Main content area

Short-term prediction of extremely hot days in summer due to climate change and ENSO and related attributable mortality

Chen, Chu-Chih, Wang, Ying-Ru, Guo, Yue-Liang L., Wang, Yu-Chun, Lu, Mong-Ming
The Science of the total environment 2019 v.661 pp. 10-17
El Nino, autocorrelation, climate change, comparative risk assessment, metropolitan areas, models, mortality, prediction, public health, summer, temperature, Taiwan
Summer days with extremely hot temperatures in Taiwan have been increasing for the past few decades, and this continuing trend is expected to worsen heat-related mortality. To mitigate the corresponding health impacts, in this study, we developed a statistical state-space model to predict the number of extremely hot days in June–September for the next year. Based on historical data from 1951 to 2017, we estimated the climate change trend after adjusting for the nonlinear lagged effect of the Niño 3.4 index. We then developed a predictive state-space model using these two primary factors and adjusting for residual autocorrelations. Validation results comparing the extremely hot days observed over 2015–2017 with predictions showed that 86% of the average prediction errors were within 4 days of the observations. To assess the health impacts, we applied the model to the projection of heat-attributable mortality (AM) in 2018 by adopting a comparative risk assessment (CRA) approach with the reference period of 2001–2010. The results showed that the Taipei metropolitan area in northern Taiwan is the most affected region with AM of 1501 deaths from all-causes, followed by Taichung in central Taiwan with 490 deaths. The prediction model and the CRA projection provide both a tool and guidance for public health administrators to address the imminent threat posed by climate change.