Apps

METexpress is a web tool for exploring how well numerical weather forecasts verify against observations. Rather than one sprawling interface, it ships a small family of apps—each pre-configured for a particular verification domain (upper air, surface, precipitation, ensembles, cyclones, and more). You pick the app that matches your question, use guided menus to choose a region, variable, level, and statistic, and METexpress assembles the plot for you. This page walks through every app and what it is for.

What an “app” is

An app in METexpress is a focused front end for a specific kind of forecast verification. Every app shares the same general interface design—a stack of guided menus on the left, a plot on the right—but each one exposes the choices specific to that app. A precipitation app, for example, offers precipitation thresholds and grid-scale controls; an ensemble app offers reliability and ROC plots that simply do not apply to a deterministic surface forecast.

The apps differ mainly along three axes:

Verification domain — What part of the atmosphere or forecast you are scoring—upper-air pressure levels, the near-surface, precipitation, ensemble spread, cyclone tracks, or convective objects.

Statistics offered — Continuous scores like RMSE, ME (mean error / bias), and correlation; contingency-table scores like CSI, FAR, POD, FBIAS, GSS, and HSS; and domain-specific metrics like ACC (anomaly correlation), CRPS, or track error.

Plot types — The shapes of analysis each app supports—time series, vertical profiles, lead-time die-off curves, threshold sweeps, reliability and ROC diagrams, and more.

Why split it into apps?

Verification questions rarely overlap. Someone scoring 500 hPa height forecasts has almost nothing in common with someone scoring a hurricane track. By specializing the menus, each app keeps only the controls and plot types that make sense for its domain—so you are not wading through irrelevant options.

Acronym glossary

The statistic acronyms used throughout this page expand as follows:

RMSE · MSE · MAE · ME — Root-mean-square error, mean-square error, mean absolute error, and mean error (bias) — continuous (scalar) error measures.

CSI · FAR · POD (PODy/PODn) · POFD — Critical Success Index, False Alarm Ratio, Probability of Detection (of yes / of no events), and Probability of False Detection — contingency-table (yes/no event) scores.

FBIAS · GSS · HSS — Frequency Bias, Gilbert Skill Score (equitable threat score), and Heidke Skill Score — more contingency-table skill scores.

FSS — Fractions Skill Score — a neighborhood (spatial-scale) verification score.

ACC — Anomaly Correlation Coefficient — how well a forecast captures departures from climatology.

CRPS · CRPSS · BS · ROC AUC — Continuous Ranked Probability Score (and its Skill Score), Brier Score, and the area under the ROC curve — probabilistic / ensemble verification measures.

How an app builds a plot

Whichever app you open, the path from question to plot is the same. The menus on the page have choices derived from the data—METexpress reads the underlying verification database to populate them, so you only ever see options that actually exist for your chosen data source. You narrow the question with each menu, and the app turns your selections into a query that returns the numbers behind the curve.

Figure 1. The app is the entry point. Your menu choices—data source, region, variable, level, threshold, and the statistic—compose a pre-defined query against the verification database, whose result the app draws as a plot.

The exact menus vary by app. Most apps expose Group, Database, Data-Source, Region, Statistic, Variable, Forecast lead time, Level, Dates, and Curve-dates; precipitation- and threshold-driven apps add a Threshold selector and controls like Interp-Method and Scale; the cyclone app swaps in Basin, Year, Storm, Truth, Storm classification, and Radius.

The eight apps

METexpress provides eight apps. Each section below names the verification question it answers, the variables and statistics it tends to surface, and the plot types it offers.

Upper Air

Plots scalar and contingency-table statistics at different pressure levels in the atmosphere—the classic upper-air verification view. Handles wind, temperature, and other scalar parameters through the vertical column.

Statistics include RMSE, bias-corrected RMSE, MSE, ME, correlation, and wind-vector metrics.

Plots: Time Series, Profile, Dieoff, ValidTime, Histogram, Contour

Anomaly Correlation

Plots anomaly correlations at different pressure levels in the atmosphere and at different heights above the ground. This is the standard measure of how well a model captures large-scale departures from climatology.

Statistics: ACC (anomaly correlation coefficient) and Vector ACC.

Plots: Time Series, Profile, Dieoff, ValidTime, Histogram, Contour

Surface

Plots scalar and contingency-table statistics at different heights above the ground—the near-surface companion to the Upper Air app. Used for surface wind, temperature, moisture, and related scalar metrics.

Shares the continuous and contingency-table statistic set with Upper Air.

Plots: Time Series, Dieoff, ValidTime, Histogram, Contour

Air Quality

Like the Surface app—scalar and contingency-table statistics at heights above the ground—but focused on variables related to air quality, such as pollutant concentrations.

Statistics include CSI, FAR, FBIAS, GSS, HSS, POD variants, POFD, RMSE, and correlation.

Plots: Time Series, Dieoff, Threshold, ValidTime, Histogram, Contour

Ensemble

Plots scalar and contingency-table statistics plus ensemble-specific metrics for multi-member ensemble model runs. The only app with probabilistic verification plots.

Statistics: RMSE, spread, ME, CRPS, CRPSS, MAE, Brier Score variants, ROC AUC, and FSS.

Plots: Time Series, Dieoff, ValidTime, Histogram, Ensemble Histogram, Reliability, ROC, Performance Diagram

Precipitation

Plots scalar and contingency-table statistics for variables relating to precipitation, where forecast skill depends strongly on accumulation thresholds and grid scale.

Statistics include CSI, FAR, FBIAS, GSS, HSS, POD variants, POFD, FSS, RMSE, and correlation.

Plots: Time Series, Dieoff, Threshold, ValidTime, GridScale, Histogram, Contour

Cyclone

Plots track and intensity verification statistics for both tropical and extratropical cyclones. Where the other apps work in fixed regions, this one works per basin, year, and storm.

Statistics include track error (along- and cross-track), MSLP (mean sea-level pressure) error, wind-speed error, and rapid-intensification metrics.

Plots: Time Series, Dieoff, ValidTime, YearToYear, Histogram

Objects

Plots skill scores and model–obs paired verification statistics for convective objects—the object-based view of how well a model places, shapes, and sizes convection.

Statistics include centroid distance, object threat/skill scores, CSI, FAR, PODy, frequency bias, and area-weighted ratios.

Plots: Time Series, Dieoff, Threshold, ValidTime

Ecosystem link

METexpress visualizes verification statistics that originate upstream in the MET verification tools and the METplus wrappers. The continuous and contingency-table scores you select here—RMSE, CSI, FBIAS, and the rest—are the same line types MET produces; METexpress reads them from a database and turns them into interactive plots.

App reference table

A side-by-side view of every app: the verification domain it targets, the kind of variables it handles, and the plot types it offers.

App	Domain / question	Typical variables	Typical plots
Upper Air	Scalar & contingency-table skill at pressure levels in the atmosphere	Wind, temperature, scalar parameters through the column	Time Series, Profile, Dieoff, ValidTime, Histogram, Contour
Anomaly Correlation	Large-scale pattern skill (anomaly vs. climatology) at pressure levels & heights	Height/geopotential and other anomaly fields	Time Series, Profile, Dieoff, ValidTime, Histogram, Contour
Surface	Scalar & contingency-table skill at heights above the ground	Surface wind, temperature, moisture	Time Series, Dieoff, ValidTime, Histogram, Contour
Air Quality	Near-surface skill focused on air-quality variables	Pollutant concentrations & related metrics	Time Series, Dieoff, Threshold, ValidTime, Histogram, Contour
Ensemble	Probabilistic & ensemble skill for multi-member runs	Ensemble fields with spread, CRPS, Brier Score, ROC AUC, FSS	Time Series, Dieoff, ValidTime, Histogram, Ensemble Histogram, Reliability, ROC, Performance Diagram
Precipitation	Scalar & contingency-table skill for precipitation variables	Accumulation amounts & intensity thresholds	Time Series, Dieoff, Threshold, ValidTime, GridScale, Histogram, Contour
Cyclone	Track & intensity skill for tropical and extratropical cyclones	Track position, MSLP, wind speed, rapid-intensification metrics	Time Series, Dieoff, ValidTime, YearToYear, Histogram
Objects	Object-based skill for convective objects (model–obs pairs)	Centroid distance, area & shape ratios, intensity ratios	Time Series, Dieoff, Threshold, ValidTime

The eight METexpress apps. Statistic and plot names are taken verbatim from the source.

Plot types, by axis

Apps share a vocabulary of plot types. The fastest way to read a METexpress plot is to know what sits on each axis. The table below gives the axes each plot type uses. (METexpress’s documentation names the plot types; the x/y assignments here follow standard verification convention — e.g. a profile puts the value on x and pressure on y.)

Plot type	x-axis	y-axis	Reads as…
Time Series	Date	Mean value of the selected parameter for that date	How skill evolves over the verification period
Profile	Mean value of the selected parameter	Pressure level	How skill varies through the vertical column
Dieoff	Forecast lead time	Selected statistic	How skill (or error) changes with increasing lead time
ValidTime	Valid UTC hour	Mean value of the selected parameter	Time-of-day (diurnal) variation in skill
Threshold	Threshold	Mean value of the selected parameter	How skill depends on the event threshold
GridScale	Grid scale	Mean value of the selected parameter	How skill depends on spatial scale (precipitation)
YearToYear	Individual years	Mean value of the selected statistic per year	Year-over-year skill (cyclone seasons)
Histogram	Statistical value bins	Number or frequency counts	The distribution of a statistic
Contour	User-chosen x-axis-parameter (e.g. forecast hour)	User-chosen y-axis-parameter (e.g. height)	A 2-D field of a statistic across two dimensions

Ensemble-only plots

The Ensemble app adds probabilistic diagrams that the deterministic apps do not have. These are the standard tools for judging whether an ensemble’s probabilities are trustworthy.

Reliability — Forecast Probability on the x-axis, Observed Relative Frequency on the y-axis. A perfect forecast falls on the diagonal—events happen as often as predicted.

ROC — False Alarm Rate on the x-axis, Probability of Detection on the y-axis. Shows the trade-off between catching events and raising false alarms.

Performance Diagram — Success Ratio (1−FAR) on the x-axis, Probability of Detection on the y-axis—a compact view that combines several contingency-table scores at once.

Ensemble Histogram — Available as a Rank Histogram, a Probability Integral Transform Histogram, or a Relative Position Histogram—different ways to check whether the spread of the ensemble matches reality.

Match the plot to the app

Not every plot type appears in every app. Threshold plots show up in the threshold-driven apps (Air Quality, Precipitation, Objects); GridScale is specific to Precipitation; YearToYear is specific to Cyclone; and the Reliability / ROC / Performance Diagram / Ensemble Histogram family is specific to the Ensemble app.

Who uses which app

If you are choosing an app, start from your question. A quick way to match a role to an app:

1. Scoring forecasts aloft? Use Upper Air for level-by-level wind and temperature skill, or Anomaly Correlation when you care about large-scale pattern skill against climatology.
2. Scoring near-surface weather? Use Surface for screen-level wind, temperature, and moisture, or Air Quality when the variables of interest are pollutant concentrations.
3. Scoring precipitation? Use Precipitation, where Threshold and GridScale plots let you see how skill depends on accumulation amount and spatial scale.
4. Working with an ensemble? Use Ensemble for spread, CRPS, Brier Score, and the reliability / ROC / performance diagrams that test probabilistic skill.
5. Tracking storms? Use Cyclone for tropical and extratropical track and intensity error, with per-basin, per-year, and per-storm views.
6. Verifying convection as objects? Use Objects for centroid distance, area and shape ratios, and object-based threat and skill scores.

Same controls, every time

Because all apps share the interface model from Figure 1, learning one app teaches you most of the others. Once you know how to pick a Data-Source, Region, Variable, Level, Threshold, and Statistic, switching apps is mostly a matter of which menus and plot types are available.

---

A derived, human-readable re-presentation — not official documentation. Sources: METexpress User’s Guide — Apps