Key points are not available for this paper at this time.
Two industrial cities in Odisha, India, namely, Rourkela and Angul, were analyzed using the machine learning–based Google Earth Engine (GEE) platform employing algorithms like random forest (RF), naïve Bayes, classification and regression tree, and k-nearest neighbor. A comparative analysis among the machine learning models indicated RF to be the best-performing model, with a minimum root-mean-square error of about 0.382, and it was thus adopted for further land-use–land cover (LULC) mapping. Continuing with RF and the data set obtained from Landsat 8 and Sentinel 2 satellites, atmospheric corrections with cloud coverage < 10% were used to analyze the LULC maps for land surface temperature, urban heat islands (UHI), and urban thermal field variance index (UTFVI), from 2016 to 2023. Using the GEE data as input for ArcMap, multiple LULC maps with preatmospheric corrections obtained through GEE were clipped for observation. From 2016 to 2023, LULC classes, like the existence of water and trees, were reduced by 26.4% and 35.38% for Angul, while by 7.42% and 18.77% for Rourkela in response to an increment in built-up areas by 35.53% and 21.47% for the same in the respective order. The corresponding effect was increased in UHI sites with middle to stronger UTFVI conditions, signifying the degradation of the local ecology. Continued conversion of UHI green zones to yellow and red zones between 2016 to 2023 implicates a prediction where, by 2030, an increment in built-up areas of about 16.75% and 11.45% and a reduction of water by 24.8% and 14.76% for Angul and Rourkela, respectively, will significantly aid the UHI zones. The present work, thus, helps in laying the groundwork for policy reformations to minimize UHI effects caused by administered urbanization and preserve water for future generations.
Nayak et al. (Tue,) studied this question.