Abstract:
The vegetation cover in Al Jabal Al Akhdar has been subjected to human and
natural pressures that have contributed to the deterioration and shrinking of the
vegetated area. Therefore, the principle goal of this dissertation was to establish
and evaluate the changes in the natural vegetation of the Al Jabal Al Akhdar
region in the period following the 2011 Libyan uprising. The thesis is comprised
of three main objectives; the first is to provide a quantitative assessment of
changes in natural vegetation cover over a period from 2004-2016, and identify
the consequent impact of human activity; the second is to investigate the impact
of climate on the natural vegetation cover; and the third objective is to evaluate
the ability of machine learning techniques to predict the natural vegetation cover
types.
GIS and remote sensing techniques and Landsat imagery, population MODIS
NDVI and climate satellite-based data have been used to achieve these
objectives, along with the ancillary data, across 53 sites in the study area. Six
classified Landsat image scenes have been used for undertaking a post-
classification comparison approach to detect the changes and the types of
changes, by the use of image processing, GIS software and spreadsheet, and
programme scripts used to detect LULC changes and determine human
activities impact. The correlaction between the ANDVI and climate factors for
each lanform, and the trends of climate factors and ANDVI for each sites in
each landform have been undertaken using statistical analysis package and
spreadsheet. Lastly the machine learning ‘J48’ algorithm, within the WEKA tool,
was applied on ANDVI, climate data, and spatial characteristics for 53 sites and
analysed statistically to test its ability to predict the natural vegetation type.
The main research findings have confirmed that from 2004-2016, natural forest
and rangelands decreased by 71,543 ha or 7.10% of the total area as a result of
urbanisation and agricultural expansion. Human activities have had more
impact than climate impact on LULC changes. The machine learning classifier
decision tree ‘J48’ algorithm was also found to have the ability to classify and
predict the natural vegetation cover type.
Finally, an evaluation was undertaken of the current distribution of natural
vegetation cover, and a forecast of future changes, utilising high-resolution
imagery is recommended. A conclusion considers how developing action plans
using tools such as those described to manage and protect the natural
vegetation cover are highly recommended.