Evaluation of Optical Remote Sensing Data in Burned Areas Mapping of Thasos Island, Greece

Mohamed Elhag,Silvena Boteva,Nese Yimaz,Jarbou Bahrawi

doi:10.1007/s41748-020-00195-1

Abstract

Forest fires are a common feature in the Mediterranean forests through the years, as a wide tract of forest fortune is lost because of the incendiary fires in the forests. The enormous damages caused by forest fires enhanced the efforts of scientists towards the attenuation of the negative effects of forest fire and consequently the minimization of biodiversity losses by searching more for the adequate distribution of attempts on forest fire prevention and, suppression. The multi-temporal Principal Components Analysis is applied to a pair of images of consecutive years obtained from Landsat-8 satellite to unconventional map and assess the spatial extent of the burned areas on the island of Thasos, Greece. First, the PCA was applied on the before fire image, and then a multi-temporal image is created from the 3rd, 4th, and 5th band of before and after images including Normalized Difference Vegetation Index to enhance the results. The results from the different steps of this analysis robustly mapped the burned areas by 82.28 ha confirmed by almost 85%. Are compared with data provided by the local forest service in order to assess their accuracy. The multi-temporal PCA outputs including NDVI (PC 4, PC %, and PC 6) give better accuracy due to its ability to distinguish the burned areas of older years and to the Normalized Difference Vegetation Index that gives better variance to the image.

Highlights

The counting of forest fires has been increasing since that to reach a maximum of 206 fires in 2016 from a minimum of 83 fires in 1984; corresponding burned areas were 6938.8 ha and 2198.8 ha, respectively (Christopoulou et al 2014; Kazanis and Arianoutsou 1996; Paula et al 2009)
These procedures only indicate how the classification strategy being employed works well in the training areas, and nothing more (Lillesand et al 2014). For this reason, proposed applications of other techniques as a mean of improving the interpretation of the error matrix among them, the Kappa coefficient were developed. The computation of this coefficient is used to determine whether the results presented in the error matrix are significantly better than a random result (Green and Congalton 2004)
The topographic map is projected according to Transverse Mercator / Spheroid GRS 1980 / Datum EGSA87

Summary

Introduction

The counting of forest fires has been increasing since that to reach a maximum of 206 fires in 2016 from a minimum of 83 fires in 1984; corresponding burned areas were 6938.8 ha and 2198.8 ha, respectively (Christopoulou et al 2014; Kazanis and Arianoutsou 1996; Paula et al 2009). There has been a substantial increase in the areas burned by wildfires, related to weather conditions (Koutsias et al 2013) especially to arson, the total number of the fire incidents varying from a minimum of 968 in 2012 to a maximum of 1443 in 2018, with corresponding areas burned of 19,613 ha and 105,450 ha, respectively (Elhag et al 2018; Turco et al 2016). Has gathered available information’s on fuel attributes conducted from the phrygana communities in Crete related to fire risk. Starting from 1980 and the number of forest fires increases, many occurred due to natural causes and some were prescribed fires (Kavouras et al 2012; Palaiologou et al 2018)

Methods

Results

Conclusion