Glossary

A

A.R.D.I.

The attribute-rich Friuli Venezia Giulia regional wildfire registry, covering the Italian side of the Karst from 1990 onward. It is one of the primary historical fire sources used for calibration and cross-validation.

AI augmentation

Machine-learning models — a U-Net spread emulator, an ensemble surrogate and a scenario narrator — layered on top of the peer-reviewed physics kernel for faster or richer outputs. The physics kernel is always available as a fallback and as the ground truth against which the AI is checked, so the approximation never silently replaces the science.

AIB (Antincendio Boschivo)

Italian for "forest fire-fighting". In Friuli Venezia Giulia it is the governance framework for wildfire prediction, prevention, suppression, monitoring and the regional fire-prevention plan, and also names the operational suitability classes used for forest roads.

AOI (Area of Interest)

A defined territory the platform models — for example the Karst or the wider cross-border domain. Each AOI fixes the coordinate system, bounding box, default resolution and the catalogue of map layers available for it.

AOI Profile

A versioned configuration bundle that declares, for one territory, the data sources, calibration state, fuel taxonomy, weather provider and allowed simulation modes. Profiles are typed as operational (locally calibrated, can serve live forecasts), benchmark-only (for evaluation against external references) or research.

ARPA FVG OSMER

The regional weather authority and meteorological station archive for Friuli Venezia Giulia, and one of the live cross-border weather sources feeding the platform.

Arrival time

A simulation output in which each map cell stores the time, measured from ignition, at which the simulated fire reaches it. It answers "when does the fire get here?" for any point, line or area.

ARSO

The Slovenian Environment Agency. It provides weather archives, the Slovenian terrain model and LiDAR point clouds used on the Slovenian side of the Karst.

B

Barrier policy

How the simulator treats infrastructure when a fire spreads. Firebreaks and railways can block the fire outright, while ordinary and forest roads act as "soft" slowdowns that reduce — but do not always stop — the spread, depending on the chosen policy.

BUI (Buildup Index)

A component of the Canadian Fire Weather Index system that represents the total amount of fuel available for combustion, used as a proxy for deeper fuel dryness in the spread calculation.

Burn probability

An ensemble output showing how often each location burned across many simulation runs — a map of likelihood rather than a single predicted perimeter.

C

Cell2Fire

An external, cellular-automata wildfire-spread simulator used as an independent reference to benchmark the platform's own predictions.

CEMS / Copernicus EMS

The Copernicus Emergency Management Service. Its rapid-mapping activations provide satellite-derived fire perimeters (such as the 2022 Karst fires) used for calibration and cross-checking.

CFFDRS (Canadian Forest Fire Danger Rating System)

The Canadian national framework of fire-weather and fire-behaviour science — the Fire Weather Index, Fire Behaviour Prediction and growth components together — developed and maintained by the Canadian Forest Service. It is the peer-reviewed scientific core that the platform's engine reimplements in a clean-room form.

CHM (Canopy Height Model)

A map of vegetation height, typically computed as the surface model minus the terrain model. It captures how tall the canopy is above the ground.

CMA-ES

Covariance Matrix Adaptation Evolution Strategy — a derivative-free optimisation method used to tune the fire model's parameters against observed fire perimeters when the response is noisy or hard to differentiate.

COG (Cloud-Optimized GeoTIFF)

A raster image format that lets clients efficiently read just the spatial window they need. It is used for terrain, fuel, arrival-time and burn-probability layers.

Copernicus

The European Union's Earth-observation programme. The platform draws on its Sentinel satellite imagery, digital elevation models, ERA5 weather reanalysis and EFFIS wildfire products.

CRS (Coordinate Reference System)

The system that defines how map coordinates correspond to real-world locations. Mismatched coordinate systems are a common cause of layers appearing shifted or misaligned.

Curing fraction

The proportion of grass fuel that has dried out and turned from green to cured. Higher curing means grass fires spread faster; it is an important seasonal input for Karst grassland fuels.

D

DEM / DTM / DSM

Digital Elevation, Terrain and Surface Models. The terrain model (DTM) describes the bare ground, while the surface model (DSM) includes the tops of trees and buildings.

Digital twin

A continuously updated virtual replica of the real Karst landscape — its terrain, fuels, infrastructure and weather — used to run simulations and support prevention and response decisions.

Dinaric

Relating to the Dinaric Alps and the Dinaric Karst — the limestone mountain belt that runs south-east from north-east Italy and Slovenia through Croatia, Bosnia and Herzegovina, Montenegro and Albania. The cross-border Italy–Slovenia Karst that the platform models is the north-western tip of that belt, which is why the standard Canadian fuel models have to be adapted to Dinaric vegetation rather than used as-is.

Dixie Fire

A major 2021 California wildfire used as a United States benchmark case for validating the platform's spread predictions against well-documented real-world data.

dNBR (Differenced Normalized Burn Ratio)

A satellite-derived burn-severity index computed from before-and-after imagery of a fire. Higher values usually indicate stronger vegetation loss and more severe burning.

DST (Decision Support Tool)

Software that turns model outputs — routing, scenarios, restoration surfaces — into information that helps operators make decisions.

E

ECMWF IFS

The Integrated Forecasting System of the European Centre for Medium-Range Weather Forecasts, used to supply forecast weather for scenario simulations.

EFFIS

The European Forest Fire Information System, part of the Copernicus Emergency Management Service. It provides European fire-danger forecasts, active-fire detection and burned-area mapping.

Ensemble envelope (p10 / p50 / p90)

A summary of many simulation runs as lower, median and upper spread bounds. The median (p50) is the central estimate; p10 and p90 frame the plausible range.

ERA5

A global atmospheric reanalysis dataset from ECMWF used as a consistent weather baseline for historical replays and for scaling beyond the station network.

ETA (Estimated Time of Arrival)

Estimated Time of Arrival — how long it takes to reach a place. On the platform it appears both as a fire's arrival time at a location and, in intervention planning, as the travel time for a response vehicle to reach the fire over the road network.

F

FARSITE / FlamMap

A widely used United States fire-behaviour and fire-growth modelling family from the Forest Service, used as an external reference for benchmarking.

FastAPI

The Python web framework used to expose the platform's simulation engine as an HTTP service. It supports asynchronous ensemble fan-out, live perimeter streaming with Server-Sent Events and machine-readable GeoJSON / QGIS-ready exports, making the engine a drop-in for the operator dashboard, Cesium 3D clients and any HTTP-speaking consumer.

FBP (Fire Behaviour Prediction)

The Canadian system that predicts rate of spread, intensity and fuel consumption for different fuel types. The platform extends it with a Karst-specific fuel set.

FFMC / DMC / DC

The three fuel-moisture codes of the Fire Weather Index system: the Fine Fuel Moisture Code (surface litter), Duff Moisture Code (loosely compacted organic layers) and Drought Code (deep, compact organic matter).

Fire registry

The consolidated cross-border record of historical fires, joining Italian and Slovenian sources together with the provenance of each record, used as ground truth for evaluation.

Firebreak / fuelbreak

A linear or strip-shaped feature — cleared, mown or otherwise managed — designed to interrupt or slow fire spread by removing or reducing the fuel available to burn.

FRP (Fire Radiative Power)

A measure of the energy a fire radiates, derived from satellite active-fire detections, used to gauge fire intensity from space.

Fuel model

A standardised description of a vegetation type's burning behaviour — its load, structure, moisture of extinction and spread characteristics. Fuel models let the simulator translate "what is growing here" into "how it burns".

Fuel moisture

The amount of water held in vegetation and litter. Drier fuel ignites and spreads more readily, so fuel moisture is a key driver of fire behaviour.

FVG (Friuli Venezia Giulia)

The Italian region forming the Italian side of the cross-border Karst project domain.

FWI / ISI

The Fire Weather Index and its Initial Spread Index component. The FWI rates overall fire-weather severity from temperature, humidity, wind and rain; the ISI combines wind and fine-fuel moisture to estimate how fast a fire would spread.

G

GeoJSON

An open standard for encoding geographic features — points, lines, polygons, fire perimeters — and their attributes in plain JSON. The simulator emits its outputs in this format so they can flow directly into web maps, dashboards and GIS tools without translation.

GeoSphere Austria

Austria's federal institute for geology, geophysics, climatology and meteorology, formed in 2023 by merging the Central Institution for Meteorology and Geodynamics (ZAMG) with the Geological Survey of Austria. As Austria's national meteorological service it issues the country's official weather and severe-weather warnings — including those carried by MeteoAlarm — for the Alpine region neighbouring the Karst.

GIS (Geographic Information System)

Software and methods for storing, analysing and visualising spatial data — terrain, fuels, infrastructure, fire perimeters and the layers built on top of them. The platform is GIS-native and produces outputs that open directly in tools such as QGIS.

GKOT

A Slovenian LiDAR point-cloud product used to derive canopy and terrain information on the Slovenian side of the Karst.

Global Fire Atlas

A satellite-based global database of individual fire events, providing perimeters, ignition points and spread patterns used as a worldwide benchmark catalogue.

Ground truth

The independently observed reality against which a model is evaluated — here, the Copernicus EMS rapid-mapping perimeters, Sentinel-2 burn-severity scenes and LSA-SAF fire-radiative-power timelines that simulated fire perimeters are scored against during calibration and benchmarking.

GWIS

The Global Wildfire Information System, a worldwide wildfire-monitoring and fire-danger service connected to the Copernicus and EFFIS workflows.

H

Hard barrier

An infrastructure feature treated as completely impassable by the fire, structurally stopping the front where it lies — in contrast to a soft barrier that only slows spread.

Hausdorff distance

A geometric measure of how far apart two shapes are — here, the largest gap between a simulated fire perimeter and the observed one. Smaller values mean the predicted and real perimeters match more closely.

Hindcast

Re-running a model over a past period using historical conditions, to test how well it reproduces what actually happened.

Huygens propagation

The method the simulator uses to advance a fire perimeter outward, treating each point on the front as a source of new spread — much like wavefronts in physics. Soft barriers act here as local slowdowns.

I

Ignition

The point and moment at which a fire starts. In the simulator it is the seed from which the modelled fire grows.

IgnitionCandidate

A possible fire start identified by fusing open-source and partner evidence in space and time. Each candidate carries a confidence level and a recommended action, from "store only" through "operator review" to launching a simulation.

IoU (Intersection over Union)

A score from 0 to 1 measuring how much a simulated fire perimeter overlaps the observed one. A value of 0 means no overlap, so it is read alongside area-error and shape diagnostics rather than on its own.

IRWIN

A United States system that assigns each wildfire incident a unique, persistent identifier shared across federal datasets, used to tie together records for benchmark fires.

ITASLO

Shorthand for the cross-border Italian/Slovenian Karst fire events used together for joint calibration of the model.

IUAV

Università IUAV di Venezia — the lead academic and coordination partner for Karst Firewall 5.0, steering the cross-border planning, policy and wildfire-risk framework.

K

K-class fuels (K01–K09)

The Karst-specific fuel taxonomy used by the platform, spanning local vegetation types such as dry grassland, low Mediterranean shrubland (garrigue), broadleaf and conifer forest, mixed woodland and dense maquis.

Karst Firewall 5.0

The cross-border Italy–Slovenia Interreg programme for wildfire prevention and a digital twin of the Italian/Slovenian Karst, of which this platform and its fire-growth engine are the technical core.

Karst priors

The set of calibrated parameter values that adapt the generic fire model to the Karst. They are versioned and selected by territory and season so that production results are reproducible and auditable.

L

LANDFIRE

A United States national geospatial programme providing consistent vegetation, fuel and fire-regime data. Its fuel grids serve as inputs for the platform's United States benchmark fires.

Layer catalogue

The published description of the map layers available for an area or a simulation run — their type, rendering, status, artifacts and quality metadata — that client applications read to know what they can display.

LDT (Local Digital Twin)

A lightweight, locally scoped digital-twin context used in the sensor and scenario-planning parts of the platform.

LiDAR

A remote-sensing technique that measures distances with laser pulses to build very detailed 3D models of terrain and vegetation. It is the source of the Karst's high-resolution terrain and canopy-height data.

LoRaWAN

A low-power, long-range radio networking standard for the Internet of Things. The electronic-nose and field sensors use it to send small readings across several kilometres on very little energy, without relying on mobile-phone coverage.

M

Manifest

A per-run audit file that records everything about a simulation — its inputs, model settings, outputs, warnings, fallbacks and calibration summary — so any result can be traced and reproduced.

MeteoAlarm

The pan-European severe-weather warning network, operated by EUMETNET, that gathers the official alerts of national meteorological services into a single colour-coded (green / yellow / orange / red) scheme. The platform listens to its feed as one of the triggers that can prompt a fire-danger nowcast when dangerous fire weather is forecast for the region.

MODIS

A NASA satellite sensor widely used in public wildfire services for active-fire detection and burned-area mapping.

Moisture coupling

A model parameter controlling how strongly fuel moisture and humidity suppress fire spread. Higher coupling makes damp conditions slow the simulated fire more sharply.

MTBS

Monitoring Trends in Burn Severity — a United States programme that maps every large wildfire since 1984 using satellite-derived burn severity, providing reference perimeters for benchmarking.

N

Nadina Lake

A 2018 boreal wildfire in British Columbia, Canada, used as the first cross-continental benchmark for the platform — a test of how the uncalibrated model performs far outside the Mediterranean Karst.

Natura 2000

The European Union network of protected nature sites. In the platform it appears as a conservation overlay that scenario planning and impact assessment must take into account.

NBAC

The National Burned Area Composite — Canada's annual nationwide compilation of wildfire perimeters, used as a cross-reference for Canadian benchmark fires.

NDVI (Normalized Difference Vegetation Index)

A satellite-derived measure of vegetation greenness and vigour from red and near-infrared bands, used to track how green — and, by comparison with the seasonal norm, how stressed — the landscape is.

Nelder-Mead

A classic derivative-free numerical optimisation method that searches for the best parameter values by iteratively reshaping a "simplex" of trial points. Used alongside CMA-ES to tune the fire model against observed perimeters when the response is noisy or not smoothly differentiable.

NIFC

The United States National Interagency Fire Center, a federal coordination body whose public services provide incident geometry and attributes for United States benchmark fires.

No-calibration baseline

A simulation run using only standard literature defaults — no local Karst tuning, flat terrain, no barriers. It establishes the model's accuracy floor before calibration is applied.

Nowcast

A simulation run for current or near-current conditions, answering "if a fire started now, where would it go?"

O

Orthophoto

An aerial or satellite photograph that has been geometrically corrected so that distances on the image are uniform and it can be used like a map.

OSINT

Open-Source Intelligence — an opt-in pipeline that gathers open and partner evidence (satellite fire detections, lightning, smoke sensors, cameras, social reports, air-quality and operational feeds) to flag possible incidents, without ever launching a response on its own.

P

Perimeter

The outline of a fire at a given moment. The simulator advances a perimeter through the landscape and reports both intermediate and final perimeters.

PMTiles

A single-file map-tile format used to deliver vector or raster layers efficiently to the web map.

PyroWISE

The physics-based, probabilistic wildfire-simulation and GIS engine at the heart of the platform, developed by Infordata. It advances a fire front across the real cross-border terrain from fuels, weather, topography and infrastructure and, by running many slightly varied simulations, reports the outcome as probabilities rather than a single line — arrival-time, perimeter and burn-probability maps for nowcasts, what-if scenarios and historical replays, each with full data provenance.

Q

QGIS

A free and open-source desktop Geographic Information System for viewing, editing and analysing spatial data. Project layers — terrain, fuels, fire perimeters and firebreaks — can be inspected in QGIS, and exported data such as Shapefiles opens there directly.

R

RAWS

Remote Automated Weather Stations — a United States network of automated stations providing hourly surface weather used to reconstruct conditions for United States benchmark fires.

Replay

A simulation mode that re-runs a historical fire or scenario, used for evaluation, calibration or communication.

ROS (Rate of spread)

How fast a fire front advances. Soft barriers reduce the local rate of spread, while calibration tunes the base rate and the influence of wind and moisture.

Rothermel model

A foundational mathematical model of surface fire spread, relating fuel, moisture, wind and slope to rate of spread. It underpins much of modern fire-behaviour prediction.

S

Scenario

A user-defined simulation for planning, what-if analysis or training, run with conditions chosen by the operator rather than the live present.

Scenario Brick

A reusable building block for composing a what-if simulation. Each brick fixes one part of the set-up — the ignition, the time horizon, the ensemble size, the fuel state or the model mode — so operators can assemble, save and re-combine scenarios from standard, repeatable pieces instead of re-entering every parameter.

Sentinel-2

A European Copernicus optical satellite constellation providing imagery for vegetation monitoring, true-colour context, NDVI and high-resolution burn-scar mapping.

SHP (Shapefile)

A widely used geospatial vector format for points, lines and polygons with their attributes — in practice a small set of files (.shp, .shx, .dbf and others). It is a common way to exchange fire perimeters, firebreaks and area boundaries with GIS tools such as QGIS.

Soft barrier

A passable feature — such as an ordinary or forest road — that slows fire spread as the front crosses it, without permanently blocking the fire downstream.

Soft roads (soft_roads)

The platform's default barrier policy for the Karst. Minor linear features — footpaths, forest tracks, dry-stone walls and ordinary roads — act as soft barriers that slow the fire and can lower its intensity as the front crosses them, but do not reliably stop it; firebreaks and railways still block it outright.

STAC (SpatioTemporal Asset Catalog)

A common standard for describing and cataloguing geospatial data products so they can be discovered and used consistently.

T

Terrain tileset

A set of 3D terrain tiles, built from the bilateral Italian and Slovenian elevation models, that enables the platform to render the Karst landscape in three dimensions with terrain, slope and hillshade overlays.

U

U-Net spread emulator

A convolutional neural network — borrowing the encoder/decoder "U-Net" architecture from image segmentation — trained to approximate the physics-based fire-spread output quickly. One of the AI-augmentation layers on top of the CFFDRS kernel, with the physics kernel always available as fallback and as the ground truth the emulator is checked against.

V

V_INCENDI_CT

A long-running Italian wildfire dataset spanning 1964 to the present with fewer attributes than A.R.D.I. It extends the Italian fire registry over a wider time span.

VIIRS

A satellite sensor used for active-fire and thermal-anomaly detection at finer spatial resolution than MODIS.

VOC (Volatile Organic Compounds)

Carbon-based gases released by vegetation and, in far greater quantity, by heating, smouldering and burning plant matter. The electronic-nose sensor network watches for the characteristic mix of VOCs that signals combustion, so a fire can be flagged in its earliest stage — often before flames or visible smoke.

W

WFIGS

The United States Wildland Fire Incident Geospatial Service, a public source of wildfire incident records — points of origin, perimeters and identifiers — used to source United States benchmark fires.

What-if run

A simulation of a hypothetical situation — "what if a fire started here, in these conditions?" — rather than the live present (a nowcast) or a past fire (a replay). What-if runs are composed from Scenario Bricks to test ignition points, weather and prevention measures such as firebreaks.

Wind coupling

A model parameter controlling how strongly wind speed and direction shape fire spread. Higher coupling stretches the simulated perimeter further downwind.

WISE / Prometheus

The Canadian Forest Service's reference wildfire-growth modelling stack. The platform's engine is a clean-room reimplementation of this public, peer-reviewed science and benchmarks against it.

Z

ZGS

Zavod za gozdove Slovenije, the Slovenian Forest Service — a source of Slovenian forest-road, firebreak and conservation layers and of the Slovenian wildfire registry.

ZRC SAZU

The Research Centre of the Slovenian Academy of Sciences and Arts — a Slovenian research partner contributing scientific, territorial and environmental expertise and cross-border data.