Projecting extreme heat-related mortality in Europe under climate change

Abstract

The assessment of health impacts of extreme hot weather under climate change is important for adaptation and mitigation actions. This thesis has developed techniques for estimating changes in heat-related mortality in Europe, with a focus on extreme daily mortality counts. The use of these techniques is illustrated through the projections of extreme elderly mortalities for London, UK and Budapest, Hungary from 2010 to 2099, using temperature projections from the perturbed physics ensemble of the regional climate model HadRM3.

The present-day relationship between daily number of deaths and temperatures at each location is modelled by Poisson generalized additive models. In order to account for possible discrepancies in climate model simulations, temperature projections from HadRM3 are calibrated by two approaches, bias correction and change factor. These are based on assumptions on the relationships in location, scale and shape between observed and modelled temperature distributions. In particular, a novel method using the Box-Cox transformation is developed to correct the bias in the upper tails of present-day simulated temperature distributions. Finally, future mortalities are projected by driving the mortality models with calibrated temperature projections.

Results of temperature calibration show that the two calibration approaches give substantially different estimates of future extreme temperatures. The estimates of 10-summer temperature return level by the two approaches differ by more than 4 degrees C over many parts of Europe in the period 2070 to 2099. For London and Budapest, the effect of this calibration uncertainty on extreme temperature projections is comparable to the effect of the uncertainty in climate model parameters which is estimated by the range of perturbed physics ensemble estimates. These two sources of uncertainties, together with the uncertainty in how the mortality-temperature relationship is modelled, contribute to large uncertainties in extreme mortality projections. Assuming constant elderly population in the future, the projected change in the 2-summer return level of number of daily elderly deaths in the period 2070 to 2099 relative to the the present-day ranges from -12% to +75% for London and from -16% to +22% for Budapest.