In consultation with the National Institute of Meteorological Research, we selected climate models and emissions scenarios to develop a set of climatologies for each of the cities, including new synoptic air mass calendars for each. We concentrated upon the frequency changes of those air masses that are historically associated with increased mortality during EHEs, specifically the dry tropical (DT) and moist tropical plus (MT+ and MT++) air masses. We compared the present numbers of offensive air mass days to those calculated using he scenarios for the two target decades in the 21st century. We compared summers for the 2000s (specifically the summers of 2006-2012) for each Korean city to see how they measured up to the numbers of health-debilitating air mass days under the various scenarios. We looked at consecutive day strings of such air masses (3 and 5 days) to see if they are becoming more frequent using the various scenarios.
We also estimated how heat-related mortality will be impacted under the various scenarios for Seoul. For example, we know that, at present, Seoul mortality increases by an average of 7 percent during DT and MT+ days. We also know, from previous collaborative research with NIMR, that the DT air mass occurs during an average summer on 7.3 percent of days in Seoul, and the MT+ average occurrence is 4.9 percent. Considering these frequencies, we have estimated that there have been over 1100 heat-related deaths during the 16 year period between 1991 to 2006, which averages to about 70 deaths per summer. Once we determined the frequency of these oppressive air masses under the various modeled scenarios, we then developed estimates of heat related mortality for Seoul under these scenarios.
We worked with two separate modeled climate scenarios, which we utilized for two different decades during the 21st century. The two scenarios, RCP 2.6 and RCP 8.5, were performed using the HadGEM2-AO global climate model, developed by the UK Met Office. The RCP 2.6 and 8.5 scenarios represent the most conservative and most aggressive scenarios in the suite, with the former assuming 420ppm of carbon dioxide by 2100 and the latter assuming 940ppm. Mean temperature increases over the next century for the RCP2.6 scenario are about 1/3 that forecast by the RCP8.5 scenario, and increases in precipitation are about half the magnitude in the former than in the latter.
An evaluation of the modeled air mass frequencies for three time periods, 2006-2012, 2041-2050, and 2091-2100, shows how the most oppressive air masses become more dominant, while the benign air masses become less prevalent. For the Busan RCP 2.6 scenarios for those time periods, it is clear that the DT, MT+, and MT++ air masses increase tremendously through the 21st century. DT, a very uncommon air mass in Busan, increases from less than 1 percent in the early 21st century to almost 3 percent by the end of the century. The increases for MT+ and MT++ are even more dramatic; the increase of MT++ is almost fortyfold in Busan, from 0.24 percent to 9.33 percent in the 2090s. For the RCP 8.5 scenario, we see an even more dramatic change in oppressive air mass frequency.
As expected, the estimated heat-related mortality in Seoul under the various scenarios shows sharp and dangerous increases. For the decade of the 2040s, the average number of excess deaths increases by about fivefold for both the RCP 2.6 and RCP 8.5 scenarios. Even more dramatic are the values for the highest years; about half the years exceed 100 deaths by a sizable margin under RCP 2.6, and almost half the years exceed 200 deaths under RCP 8.5, with two years having estimates of 400 deaths or more.
There is one very important caveat to these results. There is a significant issue involving model bias. There is no doubt that the model's inability to duplicate the observed air mass frequencies properly lessened the robustness of the results. We compared the historical modeled data over the same period (2006-2012) to the observed, with the goal of seeing how different the general frequencies in air mass type were. Some sizable differences were found. For example, the observed percent of summer days within the DT category for Daegu was 22.4; the modeled percent for the RCP 2.6 scenario was 1.31. This is clearly an unacceptable difference, particularly when we are depending upon frequencies of offensive air masses to estimate heat-related mortality. At this time, we are working with NIMR to greatly reduce the bias in the models and to redevelop the results. We are very confident that the new results will be available by the time we present this research at ICB.