In the project we applied machine/deep learning algorithms to Sentinel 2 images, in order to estimate the temperature of the Earth’s surface. By combining this data with high-resolution land cover maps and meteorological data we generated a risk map for the Urban Heat Island phenomenon, based on a multi-temporal analysis. As an end product we are able to nature-based solutions to mitigate the phenomenon. The project was validated in the cities of Naples and Milan. The expected processing is:
Albedo is the solar reflectance coefficient, which measures the percentage of radiation that a material reflects compared to the total incident solar radiation. The greater or lesser reflected radiation and the consequent lesser or greater absorption influence the temperature of a body.
Land Surface Temperature
The 10m Layer (LST) (High Spatial Resolution Land Surface Temperature map) is obtained by applying Latitudo 40’s proprietary machine learning algorithms to the Copernicus Sentinel-2 images, in order to obtain a map of the ground surface temperature to be used for the study of the Urban Heat Islands phenomenon. The map has a spatial resolution of 10 m and is provided with a frequency of no less than one survey per week.
Surface Urban Heat Island
Surface Urban Heat Islands refer to localized areas within urban environments that experience higher temperatures than surrounding rural or natural landscapes. These temperature differences are due to a combination of factors, including the alteration of land cover, the high concentration of buildings and infrastructure, and the reduction of vegetation and green spaces in cities.These maps show the occurrence of the Surface Urban Heat Island (SUHI) effect on the city’s territory. This processing describes the average difference in Land Surface Temperature observed in a specific period (at least monthly), between each location on the map and a reference temperature value, in order to visualize and evaluate the intensity of the SUHI. It is important to point out that UHI usually refers to the difference in air temperature between rural and urban areas during the night, when the latter is minimal.
Land Use Land Cover
This processing provides a classification of the types of territory subject to the analysis (e.g. meadows, buildings, trees, bodies of water, roads, etc.) and can be obtained as a manual assignment of classes based on the information present in the databases of the institutions responsible for governing the territory of interest, or using automatic image classification techniques.
Microclimatic Performance of Green Infrastructures
The role of urban vegetation in mitigating the urban heat island (UHI) effect is fundamental. Studying the cooling capacity of vegetation at the urban level and recognizing the deployment of urban green infrastructure (UGI) is a key strategy to promote a resilient environment and mitigate the UHI effect.
Park Cool Island
The Park Cool Island phenomenon can be defined as a pattern of cooler areas nestled within generally warmer urban areas. The objective of the product is to identify the best performing green areas in terms of the Park Cool Island (PCI) effect, classifying them based on size and relevant characteristics.
Heatwave Potential Risk
The elaboration of the Risk Map is generated by considering the various dangers such as Surface Urban Heat Islands, the exposure of citizens such as schools, hospitals etc., and vulnerability such as surfaces.
Urban Heat Island Vulnerable Areas
In this elaboration the most vulnerable areas of the city are represented. As can be seen from the graphic representation, this map is derived from the Surface Urban Heat Island maps and from the Heatwave Potential Risk Map.
The layer shows the capacity of the city’s green infrastructure to absorb CO2. This processing is carried out starting from a very high resolution land cover map to which the estimated surface biomass, surface biomass, subsurface biomass, soil characteristics and estimated dead organic matter are added in the application of one specific proprietary machine learning model.