Abstract

Cities are exposed to higher temperatures due to the extensive presence of man made heat sources and climate change, the latter resulting in the increase of the duration, strength and frequency heat waves. Estimating heat risk through a synthetic methodology that combines satellite images and ground data and uses machine learning for cluster analysis based on landscape, urban and social parameters is important in order to (a) recognize which areas and which populations are disproportionately exposed to heat risk (“extreme heat injustice”), especially in the event of heat waves (b) to define measures to mitigate heat risk and (c) to support urban resilience to climate extremes.

Project Description & Overview

The scope of the project is to improve urban resilience to climate extremes by mapping heat risk as a function of local climate variability, urban form and fabric, and demographic and social data that reflect socio-economic sensitivity at the neighbourhood scale. For this be accomplished, analysis of satellite data of high spatial resolution will provide maps of the surface urban heat island which along with ground data on air temperature will support the extraction of hot spot areas in the urban web. 

At a second stage, cluster analysis, with the use of Machine Learning (through the Google Earth Engine), will be performed to define which areas and which populations are disproportionately exposed to heat risk (“extreme heat injustice”), especially in the event of heat waves. Cluster analysis will take note of a selection of parameters, among others on temperature, building type and density, type and density of population, % of greenery, proximity to green areas, income, etc.

Following, landscape mapping will indicate locations where changes in the urban environment can mitigate heat risk and heat injustice: for instance, areas where impervious materials predominate could be landscaped with trees, pocket parks, and water features; areas where buildings are old and more sensitive to heat may be provided with subsidies for energy retrofitting; and buildings could develop green roofs or use cool material. In this way, cities improve their resilience to climate extremes. The project will concentrate to New York and Athens, although more cities may be added depending on the availability of data. 

Datasets

• Satellite images from Landsat, Aqua and Terra, Sentinel 2 and 3 (public and free)
• Ground data on temperature (city data bases, ERA-5)
• Census data on economy, housing, demographic dynamics, etc.

Competencies

The skill of quantitative analysis will be useful. Knowledge of Google Earth Engine and GIS will support the completion of the project. If there is no such knowledge, the sponsor will organize training sessions for the Capstone team. 

Learning Outcomes & Deliverables

A learning outcome of the project is website showcasing heat risk mapping. The prospect of developing a web=GIS will be also exploited.

Results:

Machine Learning Mapping of Heat Risk in Cities Using Satellite and Ground Data in Google Earth Engine

Students

Ahmad Nasrieh, Natalie Quah, Jiaxi Xu, and Xingjian Zhao