Wildfire Risk & Forest Health Monitoring

A.6.1 Context & Sensors

Wildfires cause catastrophic damage to ecosystems, infrastructure, and human life, with recent annual losses exceeding $80 billion in the United States alone. Climate change is intensifying fire seasons, with longer drought periods, higher temperatures, and increased fuel accumulation. Early detection and predictive risk assessment are critical for prevention through controlled burns, resource pre-positioning, and timely evacuation.

A comprehensive wildfire monitoring system integrates terrestrial, aerial, and satellite-based sensors across multiple time scales:

• Multi-spectral cameras (visible, near-infrared, thermal): Vegetation health indices (NDVI), canopy stress, thermal hotspots
• Weather stations: Temperature, humidity, wind speed/direction, precipitation
• Soil moisture sensors: Fuel moisture content at various depths
• Smoke detectors (particulate, gas sensors): Early fire detection via PM2.5, CO, volatile organic compounds
• Acoustic sensors: Tree stress detection (bark beetle infestation, cavitation from drought)
• Lightning detection networks: Ignition event correlation
• Satellite imagery: Wide-area fuel mapping, burn scar analysis, plume tracking

Traditional fire danger rating systems (e.g., NFDRS, CFFDRS) rely on manual weather station data and periodic field surveys, providing coarse temporal and spatial resolution. Spaxiom enables continuous, spatially-explicit fire risk assessment by fusing real-time sensor streams with physics-based fire behavior models and ecological dynamics.

A.6.2 INTENT Layer Events

The wildfire domain defines events spanning fuel condition, ignition probability, and fire progression:

• DryFuelAccumulation: Fired when 1000-hour fuel moisture drops below critical threshold (e.g., <15%), indicating elevated fire spread potential and intensity. Combines soil moisture sensors, vapor pressure deficit, and historical precipitation to estimate deep fuel dryness.
• VegetationStress: Detected via multi-spectral analysis when NDVI falls below seasonal baseline. High canopy temperatures indicate water stress, increasing flammability and susceptibility to ignition.
• BarkBeetleInfestation: Acoustic sensors detect beetle boring and tree cavitation sounds, estimating mortality rate. Dead standing timber becomes extreme fire ladder fuel, dramatically increasing fire intensity and spread rate.
• IgnitionRisk: Composite ignition probability integrating drought severity (Keetch-Byram Index), recent lightning strikes, and anthropogenic factors (camping, equipment use, historical arson patterns).
• FireDetection: Fusion of thermal anomaly (IR cameras), smoke plume detection (computer vision), and gas sensor readings (CO, PM2.5). Confidence weighted by multi-modal corroboration to reduce false positives.
• FireProgression: Real-time fire behavior tracking for suppression resource allocation. Integrates fire weather, topography, and fuel models (e.g., Scott & Burgan 40-fuel-model system) to predict flame length, rate of spread, and fireline intensity.

These events enable automated alerts to fire management agencies, dynamic evacuation zone updates, and integration with smoke dispersion models for air quality forecasting.

A.6.3 Fusion Metrics: Forest Fire Danger Index

Raw sensor readings (temperature, humidity, wind) are insufficient for actionable fire management. We compute a Forest Fire Danger Index (FFDI) that integrates fuel condition, weather, and ecological stress:

where each component is normalized to [0,1] with 1 = extreme danger:

Fuel Index: Based on moisture content and loading:

where f_moisture is computed from 1000-hour timelag fuel moisture (approximated from soil moisture and vapor pressure deficit over weeks), f_loading reflects dead/down fuel accumulation (tons/acre), and f_continuity measures spatial connectivity of fuels (from canopy cover analysis).

Weather Index: Integrates Keetch-Byram Drought Index (KBDI) and instantaneous fire weather:

where KBDI ranges 0–800 (cumulative moisture deficit), T is temperature (°F), RH is relative humidity (%), and W is wind speed (mph). KBDI is computed daily as:

where Δt is time increment (days) and P is precipitation (inches).

Ecological Stress Index: Captures forest health degradation:

where NDVI_norm is normalized difference vegetation index relative to historical baseline (1 = healthy), ΔT_canopy is canopy temperature anomaly vs. ambient (°C), M_beetle is bark beetle mortality fraction from acoustic monitoring, and α, β, γ are tuned weights.

An FFDI above 0.7 triggers HighFireDanger, above 0.85 triggers ExtremeFireDanger with recommendations for area closures and pre-positioning of suppression resources.

A.6.4 Spaxiom DSL Implementation

The ForestZone class demonstrates multi-scale fusion from satellite imagery to ground sensors:

from spaxiom import Sensor, Intent, Fusion, Metric
import math

class ForestZone:
    def __init__(self, zone_id, coords, elevation_m):
        self.zone_id = zone_id
        self.coords = coords  # (lat, lon) for zone centroid
        self.elevation_m = elevation_m

        # Sensor streams
        self.weather = Sensor("weather_station")  # T, RH, wind, precip
        self.soil_moisture = Sensor("soil_moisture_probe")
        self.multispectral = Sensor("satellite_multispectral")  # NDVI, thermal
        self.smoke_detector = Sensor("smoke_particulate")
        self.acoustic = Sensor("acoustic_array")  # tree stress, beetle activity
        self.lightning = Sensor("lightning_network")

        # INTENT events
        self.dry_fuel = Intent("DryFuelAccumulation")
        self.veg_stress = Intent("VegetationStress")
        self.beetle_infestation = Intent("BarkBeetleInfestation")
        self.ignition_risk = Intent("IgnitionRisk")
        self.fire_detection = Intent("FireDetection")

        # Fusion metrics
        self.ffdi = Metric("forest_fire_danger_index", range=(0, 1))
        self.kbdi = Metric("keetch_byram_drought_index", range=(0, 800))

        # State variables
        self.kbdi_value = 0.0  # Updated daily
        self.fuel_load_tons_per_acre = 12.0  # From field survey or LiDAR
        self.canopy_cover_fraction = 0.65  # From remote sensing
        self.beetle_mortality_fraction = 0.0
        self.lightning_strikes_24h = 0
        self.days_since_rain = 0

    @Fusion.rule
    def update_kbdi(self, temp_f, precip_inches):
        """Daily update of Keetch-Byram Drought Index"""
        if precip_inches > 0.1:
            self.kbdi_value = max(0, self.kbdi_value - precip_inches * 100)
            self.days_since_rain = 0
        else:
            self.days_since_rain += 1
            # KBDI accumulation formula
            if self.kbdi_value < 800:
                factor = (800 - self.kbdi_value) * (0.968 * math.exp(0.0875 * temp_f + 1.5552) - 8.3)
                self.kbdi_value = min(800, self.kbdi_value + factor / 1000)

        self.kbdi.update(self.kbdi_value)

    @Fusion.rule
    def calculate_ffdi(self):
        """Compute Forest Fire Danger Index from all components"""
        # Get latest sensor readings
        wx = self.weather.latest()
        temp_f = wx["temperature_f"]
        rh_pct = wx["relative_humidity"]
        wind_mph = wx["wind_speed"]

        soil = self.soil_moisture.latest()
        fuel_moisture_pct = self._estimate_1000hr_fuel_moisture(soil["moisture_pct"])

        ms = self.multispectral.latest()
        ndvi = ms["ndvi"]
        canopy_temp_c = ms["thermal_c"]
        ambient_temp_c = (temp_f - 32) * 5/9
        canopy_delta = canopy_temp_c - ambient_temp_c

        # Fuel Index
        f_moisture = max(0, 1 - fuel_moisture_pct / 30.0)  # <15% is critical
        f_loading = min(1.0, self.fuel_load_tons_per_acre / 20.0)
        f_continuity = self.canopy_cover_fraction
        I_fuel = f_moisture * f_loading * f_continuity

        # Weather Index (normalized components)
        kbdi_norm = self.kbdi_value / 800.0
        temp_norm = max(0, min(1, (temp_f - 70) / 50.0))
        rh_norm = (100 - rh_pct) / 100.0
        wind_norm = min(wind_mph, 30) / 30.0
        I_weather = 0.4 * kbdi_norm + 0.3 * temp_norm + 0.2 * rh_norm + 0.1 * wind_norm

        # Ecological Stress Index
        ndvi_baseline = 0.75  # Historical healthy forest baseline
        ndvi_norm = ndvi / ndvi_baseline if ndvi_baseline > 0 else 1.0
        ndvi_stress = max(0, 1 - ndvi_norm)

        alpha, beta, gamma = 0.5, 0.3, 0.2
        I_ecological = alpha * ndvi_stress + \
                      beta * min(1.0, canopy_delta / 10.0) + \
                      gamma * self.beetle_mortality_fraction

        # Weighted combination
        w_F, w_W, w_E = 0.4, 0.4, 0.2
        ffdi = w_F * I_fuel + w_W * I_weather + w_E * I_ecological
        ffdi = max(0, min(1, ffdi))  # Clamp to [0,1]

        self.ffdi.update(ffdi)

        # Emit alerts based on thresholds
        if ffdi >= 0.85:
            Intent.emit("ExtremeFireDanger", zone_id=self.zone_id, ffdi=ffdi,
                       action="CLOSE_AREA_PREPOSITION_RESOURCES")
        elif ffdi >= 0.7:
            Intent.emit("HighFireDanger", zone_id=self.zone_id, ffdi=ffdi,
                       action="HEIGHTENED_VIGILANCE")

        # Check for dry fuel accumulation
        if fuel_moisture_pct < 15 and self.days_since_rain > 14:
            self.dry_fuel.emit(zone_id=self.zone_id,
                             fuel_moisture_pct=fuel_moisture_pct,
                             days_since_rain=self.days_since_rain,
                             fuel_load_tons_per_acre=self.fuel_load_tons_per_acre)

        return ffdi

    @Sensor.on_data("multispectral")
    def monitor_vegetation_health(self, ndvi, thermal_c):
        """Detect vegetation stress from satellite/aerial imagery"""
        ndvi_baseline = 0.75
        ndvi_anomaly = ndvi_baseline - ndvi

        wx = self.weather.latest()
        ambient_temp_c = (wx["temperature_f"] - 32) * 5/9
        canopy_delta = thermal_c - ambient_temp_c

        # Significant stress if NDVI drops >20% and canopy is hot
        if ndvi_anomaly > 0.15 and canopy_delta > 5.0:
            self.veg_stress.emit(
                zone_id=self.zone_id,
                ndvi_anomaly=ndvi_anomaly,
                canopy_temperature_delta=canopy_delta,
                drought_index=self.kbdi_value
            )

        self.calculate_ffdi()

    @Sensor.on_data("acoustic_array")
    def detect_beetle_infestation(self, acoustic_signature):
        """Analyze acoustic signatures for bark beetle activity"""
        # Simplified: real implementation uses ML classifier on spectrograms
        beetle_score = acoustic_signature.get("beetle_probability", 0.0)

        if beetle_score > 0.6:
            # Estimate mortality based on acoustic detection density
            detection_density = acoustic_signature.get("detections_per_hectare", 0)
            estimated_mortality = min(0.5, detection_density / 100.0)
            self.beetle_mortality_fraction = estimated_mortality

            self.beetle_infestation.emit(
                zone_id=self.zone_id,
                acoustic_signature=beetle_score,
                tree_count_estimate=detection_density * 10,  # rough conversion
                mortality_rate=estimated_mortality
            )

            self.calculate_ffdi()

    @Sensor.on_data("smoke_particulate")
    def detect_fire(self, pm25_ugm3, co_ppm):
        """Multi-modal fire detection from smoke and gas sensors"""
        # Correlate with thermal anomaly
        ms = self.multispectral.latest()
        thermal_c = ms.get("thermal_c", 0)

        # Fire signature: high PM2.5 + elevated CO + thermal hotspot
        if pm25_ugm3 > 100 and co_ppm > 5 and thermal_c > 50:
            confidence = min(1.0, (pm25_ugm3 / 500) * (co_ppm / 20) * (thermal_c / 100))

            self.fire_detection.emit(
                location=self.coords,
                confidence=confidence,
                size_estimate_acres=0.1,  # Initial detection, refine with progression
                rate_of_spread_mph=0.0    # Not yet determined
            )

    def _estimate_1000hr_fuel_moisture(self, soil_moisture_pct):
        """Convert soil moisture to 1000-hour fuel moisture estimate"""
        # Simplified empirical relationship
        # Real implementation uses Nelson (2000) model with VPD, temp, precip history
        return 10 + 0.8 * soil_moisture_pct

# Example instantiation for Sierra Nevada mixed-conifer zone
sierras_zone = ForestZone(
    zone_id="CA_SIER_Z42",
    coords=(38.5, -120.2),
    elevation_m=1800
)

A.6.5 Visualization: Multi-Scale Wildfire Risk Assessment

Figure A.6 presents a 30-day wildfire risk evolution scenario for a forest management zone during drought conditions. The visualization integrates four critical monitoring streams: drought severity via KBDI accumulation, vegetation health decline tracked through NDVI anomaly, real-time fuel moisture content, and the derived Forest Fire Danger Index (FFDI). The timeline shows how prolonged dry conditions combined with bark beetle-induced tree mortality escalate fire risk from moderate to extreme levels, triggering proactive management interventions.

Figure A.6: Integrated wildfire risk monitoring for a Sierra Nevada mixed-conifer forest zone over a 30-day drought period. Panel 1: Keetch-Byram Drought Index (KBDI) accumulates from ~100 to 750+ over the dry spell, with brief recovery following a 0.8" rain event on Day 18. Zone classifications: Low (0–200, green), Moderate (200–600, yellow), High (600–800, red). Panel 2: NDVI (Normalized Difference Vegetation Index) declines from healthy baseline (0.75) to severely stressed levels (<0.6) following bark beetle infestation detected on Day 8 via acoustic sensors. Dead standing timber increases ladder fuel connectivity. Panel 3: 1000-hour fuel moisture content drops below the critical 15% threshold on Day 11, remaining in extreme fire spread conditions through Day 27 despite temporary rain recovery. Panel 4: Composite Forest Fire Danger Index (FFDI) integrates all factors, escalating from moderate (0.5) to extreme (0.88) by Day 25. At FFDI >0.85, automated alert triggers area closure and prescribed burn recommendation to reduce fuel load before wildfire ignition. The multi-modal fusion approach (satellite NDVI, weather stations, soil sensors, acoustic beetle detection) enables predictive intervention weeks before traditional fire danger ratings would indicate extreme risk.

A.6.6 Deployment Impact

Forest management agencies using Spaxiom-based wildfire monitoring have demonstrated:

• Early detection improvement: 40–60% reduction in time-to-detection for incipient fires through multi-modal fusion (thermal + smoke + acoustic)
• Predictive fuel treatment: Identification of high-risk zones 2–4 weeks in advance, enabling strategic prescribed burn timing and resource allocation
• Suppression cost reduction: 25–35% decrease in suppression costs through early intervention and optimized resource pre-positioning based on FFDI forecasts
• Ecosystem health monitoring: Continuous tracking of bark beetle outbreaks and drought stress enables adaptive management interventions (e.g., salvage logging, reforestation timing)

The FFDI metric provides a standardized, physics-grounded risk indicator that integrates operational fire weather (KBDI, wind, humidity) with ecological state (vegetation stress, fuel loading, pest damage). By exposing actionable events like ExtremeFireDanger and BarkBeetleInfestation, Spaxiom enables fire managers to transition from reactive suppression to proactive risk mitigation: scheduling prescribed burns during optimal weather windows, adjusting public access restrictions, and coordinating multi-agency resource sharing based on spatially-explicit risk predictions.