].complexity);

   // 10th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.p10);

   // 25th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.p25);

   // 50th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.p50);

   // 75th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.p75);

   // 90th percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.p90);

   // Specified percentile of object intensity
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.pth);

   // Sum of the object intensity values
   at.set_entry(row, c++,
                eng.obs_cluster[n].intensity_ptile.sum);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_pair_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   //
   // Done
   //

   return;
}

///////////////////////////////////////////////////////////////////////

void write_cluster_pair(ModeFuzzyEngine &eng, const Grid & grid, const int n,
                          AsciiTable &at, const int row) {
   int i, c;
   ConcatString tmp_str;

   // Write out the common header columns
   write_header_columns(eng, grid, at, row);

   c = n_mode_hdr_columns;

   // Object ID
   tmp_str.format("CF%03d_CO%03d", (n+1), (n+1));
   at.set_entry(row, c++, tmp_str);

   // Object category
   at.set_entry(row, c++, tmp_str);

   //
   // Fill the columns that don't apply with bad data values
   //
   for(i=0; i<n_mode_single_columns; i++) {
      at.set_entry(row, c++, bad_data_double);
   }

   // Distance between centroids
   at.set_entry(row, c++,
                eng.pair_cluster[n].centroid_dist);

   // Distance between boundaries
   at.set_entry(row, c++,
                eng.pair_cluster[n].boundary_dist);

   // Distance between convex hulls
   at.set_entry(row, c++,
                eng.pair_cluster[n].convex_hull_dist);

   // Difference in angles in degrees
   at.set_entry(row, c++, eng.pair_cluster[n].angle_diff);

   // Difference in aspect ratio
   at.set_entry(row, c++, eng.pair_cluster[n].aspect_diff);

   // Fcst/Obs area ratio
   at.set_entry(row, c++, eng.pair_cluster[n].fo_area_ratio);

   // Intersection area
   at.set_entry(row, c++,
                nint(eng.pair_cluster[n].intersection_area));

   // Union area
   at.set_entry(row, c++,
                nint(eng.pair_cluster[n].union_area));

   // Symmetric difference area
   at.set_entry(row, c++,
                nint(eng.pair_cluster[n].symmetric_diff));

   // Intersection over area
   at.set_entry(row, c++,
                eng.pair_cluster[n].intersection_over_area);

   // Curvature ratio
   at.set_entry(row, c++,
                eng.pair_cluster[n].curvature_ratio);

   // Complexity ratio
   at.set_entry(row, c++,
                eng.pair_cluster[n].complexity_ratio);

   // Percentile intensity ratio
   at.set_entry(row, c++,
                eng.pair_cluster[n].percentile_intensity_ratio);

   // Total interest value
   at.set_entry(row, c++,
                eng.info_clus[n].interest_value);

   //
   // Done
   //

   return;
}

////////////////////////////////////////////////////////////////////////

void justify_mode_cols(AsciiTable &at) {

   justify_met_at(at, n_mode_hdr_columns);

   return;
}

///////////////////////////////////////////////////////////////////////

void calc_fcst_clus_ch_mask(const ModeFuzzyEngine &eng, ShapeData &mask) {
   int i, x, y;
   ShapeData comp;
   Polyline poly;
   Box bb;

   if(eng.need_fcst_clus_split) {
      mlog << Error << "\ncalc_fcst_clus_ch_mask -> "
           << "should not be called with need_fcst_clus_split set to true\n\n";

      exit(1);
   }

   //
   // Initialize convex hull mask wrfdata object
   //
   mask = *(eng.fcst_clus_split);
   mask.zero_field();

   //
   // Calculate fcst convex hull mask field for composite objects
   //
   for(i=0; i<eng.collection.n_sets; i++) {
      comp = select(*eng.fcst_clus_split, i+1);
      poly = comp.convex_hull();
      poly.bounding_box(bb);

      for(x = (int) floor(bb.left()-1); x <= (int) ceil(bb.right()+1); x++) {
         for(y = (int) floor(bb.bottom()-1); y <= (int) ceil(bb.top()+1); y++) {

            if(poly.is_inside(x, y)) {
               mask.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   } // end for i

   //
   // Calculate fcst convex hull mask field for unmatched singles
   //
   for(i=0; i<eng.n_fcst; i++) {

      if(eng.collection.fcst_set_number(i+1) != -1) continue;

      comp = select(*eng.fcst_split, i+1);
      poly = comp.convex_hull();
      poly.bounding_box(bb);

      for(x = (int) floor(bb.left()-1); x <= (int) ceil(bb.right()+1); x++) {
         for(y = (int) floor(bb.bottom()-1); y <= (int) ceil(bb.top()+1); y++) {

            if(poly.is_inside(x, y)) {
               mask.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   }

   return;
}

///////////////////////////////////////////////////////////////////////

void calc_obs_clus_ch_mask(const ModeFuzzyEngine &eng, ShapeData &mask) {
   int i, x, y;
   ShapeData comp;
   Polyline poly;
   Box bb;

   if(eng.need_obs_clus_split) {
      mlog << Error << "\ncalc_obs_clus_ch_mask -> "
           << "should not be called with need_obs_clus_split set to true\n\n";
      exit(1);
   }

   //
   // Initialize convex hull mask wrfdata object
   //
   mask = *(eng.obs_clus_split);
   mask.zero_field();

   //
   // Calculate obs convex hull mask field for composite objects
   //
   for(i=0; i<eng.collection.n_sets; i++) {
      comp = select(*eng.obs_clus_split, i+1);
      poly = comp.convex_hull();
      poly.bounding_box(bb);

      for(x=(int) floor(bb.left()-1); x<=ceil(bb.right()+1); x++) {
         for(y=(int) floor(bb.bottom()-1); y<=ceil(bb.top()+1); y++) {

            if(poly.is_inside(x, y)) {
               mask.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   } // end for i

   //
   // Calculate obs convex hull mask field for unmatched singles
   //
   for(i=0; i<eng.n_obs; i++) {

      if(eng.collection.obs_set_number(i+1) != -1) continue;

      comp = select(*eng.obs_split, i+1);
      poly = comp.convex_hull();
      poly.bounding_box(bb);

      for(x=(int) floor(bb.left()-1); x<=ceil(bb.right()+1); x++) {
         for(y=(int) floor(bb.bottom()-1); y<=ceil(bb.top()+1); y++) {

            if(poly.is_inside(x, y)) {
               mask.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   }

   return;
}

///////////////////////////////////////////////////////////////////////

void calc_fcst_cluster_mask(const ModeFuzzyEngine &eng, ShapeData &comp, const int n_set) {
   int i, x, y;
   ShapeData junk;

   comp.data.set_size(eng.fcst_raw->data.nx(), eng.fcst_raw->data.ny());

   for(i=0; i<eng.n_fcst; i++) {

      if(!(eng.collection.set[n_set].has_fcst(i + 1)))  continue;

      junk = select(*(eng.fcst_split), i + 1);

      for(x=0; x<(eng.fcst_raw->data.nx()); x++) {
         for(y=0; y<(eng.fcst_raw->data.ny()); y++) {

            if( junk.is_nonzero(x, y) ) {
               comp.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   } // end for i

   comp.calc_moments();

   return;
}

///////////////////////////////////////////////////////////////////////

void calc_obs_cluster_mask(const ModeFuzzyEngine &eng, ShapeData &comp, const int n_set) {
   int i, x, y;
   ShapeData junk;

   comp.data.set_size(eng.obs_raw->data.nx(), eng.obs_raw->data.ny());

   for(i=0; i<eng.n_obs; i++) {

      if(!(eng.collection.set[n_set].has_obs(i + 1)))  continue;

      junk = select(*(eng.obs_split), i + 1);

      for(x=0; x<(eng.obs_raw->data.nx()); x++) {
         for(y=0; y<(eng.obs_raw->data.ny()); y++) {

            if( junk.is_nonzero(x, y) ) {
               comp.data.set(1.0, x, y);
            }
         } // end for y
      } // end for x
   } // end for i

   comp.calc_moments();

   return;
}

///////////////////////////////////////////////////////////////////////


double aspect_ratio_conf(double t)

{

const double tm1   = t - 1.0;
const double ratio = (tm1*tm1)/(t*t + 1.0);

return( pow(ratio, 0.3) );

}


///////////////////////////////////////////////////////////////////////