Urban Morphology, Climate, and Automated Land Use Classification approach in Latin America and Caribbean

The morphology of cities and their relationship with the urban climate has been extensively studied, constituting a major challenge at the urban scale for adaptation to climate change and global warming. This is particularly relevant due to the increasing frequency and intensity of heat waves and other extreme events that disproportionately affect the most vulnerable populations.

At the urban scale, it is essential to recognize the existence of morphoclimatic domains that create unequal living conditions based on socioeconomic status. In Latin America, and particularly in Chile, urban morphological typologies differ from the Local Climate Zones (LCZs) proposed by Steward & Oke (2012). This study presents an alternative classification of these typologies to discuss their unique characteristics, main differences, and their relationship with topographic features.

To characterize these differences, we analyze surface urban heat and cool islands using satellite data from Landsat (daytime) and MODIS and Sentinel-3 (nighttime). A major limitation in LCZ classification for Latin America and the Caribbean is the lack of labeled datasets. To address this challenge, this study explores automated detection approaches using Sentinel satellite imagery and classification models based on convolutional neural networks (HRNet, U-Net, and ResNet) using existing labeled data from Valparaíso and Santiago de Chile, two cities that have yielded promising results, revealing relationships between the 10 constructed LCZ classes and 7 socioeconomic groups ranging from high to middle and low status.

These results are discussed with the aim of refining automated classification methods that can be scaled and replicated across Latin America and the Caribbean. This research runs in parallel with the development of the Copernicus LAC-Chile Urban Atlas, one of the key initiatives aiming to provide an accurate representation of the region’s urban landscape.