Assessing the Impacts of Land-Use and Land-Cover Change on Climate Variability in Gimbo District, Southwest People’s Region, Ethiopia
No Thumbnail Available
Date
2025-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Land use and land cover change are major drivers of climate variability; however, their interaction is understudied in many regions, including southwest Ethiopia. This study investigates land use and land cover (LULC) dynamics and their influence on climate variability in the Gimbo district, a biodiversity hotspot undergoing rapid land use transitions due to population growth and agricultural expansion. Utilizing multi-temporal Landsat imagery (1989–2024) and TerraClimate datasets (temperature and rainfall), this research integrates remote sensing and statistical methods to assess changes in land surface temperature (LST) and climate trends. Six land use and land cover classes—water bodies (WB), wetlands (WL), forest land (FL), built-up area (BUA), agricultural land (AL), and tea plantation (TP)—were classified using five machine learning algorithms: Classification and Regression Tree (CART), Random Forest (RF), Support Vector Machine (SVM), Gradient Tree Boost (GTB), and K Nearest Neighbor (KNN). Cloud masking and spatial subsetting were applied to preprocess images. Accuracy assessments were conducted using overall accuracy (OAA), kappa coefficient, and F1 scores. LULC change detection was quantified through gain-loss analysis, while climate trends were evaluated using Mann-Kendall tests and Sen’s slope estimator. Pearson correlation and multiple linear regression (MLR) assessed the relationship between LULC change and climate variables. Results indicated that SVM performed best in 1989 (OAA = 84.62%, kappa coefficient = 0.81), 2010 (OAA = 89.59%, kappa coefficient = 0.87), and 2019 (OAA = 91.62%, kappa coefficient = 0.90) classifications, while RF outperformed the others in 2024, achieving an OAA of 93.37% and kappa of 0.92. Overall, SVM maintained the highest average classification accuracy (OAA = 89.11%). The LULC change detection results revealed that wetlands declined by 5.6 percentage points, representing a relative decline of 62% at an annual rate of 2.79%. Forest land saw a decrease of 25.72 percentage points, corresponding to a relative decline of 41% at a rate of 1.49% per year. Agricultural land increased by +28.84 percentage points with a relative rise of 111.5% at 2.14% per year, built-up areas grew by +0.83 percentage points with a 73.45% relative increase at 1.49% per year, and tea plantations increased by +1.67 percentage points with a 293% gain at 3.89% from 1989 to 2024. Climate analysis highlighted a statistically significant increase in mean minimum annual temperature (0.0254 °C/year, p = 0.0000201) and average annual maximum temperature (0.02355 °C/year, p = 0.00442). At the same time, MLR results showed that mean annual maximum temperature was strongly correlated with agricultural land expansion, forest loss, and tea plantation expansion, with p <0.05. Additionally, the average annual land surface temperature (LST) increased by 0.33 °C (P = 0.000062), while the mean annual rainfall exhibited a non-significant trend (0.0255 mm/year, p > 0.05). The study underscores how rapid agricultural expansion and deforestation have intensified local warming, emphasizing the need for climate-smart land management interventions. Finally, it recommends enforcing policies, strategies, and proclamations to conserve wetlands and forests while integrating high-resolution satellite data to enhance future LULC-climate analyses.
Description
Keywords
Land-Use and Land-Cover (LULC), Climate Variability, Machine Learning, Trend Analysis, Multiple Linear Regression