Show Summary Details

Page of

Subscriber: Pusan National University; date: 16 December 2018

# The Impact of Moisture and Temperature on Human Health in Heat Waves

## Summary and Keywords

Extremely high air temperatures are uncomfortable for everyone. For some segments of the population, they can be deadly. Both the physical and societal aspects of intense heat waves in a changing climate warrant close study. The large-scale meteorological patterns leading to such events lay the framework for understanding their underlying causal mechanisms, while several methods of quantifying the combination of heat and humidity can be used to determine when these patterns result in stressful conditions. We examine four historic heat waves as case studies to illustrate differences in the structure of heat waves and the variety of effects of extreme heat on humans, which are characterized in terms of demographic, geographic, and socioeconomic impacts, including mortality and economic ramifications.

Weather station data and climate model projections for the future point to an increase in the frequency and intensity of extreme heat waves as the overall climate gets warmer. Changes in the radiative energy balance of the planet are the principal culprit behind this increase. Quantifying changes in the statistics of extreme heat waves allows for examination of changes in their potential contribution to human health risk. Large-scale mortality during heat waves always occurs within a context of other factors, including public health policy, rural and urban management and planning, and cultural practices. Consequently, the impacts of heat waves can be reduced, and may in many places be manageable into the future, through implementation of such measures as public health warning systems, effective land management, penetration of air conditioning, and increased monitoring of vulnerable or exposed individuals. Given the potential for severe impacts of the more intense heat waves that are virtually certain to occur in the warmer future, it is critical that both the physical and social sciences be considered together to enable society to adapt to these conditions.

# Key Messages

• Very extreme heat waves cause many deaths across the world.

• Human-caused global warming is increasing extreme temperatures more than average temperatures. Today’s rare heat waves will be commonplace by the end of the 21st century.

• Hot and humid events can be more dangerous than hot and dry events, but the acclimation of the populace to the local climatology is the determining factor in the deadliness of a given combination of temperature and humidity.

• The very old, the very young, the sick, the poor, and those who work outside are the most vulnerable segments of the population to extreme heat waves.

• Early warning systems in both developed and developing nations are key to reducing heat wave fatalities.

• Heat wave metrics combining temperature, moisture, wind, and other factors are conceptually valid but current advisory levels are not universally applicable.

# Introduction

The effect of extremely high air temperatures on human and natural systems can be profound (Field et al., 2012). It should be plainly obvious that as global warming continues to increase due to emissions of greenhouse gases, the intensity and frequency of heat waves will also continue to increase. However, the character of the change in extremely high temperatures may differ from changes in average temperatures as the physical mechanisms causing these changes may also differ. Likewise, the response of human systems to a given severity of heat waves will also have to change to adjust to this new reality and develop new societal standards of acceptable risk. As such events become more common, we can and likely will adapt our systems to better cope with the new normal. However, there are technological, sociological, and physiological limits to adaptation, and if emissions continue unabated and the climate system responds accordingly, substantial changes to society will be necessary as a response to ever-more-frequent intense heat waves. As technological advances become available, our ability as individuals to adapt to extremes improves. The same can be said about institutional limitations as new social arrangements may emerge that allow society to cope with extremes in a more efficient way. However, improvements in both societal arrangements and technology do not necessarily benefit all members of society, as some socioeconomic sectors have more resources than others. Hence, equity becomes a central part of the patterns of adaptation to weather extremes. Impacts of heat waves include higher-than-normal hospital emergency visits (Knowlton et al., 2009; Green et al., 2010), higher-than-normal numbers of deaths (Trent, 2007), and impacts on labor productivity in urban settings (Dunne et al., 2013). The aforementioned studies are specific to high-income countries, but similar impact trends (but with different magnitudes) can be expected in lower-income countries. The analysis of the impacts of heat waves on agriculture including agricultural labor productivity is significantly absent in the research literature. Even in high-income countries, the impact of severe heat waves is disproportionally high on disadvantaged segments of the population.

A heat wave occurs when some deviation from the normal atmospheric circulation causes temperature to temporarily rise. There are a multitude of different large-scale meteorological patterns that can lead to heat waves (Grotjahn et al., 2015). For instance, in continental interiors, “blocking” patterns of localized high pressure systems can cause extended periods of stagnation. Other heat waves may be driven by strong winds. If winds are such that air is descending from high mountains to lower regions, compression by this descent from regions of low pressure to higher pressure can heat the air. Very long-duration heat waves are often associated with drought. If seasonal rainfall is low in the warm seasons, soil moisture levels can become drier than normal. The resulting decrease in evaporative cooling leads to higher air temperature that further dries the soil. While the conditions leading to each individual heat wave are unique, certain patterns are very predictable, leading to actionable forecasts of high temperatures.

The effects of the most lethal heat waves are due not only to high temperatures, but also the effects of humidity. The ability of humans to cool themselves by sweating is diminished as relative humidity increases. Hence, hot and humid conditions can be more dangerous than equivalently hot but dry conditions (Steadman, 1979a,b). That said, hot and humid are relative concepts. For example, a relatively normal summer in Jakarta or Rome could be uncomfortable in Minneapolis or Stockholm. Or temperatures may be considerably higher but more bearable due to drier conditions in cities such as Phoenix and Athens. Societies that have populated a particular geographical area for long periods of time have learned to cope with their “normal high temperatures,” as well as their relatively high or low humidities (Guirguis et al., 2014). As the global climate changes, air temperatures over land will increase in all locations in all seasons. The actual mass of atmospheric water vapor, measured as the specific humidity, will also increase because of increased evaporation over the oceans and the ability of warmer air to contain more water vapor. However, the relative humidity, the ratio of the actual water vapor relative to the fully saturated value at a given temperature, is more closely related to the human body’s ability to cool through evaporation. As described below, relative humidity will decrease in some regions as the climate warms, complicating projections of the risk to human health from future heat waves.

# Heat Wave Definitions

## Temperature

A universal quantitative definition of what constitutes an extreme heat wave is unlikely to ever be agreed upon (Perkins, 2015). Considering events of a specified temporal rarity, say the annual maximum temperature, is an oft-used approach to describe events that are very infrequent and is known to statisticians as the “block maxima” method (Coles, 2001). Such an approach (e.g., Kharin et al., 2013) is particularly useful in understanding the behavior of extreme temperatures and their future changes in a global sense as the “block,” in this case a year, can be applied uniformly across the planet. An associated formal extreme value statistical theory can be applied to the sample of annual daily maximum temperatures to fit a three-parameter distribution function. The tails of this extreme value distribution function describe the truly rare events. For instance, the 20-year return value, defined as that daily temperature exceeded once every 20 years on average, can be directly calculated once the distribution of annual daily maximum temperatures has been fitted. In this example, the return period is held fixed at 20 years. Extreme-value statistics can also be used to estimate return periods, that length of time expected on average between events of a specific temperature. Figure 1, a reproduction of figure 12.14 from chapter 12 (Collins et al., 2013) of the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), shows the changes in the 20-year return values of daily temperatures between the recent past and the end of 21st-century changes under a variety of possible future scenarios of greenhouse gas emission scenarios, as projected from contemporary climate models.

Click to view larger

Figure 1. The best estimate of CMIP5 multi-model change in the 20-year return values of annual warm temperature extremes (left panels) and cold temperature extremes (right panels) as simulated by climate models assessed in the Intergovernmental Panel on Climate Change’s (IPCC) Fifth Assessment Report (AR5). Rows show the changes in 2081–2100 relative to 1986–2005 under a scenario of low future emissions (RCP2.6, top), moderate future emissions (RCP4.5, middle panels), and high future emissions (RCP8.5, bottom) experiments. (IPCC WG1 AR5, Figure 12.14, Collins et al., 2013).

Several important inferences about projected future very hot days can be drawn from the left panel of Figure 1. Clearly, increases in extreme temperatures, similar to increases in average temperatures, will not be uniform across the planet. Also note that the panels are arranged from top to bottom in order of low to high emissions scenarios. If our emissions were to follow RCP2.6, the geoengineering scenario considered in the IPCC AR5 report that is closest to what would be required by the decision of the Conference of the Parties held in Paris in December 2015 as part of the United Nations climate change negotiation process, such rare daily high temperatures would experience modest increases of 1oC to 2oC over current levels. However, such increases could still be of critical importance to the most vulnerable individuals, as we discuss later. If society does not adopt any emissions reductions policies or geoengineering strategies, as in the RCP8.5 scenario, then such rare daily high temperatures are projected to increase by up to 7oC in some locations. In many locations, extreme temperatures will be high enough to have significant potential to cause harm to large segments of the population. That said, adaptation actions have the potential to reduce the negative impacts of weather extremes, and extreme heat in particular. Under this high emissions scenario, today’s 20-year daily high temperatures are projected to occur at least every other year in most land areas. Hence, today’s rare high temperatures will become commonplace. However, an increased frequency of heat waves is not just some danger in the distant future, as significant changes have already been documented in daily temperature extremes over the past half-century and attributed to human changes to the composition of the atmosphere (Min et al., 2013).

A key insight from extreme value theory is an estimate of the relative rarity of a particular event. As mentioned earlier, the block maxima statistical approach, used to arrive at Figure 1, is useful for gaining a sense of the global changes in extreme temperatures. However, as the range of extreme temperature varies greatly across the planet, its results may not be easily translated to human health impacts. Alternatively, the rate of the exceedance over high thresholds, known as the “peaks over threshold” or POT method, provides a mathematically equivalent but more directly relevant approach. Threshold values may be defined as a percentage of all values or defined by the impacts of exceedance. Choice of threshold value is critical to the validity of any application of extreme value statistical theory methods. The key point is that the statistical theory is based on an “extreme value sample” drawn from the largest values in the entire parent sample. In other words, extreme value theory applies only to description of the tails of the entire distribution of all temperatures. There are no set rules to guide their choice, but the thresholds must be large enough that sub-sampling of the parent data distribution is limited to its tail. Choosing a percentage-based threshold (often the 85th–95th percentile of all data) is a straightforward way of ensuring the extremeness of the selected data points. Heat waves with significant human health impacts will most likely be in the high end tail of fitted extreme value distributions and well above the thresholds used to implement the methodology.

## Humidity

As noted, human health and welfare are most impacted by the combination of high heat and relative humidity. Most of the highly cited assessment literature covering extreme temperatures in the physical climate sciences (i.e., Seneviratne et al., 2012; Hartmann et al., 2013; Collins et al., 2013; Walsh et al., 2014) is limited to descriptions of observed and projected temperature extremes without considering humidity. Conversely, the highly cited literature on the impacts of climate change (Smith et al., 2014) considers both changes in temperature and humidity on future occupational heat stress. Humans react, in part, to heat stress by sweating. The evaporation of sweat on bare skin causes the body to cool by the energy required in the phase transition from liquid to vapor. This energy is referred to as the “latent heat of evaporation.” The rate of evaporative cooling is controlled by the difference between the actual specific humidity, the amount of water vapor in the air, and the saturation specific humidity (the maximum amount of water vapor that air can contain without precipitating out). Equation 1 shows the bulk aerodynamic approximation for the rate of surface latent heat release by evaporation.

$Display mathematics$
(1)

where Q is the actual specific humidity near the evaporating surface, Qs is the saturation specific humidity, U is the near surface wind speed, ρ‎ is the density of air near the surface, L is a constant known as the specific latent heat of water and CE is another constant, sometimes called the Dalton number (Singh et al., 2005).

The saturation specific humidity is a function of air temperature and pressure known as the Clausius-Clapeyron relationship. At sea level over a broad range of air temperatures above freezing, the saturation specific humidity increases by approximately 6% per degree of warming. Hence, warm summer air can contain significantly more water vapor than cold winter air. The relative humidity, R=Q/Qs, is the ratio of the actual specific humidity to the saturation specific humidity and is an important and intuitive way to characterize the comfort of a particularly hot and humid day. From Equation 1, we can see if air is completely saturated, R = 100% and Q−Qs = 0, and there would be no evaporative cooling at all, rendering sweating ineffective as a cooling mechanism. We also note that the surface latent heat flux is an increasing function of wind speed. Hence, for a given temperature and relative humidity, sweating is less efficient at cooling when the air is stagnant than when it is windy. This wind effect is why a breeze can feel cool and bring relief when it is very hot and also explains the remarkable effectiveness of a simple fan in mitigating heat stress.

The effect of anthropogenic climate change on humidity is more complex than it is on temperature. Over the oceans, specific humidity is projected to increase at a rate controlled by the Clausius-Clapeyron relationship. In fact, very precise satellite and in situ measurements have shown that such detectible changes in specific humidity over the oceans can already be attributed to the human-induced changes in the composition of the atmosphere (Santer et al., 2007; Willett et al., 2007). As a result of this controlling mechanism, only small changes in average relative humidity over the oceans are observed or projected as temperature increases. Figure 2, a reproduction of Figure 12.21 from Working Group I of the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), shows projected middle and end of 21st-century changes in seasonal average relative humidity under the “no emissions reduction policy” scenario, RCP8.5. Over the oceans, climate models project only small increases in relative humidity. Over land regions, changes in relative humidity are more complex. In general, surface air temperature over land has warmed more due to anthropogenic climate change than it has over the ocean and is projected to continue to do so as greenhouse gas concentrations increase. Recall from the Clausius-Clapeyron relationship that the saturation specific humidity increases by about 6% per degree of warming. Hence, saturation specific humidity increases more over the land than over the oceans due to the differential warming. The actual amount of atmospheric water vapor over land is ultimately the result of transport by winds of water vapor over the oceans.1 As the increase in the water vapor over oceans is controlled by the ocean air temperature, which increases less than the land air temperature, the increase in water vapor over land is less than what the Clausius-Clapeyron relationship would have if applied locally. Thus, the average relative humidity is projected to substantially decrease over land because, although the actual specific humidity may increase, the saturation specific humidity increases at a greater rate. As we will see in one of the case studies, it is important to note that other factors also influence water vapor changes, including changes in atmospheric circulation patterns. In particular, changes in the large-scale meteorological patterns that lead to heat waves may or may not be the same as changes in seasonal averages.

Substantial relative humidity decreases are projected in the mid-latitude portions of the Northern Hemisphere summer (JJA) by the end of the 21st century (lower middle map of Figure 2). Much of North America and all of Europe are projected to experience large decreases in summer average relative humidity by this time. The drier mid-latitude regions of central and western Asia do not show as large a decrease. Part of the reason for this is that relative humidity is already low in these regions, and percentage decreases are limited.2

As the 2006 California heat wave case study discussed in “The 2006 California Heat Wave” illustrates, these projected large-scale temporally averaged changes in relative humidity may not be relevant to the severity of certain future heat waves. High humidity heat waves are often linked to anomalous local processes, such as on-shore winds, increased plant evapotranspiration (popularly referred to as “corn sweat”) (Allen et al., 1998), and other episodic sources of increased humidity during high temperature events. Climate change may impact these short-term processes in a different way than shown in Figure 2.

Click to view larger

Figure 2. Projected future changes in near-surface relative humidity from the CMIP5 models under a high emissions scenario (RCP8.5) for the December, January and February (DJF, left), June, July and August (JJA, middle) and annual mean (ANN, right) averages relative to 1986–2005 for the periods 2046–2065 (top row), 2081–2100 (bottom row). The changes are differences in relative humidity percentage (as opposed to a fractional or relative change). Hatching indicates regions where the multi-model mean change is less than one standard deviation of internal variability. Stippling indicates regions where the multi-model mean change is greater than two standard deviations of internal variability and where at least 90% of models agree on the sign of change. (IPCC WG1 AR5, Figure 12.21).

# Review of Heat Wave Metrics to Assess Human Health Impacts

As outlined above, the physiological response of humans to uncomfortably high temperatures is to cool the body by sweating, and the rate of evaporative cooling produced by sweating is a function of temperature, relative humidity, and wind speed. There are numerous heat wave metrics incorporating such information to varying degrees. In general, these metrics are presented as a single number with some empirical multivariate dependencies. The metrics can form early warning systems for the public based on their exceedance of a priori thresholds.

An easily measured index is the wet bulb temperature, Tw. Simple devices to measure this index consist of a glass thermometer with a sock immersed in water that keeps the bulb of the thermometer wet. As water evaporates from the sock, it cools the bulb, resulting in a temperature measurement less than the actual air temperature (or dry bulb temperature). The relationship of wet bulb temperature to the actual temperature and relative humidity under stagnant conditions in the shade is often read directly from a psychometric chart. Equivalent empirical formulas fitted to these charts are valid only over a limited range of temperature and relative humidity. One such formula is (Stull, 2011):

$Display mathematics$
(2)

where T is the actual temperature in oC and R is the relative humidity in percent is a good fit for –20oC < T <50oC and 5% <R <99%. The current highest instantaneous value of Tw anywhere on the planet is 31oC and is likely limited by convective instability processes. As the typical human core body temperature is around 37oC, sustained values of Tw> 35oC would be fatal due to an inability to dissipate even resting metabolic energy (Sherwood & Huber, 2010).

A related but more difficult quantity to measure is the dewpoint temperature, D. This quantity is the air temperature at which the actual specific humidity would be saturated and is generally lower than the actual air temperature. Simple dewpoint meters have a polished metal plate over which increasingly cooled air is passed until condensation is measured. While dewpoint is a useful metric in its own right, it is particularly convenient as a way to infer the actual relative humidity. The August-Roche-Magnus empirical formula for D is valid over a wide range of temperatures and relative humidity and is independent of pressure:

$Display mathematics$
(3)

which is easily inverted for relative humidity,

$Display mathematics$
(4)

where again T is in oC and R in percent.

A related but more sophisticated metric widely used in industrial settings is the wet bulb globe temperature. This metric incorporates another important factor impacting human discomfort, the amount of direct and indirect sunlight. All other things on a hot and humid day being equal, it is more uncomfortable out in the sun than it is in the shade, as sunshine directly heats the body. The wet bulb globe temperature is measured using three thermometers. In addition to wet and dry bulb thermometers, a globe thermometer, consisting of a thermometer inside of a hollow black ball approximately 15cm in diameter measures the effect of sunlight. This temperature measurement is generally higher than the actual air temperature when the sun shines. One standard definition of the wet bulb globe temperature, Twbg, with published industrial safety guidelines is

$Display mathematics$
(5)

The globe temperature, Tg is not often measured at typical weather stations. However, assuming black body radiative equilibrium, a complicated formula can be derived:

$Display mathematics$
(6)

where P is the air pressure, z is the zenith angle, fdb is the direct beam solar radiation, and fdif is the diffuse solar radiation (Dimiceli et al., 2013). The quantities are available at a subset of the global weather stations. In shady conditions or indoors, Tg is often approximated by the ambient temperature, T. The International Organization for Standardization has established workplace safety standards based on Twbg in their publication ISO 7243, Hot Environments—Estimation of the heat stress on working man, based on the WBGT-index (Parsons, 2006). Somewhat more stringent standards for athletic events and practices have been published by the American College of Sports Medicine to reflect the higher degree of exertion when playing sports (Armstrong et al., 2006).

The principal issue with using the wet bulb globe temperature as a widespread metric for deadly heat waves is that wind and the intensity of sunlight are usually not well measured throughout the day at most weather stations. Another popular metric, often used by television meteorologists, is the heat index also known as the apparent temperature based upon assumptions about human physiology and behavior, clothing, shade availability, and winds (Steadman, 1979a,b). Because of its simpler formulation, we will use the heat index to compare and contrast our four heat wave case studies in the next section. The following bi-quadratic empirical function of temperature and relative humidity is valid to within ±1.3oF (±0.7°C) of the NOAA tabular values for temperatures greater than 80oF (27°C) and relative humidity greater than 40%:

$Display mathematics$
(7)

$Display mathematics$

$Display mathematics$
(8)

Because the standards are defined by a U.S. agency, ambient temperatures in Equations (7 and 8) and HI thresholds are specified in Fahrenheit. If the air temperature is less than 70oF (21°C), we set the heat index to the temperature. Note that many other variants of heat index formulae exist for different ranges of temperature and relative humidity. Anderson et al. (2013) reviewed 21 different formulations, finding that most of them were consistent with theoretical concepts of apparent temperature, but cautioned that care must be used in selection of any particular formulation. The National Oceanic and Atmospheric Administration (NOAA) has published the advisory guidance summarized in Table 1 about prolonged exposures to heat as measured by the heat index. NOAA also warns that the effective heat index can be up to 15oF (8°C) higher in direct sunlight.

Table 1. The official NOAA heat index advisory levels.

Heat Index

Comment

Caution

80–90°F (27–32°C)

Fatigue is possible with prolonged exposure and activity. Continuing activity could result in heat cramps.

Extreme caution

90–103°F (32–39°C)

Heat cramps and heat exhaustion are possible. Continuing activity could result in heatstroke.

Danger

103–125°F (39–52°C)

Heat cramps and heat exhaustion are likely; heatstroke probable with continued activity.

Extreme Danger

over 125°F (52°C)

Heatstroke is imminent.

Our usage of the NOAA heat index and advisory statements in the case studies of “Dangerous Heat Wave Case Studies” is to illustrate the complex relationship between temperature and relative humidity and the differences between deadly heat waves. As the case studies reveal, certain combinations of high temperature and high relative humidity are not observed. The differences between the case studies clearly show that the NOAA heat index advisory statements should not be considered universal. An individual’s heat tolerance is determined by many factors including acclimation to the local climatology. Indeed, there is no clear metric from epidemiological studies linking heat wave temperature and humidity characteristics to human mortality valid across the planet. Barnett and Aström (2012) could not find statistically significant differences in correlations between U.S. mortality and measures of urban heat defined either with or without inclusion of humidity effects. On the other hand, Goldie et al. (2015) found that high nighttime humidity preceded by high daytime temperatures strongly influenced hospital admissions in the tropical city of Darwin, Australia. Vaneckova et al. (2011) compared the performance of a variety of heat wave metrics in predicting human mortality in the subtropical city of Brisbane, Australia, finding that daily average temperature outperformed daily maximum or minimum temperature on a variety of indices including humidity, wind, and solar insulation. Broadly speaking, cautionary or dangerous values of the heat index should be determined by specific local and regional climatological conditions, as well as the physiological characteristics of the local populace. As mentioned, what constitutes an extreme temperature in Paris may be normal in Karachi or Mexico City. Likewise, similar temperatures in Karachi may be more dangerous than in Hyderabad, India, due to differences in climatological values of humidity. Statistical studies relating human health to measures of heat wave intensity that are confined to single regions may be predisposed to favor simpler temperature-based metrics over complicated biometeorological indices as predictive models for a number of reasons. For instance, the available number and variety of very extreme heat waves within single regions may be too small to find multivariate statistical significance. However, because of the known physiological responses to differences in both temperature and humidity (de Dear et al., 1989; Tsutsumi et al., 2007; Kaynakli et al., 2014), multivariate heat wave indices can be of use in comparing human health effects across regions with different climatological conditions (Barreca, 2012).

It could be argued that a universal advisory of “Extreme Danger” for values of heat index or temperature over some very high threshold (for instance, 125°F or 52°C) is appropriate due to the negative health impacts to any human being from such extreme heat. But the less severe advisories in Table 1 are far less widely applicable due to the large variations in average summer temperatures across the world. Even if a set of advisory levels could be agreed upon for a particular region and heat wave metric, the relevance to individual inhabitants in that region would also vary considerably, as some groups would be more exposed or vulnerable than others. Heterogeneity in the population across age, gender, income, wellness, and other variables must be considered in assessing the impact of heat on human health.

The wide variety of approaches to public advisories, ranging from temperature values alone to complicated functions of temperature, humidity, wind, air quality, and other variables leads to a certain amount of confusion. In fact, it has been argued that temperature alone is the most effective advisory quantity as the public easily understands it (Barnett & Aström, 2012). We describe the case studies considered below in terms of the NOAA heat index, a U.S.-based metric, despite only one of the events having occurred there. This is not a statement of preference for this heat wave definition, but rather a means to compare and contrast the very different meteorological characteristics of the four heat waves. Data limitations preclude the calculation of more complex heat wave metrics, such as wet bulb globe temperature, and consideration of temperature alone does not fully differentiate these events.

# Dangerous Heat Wave Case Studies

To more clearly demonstrate the relationship between the meteorological aspects of very severe heat waves to their impacts on human health and wellbeing, we discuss four very different high temperature events. These high-impact heat waves were chosen to sample both different climatological regimes, as well as different population demographics. They were also selected for their particularly devastating consequences, and each is a rare event with a very long return period. We examine these events by using the heat index calculated from both observations and output from simulations of climate models. The HadISD observational product provides quality-controlled sub-daily3 dry bulb temperatures and dewpoint temperatures at 4,070 weather station locations worldwide suitable for long-term climate trend analyses. Data for our four case studies are taken from airport stations in the affected regions. While these stations have the advantage of the longest available data records, they tend to be in outlying areas from the highest population densities. In the urban case studies, the “heat island effect” can cause temperatures to be substantially higher in city centers than those measured at the airport. Furthermore, nighttime temperatures in city centers may not cool as much as outside near an airport due to the heat capacity of urban infrastructure and a more stagnant environment. This is especially true indoors, where it may remain considerably warmer at night than outdoors. Such effects can significantly add to the heat stress of the population if they are suitably exposed.

## The 2003 European Heat Wave

The European continent has experienced a remarkable run of hot summers since 1999, with the record for the hottest summer over almost all locations of the continent being either 2003 or 2010 (Barriopedro et al., 2011). Most days in May through August 2003 were at least 2°C warmer than average over a large area in the middle of the continent, with a moderate but long heat wave in early June 2003 and a remarkably intense 10-day-long heat wave in early August. The large-scale meteorology responsible for the heat wave was a persistent anti-cyclonic flow pattern over all of Europe for much of the summer, leading to a blocking pattern over central Europe (Black et al., 2004). The persistent stagnation dried out the soil, and temperatures soared. The heat wave was further enhanced by the human change to the composition of the atmosphere (Stott et al., 2004). Here we examine the temperature and humidity during this second heat wave in the Paris metropolitan area, which suffered an excessively high mortality rate.

Figure 3 shows a scatterplot of temperature and relative humidity at Orly Airport in Paris at the time of the maximum heat index on hot days from 1973 to 2014. The 2003 heat wave is marked by asterisks and the 1973–2014 summer average is marked by the large black dot. This method of visualizing very hot days is intended to show the climatology of extreme temperature and relative humidity and the relationship between them. It also provides a context for the extreme event in question. Scatterplots for each of the four case studies reveal different structures in the associated two-dimensional probability distribution. In all cases, there is a boundary, approximately diagonal in this particular plot, above which combinations of high temperature and high relative humidity do not occur. The details of this boundary are controlled by the local climatology. Points near that boundary are rare combinations and considered as extreme events in the two-dimensional space. That is not to say that all such events have the same impacts on human health.

Figure 3 reveals that higher temperatures generally exhibit low values of relative humidity at this weather station. The heat index reached the NOAA danger advisory stage only five times during this period, with four of those values during the heat wave in 2003 when those days were very dry at that time of day. Figure 3 also reveals a relationship between the highest values of relative humidity as a function of temperature. For instance, at 95oF (35°C), there are no instances where relative humidity exceeds 50%. However, this value of relative humidity occurred many times at 80oF (27°C). The 1973–2014 average daily maximum summer (JJA) temperature is about 75oF (24°C), and the relative humidity averages about 50% at that time of day and is indicated by the large black dot.

The top panel of Figure 4 shows the hourly values of the heat index at Orly Airport in Paris from July 4 to August 22, 2003, the hottest seven-week period that year. There were two separate periods at a caution health advisory level, with the second being longer and more severe as the daily maximum heat index was at the extreme caution level or higher for 10 days in a row. Nighttime brought relief from the heat wave, at least outside and away from the Paris city center, as the heat index and the surface air temperature (middle panel) cooled by at least 30oF (17°C) every night during the period of advisories. Daily minimum heat index values never reach into the caution advisory stage at this station over the entire HadISD dataset, including during the second 2003 heat wave. This indicates that there is substantial outside cooling at night, even on very hot days. Relative humidity increases at night during a stagnant heat wave. As the temperature drops, so does the saturation specific humidity, Qs, because of the Clausius-Clapeyron relationship. However, unless it precipitates or condenses (dew), the actual specific humidity remains fairly constant through the day. Hence, the relative humidity must be larger during the cooler parts of the day since Qs is in the denominator of its definition. The diurnal cycle of the heat index responds more to the temperature decrease than to the relative humidity increase resulting in a nocturnal decrease. However, the relative humidity increase does explain why nights may feel “sticky” during heat waves, indicating that simple metrics do not explain all components of human discomfort.

For these reasons, the diurnal cycle of relative humidity is highly inversely correlated with temperature during non-precipitating periods of a heat wave. Over this time period, the correlation at this weather station between hourly air temperature and relative humidity was −0.85. Precipitation did provide some relief on July 17 and again on July 23 at this location, with daily maximum heat indices and temperatures below the summer average (indicated by the green lines).

The average daily maximum temperature during the 2003 heat wave, 86oF (30°C), was substantially higher than the 1973–2014 summer (JJA) average of 75.1oF (23.9°C) at this location. However, the average summer relative humidity at the time of the daily maximum heat index during the heat wave was also substantially lower (37.2%) than its summer average (49.5%). Because the relative humidity is so low at the hottest part of these days, the average value of the daily maximum heat index (85.4oF or 29.7°C) is slightly lower than the coincident actual air temperature, and this was generally true during the advisory days of the heat wave. As indicated by the asterisk markers in Figure 3, the second 2003 heat wave was characterized by the highest temperatures in the HadISD record, which experienced exceptionally low relative humidity. In fact, none of the extreme caution days in 2003 exceeded relative humidity values of 40%, although such humidities occurred many times in the record. Table 1 summarizes the four case study heat waves relative to the summer averages. The 2003 heat wave is characterized not only by its high temperature and low relative humidity, but also by the anomalously high number of extreme caution and danger advisory days relative to the summer average. However, the total number of days with heat index values greater than 80oF (27°C) was not particularly out of the ordinary.

The August heat wave was exceptionally hot over most of France, and its extreme intensity resulted in an estimated 15,000 deaths in France in excess of the seasonal norm (Fouillet et al., 2006) and over 70,000 deaths in all of Europe (Robine et al., 2008). As might be expected, the highest real temperatures were in the south while the hottest anomalous temperatures were distributed fairly evenly (H’emon & Jougla, 2003; Fouillet et al., 2006). Despite this, the Ile-de-France region (Paris and suburbs) experienced a seasonal excess mortality of 2.3 times the seasonal norm, well above any other region in France; other than the center region, at 2.0, no other region was higher than 1.7 (H’emon & Jougla, 2003). There appear to be a number of reasons for this. For instance, a particularly vulnerable segment of the population was the people living on the top floors of apartment buildings. The importance of keeping indoor areas cool is illustrated by the fact that, despite a stable heat index throughout the duration of the heat wave, the first notable increase in mortality did not occur until three days after the start, with the mortality rate then peaking on the final day of the heat wave and only returning to normal four days later (Fouillet et al., 2006). In northern France (e.g., Paris), where intense heat waves are uncommon, buildings are not designed to remain cool during intense heat, with the penetration of air conditioning, for instance, being quite low. Lack of experience among Parisians in dealing with extreme heat may also have been important: Easily preventable dehydration was a leading cause of mortality. Another possible factor is that ozone levels were particularly high. The urban heat island effect, on the other hand, does not appear to be have been important, with Paris temperatures being unremarkable with respect to other areas of France, and excess mortality identical across rural and urban areas with the exception of Ile-de-France (H’emon & Jougla, 2003).

The lack of experience with intense heat seems to have also been a problem for the national public health services, which were criticized for only acknowledging the existence of both the heat wave and the excess mortality toward the end of the event (Lalande et al., 2003). The importance of various measures implemented after 2003, including public alerts and monitoring of the elderly, was illustrated in another intense heat wave in 2006 that was associated with a much lower number of excess deaths (about one-third) than might have been expected based on the 2003 and earlier events (Fouillet et al., 2008).

Click to view larger

Figure 3. Scatterplot of temperature and relative humidity at the Paris Orly Airport weather station (HadISD 071490) at the time of the maximum heat index on hot days based on the HadISD data. The JJA climatological average from 1973 to 2014 is shown by the large black dot. The 2003 European heat wave is shown by the asterisks.

Click to view larger

Figure 4. Hourly characteristics of the summer 2003 heat wave at the Paris Orly Airport weather station during the period July 4 to August 22 based on the HadISD data. Top: Heat Index. Middle: Surface air temperature. Bottom: Relative humidity. On the Heat Index plot, the yellow line indicates a caution health advisory. The orange line indicates an extreme caution health advisory. The green line is the 1973–2014 JJA average over values at the time of the daily maximum heat index.

## The 2010 Russian Heat Wave

The 2010 summer was by some measures more exceptional over Europe than the summer of 2003 (Barriopedro et al., 2011). The central feature was a heat wave that lasted up to three weeks depending on how it was measured. Unlike the 2003 European event, it lacked a clear onset and cessation, with more day-to-day variation. This event was also characterized by a persistent blocking pattern (Dole et al., 2011). Despite the absence of a long-term warming trend in the region, a human contribution to its severity has been identified (Dole et al., 2011; Otto et al., 2012). Although often referred to as the “Russian heat wave,” it actually affected all of Eastern Europe, with Asian Russia untouched. Moscow was at its center and suffered some of the more exceptional impacts, and so is the focus of this case study.

Figure 5 shows a scatterplot of temperature and relative humidity at the Moscow weather station at the time of the maximum heat index on hot days from 1973 to 2014 and the values during the 2010 heat wave. The average summer heat index values in Moscow were similar to those in Paris, but the average number of heat index advisory days was substantially lower (Table 1). Furthermore, during those advisories, temperatures were lower and relative humidity values higher than the much drier heat waves characteristic of Paris in 2003.

The top panel of Figure 6 shows the three-hourly values of the heat index at the Moscow weather station from June 21 to August 20, 2010, the hottest nine-week period. Although the highest temperatures were not as high as in Paris during 2003, the average heat index during the heat wave was higher due to the higher relative humidity. In this heat wave event, heat index values often substantially exceeded the actual air temperatures. But more notable was the exceptionally long, uninterrupted number of heat index advisory days. Furthermore, the diurnal cycle during the 2010 heat wave in Russia was often less than 20oF (11°C), resulting in higher nocturnal heat index values with a few evenings remaining in a caution advisory state when nighttime maximum relative humidity was over 90%. There was no precipitation at the Moscow weather station to lower the air temperature during this period.

The number of non-accidental excess deaths in the Moscow metropolis exceeded 10,000 during the course of the heat wave, compared to a normal rate of 300 deaths per day in a city of 11.5 million (Shaposhnikov et al., 2014). Munich Re, the reinsurance company, estimated 56,000 excess deaths from this heat wave including those outside Moscow (Munich RE, 2015). While direct exposure to heat appears to have been a major factor, pollution also appears to have been responsible for around one-third of the excess deaths, with the interaction of pollution and heat being particularly stressful for many individuals. A particular feature of the heat wave was the initiation of numerous wildfires surrounding the city. Even though these fires were small relative to the enormous fires that routinely burn in central and eastern Russia, their proximity to Moscow led to an exceptional haze within Moscow (Global Fire Monitoring Center, 2010). The heat wave was a necessary condition for triggering the fires, but it was not sufficient. Changes in government policy introduced in 2007 decreased the resources available for preventing and treating wildfires. In addition, many of the fires occurred in former peat bogs that were drained during the 20th century; these fires proved to be longer lasting and more polluting than grass or forest fires. Thus, in contrast to the Parisian heat wave, in the Moscow event, a large number of deaths resulted not only from direct exposure to heat, but also from the indirect effects of heat that were only realized through this combination of historical and contemporary policy decisions.

Click to view larger

Figure 5. Scatterplot of temperature and relative humidity at a Moscow weather station (HadISD 276120) at the time of the maximum heat index on hot days based on the HadISD data. The JJA climatological average from 1973 to 2014 is shown by the large black dot. The 2010 Russian heat wave is shown by the asterisks.

Click to view larger

Figure 6. Three hourly characteristics of the summer 2010 heat wave at the Moscow weather station during the period June 21 to August 20 based on the HadISD data. Top: Heat Index. Middle: Surface air temperature. Bottom: Relative humidity. On the Heat Index plot, the yellow line indicates a caution health advisory. The orange line indicates an extreme caution health advisory. The green line is the 1973–2014 JJA average over values at the time of the daily maximum heat index.

## The 2006 California Heat Wave

Very high temperatures were experienced over most of North America in July and August of 2006. The most severely impacted state was California. The number of directly reported deaths due to the 2006 heat wave varies from 86 to 163, depending on which method is used and which types of morbidity factors are taken into consideration (Green et al., 2010; Trent, 2007). Note that this mortality estimate is obtained by different methods than the much higher “excess death” estimates reported for the 2003 European and 2010 Russian heat waves and is not directly comparable. In California and Nevada, this heat wave was especially severe at night. Gershunov et al. (2009) found that the difference between daily maximum and minimum temperatures across California exhibited a negative trend during periods of great heat waves. They attribute this trend toward more dangerous nighttime heat waves to be a result of higher humidity values brought on by increased moisture transport from the Pacific Ocean off the coast of Baja, California. This is an important point to recognize when projecting changes in the multidimensional nature of future heat waves. While the average relative humidity may indeed decrease over most land regions in a warmer world (Figure 2), there may be circulation changes that could locally increase relative humidity despite much higher temperatures. Furthermore, changes in relative humidity during periods of very high temperatures may be very different than the seasonal average change due to differences in circulation changes specific to the large-scale meteorological patterns that cause heat waves (Grotjahn et al., 2015).

The 2006 California heat wave had negative health impacts on various segments of the state population. For example, associations between the high apparent temperature and hospital admissions for all respiratory disease, pneumonia, ischemic stroke, diabetes, and other ailments were noted (Green et al., 2010). Furthermore, the magnitudes of the impacts of the heat wave on human health vary by income, age group, and other socioeconomic factors (Green et al., 2010; Knowlton et al., 2009). Analysis of the impact of the 2006 California heat wave has been concentrated mainly on health-related topics with relatively few analyses of other societal impacts in the industrial or agricultural sectors. One important fact that is clear is that fatalities in the 2006 California heat wave were concentrated among middle-aged Latino and black males, rather than the very young and very old who died in the 2003 European and 2010 Russian heat waves or other California heat waves (H’emon & Jougla, 2003; Trent, 2007; Green et al., 2010; Shaposhnikov et al., 2014). Lax enforcement of safe labor practices and regulations could have exacerbated the death toll in high-risk socioeconomic sectors such as construction and agriculture. Labor productivity in such high-risk activities declines during heat waves (Graff Zivin & Neidell, 2014). For example, agricultural laborers are exposed to the highest heat index values of the day while doing strenuous manual labor in the fields. As a result, not only does labor productivity decline but also accidental death can increase if proper precautions are not observed.

El Centro, California, in the Imperial Valley is a very hot place, in part due to its inland location and altitude below sea level (–12m). The single HadISD station (#722810) available in the Imperial Valley does not meet all the quality-control criteria to be included among the 4,070 stations judged to be suitable for long-term trend analyses. Apart from being a decade shorter than most of the higher quality datasets, measurements are missing throughout its record. Figure 7 shows a scatterplot of temperature and relative humidity at the El Centro Airport weather station at the time of the maximum heat index on hot days from 1984 to 2014 and the values during a particularly severe portion of summer 2006. Figure 8 shows that during the 2006 California heat wave (June 24–August 1), there were no nighttime measurements at the El Centro Airport. Measurements were also missing during the weekend and over the July 4th holiday, perhaps the result of a manual system of recording the station data. As a result, this station does not inform about the effects of nocturnal heat stress. However, due to the rural nature of the Imperial Valley, no appreciable urban heat island effect was expected, and the conditions at the airport were representative of those experienced by the local outdoor population during the heat of the day. The JJA average daily maximum heat index at the local airport was nearly 105oF (41°C), somewhat above the danger advisory level. Over the available data from the period of the heat wave shown in Figure 8, this value was exceeded every day for as much as 13 hours. Relative humidity values for most of the 2006 summer were low as is typical of this very dry desert region. However, on July 25, relative humidity increased for about a week with values rising to 30% or above at the hottest time of the day. This caused a spike in the heat index to the extreme danger level on that day. The actual temperature cooled substantially immediately thereafter, but the heat index remained above average due to the high humidity. It is very likely that the unmeasured values of nocturnal heat index were also anomalously high during this week. Although values of daily maximum heat index in El Centro during the 2006 heat wave were high relative to its longer summertime average, few individual days would be considered rare. The duration, rather than the severity, of this heat wave was the outstanding characteristic of this event. It is difficult to quantify this statement due to the poor quality of the data record. However, this is consistent with larger-scale studies of the 2006 California heat wave (Gershunov et al., 2009).

Click to view larger

Figure 7. Scatterplot of temperature and relative humidity at the airport weather station (HadISD 722810) in El Centro, California, at the time of the maximum heat index on hot days based on the HadISD data. The April–November climatological average from 1984 to 2014 is shown by the large black dot. The most deadly periods of the 2006 California heat wave (July 24–August 1) are shown by the asterisks.

Click to view larger

Figure 8. Hourly characteristics of the summer 2015 heat wave at the airport weather station (HadISD 722810) in El Centro, California, during the period June 24 to August 2 based on the HadISD data. Note that nighttime, weekend, and holiday measurements are missing from the dataset. Top: Heat Index. Middle: Surface air temperature. Bottom: Relative humidity. On the Heat Index plot, the yellow line indicates a caution health advisory. The orange line indicates an extreme caution health advisory. The green line is the 1984–2014 JJA average over values at time of the daily maximum heat index.

## The 2015 Karachi Heat Wave

Because of anthropogenic climate change, 2015 was an extraordinarily warm year globally. July 2015 was the hottest July on record. As a result, many regions around the globe experienced severe heat waves. We chose Karachi for this case study because it is normally very much more humid in the summer than our other case studies and because the death toll appears to be higher than in other heat waves during the same period across the globe. Although the scientific literature analyzing the Karachi 2015 heat wave is scarce due to the short period of time since it happened, we can nevertheless discuss the impacts of the heat wave using past literature and current information, scarce as it may be. Estimates vary, but the heat wave directly claimed anywhere between 800 to 1,300 deaths in a relatively short period of time (Masood et al., 2015; BBC News, July 2, 2015). The high death toll and other related social costs were the result not only of the weather conditions at the moment the heat wave took place, but also of socioeconomic, policy, and cultural factors affecting the ability of the public and other sectors to respond to the crisis generated by extreme conditions. For example, frequent power outages limited the ability to use air conditioning for those who had access to it. Power outages also impacted the city’s water supply system, limiting access of this vital resource when it was needed the most. The absence of an efficient early warning system resulted in a late response by the government and left hospitals to deal with an unusually high number of patients with symptoms related to heatstroke. Nonetheless, aided by social media and the philanthropy of the local citizenry, major public hospitals were able to remain open (Salim et al., 2015). Although no systematic analysis of the Karachi 2015 heat wave has yet been published, it is plausible that the number of fatalities were exacerbated by the lack of planning, cultural factors associated with the observance of Ramadan, and the lack of resources needed to deal with such an extremely dangerous heat wave.

Click to view larger

Figure 9. Scatterplot of temperature and relative humidity at the Jinnah Airport weather station (HadISD 417800) in Karachi at the time of the maximum heat index on hot days based on the HadISD data. The April–November climatological average from 1973 to 2014 is shown by the large black dot. The 18 hottest days of the 2015 Karachi heat wave are shown by the asterisks.

Figure 9 shows a scatterplot of temperature and relative humidity at the Jinnah Airport weather station in Karachi at the time of the maximum heat index on hot days from 1973 to 2014 and the values during the 2015 heat wave. The scatterplot reveals that the NOAA advisory levels were poorly devised for application in the Karachi climate. The caution advisory level is routinely reached from April to November. In fact, the April to November average daily maximum heat index is at the danger level (Table 2). The 2015 heat wave was not particularly extreme when considered only by daily maximum temperatures. However, this figure reveals that the combination of heat and humidity was a rare event at the time of daily maximum heat index in that the asterisks are at the edge of the populated space of these two variables. Although an extreme value statistical theory approach can be applied in a one-dimensional sense to the heat index during 2015, a full two-dimensional approach is not yet possible with existing statistical tools, as these events are not rare in both temperature and humidity but only in the combination (Coles, 2001).

Because of the high humidity in Karachi, the diurnal temperature cycle is less pronounced than in the drier case studies. Hence, daily minimum values of the heat index often exceed the advisory levels as shown in Figure 10. Three nights during 2015 were very warm, and June 22 had the highest daily minimum heat index value on the record. The severity of the 2015 heat wave was due in part to its duration, particularly the period from June 18 to June 30 (Figure 11).

Click to view larger

Figure 10. Scatterplot of temperature and relative humidity at the Jinnah Airport weather station (HadISD 417800) in Karachi at the time of the minimum heat index on hot days based on the HadISD data. The April–November climatological average from 1973 to 2014 is below the range of this plot. The 18 hottest nights of the 2015 Karachi heat wave are shown by the large red dots.

Click to view larger

Figure 11. Hourly characteristics of the summer 2015 heat wave at the Jinnah Airport weather station in Karachi during the period June 15 to July 2 based on the HadISD data. Top: Heat Index. Middle: Surface air temperature. Bottom: Relative humidity. On the Heat Index plot, the yellow line indicates a caution health advisory. The orange line indicates an extreme caution health advisory. The green line is the 1973–2014 April–November average over values at the daily maximum heat index.

Table 2. Summary statistics of the heat wave case studies from the HadISD observational dataset. Percent of advisory days during the summer and average values of the daily maximum heat index and the surface air temperature and relative humidity at that point in the day. The top rows of the table are from the period of the case study, while the bottom rows are the summer climatological averages (summer defined as June–August for Paris, Moscow, and Fresno, April–November for Karachi). *Karachi 2015 data are based only on the first 213 days of the year.

Paris

Moscow

El Centro

Karachi

During the event

# of days HI > 80oF (27°C) caution

32 (67%)

53 (89%)

33 (100%)

164*

# of days HI > 90oF (32°C) extreme caution

14 (29%)

29 (49%)

33 (100%)

97*

# of days HI > 103oF (39°C) danger

4 (8%)

1 (2%)

29 (88%)

97*

# of days HI > 125oF extreme danger

0

0

1 (3%)

0

Daily max HI

85.4oF (29.7oC)

93.3oF (39°C)

111.3oF (44°C)

118.4oF (48°C)

T at daily max HI

86.0oF (30Co)

90.5oF (32.5°C)

107.5oF (42°C)

98.4oF (37°C)

R at daily max HI

37.2%

41.7%

22.1%

55.2%

Annual average

# of days HI > 80oF (27°C) caution

26

18

204

284

# of days HI > 90oF (32°C) extreme caution

4

2

147

216

# of days HI > 103oF (39°C) danger

0.1

0

80

85

# of days HI > 125oF (52°C) extreme danger

0

0

1.5

0.5

Summer average

Daily max HI

74.7oF (23.7°C)

75.2oF (24°C)

104.9oF (40.5°C)

99.1oF (37.3°C)

T at daily max HI

75.1oF (24°C)

75.2oF (24°C)

102.1oF (38.9°C)

91.2oF (32.9°C)

R at daily max HI

49.5%

51.9%

21.7%

52.2%

# Discussion

The four case studies considered here illustrate that deadly heat waves are complicated phenomena, in both their physical characteristics and their impacts (Table 3). Many types of meteorological causes for major heat waves are possible. Heat waves in continental interiors are often associated with persistent high-pressure systems that cause the air to stagnate. The resulting long dry periods then cause the land to dry out, reducing cooling from evapotranspiration. In cases like the 2003 European and 2010 Russian heat waves, record temperatures were experienced, but the relative humidity was quite low. The rate of record daily high temperature exceedance has been increased by anthropogenic climate change (Meehl et al., 2009) and can only be expected to continue in that direction as the overall climate warms. Future increases in average surface air temperatures will generally reduce soil moisture, especially in continental interiors (Collins et al., 2013), leading to more favorable antecedent conditions for major heat waves of this type. As a result, projected increases in extreme daily high temperatures generally exceed the increases in average temperature (Kharin et al., 2013). For such dry heat waves, other air-quality issues can compound the impacts on public health. As discussed, fires outside of the city increased the death toll during the 2010 Russian heat wave in Moscow through the additional air pollution. In a warmer future climate, conditions should generally be more conducive to fire in similar regions (Settele et al., 2014). Increased wildfire occurrence has been detected in other regions (Westerling, 2011) and in some cases is consistent with the effect of anthropogenic climate change (Gillett et al., 2004). Furthermore, air pollution from automotive and industrial emissions is projected to worsen in some areas as the climate continues to warm (Kleeman et al., 2010). Such preexisting air-quality issues undoubtedly exacerbate the public impacts of high temperatures, particularly those caused by stagnant air masses.

Table 3. Summary of health impacts of the heat wave case studies.

Paris

Moscow

El Centro

Karachi

Number of fatalitiesa

15,000–70,000

10,000–56,000

627

Approx. 1,900

Heat wave regulationsb

N/A

N/A

Rural: Yes

Urban: Yes

N/A

Early Warning System in place?c

YES

N/A

YES

NO

Notes: a Fouillet et al. (2006), Robine et al. (2008), Shaposhnikov et al. (2014), Munich Re (2015), Knowlton et al. (2009), and Masood et al. (2015).

b Reid et al. (2009).

c Kovats and Ebi (2006) and Masood et al. (2015).

Differences in the physical properties of major heat waves certainly affect how deadly their effects are. However, the characteristics of the affected population ultimately determine who suffers the most. In the urban case studies considered here, the very young and especially the very old were most at risk. Many of the urban heat wave fatalities could be prevented by changes in social behavior and policy. Early public warnings of the imminent dangers are the most important among such changes. Inexpensive early public warning systems, such as implemented in Philadelphia, have been shown to be effective at saving the lives of the elderly (Ebi et al., 2004). To be effective, early warning systems must take into account how to communicate with those most at risk and provide the targeted information so they can obtain the help that they need. Delivery systems of such information may range from modern technological solutions (text messages, email, reverse 911 calls, etc.) to the very simple, such as a truck with a loudspeaker, depending on the needs and resources of the local population. Clearly, the design of effective early heat wave warning systems may be very different in low-income countries than in high-income countries, as well as very different in large cities than in rural communities.

In the two case studies closer to the ocean (California in 2006 and Pakistan in 2015), humidity played an important role. The temperature and humidity data from Karachi (Figure 9) reveal that it was the combination of the two fields that created a rare and deadly event during 2015. This was especially true in the 2015 Karachi event as daily maximum air temperatures were at least 4oC below record temperatures. These more humid heat waves will likely not have decreased future relative humidity because of their coastal proximity and may even experience an increase in relative humidity because of local changes in atmospheric circulation (Gershunov et al., 2009). As in continental interiors, extreme temperatures will continue to increase as the overall climate warms, although perhaps at a somewhat lower rate. Multidimensional extreme value statistical methods potentially offer insight into these changes, but classical techniques are not applicable to such situations unless both variables are extreme by themselves alone (Coles, 2001). Although much current research in statistics is being directed toward this problem (Weller, 2013; Davison & Huser, 2015), practical interpretation methods of multidimensional statistics must also be developed.

Most of the fatalities in the 2006 California heat wave occurred among middle-aged male agricultural workers (Trent, 2007). Regulations aimed at preventing occupational heat-related illness and deaths vary from country to country and even within countries. For example, in the United States, only two states, California and Washington, have explicit heat farm labor regulations. Both states require employers to provide shade and fresh water so that workers can seek relief from the heat. In addition, employers are required to provide information and training so that more informed decisions can be made by the workers themselves. Although the U.S. OSHA regulations are more specific, agricultural workers work outside and are exempt from many of these regulations. That said, it is suggested that employers monitor workers’ stress levels. If a worker’s core temperature exceeds 37.6oC and his or her heartbeat rate exceeds 110 beats per minute, then the length of the next work period should be reduced by one-third. The difficulty with this type of regulation is that many agricultural workers work at piece rates (number of units harvested or some similar rate), and there is no such a thing as “the next work period.” Cultural issues also play a role in the impact that heat stress has on a worker’s body; agricultural workers, many of whom are undocumented, may be reluctant to report heat stress for fear of either losing their jobs or having the immigration authorities learn about their situation should they go to a hospital or submit a report.

The case studies considered here reveal that the fixed thresholds defining the official NOAA heat index advisory levels of caution, extreme caution, danger, and extreme danger are not generally applicable. The original development of the heat index was based on physiological data from healthy U.S. college students (Steadman, 1979a) and does not reflect the broader U.S. population, much less those in different climates around the world. People are generally acclimated to the normal ranges of temperature and humidity conditions where they live. It is the large excursions from that normal range that are dangerous, and a more appropriate index and set of advisory levels should measure those excursions. To some extent, local advisories implicitly account for this as different countries may set different levels appropriate to local climatology. Clearly, future work is necessary in this regard, but we suggest that deviations from the normal summer conditions placed in the context of some measure of variability would be more widely applicable (Guirguis et al., 2014) and allow for more considered comparisons of heat waves in different locations. Metrics based on the relative rarity of heat index values obtained from extreme value statistical methods could also serve in this manner. Last, some stratification of advisory warnings based on the variations between healthy people and the more vulnerable members of society could also be more effective in alerting the public to the specific dangers faced during a heat wave.

Early warning systems for extreme events associated with climate change can reduce their negative impact. Heat wave early warning systems have been shown to reduce both morbidity and mortality (Ebi et al., 2004; Kovats & Ebi, 2006; Fouillet et al., 2008), and are now part of the policy implementation portfolio intended to reduce the health impacts of heat waves (Luber & McGeehin, 2008; Knowlton et al., 2009; Watts et al., 2015). In addition, early warning systems can be part of a plan that reduces infrastructure damage and losses in labor productivity that result from heat waves (Dunne et al., 2013). It is important to note that just as the impact of heat waves is highly contextualized, the design and implementation of early warning systems must also align with the socioeconomic reality of each country, region, and even municipality or local administrative unit.

The influence of anthropogenic climate change on heat waves is quite clear. For instance, the chances of the hot 2003 European summer were at least doubled by human interference with the climate system (Stott et al., 2004). But by 2014, the probability of reaching the 2003 temperatures in Europe had increased to at least a factor of 10 times higher than it would have been in a climate absent of human interference (Christidis et al., 2014). As the IPCC projections in Figure 1 imply, the future frequency of very severe heat waves as measured by temperature will increase substantially at all locations, even under the most aggressive greenhouse gas emission scenarios considered by the IPCC, resulting in today’s rare temperatures becoming commonplace by the end of this century (Tebaldi & Wehner, 2016). The best estimates of the increase from the beginning of this century in 20-year annual maximum temperatures averaged over all land are 3oC by the middle of this century and 6oC at its end in the absence of aggressive greenhouse gas emission reduction policies (Kharin et al., 2013). However, the effect on human health will likely be partly tempered by the projected decrease in relative humidity, at least in continental interiors. Multi-model projections of future changes in extreme heat stress, a similar quantity to the heat index presented here, are a few degrees less than for extreme temperatures at the end of this century due to projected decreases in relative humidity (Fischer & Knutti, 2013). Because the joint behavior of temperature and relative humidity is subject to thermodynamical constraints, uncertainty arising from differences in climate models in future extreme heat stress is lower than it is for temperature alone (Fischer & Knutti, 2013).

The four case studies examined in this article were chosen in part because of their large impacts on human health, but also because they were very rare events with long return periods. In part, their deadliness was due to this rarity, as social systems are ill-equipped to deal with such infrequent occurrences. As climate change causes these classes of heat waves to become more common, societies are expected to adapt, and such events of equivalent magnitude should become less deadly than they are now. However, the significantly higher temperatures of equally rare heat waves of the future are of grave concern.

People often say, “It’s not the heat, it’s the humidity.” This reflects the significant damages to human health that higher moisture can inflict during heat waves. However, humidity is only part of the story. Other important factors such as air quality and winds also determine the danger to any particular individual. The greatest risk to health comes from the most unusual heat waves, whether they are caused by very high heat or a rare combination of factors. The impacts on a community at large are determined by its demographics, its access to air conditioning, and its administrative response to heat waves. Many of the deaths associated with the heat waves of the four case studies considered here were preventable. Early warning and education about the risks of severe heat waves are straightforward and cost-effective methods to reduce mortality. As anthropogenic climate change makes severe heat waves much more common, such actions will become even more important.

# Acknowledgments

This work was supported by the Regional and Global Climate Modeling Program of the Office of Biological and Environmental Research in the Department of Energy Office of Science under contract number DE-AC02-05CH11231 and the National Science Foundation grant no. 000237060 under the Earth System Model (EaSM2) program. This article was prepared as an account of work sponsored by the U.S. government. While this article is believed to contain correct information, neither the U.S. government nor any agency thereof, nor the Regents of the University of California, nor any of their employees, makes any warranty, express or implied, or assumes any legal responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe on privately owned rights. Reference herein to any specific commercial product, process, or service by its trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the U.S. government or any agency thereof, or the Regents of the University of California. The views and opinions of authors expressed herein do not necessarily state or reflect those of the U.S. government or any agency thereof or the Regents of the University of California.

## References

Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop evapotranspiration—Guidelines for computing crop water requirements, FAO irrigation and drainage paper 56, Food and Agriculture Organization of the United Nations, Rome.Find this resource:

Anderson, G. B., Bell, M. L., & Peng, R. D (2013). Methods to calculate the heat index as an exposure metric in environmental health research. Environmental Health Perspectives, 121, 1111–1119.Find this resource:

Armstrong L. E., Epstein, Y., Greenleaf, J. E., Haymes, E. M., Hubbard, R. W., Roberts, W. O., & Thompson, P. D. (2006). American College of Sports Medicine position stand: Heat and cold illnesses during distance running. Medicine and Science in Sports and Exercise, 28, i–x.Find this resource:

Barreca, A. (2012). Climate change, humidity, and mortality in the United States. Journal of Environmental Economics and Management 63, 19–34.Find this resource:

Barnett A. G., & Aström, C. (2012). Commentary: What measure of temperature is the best predictor of mortality?. Environmental Research, 118, 149–151.Find this resource:

Barnett A. G., Tong, S., & Clements, A. C. (2010). What measure of temperature is the best predictor of mortality? Environmental Research, 110, 604–611.Find this resource:

Barriopedro, D., Fischer, E. M., Luterbacher, J., Trigo, R. M., & Garca-Herrera, R. (2011). The hot summer of 2010: Redrawing the temperature record map of Europe. Science 332, 220–224.Find this resource:

BBC News (July 2, 2015). Why did so many die in Karachi’s heatwave?.

Black, E., Blackburn, M., Harrison, G. B., Hoskins, B., & Methven, J. (2004). Factors contributing to the summer 2003 European heatwave. Weather, 59, 217–223.Find this resource:

Christidis, N., Jones, G., & Stott, P. A. (2014). Dramatically increasing chance of extremely hot summers since the 2003 European heatwave. Nature Climate Change, 5, 46–50.Find this resource:

Coles, S. (2001). An Introduction to Statistical Modeling of Extreme Values. London: Springer.Find this resource:

Collins, M., Knutti, R., Arblaster, J. Dufresne, J.­L., Fichefet, T., Friedlingstein, P., et al. (2013) Long-term climate change: Projections, commitments and irreversibility. In T. F. Stocker, D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, et al. (Eds.), Climate change 2013: The physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, U.K.: Cambridge University Press.Find this resource:

Davison, A. C., & Huser, R. (2015). Statistics of extremes. Annuual Review of Statistics and Its Application. 2, 203–235.Find this resource:

de Dear, R. J., Knudsen, H. N., & Fanger, P. O. (1989). Impact of air humidity on thermal comfort during step-changes. ASHRAE Transactions, 95(2), 336–350.Find this resource:

Dimiceli, V. E., Piltz, S. F., & Amburn, S. A. (2013). Black globe temperature estimate for the WBGT index, in IAENG transactions on engineering technologies. Lecture Notes in Electrical Engineering, 170, 323–334.Find this resource:

Dole, R., Hoerling, M. Perlwitz, J., Eischeid, J., Pegion, P., Zhang, T., et al. (2011). Was there a basis for anticipating the 2010 Russian heat wave?. Geophysical Research Letters, 38, L06702.Find this resource:

Dunne, J. P., Stouffer, R. J., & John, J. G. (2013). Reductions in labour capacity from heat stress under climate warming. Nature Climate Change, 3, 563–566.Find this resource:

Ebi, K. L., Teisberg, T. J., Kalkstein, L. S. Robinson, L., & Weiher, R. F. (2004). Heat watch/warning systems save lives: Estimated costs and benefits for Philadelphia 1995–98. Bulletin of the American Meteorological Society, 85, 1067–1073.Find this resource:

Field, C. B., Barros, V., Stocker, T. F., Qin, D., Dokken, D. J., Ebi, K. L., et al. (Eds.). (2012). Managing the risks of extreme events and disasters to advance climate change adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. Cambridge, U.K.: Cambridge University Press.Find this resource:

Fischer, E. M., & Knutti, R. (2013). Robust projections of combined humidity and temperature extremes, Nature Climate Change, 3, 126–130.Find this resource:

Fouillet, A., Rey, G., Laurent, F., Pavillon, G., Bellec, D., Guihenneuc-Jouyaux, D., et al. (2006). Excess mortality related to the August 2003 heat wave in France. Inter. Archives of Occupational and Environmental Health, 80, 16–24Find this resource:

Fouillet, A., Rey, G., Wagner, V., Laaidi, K., Empereur-Bissonnet, P., Le Tertre, A., et al. (2008). Has the impact of heat waves on mortality changed in France since the European heat wave of summer 2003? A study of the 2006 heat wave. International Journal of Epidemiology, 37, 309–317.Find this resource:

Gershunov, A., Cayan, D. R., & Iacobellis, S. F. (2009). The great 2006 heat wave over California and Nevada: Signal of an increasing trend. Journal of Climate, 22, 6181–6203.Find this resource:

Gillett, N. P., Weaver, A. J., Zwiers, F. W., & Flannigan, M. D. (2004). Detecting the effect of climate change on Canadian forest fires. Geophysical Research Letters, 31(18), L18211.Find this resource:

Global Fire Monitoring Center. (2010). Preliminary assessment of the fire situation in Western Russia in 2010. International Forest Fire News, 40, 20–42.Find this resource:

Goldie, J., Sherwood, S. C., Green, D., & Alexander, L. (2015). Temperature and humidity effects on hospital morbidity in Darwin, Australia. Annals of Global Health, 81, 333–341.Find this resource:

Graff Zivin, J., & Neidell, M. (2014). Temperature and the allocation of time: Implications for climate change. Journal of Labor Economics, 32, 1–26.Find this resource:

Green, R. S., Basu, R., Malig, B., Broadwin, R., Kim, J. J., & Ostro, B. (2010). The effect of temperature on hospital admissions in nine California counties. International Journal of Public Health, 55, 113–121.Find this resource:

Grotjahn, R., Black, R., Leung, R., Wehner, M. F., Barlow, M., Bosilovich, M., et al. (2015). North American extreme temperature events and related large scale meteorological patterns: Statistical methods, dynamics, modeling, and trends. Climate Dynamics, 46(3), 1151–1184.Find this resource:

Guirguis, K., Gershunov, A., Tardy, A., & Basu, R. (2014). The impact of heat waves on human health in California. Journal of Applied Meteorology and Climatology, 53, 3–19.Find this resource:

Hartmann, D. L., Klein Tank, A. M. G., Rusticucci, M., Alexander, M. V., Brönnimann, S., Charabi, Y., et al. (.2013). Observations: Atmosphere and surface. In T. F. Stocker, D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, et al. (Eds.), Climate change 2013: The physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, U.K.: Cambridge University Press.Find this resource:

H’emon, D., & Jougla, E. (2003). Surmortalit’e li’ee a la canicule d’aout 2003—Raport d’etape: Estimation de la surmortalit’e et principales caract’eristiques ‘epid’emiologiques, Institut national de la sant’e et de la recherche m’edicale, September.Find this resource:

Kaynakli, O., Mutlu, M., Atmaca, I., Kilic, M., Dincer, I., Midilli, A., & Kucuk, H. (2014). Investigation of humidity effects on the thermal comfort and heat balance of the body, in progress in exergy, energy, and the environment. In I. Dinçer, A. Midilli, & H. Kucuk (Eds.), Progress in exergy, energy and the environment (1st ed., pp. 421–434). Springer International Publishing.Find this resource:

Kleeman, M. J., Chen, S. H., & Harley, R. A. (2010). Report to the California air resources board, project #04-349.

Kharin, V. V., Zwiers, F. W., Zhang, X., & Wehner, M. (2013). Changes in temperature and precipitation extremes in the CMIP5 ensemble. Climatic Change, 119, 345–357.Find this resource:

Knowlton, K., Rotkin-Ellman, M., King, G., Margolis, M. G., Smith, D., Solomon, G., et al. (2009). The 2006 California heat wave: Impacts on hospitalizations and emergency department visits. Environmental Health Perspectives, 117, 61–67.Find this resource:

Kovats, R. S., & Ebi, K. L. (2006). Heatwaves and public health in Europe. European Journal of Public Health 16, 592–599.Find this resource:

Lalande, F., Legrain, S., Valleron, A. J., & Meyniel, D. (2003). Mission d’expertise et d’evaluation du systeme de sant’e pendant la canicule 2003, Minist`ere de la sant’e, de la famille, et des personnes handicap’ees. Paris.Find this resource:

Luber, G., & McGeehin, M. (2008). Climate change and extreme heat events. American Journal of Preventive Medicine, 35, 429–435.Find this resource:

Masood, I. Z., Majid, S., Sohail, A., Zia, S., & Raza, S. (2015). The deadly heat wave of Pakistan, June 2015. International Journal of Occupational and Environmental Medicine, 6(4), 247–248.Find this resource:

Meehl, G. A., Tebaldi, C., Walton, G., Easterling, D., & McDaniel, L. (2009). Relative increase of record high maximum temperatures compared to record low minimum temperatures in the U.S.. Geophysical Research Letters, 36, L23701.Find this resource:

Min, S.­K., Zhang, X., Zwiers, F., Shiogama, H., Tung, Y. S., & Wehner, M. (2013). Multi-model detection and attribution of extreme temperature changes. Journal of Climate, 26, 7430–7451.Find this resource:

Otto, F. E. L., Massey, N., vanOldenborgh, G. J., Jones, R. G., & Allen, M. R. (2012). Reconciling two approaches to attribution of the 2010 Russian heat wave. Geophysical Research Letters, 39, L04702.Find this resource:

Parsons, K. (2006), Heat stress standard ISO 7243 and its global application. Industrial Health, 44, 368–379.Find this resource:

Perkins, S. E. (2015). A review on the scientific understanding of heatwaves—their measurement, driving mechanisms, and changes at the global scale. Journal of Atmospheric Research, 164, 242–267.Find this resource:

Reid, C. E., O’Neill, M. S., Gronlund, C. J., Brines, S. J., Brown, D. G., Diez-Roux, A. V., & Schwartz, J. (2009). Mapping community determinants of heat vulnerability. Environmental Health Perspectives, 117, 1730–1736.Find this resource:

Robine, J. M., Cheung, S. L. K., Le Roy, S., Van Oyen, H., Griffiths, C., Michel, J. P., & Herrmann, F. R. (2008). Death toll exceeded 70,000 in Europe during the summer of 2003. Comptes Rendus Biologies, 331(2), 171–178.Find this resource:

Salim, A., Ahmed, A., Ashraf, N., & Ashar, M. (2015). Deadly heat wave in Karachi, July 2015: Negligence or mismanagement? International Journal of Occupational and Environmental Medicine, 6, 249.Find this resource:

Santer, B. D., Mears, C., Wentz, F. J., Taylor, K. E., Gleckler, P. J., Wigley, T. M. L., et al. (2007). Identification of human-induced changes in atmospheric moisture content. Proceedings of the National Academy of Sciences, 107, 15248–15253Find this resource:

Seneviratne, S. I., Nicholls, N., Easterling, D., Goodess, C. M., Kanae, S., Kossin, J., et al. (2012). Changes in climate extremes and their impacts on the natural physical environment In C. B. Field, V. Barros, T. F. Stocker, D. Qin, D. J. Dokken, K. L. Ebi, et al. (Eds.), Managing the risks of extreme events and disasters to advance climate change adaptation [A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change (IPCC)] (pp. 109–230). Cambridge, U.K.: Cambridge University Press.Find this resource:

Settele, J., Scholes, R., Betts, R., Bunn, S., Leadley, P., Nepstad, D, et al. (2014). Terrestrial and inland water systems. In C. B. Field, V. R. Barros, D. J. Dokken, K. J. Mach, M. D. Mastrandrea, T. E. Bilir, et al. (Eds.), Climate change 2014: Impacts, adaptation, and vulnerability. Part A: Global and sectoral aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (pp. 271–359). Cambridge, U.K.: Cambridge University Press.Find this resource:

Shaposhnikov, D., Revich, B., Bellander, T., Bedada, G. B., Bottai, M., Kharkova, T., et al. (2014). Mortality related to air pollution with the Moscow heat wave and wildfire of 2010. Epidemiology, 25, 359–364.Find this resource:

Sherwood, S. C., & Huber, M. (2010). An adaptability limit to climate change due to heat stress. Proceedings of the National Academies of Science, 107, 9552–9555.Find this resource:

Singh, R., Joshi, P. C., & Kishtawal, C. M. (2005). A new technique for estimation of surface latent heat fluxes using satellite-based observations. Monthly Weather Review, 133, 2692–2710.Find this resource:

Smith, K. R., Woodward, A., Campbell-Lendrum, D., Chadee, D. D., Honda, Y., Liu, Q., et al. (2014). Human health: Impacts, adaptation, and co-benefits. In C. B. Field, V. R. Barros, D. J. Dokken, K. J. Mach, M. D. Mastrandrea, T. E. Bilir, et al. (Eds.), Climate change 2014: Impacts, adaptation, and vulnerability. Part A: Global and sectoral aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (pp. 709–754). Cambridge, U.K.: Cambridge University Press.Find this resource:

Steadman, R. G. (1979a). The assessment of sultriness. Part I: A temperature-humidity index based on human physiology and clothing science. Journal of Applied Meteorology, 18, 861–873.Find this resource:

Steadman, R. G. (1979b). The assessment of sultriness. Part II: Effects of wind, extra radiation and barometric pressure on apparent temperature. Journal of Applied Meteorology, 18, 874–885.Find this resource:

Stott, P. A., Stone, D. A., & Allen, M. R. (2004). Human contribution to the European heatwave of 2003. Nature, 432, 610–613.Find this resource:

Stull, R. (2011). Wet-Bulb temperature from relative humidity and air temperature. Journal of Applied Meteorology and Climatology, 50, 2267–2269.Find this resource:

Tebaldi, C., & Wehner, M. (2016). Benefits of mitigation for future heat extremes under RCP4.5 compared to RCP8.5. Climatic Change, 5, 1–13.Find this resource:

Trent, R. B. (2007). Review of July 2006 heat wave related fatalities in California (manuscript). California Department of Health Services.Find this resource:

Tsutsumi, H., Tanabe, S., Harigaya, J., Iguchi, Y., & Nakamura, G. (2007). Effect of humidity on human comfort and productivity after step changes from warm and humid environment. Building and Environment, 42, 4034–4042.Find this resource:

Vaneckova, P., Neville, G, Tippett, V, Aitken, P., Fitzgerald, G., & Tong, S. (2011). Do biometeorological indices improve modeling outcomes of heat-related mortality? Journal of Applied Meteorology and Climatology, 50, 1165–1176.Find this resource:

Walsh, J., Wuebbles, D., Hayhoe, K., Kossin, J., Kunkel, K., Stephens, G., et al. (2014). Our changing climate: Climate change impacts in the United States: The third national climate assessment. J. M. Melillo, T. C. Richmond, & G. W. Yohe (Eds.), U.S. Global Change Research Program.Find this resource:

Watts, N., Adger, W. N., Agnolucci, P., Blackstock, J., Byass, P., Cai, W., et al. (2015). Health and climate change: Policy responses to protect public health. Lancet, 386, 1861–1914.Find this resource:

Weller, G. B. (2013). Joint tail modeling via regular variation with applications in climate and environmental studies (PhD diss.). Colorado State University, Fort Collins, CO.Find this resource:

Westerling, A., Bryant, B., Preisler, H., Holmes, T., Hidalgo, H., Das, T., & Shrestha, S. (2011). Climate change and growth scenarios for California wildfire. Climatic Change, 109, 445–463.Find this resource:

Willett, K. M., Gillett, N. P., Jones, P. D., & Thorne, P. W. (2007). Attribution of observed surface humidity changes to human influence. Nature, 449, 710–713.Find this resource:

## Notes:

(1.) Note that at any instant in time, water vapor over land may also be provided locally by evaporation from the soil; however, the ultimate source of that water was also the oceans.

(2.) The careful reader will note that the dry regions of Southern Europe exhibit a very pronounced decrease. In this region, the effect of the expansion of the Hadley cell, a change in the general circulation of the atmosphere, has a profound drying effect on all aspects of the hydrological cycle. This is but one example of the complicated interactions of the multiple mechanisms of climate change due to global warming.

(3.) Up to hourly at some stations.