The COVID-19 pandemic drastically altered human behavior globally, particularly in major cities. Lockdowns and stay-at-home advisories significantly reduced daytime populations in city centers, as observed in Tokyo's office districts where the population decreased by over 50% due to increased teleworking. While previous studies highlighted improvements in air quality and reductions in greenhouse gas emissions during lockdowns, the impact on local urban climates, specifically air temperature (T), anthropogenic heat emission (Q_H), and electricity consumption (EC), remained largely unexplored. The challenges in isolating the pandemic's signal from natural temperature variability in observational data necessitate numerical simulations using urban parameterizations. Existing models, such as single-layer urban canopy models (UCMs), handle Q_H but lack direct representation of human behavior. This study overcomes this limitation by integrating a coupled UCM with a building energy model (BEM) (UCM-BEM), allowing for the direct incorporation of real-time population and traffic data (social big data) to simulate the effects of behavioral changes on urban climate. This builds upon previous work, which estimated the impact of stay-at-home advisories on urban climate but lacked the detail and comprehensive spatial coverage of this study. Understanding the impact of behavioral changes on urban climate is crucial given the projected increase in urban populations and the growing concerns surrounding urban warming and the urban heat island effect. This study aims to evaluate the potential of altered human behavior as a cost-effective 'soft-type' climate change adaptation strategy in comparison to more traditional 'hard-type' approaches like greening and cool roofs.

Existing research demonstrates the impact of COVID-19 lockdowns on global and local environments. Studies show improvements in air quality (e.g., reductions in NO_x and black carbon) and decreases in greenhouse gas emissions (e.g., CO₂). However, research on the pandemic's effect on urban microclimates remains limited. While some studies have noted changes in land surface temperature (LST) and urban heat islands during lockdowns, these studies faced methodological challenges, such as distinguishing pandemic-induced changes from natural climate variability. The use of numerical models coupled with urban parameterizations offers a promising approach to overcome these limitations. Previous efforts have employed urban canopy models (UCMs) coupled with building energy models (BEMs) to assess the impact of anthropogenic heat on urban temperature. However, these studies often lack the integration of real-time human behavior data, hindering the accurate estimation of the pandemic's effects. This study directly addresses this gap by integrating high-resolution social big data with a sophisticated UCM-BEM model.

This study employed a novel approach that integrates real-time population and traffic data (social big data) with an advanced urban climate model. The model used was the Advanced Research WRF (ARW) version 3.7.1, coupled online with a building energy model (CM-BEM), known as WRF-CM-BEM. The model domain covered the Tokyo Metropolitan Area (TMA) with varying grid resolutions (25 km, 5 km, and 1 km). High-resolution Mobile Spatial Statistics (MSS) data from NTT Docomo, Inc., providing hourly population statistics with 500 m spatial resolution, were utilized to represent population density changes during the stay-at-home advisories (April 18th - May 14th, 2020) compared to the same period in 2019. Traffic data from the Japan Road Traffic Information Center were also incorporated to assess changes in traffic volume. The model was calibrated against existing data and simulations. For both the spring (April-May) and summer (July-August) periods, the model incorporated these changes in human behavior, adjusting parameters such as the number of building occupants, appliance electricity consumption (baseload EC), air conditioning (AC) operation schedules, and traffic-related anthropogenic heat (Q_F,TRA) based on the ratios of COVID-19 period data to pre-pandemic data (equations 4 and 5 in the paper). Simulations were run for both a 'No-COVID' scenario (using pre-pandemic data) and a 'COVID' scenario (incorporated the observed changes in human behavior). The key variables analyzed include changes in electricity consumption (ΔEC), anthropogenic heat flux (ΔQ_F), and near-surface air temperature (ΔT). Simple linear regression equations were developed (equations 1-3 in the paper) to relate population change (ΔP) to ΔEC, ΔQ_F, and ΔT, enabling the estimation of these impacts for other Japanese cities and potentially cities with similar characteristics.

The analysis revealed substantial changes in urban climate variables in Tokyo during the COVID-19 stay-at-home advisories. In the office district, daytime electricity consumption (EC) decreased by approximately 70%, and anthropogenic heat flux (Q_F) reduced by approximately 67% compared to the pre-pandemic period. This reduction in anthropogenic heat led to a decrease in daytime near-surface air temperature (ΔT) of approximately -0.21°C in the central Tokyo office area. The temperature decrease was correlated with population reduction in the office district. In contrast, residential areas showed increases in EC and in some cases Q_F due to increased daytime populations, but changes in temperature were not statistically significant. These relationships between population changes and climate variables (ΔEC, ΔQ_F, ΔT) were quantified using linear regression equations (Equations 1-3). Applying these equations to data from other major Japanese cities revealed similar patterns of decreased EC, Q_F, and T in city centers. The study further estimated a reduction in daily total CO₂ emissions in Tokyo's 23 wards by 8.2% during the spring period and 7.9% during the summer period, attributed to the pandemic's impact on human behavior. Summer simulations showed an even more substantial impact, with a temperature decrease in central Tokyo of -0.3°C, equating to 30% of the past GHG-induced warming in Tokyo, highlighting the potential of this effect in the context of climate change mitigation. The results from the study are consistent with temperature decreases estimated by other studies that employed observational data and different methodologies.

The findings of this study demonstrate the substantial impact of changes in human behavior on urban climate. The reduction in air temperature observed in Tokyo's city center during the COVID-19 pandemic is significant, particularly considering its consistency with other studies employing diverse methodologies. The developed linear regression equations provide a simplified approach to estimating these climate impacts across various cities, offering a valuable tool for urban planning and climate change mitigation strategies. The study's results strongly suggest that changes in human behavior, such as the widespread adoption of teleworking, can serve as a cost-effective 'soft' adaptation strategy to mitigate urban heat island effects and reduce the demand for air conditioning. The magnitude of temperature change observed during the pandemic is comparable to that achieved through traditional ‘hard’ mitigation methods like greening and cool roofs, but without the significant financial investment. Furthermore, the model accurately replicates observed weekday-weekend temperature differences, underscoring the reliability of its findings and the potential for broader applications.

This study successfully quantified the impact of COVID-19-induced changes in human behavior on urban climate in Tokyo, demonstrating a significant reduction in air temperature in the city center. The proposed methodology integrating social big data with an urban climate model offers a robust tool for future studies and urban planning. The results highlight the potential for shifts in human behavior, such as remote work, to serve as a cost-effective climate change adaptation strategy in urban environments. Future research should investigate the combined effects of behavioral changes and traditional mitigation approaches, as well as explore the long-term sustainability and societal implications of such behavioral shifts.

While the study provides a valuable contribution, some limitations should be noted. The study's focus is on Tokyo and its surrounding areas, which may limit the generalizability of findings to other cities with different characteristics. The analysis relies on data from a specific period during the pandemic, and the observed changes in human behavior might not be entirely representative of long-term trends. Additionally, the model's accuracy depends on the reliability and resolution of the social big data employed, and future research could benefit from exploring alternative sources or methods to improve the data's representation of human activities.