Mortality algorithm development
Correlating our mortality data with the meteorological data for the offensive air masses (DT, or dry tropical, MT + , or moist tropical +), we arrived at the following algorithm for Los Angeles County utilizing only statistically significant variables at the p < 0.05 level:
$$\% MORT = -1.426 + 0.363 NFPTS + 5.219 DT + 1.609 MT + 0.057 AT05$$
where % MORT is the percent change in mortality from the baseline value (we consider this heat-related mortality), NFPTS is the Nairn-Fawcett Extreme Heat Factor (Nairn and Fawcett 2014), which evaluates heat in three consecutive day increments and determines whether the period before the heat wave has been hot or comfortable (which could have a significant difference on response), DT is a dummy variable which is added just for the DT air mass days, which is a highly transparent air mass with minimum cloud cover and significant solar radiation income, MT is a dummy variable for MT + which is added just for the MT + days, and AT05 is 5AM apparent temperature. Utilizing this algorithm, which is based upon DT and MT + heat events spanning the period 1985–2010, we noted that during an average 5-day Los Angeles County EHE during this timeframe, excess mortality is 4.1% on the first day of the event. In the case of a 5-day heat event, like two of the EHEs we evaluated, it increases to 11.9% on the fifth day of the event.
This mortality algorithm was applied to the baseline (or control) meteorological conditions during the four evaluated heat events, as well as the four cases of mitigation scenarios presented in Table 4.
Meteorological and health impacts of the mitigation scenarios
We saw clear changes in air temperature and dewpoint temperature across all four heat events utilizing the four mitigation scenarios, consistent with studies conducted in other cities (Kalkstein et al. 2020). Temperatures mostly showed decreases in the range of 1–2 °C (1.8–3.6 °F; Table 5 and 6), while dewpoint temperatures showed similar increases in magnitude. Cases 1 and 3, which have more modest urban tree cover increases than the other scenarios, show smaller changes than cases 2 and 4 with the most aggressive tree canopy increases. This is intuitive, since added tree cover would add water vapor into the atmosphere through evapotranspiration, thus increasing the dewpoint temperature. Nevertheless, in general, the largest decreases in temperature also occur in cases 2 and 4, sometimes exceeding 2.5 °C (4.5 °F), especially during the August 2009 event (Table 6), when maximum temperatures reached 36–40 °C (97–104 °F), depending on the location.
Table 5 Changes in meteorology for the June 2008 heat wave. All four scenario cases are presented. Delta T is the change in temperature (°C) from the baseline. Delta Td is the change in dewpoint temperature (°C) from the baseline. Increasingly dark (blue) color represents greater reductions; increasingly dark (orange) color in bold italics represents greater increases Full size table
Table 6 Changes in meteorology for the August 2009 heat wave Full size table
Besides evapotranspiration increases, some of the increases in dewpoint temperature are physically attributed to the cooling temperatures themselves, especially for cases 2 and 4. When temperatures are cooled, vertical motion of the atmosphere is inhibited, and the dispersal of near-surface moisture is therefore less efficient. Thus, moisture from sources such as car exhaust, air conditioning, and even from trees is less likely to be dispersed vertically and more likely to accumulate near the ground.
However, the decrease in air temperature is more important in terms of human well-being than the accompanying increase in dewpoint temperature. The apparent temperature (National Oceanic and Atmospheric Administration 2021), which is the perceived temperature by humans and represents the combined impacts of thermal and moisture characteristics in the atmosphere (sometimes called the “heat index”) is impacted more by a drop in temperature than a rise in dewpoint temperature. For example, an air temperature of 40 °C (104 °F), coupled with a dewpoint temperature of 20 °C (68 °F) yields an apparent temperature of 44 °C (111 °F). If the temperature is dropped to 37 °C (99 °F) and the dewpoint temperature is raised to 22 °C (72 °F) — something that is common within the scenarios we modeled for this study — the apparent temperature drops to 42 °C (107 °F). Thus, the air temperature plays a more important role in human perceived conditions than does dewpoint temperature.
The impact of these thermal changes has a significant effect upon human mortality, as indicated by our modeling (Table 7 and 8). The June 2008 EHE provides a good example (Table 7), and mortality increases above the baseline are reduced for each scenario, particularly case 4. As the heat event continues, the mortality increase becomes greater; for the baseline, the percentage increase is 1.2% for the first day of the event to 13.5% for the fourth consecutive day, a common result for all of our modeling using the aforementioned mortality algorithm. On average, approximately 150 people die daily during summer in Los Angeles County from all causes (Los Angeles County Department of Public Health 2015). Thus, a 13.5% increase in mortality represents about 20 extra deaths from heat on June 22, 2008, an unfortunate and considerable number of deaths. If we sum the percentage increases for all 4 days of the EHE, the mean 6.9% increase represents about 41 extra deaths which we relate to heat during the entire event (6.9% of 600 total deaths equals 41 excess deaths during the 4-day heat event). For June 19–20, 2008, the baseline indicates that an MT air mass was present for those days. There was an air mass change to DT on June 21, while a transition air mass (a change from one air mass to the next; indicates a cold front passage) was present on June 22.
Table 7 Changes in air mass type, apparent temperature (AT), and mortality for June 2008 EHE. 5AM and mean daily apparent temperature are displayed for each day during the EHE. “Increase in mortality %” represents percent increase in excess mortality over the daily mortality standardized value. The mean increase for all the EHE days is shown at the second row from the bottom. The net decrease in heat-related mortality from the baseline is shown in the bottom row. “SSC Type” shows air mass type; darker cells in bold italics show actual changes in air mass type due to a significant meteorological change Full size table
Table 8 Changes in air mass type, apparent temperature (AT), and mortality for September 2010 EHE Full size table
For case 1 on June 19, there was no reduction in excess mortality although apparent temperatures were somewhat lower (1.2% above the baseline). Yet, reductions can be seen for the other 3 days of the heat wave: on the 20th, from 1.9% in the baseline to 1.7%; on the 21st, from 11% in the baseline to 8.5%; and on the 22nd, from 13.5% in the baseline to 12.1%. Thus, for the entire 4-day heat event period, case 1 produced a 1 percentage point decline in excess mortality, from 6.9 to 5.9%. This is a 15% decrease in heat-related mortality, and represents about 6 saved lives (from 41 excess deaths to 35 deaths). In contrast, case 2 only reduced excess mortality by 8% (6.9–6.4%) when compared to the baseline. Case 3 did slightly better than case 2, but case 4, the most aggressive case in terms of increasing tree cover and albedo, reduced excess mortality by 18%, or about 8 deaths (from 41 to 33). We find these results to be encouraging, as they indicate that heat-related deaths could be reduced significantly in a heat event of this type.
The September 2010 event (Table 8), a dry Santa Ana situation, had even more dramatic outcomes. Most of the days during this EHE were DT (dry tropical), the air mass type that kills the most people in Los Angeles. On September 26, the beginning of the heat event, there was an actual air mass change under cases 2, 3, and 4, from DT to a more benign dry moderate (DM) air mass. Such air mass changes are rare in similar evaluations of cities. This change has a great impact on reducing heat-related mortality, as can be seen for cases 2, 3, and 4 on September 26, where no heat-related mortality was estimated. During this EHE, the mean percentage reduction on the days when excess mortality was estimated dropped by 29% for case 4 (8.8% is a 29% reduction from 12.4%), which is the equivalent of saving 23 lives during that heat event (from 78, based on our algorithm for this heat event, to 55 deaths). This result was among the most encouraging we have seen for such heat wave analysis in any large urban area.
The July 2006 heat wave was hot and humid, showing some monsoonal influence with MT + air masses dominating. Temperatures on the 22nd and 23rd approached 38 °C (100 °F) and dewpoints were near 16 °C (61 °F). The results were somewhat similar to the two previous heat waves evaluated: the temperature decreases under the four scenarios were rather significant, with decreases ranging generally from 0.5 to 2 °C (0.9–3.6 °F) for cases 1, 2, and 3, and generally 0.5 to 1.0 °C (0.9–1.8 °F) lower than that for case 4. Dewpoint simulations showed similar increases to other heat waves in the analysis, varying from slight decreases to increases up to 2 °C (3.6 °F). Notably, the greatest temperature decreases and dewpoint increases were observed at night, when thermal changes often have the greatest impact on human health.
Three of the 5 days during this heat event demonstrated some air mass change. For July 22 and 26, all four cases resulted in changes from MT + to MT, a more benign air mass. On July 25, we saw a change from MT + to MT for case 4, our most aggressive scenario. Mortality percentages diminished, particularly for case 4. In fact, the decreases in excess mortality range from 9% for case 2 to 18% for case 4, certainly significant decreases attributed to the modeled alterations in albedo and vegetation.
The August 2009 heat event was somewhat drier; unlike the 2006 event, there were some DT and DM days and dewpoint temperatures were generally below 10 °C (50 °F). As is often the case when drier heat events are present, the results were more variable than the more humid events. Temperature decreases usually exceeded 1.0 °C (1.8 °F), and frequently approached 2.0 °C (3.6 °F), especially for case 4. In this heat wave, we saw instances that exceeded a 3 °C (5.4 °F) decrease, which is especially large. Drier air masses possess a lower specific heat than more moist air masses, which permits them to gain or lose energy at a faster rate. Thus, it is possible to see these more extreme results, with greater daily swings. This is particularly the case for dewpoint temperature, which shows up to an almost 8 °C (14.4 °F) swing between increases and decreases during the heat event. We have closely examined these large dewpoint swings, and there is nothing that we observed to consider that these are not feasible, based upon the scenarios and modeling that we used. However, we think that results from this August 2009 heat event should be observed with greater caution than the other EHEs because of the high variability in the results.
For the 2009 event, we also saw some air mass changes. Three MT days changed to DM, a generally cooler and more comfortable air mass. There were also decreases in excess mortality percentage, but the increases in mortality from the heat were much smaller for this event than for the other three. Temperatures for this event were the lowest for all four heat waves evaluated, hence the lower mortality increases. Thus, the percentage decreases, though very large, are to be regarded with some perspective, noting that overall heat deaths were lower.
Local effects on climate change
The results of this study suggest that we have existing technologies to significantly lessen the impacts of heat on negative health outcomes. However, there are also some important climate change implications that arise from our analysis.
A typical approach to evaluating the impact of climate change upon excess mortality is to apply climate models to mortality algorithms and determine how many additional deaths would occur under the various emissions scenarios (e.g., Sheridan et al. 2012). In this study, we departed from the approach and attempted something quite different. Rather, we estimated how many years of climate change–induced warming we could potentially delay if the four albedo/canopy cover scenarios offered in this evaluation were implemented. We modeled the four cases under both business-as-usual and moderate mitigation scenarios at the Los Angeles County level by utilizing the Representative Concentration Pathways (RCP) models 8.5 and 4.5 (van Vuuren et al. 2011; Schwalm et al. 2020), approved by the Intergovernmental Panel on Climate Change (IPCC). The results are intuitive and allow us to accept our hypothesis that these cooling scenarios can potentially delay climate change–induced warming in Los Angeles by a number of years or even decades (Fig. 1).
Fig. 1 Years of delay of climate change–induced warming under the four albedo/canopy cover scenarios. Bars indicate the number of years of delay that would result from each case under either RCP 8.5 or 4.5 Full size image
To accomplish this, we initially determined the mean reductions in maximum temperature for Los Angeles County, using the same tree cover/albedo prescriptions that we used in the rest of the study. The mean reduction for each case is 1.09 °C (1.8°F) for case 1, slightly less than 1 °C for cases 2 and 3, and a considerably larger 1.71 °C (3 °F) for case 4. We examined the 90th percentile of daily maximum temperature for the entire year and then just for summer (May through September). Using modeled data for the years 1950–2099, the average temperature increases under the business-as-usual and moderate mitigation scenarios were + 0.039 °C and + 0.016 °C per year, respectively. We then divided the average temperature reduction of the four tree cover and albedo cases by those average annual temperature increases to determine how many years of warming could be delayed. For example, implementing case 4 (high tree cover and high albedo) would reduce temperatures by an average 1.7 °C, so we find that 1.7 / 0.039 = 43 years of possible delay. This means that climate change–caused warming could be potentially delayed approximately 43 years relative to a business-as-usual emissions scenario (RCP8.5) if tree cover and albedo were to be increased aggressively. In this example, Angelenos could experience a climate in the year 2063 that was like the climate in year 2020. For the moderate mitigation scenario (RCP4.5), the delays would be greater, related to the lesser slope of temperature increase in that model. Assuming the case 1 example above, the 1.09 °C (nearly 2 °F) decrease would delay the warming by 69 years (1.09/0.016), since the slope of temperature increase for the RCP4.5 model is less than half of RCP8.5. Thus, if we could meet the emissions demands of RCP4.5, the effectiveness of the cooling will yield an even greater delay in climate change–caused warming. We believe this is a novel way to estimate how urban cooling scenarios based upon presently-available technologies can potentially mitigate climate change–induced warming effects in urban areas.