Those living in warmer climates lose more sleep per degree of temperature rise
The elderly, women, and residents of lower-income countries are impacted most
Insufficient sleep is a risk factor for several adverse physical and mental outcomes. A lack of sleep has been associated with reduced cognitive performance, diminished productivity, compromised immune function, adverse cardiovascular outcomes, depression, anger, and suicidal behavior. High ambient temperatures have been associated with reduced subjective sleep quality, but little is known regarding the influence of outdoor weather conditions and rising outdoor temperatures on objective measures of sleep globally. When linked with global weather and climate measurements, sleep-tracking data from wristbands reveal that warmer nighttime temperatures do indeed harm sleep, with unequal effects. The elderly, residents of lower-income countries, females, and those already living in hotter climates are disproportionately impacted. Further analysis reveals that elevated ambient temperatures may already be impairing human sleep globally. Without further adaptation, and should greenhouse gas concentrations not be stabilized until the end of the century, each person could be subjected to an average of 2 weeks of temperature-attributed short sleep each year.
Ambient temperatures are rising worldwide, with the greatest increases recorded at night. Concurrently, the prevalence of insufficient sleep is rising in many populations. Yet it remains unclear whether warmer-than-average temperatures causally impact objective measures of sleep globally. Here, we link billions of repeated sleep measurements from sleep-tracking wristbands comprising over 7 million sleep records (n = 47,628) across 68 countries to local daily meteorological data. Controlling for individual, seasonal, and time-varying confounds, increased temperature shortens sleep primarily through delayed onset, increasing the probability of insufficient sleep. The temperature effect on sleep loss is substantially larger for residents from lower-income countries and older adults, and females are affected more than males. Those in hotter regions experience comparably more sleep loss per degree of warming, suggesting limited adaptation. By 2099, suboptimal temperatures may erode 50–58 h of sleep per person-year, with climate change producing geographic inequalities that scale with future emissions.
Summarizing our empirical results, we find that adults fall asleep later, rise earlier, and sleep less during hot nights. Deviating from the results of laboratory studies that constrained adaptive behavior, we show that increases in nighttime temperature reduce time slept across the global temperature distribution, with effects increasing in magnitude as temperatures become hotter. The effect of a 1°C increase in minimum temperature among the elderly is over twice the effect observed in other age groups. Further, the effect is nearly three times as large among globally poorer individuals as it is among individuals in richer nations and is significantly larger in females as compared with males. We do not find evidence of sleep adaptation to warmer temperatures within days, between days, across summer months, or between climate regions. Indeed, the sleep impact per degree of temperature increase in warmer locations is significantly larger than in colder locations. Our results imply that suboptimal ambient temperatures likely already erode human sleep considerably early in the 21century. Coupling our model estimates with downscaled climate model output, we project that climate change may exacerbate global environmental inequalities by disproportionately eroding sleep in the warmest regions, with differential societal sleep impacts scaling with future atmospheric greenhouse gas concentrations. We verify that our primary conclusions are robust to alternative sample inclusion criteria, meteorological data, temporal controls, and outcome measures ( Tables S6–S20 S49 , and S50 Figures S2 and S3 ). Further, our modeling framework controls for any unobserved, fixed device characteristics, and we confirm that the period and frequency of sleep-tracking wristband use does not alter our primary results ( Experimental procedures Tables S32 and S33 ).
To investigate whether ambient temperature alters sleep, we pair our sleep observations of nighttime sleep duration (total sleep time) and timing (sleep onset, midsleep, and offset) with geolocated meteorological and climate data ( Figures 1 A and 1B; Experimental procedures ). We specify multivariate fixed-effects panel models—derived from the climate econometrics literature—with individual repeated measures, using as good as random variation in meteorological variables relative to local averages to estimate the total effect of ambient nighttime temperature on individual sleep outcomes ( Tables S6 and S7 ). An advance of the present study is that our dataset allows us to control for all stable individual characteristics and leverage within-person fluctuations in both weather exposures and sleep outcomes to isolate the plausibly causal effect of nighttime temperature on our person-level sleep outcomes while controlling for other potentially confounding individual-level, calendar-date-specific, and subnational administrative region-by-month spatiotemporal factors that might otherwise bias inference between temperature exposures and sleep outcomes. Importantly, this statistical model also controls for location-by-date historical climate normals and cloud-cover alterations in daylight, removing the potential confounding effect of seasonality from our analyses ( Tables S6–S8 and S30 ). Thus, whereas sleep laboratory research in this setting typically manipulates ambient room temperatures while limiting behavioral adaptation, the present study seeks to instead estimate the total effect of quasi-random changes in outside ambient temperatures on sleep patterns, allowing for habitual behavioral adjustments to temperature, including possible responses to the environmental information conveyed by outdoor conditions. This latter point is important for studying temperature-sleep relationships under ecologically valid circumstances, because even awareness of outdoor ambient conditions while indoors may impact sleep behavior at night.
Fixed effects models versus mixed effects models for clustered data: reviewing the approaches, disentangling the differences, and making recommendations.
In contrast to the limited precision and resolution of the subjective and indirect measures employed by previous studies—even the largest of which only used data from one country—the global reach of sleep-tracking wristbands holds promise for understanding the environmental determinants of human sleep. Here, we draw on a large-scale sleep dataset of over 10 billion sleep observations registered from 2015 to 2017, comprising 7.41 million repeated daily sleep records spanning 68 countries using accelerometry-based sleep-tracking wristbands linked to a smartphone application ( Figures 1 B and 1D ). This sleep dataset replicates established age, interregional, and socio-temporal sleep characteristics ( Experimental procedures Tables S1 and S2 Figure 1 C). Accelerometry-based sleep tracking devices are increasingly ubiquitous and particularly well suited for large-scale observational studies,offering several empirical advantages over previous research designs. In situ sleep measures from sleep-tracking wristbands provide dynamic spatial and temporal reference information for precise merging with meteorological data across diverse geographic regions, enabling the study of the effect of temperature on within-individual changes across the entire sleep period. Moreover, objective measures of total sleep duration can be used to investigate whether temperature affects the probability of obtaining short sleep, following standard definitions.
(D) Annual total number of nighttime sleep observations collected over the 2-year period from September 2015 through October 2017, in millions.
(C) Plot showing regular and dynamic temporal patterns in within-individual sleep duration deviation from average (in hours) over the 2016 calendar year. Each daily measure corresponds to the mean of all within-individual nightly sleep deviations for active users on that day. Recurring weekend peaks (above zero) and weekday valleys (below zero) reflect the imbalanced temporal structure of the adult working week—whereby sleep reduction during weekdays is partially compensated for on weekends with oversleep.
(B) World map depicting the country-level count of accelerometry-based sleep-tracking wristband users included in this study, spanning 68 countries from all continents except for Antarctica. Countries with relatively more users appear as darker shades of green.
(A) Plotted map of weather stations from the Global Historical Climatology Network-Daily (GHCND). Each blue dot represents one station.
Far less is known about the influence of outdoor ambient temperatures and meteorological conditions on adult sleep in real-world settings.Evidence from self-report studies indicates that the prevalence of reported sleep deficiencies increases in warm weather.The largest of these studies pooled data in the United States from nationally representative health surveys and found that higher monthly nighttime temperature anomalies increased self-reported nights of insufficient sleep during the previous month. However, retrospective self-reported sleep outcomes are notoriously imprecise, unreliable, and have been shown to have questionable internal validity.Thus, it remains an open question whether, and to what extent, ambient thermal and weather conditions might affect objective repeated measures of individual sleep duration and timing across a global adult population.
Agreement between simple questions about sleep duration and sleep diaries in a large online survey.
Self-reported and measured sleep duration: how similar are they?.
The effect of high indoor temperatures on self-perceived health of elderly persons.
Temperature and mental health: evidence from the spectrum of mental health outcomes.
Climate change and sleep: a systematic review of the literature and conceptual framework.
Prior research investigating the influence of ambient temperature on sleep in adults has been largely restricted to short controlled laboratory studies or imprecise self-report surveys. Humans and other mammals have developed both neurophysiological and behavioral processes to coordinate rhythms of thermoregulation and sleep, presumably to conserve energy expenditure.In humans, the maximal rate of core body cooling is strongly correlated with sleep onset, and sleep propensity peaks near the minimum of the core body temperature rhythm.Preceding sleep onset, increased blood flow to the distal skin and extremities enables cooling of the core body temperature.Both skin and core body temperatures become more sensitive to environmental temperature during sleep, and duration of wakefulness has been shown to increase when temperatures warm or cool outside of the thermoneutral zone—the range of ambient temperatures where the body can maintain its core temperature only through regulating dry heat loss (skin blood flow)—albeit under controlled conditions that constrain human adaptation.
Sleep under extreme environments: effects of heat and cold exposure, altitude, hyperbaric pressure and microgravity in space.
The thermophysiological cascade leading to sleep initiation in relation to phase of entrainment.
Natural sleep and its seasonal variations in three pre-industrial societies.
The energy allocation function of sleep: a unifying theory of sleep, torpor, and continuous wakefulness.
Converging evidence suggests that climate change is challenging human mental health and cognitive functioning, although the behavioral mechanisms remain unclear.Recent research based on self-reported data—limited to the United States—suggests that sleep may constitute one such pathway.Regular and sufficient sleep supports human physical and mental health.Short sleep duration is associated with reduced cognitive performance,diminished productivity,increased absenteeism,compromised immune function,and elevated risk of hypertension, adverse cardiovascular outcomes,mortality,depression, anger, and suicidal behaviors.Acute sleep restriction delays reaction times,increases accident risk,inhibits the neural encoding of new experiences to memory,and limits the clearance of neurotoxic metabolites from the brain linked to aging and neurodegenerative diseases.Nevertheless, growing proportions of industrialized populations do not obtain adequate sleep, a development attributed to lifestyle and environmental changes, but not yet fully understood.Concurrently, nighttime ambient temperatures are increasing due to both anthropogenic climate change and the expansion of urban heat islands.To inform policy, planning, and practice, more information is needed about the environmental factors that curtail or promote sufficient sleep globally, particularly the role played by outdoor ambient temperature.
Climate change and sleep: a systematic review of the literature and conceptual framework.
Updated analyses of temperature and precipitation extreme indices since the beginning of the twentieth century: the HadEX2 dataset.
Trends in self-reported sleep duration among US adults from 1985 to 2012.
A deficit in the ability to form new human memories without sleep.
Temperature and mental health: evidence from the spectrum of mental health outcomes.
Climate change and sleep: a systematic review of the literature and conceptual framework.
Identifying and preparing for the mental health burden of climate change.
Association between ambient heat and risk of emergency department visits for mental health among US adults, 2010 to 2019.
The 2021 report of the Lancet Countdown on health and climate change: code red for a healthy future.
Learning is inhibited by heat exposure, both internationally and within the United States.
Threats to mental health and wellbeing associated with climate change.
The case for systems thinking about climate change and mental health.
These globally averaged, population-weighted estimates mask considerable spatial heterogeneity in impacts, with projected geographic inequalities in temperature-driven sleep loss ( Figures 5 A–5D ) expected to increase over time. Without further adaptation by the end of the century, residents in the warmest areas are projected to experience over 23 h of additional temperature-driven sleep loss per year by 2099 under a high GHG concentration scenario and 8.5 h of additional sleep loss under an intermediate stabilization scenario ( Figures 5 A–5D and S6 ), net of existing temperature-attributed excess sleep loss apparent in 2010 ( Figure 4 Experimental procedures ). Similarly, future warming is projected to unequally increase temperature-driven short sleep. Disparities in estimated net sleep erosion increase both over time and across space between warmer and colder regions under all scenarios ( Figures 5 E–5H), but differential impacts are projected to be more modest under an increasingly plausible scenario in which atmospheric GHG concentrations stabilize by the end of the century (RCP4.5). By the end of the 21century, adults in the warmest regions are expected to experience approximately 3 additional nights of short sleep per year due to rising nighttime temperatures under the more moderate (RCP 4.5) scenario compared with upwards of 7 additional nights of short sleep under a “no policy” increasing GHG concentration (RCP 8.5) scenario. Critically, these projected changes are net of estimated temperature-attributed sleep impacts at the beginning of the 21century ( Figure 4 ), and our results indicate that suboptimal, warmer ambient temperatures likely already contribute to insufficient sleep globally ( Figure 2 D). Importantly, our historical estimates underlying these projections may also be conservative, since the majority of data arise from high-income countries and are skewed toward a middle-aged, male demographic ( Experimental procedures ). Indeed, our subgroup analyses indicate that future sleep loss may be larger by a factor of ∼3 for lower-income countries, a factor of ∼2 for demographically older populations and marginally higher for women ( Figure 3 ). Moreover, the cumulative impact of lagged temperature effects likely exceeds the contemporaneous estimates used in these projections ( Results Tables S39–S41 ). Future planetary-scale research is needed that systematically investigates the impact of rising temperatures and other climate hazards on the sleep outcomes of vulnerable populations, particularly those residing in low-income countries and communities.
(E–H) Global impact maps projecting the individual excess count of temperature-attributed nights of short sleep by 2050 and 2099 (net of 2010 temperature-attributed sleep loss) under a midcentury stabilized GHG concentration scenario (RCP4.5, left) and unmitigated increasing GHG concentration scenario (RCP8.5, right). Grid cells depict the additional annual number of short (<7 h) sleep nights per person for the specified 25 × 25 km area above the 2010 baseline. Brown colors indicate areas with relatively lower impacts on insufficient sleep, orange colors depict areas with moderate impacts, and red colors represent regions with more severe impacts. Planetary-scale societal sleep impacts due to ambient temperature accrue unevenly across regions and over time, with higher GHG concentrations leading to more pervasive severe impacts.
(A–D) World maps projecting annual individual temperature-attributed net sleep loss by 2050 and 2099 (net of 2010 temperature-driven sleep loss) under intermediate (RCP4.5, left) and high GHG concentration (RCP8.5, right) scenarios. Each colored grid cell represents the additional per-person annual sleep loss projected for the corresponding 25 × 25 km area using the ensemble average of 21 NASA bias-corrected and statistically downscaled CMIP5 models. Darker purple colors represent areas with the lowest projected annual net sleep losses from the 2010 baseline, while lighter yellow colors signal areas with the largest annual sleep reductions. Projections are only shown for countries in the dataset; other countries shown in gray. Geographic inequality in the magnitude of climate-change-driven sleep loss is evident already by 2050 and becomes more pronounced by the end of the century, with global inequalities scaling with the level of future emissions.
Finally, abiding by the assumption that future sleep loss will respond to projected changes in nighttime minimum temperature as they have responded to them in the recent past, we construct projections for the impact of ambient warming on cumulative annual individual sleep loss and temperature-attributed short sleep for all countries within our dataset ( Figure 1 B). Consistent with the climate impacts literature, we draw upon gridded data from 21 global climate models run under both end-of-century stabilization (Representative Concentration Pathway 4.5 [RCP4.5]) and increasing (RCP8.5) atmospheric greenhouse gas concentration scenarios and link downscaled, nighttime temperature projections with our spline regression model to compute the per-person average annual excess sleep loss expected at the beginning, middle, and end of the century ( Figures 4 A and 4B ; Experimental procedures ). Taking a world-population-weighted average across all grid-cell-level daily time series and each of the 21 climate models, we estimate that annual individual excess sleep loss due to suboptimal nighttime temperature was likely considerable near the beginning of the 21century, with temperatures eroding an estimated 44 excess hours of sleep per person annually on average ( Figure 4 A) and contributing approximately 11 additional nights of short sleep per person annually, based on downscaled climate simulation data for 2010 ( Figure 4 B). Total annual sleep loss due to warming nighttime temperatures may steadily increase by midcentury, with yearly losses becoming markedly larger by 2099 under an increasing greenhouse gas (GHG) scenario but only moderately larger under a scenario in which atmospheric GHG concentrations stabilize by the end of the century ( Figure 4 A). Mean projected temperature-attributed individual excess sleep loss in 2099 varies from ∼50 (range: 46.7–53.6) h per year in a stabilized GHG concentration scenario (RCP4.5) to ∼58 (range: 52.7–65.2) h in an increasing GHG concentration scenario (RCP8.5). Sleep erosion is projected to exact growing societal sleep impacts, with the number of temperature-attributed short nights of sleep estimated to increase from ∼11 per person per year in 2010 to ∼12 short nights per year by 2050 under an intermediate RCP4.5 scenario. By the end of the century, nighttime temperatures may contribute to ∼13 (range: 11.9–13.8) short nights of sleep per person per year under the stabilized RCP4.5 pathway and over ∼15 (range: 13.6–17.0) short nights of sleep per year under the increasing GHG concentration (RCP8.5) pathway.
(B) Mean individual projections of the cumulative annual count of short (<7 h) sleep nights per person due to nighttime ambient temperature in 2010, 2050, and 2099. Rain cloud plots depict the distributions of projected annual sleep impacts across 21 bias-corrected and statistically downscaled CMIP5 climate models under RCP4.5 (purple) and RCP8.5 (orange) warming scenarios. Vertical marks beneath each distribution depict CMIP5 climate-model-specific projection estimates, with the median model projections shown as darkened marks.
(A) Global population-weighted average individual-level projections for the impact of elevated nighttime temperatures on hours of sleep loss under a midcentury stabilized atmospheric GHG concentration scenario (RCP4.5 in purple) and an increasing GHG concentration climate-change scenario (RCP8.5 in orange). Each line represents the estimated annual total per-person excess sleep loss due to suboptimal ambient temperature for a different downscaled climate model projection, averaged across all country-level pixels within the dataset. The dark colored lines plot the scenario-specific ensemble mean projected loss across 21 statistically downscaled global CMIP5 climate models. Sleep loss increases over time due to projected warming across all countries.
Since prior research suggests that people may be able to physiologically or behaviorally acclimatize to warmer temperatures over relatively short periods of time,we further assess possible intra-annual and inter-day sleep adaptation to ambient temperature. First, we test whether human sleep responds differently to nighttime temperature increases experienced during the first month of summer—when nights with locally hotter temperatures are relatively newer—versus the last month of summer when they are more familiar.While short-run acclimatization would be apparent if the effect of temperature on sleep duration diminishes from the first to the last month of summer, we instead find evidence that nighttime temperatures appear to incur similar to marginally more sleep loss near the end of summer, when warmer temperatures are relatively less novel ( Experimental procedures Table S38 ). Second, after a given temperature exposure, people may physiologically adapt or otherwise shift when they get sleep via intertemporal substitution across days. For instance, short-term reductions in sleep due to temperature may in turn increase homeostatic sleep pressure that facilitates subsequent recovery of initial sleep loss.Alternatively, previous days’ thermal conditions may further disrupt sleep via delayed impacts not captured by our contemporaneous effect estimates. To account for this, we estimate a distributed lag model that includes lagged minimum temperature terms from the previous 7 days. We find that an increase in nighttime temperature produces additional delayed sleep loss: the sum of the contemporaneous and lagged coefficients is ∼30% larger than the contemporaneous effect of temperature alone ( Table S39 ). Coefficients remain negative through the 5lag, with cumulative sleep loss growing up until a small partial rebound on days 6 and 7. These results persist when including lags for all weather variables or including temperature lags for each of the preceding 14 days ( Tables S39 and S40 ), indicating that a rise in ambient temperature yields cumulatively larger net sleep loss rather than delayed sleep substitution.
Temperature and the allocation of time: implications for climate change.
Temperature and the allocation of time: implications for climate change.
Those who reside in warmer areas with generally higher temperatures may respond to temperature increases differently than those living in colder regions.To investigate whether the effect of temperature differs by ambient climate context, we conducted a heterogeneity analysis by decile of average local minimum temperature over the 2015–2017 period ( Experimental procedures Figure 3 E). We find that, compared with the coldest average temperature decile, marginal effects of minimum temperature on sleep loss are significantly larger for residents of warmer deciles (4–10deciles). Those living in hotter regions experience comparably more sleep loss per degree of warming, suggestive of limited adaptation in warmer climates. These results appear consistent with our binned temperature specifications ( Figure 2 A), showing that temperature increases at colder temperatures yield smaller effect sizes.
Valuing the Global Mortality Consequences of Climate Change Accounting for Adaptation Costs and Benefits, Report No. 27599.
To determine whether increases in minimum temperatures impact human sleep differently over the course of the year, we inspect the marginal effect of temperature on sleep loss across each season, accounting for hemispheric differences in seasonality. Nighttime temperature increases result in sleep loss throughout the year ( Figure 3 D). Consistent with the annual temperature distribution, we find that rising nighttime temperatures decrease sleep duration the most during summer months (coefficient: −0.55), followed by fall (coefficient: −0.35), spring (coefficient: −0.25), and winter (coefficient: −0.20) months (all coefficients are significantly different from zero at the p < 0.01 level). The per-degree effect of an increase in nighttime temperature on sleep loss during the summer is nearly three times larger than in the winter. Our results provide further evidence that temperature increases impart the largest losses on human sleep when nighttime temperatures are already elevated ( Figure 2 A; Table S21 ).
Since access to infrastructure, cooling technologies, and other unobserved environmental resources may plausibly modify the extent to which temperature impacts sleep, we further test whether our results differ across country-income levels. We find that the effect of minimum nighttime temperature on human sleep loss is substantially larger for people residing within lower-middle-income countries (coefficient: −0.85) compared with countries with higher income levels ( Figure 3 C; Tables S25 and S26 ). The negative effect of nighttime temperature on sleep duration is 2.8 times greater (p = 0.087) for residents in lower-middle-income countries compared with those from high-income countries (coefficient: −0.30) and 3.6 times greater (p = 0.057) compared with upper-middle-income countries (coefficient: −0.23). Collectively, these results provide initial evidence that countries from all observed income levels are sensitive to the effect of ambient nighttime temperature on sleep, but the amount of sleep loss per degree increase may be disproportionately larger for people in lower-middle-income countries.
Under identical conditions, females’ core body temperatures decrease earlier in the evening compared with males,possibly exposing females to higher environmental temperatures around their time of habitual sleep onset. Females have also been shown to have greater subcutaneous fat thickness, which might impair nocturnal heat loss.Comparing the effect of minimum temperature on sleep duration between sexes reveals that the per-degree negative impact of nighttime temperature rise is significantly (p < 0.01) but only slightly larger for females (coefficient: −0.34) than males (coefficient: −0.27) in our dataset ( Figure 3 B). This finding adds to evidence that females may be more predisposed to adverse heat effects on health than males.
Individual and environmental demographic factors may modify the impact of temperature on sleep. Older adulthood is marked by an attenuated thermoregulatory response to suboptimal environmental temperatures, earlier sleep timing, and reduced total sleep duration.Such age-related developments may increase the nocturnal sensitivity of the elderly to higher ambient temperatures, possibly challenging sleep demand. We find that older adults (>65) are markedly more sensitive to exogenous increases in nighttime ambient temperature than mid-aged adults and young adults ( Figure 3 A ). The per-degree effect of nighttime temperature on lost sleep for older adults (coefficient: −0.61) is over two times (p < 0.01) the effect estimated for mid-age adults (coefficient: −0.28). These results add to increasing evidence of the age-related ambient temperature sensitivity of sleep.To explore the emergence of heightened temperature sensitivity in later life, we run an alternative specification featuring smaller age groups for every 10-year increment above 30 years of age. We show that heightened temperature sensitivity may emerge rapidly after age 60 and increase further beyond age 70 ( Table S23 ).
(E) Marginal effects of a 1°C increase by average minimum temperature decile over the 2015–2017 period, from coldest (dark blue) to warmest (red) locations. Temperature increases exert larger impacts in warmer regions compared with colder regions. Error bars represent 95% confidence intervals. ∗∗∗∗p < 0.001, ∗∗∗p < 0.01, ∗∗p < 0.05, and ∗p < 0.1.
(D) The marginal effects of nighttime minimum temperature by season of the year on sleep loss. Temperature increases are associated with the greatest sleep losses during summer nights, followed by fall, spring, and winter nights.
(C) Plot of the marginal effects of nighttime minimum temperature by country-level gross national income (GNI). The effect of temperature on sleep loss is substantially larger for people residing within lower-middle-income countries (n = 995 adults; n = 14,639 observations) compared with upper-middle- (n = 5,910 adults; n = 274,488 observations) and high-income countries (n = 38,675 adults; n = 4,116,044 observations).
(B) The marginal effects of temperature by sex on sleep loss. Those who identify as female (n = 13,302 adults; n = 1,279,271 observations) lose more sleep per degree increase in minimum temperature compared with those who identify as male (n = 29,811 adults; n = 3,125,900 observations).
(A) The marginal effect of temperature by age category on sleep loss produced by interacting age group with nighttime minimum temperature within our primary model specification ( Experimental procedures Equation 2 ). The marginal effect of increasing temperature by 1°C on sleep loss is nearly twice the magnitude for older adults (n = 1,289 adults; n = 155,922 observations) compared with mid-aged adults (n = 39,460 adults; n = 4,078,623 observations) and young adults (n = 2,364 adults; n = 170,626 observations).
To investigate how the entire sleep period responds to temperature-driven sleep loss, we construct separate flexible models to predict sleep onset, midsleep, and offset timing. Drawing on these combined estimates, we show that rising temperatures compress the human sleep period through both a larger delay in sleep onset and a moderate advance in sleep offset. As minimum temperatures rise above −10°C, delays in sleep onset curtail sleep duration ( Figures 2 A–2C and 2E) and marginally delay midsleep. By contrast, nighttime temperature increases advance sleep offset timing when temperatures exceed 15°C. Thus, larger declines in sleep duration at warmer nighttime temperatures are jointly driven by both delays in sleep onset and advances in sleep offset, constricting the human sleep period and slightly delaying midsleep ( Tables S12–S14 ).
Our finding that human sleep is unidirectionally sensitive to increasing ambient temperatures across the temperature distribution differs from previous experimental studies that found reductions in sleep under both high and low environmental temperatures.Instead, our within-person global analysis uncovers a similar functional relationship as those identified by prior national survey analyses using subjective measures of sleep.In real-world settings, humans appear to be better at adapting their surroundings to obtain sufficient sleep under cooler outside conditions, whereas sleep loss increases with rising ambient temperatures. Since other meteorological factors may also influence sleep, we use our primary flexible model specification ( Experimental procedures Equation 1 ) to estimate the human sleep response to changes in weather. Sleep loss increases further as a function of the diurnal temperature range—the difference between daily maximum and minimum temperature. This result is directionally consistent with the diurnal temperature range attributed mortality response identified by a recent multi-country analysis.Since our specified model controls for other weather variables, including cloud cover and relative humidity, two plausible explanations are that indoor environments may retain heat gained during the day or that daytime heat may impart physiological demands that extend into the sleep period. Importantly, diurnal temperature range is projected to increase annually over Europeand separately across most other regions during summer months under a high-emissions, climate- change scenario.By contrast, high levels of precipitation, wind speed, and cloud cover each marginally increase sleep duration ( Figure S1 ). Compared with moderate levels of relative humidity, both low and high levels reduce sleep, with the former producing greater sleep reduction, providing initial evidence that dry conditions may curtail sleep.
Temperature and mental health: evidence from the spectrum of mental health outcomes.
Sleep under extreme environments: effects of heat and cold exposure, altitude, hyperbaric pressure and microgravity in space.
Our results are robust, even when employing more extreme thresholds of short sleep, including <6 h and <5 h, demonstrating that marginal losses in total sleep time with rising temperatures predisposes people to insufficient sleep attainment ( Figure 2 D; Tables S9 and S10 ). Further, since prior evidence from aggregated mobile phone calling data suggests that people may compensate for seasonal sleep reductions during the summer with afternoon naps, we check that our primary results are robust to replacing nighttime sleep duration with 24-h sleep duration.Contrary to the hypothesis that total sleep time might be conserved, we find that including daytime sleep actually slightly increases the effect size of temperature on within-individual sleep loss within our sample ( Tables S8 and S29 ). Moreover, constraining our sample to only include high-income countries does not alter our primary results ( Tables S49 and S50 ).
The results of our binned temperature regressions indicate that exogenous increases in nighttime ambient temperature reduce adult sleep duration across nearly the entire observed temperature distribution ( Figures 2 A and S2 ). Climate change is projected to continue to increase the magnitude and frequency of extreme nighttime temperatures beyond the recent historical record. Our data indicate that, on very warm nights (>30°C), sleep declines by 14.08 min (−10.61 to −17.55) compared with nights with the lowest temperature-attributed sleep loss in our sample. Increasing nighttime temperatures amplify the estimated probability of obtaining a short night of sleep, measured with multiple standard definitions for insufficient sleep.The probability of sleeping less than 7 h increases gradually up to 10°C, before increasing at an elevated rate. Nighttime minimum temperatures greater than 25°C increase the probability of getting less than 7 h of sleep by 3.5 percentage points compared with the temperature baseline of 5°C–10°C ( Figure 2 D). However, our results show that the optimum nighttime ambient temperature for sufficient sleep may be considerably lower than this baseline, with nighttime heat inducing short sleep across most of the temperature distribution. Providing scale for this estimated relationship, exposure to nighttime temperatures exceeding 25°C, if extrapolated for an equivalent population of 100,000 adults across a single night, would result in 4,600 additional individuals obtaining a short <7-h night of sleep compared with the estimated optimum minimum nighttime temperature.
(D) A plot of the predicted change in the probability of obtaining a short night of sleep across each minimum temperature bin. As temperature increases above 5°C, the probability of obtaining a short night of sleep—measured with three standard criteria—also increases.
(C) Nighttime temperature increases above −10°C marginally delay midsleep—the midpoint of the human sleep period—although the magnitude of change at higher temperatures is smaller than concomitant changes in sleep onset and offset.
(B) High nighttime temperatures significantly compress the human sleep period, primarily through a delay in sleep onset and a marginally smaller advance in sleep offset. Sleep offset advances under higher nighttime temperatures >15°C, while very cold temperatures below −10°C delay offset timing.
(A) Plot of the relationship between increases in nighttime minimum temperature and the average within-individual change in sleep duration for each temperature bin. As minimum temperatures rise, sleep duration decreases with a steeper linear decline when temperatures exceed 10°C. Shaded regions represent 95% confidence intervals computed using heteroskedasticity-robust standard errors clustered on the first administrative division level. Histograms plot the distribution of observed nighttime temperatures across millions of sleep observations. We confirm sufficient observational support across all temperature bins ( Table S5 ).
Discussion
In summary, we provide extensive evidence that human sleep is sensitive to nighttime ambient temperature, posing an additional climate-change-related threat to global public health and human well being. Increases in nighttime minimum temperature reduce sleep duration and increase the probability of obtaining insufficient sleep via the constriction of the human sleep period, primarily by delaying when people fall asleep. The effect of nighttime temperature on sleep loss is amplified for lower-income countries, older adults, and females. Our results suggest that temperature-driven sleep loss is evident across demographics, and increasing temperatures lead to some within-person sleep loss across all seasons, with the largest losses during the warmest months and on nights when minimum temperatures exceed 10°C. We do not find evidence of short-term acclimatization of sleep to warmer temperatures via intra-day, inter-day, or intra-annual substitution, and the marginal effect of increasing temperature is even larger for those already living in globally warmer regions compared with those residing in colder areas. Taken together, we find limited evidence of human sleep adaptation to hotter temperatures. We estimate that suboptimal nighttime temperatures likely already inflict considerable individual sleep loss early in the 21st century, and thus, increasing nighttime temperatures may further erode human sleep into the future. The burden of future warming will not be evenly distributed, barring further adaptation and mitigation, with people living in hotter climates expected to lose considerably more hours of sleep per year by 2099, contributing to societal impacts that scale with the level of future atmospheric GHG concentrations. Taken together, our results demonstrate that temperature-driven sleep loss likely has and may continue to exacerbate global environmental inequalities.
1 Berry H.L.
Waite T.D.
Dear K.B.G.
Capon A.G.
Murray V. The case for systems thinking about climate change and mental health. , 4 Evans G.W. Projected behavioral impacts of global climate change. , 7 Romanello M.
McGushin A.
Di Napoli C.
Drummond P.
Hughes N.
Jamart L.
Kennard H.
Lampard P.
Solano Rodriguez B.
Arnell N.
et al. The 2021 report of the Lancet Countdown on health and climate change: code red for a healthy future. , 9 Obradovich N.
Minor K. Identifying and preparing for the mental health burden of climate change. , 10 Hwong A.R.
Wang M.
Khan H.
Chagwedera D.N.
Grzenda A.
Doty B.
Benton T.
Alpert J.
Clarke D.
Compton W.M. Climate change and mental health research methods, gaps, and priorities: a scoping review. , 45 Carleton T.A.
Hsiang S.M. Social and economic impacts of climate. , 57 Hajat S.
O’Connor M.
Kosatsky T. Health effects of hot weather: from awareness of risk factors to effective health protection. , 58 Patz J.A.
Frumkin H.
Holloway T.
Vimont D.J.
Haines A. Climate change. , 59 Watts N.
Amann M.
Arnell N.
Ayeb-Karlsson S.
Belesova K.
Boykoff M.
Byass P.
Cai W.
Campbell-Lendrum D.
Capstick S.
et al. The 2019 report of the Lancet Countdown on health and climate change: ensuring that the health of a child born today is not defined by a changing climate. , 60 Vergunst F.
Berry H.L. Climate change and children’s mental health: a developmental perspective. , 61 Liu J.
Varghese B.M.
Hansen A.
Xiang J.
Zhang Y.
Dear K.
Gourley M.
Driscoll T.
Morgan G.
Capon A.
Bi P. Is there an association between hot weather and poor mental health outcomes? A systematic review and meta-analysis. , 62 Vicedo-Cabrera A.M.
Scovronick N.
Sera F.
Royé D.
Schneider R.
Tobias A.
Astrom C.
Guo Y.
Honda Y.
Hondula D.M.
et al. The burden of heat-related mortality attributable to recent human-induced climate change. , 63 Ebi K.L.
Vanos J.
Baldwin J.W.
Bell J.E.
Hondula D.M.
Errett N.A.
Hayes K.
Reid C.E.
Saha S.
Spector J.
Berry P. Extreme weather and climate change: population health and health system implications. , 64 Middleton J.
Cunsolo A.
Jones-Bitton A.
Wright C.J.
Harper S.L. Indigenous mental health in a changing climate: a systematic scoping review of the global literature. 5 Park R.J.
Goodman J.
Behrer A.P. Learning is inhibited by heat exposure, both internationally and within the United States. , 6 Burke M.
González F.
Baylis P.
Heft-Neal S.
Baysan C.
Basu S.
Hsiang S. Higher temperatures increase suicide rates in the United States and Mexico. , 7 Romanello M.
McGushin A.
Di Napoli C.
Drummond P.
Hughes N.
Jamart L.
Kennard H.
Lampard P.
Solano Rodriguez B.
Arnell N.
et al. The 2021 report of the Lancet Countdown on health and climate change: code red for a healthy future. , 8 Nori-Sarma A.
Sun S.
Sun Y.
Spangler K.R.
Oblath R.
Galea S.
Gradus J.L.
Wellenius G.A. Association between ambient heat and risk of emergency department visits for mental health among US adults, 2010 to 2019. , 13 Mullins J.T.
White C. Temperature and mental health: evidence from the spectrum of mental health outcomes. , 53 Obradovich N.
Migliorini R.
Paulus M.P.
Rahwan I. Empirical evidence of mental health risks posed by climate change. , 65 Baylis P.
Obradovich N.
Kryvasheyeu Y.
Chen H.
Coviello L.
Moro E.
Cebrian M.
Fowler J.H. Weather impacts expressed sentiment. , 66 Parks R.M.
Bennett J.E.
Tamura-Wicks H.
Kontis V.
Toumi R.
Danaei G.
Ezzati M. Anomalously warm temperatures are associated with increased injury deaths. , 67 Zhang P.
Deschenes O.
Meng K.
Zhang J. Temperature effects on productivity and factor reallocation: evidence from a half million Chinese manufacturing plants. , 68 Dasgupta S.
van Maanen N.
Gosling S.N.
Piontek F.
Otto C.
Schleussner C.F. Effects of climate change on combined labour productivity and supply: an empirical, multi-model study. , 69 Park R.J.
Goodman J.
Hurwitz M.
Smith J. Heat and learning. , 70 Lu P.
Zhao Q.
Xia G.
Xu R.
Hanna L.
Jiang J.
Li S.
Guo Y. Temporal trends of the association between ambient temperature and cardiovascular mortality: a 17-year case-crossover study. , 71 Wang J.
Obradovich N.
Zheng S. A 43-million-person investigation into weather and expressed sentiment in a changing climate. , 72 Baylis P. Temperature and temperament: evidence from twitter. , 73 Garg T.
Jagnani M.
Taraz V. Temperature and human capital in India. 13 Mullins J.T.
White C. Temperature and mental health: evidence from the spectrum of mental health outcomes. , 15 Killgore W.D.S. Effects of sleep deprivation on cognition. , 18 Barnes C.M.
Watson N.F. Why healthy sleep is good for business. , 19 Irwin M.R. Why sleep is important for health: a psychoneuroimmunology perspective. , 20 Cappuccio F.P.
Cooper D.
D’Elia L.
Strazzullo P.
Miller M.A. Sleep duration predicts cardiovascular outcomes: a systematic review and meta-analysis of prospective studies. , 21 Jackson C.L.
Redline S.
Emmons K.M. Sleep as a potential fundamental contributor to disparities in cardiovascular health. , 23 Goldstein A.N.
Walker M.P. The role of sleep in emotional brain function. , 24 Bernert R.A.
Kim J.S.
Iwata N.G.
Perlis M.L. Sleep disturbances as an evidence-based suicide risk factor. , 25 Czeisler C.A.
Wickwire E.M.
Barger L.K.
Dement W.C.
Gamble K.
Hartenbaum N.
Ohayon M.M.
Pelayo R.
Phillips B.
Strohl K.
et al. Sleep-deprived motor vehicle operators are unfit to drive: a multidisciplinary expert consensus statement on drowsy driving. 23 Goldstein A.N.
Walker M.P. The role of sleep in emotional brain function. , 74 Minkel J.D.
Banks S.
Htaik O.
Moreta M.C.
Jones C.W.
McGlinchey E.L.
Simpson N.S.
Dinges D.F. Sleep deprivation and stressors: evidence for elevated negative affect in response to mild stressors when sleep deprived. 23 Goldstein A.N.
Walker M.P. The role of sleep in emotional brain function. , 24 Bernert R.A.
Kim J.S.
Iwata N.G.
Perlis M.L. Sleep disturbances as an evidence-based suicide risk factor. 20 Cappuccio F.P.
Cooper D.
D’Elia L.
Strazzullo P.
Miller M.A. Sleep duration predicts cardiovascular outcomes: a systematic review and meta-analysis of prospective studies. , 21 Jackson C.L.
Redline S.
Emmons K.M. Sleep as a potential fundamental contributor to disparities in cardiovascular health. , 22 Cappuccio F.P.
D’Elia L.
Strazzullo P.
Miller M.A. Sleep duration and all-cause mortality: a systematic review and meta-analysis of prospective studies. 15 Killgore W.D.S. Effects of sleep deprivation on cognition. , 16 Krause A.J.
Simon E.B.
Mander B.A.
Greer S.M.
Saletin J.M.
Goldstein-Piekarski A.N.
Walker M.P. The sleep-deprived human brain. 25 Czeisler C.A.
Wickwire E.M.
Barger L.K.
Dement W.C.
Gamble K.
Hartenbaum N.
Ohayon M.M.
Pelayo R.
Phillips B.
Strohl K.
et al. Sleep-deprived motor vehicle operators are unfit to drive: a multidisciplinary expert consensus statement on drowsy driving. 19 Irwin M.R. Why sleep is important for health: a psychoneuroimmunology perspective. Our results carry significant implications for adaptation planning, policy, and research. Growing evidence faults increases in temperature with societal impacts to public health, behavior, and mental well being, although the causal mechanisms have remained poorly characterized.Insufficient sleep increases the risk of many of the same negative physiological, behavioral, social, and economic outcomes shown to increase with high temperatures.Thus, sleep may act as a key biobehavioral mechanism between ambient temperature and adverse human outcomes, with implications for human performance and productivity as well as physical and mental health.For instance, by elevating the probability of short sleep, high ambient temperatures may predispose susceptible segments of society to worsened affect,anger and aggression,hypertension and adverse cardiovascular outcomes,diminished cognitive performance,elevated risk of accidents and injuries,and compromised immune system functioning.While further research should seek to clarify this hypothesis, addressing the nocturnal impact of rising ambient temperatures on human sleep may be an efficient early intervention to reduce downstream adverse behavioral and developmental impacts linked to insufficient sleep. Through the use of consistently measured sleep records registered by sleep-tracking wristbands, our findings indicate that elevated temperatures drive sleep loss primarily by delaying when people fall asleep, providing a specific target for future adaptive interventions that seek to attenuate the impact of nighttime heat.
31 Salamanca F.
Georgescu M.
Mahalov A.
Moustaoui M.
Wang M. Anthropogenic heating of the urban environment due to air conditioning. , 58 Patz J.A.
Frumkin H.
Holloway T.
Vimont D.J.
Haines A. Climate change. , 75 Biardeau L.T.
Davis L.W.
Gertler P.
Wolfram C. Heat exposure and global air conditioning. , 76 Davis L.W.
Gertler P.J. Contribution of air conditioning adoption to future energy use under global warming. 77 Tuholske C.
Caylor K.
Funk C.
Verdin A.
Sweeney S.
Grace K.
Peterson P.
Evans T. Global urban population exposure to extreme heat. , 78 Wang J.
Chen Y.
Liao W.
He G.
Tett S.F.B.
Yan Z.
Zhai P.
Feng J.
Ma W.
Huang C.
Hu Y. Anthropogenic emissions and urbanization increase risk of compound hot extremes in cities. 59 Watts N.
Amann M.
Arnell N.
Ayeb-Karlsson S.
Belesova K.
Boykoff M.
Byass P.
Cai W.
Campbell-Lendrum D.
Capstick S.
et al. The 2019 report of the Lancet Countdown on health and climate change: ensuring that the health of a child born today is not defined by a changing climate. , 79 Hoffman J.S.
Shandas V.
Pendleton N. The effects of historical housing policies on resident exposure to intra-urban heat: a study of 108 US urban areas. , 80 Schell C.J.
Dyson K.
Fuentes T.L.
Des Roches S.
Harris N.C.
Miller D.S.
Woelfle-Erskine C.A.
Lambert M.R. The ecological and evolutionary consequences of systemic racism in urban environments. , 81 Nardone A.
Rudolph K.E.
Morello-Frosch R.
Casey J.A.
Casey Joan A. Redlines and greenspace: the relationship between historical redlining and 2010 greenspace across the United States. Interestingly, a corollary to our results is that ambient cooling interventions may be able to promote sleep gain ( Figure 2 A). Although access to air conditioning may partially buffer the effect of high ambient temperatures ( Figure 3 C), these same adaptive technologies can potentially exacerbate the unequal burdens of both global and local warming, through increased GHG emissions and ambient heat displacement.Moreover, continued urbanization is expected to further amplify ambient heat exposure.Heat-resilient planning, environmental design, and biopsychosocial interventions may be needed to equitably protect the world’s urban population centers and vulnerable communities from differential exposure to magnified nighttime temperatures.
76 Davis L.W.
Gertler P.J. Contribution of air conditioning adoption to future energy use under global warming. Several considerations should be taken into account when interpreting our results. First, global access and adoption of wearable devices is not geographically or demographically uniform. Our dataset contains more people who are middle-aged, male, and from high- and upper-middle-income countries ( Figure 1 B; Experimental procedures ). Given that nighttime temperature effects are larger for females, the elderly, and lower-middle-income countries in our sample, the magnitude of our primary effect estimates and projections is likely conservative. Sleep-tracking wristband ownership may also be associated with unobserved demographic factors, including higher socioeconomic status, physiological resilience, and access to cooling technologies, possibly reducing the accuracy of our estimates—especially in lower-middle-income countries.
82 Blanc E.
Schlenker W. The use of panel models in assessments of climate impacts on agriculture. 35 Yetish G.
Kaplan H.
Gurven M.
Wood B.
Pontzer H.
Manger P.R.
Wilson C.
McGregor R.
Siegel J.M. Natural sleep and its seasonal variations in three pre-industrial societies. , 39 van Loenhout J.A.F.
le Grand A.
Duijm F.
Greven F.
Vink N.M.
Hoek G.
Zuurbier M. The effect of high indoor temperatures on self-perceived health of elderly persons. , 83 Hausman J. Mismeasured variables in econometric analysis: problems from the right and problems from the left. 55 Graff Zivin J.
Neidell M. Temperature and the allocation of time: implications for climate change. , 84 Obradovich N.
Fowler J.H. Climate change may alter human physical activity patterns. , 85 Heaney A.K.
Carrión D.
Burkart K.
Lesk C.
Jack D. Climate change and physical activity: estimated impacts of ambient temperatures on bikeshare usage in New York city. , 86 Chan N.W.
Wichman C.J. Climate change and recreation: evidence from North American cycling. , 87 Bernard P.
Chevance G.
Kingsbury C.
Baillot A.
Romain A.J.
Molinier V.
Gadais T.
Dancause K.N. Climate change, physical activity and sport: a systematic review. Second, the deconvolution process used in our analyses likely mechanically biases our effect estimates toward zero.Nevertheless, we observe consistent ambient-temperature effects on sleep—even for people living in industrialized societies and high-income countries with plausible access to air conditioning ( Figure 3 C). Moreover, station-based measures of ambient temperature may differ from actual temperature exposures where people live, likely attenuating the magnitude of our empirical estimates of the relationship between temperature and sleep.As such, results from our ecological study reflect the total effect of ambient outdoor temperature on human sleep duration and timing, including all sleep-adjacent behavioral effects. For instance, temperature-altered physical activities—previously shown to be sensitive to ambient thermal conditions—may subsequently impact human sleep. Indeed, in an alternative model specification where we simultaneously include maximum and minimum temperature as explanatory variables, we find that daytime maximum temperatures may result in greater sleep loss ( Table S45 ). However, we interpret this exploratory result cautiously since the specification risks introducing multicollinearity due to serial correlation between daily minimum and maximum temperature values. Future multi-country studies with paired person-level physical activity outcomes are needed to assess whether altered physical activity may be implicated in the causal pathway linking temperature and sleep outcomes, including the delay in sleep onset that we identify.
37 Buguet A. Sleep under extreme environments: effects of heat and cold exposure, altitude, hyperbaric pressure and microgravity in space. 34 Harding E.C.
Franks N.P.
Wisden W. The temperature dependence of sleep. 12 Rifkin D.I.
Long M.W.
Perry M.J. Climate change and sleep: a systematic review of the literature and conceptual framework. 88 Min K b
Lee S.
Min J.Y. High and low ambient temperature at night and the prescription of hypnotics. Third, the current study primarily measures changes in sleep duration and timing, which does not convey how the observed decline in sleep duration impacts underlying sleep physiology. Controlled experiments with human subjects have shown that rapid eye movement (REM) and non-REM (NREM) sleep decrease when people are exposed to high environmental temperatures.Yet it remains unclear how ambient temperature modulates human sleep architecture and other neurobehavioral correlates of restorative sleep in real-world settings globally.If people respond differently outside of the sleep laboratory, for instance, via adaptive improvements to sleep quality, then the consequences of the sleep loss we identify may be partially offset. In an initial exploratory analysis (mirroring our analytical approach outlined in Equation 1 ), we investigate the influence of ambient temperature on sleep interruption—the probability that an individual wakes up one or more times during the nighttime sleep period. We do not find evidence that an increase in ambient temperature significantly reduces or otherwise alters registered nighttime awakening ( Table S51 ), suggesting that temperature-driven reductions in sleep quantity may not be compensated for with improvements in sleep quality. However, similar to wrist actigraphy, accelerometer-based activity-tracking wristbands may underestimate nighttime awakenings, suggesting that the sleep impact estimates in this study may be conservative. Future in situ research should further investigate whether sleep fragmentation is also sensitive to ambient weather conditions. Moreover, global research is needed to understand the impact of ambient temperature on sleep disordersand coping behaviors.
75 Biardeau L.T.
Davis L.W.
Gertler P.
Wolfram C. Heat exposure and global air conditioning. Fourth, although our sample includes data from 68 countries spanning all populated continents, it has sparse coverage for large parts of Africa, Central America, South America, and the Middle East—regions that already rank among the warmest in the world ( Figure 5 ). Climate projections indicate that many of the countries within these regions will be disproportionately exposed to some of the highest ambient temperatures and most cooling degree days, warranting future study.