Machine learning-based assessment and simulation of land use modification effects on seasonal and annual land surface temperature variations

Mudassir Khan; Muhammad Qasim; Adnan Ahmad Tahir; Abida Farooqi

doi:10.1016/j.heliyon.2023.e23043

Machine learning-based assessment and simulation of land use modification effects on seasonal and annual land surface temperature variations

Heliyon. 2023 Nov 29;9(12):e23043. doi: 10.1016/j.heliyon.2023.e23043. eCollection 2023 Dec.

Authors

Mudassir Khan¹, Muhammad Qasim², Adnan Ahmad Tahir³, Abida Farooqi¹

Affiliations

¹ Department of Environmental Sciences, Quaid-i-Azam University, Islamabad, Pakistan.
² Department of Environmental and Conservation Sciences, University of Swat, Pakistan.
³ Department of Environmental Sciences, COMSATS University Islamabad (CUI), Abbottabad Campus, Abbottabad 22060, Pakistan.

Abstract

Rapid urban sprawl adversely impacts the local climate and the ecosystem components. Islamabad, one of South Asia's green and environment-friendly capitals, has experienced major Land Use Land Cover (LULC) changes over the past three decades consequently, elevating the seasonal and annual Land Surface Temperature (LST) in planned and unplanned urban areas. The focus of this study was to quantify the fluctuations in LULC and LST in planned and unplanned urban areas using Landsat data and Machine Learning algorithms involving the Support Vector Machine (SVM) over the 1990-2020 data period. Moreover, hybrid Cellular Automata-Markov (CA-Markov) and Artificial Neural Network (ANN) models were employed to project the future changes in LULC and annual LST, respectively, for the years 2035 and 2050. The findings of the study reveal a distinct difference in seasonal and annual LST in planned and unplanned areas. Results showed an increase of ∼22 % in the built-up area but vegetation and bare soil decreased by ∼10 % and ∼12 %, respectively. Built-up land showed a maximum annual mean LST followed by bare-soil and vegetative surfaces. Seasonal analysis showed that summer months experience the highest LST, followed by spring, autumn and winter. Future projections revealed that the built-up areas (∼27 % in 2020) are likely to increase to ∼37 % and ∼50 %, and the areas under the highest annual mean LST class i.e., ≥28 °C are likely to increase to ∼19 % and ∼21 % in planned, and ∼38 % and ∼42 % in unplanned urban areas for the years 2035 and 2050, respectively. Planned areas have better temperature control with urban green spaces, and controlled infrastructure. The Capital Development Authority of Islamabad may be advised to control the expansion of built-up areas, grow urban forests, and thus mitigate the possible Urban Heat Island (UHI) effect.

Keywords: Artificial neural network; Cellular automata-markov; Machine learning; Urban forest; Urban heat island; Urbanization.