soumyasrikesani/africa-wildfire-viirs
Domain:
environment and energygeospatialRecord type:
project
Creator:
sou
Host:
This is a case study on wild fires in Africa used to know the importance or unnoticed Wildfire encroachments in the Lowlands and ecoregion in African Forests
# Wildfire Encroachment into African Moist Tropical Forests
**DSA 5900 Professional Practicum | University of Oklahoma | Spring 2026**
**Team:** Soumya Sri Kesani, Poorna Chandan Reddy Pandem
**Sponsor:** Dr. Michael Wimberly and Dr. Gopichandh Danala, Data Institute for Societal Challenges (DISC)
**Supervisor:** Dr. Matt J. Beattie
---
## Overview
Wildfires burning beneath the closed canopy of West Africa's moist tropical forests are largely invisible to standard global monitoring tools. This project builds a spatiotemporal detection and mapping pipeline to close that discovery gap across the **East-West Guinean Lowland Forests** ecoregion.
Using NASA VIIRS SNPP 375m active fire detections (2012-2024), we engineer a graph-based clustering engine that groups raw thermal hotspots into contiguous fire events. These events are converted into burned area polygons and interpolated onto 375m raster surfaces of burn date and Fire Radiative Power (FRP). The pipeline is benchmarked against MODIS MCD64A1 and validated against Landsat-derived dNBR masks from Google Earth Engine.
---
## Key Results
| Metric | Value |
|--------|-------|
| Study Period | 2012-2024 |
| Total Detections Processed | ~1.85 million |
| Optimal Config | D = 2000m, T = 2 days, NH confidence |
| Mean Kappa (6 validation sites) | 0.588 |
| Mean Precision | 0.856 |
| Discovery Gain over MODIS | 14.1% (128,067 Ha) |
| Active Encroachment Frontiers | Tinte Bepo (p=0.039), Opro River (p=0.047) |
| Peak Fire Year | 2015 |
| Lowest Fire Year | 2021 |
---
## Repository Structure
```
africa-wildfire-viirs/
│
├── process_africa_vaf.R # Data ingestion from NASA FIRMS
├── Read_africa_vaf.R # Data reading and rasterization
├── viirs_fire_interpolation.R # Core clustering and interpolation pipeline
├── MK.R # Mann-Kendall trend analysis
├── seasonality.R # STL decomposition and seasonal analysis
├── burnvalidation.R # dNBR-based …