Evaluation of environmental and climatic impacts of sand dune movement using geographic object-based image analysis and machine learning

Abstract

This study focuses on the identification and extraction of sand dunes, analysis of morphometric features, and prediction of active processes in desert areas, and employs Geographic Object-Based Image Analysis (GEOBIA) in desert environments. To achieve this goal, this study utilizes a combination of self-organizing maps (SOM), analytical hierarchy process (AHP), and decision tree (DT) algorithms to predict erosion risks caused by sand dune migration. This study also evaluates the relationship between sand dune morphometric features and active processes in diverse desert environments based on these features. Furthermore, to forecast climatic parameters and their relationship with sand dune morphometric features, the Long Short-Term Memory (LSTM) model is utilized. The results of the neural network analysis based on SOM indicate that transverse and barchan dunes are more susceptible to erosion hazards due to their irregular shape and broad distribution. Star dunes, however, exhibit higher stability and density, and this renders them less vulnerable. Moreover, the results of the correlation between dune morphometry and climatic parameters indicate higher hazards from barchan, star, and seif dunes due to their greater heat absorption than other types of dunes. Finally, climatic parameter prediction via the LSTM method, with R² accuracy of 0.8 and 0.9 and RMSE values of 0.11 for wind speed and 0.26 for temperature, respectively, suggests that the increase in aridity and wind speed in desert areas will intensify the migration of dunes in the future. Therefore, predicting environmental risks caused by sand dune movements and active processes in desert areas based on their morphometric characteristics facilitates better management of desert areas.