Application Screenshots

Screenshots will be added before final submission. Representative placeholders are shown below indicating the intended content of each view.
Main 3D View — Paraguay Fire Events Cesium globe with active fire markers, smoke particles, and burn scars rendered on photorealistic terrain. Gran Chaco region, February–March 2026.
Timeline Animation Panel Day-by-day timeline slider showing fire spread progression. Burn scar accumulation visible across multiple days of the study period.
Minimap Weather Overlay Orthographic top-down minimap with temperature, wind arrow, and cloud cover layers simultaneously enabled over the study region.
Unity Editor Pipeline Tool WildfirePipeline editor window showing boundary controls, date range, FIRMS collection selector, and run summary log.
Fly-Through Export with Annotations In-progress fly-through recording showing an annotation card overlaid on the 3D scene with a waypoint sequence over Portugal.
In-Game Settings Panel Runtime pause settings menu showing toggles for smoke, wind arrows, connector fires, FPS counter, confidence threshold, and performance preset.

Quantitative Results

The following figures summarise the performance and data characteristics observed during platform validation across the four study areas.

Study Area Date Range (test) Raw Detections After Confidence Filter Final Markers
Paraguay (Gran Chaco) 01 Feb – 10 Mar 2026 13,279 ~9,800 ~3,200 (clustered)
Portugal Summer season sample TBD TBD TBD
Eastern Australia (NSW/QLD) Black Summer 2019–20 TBD TBD TBD
Hawaii Aug 2023 (Lahaina) TBD TBD TBD

TBD values will be completed in the final thesis text. Paraguay figures are from the primary validation run documented in the methodology chapter.

Implemented System Components

Data Pipeline (Unity Editor)

  • FIRMS Area API client with automatic 5-day chunking and deduplication
  • Shapefile writer (.shp / .dbf / .shx / .prj) for fire points and boundary polygons
  • CSV report with per-detection attributes (lat, lon, FRP, brightness, confidence, date)
  • Run summary log with total counts, centroid coordinates, and date range
  • Preset region library (Paraguay, Portugal, Australia, Hawaii) with centroid preview

Core Visualisation Engine

  • WGS84 → ECEF → Unity world-space coordinate pipeline via Cesium for Unity
  • Terrain-readiness spawn gate with exponential back-off (up to 15 retries, 0.5–2 s)
  • Fire clustering: radius-based merge with sqrt(count) scale + tiny-fire merge stage
  • Burn scar projection with terrain vertex conformance and geographic distance merging
  • Day-by-day timeline with incremental coroutine construction and configurable playback speed
  • Confidence filtering with auto-fallback to unfiltered load if result would be empty

Weather and Environmental Overlays

  • Temperature layer: JSON loader, blue–red colour ramp, optional location merging, TextMesh labels
  • Wind layer: directional arrow grid with configurable stride-based update for CPU savings
  • Cloud cover layer: disc markers with clear–overcast colour ramp, transparency support
  • All layers registered on the orthographic minimap; togglable independently at runtime

Camera and Presentation Tools

  • Fly-through path system: ECEF waypoints, configurable speed, CesiumOriginShift suspension during playback
  • Annotation system: per-waypoint-range text and image cards, draggable during recording, persisted screen position
  • MP4 and GIF export with configurable resolution, FPS, and GIF width
  • Camera-look gate: simulation advances only when viewport aims at active fire bounds

User Interface

  • Main menu with preset selection, date range entry, and performance preset toggle
  • Orthographic minimap with zoom, pan, and per-layer visibility controls
  • In-game pause settings panel: smoke, wind, connector fires, FPS counter, seconds-per-day, confidence threshold
  • FPS counter overlay persisted via PlayerPrefs
  • Runtime performance preset (Quality / Performance) affecting render scale, particle budgets, and LOD

Limitations and Future Work

  • No urban / land-use mask — Industrial heat sources in city centres can produce false-positive detections. A future version should apply an urban footprint mask to filter out known non-wildfire hotspots.
  • Single-ray terrain snapping — The current terrain height query uses a single vertical raycast. On steep volcanic terrain (Hawaii) or deep ravines, a multi-ray fan approach would improve surface adhesion.
  • No fire spread simulation model — The platform animates historical detections chronologically but does not model or predict fire spread. Integration of a physics-based spread model (e.g., Rothermel) is planned as future work.
  • Weather data resolution — Open-Meteo point-based weather data has coarser spatial resolution than the fire detections. Higher-resolution gridded weather products (e.g., ERA5 reanalysis) would improve overlay accuracy.
  • Mobile and web deployment — The current build targets desktop Windows. A WebGL or mobile build would broaden accessibility for emergency management stakeholders.