Introduction to Burn Severity Mapping

Forest fires are a natural part of many ecosystems, but their intensity and impact can vary greatly. **Burn severity** is a measure of the ecological change caused by a fire. Mapping it is essential for post-fire management, helping foresters to prioritize areas for erosion control, monitor natural regeneration, and plan reforestation efforts. This is typically done by calculating the **Normalized Burn Ratio (NBR)**, an index that uses the Near-Infrared (NIR) and Short-Wave Infrared (SWIR) bands, which are sensitive to changes in vegetation health and soil moisture. By comparing the NBR from before and after a fire, we can create a **differenced NBR (dNBR)** map that quantifies the severity of the burn.

Analysis Workflow

Map Visualization

The map below shows the classified burn severity. Colors range from green (unburned) and yellow (low severity) to red and purple (high severity), clearly delineating the fire's impact on the landscape.

GEE Code Snippet

This script processes pre- and post-fire imagery to calculate and classify the differenced Normalized Burn Ratio (dNBR), and adds a legend to the map.

// NOTE: You must define a 'boundary' FeatureCollection before running.
var prefire_start = '2019-01-01';
var prefire_end = '2020-02-15';
var postfire_start = '2020-02-20';
var postfire_end = '2020-02-22';
var platform = 'S2'; // Sentinel-2 or Landsat 8

var ImCol = (platform == 'S2') ? 'COPERNICUS/S2' : 'LANDSAT/LC08/C01/T1_SR';
var area = ee.FeatureCollection(boundary);
Map.setCenter(101.7784, 16.87633, 12);

// Filter image collections for pre- and post-fire periods.
var prefireImCol = ee.ImageCollection(ImCol).filterDate(prefire_start, prefire_end).filterBounds(area);
var postfireImCol = ee.ImageCollection(ImCol).filterDate(postfire_start, postfire_end).filterBounds(area);

// Cloud masking functions for S2 and L8.
function maskS2sr(image) {
  var qa = image.select('QA60');
  var cloudBitMask = 1 << 10;
  var cirrusBitMask = 1 << 11;
  var mask = qa.bitwiseAnd(cloudBitMask).eq(0).and(qa.bitwiseAnd(cirrusBitMask).eq(0));
  return image.updateMask(mask).copyProperties(image, ["system:time_start"]);
}
// ... [L8 mask function would be here] ...

// Apply cloud mask.
var prefire_CM_ImCol = prefireImCol.map(maskS2sr);
var postfire_CM_ImCol = postfireImCol.map(maskS2sr);

// Create mosaics and calculate NBR.
var pre_cm_mos = prefire_CM_ImCol.mosaic().clip(area);
var post_cm_mos = postfire_CM_ImCol.mosaic().clip(area);
var preNBR = pre_cm_mos.normalizedDifference(['B8', 'B12']);
var postNBR = post_cm_mos.normalizedDifference(['B8', 'B12']);

// Calculate dNBR.
var dNBR = preNBR.subtract(postNBR).multiply(1000);

// Define SLD style for classification visualization.
var sld_intervals =
  '<RasterSymbolizer>' +
    '<ColorMap type="intervals" extended="false" >' +
      '<ColorMapEntry color="#7a8737" quantity="-250" label="-250" />' +
      '<ColorMapEntry color="#acbe4d" quantity="-100" label="-100" />' +
      '<ColorMapEntry color="#0ae042" quantity="100" label="100" />' +
      '<ColorMapEntry color="#fff70b" quantity="270" label="270" />' +
      '<ColorMapEntry color="#ffaf38" quantity="440" label="440" />' +
      '<ColorMapEntry color="#ff641b" quantity="660" label="660" />' +
      '<ColorMapEntry color="#a41fd6" quantity="2000" label="2000" />' +
    '</ColorMap>' +
  '</RasterSymbolizer>';
Map.addLayer(dNBR.sldStyle(sld_intervals), {}, 'dNBR classified');

// ... [Legend and Area Calculation code follows] ...

Analysis & Results

Reference