An investigation of the use of remote sensing and climatic data for improved pre-fire management planning in Lebanon

Abstract

Forest fires represent a damaging threat to Mediterranean forests. Like many other Mediterranean countries, forest fires have been destroying Lebanons natural resources. In 2009, a National Strategy for Forest Fire management was endorsed by the Government of Lebanon aiming at the reduction of forest fires in the country while allowing for fire regimes that are socially, economically, and ecologically sustainable. Fuel, weather, and topography are considered critical to fire management, and are the main factors that influence the risk of fire ignition. In the present study, Object-Based Image Analysis (OBIA) was used to map fuel types, fire ignition risk, and fire sensor deployment sites. The aim of this work was to investigate the role of advanced remote sensing and climatic data for improved pre-fire management planning in Lebanon. The specific objectives were: 1. To investigate the potential use of multispectral images (LANDSAT and ASTER) for fuel type mapping and Very High Spatial Resolution (VHR) (SPOT) images for fire ignition risk mapping at the regional level in North Lebanon. 2. To investigate the use of OBIA of VHR SPOT imagery for optimal fire sensor distribution at the local level in North Lebanon. 3. To investigate temporal fire activity in Lebanon and its relationship to climatic variability. First, the OBIA approach was used for the generation of three thematic fuel type maps of North Lebanon based on an adapted Prometheus fuel type classification system (a total of six classes). Field collected data were used to train the classification models. Initially a LANDSAT TM image was segmented and then classified, the results of the LANDSAT classification showed a low overall accuracy of 50% when using six different fuel types. This required the amendment of the classification scheme. A second LANDSAT classification scheme was developed by merging fuel type classes into three categories, and an improved overall accuracy of 75% was obtained. On the other ha a multispectral ASTER image was classified taking into account the adapted Prometheus fuel type classification system. This classification, with a 70% of overall accuracy, provided enough spectral information to be able to recognize the main fuel type classes of the Prometheus classification system in a semi arid environment of the Eastern Mediterranean. It was concluded that the use of OBIA and ASTER imagery could present an affordable and operational approach for fuel type mapping in Lebanon. Subsequently, the ASTER resulting fuel type map was integrated with other topographic and static parameters into the generation of a VHR (SPOT) based fire ignition risk map of North Lebanon. The 3 bands of the SPOT image, the aspect and slope layers of the DEM, a thematic fuel type map, and a road network were imported for use in the OBIA approach. The final results revealed that, 62% of the forested area of the study area of North Lebanon is at very high and high risk of fire ignition. Also, the OBIA of VHR (SPOT) imagery was used to map fuel types, fire ignition risk, and to develop a deployment map of forest fire sensor nodes at the local level. The combined use of fuel type map and fire risk map allowed the generation of a fire sensor deployment map based on multi-criteria decision analysis. The final results showed what is expected to be the most appropriate sites for the deployment of three different fire sensor nodes at the local level. This is the first of its kind study for the deployment of wireless sensor networks for forest fire detection. Simultaneously, fire seasonality was determined using fire occurrence data from 2001 to 2011. Climatic variables which included daily temperature, relative humidity, precipitation and wind velocity for the past ten years were studied for correlation with fire occurrence. Main results showed a likely increase in the length of the fire season during the past decade.