Quantifying large area land use and land cover change in Sri Lanka and its impact on elephant habitat using time series remote sensing data

Land use and land cover change (LULCC) is a key aspect of global environmental change. Exponential human population growth and intensification of human land use in the recent past is resulting in many changes in ecosystem functions and services at all scales from local to global. Monitoring and quantifying these changes are essential for environmental management and conservation. Remote sensing / earth observation has been applied for this purpose for decades as it provides the ability to detect change over large areas. In Sri Lanka, the country's landscape has been evolving over an extended period due to human (socioeconomic, political) and natural (biophysical) causes. Following the introduction of an open economic policy in the late 1970s, multiple socioeconomic and political changes have taken place. These have resulted in the establishment of multi-purpose river basin development projects, transportation and highway development projects, agricultural expansion through irrigation, and the emergence of new urban centres and expansion of existing ones. A civil war also afflicted the country from 1980 to 2009. Northern and Eastern Sri Lanka was severely affected, and during some phases of the war, the entire country was affected and experienced negative economic development. LULCC records for these periods are missing or extremely limited. When the 30 years of civil war ended in 2009, LULCC intensified around the country, due to a range of post-war development activities and resettlement. The LULCC occurring in this dynamic landscape is yet to be explored and quantified. The first phase of this research aimed to test the applicability of Landsat time series data for quantifying LULCC in Sri Lanka over the last 25 years. A land cover change map for the period was developed using 4,440 image tiles covering nine Landsat scenes. The LandTrendr change detection algorithm, utilising the normalised burn ratio (NBR) and normalised difference vegetation index (NDVI), was used to develop spectral trajectories over this period. A land cover change map was created with the Random Forest (RF) method, using 2,117 manually interpreted reference pixels, of which 75% were used for training and 25% for validation. Three change classes were defined: unchanged, disturbed but unchanged, and changed. A classification process was then performed using RF, with the model achieving 94% overall accuracy. In the second phase, a Sri Lanka-wide analysis was conducted to explore the role of LULCC in relation to human elephant conflict (HEC) using official government data and the land cover change dataset (1993-2018) developed in the first stage of the research. HEC has grown in Sri Lanka in the last two decades and become a key environmental issue. Sri Lanka accounts for 10% of the total Asian elephant population and 2% of Asian elephant habitat. Asian wild elephants are now an endangered species due to the continuous threat of habitat loss and fragmentation. Sri Lanka now also has the highest elephant death rate and second highest number of HEC-related human deaths in elephant-range countries. A total of 18,000+ HEC incidents (human death, elephant death, human injury, property damage) for 2003 to 2018 were collected from the Department of Wildlife Conservation of Sri Lanka. This thesis investigated rates of HEC over time and compared these to rates of LULCC over the same period. It also presents spatial analytics of HEC and LULCC, as well as determining hotspots of HEC and LULCC using a kernel density estimator. Annual HEC incidents were found to broadly increase in line with land use change events (r = 0.43, p < 0.05). We found a strong negative correlation between HEC location and distance to LULCC events. In total, 98% of HEC was found to be within 1 km of an area that experienced recent LULCC. The third stage of the study explored the impact of LULCC on the performance of the protected area network in Sri Lanka. Since the introduction of Aichi Biodiversity Target 11, under the Convention on Biological Diversity (2010), Sri Lanka has increased its protected area network to 632, with 29% of its terrestrial land area now under protection. Although these protected areas exist on paper, there is reason for concern over their performance, with LULCC adversely affecting them. This study assessed the 632 terrestrial protected areas in Sri Lanka. The evaluations of protected areas were based on four criteria: amount of LULCC, human influence, low fragmentation, and high connectivity. Five variables were selected to represent these, following the literature: total LULCC, Human Footprint Index, determined by perimeter area ratio, Euclidian nearest neighbour distances, and effective mesh size. All variables were given equal weight and classified into five performance levels. Results showed that 78% of the protected areas performed moderately well, while 14% performed highly and 8% scored low. The latter part of this analysis compared performance score against protected area types of the International Union for Conservation of Nature (IUCN), and showed that strict nature reserves (IUCN type Ia) and wilderness areas (IUCN type Ib) performed above average, while national parks (IUCN type II) performed below average. The final stage of the research explored whether protected areas in Sri Lanka meet the requirements to protect the endangered Asian elephant (Elephas maximus). The results of the protected area performance analysis extracted 400 protected areas in Sri Lanka's HEC region. First, a spatial analysis was performed to find the proximity of HEC incidents to protected areas. The number of patches and mean patch size were derived using Fragstats (version 4.2). The study found that 86% of elephant death incidents occurred within a 5 km radius of protected areas, with a strong negative correlation with distance from protected areas (r = -0.94, p < 0.05); 43% of HEC incidents and 23% of elephant deaths occurred inside protected areas. These areas were also found to fragment over time and elephant deaths showed a strong positive correlation with fragmentation (r = 0.88, p < 0.05). Wildlife regions that experienced high LULCC also experienced a greater number of elephant deaths, with a moderate positive correlation (r = 0.54, p < 0.05). The research concluded with a synthesis on the future LULCC trends in Sri Lanka and the implications for protected area management and HEC.