Skip to content

About

FireSignal is an interactive platform that uses NASA satellite data to explore, analyze, and compare historical fire activity across India.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

29 Commits

Folders and files

Repository files navigation

FireSignal

Satellite Active-Fire Observatory & Cross-Sensor Harmonization for India

License: MIT NASA Space Apps Challenge 2026 React 19 TypeScript 7 Vite 8 Tailwind CSS 4

FireSignal is an interactive scientific exploration platform and empirical harmonization engine developed for the NASA Space Apps Challenge 2026 challenge: "Harmonization of MODIS and VIIRS Hot Spots".

The platform explores 23 years of satellite active-fire observations across India, provides an interactive vector cartographic observatory for 36 administrative regions, and evaluates empirical cross-sensor harmonization models that align modern high-resolution VIIRS observations with the long-term MODIS climate record.


Canonical Application Views

FireSignal is structured around four primary client-side analytical views accessible directly via URL hash navigation:

View Canonical Route Description
Geographic Observatory /#observatory Interactive SVG map of India (Albers equal-area projection) displaying regional occupied grid-cell-days (OGCD) across 36 states and union territories. Features an annual/monthly temporal scrubber (2003–2025), sensor toggle (MODIS vs. VIIRS), regional inspection panel, choropleth thresholds, and cached return navigation.
Historical Patterns /#historical-patterns 23-year national daily time-series analysis (2003–2025), interactive calendar heatmaps, annual totals, and multi-year seasonal progression curves.
Sensor Laboratory /#sensor-laboratory Cross-sensor empirical harmonization evaluation, comparing raw VIIRS against national and regional transfer models (global and seasonal adjustments) across chronological training, validation, and held-out test splits.
Research & Methods /#research-methods Comprehensive methodological framework, mathematical formulations of OGCD, spatial projection rationale (EPSG:6933), validation protocols, and scientific caveats.

Research Motivation & Problem Statement

Satellite remote sensing of active fires relies on thermal anomalies detected in middle- and thermal-infrared spectral channels. For over two decades, the MODIS sensors aboard NASA's Terra and Aqua satellites provided the global scientific standard for fire monitoring (1 km spatial resolution at nadir). In 2011, the VIIRS instrument aboard the Suomi-NPP satellite introduced higher spatial resolution (375 m at nadir), wider swath coverage without edge-of-scan pixel degradation, and substantially enhanced sensitivity to smaller, cooler fires.

Because VIIRS has a nominal nadir footprint area approximately 7.1 times smaller than MODIS (0.1406 km² vs. 1.0 km²), exhibits distinct off-nadir pixel expansion characteristics across the swath, detects fires at lower radiant heat thresholds, and passes overhead at different local solar times, raw MODIS and VIIRS hotspot detections cannot be treated as directly interchangeable measurements. A naive concatenation of their records introduces an artificial multi-fold jump in apparent fire activity after 2012.

FireSignal addresses this discontinuity by:

  1. Formulating a common, physically consistent analytical unit: Occupied Grid-Cell-Days (OGCD) on an equal-area 5 km grid (EPSG:6933).
  2. Empirically evaluating cross-sensor statistical transfer models (global and monthly seasonal adjustments at national and state levels) on held-out test years (2024–2025).
  3. Making sensor differences and harmonization trade-offs interactive and transparent through the web-based Geographic Observatory.

Core Scientific Concepts

1. Occupied Grid-Cell-Days (OGCD)

To mitigate double-counting caused by multiple detections within a single pixel footprint or multiple satellite passes on the same day, FireSignal aggregates detections into Occupied Grid-Cell-Days (OGCD):

$$\text{OGCD}(c, d, s) = \begin{cases} 1 & \text{if } \ge 1 \text{ detection by sensor } s \text{ in cell } c \text{ on date } d \ 0 & \text{otherwise} \end{cases}$$

A grid cell contributes at most one unit of activity per sensor per calendar day, regardless of whether one or ten thermal anomalies were flagged within that cell.

2. Equal-Area Coordinate Reference System (EPSG:6933)

Standard equirectangular (latitude/longitude) grids suffer from significant poleward area distortion. FireSignal projects all active-fire point detections into the Cylindrical Equal-Area / EASE-Grid 2.0 Global projection (EPSG:6933) using nominal 5,000 m $\times$ 5,000 m ($25\text{ km}^2$) grid cells. Every grid cell represents an equal surface area across all latitudes in India.

3. What FireSignal Measures (and What It Does Not)

Important

  • FireSignal measures: Detected thermal anomaly spatial density on a discrete equal-area spatio-temporal grid (OGCD).
  • FireSignal does NOT measure: Confirmed individual wildfires, total burned area (hectares), fire severity, fire radiative energy (FRE), or ground-truth fire perimeters.
  • Absence of detection: Zero OGCD in a cell indicates that no satellite detection satisfied quality and cloud-clearing thresholds. It does not prove the absence of active fires, as persistent cloud cover, heavy smoke, and orbital gaps obscure surface thermal anomalies.
  • Statistical reference: MODIS serves as the historical baseline reference, not as absolute physical ground truth. Statistical harmonization does not constitute physical instrument recalibration.

Satellite Data Sources & Sensor Characteristics

FireSignal ingests active-fire data from the NASA Fire Information for Resource Management System (FIRMS):

Sensor Platform Spatial Resolution (Nadir) Archive Period Used Records in India Archive Primary Overpass (Equator)
MODIS Terra & Aqua 1,000 m 2003–2025 (23 yrs) 1,801,292 10:30 & 22:30 (Terra) / 13:30 & 01:30 (Aqua)
VIIRS Suomi-NPP 375 m 2013–2025 (13 yrs) 7,634,234 13:30 & 01:30

Sensor Discrepancy Baseline (March–May 2023 India Experiment)

During the intensive 3-month peak fire season baseline experiment (March 1 to May 31, 2023):

  • Raw Thermal Detections: MODIS detected 37,269 raw hotspots, whereas VIIRS detected 313,980 raw hotspots (8.42$\times$ higher).
  • 5 km Spatial Aggregation (OGCD): MODIS recorded 25,700 OGCD, while VIIRS recorded 160,672 OGCD (6.25$\times$ higher).
  • Daily Co-occurrence: National daily total OGCD correlated strongly ($r = 0.8841$), indicating high temporal synchrony despite scale discrepancies.
  • Day-to-Day Change Correlation: Day-to-day activity changes exhibited a moderate correlation ($r = 0.4776$) with 65.93% directional agreement.
  • Diurnal Divergence: The VIIRS-to-MODIS OGCD ratio was 5.66$\times$ during daytime and jumped to 12.16$\times$ during nighttime, reflecting VIIRS's heightened capability to detect cool, small nighttime agricultural burns invisible to MODIS.

Statistical Harmonization & Evaluation

FireSignal evaluates empirical models designed to predict expected MODIS-equivalent OGCD from raw VIIRS observations.

Chronological Data Partitions

To prevent data leakage and simulate operational deployment, models are trained and tested on non-overlapping chronological splits:

  • Training Period: 2013–2021 (9 years)
  • Validation Period: 2022–2023 (2 years)
  • Held-Out Test Period: 2024–2025 (2 years reserved strictly for final reporting)

Evaluation Results on Held-Out Test Data (2024–2025)

  1. National Harmonization Models:

    • Raw VIIRS Baseline: High systematic offset due to resolution divergence.
    • Global Adjustment: Proportional scaling based on training ratio.
    • Seasonal Adjustment: Month-of-year adjustment factors capturing agricultural burn seasons. Reduced national monthly Mean Absolute Error (MAE) by 19.7% and Root Mean Squared Error (RMSE) by 24.6% compared to global scaling.
  2. Regional Harmonization Models (State Level):

    • Evaluated across 28 Indian administrative regions with at least 12 paired monthly observations in the held-out test set.
    • Regional Seasonal Adjustment: Achieved an MAE of 43.56 OGCD and RMSE of 107.65 OGCD.
    • Outperformed national seasonal adjustment with 38.6% lower MAE and 38.5% lower RMSE.
    • Performance improved in 24 out of 28 regions; 4 regions with sparse or highly erratic fire activity showed slight degradation, illustrating the limits of purely local statistical adjustments in low-signal regimes.

Technology Stack

Layer Technologies Purpose
Frontend Framework React 19, TypeScript 7, Vite 8 Reactive component tree, strict type safety, fast bundling
Styling & Design System Tailwind CSS 4, Lucide Icons Dark Observatory design tokens, accessible contrast, responsive layout
Cartography & Charts Custom SVG Albers Projection, Recharts Interactive India ADM1 choropleth, annual timelines, anomaly sparkbars
Static Analysis & Testing Node.js Test Runner, Oxlint, TypeScript 7 43 automated regression tests, zero lint warnings, full build verification
Scientific Processing Python 3, NumPy, pandas, GeoPandas, SciPy, pyproj Equal-area spatial binning, point-in-polygon joins, regression pipelines
Geographic Simplification Mapshaper Topology-preserving 15% Douglas–Peucker simplification for web performance
Production Hosting Vercel Edge Network Global CDN delivery of precomputed static datasets and SPA bundle

Repository Architecture

firesignal/
├── frontend/                     # React 19 + TypeScript + Vite web application
│   ├── public/data/              # Precomputed browser JSON datasets and simplified GeoJSON
│   │   ├── regional_monthly_5km.json       # 36-region monthly OGCD (2003–2025)
│   │   ├── historical_calendar_5km.json    # National daily OGCD time series
│   │   ├── india_adm1_simplified.geojson   # Topology-cleaned 36-region boundaries
│   │   └── harmonization_results.json      # National & regional evaluation metrics
│   ├── src/                      # UI architecture & components
│   │   ├── GeographicObservatory.tsx       # Interactive vector map & state inspector
│   │   ├── observatoryAggregations.ts      # Client-side spatial data aggregation & cache
│   │   ├── HistoricalCalendar.tsx          # 23-year historical time series visualization
│   │   ├── HarmonizationResults.tsx        # Statistical model comparison laboratory
│   │   ├── RegionalHarmonizationResults.tsx# State-by-state model evaluations
│   │   ├── navigation.ts                   # Hash routing and canonical route normalization
│   │   └── App.tsx                         # Navigation bar, view routing, layout shell
│   └── test/                     # Automated frontend test suite (43 tests)
│       ├── navigation.test.ts              # Route parsing & fallback contracts
│       ├── observatory_aggregations.test.ts# Metric calculations & leap-year logic
│       ├── observatory_cache.test.ts       # In-memory return navigation cache tests
│       └── scientific_data_contracts.test.ts# GeoJSON schemas, sensor years, OGCD limits
├── pipeline/                     # Python data pipelines and harmonization experiments
│   ├── regional_pipeline.py      # Spatial point-in-polygon join & 5 km grid aggregation
│   ├── harmonization_experiment.py # National model training, cross-validation & testing
│   ├── regional_harmonization_experiment.py # Regional model evaluation
│   ├── historical_pipeline.py    # Builds 2003–2025 national daily historical calendar
│   └── run_all.py                # Pipeline execution coordinator
├── tests/                        # Python automated test suite (5 tests)
│   └── test_scientific_contracts.py # Python contract validation against JSON data
├── data/                         # Data documentation and source references
│   └── boundaries/               # geoBoundaries README, download links, and licensing
├── LICENSE                       # MIT License for original FireSignal source code
├── THIRD_PARTY_NOTICES.md        # Upstream provenance, boundary terms & software notices
├── METHODOLOGY.md                # Comprehensive scientific documentation and equations
└── README.md                     # Project overview, quickstart, and evaluation

Local Development Setup

Prerequisites

  • Node.js: ^20.19.0 || >=22.12.0 (verified Vite 8 engine requirement)
  • npm: v10.0.0 or higher

Note

Running and exploring the web application does not require Python or raw satellite data downloads. The frontend reads precomputed, verified JSON datasets directly from frontend/public/data/.

1. Install and Start Development Server

# Clone the repository
git clone https://github.com/pushkarscripts/firesignal.git
cd firesignal

# Navigate to frontend and install dependencies
cd frontend
npm install

# Start the Vite development server
npm run dev

Open your browser at http://localhost:5173/ (or the port indicated in terminal output).

2. Production Build and Preview

To compile the production bundle with static type checking:

npm --prefix frontend run build
npm --prefix frontend run preview

Testing & Code Quality

FireSignal includes comprehensive automated regression suites covering routing, spatial calculations, cache lifecycles, and scientific data contracts:

1. Frontend Automated Tests (43 Tests)

Executed via the native Node.js test runner:

npm --prefix frontend run test

Verifies route hash parsing, leap-year days calculation, missing vs. zero state classification, return-navigation cache idempotency, and boundary GeoJSON 1-to-1 name matching.

2. Python Scientific Contract Tests (5 Tests)

Executed via Python unittest:

python -m unittest discover -s tests

Verifies sensor year bounds (MODIS 2003–2025, VIIRS 2013–2025), EPSG:6933 5 km metadata specifications, and data integrity of exported JSON artifacts.

3. Static Analysis & Linting

Run the fast Rust-based linter oxlint:

npm --prefix frontend run lint

Reproducing the Scientific Pipeline (Optional)

Users wishing to regenerate the raw spatial aggregations from source satellite archives can execute the Python processing pipeline.

1. Python Environment Setup

Requirements: Python 3.10+ with spatial and scientific libraries.

python -m venv .venv
source .venv/bin/activate   # On Windows: .venv\Scripts\activate
pip install numpy pandas geopandas pyproj scipy shapely matplotlib

2. Download Raw Datasets

  1. Administrative Boundaries: Retrieve the India ADM1 GeoJSON from the geoBoundaries API according to instructions in data/boundaries/README.md and save to data/boundaries/india_adm1.geojson.
  2. NASA FIRMS Archives: Request and download raw historical MODIS C6.1 and VIIRS SNPP C2 active-fire archives for India from NASA FIRMS and place ZIP files in data/raw/.

3. Run Pipeline Scripts

# Generate regional 5 km aggregations (point-in-polygon join)
python pipeline/regional_pipeline.py --full --chunk-size 100000

# Execute national harmonization experiment
python pipeline/harmonization_experiment.py

# Execute regional harmonization experiment
python pipeline/regional_harmonization_experiment.py

# Generate historical 23-year calendar dataset
python pipeline/historical_pipeline.py

Scientific Limitations

  1. Not Ground Truth: MODIS is an observational reference with its own omission and commission errors, not physical ground truth.
  2. Not Physical Calibration: Statistical harmonization models are empirical transfer functions; they do not alter sensor optics, radiative transfer physics, or physical detection thresholds.
  3. Not Burned Area or Severity: OGCD measures detected hotspot spatial occupancy. It cannot be converted to burned area or wildfire severity without fuel models, fire progression analysis, and surface validation.
  4. Cloud and Smoke Obscuration: Optical sensors cannot detect active fires through dense convective cloud cover or thick smoke plumes. Zero OGCD does not prove zero fire activity.
  5. Resolution Discrepancy: The footprint area disparity between MODIS (1,000 m at nadir, 1.0 km²) and VIIRS (375 m at nadir, ~0.1406 km²; a nominal ~7.1:1 nadir area ratio), combined with differing off-nadir pixel growth rates, can enable VIIRS to detect smaller or lower-intensity agricultural fires that MODIS may not detect.
  6. Selection Bias in Regional Tests: Regional evaluations exclude months without paired observations, which may under-represent cloud-covered or low-activity months.
  7. Sparse Data Regimes: In states with infrequent fire activity (e.g., small coastal territories or high-altitude regions), statistical harmonization ratios are sensitive to small sample sizes.
  8. Historical Boundary Demarcation: Administrative boundaries reflect a 2011 Census reference snapshot from geoBoundaries/DataMeet, which may differ from current state and union territory definitions.
  9. No Independent Ground Truth Validation: The statistical models have been cross-validated across chronological satellite partitions, but have not been evaluated against independent ground-based perimeter GPS surveys.
  10. Non-Operational Scope: FireSignal is designed strictly for exploratory scientific research and retrospective climate analysis, not for operational emergency dispatch or real-time wildfire suppression.

License & Third-Party Attribution

Software License

Original FireSignal application source code, cartographic visualization logic, and Python pipelines are licensed under the MIT License (LICENSE).
Copyright (c) 2026 Pushkar.

Third-Party Data & Spatial Provenance

External datasets retain their respective upstream licenses and terms (see THIRD_PARTY_NOTICES.md):

  • Satellite Active-Fire Observations: Provided by NASA FIRMS / ESDIS Project. NASA Earth Science data is open and unrestricted; works of the US Federal Government are in the public domain domestically (17 U.S.C. § 105).
  • Administrative Boundaries: Sourced from geoBoundaries (gbOpen) under CC BY 4.0, with underlying boundary attribution to DataMeet India and the Election Commission of India (2011 reference snapshot) under CC BY 2.5 IN.
  • Simplified Vector Asset (frontend/public/data/india_adm1_simplified.geojson): An adapted derivative work simplified via Mapshaper for web cartography, distributed under CC BY 2.5 IN / CC BY 4.0 terms, and excluded from the software MIT license.
  • Third-Party Libraries: React, Recharts, Tailwind CSS, Lucide, and Vite are licensed under permissive open-source licenses (MIT/ISC).

Non-Endorsement

Reference to NASA, NASA FIRMS, geoBoundaries, DataMeet, or the Election Commission of India does not constitute or imply official sponsorship, certification, or endorsement of FireSignal.


Citation & Academic Reference

If you use FireSignal, its analytical methodology, or its harmonization findings in your research or project evaluation, please cite:

@misc{firesignal2026,
  author       = {Pushkar},
  title        = {FireSignal: Satellite Active-Fire Observatory and Cross-Sensor Harmonization for India},
  year         = {2026},
  publisher    = {GitHub},
  howpublished = {\url{https://github.com/pushkarscripts/firesignal}},
  note         = {NASA Space Apps Challenge 2026}
}

Upstream Scientific Citations

  • MODIS Active Fire Algorithm: Giglio, L., Schroeder, W., & Justice, C. O. (2016). The collection 6 MODIS active fire detection algorithm and fire products. Remote Sensing of Environment, 178, 31–41. https://doi.org/10.1016/j.rse.2016.02.054
  • VIIRS 375 m Active Fire Algorithm: Schroeder, W., Oliva, P., Giglio, L., & Csiszar, I. A. (2014). The New VIIRS 375 m active fire detection product—Algorithm description and initial assessment. Remote Sensing of Environment, 143, 85–96. https://doi.org/10.1016/j.rse.2013.12.008
  • geoBoundaries Global Database: Runfola, D., Anderson, A., Baier, H., Cauley, M., Chen, D., Diloreto, E., et al. (2020). geoBoundaries: A global database of political administrative boundaries. PLOS ONE, 15(4), e0231866. https://doi.org/10.1371/journal.pone.0231866

Author & Acknowledgments

  • Lead Developer & Researcher: Pushkar (@pushkarscripts)
  • Challenge: NASA Space Apps Challenge 2026 — Harmonization of MODIS and VIIRS Hot Spots
  • Acknowledgments: Sincere gratitude to the NASA FIRMS and ESDIS teams for open-access remote sensing archives, the William & Mary geoLab for the geoBoundaries database, and DataMeet India for open spatial data advocacy.

About

FireSignal is an interactive platform that uses NASA satellite data to explore, analyze, and compare historical fire activity across India.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages