Mapping coastal transformations with a novel Cellular Automata-Markov-Random forest framework for land use change...

Area of Science:

• Environmental Science
• Geospatial Analysis
• Coastal Management

Background:

• Coastal areas face dynamic changes from natural processes and human activities.
• Accurate prediction of shoreline and land use/land cover (LULC) changes is crucial for sustainable coastal management.
• Oman's coastal regions are particularly vulnerable to these dynamic changes.

Purpose of the Study:

• To develop and evaluate a hybrid modeling framework combining CA-Markov and machine learning for enhanced LULC and shoreline change projections in Oman.
• To assess the predictive performance of different hybrid models against the traditional CA-Markov model.
• To provide accurate projections for future coastal LULC and shoreline dynamics to support sustainable management strategies.

Main Methods:

• Delineation of coastlines using multi-temporal Landsat images (1997-2024) and the Normalized Difference Water Index.
• Quantification of coastal erosion and accretion rates using End Point Rate and Linear Regression Rate analyses.
• Evaluation of four models (CA-Markov, CA-Markov+XGBoost, CA-Markov+CART, CA-Markov+RF) for future LULC prediction, with CA-Markov+RF showing superior performance.

Main Results:

• Significant spatial variability in shoreline changes observed between 1997 and 2024, with notable erosion in Rakhyut (-1.81 m/year) and accretion in Bawshar (1.41 m/year).
• Rapid urban expansion detected, particularly in Muscat, where built-up area increased from 10.31 km² (1997) to 116.41 km² (2015).
• The hybrid CA-Markov+RF model achieved the highest predictive accuracy (0.935) compared to CA-Markov (0.905), demonstrating the effectiveness of machine learning integration.

Conclusions:

• The hybrid CA-Markov+RF model significantly enhances the accuracy of LULC and shoreline change projections in vulnerable coastal areas.
• Future projections (2033) indicate continued urban growth in Salalah and Sohar, with potential reductions in vegetation cover in arid zones.
• The findings underscore the importance of advanced modeling techniques for effective coastal zone management and sustainable development planning.