On the Electrification of Winter Season in Cold Climate Megacities—The Case of New York City

The building sector is a prime contributor of energy consumption and the emission of greenhouse gases (GHG) and is therefore a good starting place to address the path of decarbonization. The northeastern market of New York City (NYC) recently passed a bill to decarbonize the building sector by 2050, starting in 2027 (i.e., Local Law 97) [1]. With a housing sector of one million buildings, mostly 1–4 family houses, accounting for 52.9 million metric tons of carbon dioxide emissions in 2014 alone, NYC stands as a model city to accelerate our sustainability efforts. Fossil fuels used in both space heating and cooling account for 60% of GHG emissions from buildings and 30% of citywide GHG emissions [2]. For NYC, approximately 6000 buildings have adopted cleaner forms of energy, as of 2017. It is further expected that the population of NYC will expand by 100,000 buildings and over 9 million occupants by 2050 from the current 8 million. Such rapid transformations will have immediate consequences in the urban environment and in the energy infrastructure that needs to be urgently researched.

Delving into the broader environmental implications, the role of mesoscale dynamics comes to the forefront. This domain, characterized by spatial scales extending horizontally from a few kilometers to several hundred kilometers, and vertically from a few tens of meters to the full depth of the troposphere, underpins key interactions in the urban environment [3]. At this scale, the interactions between diverse land use types, surface characteristics, complex terrain, atmospheric composition, water bodies, and the urban canopy are crucial. Of these factors, the urban canopy is particularly noteworthy, as it plays a significant role in shaping the environmental conditions within urban areas. Diverging from natural environments, urban regions exhibit unique surface–atmosphere interactions. These interactions are shaped by factors such as extensive building coverage, the high density of vertical surfaces of varying heights, and the use of materials with a high heat capacity. The urban heat island (UHI) effect, a distinct feature of urban environments, arises from the modification of land–air processes within built environments. This phenomenon is primarily attributed to factors such as the use of high heat capacity materials, the presence of densely clustered vertical structures with varying heights, limited green spaces, and the release of anthropogenic heat [4–6]. In summer seasons, urban regions can experience temperature increases of up to 10 °C [5,7]. During heatwave events, the UHI effect can further amplify these localized heat extremes, attributable to urban surface dryness, heat wave–associated low wind speeds, and enhanced heat storage and generation [8–15]. Amid these urban characteristics, an additional compounding factor emerges: the waste heat generated by air conditioning (AC) systems. While these systems provide necessary interior temperature control, they inadvertently increase outdoor temperatures, intensifying the UHI effect. This amplification cycle ultimately drives a greater demand for cooling energy, thus leading to an increase in waste heat emissions, further perpetuating the UHI phenomenon. During the winter seasons, the UHI effect can warm the urban environment by up to 3 °C [16,17], leading to a reduction in energy demands for heating [17]. Reference [18] highlights several physical processes underpinning this intensification, including (1) reduced snow cover in urban areas due to human-led removal activities, (2) the thermal heat storage in the urban structures that gets released amid cold waves, and (3) the emission of heat from buildings due to anthropogenic space heating. However, the winter UHI may also be associated with poorer air quality due to increased pollutant concentrations in the urban atmosphere. Furthermore, around 55% of the world’s population currently resides in urban areas. This is projected to increase to 68% by 2050 [19], thus raising the stakes for these deep decarbonization efforts.

Addressing the intricate interplay between weather, buildings, and energy demand necessitates an urbanized version of the Weather and Research Forecasting (WRF) model, a nonhydrostatic mesoscale numerical weather prediction system that serves as a powerful tool for both weather research and forecasting [20]. This setup couples the WRF model to a multilayer building environment parameterization (BEP) [21] and a building energy model (BEM) [22] that has been extensively used to study urban weather, climate, and energy demand. Hereafter, the WRF model coupled with BEP + BEM will be denoted as urban WRF or uWRF. Given the current computational limitations, it is not feasible to model every specific building within a meteorological grid cell; however, a BEM that is overly simplistic may fail to accurately capture critical interactions between the built environment and the atmosphere. To address these complexities, we utilize a BEM that explicitly resolves several crucial elements, including the diffusion of heat through walls, roofs, and floors; the heat generation associated with occupants and equipment; the exchange of radiation between indoor surfaces; and the specific energy consumption tied to the functioning of heating, ventilation, and air conditioning (HVAC) systems. The BEM is configured to characterize the general physical attributes of an ideal building, representative of those within the grid cell. This representation relies on the use of urban canopy parameters (UCPs) specific to the city being modeled.

One of the earliest evaluations of this model setup was carried out in the Phoenix metropolitan area. During this assessment, the modeled electrical demand from uWRF was compared with actual observed data on AC electrical demand recorded over a series of summer heat events. They demonstrated the model’s capability of replicating the observed diurnal profile of AC electrical demand [23]. Reference [18] examined the impact of urban air temperature dynamics and AC electrical demand for Osaka, Japan, a large Asian megacity, using one year’s worth of evaluation data. The model demonstrated its ability to replicate the diurnal and spatial variations of surface air temperature and AC electrical demand for multiple electric power substations within the city. They showed that BEP + BEM could serve as an effective tool to reproduce the year-round urban air temperature and electricity demand in a typical urban grid. For the case of NYC, Refs. [24,25] used a high-resolution UCP dataset to represent the city’s urban form and evaluated the model’s meteorological performance against surface observations. They found that including the thermal and drag effects of buildings improved model simulations over urban regions, providing better spatially-distributed estimates of surface air temperatures and 10-meter wind speeds when compared to point-source measurements. Reference [26] continued this work and reported on a methodology to forecast peak cooling loads at the city scale in NYC. Through a detailed evaluation against observed load values, they were able to account for a significant amount of load variability in the forecast outputs during extreme heat events, with a statistically significant correlation factor of 0.81 and a root-mean-square error (RMSE) of 669 MW. Furthermore, Ref. [27] explored the impacts of increasing residential AC adoption as an adaptive measure to extreme heat in NYC, reporting on the potential trade-offs in extreme heat adaptation strategies for cities that may be necessary in the context of increasing extreme heat events.

In contrast to the summer season, there have been limited studies conducted on the city-scale energy demands during the winter season. Xu et al. [17] carried out a comprehensive analysis in Beijing, China, using the BEP-BEM model in an offline mode, without the integration of the mesoscale model with urban physics. The results of the study revealed that the winter UHI effect could considerably affect building energy requirements in urban regions. Throughout the period of investigation, the average UHI intensity peaked at 3.41 K, and this phenomenon was shown to diminish the heating energy demand by roughly 20% in comparison to suburban areas. This work not only offered vital insights into the simulation of single-building energy dynamics but also demonstrated consistency between a coupled BEP + BEM model and a standalone building energy simulation under winter conditions, verifying the BEP + BEM’s ability to represent energy demands under winter conditions. Building on this localized study, Ref. [28] broadened the scope to the city level, evaluating the ability of uWRF to model building energy consumption in Bolzano, Italy, during the winter season. Leveraging observations for different building types, the study found that the uWRF setup could successfully reproduce the spatial distribution of energy consumption across the city. Moreover, the findings underscored the importance of factors such as building height and the building surface area to plan area ratio in determining energy consumption patterns. They also found that the model accurately captured the daily energy consumption patterns in residential areas but encountered difficulties in replicating demand peaks in commercial–industrial areas. The study thus highlighted the importance of precise representation of urban morphology in mesoscale meteorological models to obtain trustworthy estimates of citywide building energy consumption. Reference [16] broadened the geographic focus to include 12 U.S. cities, providing further insights into the role of UHIs during cold waves. The study demonstrated that UHIs not only enhanced urban warmth in winter but also became more pronounced during cold spells, with anthropogenic heating contributing over 30% to this increase. The study emphasized the potential of UHIs as buffers against extreme cold, reducing heating demands and mitigating cold-related hazards, a theme that resonated with the findings on Beijing’s winter UHI and its impact on heating demand. However, it also cautioned against the potential adverse effects of standard UHI mitigation efforts during cold periods, indicating a need for more nuanced approaches to urban heat mitigation policies.

In this current investigation, we delve into the broader impacts of transitioning the heating infrastructure to complete citywide building electrification as an adaptive measure for urban decarbonization, with NYC serving as a representative case study. Focusing on the scenario of a full shift from natural gas space heating to electrified heating through air-source heat pump (ASHP) systems, we build on existing research on the UHI effects, winter energy demands, and the complex interplay between weather, buildings, and energy consumption. The scarcity of winter uWRF meteorological evaluations, particularly within NYC, highlights a significant gap in the existing body of knowledge. Furthermore, city-scale heating energy analysis linked with boiler waste heat has not been previously conducted within this context, and a transition to 100% ASHP systems represents a novel area of research. The objective of this work is to deliver a comprehensive understanding of how electrification might reshape energy infrastructure and alter environmental phenomena at the regional scale. By zeroing in on NYC’s unique urban landscape, this project offers insights that can steer similar initiatives in other urban areas, promoting sustainable urban growth amid rising energy demands and environmental concerns.

The multilayer BEP parameterization is designed to represent the impact of urban morphology on local atmospheric conditions. This representation considers the effect of both horizontal surfaces (including roofs and streets) and vertical surfaces (such as walls) on key meteorological variables such as wind speed, temperature, and turbulent kinetic energy. These multifaceted interactions are visually illustrated in the left portion of Fig. 1. This results in significant effects on the thermodynamic structure of the lower part of the urban boundary layer. In parallel, the BEM calculates urban heat fluxes originating from the heat exchanges between buildings and their surroundings. These exchanges span heat transfer across walls, floors, roofs, solar radiation heat exchange through windows, and the impacts of HVAC systems [22]. The waste heat effects from the HVAC system are integrated into the mesoscale equations, resulting in modifications to the ambient outdoor conditions. These alterations set in motion a feedback mechanism between the ambient conditions and anthropogenic heat releases. It is important to note that the integration of winter anthropogenic heat outputs from the boiler system is not included in the default uWRF BEM code. Therefore, we have incorporated the necessary adjustments into the code, which will be discussed in the next paragraph. The BEM parametrization is demonstrated in the right part of Fig. 1. In general, this framework represents a simplified version of more intricate BEMs, such as EnergyPlus [29]. Traditional BEMs are highly detailed and complex, but they are not designed to simulate the total impact of multiple buildings [30] or to effectively couple with atmospheric models to generate dynamic feedback [31]. To overcome these limitations and represent aggregate building energy demand and external environmental interactions efficiently, this framework adopts a “building-averaged” approach. In this work, we use a high-resolution UCP dataset to represent the complex urban building morphological features in NYC, providing a comprehensive yet efficient simulation without the need to resolve each building individually. Its computational efficiency means coupling with an atmospheric model is feasible, thus reducing computational costs while maintaining relevance for urban studies [18].

Here, Q_B = ρC_pV_B (JK⁻¹ ) and V_B(m³) represent the overall room heat capacity and total room volume of the indoor air on a given floor, respectively. The model calculates the total sensible heat load H_in (W) and total latent heat load E_in(W) using heat transfer equations that span the building envelope. These calculations consider not only the sensible heat exchange through ventilation but also the internal heat generation from sources such as equipment and occupants. The methodology also incorporates the latent heat load through the changes in water vapor concentration induced by ventilation, including an evaporative component related to the occupants. Comprehensive details on this formulation can be referenced in Refs. [17,22,24,34,35], where each element of the heat transfer equations and internal loads is explained.

In contrast, the ASHP system does not lose heat due to inefficiencies like boilers; thus, it can be assumed zero or H_Loss,ASHP(sys) = 0. Moreover, to accurately account for the energy drawn from the surroundings by the ASHP system, the E_in term in Eq. (1) must be adjusted to subtract the energy consumed by the evaporators. However, this aspect will not be covered in the present study. Future research will investigate this component, which is anticipated to reveal an even more substantial cooling effect resulting from ASHP operation.

In this study, we employed WRF version 4.2.2 and configured the model using three two-way nested domains with horizontal resolutions of 9, 3, and 1 km, as depicted in Fig. 2(a). The simulation period spanned from January 1 to 31, 2021. To achieve optimal results and consider our computational resources, we set up the model to run in consecutive 4-day segments, with a 1-day spin-up period for each segment. Furthermore, this approach allowed us to frequently reinitialize the initial conditions based on reanalysis data, thereby ensuring a precise representation of the continuously evolving atmospheric conditions. Data from this spin-up day were excluded from each segment of the results. The remaining 3 days of each interval were retained and subsequently concatenated to create a comprehensive, month-long dataset. The model was set up with 51 vertical levels, with 15 levels situated within the lowest 300 m (each level being roughly 10–20 m) of the atmosphere to accurately resolve processes within the urban boundary layer. The highest resolution domain (1 km), D03 (Fig. 2(a)), contained the NYC Metropolitan area, which also included parts of New Jersey and Connecticut adjacent to the city. A time-step of 45 s was used for the outer domain, based on the recommendations from the WRF user guide (which suggests 6 × dx of the coarsest domain in km; we used 5 × dx) [20]. Model physics, adopted from previous work in NYC [26,27,37], included the Noah land surface model [32], the Rapid Radiative Transfer Model for Global Circulation Models (RRTMG) for longwave radiation [38], the RRTMG scheme for shortwave radiation [38], the Mellor–Yamada–Janjic planetary boundary layer scheme [39], and the Aerosol aware Thompson microphysics scheme [40]. Initial and boundary conditions were taken from the North American Regional Reanalysis (NARR) [41], which provided the atmospheric and surface data fields at 32 km resolution for the contiguous United States.

We utilized the Primary Land Use Tax Lot Output (PLUTO) dataset, provided by the NYC Department of City Planning through the NYC open data initiative, to generate the UCPs for the uWRF model. PLUTO, a comprehensive vector-format building dataset, contained more than 1 million polygon feature classes and information about the city’s buildings and urban land classes, which were essential for accurately representing the building morphological features in NYC. Within this dataset, buildings were categorized into three distinct classes: low-intensity residential, high-intensity residential, and commercial or industrial. The UCPs were then calculated based on attributes such as footprint area, number of floors, and use type. Various urban parameters, including building area fraction, building surface area-to-height ratio, land class, and height, were calculated from the PLUTO dataset on a building lot basis. These values were then aggregated to conform to the 1 km resolution of domain d03, as detailed by Ref. [25] and shown in Figs. 2(c)–2(e). The incorporation of PLUTO data into the uWRF model system ensured a more realistic representation of the city’s urban landscape. For the building physical parameters, we referenced Ref. [35], where the parameters are explicitly dependent on the building class and were derived following ASHRAE and U.S. Department of Energy standards. A summary of the main model options is shown in Table 1.

Two case studies were used to evaluate the environmental and infrastructural implications of increasing electrified winter heating adoption to 100% in NYC. Each of these cases used a fully resolved urbanized WRF setup with differences in the heating method, from boiler to electricity. The first case, referred to onward as the boiler case, simulated 100% natural gas space heating under the assumption that gas heating systems were used to maintain a setpoint temperature of 20 °C in every building in NYC and used an eff of 0.8. The second case, referred to onward as the ASHP case, assumed 100% adoption of ASHP systems for each building and used a COP of 3.0. The operation schedule for both cases encompasses the entire day (0000–2400). Both these cases are summarized in Table 2.

Finally, the gridded model outputs for the heating load and energy demand per unit area (W/m²) were multiplied by the actual grid spacing area (1 km²) to provide the heating load and building energy demand at each grid point. Subsequently, the results for all grid points across NYC were aggregated and summed to generate the total energy demand at the city scale. It is important to distinguish between the two energy demands considered: the boiler energy demand, which signifies the total amount of fuel energy required to meet the heating demand, and the ASHP energy demand, representing the total electrical energy demand required for the same purpose. Owing to its less efficient nature, the boiler energy demand will inherently be much higher when compared to the ASHP electrical energy demand. This approach allowed for a comprehensive understanding of the energy consumption patterns and spatial distribution throughout the entire city.

Some key limitations to this study stem from the assumptions made in our modeling approach. The use of a constant COP, without considering variations in the COP curve over time, overlooks the fact that ASHP heat pump efficiency depends on outdoor temperatures. This omission may lead to a less accurate representation of the heating system’s performance across varying climatic conditions, thus potentially limiting the applicability of our findings to broader contexts or extreme weather scenarios. Additionally, the exclusion of anthropogenic heat from traffic and industrial processes may result in an underprediction of the waste heat effect. Finally, the model’s constraint to a 1-month simulation period highlights the need for future studies to incorporate a more extended analysis, enhancing the insights into the long-term impacts of heating electrification in urban environments.

Model performance was evaluated against surface observations from the Automated Surface Observing System network (sites included: JFK and LGA) and the New York State Mesonet network (sites included: MANH, BRON, and STAT) [42]. Hourly outputs of temperature, wind speed, and wind direction were compared at five different stations (as shown in Fig. 2(b)) using different suitable performance metrics. These metrics included the RMSE, mean absolute error (MAE), and the correlation coefficient (R²) to measure the degree of error in comparison to the observations. Figure 2(e) also highlights the five different boroughs that make up the NYC region, including Manhattan, Bronx, Brooklyn, Staten Island, and Queens.

The uWRF model demonstrated robust performance in capturing diurnal temperature variations and extreme temperature events throughout January. As depicted in Fig. 3, the model accurately represented both maximum and minimum daily temperature variation across the month at each station. This ability to reproduce temperature fluctuations highlights the model’s reliability in simulating the thermal winter conditions in urban environments. Table 3 presents the evaluation metrics that further showcase the uWRF model’s strong performance in predicting temperature. The R² values represent the proportion of the variance in the observed data that can be predicted by the model and were high for temperature, ranging from 81% at the JFK site to 88% at the BRON site. This suggests that the model could account for a significant portion of the temperature variability observed at these stations. RMSE values for temperature were relatively low, with the highest value of 1.87 °C at JFK and the lowest value of 1.45 °C at BRON. These values indicate the typical deviation between the model’s predicted temperatures and the actual observed temperatures, with lower RMSE values implying a smaller overall error in temperature prediction. Additionally, the MAE values for temperature were relatively small, ranging from 1.15 °C at BRON to 1.5 °C at JFK and LGA. MAE represents the average absolute difference between the predicted and observed values, further supporting the model’s accuracy in simulating temperature during the winter month of January.

The uWRF model also demonstrated reasonable performance in simulating wind speed and direction. At LGA and JFK, the uWRF model tended to overpredict wind speeds during peak times, as shown in the time series Fig. 3 between January 17 and 24. However, the model accurately captured the transition to minimum wind speeds at these locations. This suggests that the model was proficient at simulating wind speed changes despite the slight overestimation during peak periods. Conversely, for the BRON and MANH sites, the model tended to underestimate peak wind speeds, highlighting a potential area for improvement, which can be attributed to the urban canopy parameters and further tuning. Despite this underestimation, the overall performance of the model in simulating wind speed patterns remained good. The STAT station exhibited the best performance among the evaluated locations, with the model accurately capturing both minimum and maximum wind speeds as illustrated in the time series Fig. 3. This strong performance at the STAT station further supports the model’s ability to simulate wind speed, even though improvements can still be made for certain locations and times. As shown in Table 3, the R² values for wind speed range from 50% at the BRON and MANH sites to 68% at the STAT site, indicating that the model can explain a moderate to substantial proportion of the observed wind speed variability. The RMSE values for wind speed were relatively low, with the highest value of 2.74 m/s at LGA and the lowest value of 1.16 m/s at STAT. The MAE values for wind speed were also relatively small, ranging from 0.88 m/s at STAT to 2.1 m/s at LGA, further supporting the model’s skill in simulating wind speed. For wind direction, the R² values ranged from 30% at the BRON site to 44% at the MANH and STAT sites, illustrating the ability to account for a moderate portion of the observed wind direction variability. Although there is potential for enhancement in wind direction predictions, these results still demonstrate the model’s ability to capture general wind patterns during the winter month. The RMSE values for wind direction varied between 46.74 deg at JFK to 57.42 deg at LGA, while the MAE values varied from 28.22 deg at JFK to 36.95 deg at BRON. These results show that the model has an acceptable level of accuracy in simulating wind direction, considering the complex urban environment and associated challenges in predicting wind patterns.

Overall, the results presented above compared favorably with recent research on summer heat wave events in NYC [13,25]. While previous studies have primarily focused on summer conditions, our new findings contribute valuable insights into the winter season, making them among the few studies to explore the performance of urban weather models during colder months.

The uWRF model simulations revealed the potential meteorological impacts of transitioning from boiler systems to heat pumps in NYC. These results showed that adopting heat pumps could lead to a significant reduction in the surface temperature across the domain, particularly during nighttime hours when heating demand was at its peak and traditional boiler systems release substantial amounts of heat into the atmosphere. Figure 4 showcases the average air temperature difference in blue, the domain variability shaded in gray, and the peak difference between the two scenarios in orange. The average difference was about 0.22 °C for the month of January 2021. This temperature difference can be attributed to the lower heat emissions from ASHP systems compared to gas heating systems, which ultimately led to a reduced UHI effect because of the sensible heat flux released into the environment. The spatial effects of this transition were further examined during specific time snapshots. As illustrated in Fig. 5, cases (a) and (b) represent the temperature difference during the early morning hours, where an average temperature change of approximately 1–1.5 °C was observed uniformly distributed across the city. Cases (c) and (d), on the other hand, depict instances when the peak temperature difference reached about 2.5–3 °C and could be observed primarily over Queens. Upon analyzing the monthly average subplot of Fig. 5(e), distinct temperature variations across different boroughs were noticed. Queens and Staten Island exhibited the most significant changes on average, while the Bronx displayed the lowest temperature reduction. This could be attributed to the Bronx’s inland position relative to the other boroughs.

Cross-sectional panels of potential temperature are shown in Fig. 6, with the planetary boundary layer height (PBLH) colored in pink and further used to explore the impacts on the built environment for each case. The heat pump case generally exhibited lower potential temperature values in the lower atmosphere, particularly in the urban core as demonstrated in the first 400 m of elevation. For this specific time of January 12, 2021, 8:00 a.m. EST, results show a difference of about 1 K (potential temperature) higher over the entire surface of the cross section when comparing the boiler case to the ASHP case, leading to a deeper boundary layer in the Bronx to central Queens. The decrease in potential temperature is expected to lead to a more stable atmospheric boundary layer, reducing the likelihood of turbulence and vertical mixing. When looking at the PBLH cross section and surface maps in Fig. 6(b), we observed a distinct spatial pattern in PBLH across NYC. In the heat pump case, PBLH values were generally lower, indicating a more stable boundary layer. The lower PBLH values in the heat pump case were particularly evident in densely populated areas, such as Manhattan and the Bronx, where the urban heat island effect was most pronounced. Conversely, PBLH values were relatively similar between the two cases in the less densely populated areas, such as Queens and Staten Island. Figure 6(b) also includes a spatial difference map between the two cases, further supporting the previous results. The decrease in PBLH associated with the adoption of heat pump systems could lead to a reduction in vertical mixing and dispersion of pollutants in the urban environment. These findings emphasize the complex spatial impacts of heating system transitions on local meteorology and urban environments, providing valuable insights for future research and policy development.

The implications of transitioning from natural gas boilers to heat pumps on energy demand were further explored, with the results showcased in Figs. 7 and 8. A critical aspect of this analysis was to determine the total energy demand that could emerge during such a transition. This was done by aggregating the energy consumption of all buildings in the domain, resulting in a time series plot for both the boiler and ASHP systems in Fig. 7. Additionally, the total NYC heat load was provided for reference as a dashed black line. The highest energy demand was observed on January 29, 2021, at 6:00 a.m. EST, with the boiler system requiring approximately 42,700 MW of energy (equivalent electricity from gas), in stark contrast to the ASHP system, which had a peak demand of around 11,400 MW. Furthermore, Fig. 7 provides an average diurnal profile for each heating system. On average, the peak energy demand occurred at 7 a.m. EST, with the ASHP system reaching nearly 5800 MW, compared to the boiler system’s peak demand of approximately 21,500 MW. The lowest energy demand occurred at 2:00 p.m. EST, with the ASHP system dropping to nearly 1200 MW, compared to the boiler system’s minimum of about 4400 MW. These findings highlight the potential for ASHP systems to substantially reduce energy consumption while maintaining the required heating capacity.

Three cases were selected during peak demand for the ASHP system and are shown in Fig. 8, which includes an aggregated monthly average subplot in Fig. 8(d). As expected, the highest density of energy consumption occurred in lower/mid-Manhattan, with a peak energy demand of about 40 W/m² (case (d) on January 29 at 6:00 a.m. EST). On average, this same area experienced a demand of about 15 W/m². The spatial variability of the demand remained very similar across each case but did differ in magnitude as each case progressed and increased. The city currently consumes far less electricity during winter seasons, with peak energy demand reaching close to 8 GW in the summertime, primarily due to air conditioning, as reported in Ref. [26]. The large-scale adoption of ASHP could result in a shift of NYC’s peak power demand from summer to winter, with the highest grid congestion likely to occur in early January mornings rather than July afternoons, as demonstrated in Fig. 7. This projected demand will have significant implications on the delivery of electricity, particularly in the context of increasing expectations for clean energy sources. Addressing this increased demand will necessitate infrastructure upgrades and the integration of renewable energy sources to minimize the environmental impact of the heightened electricity consumption. Moreover, the transition to ASHP may also require improvements in building insulation and energy efficiency, as the performance of heat pumps can be affected by the building envelope. These enhancements could lead to additional energy savings and further reduction of greenhouse gas emissions, contributing to a more sustainable urban environment.

This work explored the effects of adopting electrified space heating systems in a densely populated city like NYC, driven by the need for accelerated policies to achieve deep decarbonization in urban environments. Buildings in such cities represent the largest source of greenhouse gas emissions, as evidenced by NYC, where buildings account for over 70% of all greenhouse gas emissions. By examining the implications of transitioning from conventional heating methods to electric-based systems, we provide valuable insights into the potential benefits and challenges associated with this transformation, offering a better understanding of how cities can effectively tackle their emissions while addressing the growing demand for sustainable energy solutions. Buildings play a significant role in the UHI effect in these cities, with anthropogenic emissions being key contributors. As such, transitions in heating systems may have profound implications on both the environment and energy infrastructure. Traditionally, winter space heating in cold climates relies on natural gas or oil-based heating systems, which directly contribute to urban heat, GHG, and other pollutant emissions resulting from the combustion process. Transitioning to electrified systems could lead to a reduction in GHG and other pollutant emissions while increasing electricity demand, which could potentially be met by clean energy sources such as solar or wind power.

Thus, there is a need to develop and adopt techniques and methods that will enable a comprehensive investigation of the implications of these deep decarbonization adoptions.

In this study, we used a mesoscale WRF model coupled with the BEP + BEM parametrization to explore the complex interactions between urban surfaces, local meteorological conditions, and building energy consumption. Two different scenarios were used to evaluate the environmental and infrastructural implications of increasing electrified winter heating adoption to 100% in NYC. The first case simulated 100% natural gas space heating under the assumption that gas heating systems were used to maintain a setpoint temperature of 20 °C in every building in NYC and used an eff of 0.8. The second case assumed 100% adoption of ASHP systems for each building and used a COP of 3.0. The uWRF model displayed a strong performance in simulating temperature, wind speed, and wind direction during January in NYC. The model accurately captured diurnal temperature variations, extreme temperature events, and the overall thermal winter conditions in urban environments. The uWRF model exhibited a strong thermal performance, with average R² values of 85%, average RMSE of 1.69 °C, and average MAE of 1.37 °C, demonstrating its accuracy and reliability in capturing the diurnal temperature variations during the reporting period.

The uWRF model simulations provided valuable insights into the potential meteorological and energy infrastructure impacts of transitioning from boiler systems to ASHP in NYC. Results showed that adopting to ASHP could lead to an average surface temperature reduction of 0.22 °C during January, with some areas experiencing reductions of up to 2.5–3 °C. These temperature reductions contribute to a more stable atmospheric boundary layer, particularly during peak heating demand periods, and help reduce the urban heat island effect. Regarding energy consumption, the study revealed that ASHP has the potential to substantially reduce energy demand compared to traditional boiler systems. The peak energy demand for the ASHP system reached nearly 5800 MW, while the boiler system’s peak demand was approximately 21,500 MW. Moreover, the lowest energy demand for the ASHP system was around 1200 MW, compared to the boiler system’s minimum of about 4400 MW. However, the large-scale adoption of heat pumps may shift NYC’s peak power demand from summer to winter, with the highest grid congestion likely to occur in early January mornings. This projected demand will have significant implications for the delivery of electricity, particularly considering the increasing expectations for clean energy sources. Addressing this increased demand will necessitate infrastructure upgrades, the integration of renewable energy sources, and improvements in building insulation and energy efficiency.

This initial study serves as a foundation for expanding research on decarbonization processes in large, densely populated cities worldwide. Future investigations could encompass the partial adoption of electrified ASHP systems, the capacity of the energy infrastructure to handle such transitions, and the implementation of other services like hot water, cooking, and transportation through electrified energy services. Notably, we aim to explore these scenarios in upcoming studies, focusing on the potential impacts on air quality due to the reduced use of combustion processes. We encourage fellow researchers to engage in this vital area of research, as understanding the implications of widespread electrification and decarbonization is crucial for promoting sustainable urban development and reducing greenhouse gas emissions on a global scale.

This study is supported by The National Oceanic and Atmospheric Administration (NOAA) – Cooperative Science Center (CSC) for Earth System Sciences and Remote Sensing Technologies (CESSRST-II) under the Cooperative Agreement Award #NA22SEC4810016. The authors would like to thank the NOAA Office of Education, The Educational Partnership Program with Minority Serving Institutions (NOAA-EPP/MSI) and the NOAA-CESSRST-II for full fellowship support for Harold Gamarro. The statements, findings, conclusions and recommendations are those of the author(s) and do not necessarily reflect the views of NOAA. The research was also possible due to funding from the U.S. National Science Foundation (Grant No. EEC 2113874). The research was also possible due to funding from Brookhaven National Laboratory – Laboratory Directed Research and Development Grant No. 23-045. This research was made possible by the New York State (NYS) Mesonet. Original funding for the NYS Mesonet was provided by Federal Emergency Management Agency grant FEMA-4085-DR-NY, with the continued support of the NYS Division of Homeland Security & Emergency Services; the State of New York; the Research Foundation for the State University of New York (SUNY); the University at Albany; the Atmospheric Sciences Research Center (ASRC) at the University at Albany; and the Department of Atmospheric and Environmental Sciences (DAES) at the University at Albany.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

l =

latent heat of evaporation (J/kg)

q_Vr =

specific humidity of the indoor air (kg/kg)

C_p =

specific heat of air (J/kg/K)

H_Loss,Boiler =

heat loss from buildings in winter (W)

Q_AH =

waste heat released by heating system (W)

Q_B =

overall room heat capacity (J/K)

T_r =

indoor room temperature (K)

V_B =

overall room volume (m³)

eff/COP =

coefficient of performance/efficiency for heating system

E_H,ASHP/E_H,Boiler =

heating energy demand (W)

H_in/E_in =

sensible/latent heat load in one room (W)

H_out/E_out =

sensible/latent heat added by the heating system (W)

ρ =

air density (kg/m³)

AC =

air conditioning

ASHP =

air-source heat pump

BEM =

building energy model

BEP =

building environment parameterization

COP =

coefficient of performance

GHG =

greenhouse gases

HVAC =

heating, ventilation, and air conditioning

MAE =

mean absolute error

NARR =

North American Regional Reanalysis

NYC =

New York City

PBLH =

planetary boundary layer height

   PLUTO =

Primary Land Use Tax Lot Output

RMSE =

root mean square error

R² =

correlation coefficient

UCP =

urban canopy parameters

UHI =

urban heat island

WRF =

Weather Research and Forecasting