Research Article: The use of climate information to estimate future mortality from high ambient temperature: A systematic literature review

Date Published: July 7, 2017

Publisher: Public Library of Science

Author(s): Michael Sanderson, Katherine Arbuthnott, Sari Kovats, Shakoor Hajat, Pete Falloon, Juan A. Añel.


Heat related mortality is of great concern for public health, and estimates of future mortality under a warming climate are important for planning of resources and possible adaptation measures. Papers providing projections of future heat-related mortality were critically reviewed with a focus on the use of climate model data. Some best practice guidelines are proposed for future research.

The electronic databases Web of Science and PubMed/Medline were searched for papers containing a quantitative estimate of future heat-related mortality. The search was limited to papers published in English in peer-reviewed journals up to the end of March 2017. Reference lists of relevant papers and the citing literature were also examined. The wide range of locations studied and climate data used prevented a meta-analysis.

A total of 608 articles were identified after removal of duplicate entries, of which 63 were found to contain a quantitative estimate of future mortality from hot days or heat waves. A wide range of mortality models and climate model data have been used to estimate future mortality. Temperatures in the climate simulations used in these studies were projected to increase. Consequently, all the papers indicated that mortality from high temperatures would increase under a warming climate. The spread in projections of future climate by models adds substantial uncertainty to estimates of future heat-related mortality. However, many studies either did not consider this source of uncertainty, or only used results from a small number of climate models. Other studies showed that uncertainty from changes in populations and demographics, and the methods for adaptation to warmer temperatures were at least as important as climate model uncertainty. Some inconsistencies in the use of climate data (for example, using global mean temperature changes instead of changes for specific locations) and interpretation of the effects on mortality were apparent. Some factors which have not been considered when estimating future mortality are summarised.

Most studies have used climate data generated using scenarios with medium and high emissions of greenhouse gases. More estimates of future mortality using climate information from the mitigation scenario RCP2.6 are needed, as this scenario is the only one under which the Paris Agreement to limit global warming to 2°C or less could be realised. Many of the methods used to combine modelled data with local climate observations are simplistic. Quantile-based methods might offer an improved approach, especially for temperatures at the ends of the distributions. The modelling of adaptation to warmer temperatures in mortality models is generally arbitrary and simplistic, and more research is needed to better quantify adaptation. Only a small number of studies included possible changes in population and demographics in their estimates of future mortality, meaning many estimates of mortality could be biased low. Uncertainty originating from establishing a mortality baseline, climate projections, adaptation and population changes is important and should be considered when estimating future mortality.

Partial Text

Warming of the Earth’s climate is now unequivocal; global average temperatures have risen by 0.85°C between 1880 and 2012 [1]. Increases in temperature over land areas are almost always higher than global average increases but vary between different regions of the Earth [2,3]. The frequency of heat waves has also increased in many continents [1]. Global mean temperature is projected to increase by about 1.6 to 2.6°C above the preindustrial period by the 2050s, depending on the scenario used [1]. Using median values, projected temperature increases for Europe and America are between 2 and 4°C for the 2050s (relative to present-day climate). Higher increases are projected over much of Asia and Australia [3].

Initially, 608 articles were identified from the literature searches after duplicate entries had been removed. Thirteen additional articles were selected from examination of reference lists and citing literature. After screening the titles and abstracts, the full texts of 130 articles were examined in detail. Of these articles, 63 were found to contain quantitative estimates of future heat-related mortality and so were selected for the systematic review (Fig 1). The locations studied, time periods, climate models, emissions scenarios and treatment of adaptation are summarised in Table 1, together with the meteorological variables used in each study. Further technical details of each study, specifically the variables used to model mortality, calibration methods, time of year considered and consideration of changes in population are shown in Table 2. The locations of the cities studied are shown in S1 Fig, and the time periods considered in each article are illustrated in S2 Fig.

Sixty three papers estimating future mortality from high temperatures and heat waves have been reviewed. These studies used a wide range of surface observations and climate model projections (Tables 1 and 2). All the studies indicate that heat-related mortality would increase under a warming climate. The projected impacts of climate change on mortality are highly dependent on the future scenarios and climate models chosen. The majority of studies have used a small number of climate simulations without considering where they lie within the range of future projections. The difficulty in obtaining and processing data from climate models is likely to have been one factor in older studies, although access to climate model data has greatly improved in recent years. The use of climate information from a small number of models means future mortality estimates could be biased low or high. Ideally, all available simulations would be used to estimate future mortality. Alternatively, a subset of the climate models could be selected which captures key regional climate processes and the range of possible changes in climate [18,53]. Calibration of data from climate models is required to reduce the impacts of any biases. It should be noted that no calibration method will remove all deficiencies in the modelled climate, such as over- or under-prediction of the persistence of periods of hot and cold weather.

Heat-related mortality is now recognised as a serious issue which is likely to increase in severity as the climate warms. Studies of future mortality would benefit from more interdisciplinary collaborations to improve the quality of research and results and also to broaden the interest and readership of work that is likely to be important for public policy across a number of sectors. There are very few studies of the effects of warm temperatures on mortality outside of high-income countries. More studies are needed in middle and low income countries, and in sub-tropical and tropical areas. Additionally, there are very few studies specifically estimating future mortality in rural populations.

There are many different factors to be considered when estimating future heat-related mortality. Much depends on what data are available for any given area and the purpose of the research, for example, exploring worst case scenarios and impacts of different policy options. One important recommendation from this review is transparency in reporting, ensuring the data used fit the purpose of each study and any limitations are reported.