Sleep is crucial for health, mood, cognitive performance, work quality, and social life. Many factors influence sleep, including mood, personality, sleep quality, chronotype, and demographics. Sleep behavior is influenced by circadian processes, a 24-hour cycle relying on internal timekeeping cells that synchronize with environmental cues like light and temperature. Animal models show seasonal adjustments in the circadian pacemaker, influencing wake duration. However, the impact of seasons on sleep in humans, especially in environments with artificial light, remains unclear. Previous studies examining seasonal effects on sleep have produced mixed findings, likely due to methodological differences and limitations. These differences include study settings (laboratory vs. observational), datasets (wearable, self-report, cell phone data), populations studied (geographic location, age, chronotype), and temporal resolution. Many studies had small sample sizes and short observation periods, while larger studies often relied on self-report data, introducing potential biases. This study aimed to investigate the detailed effects of seasons and weather on sleep by addressing these methodological limitations. It used objective, continuous, and longitudinal measures of sleep from wearable devices in a large sample of individuals across the U.S., controlling for demographic and psychological factors and incorporating daily weather data.

Existing research on the relationship between seasons and sleep presents conflicting results. Laboratory studies, while providing controlled environments, often use small sample sizes and short observation periods, limiting generalizability. Studies using wearable sensors offer objective measures but frequently suffer from limited sample sizes and short observation durations. Self-report and large-scale data repository studies, although boasting substantial sample sizes, often lack detailed sleep timing data and are susceptible to recall bias. Studies examining pre-industrial societies have revealed longer sleep durations during winter, highlighting the potential influence of artificial light on seasonal sleep patterns. The inconsistencies across these studies suggest the need for a large-scale, long-term study using objective and continuous measurements to understand the impact of seasons and weather on sleep, controlling for other confounding factors.

This study involved 216 participants across the U.S. who wore Garmin VivoSmart 3 fitness bands for a year (February 5, 2018, to March 15, 2019), providing continuous sleep data. Data on sleep duration, bedtime, and wake time were collected via the Garmin Health API. Daily weather data, including 14 variables, were gathered from the World Weather Online API using participants' zip codes. Principal component analysis reduced the weather data to three principal components: temperature, wind, and humidity/cloud cover. Day length was calculated from sunrise/sunset times. Participants also completed a survey to collect demographic and psychological trait data (age, sex, organization, supervisory role, latitude, longitude, affect balance, personality traits (Big Five Inventory), sleep quality (PSQI), and chronotype (MEQ)). Data cleaning excluded travel days, the week following daylight saving time changes, and participants with less than 50% of weekday data in any season. Mixed linear-effects models were used to analyze the data, with random intercept effects for participants. Three nested models were compared: a baseline model, a model including demographic and psychological traits, a model adding seasonal variables, and a full model adding weather variables. Standardized beta coefficients, confidence intervals, pseudo R² values, and BIC were used to assess model fit and variable importance.

The study found modest but statistically significant seasonal and weather effects on sleep. For sleep duration, day length and the spring season were significant predictors. Longer day lengths were associated with shorter sleep durations (3.6 min decrease per extra hour of daylight). Spring was associated with a 12.6-min decrease in sleep duration compared to winter. For bedtime, day length, seasons (spring and summer), temperature, age, openness (personality trait), and chronotype were significant predictors. Longer day lengths resulted in earlier bedtimes (1.8 min earlier per hour), while spring and summer delayed bedtimes (4.2 min and 6 min respectively compared to winter). Higher temperature was associated with later bedtimes (0.6 min later per unit increase). For wake time, day length, seasons (fall, spring, and summer), chronotype, openness, and temperature were significant predictors. Longer day lengths resulted in earlier wake times (5.4 min earlier per hour). Fall and spring showed earlier wake times (1.8 min and 8.4 min earlier, respectively, compared to winter), while summer delayed wake times (7.8 min later compared to winter). The study revealed that differences in sleep duration might be more driven by differences in wake time rather than bedtime.

This study, using a large sample and objective, continuous measures, replicates and extends prior findings. The significant negative effects of spring and day length on sleep duration are consistent with some previous research. Findings on seasonal effects on bed and wake times also align with some previous studies. The modest effects of temperature on bed and wake times suggest that while temperature extremes may significantly impact sleep, even moderate temperature changes have a small, measurable influence. The study highlights the importance of considering both seasonal and daily-level variables, like day length and temperature, in future studies of sleep. The relatively large sample size enhances the generalizability of findings to an adult population of college-educated information workers.

This study demonstrates that seasons have a modest but statistically significant effect on sleep patterns, even in industrialized settings, and after accounting for demographic and psychological factors. The findings highlight the value of long-term, continuous, objective sleep measurement and the importance of considering day length and temperature as variables in future sleep research. Future research could explore seasonal effects on weekend sleep, examine sleep in other populations, and investigate interventions to mitigate negative seasonal sleep impacts.

The study sample may be biased towards individuals with better compliance due to data filtering, and generalizability is limited to similar populations (college-educated information workers). Exogenous factors (e.g., stressful life events) were not controlled for, and the focus was on weekday sleep only. The study used weather data rather than personally sensed environmental data, potentially missing finer-grained details on participants’ exposure. Finally, while daylight savings time changes were adjusted for, their influence on sleep may not have been fully accounted for.