Work

Fusion of deep neural networks and infrared wave analysis for climate pattern prediction and shifting sand hazard Assessment

Year: 2025

Type: article

Abstract: Since shifting sands pose a serious threat to surrounding areas, it is important to forecast their spatial trends in desert regions for better management of these areas. Given the importance of this issue, this study develops a hybrid approach to distinguish sand dunes from other desert features with deep neural networks and remote sensing techniques at different scales in the Nafud Desert in Saudi Arabia and the Registan Desert in Afghanistan. The study also models shifting sand dunes by combining morphometric features and infrared waves to predict their spatial boundary changes and associated risks. Moreover, the analysis of the relationship between morphometric features and climate parameters forecasts future spatial shifts of sand dunes. The results show that the hybrid deep learning model of Very Deep Super-Resolution (VDSR) and Laplacian effectively enhances image resolution for precise sand dune boundary extraction. Furthermore, the integration of topographic and spectral features into a new hybrid model, TPI, and K-Mean++ enables accurate prediction of spatial boundary changes in desert areas. The combination of deep neural networks and multi-resolution segmentation (MRS) identifies the optimal scale for sand dune separation as 90 and 100 m. Moreover, the analysis of spectral reflectance combined with the CA-Markov model and climate parameters indicates that infrared waves can predict shifting sand dunes with high accuracy and highlight them as a potential threat to surrounding lands. Morphometric-climate analysis further reveals how these shifts may evolve, aiding future hazard mitigation. This integrated approach advances our understanding of sand dune-climate interactions and offers a robust framework for managing desertification threats in vulnerable regions.

Source: International Journal of Applied Earth Observation and Geoinformation

DOI: 10.1016/j.jag.2025.104649

Authors: Huayu Lu, Marzieh Mokarram

Institutions: Nanjing University, Shiraz University

Language: English

Topic: Aeolian processes and effects

Subfield: Earth-Surface Processes

Field: Earth and Planetary Sciences

Domain: Physical Sciences

SDGs: Climate action, Life on land

Open Access status: gold

Estimated APC paid: $3,550