UserScript: Calculate the Difficulty Index

model_applications/medium_range/ UserScript_fcstGEFS _Difficulty_Index.conf

Scientific Objective

This use case calls the UserScript wrapper to run a user provided script that calculates the difficulty index for windspeed. This use case allows for the user to change a variety of variables needed to run the difficulty index (i.e. threshold start and units) so that user can run the script at different thresholds without needing to alter the code. This script run by the use case uses METcalcpy to provide the difficulty index calculation and METplotpy to provide the plotting capability.

The difficulty index was developed by the Naval Research Lab (NRL). The overall aim of the difficulty index is to graphically represent the expected difficulty of a decision based on a set of forecasts (ensemble) of, e.g., significant wave height as a function of space and time. There are two basic factors that can make a decision difficult. The first factor is the proximity of the ensemble mean forecast to a decision threshold, e.g. 12 ft seas. If the ensemble mean is either much lower or much higher than the threshold, the decision is easier; if it is closer to the threshold, the decision is harder. The second factor is the forecast precision, or ensemble spread. The greater the spread around the ensemble mean, the more likely it is that there will be ensemble members both above and below the decision threshold, making the decision harder. (A third factor that we will not address here is undiagnosed systematic error, which adds uncertainty in a similar way to ensemble spread.) The challenge is combining these factors into a continuous function that allows the user to assess relative risk.


This use case calculates the difficulty index for windspeed using NCEP GEFS ensemble data. The data is composed of 30 ensemble members that have been compiled and compressed into one .npz file.

  • Variables required to calculate the difficulty index: Levels required: 10-m #. v- component of wind #. u- component of wind #. Windspeed #. Latitude #. Longitude

  • Forecast dataset: NCEP GEFS 30 member Ensemble - Initialization date: 20191208 - Initialization hours: 12 UTC - Lead times: 60 - Format: Grib2 - Resolution: 0.5 degree

METplus Components

This use case runs the UserScript wrapper tool to run a user provided script, in this case,

METplus Workflow

This use case loops by process which means that each tool is run for all times before moving to the next tool. The tool order is as follows:


This example loops by initialization time (with begin, end, and increment as specified in the METplus UserScript_fcstGEFS_Difficulty_Index.conf file).

1 initialization time will be run over 1 lead time:

Init: 20201208_12Z
Forecast lead: 60

METplus Configuration

METplus first loads all of the configuration files found in parm/metplus_config, then it loads any configuration files passed to METplus via the command line with the -c option, i.e. -c parm/use_cases/model_applications/medium_range/UserScript_fcstGEFS_Difficulty_Index.conf


# List of applications to run

# time looping - options are INIT, VALID, RETRO, and REALTIME
# If set to INIT or RETRO:
# If set to VALID or REALTIME:

# Format of INIT_BEG and INIT_END using % items
# %Y = 4 digit year, %m = 2 digit month, %d = 2 digit day, etc.
# see for more information
# %Y%m%d%H expands to YYYYMMDDHH

# Start time for METplus run - must match VALID_TIME_FMT
INIT_BEG = 2020120812

# End time for METplus run - must match VALID_TIME_FMT
INIT_END = 2020120812

# Increment between METplus runs (in seconds if no units are specified)
#  Must be >= 60 seconds

# List of forecast leads to process for each run time (init or valid)
# In hours if units are not specified
# If unset, defaults to 0 (don't loop through forecast leads)

# Order of loops to process data - Options are times, processes
# Not relevant if only one item is in the PROCESS_LIST
# times = run all wrappers in the PROCESS_LIST for a single run time, then
#   increment the run time and run all wrappers again until all times have
#   been evaluated.
# processes = run the first wrapper in the PROCESS_LIST for all times
#   specified, then repeat for the next item in the PROCESS_LIST until all
#   wrappers have been run
LOOP_ORDER = processes

# list of strings to loop over for each run time.
# value for each item can be referenced in filename templates with {custom?fmt=%s}


USER_SCRIPT_INPUT_DIR = {INPUT_BASE}/model_applications/medium_range/diff_index

USER_SCRIPT_INPUT_TEMPLATE = {USER_SCRIPT_INPUT_DIR}/wndspd_GEFS_NorthPac_5dy_30mem_{init?fmt=%Y%m%d%H}.npz

USER_SCRIPT_OUTPUT_DIR = {OUTPUT_BASE}/model_applications/medium_range/diff_index

USER_SCRIPT_COMMAND = {PARM_BASE}/use_cases/model_applications/medium_range/UserScript_fcstGEFS_Difficulty_Index/


# Difficulty index specific variables










DIFF_INDEX_FIG_BASENAME = {USER_SCRIPT_OUTPUT_DIR}/wndspd_GEFS_NorthPac_5dy_30mem_difficulty_index

MET Configuration

There are no MET tools used in this use case.

Python Embedding

This use case uses a Python embedding script to read input data


#!/usr/bin/env python3

Load fieldijn from npz file created with
helper function, compute ensemble mean and spread, compute
difficulty index for a set of thresholds, plot and save the results.
Author: Bill Campbell, NRL and Lindsay Blank, NCAR

Taken from original but replacing with METcalcpy and METplotpy.

import os
import sys
import numpy as np
import matplotlib.pyplot as plt
from metcalcpy.calc_difficulty_index import forecast_difficulty as di
from metcalcpy.calc_difficulty_index import EPS
from metcalcpy.piecewise_linear import PiecewiseLinear as plin
import metplotpy.plots.difficulty_index.mycolormaps as mcmap
from metplotpy.plots.difficulty_index.plot_difficulty_index import plot_field

def load_data(filename):
    """Load ensemble data from file"""
    loaded = np.load(filename)
    lats, lons = (loaded['lats'], loaded['lons'])
    fieldijn =

    return lats, lons, fieldijn

def compute_stats(field):
    """Compute mean and std dev"""
    mu = np.mean(field, axis=-1)
    sigma = np.std(field, axis=-1, ddof=1)

    return mu, sigma

def compute_wind_envelope():
    Computes piecewise linear envelope for winds in knots.

    Piecewise linear object

    # Envelope for version 6.1, the default
    xunits = 'kn'
    A6_1_name = "A6_1"
    A6_1_left = 0.0
    A6_1_right = 0.0
    A6_1_xlist = [5.0, 28.0, 34.0, 50.0]
    A6_1_ylist = [0.0, 1.5, 1.5, 0.0]
    Aplin =\
            plin(A6_1_xlist, A6_1_ylist, xunits=xunits,
                    right=A6_1_right, left=A6_1_left, name=A6_1_name)

    return Aplin

def compute_difficulty_index(field, mu, sigma, thresholds, Aplin):
    Compute difficulty index for an ensemble forecast given
    a set of thresholds, returning a dictionary of fields.
    dij = {}
    for threshold in thresholds:
        dij[threshold] =\
            di(sigma, mu, threshold, field, Aplin=Aplin, sigma_over_mu_ref=EPS)

    return dij

def plot_difficulty_index(dij, lats, lons, thresholds, units):
    Plot the difficulty index for a set of thresholds,
    returning a dictionary of figures
    myparams = {'figure.figsize': (8, 5),
                'figure.max_open_warning': 40}
    figs = {}
    cmap = mcmap.stoplight()
    for threshold in thresholds:
        if np.max(dij[threshold]) <= 1.0:
            vmax = 1.0
            vmax = 1.5
        figs[threshold] =\
                          lats, lons, vmin=0.0, vmax=vmax, cmap=cmap,
                          xlab='Longitude \u00b0E', ylab='Latitude',
                          clab='thresh={} {}'.format(threshold, units),
                          title='Forecast Decision Difficulty Index')

    return figs

def save_difficulty_figures(figs, save_thresh, units):
    Save subset of difficulty index figures.
    fig_fmt = os.environ.get('DIFF_INDEX_FIG_FMT')
    fig_basename = os.environ.get('DIFF_INDEX_FIG_BASENAME')

    # create output directory if it does not already exist
    output_dir = os.path.dirname(fig_basename)
    if not os.path.exists(output_dir):

    for thresh in save_thresh:
        thresh_str = '{:.2f}'.format(thresh).replace('.', '_')
        fig_name = (fig_basename + thresh_str +
                    '_' + units + '.' + fig_fmt)
        print('Saving {}...\n'.format(fig_name))
        figs[thresh].savefig(fig_name, format=fig_fmt)

def plot_statistics(mu, sigma, lats, lons, units='feet'):
    """Plot ensemble mean and spread, returning figure handles"""
    cmap = mcmap.spectral()
    mu_fig =\
        plot_field(mu, lats, lons, cmap=cmap, clab=units,
                      vmin=0.0, vmax=np.nanmax(mu),
                      xlab='Longitude \u00b0E',
                      title='Forecast Ensemble Mean')
    sigma_fig =\
        plot_field(sigma, lats, lons, cmap=cmap, clab=units,
                      vmin=0.0, vmax=np.nanmax(sigma),
                      xlab='Longitude \u00b0E',
                      title='Forecast Ensemble Std')

    return mu_fig, sigma_fig

def save_stats_figures(mu_fig, sigma_fig):
    Save ensemble mean and spread figures.

    fig_fmt = os.environ.get('DIFF_INDEX_FIG_FMT')
    fig_basename = os.environ.get('DIFF_INDEX_FIG_BASENAME')
    mu_name = fig_basename + 'mean.' + fig_fmt
    print('Saving {}...\n'.format(mu_name))
    mu_fig.savefig(mu_name, format=fig_fmt)
    sigma_name = fig_basename + 'std.' + fig_fmt
    print('Saving {}...\n'.format(sigma_name))
    sigma_fig.savefig(sigma_name, format=fig_fmt)

def main():
    Load fieldijn from npz file created with
    helper function, compute ensemble mean and spread, compute
    difficulty index for a set of thresholds, plot and save the results.

    filename = os.environ.get('DIFF_INDEX_INPUT_FILENAME')
    lats, lons, fieldijn = load_data(filename)
    # Convert m/s to knots
    units = os.environ.get('DIFF_INDEX_UNITS')
    mps2kn = 1.94384
    fieldijn = mps2kn * fieldijn
    # Ensemble mean, std dev
    muij, sigmaij = compute_stats(fieldijn)
    # Windspeed envelope
    Aplin = compute_wind_envelope()
    # Difficulty index for a set of thresholds
    #thresholds = np.arange(os.environ.get('DIFF_INDEX_THRESH_START'), os.environ.get('DIFF_INDEX_THRESH_END'), os.environ.get('DIFF_INDEX_THRESH_STEP'))
    start = float(os.environ.get('DIFF_INDEX_THRESH_START'))
    stop = float(os.environ.get('DIFF_INDEX_THRESH_END'))
    step = float(os.environ.get('DIFF_INDEX_THRESH_STEP'))
    thresholds = np.arange(start, stop, step)
    dij = compute_difficulty_index(fieldijn, muij, sigmaij, thresholds, Aplin=Aplin)
    # Plot and save difficulty index figures
    figs = plot_difficulty_index(dij, lats, lons, thresholds, units)
    save_start = float(os.environ.get('DIFF_INDEX_SAVE_THRESH_START'))
    save_stop = float(os.environ.get('DIFF_INDEX_SAVE_THRESH_STOP'))
    save_step = float(os.environ.get('DIFF_INDEX_SAVE_THRESH_STEP'))
    save_thresh = np.arange(save_start, save_stop, save_step)
    save_difficulty_figures(figs, save_thresh, units)
    # Plot and save ensemble mean, std_dev
    mu_fig, sigma_fig =\
        plot_statistics(muij, sigmaij, lats, lons, units=units)
    save_stats_figures(mu_fig, sigma_fig)

if __name__ == '__main__':

Running METplus

This use case can be run two ways:

1) Passing in UserScript_fcstGEFS_Difficulty_Index.conf, then a user-specific system configuration file: \
-c /path/to/METplus/parm/use_cases/model_applications/medium_range/UserScript_fcstGEFS_Difficulty_Index.conf \
-c /path/to/user_system.conf
  1. Modifying the configurations in parm/metplus_config, then passing in UserScript_fcstGEFS_Difficulty_Index.conf: \
    -c /path/to/METplus/parm/use_cases/model_applications/medium_range/UserScript_fcstGEFS_Difficulty_Index.conf

The former method is recommended. Whether you add them to a user-specific configuration file or modify the metplus_config files, the following variables must be set correctly:

  • INPUT_BASE - Path to directory where sample data tarballs are unpacked (See Datasets section to obtain tarballs). This is not required to run METplus, but it is required to run the examples in parm/use_cases

  • OUTPUT_BASE - Path where METplus output will be written. This must be in a location where you have write permissions

  • MET_INSTALL_DIR - Path to location where MET is installed locally

and for the [exe] section, you will need to define the location of NON-MET executables. If the executable is in the user’s path, METplus will find it from the name. If the executable is not in the path, specify the full path to the executable here (i.e. RM = /bin/rm) The following executables are required for performing series analysis use cases:

Example User Configuration File:

INPUT_BASE = /path/to/sample/input/data
OUTPUT_BASE = /path/to/output/dir
MET_INSTALL_DIR = /path/to/met-X.Y

RM = /path/to/rm
CUT = /path/to/cut
TR = /path/to/tr
NCAP2 = /path/to/ncap2
CONVERT = /path/to/convert
NCDUMP = /path/to/ncdump

Expected Output

A successful run will output the following both to the screen and to the logfile:

INFO: METplus has successfully finished running.

Refer to the value set for OUTPUT_BASE to find where the output data was generated. Output for this use case will be found in a directory relative to OUTPUT_BASE. There should be a list of files that have the following format:


Where THRESH isa number between DIFF_INDEX_SAVE_THRESH_START and DIFF_INDEX_SAVE_THRESH_STOP which are defined in UserScript_fcstGEFS_Difficulty_Index.conf.



  • UserScriptUseCase

  • MediumRangeAppUseCase

  • NRLOrgUseCase

Navigate to the METplus Quick Search for Use Cases page to discover other similar use cases.

sphinx_gallery_thumbnail_path = ‘_static/medium_range-UserScript_fcstGEFS_Difficulty_Index.png’

Total running time of the script: ( 0 minutes 0.000 seconds)

Gallery generated by Sphinx-Gallery