Heatwaves risk assessment for Australia using variants of the Cox model

Jason West 1 Bureau of Meteorology, Australia

Search for other papers by Jason West in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0003-3271-3155
Restricted access

Abstract

We investigated the influence of key factors on the hazards associated with the occurrence of temporal clustering of heatwaves using variants of the Cox regression model. The use of survival analysis as a statistical framework to infer characteristics of probability distribution functions is a powerful method for modelling the full distribution of climate variables rather than confining the analysis to simple statistical parameters like mean and variance. This approach allows for the assessment of the probability of extremes that are more informative than central tendency measures and constitutes the core of probabilistic risk assessments. Variants of the Cox model account for non-stationary patterns resulting from external influences such as climate change for situations where events do not depend on the observed time from last occurrence, nor on the number of events previously observed. We apply this concept to the onset of heatwaves to assess their effect by location and time. We use observed maximum daily temperatures across 28 urban and rural areas in Australia for the period 1956–2022, along with monthly observations for the Oceanic Niño Index (ONI) sea surface temperature anomalies and the Indian Ocean Dipole (IOD) sea surface temperature index as climate covariates to estimate the probability of heatwave occurrence over time after an initial time reference. We find that location (urban/rural) and climate covariates derived from ONI and IOD are significant indicators of heatwave hazards, while epoch (pre-1980/post-1980) exhibited only a moderate difference. Despite the urban heat island effect, urban locations are less prone to extremes in heatwave frequency than rural locations.

© 2025 American Meteorological Society. This is an Author Accepted Manuscript distributed under the terms of the default AMS reuse license. For information regarding reuse and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding Author: Bureau of Meteorology, 16/32 Turbot St, Brisbane, QLD, 4000, Australia. Email: jason.west@bom.gov.au P: +61 438 389 162.

Abstract

We investigated the influence of key factors on the hazards associated with the occurrence of temporal clustering of heatwaves using variants of the Cox regression model. The use of survival analysis as a statistical framework to infer characteristics of probability distribution functions is a powerful method for modelling the full distribution of climate variables rather than confining the analysis to simple statistical parameters like mean and variance. This approach allows for the assessment of the probability of extremes that are more informative than central tendency measures and constitutes the core of probabilistic risk assessments. Variants of the Cox model account for non-stationary patterns resulting from external influences such as climate change for situations where events do not depend on the observed time from last occurrence, nor on the number of events previously observed. We apply this concept to the onset of heatwaves to assess their effect by location and time. We use observed maximum daily temperatures across 28 urban and rural areas in Australia for the period 1956–2022, along with monthly observations for the Oceanic Niño Index (ONI) sea surface temperature anomalies and the Indian Ocean Dipole (IOD) sea surface temperature index as climate covariates to estimate the probability of heatwave occurrence over time after an initial time reference. We find that location (urban/rural) and climate covariates derived from ONI and IOD are significant indicators of heatwave hazards, while epoch (pre-1980/post-1980) exhibited only a moderate difference. Despite the urban heat island effect, urban locations are less prone to extremes in heatwave frequency than rural locations.

© 2025 American Meteorological Society. This is an Author Accepted Manuscript distributed under the terms of the default AMS reuse license. For information regarding reuse and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding Author: Bureau of Meteorology, 16/32 Turbot St, Brisbane, QLD, 4000, Australia. Email: jason.west@bom.gov.au P: +61 438 389 162.
Save