Author ORCID Identifier:
Date of Graduation
5-2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Environmental Dynamics (PhD)
Degree Level
Graduate
Department
Environmental Dynamics
Advisor/Mentor
Aly, Mohamed
Committee Member
Ruhl-Whittle, Laura
Second Committee Member
Cheng, Linyin
Keywords
Arkansas; Climate Change; Land Surface Temperature; Land Use/Cover Change; Machine Learning; Urbanization
Abstract
Urbanization is rapidly transforming the landscape of Arkansas, intensifying land surface temperature (LST) and exacerbating urban heat island (UHI) effects across the region. This dissertation presents a comprehensive, three-part investigation into the interactions between land use and land cover (LULC) change, LST dynamics, and urban thermal risk using remote sensing data and machine learning methods. The research focuses on three urban centers in Arkansas—Fort Smith, Little Rock, and Northwest Arkansas, covering the period from 2001 to 2031. The first study provides a systematic review of 81 peer-reviewed articles that apply machine learning to model the relationship between LULC and LST. Following PRISMA guidelines, the review identifies common modeling techniques, key spectral indices (e.g., NDVI, NDBI), spatial and temporal patterns, and methodological gaps. It highlights the dominance of Landsat-based studies, the popularity of Random Forest and ANN for prediction tasks, and a growing interest in integrating LULC-LST models into urban climate resilience planning. The study also emphasizes the importance of model interpretability and the need for spatially explicit forecasts in underrepresented regions. Building on this conceptual foundation, the second study applies the CA–Markov model to project LULC changes in the three Arkansas cities between 2001 and 2031. Using reclassified NLCD data and transition probability matrices, the study simulates future land conversion patterns. Results show substantial urban expansion, particularly in Northwest Arkansas, at the expense of vegetation and agricultural land. Model accuracy reaches 91.9%, confirming the suitability of CA–Markov for regional LULC forecasting. The projections reveal spatial patterns of urban sprawl that form the basis for later surface temperature modeling. The third and final study leverages Landsat imagery and XGBoost regression to predict seasonal LST for 2026 and 2031. Using biophysical predictors—including NDVI, NDBI, brightness temperature, and spatial coordinates—the model achieves high predictive accuracy (R² = 0.74–0.78; RMSE ≤ 1.46 °C). The results indicate significant thermal intensification across all cities, with the 40–45 °C temperature range projected to dominate by 2031. Drought indices (SPI and SMI) are also incorporated to assess moisture stress trends and their influence on urban thermal exposure. The findings identify high-risk heat zones and support targeted interventions such as green infrastructure, zoning reform, and reflective surface deployment. Overall, this dissertation integrates spatial modeling, remote sensing, and machine learning to deliver a data-driven framework for urban heat risk assessment and climate-resilient planning in mid-sized American cities. The methods and insights developed herein are scalable to other rapidly urbanizing regions facing similar environmental challenges.
Citation
Vahid Belarghou, R. (2026). Modeling Urban Expansion and Heat Risk in Arkansas Using Remote Sensing and Machine Learning: A Multi-Scale Assessment of Land Use Change and Surface Temperature Dynamics. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/6206
