Harnessing geospatial and machine learning technologies to simulate ecosystem services and greenhouse gas emissions in a dynamic watershed system of Haramaya Lake, Ethiopia

Abstract

The increased land use change, coupled with weak regulatory mechanisms, has increased the deterioration of Haramaya lake Watershed, thus contributing to ecosystem service value (ESV) decline and greenhouse gas (GHG) emissions. However, the information about the observed and predicted impacts of land use changes on the ESV changes and GHG emissions in the Haramaya Lake Watershed is limited. This study thus followed an interdisciplinary approach, integrating three pressing global change factors: land use/land cover (LULC) change, ESV changes, and GHG emissions to understand their historical and future impacts on Haramaya Lake Watershed using machine learning. Support Vector Machine (SVM) algorithm was used for historical LULC change analysis (1985–2024) and a machine learning-based Artificial Neural Network (ANN) was employed for simulation (2063). The ESVs changes were evaluated using the benefit transfer method based on the LULC data for the study periods and corresponding modified ESV coefficients derived from empirical studies. To estimate GHG emissions, the study followed the Inter-governmental Panel on Climate Change’s (IPPC) National GHG Inventories guideline. The results revealed a significant LULC shift between 1985 and 2024, characterized by a marked expansion of Khat (+ 9.98%) and built-up area (+ 5.03%) at the expense of cropland (-10.99%) and vegetation (-5.45%). Under a business-as-usual scenario simulated by the validated ANN model, these pressures are projected to continue, with cropland expanding to cover 45.22% and built-up area 20.22% of the watershed by 2063. The model predicts a decline in the wetland from 5.63% to 4.59%. The overall ESV was increased by 2.5% between 1985 and 2024, but projected to decline by 41.95% by 2063. The carbon sequestration capacity of Haramaya Lake Watershed exhibited a declining trend. Except for vegetation land use class, other land use types primarily contributed and projected to contribute to carbon emissions. The results suggest immediate implementation of integrated watershed management strategies, including land use zoning, sustainable agricultural practices, and alternative livelihood strategies.