Abstract:
Precise classification of land use and land cover (LULC) is essential for sustainable
resource management and understanding the climate impact of such changes, facilitated
by cloud computing platforms like Google Earth Engine (GEE) and their extensive pre processed data, empowering Machine Learning and Deep Learning techniques for
spectral segmentation of remote sensing data. This study compares four ML (Random
Forest (RF), Support Vector Machine (SVM), Classification and Regression Trees
(CART), Gradient Tree Boosting (GTB)) algorithms for LULC classification in GEE
followed by impact of LULC change on LST and socio-economic conditions. SVM and
RF provided the most promising results, providing an accuracy of more than 81% for
each year (2002, 2013, and 2022), while GTB and CART underperformed, especially
in segregation of water, forest, and shadows, providing an accuracy of 76 – 81%.
Spectral indices were derived using convolution in GEE. Finally, correlation analyses
between spectral indices, topographical variables and Land Surface Temperature (LST)
reveals a positive relationship between LST, Normalized Difference Built-Up Index
(NDBI), Urban Density Index (UDI), and Hillshade and a negative correlation between
Normalized Difference Vegetation Index, Normalized Difference Water Index
(NDWI), slope, aspect and elevation. Survey findings indicate urbanization and
deforestation have raised temperatures and increased extreme weather events. While
47% view land use changes positively, 36% report negative impacts. 66% note reduced
water access and heightened resource competition. Other effects include increased
energy demand (36%), rising property values (67%), improved agriculture (47%),
higher traffic (64%), and increased housing costs (51%). Healthcare and education
access are affected for 26%, and job opportunities for 12%. The study concludes that
land use changes elevate Land Surface Temperature (LST), impacting human lives.
