Burn Severity Mapping

Monitoring and predicting the severity of burnt areas using remotely sensed satellite data.

by Kate Doyle
06/09/2018

Figure 1: True colour Sentinel 2 images of a) pre-fire and b) post fire.
Note the artefact left by the aeroplane in Figure 1a.

On June 24th 2018 a large moor fire began on Saddleworth Moor, an upland heath area outside Manchester. The blaze was finally extinguished three weeks later on July 18th. Satellite imagery from this time can be useful to assess the consequences of fires such as these. In this case, two Sentinel-2 images were taken - showing a pre-burn image of the moor (Figure 1a) and a post burn image of the moor (Figure 1b) taken on 27th May 2018 and 4th July 2018 respectively.

Figure 1: True colour Sentinel 2 images of a) pre-fire and b) post fire.
Note the artefact left by the aeroplane in Figure 1a.
Figure 2: NDVI of Saddleworth of a) pre-fire and b) post fire

One way to measure the impact of fire is to look at the Normalised Differential Vegetation Index (NDVI) which shows a measure of live green vegetation. A higher NDVI shows healthier vegetation while less or no vegetation corresponds to a lower NDVI value. The pre-fire (Figure 2a) and post fire (Figure 2b) images differ dramatically, with a clear delineation of the fire extent visible in the post-fire image where NDVI is lower. Prior to the fire the vegetation has a high NDVI and therefore is healthy.

Figure 2: NDVI of Saddleworth of a) pre-fire and b) post fire

Fire fuel (Figure 3) was calculated using the ENVI fire tool and shows the burn hazards and distribution of fire fuels before the fire - in this case 27th May 2018. Fire fuel maps such as this can be useful to prepare for fires, as they can show areas which are most susceptible to being burned. Areas with higher fuel usually contain drier materials while lower fuels are greener and lusher.

Burn severity post fire was calculated using a Normalised Burn Ratio (NBR) for the pre and post fire images individually with the calculation being:

$$NBR = ({NIR - SWIR \over NIR + SWIR})$$

From this Differenced Normalised Burn Ratio (dNBR) between images is calculated:

$$dNBR = NBR_{pre-fire} - NBR_{post-fire}$$

The output of this is a dNBR (Figure 4) image showing the difference between pre and post fire images, with a higher dNBR being indicative of more severe fire damage while areas with a negative dNBR is suggestive of regrowth following the fire. The subsequent dNBR image was then classified using the burn severity classification proposed by the United States Geological Survey (Table 1) which produces a thematic burn severity layer. From this initial estimation other parameters such as soil-burn severity can be estimated as well as implementing restoration plans and, if the data is available, a comparison with historic burn areas.

Severity Level
Enhanced Regrowth - High (post-fire)
Enhanced Regrowth - High (pre-fire)
Unburned
Low Severity
Moderate - Low Severity
Moderate - High Severity
High Severity
dNBR Range (Not Scaled)
-0.500 to -0.251
-0.250 to -0.101
-0.100 to +0.099
+0.100 to +0.269
+0.270 to +0.439
+0.440 to +0.659
+0.660 to +1.300
Table 1: USGS severity level classification with corresponding dNBR value
Figure 3: Fire fuel Saddleworth Moor
Figure 4: Burn Severity Map of Saddleworth Moor

There does not appear to be a significant correlation between the fire fuel and subsequent burn severity. However, in order to monitor future fires in the area, or look at the possibility for other moor fires, the data could be useful in inferring trends in fire spread and looking for common factors that may have caused a higher burn severity.

Remotely sensed imagery can be useful to infer trends both pre and post fire as well as allowing near real-time monitoring of burn events, although it should be noted that to gain the most accurate results, field studies should also be carried out.