JTWeatherNow | Forecast Validation

ConnectingWaiting for observational feeds

How is the forecast doing?

Start with the map for a simple answer. Green locations are within the configured operational tolerance; blue and red show where the forecast ran low or high. Open the other views only when you want more detail.

Selected cycle—

Verified through—

Stations used—

Selected detail network—

Forecast selection

Choose the forecast cycle

Overview and Surface Stations also use the weather-variable and unit controls. Upper Air has its own vertical-variable selector.

Forecast cycle Weather variable Display units

Reading the latest verification…

Start here: Read the summary, then use the colored map to see where the forecast was on target, too low, or too high.

Primary validation view

Where is the forecast performing well — and where is it missing?

Map coverageMap period

Plain-language forecast summary

How did the forecast perform?

—

Loading station-level verification…

Map colors

Within operational toleranceUnderforecastOverforecastLarge but mixed errors

Operational tolerance is loading…

How is the green range determined?

The green band is a configurable operational screening tolerance, not an official NWS pass/fail standard. For the latest verified hour, the absolute forecast-minus-observation error must be within the tolerance. For a multi-hour period, MAE must be within the tolerance. Locations outside it are labeled under- or overforecast from the bias; amber is used when large errors occur in both directions and the bias nearly cancels.

The current defaults are practical starting values: temperature ±2.0°C, dewpoint ±2.5°C, wind ±5 kt, sea-level pressure ±3 hPa, SST ±1.5°C, and 1-hour precipitation ±2.5 mm. These can be changed in the validation configuration. They have not yet been calibrated from a long JTWeatherNow historical-error climatology.

Map colors describe forecast error for the selected cycle, network, variable and period. Positive bias means the forecast was higher than observed; negative bias means lower.

Forecast journey

Which forecast is being scored?

—

Loading the selected forecast timeline…

Forecast archived—

The original forecast hours preserved for later verification

Verified so far—

Hours whose valid times have occurred and matched observations

Newer forecast cycles—Issued later and scored separately — they never overwrite this forecast

Surface verification

Explore networks or individual stations

Use this view when you want more than the map summary. Network charts show broad skill; Station Explorer opens the full time series for one public observing site.

Step 1 · Observation network

Where should the forecast be checked?

Airport, mesonet and marine observations remain separate so dense networks or different environments do not hide one another.

Live upper-air observations · NOAA/NCEP MADIS weather balloons

Operational radiosonde verification

Operational WRF

This verifies the same JTWeatherNow Operational WRF 14 km forecast selected above. Radiosonde profiles are downloaded automatically on Forecaster3, matched to preserved WRF profiles at standard pressure levels, and scored by forecast day. F00 is not counted as Day 1 skill.

Plain-language guide

Lower error means the operational forecast more closely matched the observed weather-balloon profile. Sounding times are shown in UTC first (Z time) so launch times remain unambiguous across time zones. Your browser-local time is shown second in Station Explorer.

200–300 hPaUpper troposphere

500 hPaMiddle troposphere

700–850 hPaLower troposphere

925–1000 hPaNear surface

Network vertical profile

MAE by pressure level

—

How to read it: pressure decreases upward. 925–1000 hPa is near the surface, 500 hPa is mid-atmosphere, and 200–300 hPa is upper troposphere. Farther left means smaller MAE and a closer forecast match.

Day 1–5 upper-air skill

Radiosonde MAE by forecast day

A day appears only after a matching radiosonde launch has occurred and been ingested. Launch times are handled in UTC and are not limited to exactly 00Z/12Z.

Numerical profile

Pressure-level statistics

Level	Stations	Samples	Bias	MAE	RMSE

On smaller screens, scroll the table sideways to see Bias, MAE, and RMSE.

Sounding station explorer

Observed profile vs operational WRF

Choose a radiosonde station and launch time to inspect the vertical profile directly.

Time standard: UTC (Z) is shown first. Your browser-local time is shown second; it is not assumed to be the sounding station's local time.

StationSounding time

No operational sounding match is available for this cycle yet.

Advanced verification

Research diagnostics, data sources & methodology

These products are intentionally separated from the basic forecast view. They include experimental/certified WRFDA comparisons, provenance and technical metric definitions.

Experimental/certified upper-air comparison

WRFDA vs control radiosonde diagnostics

Certified experiment

These are not the live operational sounding scores above. They are imported from the established certified WRFDA-vs-control verifier and remain separate so experimental results cannot be confused with operational forecast performance.

Vertical profile

MAE by pressure level

MAE

Pressure decreases upward: 1000 hPa is near the surface, while 200–300 hPa represents the upper troposphere.

Improvement by level

WRFDA minus control benefit

Positive values follow the certified verifier convention and indicate lower WRFDA error than control.

Numerical profile

Pressure-level statistics

Level	WRFDA MAE	Control MAE	Improvement	N

Observation provenance

What is being checked?

Live METAR, mesonet, marine, and radiosonde observations feed operational verification. Radar/lightning and experimental WRFDA-vs-control diagnostics remain separate certified products.

Selected metric table

Period	Stations	Samples	Bias	MAE	RMSE	Correlation

How to read this page

MAE is the typical absolute error; lower is better. RMSE emphasizes larger misses; lower is better. Bias shows whether JTWeatherNow tends to forecast too high or too low; zero is ideal. Correlation indicates how closely forecast and observed changes move together; closer to 1 is stronger.

F00 is excluded from Day 1. Observations are quality controlled and matched in space and time to preserved WRF output. Airport, mesonet and marine networks remain separate because they sample different environments and densities. Sea-level pressure uses a native WRF SLP field when available; otherwise it is diagnosed from WRF model pressure, temperature, moisture and geopotential height before comparison with observed sea-level pressure.

Station-level MADIS details remain unpublished by default; its network-level verification remains fully available.

RAOB / radiosonde verification now has two clearly separated layers: live NOAA/NCEP MADIS soundings matched directly to the operational WRF, plus optional certified WRFDA-vs-control experiment diagnostics. The live section shows network and station vertical profiles graphically and numerically.

How is the forecast doing?

Choose the forecast cycle

Where is the forecast performing well — and where is it missing?

How did the forecast perform?

Which forecast is being scored?

Explore networks or individual stations

Where should the forecast be checked?

Detailed network charts

What do the observations say?

Forecast performance by day

Does skill change from Day 1 to Day 5?

Rolling Day 1–5 performance

Recent forecast history

Network-average forecast and observed values

Forecast error by lead hour

Selected score by forecast day

Stations contributing by day

How forecast errors are distributed

High vs low tendency by lead hour

Inspect one observing site in detail

JTWeatherNow compared with observations

Forecast minus observed

Forecast vs observed scatter

Station MAE by forecast day

Observation time offset

Operational radiosonde verification

MAE by pressure level

Radiosonde MAE by forecast day

Pressure-level statistics

Observed profile vs operational WRF

Research diagnostics, data sources & methodology

WRFDA vs control radiosonde diagnostics

MAE by pressure level

WRFDA minus control benefit

Pressure-level statistics

Radar, precipitation & convection skill

WRFDA vs control

WRFDA improvement relative to control

What is being checked?

Metrics & methodology

Selected metric table

How to read this page