Updating operational status…
Refreshing the latest recorded activity.
Last detection review · —Queue status unconfirmed
Next cycle unconfirmed
Measured data available · hourly PM10
Latest measured data in saved inputs. Corridor coverage is checked station by station.
Reportable impacts
Current valid forecastsUpdating forecasts…
LIVE OPERATIONS
Active wildfire plumes
All active operational plumes
LIVE OPERATIONS
Incident
Spain, Portugal, Italy, France & Switzerland
Execution & scientific detailsScientific gates, integrity and final-product availabilityDETAILS ▾
Checking scientific state…
Loading map…
0 active fires
0 active particles
0 positive corridor stations
ARCHIVE
Historical Forecasts & Validation
5 YEAR ARCHIVE
October 2026
Mo
Tu
We
Th
Fr
Sa
Su
Archive unavailable
Select a day with archived forecasts.
PROSPECTIVE EEA VALIDATION
Model Performance
Frozen forecasts verified against subsequent EEA E2a/UTD observations
PENDING
RMSE
—
Median across physical fire events
MAB
—
Mean absolute bias · MAE-equivalent
Willmott IOA
—
Classical index of agreement d
FAC2
—
0.5 ≤ prediction / observation ≤ 2
EU MQI 24h
—
EU 2024/2881 prospective screening
RMSE, MAB, Willmott IOA and FAC2: issue metrics are averaged within each physical fire event, then the median across evaluable physical fire events is reported.
Events improved by MAE · —Median MAE skill · —Median RMSE skill · —Median ΔPearson r · —Evaluable forecasts · —Physical events · — / —EEA support · — receptor-hours
EU MQI 24h is a prospective Directive 2024/2881 screening: each forecast issue is evaluated independently from T0+1 to T0+24, with at least 18 identical valid PlumeTrace–EEA hourly pairs. Issues are never mixed. Measurement uncertainty uses the Directive DQO envelope; no method-specific uncertainty is assumed. This is not a formal regulatory compliance claim. Scientific receptor eligibility remains peak Fire > 1 µg/m³.
CURRENT R6 CASE
Operational case state
PT
No current R6 case
The next operational case will appear here automatically.
INFO
How it works
PlumeTrace is a receptor-oriented wildfire air-quality forecasting system designed to estimate when, where and by how much wildfire smoke may affect populated areas. It combines satellite fire detection, numerical weather prediction, atmospheric composition data and ground-based air-quality observations within a single prospective forecasting workflow.
For each detected fire, PlumeTrace simulates atmospheric transport using a Lagrangian approach driven by ICON-EU meteorological forecasts, including the evolving wind field responsible for plume transport and dispersion. The system identifies the downwind transport corridor, screens the monitoring network for physically connected stations and selects eligible urban receptors where a quantitative PM₁₀ forecast can be produced.
A key scientific concept is the Lagrangian positive corridor. Rather than assuming that every monitoring station near a fire is potentially affected, PlumeTrace identifies the portion of the atmosphere that is dynamically connected to the source according to the simulated transport. Only receptors lying within this physically consistent corridor can be considered candidates for a wildfire contribution.
The continuity of transported particle families — the Cohorts Chain — is used to preserve the causal connection between the wildfire source, intermediate locations and the final receptor. This helps distinguish physically consistent plume transport from coincidental geographical proximity.
Scientific hard gates are applied before a quantitative forecast is published. They require causal and physical support for the transported signal and restrict final quantitative predictions to eligible urban receptors. If these conditions are not met, PlumeTrace does not claim a wildfire impact.
To represent the non-fire component of air pollution, PlumeTrace also uses Copernicus Atmosphere Monitoring Service (CAMS) information, including large-scale atmospheric composition and mineral-dust contributions. These data are combined with station-specific background information so that the forecast can distinguish between the expected ambient PM₁₀ concentration and the additional contribution attributable to wildfire smoke.
Forecasts are then evaluated prospectively against independent EEA E2a/UTD observations. Each forecast issue is treated independently: observations available after the forecast has been issued are used only for verification and never to retrospectively recalibrate the prediction.
Model performance is monitored using standard indicators including RMSE, Mean Absolute Bias (MAB), Willmott Index of Agreement and FAC2, together with specific metrics designed to quantify whether explicitly including the wildfire contribution improves the forecast.
EU Modelling Quality Indicator
For PM₁₀, a sampling point passes the screening when MQI ≤ 1. The Directive quality objective requires this condition to be achieved at at least 90% of eligible sampling points.
In PlumeTrace, the 24-hour MQI is reported as a prospective Directive-based screening indicator. Each forecast issue is evaluated independently over its own T0+1 to T0+24 h window, using at least 18 valid paired forecast–observation hours. Different forecast issues are never mixed. This should be interpreted as a model-quality screening indicator, not as a formal regulatory compliance statement.
ABOUT PLUMETRACE
We tell the story of the air.
Where it comes from, how it moves, what it carries — and where it matters.
PlumeTrace follows the journey of wildfire emissions through the atmosphere and translates a complex physical process into something easier to understand: what may happen to the air people actually breathe.
We combine observations, atmospheric transport modelling and independent environmental data to reconstruct that journey. Rather than simply asking where smoke could travel, we look for evidence connecting a fire, the atmosphere along its path, and the places potentially affected.
We observe. We trace. We connect the evidence.
And when the atmospheric evidence is not strong enough, PlumeTrace does not claim an impact.