Assessment of Small-Extent Forest Fires in Semi-Arid Environment in Jordan Using Sentinel-2 and Landsat Sensors Data

The objective of this study was to evaluate the separability potential Sentinel-2A (MultiSpectral Instrument, MSI) and Landsat (Operational Land Imager, OLI Thermal Infrared Sensor, TIRS) derived indices for detecting small-extent (<25 ha) forest fires areas severity degrees. Three remote sensing [differenced Normalized Burn Ratio (dNBR), differenced Different Vegetation Index (dNDVI), surface temperature (dTST)] were used at three sites located in Northern Jordan; Ajloun (total burned area 23 ha), Dibbeen (burned 10.5), Sakeb 15 ha). Compared ground reference data, Sentinel-2 MSI able delimit fire perimeter more precisely than Landsat-8. accuracy (area coincidence) 7%–26% higher that Landsat-8 across sites. In addition, reduced omission by 28%–43% commission 6%–38% compared sensors. Higher attributed spatial resolution lower mixed pixel problem (mixed pixels within Sentinel-2, 8.5%–13.5% vs. 31%–52% OLI). dNBR had (higher coincidence values less commission) dNDVI dTST. terms degrees, (the best index candidate) from both satellites sensors only capable severe spots “severely-burned” with producer >70%. fact, dNBR-Sentinel-2/Landsat-8 overall Kappa coefficient classifying degree 70% studied sites, except Sentinel-dNBR (72.5%). conclusion, promise delimitate perimeters small-scale areas, but further remotely-sensed techniques are require (e.g., Landsat-Sentinel data fusion) improve severity-separability potential.