MTG FRP Fire Progression Analyzer with Intensity Estimation

December 15, 2025 · View on GitHub

A Python script for creating hourly and cumulative fire progression polygons from MTG (Meteosat Third Generation) FRP (Fire Radiative Power) data, with calibration against reference burned areas and fire intensity estimation.

https://github.com/user-attachments/assets/198a031a-720d-4a23-9ee9-2cea6e173c95

  • Fire Intensity based on Radiative Method: I_rad = FRP / (L × X_r)

Fire_Intensity

  • Fire Intensity based on Traditional Method: I_trad = H × w × v

Fire_Intensity_Byram

📋 Description

This script processes MTFRPPixel fire detection data to create temporal progression polygons of wildfires. It supports both cumulative and hourly progression modes, with advanced filtering, calibration capabilities, and now includes fire intensity estimation using radiative fraction approach.

✨ Features

  • Dual Processing Modes: Cumulative progression and hourly progression for analysis and animations
  • Non-Overlapping Progression: Option for print layouts showing only new burned areas each hour
  • Advanced Filtering: FRP thresholding, spatial density filtering (DBSCAN), cluster size filtering and minimum area filtering
  • Flexible Hull Generation: Concave hull with fallback to convex hull
  • Calibration System: Automatic parameter calibration using reference burned areas
  • Temporal Continuity: Fill missing hours to maintain temporal (hourly) progression
  • Area Correction: Negative buffer (shrink) to mitigate area overestimation
  • Multiple Outputs: Cumulative and hourly polygons, daily summaries, calibrated versions and non-overlapping progression
  • NEW: Byram Fire Intensity Calculation: Two methods (radiative and traditional) with validation
  • NEW: Propagation Speed Analysis: 90th percentile distance for robust speed estimation
  • NEW: Fire Front Length Estimation: MTG-resolution corrected length calculation
  • NEW: Radiative Efficiency: Cross-validation between intensity methods

🛠️ Installation

Prerequisites

# Install Python dependencies (shapely 2.x)
pip install pandas geopandas shapely scikit-learn numpy

Download Script

git clone https://github.com/PedroVenancio/mtg-frp-fire-progression.git
cd mtg-fire-progression

🚀 Usage

Basic Examples

Process fire progression with default parameters:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix frp_fires

Process with calibration using reference burned areas:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix calibrated_frp_fires --reference_areas reference_burned_areas.gpkg

Process only cumulative progression:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix cumulative_only --no_hourly

Process without non-overlapping progression:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix no_overlap --no_non_overlapping

Adaptive FRP filtering (keep only top 85% of FRP values):

python mtg_frp_fire_progression.py --input fire_data.gpkg --frp_threshold_method adaptive --frp_quantile_threshold 0.15

Fixed FRP threshold (keep only FRP ≥ 20 MW):

python mtg_frp_fire_progression.py --input fire_data.gpkg --frp_threshold_method fixed --min_frp 20

Custom FRP filtering and clustering:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix custom --frp_threshold_method fixed --min_frp 15 --min_cluster_size 5 --density_eps 500

Advanced Examples with Intensity Calculation

Calculate Byram fire intensity with default parameters:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix fire_intensity --calculate_intensity

Calculate intensity with custom radiative fraction (0.17), mtg_correction (0.65) and fuel type (forest):

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix custom_intensity \
    --calculate_intensity \
    --radiative_fraction 0.17 \
    --fuel_type forest \
    --mtg_correction 0.65

Custom minimum area non-overlapping (1 ha) with intensity calculation:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix intensity_print \
    --min_area_non_overlapping 1.0 \
    --calculate_intensity \
    --radiative_fraction 0.16

Complete workflow with calibration and intensity calculation:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix complete_analysis \
    --reference_areas reference_burned_areas.gpkg \
    --min_area_non_overlapping 0.5 \
    --calculate_intensity \
    --radiative_fraction 0.15 \
    --heat_value 18000 \
    --fuel_type shrub

Advanced Examples

Fine-tune hull generation and area correction:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix tuned \
    --ratio 0.05 \
    --buffer_distance 100 \
    --shrink_distance 40 \
    --frp_threshold_method adaptive \
    --frp_quantile_threshold 0.2

Complete workflow with calibration and custom non-overlapping area:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix complete \
    --reference_areas reference_burned_areas.gpkg \
    --min_area_non_overlapping 1.0 \
    --buffer_distance 80 \
    --shrink_distance 30

Skip calibration even with reference data:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix no_calib \
    --reference_areas reference.gpkg \
    --skip_calibration

Process with start hour instead of end hour:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix start_hour --use_start_hour

Disable missing hour filling:

python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix no_fill --no_fill_missing

📊 Parameters

Required Parameters

  • --input: Input FRP data file (GeoPackage). Expects a GeoPackage generated by MTG/MTFRPPixel Data Processor
  • --output_prefix: Prefix for all output files

Core Processing Parameters

  • --fire_id_column: Column identifying each fire (default: 'fire_id')
  • --buffer_distance: Buffer distance in meters (default: 80)
  • --ratio: Concave hull ratio parameter 0-1 (default: 0.08)
  • --min_cluster_size: Minimum cluster size for polygons (default: 3)
  • --density_eps: DBSCAN epsilon distance in meters (default: 300)
  • --shrink_distance: Negative buffer distance in meters (default: 30)

FRP Filtering Parameters:

  • --frp_threshold_method: FRP threshold method: 'fixed' or 'adaptive' (default: adaptive)
  • --min_frp: Minimum FRP value for filtering (default: 10)
  • --frp_quantile_threshold: Quantile for adaptive FRP threshold (default: 0.15)

FRP filtering uses mutually exclusive parameters depending on the method:

  • When using --frp_threshold_method adaptive (default):
    • Uses only --frp_quantile_threshold (default: 0.15 = 15th percentile)
    • Ignores --min_frp completely
    • Example: --frp_threshold_method adaptive --frp_quantile_threshold 0.2 (keep top 80% of FRP values)
  • When using --frp_threshold_method fixed:
    • Uses only --min_frp (default: 10.0 MW)
    • Ignores --frp_quantile_threshold completely
    • Example: --frp_threshold_method fixed --min_frp 15 (keep FRP ≥ 15 MW)

Non-Overlapping Progression Parameters

  • --no_non_overlapping: Disable non-overlapping progression (enabled by default)
  • --min_area_non_overlapping: Minimum area in hectares for non-overlapping polygons (default: 0.5)

Byram Intensity Parameters (NEW)

  • --calculate_intensity: Calculate Byram fire intensity (propagation speed and FRP in new area)
  • --radiative_fraction: Radiative fraction for Byram intensity calculation. Xr = 0.15-0.20 for wildfires (based on Wooster et al., 2005; Johnston et al., 2017) (default: 0.15)
  • --heat_value: Heat content in kJ/kg for traditional Byram intensity calculation (default: 20000, average for portuguese fuel models (Fernandes et al.). Typical range 15000-21000 kJ/kg)
  • --fuel_consumption: Fixed fuel consumption in kg/m². If None, uses speed-adjusted values (default: None)
  • --fuel_type: Fuel type for continuous consumption model: 'grass', 'shrub', or 'forest' (default: shrub)
  • --mtg_correction: Correction factor for MTG 1km resolution artifacts. Typical values: 0.6-0.7. Lower values reduce L more aggressively (default: 0.6)

Processing Flags

  • --no_cumulative: Disable cumulative progression
  • --no_hourly: Disable hourly progression
  • --no_fill_missing: Disable filling missing hours
  • --no_density_filter: Disable density filtering
  • --use_start_hour: Use start hour instead of end hour

Calibration Parameters

  • --reference_areas: Reference burned areas file for calibration
  • --skip_calibration: Skip calibration even if reference areas provided

🗂️ Output Structure

Generated Files

  • {prefix}_cumulative_initial.gpkg - Initial cumulative progression
  • {prefix}_hourly_initial.gpkg - Initial hourly progression
  • {prefix}_cumulative_calibrated.gpkg - Calibrated cumulative progression (if calibration applied)
  • {prefix}_hourly_calibrated.gpkg - Calibrated hourly progression (if calibration applied)
  • {prefix}_daily_cumulative_initial.gpkg - Daily summary of cumulative progression
  • {prefix}_daily_hourly_initial.gpkg - Daily summary of hourly progression
  • {prefix}_daily_cumulative_calibrated.gpkg - Calibrated daily summary (if calibration applied)
  • {prefix}_daily_hourly_calibrated.gpkg - Calibrated daily summary (if calibration applied)

Non-Overlapping Progression Files

  • {prefix}_non_overlapping_hourly_initial.gpkg - Hourly non-overlapping progression (initial)
  • {prefix}_non_overlapping_hourly_calibrated.gpkg - Hourly non-overlapping progression (calibrated)
  • {prefix}_daily_non_overlapping_initial.gpkg - Daily non-overlapping summary (initial)
  • {prefix}_daily_non_overlapping_calibrated.gpkg - Daily non-overlapping summary (calibrated)

Byram Intensity Files (NEW)

  • {prefix}_byram_improved_initial.gpkg - Cumulative progression with fire behavior and Byram intensity columns (initial)
  • {prefix}_byram_improved_calibrated.gpkg - Cumulative progression with fire behavior and Byram intensity columns (calibrated)
  • {prefix}_fire_metrics_improved_initial.csv - Detailed fire metrics CSV (initial)
  • {prefix}_fire_metrics_improved_calibrated.csv - Detailed fire metrics CSV (calibrated)
  • {prefix}_propagation_vectors_90th_percentile.gpkg - Propagation points using 90th percentile distance
  • {prefix}_propagation_vectors_max.gpkg - Propagation points using maximum distance
  • {prefix}_non_overlapping_hourly_byram_improved_initial.gpkg - Non-overlapping progression with intensity (initial)
  • {prefix}_non_overlapping_hourly_byram_improved_calibrated.gpkg - Non-overlapping progression with intensity (calibrated)

Data Structure

Files contain:

  • fire_id: Fire identifier
  • datetime_hour: Hour in UTC
  • datetime_display: Display hour (end hour by default)
  • geometry: Progression polygon
  • area_ha: Area in hectares
  • n_points_current_hour: Points in current hour
  • n_points_cumulative: Cumulative points (cumulative version)
  • frp_current_hour: FRP in current hour
  • frp_cumulative: Cumulative FRP (cumulative version)
  • hull_type: Type of hull used ('concave', 'convex', etc.)
  • data_status: Data origin ('original', 'filled')
  • shrink_applied: Whether negative buffer was applied
  • shrink_distance_m: Shrink distance in meters

Non-Overlapping Specific Fields

  • area_ha_new: New area burned in this time step (hectares)
  • area_ha_cumulative: Total cumulative area up to this time step
  • progression_type: 'non_overlapping' or 'daily_non_overlapping'

NEW: Byram Intensity Fields

  • propagation_distance_m: 90th percentile propagation distance (meters) - robust metric
  • propagation_distance_max_m: Maximum propagation distance (meters) - absolute maximum
  • propagation_distance_reduction_pct: Reduction percentage (90th vs max) - data quality indicator
  • propagation_speed_ms: Propagation speed (rate of spread) (m/s)
  • propagation_speed_kmh: Propagation speed (rate of spread) (km/h)
  • fire_front_length_m: Fire front length with MTG correction (meters)
  • frp_new_area_mw: FRP in new burned area only (MW)
  • n_points_new_area: Number of FRP points in new area
  • heat_value_kj_kg: Heat content used in calculation (from --heat_value parameter)
  • byram_radiative_intensity_kw_m: Radiative Byram intensity (kW/m) - I = FRP/(L×Xr)
  • byram_traditional_intensity_kw_m: Traditional Byram intensity (kW/m) - I = H×w×v
  • radiative_efficiency: Ratio radiative/traditional intensity (validation metric)
  • implied_radiative_fraction: Implied Xr that would equalize both methods
  • intensity_class: Radiative intensity classification
  • traditional_intensity_class: Traditional intensity classification
  • adjusted_fuel_consumption_kg_m2: Speed-adjusted fuel consumption

🔧 Processing Details

Algorithm Overview

  1. Data Filtering:

    • FRP thresholding (fixed or adaptive)
    • Spatial density filtering (DBSCAN)
    • Cluster size filtering

  2. Temporal Processing:

    • UTC timezone enforcement
    • Hourly binning
    • Missing hour filling (optional)

  3. Polygon Generation:

    • Concave hull with progressive ratio
    • Convex hull fallback
    • Positive buffer for smoothing
    • Negative buffer for area correction

  4. Propagation Analysis (NEW):

    • Calculate 90th percentile propagation distance (robust)
    • Estimate fire front length with MTG correction
    • Calculate propagation speed (rate of spread) (m/s and km/h)
    • Generate propagation vectors for visualization

  5. Intensity Calculation (NEW):

    • Radiative Method: I_rad = FRP / (L × X_r)
    • Traditional Method: I_trad = H × w × v
    • Cross-validation through radiative efficiency
    • Intensity classification (Low to Extreme)

  6. Non-Overlapping Progression:

    • Calculate geometric difference between consecutive time steps
    • Show only new burned areas for each hour
    • Apply minimum area filter (default: 0.5 ha)
    • Create daily summaries of new burned areas

  7. Calibration:

    • Compare with reference burned areas
    • Calculate overestimation ratios
    • Suggest optimal shrink distance

Byram Intensity Calculation Details

The intensity calculation system implements:

Radiative Method (based on Wooster et al., 2005; Johnston et al., 2017):

I_rad = FRP / (L × X_r)

Where:

  • I_rad = Radiative Byram intensity (kW/m)
  • FRP = Fire Radiative Power in new area (converted to kW)
  • L = Fire front length (m) - MTG corrected
  • X_r = Radiative fraction (default: 0.15, range: 0.15-0.20)

Traditional Method (Byram, 1959):

I_trad = H × w × v

Where:

  • I_trad = Traditional Byram intensity (kW/m)
  • H = Heat content (20000 kJ/kg for Mediterranean fuels)
  • w = Fuel consumption (kg/m²) - speed-adjusted or fixed
  • v = Propagation speed (m/s)

Key Innovations:

  • 90th percentile distance: Uses 90th percentile instead of maximum for robust propagation
  • MTG correction: Applies 0.6 factor to fire front length for 1km resolution artifacts
  • Speed-adjusted fuel consumption: Fuel consumption varies with propagation speed
  • Validation metrics: Radiative efficiency and implied X_r for cross-validation

Calibration Process

The calibration system:

  • Matches fires between FRP data and reference areas
  • Calculates overestimation percentages
  • Suggests optimal shrink_distance parameter
  • Heuristic: 10 meters shrink per 10% overestimation
  • Limits suggested shrink between 20-80 meters

Non-Overlapping Progression (ENABLED BY DEFAULT)

The non-overlapping progression is now enabled by default because it provides cleaner visualization for most use cases:

  • Shows only the new area burned in each time step
  • Ideal for print layouts and progression visualization
  • Filters out small polygons (< 0.5 ha by default)
  • Maintains temporal sequence without overlaps
  • Provides cleaner visualization of fire spread patterns

To disable this feature, use: --no_non_overlapping

💡 Usage Tips

For Best Results

Data Preparation:

  • Get FRP data from MTG/MTFRPPixel Data Processor
  • Ensure consistent fire_id between FRP and reference data
  • Use projected CRS (meters) for accurate distance calculations
  • Clean invalid geometries before processing

Parameter Tuning:

  • Start with ratio=0.08 (for concave hull) and adjust based on point density
  • Use adaptive FRP threshold for datasets with varying fire intensities
  • Set min_cluster_size=2 for detecting small fires
  • Increase density_eps for more dispersed fire patterns
  • Use min_area_non_overlapping=1.0 for cleaner print layouts

Intensity Calculation:

  • Use --calculate_intensity for fire behavior analysis
  • Adjust --mtg_correction based on landscape complexity (0.5-0.7)
  • Monitor propagation_distance_reduction_pct for data quality
  • Use --fuel_type appropriate for your vegetation (grass, shrub or forest)
  • Validate with radiative_efficiency (should be 0.1-0.3)

Memory Management:

  • Process large datasets in batches by fire_id
  • Use --no_hourly or --no_cumulative to reduce output size
  • Monitor RAM usage with very large input files

QGIS Integration

  1. Visualization:

    • Style by datetime_hour or datetime_display for temporal visualization and animation
    • Use graduated symbols for area_ha, byram_radiative_intensity_kw_m, byram_traditional_intensity_kw_m
    • Filter by data_status to identify filled hours

  2. Animation:

    • Use cumulative version with datetime_hour or datetime_display for temporal animation using QGIS Temporal Controller Panel
    • Daily summaries provide cleaner animation frames for large time frames

  3. Print Layouts:

    • Use non-overlapping progression for clear progression maps
    • Each time step shows only new burned area
    • Perfect for static maps showing fire spread sequence
    • Color by time steps for hourly / daily progression
    • Color by intensity class for behavior analysis [Low (<500 kW/m); Moderate (500-2000 kW/m); 'High (2000-4000 kW/m)'; 'Very High (4000-10.000 kW/m)'; 'Extreme (>10.000 kW/m)']

  4. Analysis:

    • Hourly version for detailed temporal analysis
    • Compare calibrated vs initial versions for accuracy assessment
    • Intersect with reference data to get better statistics and make it closer to the reality
    • Use intensity metrics for fire behavior studies

🐛 Troubleshooting

Common Issues

No polygons created:

  • Check FRP filtering thresholds
  • Verify minimum cluster size
  • Ensure valid fire_id values (not -1 or NaN)

Overestimation issues:

  • Increase shrink_distance
  • Use calibration with reference areas
  • Adjust ratio parameter for tighter hulls

Memory errors:

  • Process fires individually
  • Increase min_cluster_size to reduce polygon count
  • Use --no_fill_missing to reduce output size

Temporal gaps:

  • Enable fill_missing_hours (default)
  • Check input data temporal continuity
  • Verify timezone handling

Too many small polygons in non-overlapping output:

  • Increase min_area_non_overlapping (try 1.0 or 2.0 ha)
  • Check if shrink_distance is appropriate
  • Verify FRP filtering thresholds

Intensity calculation issues:

  • Check propagation_distance_reduction_pct - values > 50% indicate data issues
  • Adjust mtg_correction if fire front lengths seem unrealistic
  • Verify if radiative_efficiency is in plausible range
  • Check if fuel type matches the burned vegetation

Performance Optimization

For large datasets:

# Process with stricter filtering
python mtg_frp_fire_progression.py --input large_data.gpkg --output_prefix optimized \
    --min_cluster_size 5 \
    --min_frp 20 \
    --no_hourly

For high precision:

# Use tighter parameters and calibration
python mtg_frp_fire_progression.py --input precise_data.gpkg --output_prefix precise \
    --ratio 0.05 \
    --buffer_distance 50 \
    --reference_areas high_quality_reference.gpkg

For print-ready progression maps:

# Generate clean non-overlapping progression
python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix print_ready \
    --min_area_non_overlapping 1.0 \
    --buffer_distance 80 \
    --shrink_distance 40

For detailed fire behavior analysis:

# Generate complete intensity analysis
python mtg_frp_fire_progression.py --input fire_data.gpkg --output_prefix behavior \
    --reference_areas reference_burned_areas.gpkg \
    --calculate_intensity \
    --radiative_fraction 0.16 \
    --fuel_type forest \
    --mtg_correction 0.65 \
    --heat_value 19000

📝 Complete Workflow Example

# 1. Process with calibration, non-overlapping, and intensity calculation
python mtg_frp_fire_progression.py --input mtg_frp_data.gpkg --output_prefix fire_analysis \
    --reference_areas sentinel2_burned_areas.gpkg \
    --calculate_intensity \
    --radiative_fraction 0.15 \
    --heat_value 18000 \
    --fuel_type shrub \
    --mtg_correction 0.6

# 2. Analyze results in QGIS
# Files: fire_analysis_*.gpkg and fire_analysis_*.csv

# 3. Analyze results in DBeaver / Calc
# Files: fire_analysis_*.gpkg and fire_analysis_*.csv
# Create graphs with hourly FRP and compare with fire weather indices and other metrics

# 4. Compare intensity methods in QGIS
# Open fire_analysis_non_overlapping_hourly_byram_improved_calibrated.gpkg
# Compare byram_radiative_intensity_kw_m vs byram_traditional_intensity_kw_m
# Check radiative_efficiency for validation
# Style by propagation_speed_kmh for rate of spread visualization

# 5. Create temporal animation in QGIS
# Use _cumulative_calibrated outputs with datetime_hour for animation

# 6. Create print layouts in QGIS
# Use non_overlapping_hourly_byram_improved_calibrated for hourly / daily progression and intensity maps

🔒 Data Requirements

Input Data (MTG/MTFRPPixel Data Processor)

  • FRP data with geometry and fire_id column
  • Valid datetime information (acquisition_datetime or ACQTIME)
  • FRP values for filtering

Reference Data (for calibration)

  • Burned area polygons with matching fire_id
  • Same spatial extent as FRP data
  • Preferably from high-resolution sources (Sentinel-2, Landsat)

📄 License

This project is distributed under the MIT License. See LICENSE file for details.

🙋 Support

For issues and questions:

  1. Check troubleshooting section
  2. Create GitHub issue
  3. Contact project maintainer

📚 References

Core Algorithms

Intensity Calculation

GIS Integration

Notes:

  • This tool is designed to work with MTG FRP data, but can be adapted for other fire detection datasets with similar structure.
  • Intensity calculation methods are based on literature but have several assumptions that need to be validated for different fire regimes. Byram Intensity Parameters can be used to fine tune the outputs, using field observations or better fire behaviour data.
  • Fire intensity values should be considered initial estimations to identify different stages of fires during their lifespan and to compare the fire behaviour of diferent wildfires.