Using satellite image data for detecting forest burn severity and evaulating post-fire temporal status: Mediterrenean region sample

The area consisting of various types of trees, bushes and plants is called forest. Forests are usually located near shores where they can meet their need of water more efficiently. Bringing numerous economic and social benefits, like a remedy to pollution, forests are a crucial natural resource. Forests are important for balancing our climate in general. Plants also play a crucial role in the purification of air. When breathing, they absorb carbon dioxide and release oxygen. They were the reason why life outside of water became possible in the first place. Plants and forests help to enrich our soil by recycling nutrients that are obtained from dead leaves and small animals on the forest floor. . Furthermore, they have a rather vital place in the ecosystem for the enviromental and social balance. Effects of forest fires and implications are one of the most important natural disasters all over the world. There are three conditions that need to be present in order for a wildfire to burn, which firefighters refer to as the fire triangle: fuel, oxygen, and a heat source. Fuel is any flammable material surrounding a fire, including trees, grasses, brush, even homes. The greater an area's fuel load, the more intense the fire. Air supplies the oxygen a fire needs to burn. Heat sources help spark the wildfire and bring fuel to temperatures hot enough to ignite. The prediction and forecasting of the fire caracteristics is fundamantel for fire fighting. When the statistical data analyzed it is observed that forest fires have shown variation, but in parallel to the population growth a number of forest fires has increased widely in recent years. This causes erosions, landslides, desertification and mass loss. In addition, after forest fires, renewal of forests and vegetation crucial for land management. Classical methods used for detection of burned area and burn severity require a long and challenging process in terms of time and cost factors. Thanks to the advanced techniques used in the field of Remote Sensing, burned area and burn severity can be determined with high accuracy. Remote Sensing techniques usually use various spectral indexes which emphasize changes induced by fire in vegetation spectral behaviour to determine fire charecteristics precisely. Remote Sensing data can assist fire management as three stages relative to fire occurence : (i) before the fire which related to vegetation biomass, (ii) during the fire which realted to near-real time location of fires, (iii) after the fire which related to assesment of burned areas. In this study, the three massive forest fires that occured in the Mediterannean district of our country have been observed. The pre-fire and post-fire continuum were sighted with the help of satallite images. Within this scope, by using satallite images, the burn severity was detected. Furthermore, rehabilitation studies were observed. In Turkey, it has not occured to one to put on a study on fire damage detection and burnt severity determination by using satallite images. The main purpose of this thesis, is to accurately and quickly detect the burn area and the burnt severty with the help of satallite images. In addition, a healthy observation of the foliation process of the rehabilitaion studies was aimed. In accordance with this purpose, the pre-fire and post-fire Landsat7 ETM + and MODIS Mod 09A (Surface Reflectance) satellite images and spatial autocorrelation statistics such as local Moran's I and Getis-Ord local Gi indexes belonging to Antalya-Taşağıl, Mersin-Gülnar and Hatay-Samandağ were used together to detect the burned area and the burn ratio with the help of NBR (Normalized Burn Ratio) and dNBR (Differenced Normalized Burn Ratio) index. In this study, 34 Landsat 7 ETM+ images and 6 MODIS Mod09A images used for evaluation of three forest fire period. Images for the initial assessment are taken directly before the fire (pre-fire image) and after the fire (post- fire image) as possible. The difference between those two images will show how much vegetation there was before the fire and how much was left after the burn thereby analyzing all the vegetation that was burned. Burn severity classification categories were divided into seven classes according to USGS (United States Geological Survey) standarts. Local Moran's I and Getis-Ord Local Gi wee calculated using dNBR as intensity. In the following process, the NDVI (Normalized Difference Vegetation Index) images were obtained and the rehabilitation studies were observed temporally. In this context, NDVI images taken just before the fire started (pre-fire image) and are compared other NDVI images taken after the fire (post-fire image) season until the year 2014. The spatial distribution of post-fire NDVI is mainly influenced by vegetation regeneration. Satellite images have been used as an important source for identification of fire damages and risks in fire management field. This study was performed using pre-fire and post-fire satellite images to determine burned area, burn severity and post-fire regeneration in Mediterranean Region, Turkey where serious forest fires reoccurred. In the end of the study, for Antalya -Taşağıl fire, with using MODIS dNBR image 20 479 ha, Landsat 7 ETM + dNBR image 16 996 ha, Landsat 7 ETM + local Moran's I technique 12 212 ha, Landsat 7 ETM + Getis-Ord local Gi technique 15 242 ha burned area results were detected. These results showed that, using Getis-Ord local Gi technique's result is the closest result to the General Directorate of Frorestry (GDF) datas. Forest rehabilitation works applied after fires, has been observed succesful in the majority of the area. For Mersin-Gülnar fire, with using MODIS dNBR image 7812 ha, Landsat 7 ETM + dNBR image 5388 ha, Landsat 7 ETM + local Moran's I technique 4262 ha, Landsat 7 ETM + Getis-Ord local Gi technique 5271 ha burned areas results were detected. These results showed that, using Getis-Ord local Gi technique's result is the closest result to the GDF datas for this fire, too. However, for this fire, forest rehabilitation works applied after fires, has been observed poor germination, and rehabilitation works results has been discovered worse than expected. For Hatay-Samandağ fire, with using MODIS dNBR image 1690 ha, germination Landsat 7 ETM + dNBR image 1162 ha, Landsat 7 ETM + local Moran's I technique 1074 ha, Landsat 7 ETM + Getis-Ord local Gi technique 15 242 ha burned areas results were detected. These results showed that, using Getis-Ord local Gi technique' s result is the closest result to the GDF datas for this fire, too. In the last two-year period, rehabilitation works applied has been observed benefits in the northern and southern sections of forest.