Spatio-Temporal Modeling of Land-Use/Land-Cover Change and Land Surface Temperature Using SVM and CA-Markov in Dilla...

Urban expansion in Dilla Town, Ethiopia, has progressively reshaped local thermal conditions by replacing vegetation and cropland with impervious built-up surfaces. This study analyses how three decades of land-use/land-cover (LULC) change have modified land surface temperature (LST) and projects future thermal patterns. Landsat TM/ETM+/OLI imagery from 2001, 2011, and 2021 was classified into five LULC classes using a Support Vector Machine (overall accuracy 90.4-92.4%; κ = 0.88-0.90). LST was retrieved from thermal bands via a single-channel algorithm and related to NDVI and the Normalized Difference Built-up Index (NDBI). Between 2001 and 2021, the built-up area increased by ∼5.34 × 10⁶ m², while mixed forest and cropland decreased by 3.14 × 10⁶ m² and 2.92 × 10⁶ m², respectively. Over the same period, LST rose from 23.17°C to 33.80°C to 24.42°C to 40.18°C, and mean LST increased from 27.47°C to 30.61°C. In 2021, LST showed a strong positive correlation with NDBI (R² ≈ 0.64) and a negative correlation with NDVI (R² ≈ 0.36). A CA-Markov model, validated with a Kappa Index of Agreement of up to 0.95, predicts continued urban expansion and a marked shift toward hotter LST classes by 2031, emphasizing the need for greening, wetland conservation, and heat-aware urban planning.