Scholarlink

<p>Evaluating and projecting Land Use Land Cover (LULC) changes using spatiotemporal data in the Dechatu River Catchment (DRC), eastern Ethiopia, is vital for environmental monitoring and catchment planning, as the area is highly susceptible to flash floods and erosion that impact the city of Dire Dawa. This study aimed to analyze LULC classification, examine spatiotemporal changes from 1995 to 2024, and project future land use patterns for 2050 using multi-temporal satellite imagery processed with the Random Forest (RF) machine learning algorithm in Google Earth Engine. It projected the future changes for the year 2050 by integrating the QGIS MOLUSCE plugin with cellular automata (CA) and artificial neural network (ANN) techniques. While robust machine-learning and cellular-automata methods were applied, uncertainties persist due to classification errors, driver-variable quality, spatial resolution limits, and the assumption that recent transition trends remain stationary in the future. Five LULC classes, built-up, agricultural land, forest, shrubland, and barren land, were identified for 1995, 2010, and 2024. Overall classification performance improved over time, reaching 91.35% overall accuracy and 88.41% Kappa in 2024. Results revealed consistent expansion of built-up and agricultural areas by 197% and 48.44%, respectively, while forest, shrubland, and barren land experienced significant declines of 33.45%, 38.67%, and 60.45% between 1995 and 2024. Projections indicate the continued dominance of agriculture (66.71%), built-up areas rise to 8.09%, and a decline in forest (7.39%) and shrublands (15.49%) (net increases of 14.46% agriculture and 25.62% built-up from 2024; net decreases of 31.43% forest, 28.92% shrubland and − 14.33% barren land). Qualitative result from FGDs and KIIs suggest that the changes are driven by multiple factors, such as agricultural expansion, population growth, fuelwood collection, urban development, charcoal production, informal settlements, and policy interventions. This study provides insights into LULC dynamics in DRC and provides a valuable foundation for sustainable evidence-based planning and watershed management to overcome hazards such as flash flooding as well as sustainable restoration of degraded landscapes of the catchment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-temporal modelling and projection of land use/land cover dynamics in the Dechatu River Catchment, Ethiopia using Google Earth Engine and machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月25日

查看全文
  • Abdulbasit Hussein,
  • Samuel Feyissa,
  • Ketema Bekele,
  • Sinteyahu Workeneh,
  • Kibebew Kibret

摘要

Evaluating and projecting Land Use Land Cover (LULC) changes using spatiotemporal data in the Dechatu River Catchment (DRC), eastern Ethiopia, is vital for environmental monitoring and catchment planning, as the area is highly susceptible to flash floods and erosion that impact the city of Dire Dawa. This study aimed to analyze LULC classification, examine spatiotemporal changes from 1995 to 2024, and project future land use patterns for 2050 using multi-temporal satellite imagery processed with the Random Forest (RF) machine learning algorithm in Google Earth Engine. It projected the future changes for the year 2050 by integrating the QGIS MOLUSCE plugin with cellular automata (CA) and artificial neural network (ANN) techniques. While robust machine-learning and cellular-automata methods were applied, uncertainties persist due to classification errors, driver-variable quality, spatial resolution limits, and the assumption that recent transition trends remain stationary in the future. Five LULC classes, built-up, agricultural land, forest, shrubland, and barren land, were identified for 1995, 2010, and 2024. Overall classification performance improved over time, reaching 91.35% overall accuracy and 88.41% Kappa in 2024. Results revealed consistent expansion of built-up and agricultural areas by 197% and 48.44%, respectively, while forest, shrubland, and barren land experienced significant declines of 33.45%, 38.67%, and 60.45% between 1995 and 2024. Projections indicate the continued dominance of agriculture (66.71%), built-up areas rise to 8.09%, and a decline in forest (7.39%) and shrublands (15.49%) (net increases of 14.46% agriculture and 25.62% built-up from 2024; net decreases of 31.43% forest, 28.92% shrubland and − 14.33% barren land). Qualitative result from FGDs and KIIs suggest that the changes are driven by multiple factors, such as agricultural expansion, population growth, fuelwood collection, urban development, charcoal production, informal settlements, and policy interventions. This study provides insights into LULC dynamics in DRC and provides a valuable foundation for sustainable evidence-based planning and watershed management to overcome hazards such as flash flooding as well as sustainable restoration of degraded landscapes of the catchment.