Analysis of land use/land cover change (LULCC) and debris flow risks in Adama district, Ethiopia, aided by numerical simulation and deep learning-based remote sensing

• Published in: Stochastic environmental research and risk assessment Vol. 37; no. 12; pp. 4893 - 4910
• Main Authors: Bojer, Amanuel Kumsa, Ahmed, Muhammed Edris, Bekalo, Desta Jula, Debelee, Taye Girma, Al-Quraishi, Ayad M. Fadhil, Deche, Almaz
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2023; Springer Nature B.V
• Subjects: Aquatic Pollution; Chemistry and Earth Sciences; Computational Intelligence; Computer Science; Debris flow; Deep learning; Detritus; Earth and Environmental Science; Earth Sciences; Environment; Environmental risk; Flow simulation; Grasslands; Land cover; Land use; Landslides; Math. Appl. in Environmental Science; Mathematical models; Natural resources; Numerical methods; Original Paper; Particle methods (mathematics); Physics; Probability Theory and Stochastic Processes; Remote sensing; Risk; Risk assessment; Shrublands; Simulation; Statistics for Engineering; Sugarcane; Waste Water Technology; Water Management; Water Pollution Control; Ethiopia; Manning’s coefficient; Deep learning models; Debris flow; Numerical simulation model; LULCC; Ethiopia
• ISSN: 1436-3240; 1436-3259
• DOI: 10.1007/s00477-023-02550-w

Detecting land use/land cover change (LULCC) and assessing the risk of slope failure and debris flow has been a worldwide concern. This study is the first in Adama District, Ethiopia, to use deep learning (DL)-based remote sensing to assess LULCC and predict the risk of slope failures and debris flows using numerical simulation methods. This study uses DL and remote sensing to analyse the spatiotemporal changes in LULC and landslide sites. The enhanced detection of debris flow susceptibility areas enabled the precise prediction of these areas’ location and sphere of influence and the precise evaluation of debris flow risk. This led to a reduction in the losses caused by such disasters. Changes in the six classes of LULC were assessed with an overall accuracy of above 87% and an overall kappa statistic of 85%. The results revealed a decreased trend in grassland, shrubland, and bareland over 30 years (1991–2021) of − 31.03 km 2 , − 38.15 km 2 , and − 114.19 km 2 , respectively. Also, a recent analysis of land-use maps from the past three decades reveals that the built-up area has increased significantly, from 0.95% to 5.64%. In contrast, shrubland has decreased notably, from 12.01 to 7.78% since 2021. These changes suggest that human activity significantly impacts the landscape, and that more needs to be done to protect our natural resources. The depth-integrated particle method flow simulation technique reveals high landslide risk in Adama City and Wonji sugar cane fields, aiding decision-makers in reducing damage and limiting over-cultivation in high-risk areas.