Heat wave risk assessment in Bologna using Spatio-temporal artificial intelligence

Leveraging LSTM enhanced by the hippopotamus optimization algorithm

Datos Bibliográficos

ID21234229
AutoresAniseh Saber (0000-0002-1714-7952, University of Bologna), Claudia De Luca (0000-0001-6836-4298, University of Bologna), Ali Pourzangbar (0000-0002-6053-1669, Karlsruhe Institute of Technology, autor de correspondencia), Simona Tondelli (0000-0003-0891-7852, University of Bologna), Mary Lou Bell (0000-0002-3965-1359, Yale University), Michelle L Bell
Año2025
Volumen131
Páginas106671
Fecha de publicación2025-09-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaSustainable Cities and Society (JOURNAL)
Identificadores de la revistaISSN: 2210-6707 • E-ISSN: 2210-6715
EditorialElsevier BV (PUBLISHER)
DOI10.1016/j.scs.2025.106671
OpenAlexW4413110111
IdiomaEN
Citas recibidas3
Referencias citadas81

An LSTM-Hippopotamus Algorithm framework quantifies heatwave risk in Bologna. • Hazard, exposure, and vulnerability are integrated at tract level from 2014 to 2023. • Risk projections are made under RCP 4.5, 6.0, and 8.5 climate scenarios. • EVI and access to green areas are critical for urban resilience planning. • Findings support Bologna’s SECAP and climate adaptation strategies. This study proposes a spatio-temporal machine learning framework that integrates a Long Short-Term Memory network with the nature-inspired Hippopotamus Optimization Algorithm for Heat Wave Risk (HWR) assessment in Bologna, Italy, where HWR refers to the combined threat posed by the intensity, frequency, and duration of extreme heat events and their impact on exposed and vulnerable populations. We utilized environmental, infrastructure, and demographic data spanning from 2014 to 2023. The heat wave model was developed using 14 key factors covering hazard, exposure, and vulnerability under Representative Concentration Pathways (RCPs) 4.5, 6.0, and 8.5. Inspection of the results revealed temperature, distance to public transportation, local climate zone, and Enhanced Vegetation Index (EVI) as critical factors influencing HWR. Performance evaluations employing statistical indices and a confusion matrix affirm the model’s robust predictive capabilities. The developed model accurately distinguished between risk categories, achieving class-wise accuracies of 77.76 % for the 'Very Low' risk class and 78.72 % for the 'High' risk class, both of which were considered satisfactory. The Partial Dependence Plot approach is employed to interpret the behavior of the developed machine learning model, revealing that high temperatures, high relative humidity, limited vegetation cover, and increased distance from critical services such as healthcare and public transportation strongly influence the predicted HWR. Notably, projections under RCP 8.5 predict a significant increase in 'Very High' risk areas from 34 % in 2023 to 65 % by 2050. The study also confirms that enhancements in EVI and reduced proximities to green areas significantly mitigate HWR. These results emphasize the importance of incorporating targeted green infrastructure in urban planning to enhance resilience against heat waves, providing essential insights for urban planners and policymakers

Algorithm · Biology · Climate change · Heat wave · Hippopotamus · Machine learning · Mathematical optimization · Optimization algorithm · Climate Change and Health Impacts · Computer Science · Mathematics · Urban Heat Island Mitigation · Wind and Air Flow Studies · Artificial Intelligence · Ecology