Object-based Local Climate Zone classification to investigate spatiotemporal dynamics of Surface Urban Heat Islands in temperate cities

The Local Climate Zones (LCZ) classification system has been applied in Brazil's southernmost capital, utilizing Geographic Object-Based Image Analysis segmentation to investigate spatiotemporal Surface Urban Heat Island phenomenon, represented here by area of elevated surface temperature across all seasons for the seasonal LST composites (2000/2004, 2010/2011, and 2019/2022) associated with the LCZ reference years 2002, 2010, and 2023. Time-series temperature averages for each season were derived from Landsat images, while high-resolution images and the Random Forest algorithm were employed for mapping. Main results identified LCZ classes 2, 3, 6, A, B, C, E, and G in the study area, with the first three belonging to the built category and the remainder to the land cover category. In addition, LCZ 6 exhibits a noteworthy increase in its total area by 42.18% from 2010 to 2023, while LCZ A experienced a decline of 5.2 percentage-points from 2002 to 2010, a decrease of 21.22%. Furthermore, temperature data from LCZ 3 reached approximately 40 °C in in the most recent LST composite period associated with the 2023 LCZ map, marking the highest average temperature observed during the summer. The classification accuracy of the 2023 LCZ map was deemed satisfactory at 85.7%, and statistical analyses confirmed significant differences in temperature averages. Consequently, the proposed methodology proves to be effective for detecting Land Surface Temperatures in regions with diverse characteristics, facilitating the identification of areas that require mitigation measures for high temperatures, and aligning with Sustainable Development Goal 11.