Comprehensive modelling of coastal landuse and land cover (LULC) dynamics and future changes using random forest, GEE, and MLP-Markov chain integration: a geo-computational approach.

Coastal landuse / land cover (LULC) faces dynamic transformations driven by natural and human activities. This study investigates decadal LULC dynamics and their future projections (2034, 2044, and 2054) along the Ernakulam-Alappuzha coastal stretch of Kerala, India, using Google Earth Engine (GEE)- based time-series Landsat ETM + and OLI images from 2004, 2015, and 2024. GIS-based machine learning (ML) techniques, including Random Forests (RF), Support Vector Machines (SVM), and Classification and Regression Trees (CART), were used to extract LULC features. The spectral indices, such as NDVI (Normalized Difference Vegetation Index), SAVI (Soil-Adjusted Vegetation Index), MNDWI (Modified Normalized Difference Water Index), NDBI (Normalized Difference Built-up Index), NDBaI (Normalized Difference Bareness Index), BSI (Bare Soil Index), and UI (Urban Index), were incorporated to enhance the future LULC trends. Among all ML algorithms, the RF achieved the highest classification accuracy with 95.46%, 92.38%, and 93.94% for 2004, 2015, and 2024, respectively. The LULC patterns indicate pronounced urban expansion, with built-up areas increasing from 109.93 km2 (6.47%) in 2004 to 373.49 km2 (21.98%) in 2024, largely replacing plantations, which declined from 648.94 km2 (38.17%) to 397.03 km2 (23.36%). The results reflect an accelerating urbanization trend and a gradual decline of natural and semi-natural coastal land covers. Future LULC projections for 2034, 2044, and 2054, simulated using the Multi-Layer Perceptron-Markov Chain (MLP-MC) model in the Land Change Modeller (LCM) of TerrSet v.20, reveal significant transformations in the Ernakulam-Alappuzha coastal stretch. Validation of the 2024 predicted map against the actual LULC map produced a Kappa value of 0.71, confirming the model's reliability. The projections indicate a dominant expansion of built-up areas, from 373.49 km2 (21.98%) in 2024 to 747.52 km2 (43.99%) by 2054, with the most rapid increase occurring between 2024 and 2034 (147.58 km2, 39.51%, 14.76 km2/year). The changes of LULC reflect the trend of future patterns which accelerate urban growth, primarily through the conversion of plantations, fallow land, vegetation, and wetlands. In contrast, natural and semi-natural land covers are expected to decline markedly, with the steepest changes in the first decade (2024-2034) and slower, continued transformations thereafter. These patterns suggest ongoing urban dominance and degradation of coastal ecosystems, underscoring the need for strategic land-use planning and sustainable coastal management practices.