Burnt area detection in PDN using two post fire images

Abstract

The dynamic of open fire phenomena makes remote sensing from satellite a good tool for both fire detection and burn area assessment. Multi-temporal approaches usually relies on a couple of multispectral images acquired the former before and the latter after the fire event to derive vegetation maps (e.g. by using threshold techniques on the Normalised Differential Vegetation Index-NDVI, Normalized Burn Ratio -NBR) However, in countries such as UK a pre-and post fire pair of cloud free images from VNIR operating satellite may be not available. This work focuses on detecting a burned area when only two post fire images are available. The study area is Dovestones- Stalybridge that was affected by a fire on 10th April 2016. Data acquired by Landsat OLI (30m x 30m pixels) on 4 May and 5 June 2016 and Sentinel 2A (10 x 10 m pixels) on 20 April 2016 and 19 July 2016 were used to derive dNDVI - differential NDVI- map by using two post images. A validation fieldtrip was conducted on July 2016 to identify re-growing and burn patches to validate the obtained results. The results are promising and the techniques can be used to complement burn area detection retrieved by using new SAR based methods or more traditional methods based on dNDVI or dNBR evaluated on a pre- post pair images. Work is in progress as more test cases are needed to derive a threshold approach.