Lab 30 - Wildfire Analysis | Remote Sensing Course

What You'll Learn

Calculate NBR from Landsat NIR and SWIR bands
Compute dNBR (delta NBR) from pre- and post-fire imagery
Classify burn severity using USGS thresholds
Summarize burned area by severity class within fire perimeters
Export burn severity maps and statistics

Why This Matters

Burn severity mapping supports wildfire management:

Recovery planning: High-severity areas need different restoration strategies
Erosion risk: Burned areas are vulnerable to debris flows and floods
Carbon accounting: Fires release stored carbon to the atmosphere
Insurance/damage assessment: Quantify impacts for recovery funding

Before You Start

Prerequisites: Understand spectral indices and image differencing.
Estimated time: 75 minutes
Materials: Earth Engine account and a known fire event with perimeter data.

Key Terms

NBR (Normalized Burn Ratio): A spectral index using NIR and SWIR bands that is sensitive to vegetation moisture and burn severity: (NIR - SWIR2) / (NIR + SWIR2)
dNBR (delta NBR): The difference between pre-fire and post-fire NBR. Positive values indicate burn severity.
Burn Severity: The degree to which fire has altered an ecosystem, from unburned to high severity.
MTBS: Monitoring Trends in Burn Severity - a USGS/USFS program providing burn severity data for the US.

USGS Burn Severity Thresholds

Class	dNBR Range	Description
0	< 0.1	Unburned/Low regrowth
1	0.1 - 0.27	Low severity
2	0.27 - 0.44	Moderate-low severity
3	0.44 - 0.66	Moderate-high severity
4	> 0.66	High severity

Step 1: Define AOI and Date Ranges

Set up your study area and pre/post fire dates:

// AOI and dates (adjust for your fire)
var aoi = ee.Geometry.Rectangle([-118.2, 34.1, -117.5, 34.4]);
Map.centerObject(aoi, 10);

// Pre-fire: before ignition
var preStart = '2020-07-01';
var preEnd   = '2020-08-31';

// Post-fire: after containment, before vegetation recovery
var postStart = '2020-10-01';
var postEnd   = '2020-11-30';

Step 2: Calculate NBR

Create a function to mask clouds and calculate NBR:

// Mask clouds and shadows using QA_PIXEL
function maskL2(img) {
    var qa = img.select('QA_PIXEL');
    var cloud = qa.bitwiseAnd(1 << 3).neq(0);
    var shadow = qa.bitwiseAnd(1 << 4).neq(0);
    var mask = cloud.or(shadow).not();
    return img.updateMask(mask);
}

// Calculate NBR from Landsat
function prep(img){
    var sr = maskL2(img);
    var nir = sr.select('SR_B5').multiply(0.0000275).add(-0.2);
    var sw2 = sr.select('SR_B7').multiply(0.0000275).add(-0.2);
    var nbr = nir.subtract(sw2).divide(nir.add(sw2)).rename('NBR');
    return nbr.copyProperties(img, img.propertyNames());
}

// Load Landsat 8 and create pre/post NBR composites
var L = ee.ImageCollection('LANDSAT/LC08/C02/T1_L2').filterBounds(aoi);
var pre  = L.filterDate(preStart, preEnd).map(prep).median().clip(aoi);
var post = L.filterDate(postStart, postEnd).map(prep).median().clip(aoi);

Map.addLayer(pre, {min:-1, max:1, palette:['#1a9850','#ffffbf','#d73027']}, 'NBR pre');
Map.addLayer(post,{min:-1, max:1, palette:['#1a9850','#ffffbf','#d73027']}, 'NBR post');

Step 3: Calculate dNBR and Classify Severity

// Calculate delta NBR
var dnbr = pre.subtract(post).rename('dNBR');
Map.addLayer(dnbr, {min:-0.2, max:1, palette:['#2c7bb6','#ffff8c','#d7191c']}, 'dNBR');

// Classify burn severity using USGS thresholds
var sev = dnbr.expression(
  'b(0) <= 0.1 ? 0 : ' +
  'b(0) <= 0.27 ? 1 : ' +
  'b(0) <= 0.44 ? 2 : ' +
  'b(0) <= 0.66 ? 3 : 4'
).rename('severity');

Map.addLayer(sev, {
    min: 0, max: 4, 
    palette: ['#1a9641','#ffffbf','#fdae61','#d7191c','#a50026']
}, 'Severity classes');

Step 4: Summarize Area by Severity Class

// Fire perimeter (replace with actual perimeter if available)
var perimeter = aoi; // Use actual fire perimeter FeatureCollection

// Calculate area (hectares) by severity class
var areaImg = ee.Image.pixelArea().divide(10000).addBands(sev);
var byClass = areaImg.reduceRegion({
  reducer: ee.Reducer.sum().group({groupField: 1, groupName: 'severity'}),
  geometry: perimeter, 
  scale: 30, 
  maxPixels: 1e9
});

print('Area (ha) by severity class', byClass);

Step 5: Export Results

// Convert grouped dictionary to FeatureCollection for CSV export
var groups = ee.List(byClass.get('groups'));
var table = ee.FeatureCollection(groups.map(function(g){
    g = ee.Dictionary(g);
    return ee.Feature(null, {
        severity: g.get('severity'),
        area_ha: g.get('sum')
    });
}));

// Export table as CSV
Export.table.toDrive({
    collection: table,
    description: 'burn_severity_area_by_class',
    fileFormat: 'CSV'
});

// Export dNBR as GeoTIFF
Export.image.toDrive({
    image: dnbr,
    description: 'dnbr_geotiff',
    region: aoi,
    scale: 30,
    maxPixels: 1e13
});

Check Your Understanding

Why do we use SWIR bands in the NBR calculation?
What would a negative dNBR value indicate?
Why is timing of post-fire imagery important?
How might topography affect burn severity patterns?

Troubleshooting

Problem: Pre-fire NBR shows burn scars

Solution: Check your pre-fire dates - they may overlap with an earlier fire. Adjust date range.

Problem: High dNBR values appear in water bodies

Solution: Add a water mask using the JRC Global Surface Water dataset.

Problem: Clouds in post-fire composite

Solution: Extend the date range or use a different cloud masking approach.

Pro Tips

Use MTBS data: Compare your results to official MTBS burn severity maps
Consider RdNBR: Relativized dNBR normalizes by pre-fire NBR for more consistent results
Multiple fires: Use fire perimeter datasets (NIFC, CalFire) for specific events
Time series: Track post-fire vegetation recovery over multiple years

?? Lab Submission

Subject: Lab 30 - Wildfire Analysis - [Your Name]

Submit:

Map layers: NBR pre/post, dNBR, and severity classes
Table: area (hectares) by severity class within fire perimeter
Short discussion: Note any unburned/low-severity islands and likely drivers
EE script URL and exported CSV