Anthropogenic heating of the urban environment due to air conditioning
Abstract
This article investigates the effect of air conditioning (AC) systems on air temperature and examines their electricity consumption for a semiarid urban environment. We simulate a 10 day extreme heat period over the Phoenix metropolitan area (U.S.) with the Weather Research and Forecasting model coupled to a multilayer building energy scheme. The performance of the modeling system is evaluated against 10 Arizona Meteorological Network weather stations and one weather station maintained by the National Weather Service for air temperature, wind speed, and wind direction. We show that explicit representation of waste heat from air conditioning systems improved the 2 m air temperature correspondence to observations. Waste heat release from AC systems was maximum during the day, but the mean effect was negligible near the surface. However, during the night, heat emitted from AC systems increased the mean 2 m air temperature by more than 1°C for some urban locations. The AC systems modified the thermal stratification of the urban boundary layer, promoting vertical mixing during nighttime hours. The anthropogenic processes examined here (i.e., explicit representation of urban energy consumption processes due to AC systems) require incorporation in future meteorological and climate investigations to improve weather and climate predictability. Our results demonstrate that releasing waste heat into the ambient environment exacerbates the nocturnal urban heat island and increases cooling demands.
1 Introduction
Cooling demand due to air conditioning (AC) systems can consume more than 50% of the total electricity demand during extreme heat events in semiarid urban environments, with maximum consumption up to 65% of total electricity demand during peak late afternoon hours [Salamanca et al., 2013]. Cooling demands for rapidly expanding urban areas are likely to increase considerably during the next several decades due to local warming induced by expanding buildup environments coupled with greenhouse gas-induced global climate change [Georgescu et al., 2013; Argueso et al., 2013]. Examination of current AC electricity demand and assessment of the additional heating contribution from this anthropogenic forcing to the ambient environment is therefore necessary, particularly in warm urban environments, to address future energy needs for a rapidly urbanizing planet in a sustainable manner.
Recent investigations have demonstrated that summertime U.S. energy demand will increase under conditions of future climate change due to the growing use of AC systems and increased ambient temperatures [Sailor, 2001; Sailor and Pavlova, 2003]. These investigations are based on linear empirical adjustments that relate current climate parameters (e.g., cooling degree days, heating degree days, and wind speed) to energy consumption and assume that only climatic factors (i.e., climatic variables) can change when energy demands are projected. As a result, predicted energy consumption is likely to be incorrect. Empirical adjustments describe the mathematical dependence of the energy consumption in terms of independent climatic variables. These experimental models are based on present urban scenarios and consequently are more appropriate for short- or medium-term forecasting of energy consumption. An advantage of a physics-based modeling system (relative to previous empirically based models) is that urban scenarios and climatic factors are taken into account when the energy consumption is calculated. For example, the electricity consumption of the AC systems (and their contribution to the outdoor environment) depends on the urban morphology that characterizes the urban area. The same volume of indoor air that is cooled can be distributed in multiple ways (e.g., taller buildings with a relatively smaller geographical footprint versus smaller buildings spread over larger expanses of land), each one potentially leading to different AC electricity consumption. Therefore, projected urban scenarios cannot be analyzed using empirical adjustments based on current morphological characteristics, and a physics-based modeling system should be considered instead. In other words, the use of a physically based dynamic approach, as utilized here, is necessary given the inherent feedbacks associated with the system of interest. For example, a warmer environment will lead to increased AC demand, which in turn will output additional waste heat into the environment, leading to further increase in AC demand, resulting in a positive feedback loop whose characterization cannot be represented via the use of present empirical approaches for future projections.
The present work uses a physics-based numerical model to evaluate the impact of sensible heat emission from AC systems on air temperature and to investigate their electricity consumption during a summertime extreme heat period. Summertime extreme heat days are projected to become more frequent and intense as a result of climate change [Miller et al., 2007; Tebaldi et al., 2006]. Reliable models are needed for projecting energy demands that can help to inform and assist power utility companies concerning upcoming cooling requirements and vulnerabilities to climate change and extreme weather [U.S. Department of Energy, 2013].
General AC systems absorb heat (cooling the indoor air) from the interior of the buildings and release heat into the surrounding outdoor environment. We show that release of waste heat raises the outdoor temperature and as a result increases the electricity consumption needed for cooling. The analysis of this feedback requires a two-way coupling between a building energy model (BEM) and an atmospheric model. Detailed BEMs have been developed, emerging from the engineering community (e.g., U.S. Department of Energy EnergyPlus; http://apps1.eere.energy.gov/buildings/energyplus), but without dynamic interaction with the outdoor environment. Although these sophisticated models represent the state-of-the-art in building energy analysis, they require a large number of input parameters to describe a particular building, making them unsuitable for urban climate research. For these reasons, more simple BEMs were developed by the meteorological community to model the two-way interactions between urban climate and building energy use. The number of parameters required to describe a building were reduced significantly. Thanks to this simplification, the interactions between a city and the overlying atmosphere could be modeled.
The atmospheric and building energy models are dynamically coupled. The atmospheric model supplies the BEM with the outdoor air temperature, outdoor air humidity, and the boundary conditions for the temperature calculation for building walls and roofs. The BEM then provides the atmospheric model with the heat fluxes associated with the consumption of energy within the buildings (e.g., AC systems). This two-way interaction is carried out through the urban canopy parameterization (UCP) [Masson, 2000; Kusaka et al., 2001; Martilli et al., 2002]. Detailed UCPs account for radiation trapping and shadowing effects that occur in the urban canyon due to increased obstruction of open sky from buildings. Horizontal and vertical surfaces (roofs, roads, and building walls) are treated separately allowing a better representation of the urban domain. However, the generation of heat inside the buildings and the exchanges with the exterior were not explicitly resolved. Alternatively, a source term of heat estimated from energy consumption databases was added to the sensible heat flux from the urban surfaces [Kusaka et al., 2001], but without considering the spatial variability of the energy demand.
The next stage in the evolution of urban meteorological modeling observed the integration of BEMs into UCPs to analyze the effect of AC systems on air temperature [Kikegawa et al., 2003; Kondo and Kikegawa, 2003]. Kikegawa et al. [2003] simulated (for a district of Tokyo, Japan) a reduction of 1.3°C for the daily mean air temperature when the waste heat originated by the AC systems was removed from the atmosphere. The simulated temperature reduction produced an electricity consumption saving close to 5.8%. More recently, Salamanca et al. [2010] developed a new BEM and computed an electricity consumption saving of 3.6% (for the commercial areas of Madrid, Spain), when the waste heat from the AC systems was eliminated from the outdoor environment [Salamanca et al., 2012].
In this paper we characterize diurnal variation of anthropogenic heating due to AC systems. The effect of AC systems is not constant during the day. High outdoor temperatures increase waste heat release and electricity consumption in regions where their use is prevalent. However, it is the planetary boundary layer height and not the outdoor temperature that determines the impact of the AC systems on the region's climatology. de Munck et al. [2013] simulated an increase of up to 1°C during the night (for central Paris, France), when dry AC systems were considered. They found that the effect of the AC systems was more important during the night due to the limited vertical extent (a few hundred meters) of the urban boundary layer. During the day, the urban boundary layer height can reach several kilometers, and the effect of AC systems can be small near the surface. Salamanca et al. [2012] reported the same occurrence for the city of Madrid (Spain); the effect was stronger at the end of the day and during the night. Li et al. [2013] reported the same phenomenon for the city of Singapore; the anthropogenic heat played an important role in the region's climatology from late afternoon to early morning when solar radiation did not dominate the surface energy balance.
The present paper evaluates the anthropogenic contribution of AC systems on air temperature and examines their electricity consumption for the rapidly expanding Phoenix metropolitan area, one of the largest metropolitan areas in the U.S., and the largest in the Colorado River Basin. Phoenix is located within the semiarid Sonoran desert and because of its harsh summertime conditions makes considerable use of AC systems. Due to the prevalence of meteorological facilities [Chow et al., 2012], the metropolitan area is an ideal test bed for urban meteorological research. The effect of AC systems has been analyzed with a BEM integrated in a multilayer building effect parameterization (BEP) [Salamanca et al., 2010; Martilli et al., 2002] coupled to the Weather Research and Forecasting (WRF) model. In previous work, Salamanca et al. [2013] validated the performance of BEP + BEM coupled to the WRF model and compared diurnal profiles of modeled versus observed AC electricity consumption (supplied by an electric utility company) for the Phoenix metropolitan area. The excellent agreement obtained in the aforementioned work provides confidence in the model's ability to assess the effect of AC systems on regional air temperature and to investigate their spatially explicit electricity consumption. In this paper, we first assess the capability of the multilayer building energy parameterization to reproduce the evolution of the diurnal cycle of near-surface climatology. We then analyze the effect of AC systems on air temperature and examine their electricity consumption.
2 Methodology
To evaluate the ability of BEP + BEM in conjunction with the Noah land surface model [Chen and Dudhia, 2001; Ek et al., 2003] to reproduce near-surface climatology, a 10 day clear sky period was simulated. The event covered 10 extreme heat days (EHD) from 10 to 19 July (2009), and we assumed that the use of AC systems encompassed the entire metropolitan area. Following previous criteria to identify extreme heat events [Grossman-Clarke et al., 2010], the last 4 days (from 16 to 19 July) can be considered an extreme heat event. In fact, 16 to 20 July 20 (2009) is the most recent extreme heat event in the region.
Four simulations were carried out to analyze the contribution of AC systems on air temperature. In the first simulation, the total floor area was considered air conditioned and the total waste heat from the AC systems was rejected into the atmosphere (hereafter denoted as AC100% simulation). For the second and third cases, only 65% and 35% (henceforth denoted as AC65% and AC35% simulations, respectively) of the total waste heat were rejected into the atmosphere, and finally, for the fourth case (henceforward denoted as AC_NoOut simulation) anthropogenic heat from AC systems was assumed not to be released outside into the atmosphere (although the AC systems were still working, maintaining a cooled indoor temperature). The ability to reproduce the near-surface climatology was evaluated by comparing WRF-simulated air temperature, wind speed, and wind direction against observations available at hourly frequency (see section 3.1).
2.1 Site Description
Phoenix metropolitan area is a rapidly urbanizing region in the Sonoran desert of the southwestern United States. Phoenix is located at 33.45°N, 112.07°W within the northern portions of the Sonoran desert, in the center of the dry Salt River Valley in Arizona. It has an arid climate with extremely hot summers and mild winters. July is the warmest month of the year with a mean maximum temperature of 41.4°C, and a mean minimum exceeding 27°C [Middel et al., 2012]. During the previous decade the population grew by 29% to 4.1 million people [U.S. Census Bureau, 2010] making it one of the largest metropolitan areas in the United States.
2.2 Numerical Experiments
We use the nonhydrostatic (V3.4.1) version of the Weather Research and Forecasting (WRF) model [Skamarock et al., 2008] coupled to the Noah land surface model [Chen and Dudhia, 2001; Ek et al., 2003] to evaluate near-surface climatology. The Noah land surface model has one canopy layer and the following prognostic variables: soil moisture and temperature in the soil layers, water stored in the vegetation canopy, and snow stored on the ground. The multilayer building energy scheme BEP + BEM was applied to the fraction of grid cells with built cover, and the Noah land surface model to the fraction of natural surfaces. BEM is a building energy model [Salamanca et al., 2010] integrated in the multilayer building effect parameterization (BEP) [Martilli et al., 2002] that takes into account the exchanges of energy between the buildings and the surrounding atmosphere as well as the impact of AC systems. In BEM, a building is treated as a pile of boxes, each box representing a particular floor. Buildings of several floors can be considered, and the time evolution of indoor air temperature (and humidity) is estimated separately for each floor. Natural ventilation, shortwave radiation penetrating through the windows, and heat generated by occupants and equipment are considered in the model. When the indoor temperature reaches a fixed target value, all the extra sensible heat (to maintain the indoor temperature constant) is extracted by the AC systems as well as the corresponding heat flux associated with their electricity consumption.
WRF simulations were conducted with initial and boundary conditions obtained from the National Centers for Environmental Prediction Final Analyses data (number ds083.2) with a spatial resolution of 1° × 1° and a temporal resolution of 6 h. The horizontal domain was composed of four two-way nested domains (Figure 1a) with 120 × 100, 163 × 157, 151 × 151, and 241 × 211 grid points, and a grid spacing of 27, 9, 3, and 1 km, respectively. The vertical dimension included 40 levels, with 14 within the lowest 1.5 km to better resolve urban planetary boundary layer processes. The planetary boundary layer was parameterized with the one-and-a-half-order closure [Bougeault and Lacarrere, 1989] turbulent scheme. The selected radiation parameterizations were the Dudhia shortwave radiation scheme [Dudhia, 1989] and the Rapid Radiative Transfer Model longwave parameterization [Mlawer et al., 1997]. The microphysics package was WSM3 [Hong et al., 2004], and no cumulus cloud scheme was considered in the two inner domains.
The U.S. Geological Survey 30 m 2006 National Land Cover Data set [Fry et al., 2011] was used to represent modern-day land use-land cover within the Noah land surface model for the urban domain. Based on the packing density of buildings, three different urban classes were defined (Figure 1b). They describe the morphology of the city: commercial or industrial (COI), high-intensity residential (HIR), and low-intensity residential (LIR). Building parameters (Table 1) were extracted from Burian et al. [2002], who summarize morphological characteristics for an area centered on the downtown of Phoenix. For the urban core, an internal layer of 0.06 m of insulating material (thermal conductivity of 0.09 W m−1 K−1 and heat capacity of 0.382 × 106 J m−3 K−1) was considered for roofs and vertical walls to describe the city. The thermal properties for the different urban surfaces (roofs, roads, and building walls) are detailed in Table 2 and correspond to standard building materials [Clarke et al., 1991]. For the AC model, the target internal temperature was set to 25°C [Kikegawa et al., 2003; Ohashi et al., 2007; Salamanca et al., 2011, 2012; Bueno et al., 2012; Lemonsu et al., 2013] and the coefficient of performance to 4.5 [Salamanca et al., 2013] for the four simulations. We assume that all anthropogenic heat emission from AC systems is rejected as sensible heat. This assumption is fully justified considering that AC systems that reject latent heat into the outside environment are normally installed in large office buildings (The ratio of sensible to latent heat emission from AC systems is commonly 3:2 in commercial areas [Ohashi et al., 2007].) in commercial areas, and these areas covered just the 1% of the urban domain.
| Parameter | COI (Commercial or Industrial) | HIR (High-Intensity Residential) | LIR (Low-Intensity Residential) |
|---|---|---|---|
| Urban fractionaa Georgescu et al. [2011]. |
0.95 | 0.85 | 0.70 |
| Building plan area fraction | 0.333 | 0.250 | 0.286 |
| Percent of buildings of 5 m of height | 50 | 80 | 90 |
| Percent of buildings of 10 m of height | 30 | 20 | 10 |
| Percent of buildings of 20 m of height | 8 | 0 | 0 |
| Percent of buildings of 30 m of height | 12 | 0 | 0 |
- a Georgescu et al. [2011].
| Surface | Thermal Conductivity of the Material (W m−1 K−1) | Specific Heat of the Material (× 106 J m−3 K−1) | Emissivity of the Surface | Albedo of the Surface | Thickness (m) |
|---|---|---|---|---|---|
| Roof | 0.67 | 1.32 | 0.90 | 0.20 | 0.1575 |
| Wall | 0.67 | 1.32 | 0.90 | 0.20 | 0.1575 |
| Road | 0.74 | 1.40 | 0.95 | 0.125 | 0.5375 |
2.3 Data for Model Evaluation
To evaluate the ability of WRF to reproduce the near-surface climatology, 11 weather stations available at hourly frequency (Figure 1b) were utilized. One weather station is maintained by the National Weather Service (NWS), and the Arizona Meteorological Network (AZMET) maintains the rest. Three stations were classified as urban (Waddell, Mesa, and Sky Harbor Airport) and eight as rural. Waddell was classified as LIR and Mesa and Sky Harbor Airport as HIR. The two urban AZMET weather stations (Phoenix Encanto and Phoenix Greenway) were not considered because they are located on golf courses and are not a true representation of the urban environment that is analyzed here. Although both stations were classified as urban by the WRF model at 1 km resolution, their climatology is governed by the impact of the city and the microclimate characteristic of heavily irrigated green areas. WRF-simulated hourly output frequency allowed direct comparison against observations for 2 m air temperature (T2 m), 10 m wind speed (WS10 m), and 10 m wind direction (WD10 m).
3 Results and Discussion
3.1 Evaluation of Diurnal Cycle of Near-Surface Climatology
The four WRF experiments produced notable statistical correspondence to observations for T2 m, WS10 m, and WD10 m at the urban and rural stations, and as a starting point only the AC100% simulated results are presented here. The daily evolution of near-surface temperature, including maximum and minimum temperatures, was satisfactorily simulated for the 10 day EHD period at both urban and rural stations (Table 3); although a warm bias was apparent in the second half of the period (Figure 2). WRF-modeled root-mean-square error was below 1.74°C for both urban and rural locales. WRF reproduced the observed variability of near-surface wind speed (Figure 3), although there was an almost systematic overestimation in rural areas, a bias that has been noted in previous work [Grossman-Clarke et al., 2010]. WRF-simulated root-mean-square error remained below 1.67 m s−1, with similar performance for both urban and rural stations (Table 3). Typical clear sky days produce predominantly daytime westerly winds, and nocturnal easterly flow from the higher northeast-east terrain, downslope to Phoenix [Brazel et al., 2005]. WRF was able to capture the diurnal cycle of this topographically induced complex flow in the region for both urban and rural stations (Figure 4), although correspondence to observations was better for rural areas. The results presented here demonstrate that WRF was able to reproduce satisfactorily the near-surface climatology for the 10 day extreme heat period. Other recent WRF urban studies reported similar agreement for other events [e.g., Grossman-Clarke et al., 2010; Kusaka et al., 2012; Grawe et al., 2012; Kim et al., 2013; Giovannini et al., 2013].
, Mean Absolute Error 
, and Mean Absolute Percentage Error 
for the Near-Surface Variables T2 m(°C), WS10 m(m s−1), and WD10 m(°) at the Urban and Rural Stations for the AC100% Experiment| Rural | Urban | Rural | Urban | Rural | Urban | |
|---|---|---|---|---|---|---|
| 10 day EHD Period | T2 m (°C) | T2 m (°C) | WS10 m (m s−1) | WS10 m (m s−1) | WD10 m (deg) | WD10 m (deg) |
| RMSE | 1.700 | 1.736 | 1.668 | 1.641 | 47.841 | 84.108 |
| MAE | 1.305 | 1.426 | 1.355 | 1.275 | 38.252 | 62.911 |
| MAPE | 0.038 | 0.039 | 0.846 | 1.014 | 0.214 | 0.536 |
3.2 Evaluation of AC Contribution on Air Temperature
To evaluate the contribution of AC systems on air temperature, the WRF AC100%, AC65%, and AC35% simulations were compared against the AC_NoOut experiment. Figure 5 shows the mean 2 m air temperature differences (T2 m(AC100%) − T2 m(AC_NoOut)) averaged over the entire 10 day EHD July 2009 period. During the night (from 8 P.M to 5 A.M.), the effect of the AC systems was significant, increasing the mean 2 m air temperature between 1°C and 1.5°C for most of the urban area. The impact was not homogeneous with reduced warming (0.25–0.5°C) near the limits of the metropolitan area. During daylight hours (from 6 A.M. to 7 P.M.), the effect of AC systems was less important, with near-surface temperature differences ranging between 0.25°C and 0.5°C (Figure 5b).
BEP + BEM was able to reproduce accurately the diurnal profile of AC electricity consumption when the ratio of air-conditioned floor area to total floor area was set to 0.65 for the city of Phoenix [Salamanca et al., 2013]. Figure 6 shows the contribution of the AC systems when this assumption was used (or equivalently when 65% of the total waste heat was ejected into the atmosphere). During the night, heat rejected from AC systems increased the mean 2 m air temperature between 0.5°C and 1°C for most of the metropolitan area. However, during the day, the effect of AC systems was negligible near the surface. Finally, Figure 7 shows the mean 2 m air temperature differences when the ratio of air conditioned to total floor area was set to 0.35. During nighttime hours, the areas with an increase between 0.25°C and 0.5°C dominated the region, although some neighborhoods still showed differences from 0.5°C to 0.75°C. During the day, the effect of the AC systems was again insignificant. These results demonstrate that AC systems played an important role in exacerbating the nocturnal urban heat island. The observed maximum nighttime urban-rural temperature differences ranged between 5°C and 7°C during the period analyzed, and strategies that avoid release of waste heat into the ambient environment should be considered.
Our simulations demonstrate that heat rejected from AC systems modifies the thermal stratification of the urban boundary layer. Figure 8 shows the mean vertical profile of the potential temperature at Sky Harbor Airport weather station (see Figure 1b) close to Phoenix's downtown, averaged for the 10 day EHD period for the AC65% and AC_NoOut simulations. After sunset (at 10 P.M. local time), the potential temperature difference ϑ(~250 m) − ϑ(0 m) for the AC65% experiment was smaller (~1 K) than the potential temperature difference, at the same elevations; for the AC_NoOut experiment, demonstrating that waste heat release promotes enhanced vertical mixing (Figure 8a). The turbulent kinetic energy (TKE) for the AC65% simulation was greater than the TKE for the AC_NoOut experiment in this thin layer close to the ground (Figure 9). Some hours later (at 3 A.M. local time), while the nocturnal mixing layer evolves, differences (ϑAC65%(0 m) − ϑAC_NoOut(0 m)) of ~0.75 K were computed close to the ground. The potential temperature profiles (Figure 8a) illustrate a strong stable stratification that could not be confirmed by observations because experimental data were not available. Although nighttime stable conditions are not very common in urban areas, previous work of Grossman-Clarke et al. [2007] reported a similar stratification corroborated by observations above 350 m from the surface during a 20 day period of June 2001 at the same location. During the morning, the mean potential temperature profiles were similar for both simulations (Figure 8b). The rest of the day, the mean AC65% potential temperature profiles were warmer than the corresponding AC_NoOut vertical profiles (by ~0.5 K at 3 P.M.), but a clear impact on the stability of the urban boundary layer was not simulated.
3.3 Evaluation of AC Contribution on Surface Heat Fluxes
The AC systems increased the sensible heat fluxes exchanged with the atmosphere. Simulated surface sensible heat fluxes averaged over the LIR urban grid cells for the entire 10 day EHD period are shown in Figure 10a. The sensible heat fluxes are positive when directed away from the surface into the atmosphere, and negative when directed toward (or into) the surface. The waste heat ejected by AC systems increased the total sensible heat flux beyond 420 W m−2. Although the maximum contributions were observed during the afternoon and evening (Figure 10b), the near-surface mean temperature was considerably affected only during the night (Figures 5-7). Previous work [e.g., de Munck et al., 2013; Salamanca et al., 2012] has reported that the effect of AC systems on regional climatology was more important during the night due to the limited depth of the urban boundary layer. A smaller quantity of sensible heat ejected during the night can increase the air temperature more relative to a greater quantity released during the day. Figure 11a shows the computed surface sensible heat flux but this time averaged over the HIR urban grid cells. The heat fluxes ejected by AC systems (Figure 11b) are greater than the corresponding heat fluxes computed for the LIR areas because for the former urban grid cells; the volume of indoor air conditioned is greater. The maximum total sensible heat was reached 1 h after noon and was approximately 480 W m−2 for the AC100% simulation. Finally, Figure 12a shows the surface sensible heat fluxes computed for the COI urban grid cells. Commercial areas have taller buildings and greater urban fractions than residential areas leading to larger surface sensible heat fluxes. The maximum computed sensible heat was 520 W m−2 (AC100% simulation) due to the contribution of the AC systems (50–60 W m−2) at this time of the day (Figure 12b).
Although negative values of sensible heat flux at night are usually not significant in urban areas, Grimmond and Oke [1995] reported (using observations) comparable negative values during a summertime period for a suburb in Tucson (AZ), which demonstrates that negative sensible heat flux at night does occur.
3.4 Effect of the AC Systems in the Prediction of the Air Temperature
In this section we analyze how waste heat release affected the prediction of near-surface air temperature. To address this issue all available urban weather stations were considered: the two AZMET weather stations Mesa and Waddell and the Sky Harbor Airport weather station maintained by the NWS and located close to Phoenix downtown (Figure 1b). Figures 5-7 have shown that the mean impact of the AC systems was significant during the night and negligible during the day; consequently, a considerable effect was expected only during the nighttime period.
Figure 13 shows a scatterplot of mean modeled versus mean observed 2 m air temperature averaged for the entire 10 day EHD period during nighttime hours (from 8 P.M. to 5 A.M.) at the three aforementioned urban stations. The small effect observed at Waddell was due to its proximity to the boundary of the metropolitan area. The AC_NoOut simulation underestimated the mean near-surface air temperature for all stations, and as a result, the “extra” heat added by the AC systems improved prediction of the air temperature. During the day (from 6 A.M. to 7 P.M.), WRF simulation (AC_NoOut) overestimated slightly the air temperature at Waddell and slightly underestimated the air temperature at Mesa and Sky Harbor Airport. In general, the extra heat added by the AC systems improved air temperature prediction, but the improvement was smaller during this period of time. This assertion is shown in Table 4, which summarizes the root-mean-square errors and the mean absolute errors calculated for the four simulations and both periods of time. During the night, the root-mean-square error was reduced by 0.86°C for the AC100% experiment and ~0.5°C for the AC65% simulation compared to the root-mean-square error of the AC_NoOut experiment. However, during the day, the improvement was less significant (0.19°C and 0.13°C for the AC100% and AC65% experiments, respectively) when the waste heat was added to the outdoor environment due to correspondingly smaller effect (Figures 5-7) during this period of time.
| Simulation | RMSE (°C) | MAE (°C) | ||
|---|---|---|---|---|
| Daytime | Nighttime | Daytime | Nighttime | |
| AC100% | 0.291 | 0.907 | 0.255 | 0.713 |
| AC65% | 0.347 | 1.279 | 0.336 | 1.042 |
| AC35% | 0.354 | 1.463 | 0.336 | 1.217 |
| AC_NoOut | 0.480 | 1.770 | 0.427 | 1.484 |
- a The three urban weather stations Mesa, Sky Harbor Airport, and Waddell indicated in Figure 1b were used for the calculations.
3.5 Analysis of Urban AC Electricity Consumption
Salamanca et al. [2013] demonstrated that BEP + BEM was able to reproduce satisfactorily the observed diurnal profile of Phoenix's AC electricity consumption provided by an electric utility company. However, in the previous work, the spatial distribution of the cooling demand was not assessed. Figure 14 shows the spatial distribution of diurnal mean AC electricity consumption (AC65% experiment) averaged for the entire 10 day EHD period in July 2009. The minimum AC consumption was observed shortly after sunrise (Figure 14a) and was ~2100 MW when the entire urban domain (3285 km2) was aggregated. Maximum electricity consumption happens during late afternoon hours (~4 P.M.–~6 P.M.), with use of AC systems peaking at ~5 P.M. The peak was ~8000 MW (Figure 14b) emphasizing the considerable cooling requirements for the Phoenix metropolitan area. Comparison between the AC65% and AC_NoOut experiments produced a mean energy saving of 1277 MWh by day due to the difference in AC electricity consumption.
4 Conclusions
In this paper, the effect of the AC systems has been analyzed with the multilayer building energy parameterization BEP + BEM coupled to the atmospheric WRF model for the Phoenix metropolitan area. In previous work, BEP + BEM was validated by comparing computed AC electricity consumption against observationally derived AC electricity consumption provided by an electric utility company. The excellent agreement concerning both diurnal profiles gave us confidence to assess the contribution of AC systems on air temperature and to examine spatial distribution of electricity consumption.
Three different ratios of air-conditioned floor area to total floor area have been investigated. When the total floor area was air conditioned (this case can be considered as an upper limit of the impact of AC systems), the waste heat release from AC systems increased the mean nighttime temperature 1–1.5°C, amplifying considerably the magnitude of the urban heat island. When the ratio was reduced to 0.65, somewhat more realistic according to preceding AC electricity consumption estimates, the mean nighttime temperature was increased between 0.5°C and 1°C, still showing a significant effect in the region. Finally, when the ratio was reduced to 0.35, the increase was 0.25–0.5°C across most of the built environment. During the day, although sensible heat fluxes from the AC systems are maximum, the mean effect was not significant near the surface. It was somewhat evident only when the total floor area was considered air conditioned. Considering that the AC65% experiment is the most representative scenario for Phoenix, we conclude that the AC systems increased the mean nighttime 2 m air temperature up to 1°C over some urban locations, although for most areas the increase was between 0.5°C and 0.75°C. Waste heat release favored nocturnal vertical mixing and contributed to exacerbate the urban heat island. In polluted environments waste heat release could be regarded as favorable since it might reduce the concentration of pollutants near the ground, but a detailed analysis is beyond the scope of the present work.
The effect of AC systems in the prediction of near-surface air temperature was also analyzed. In general, accounting for waste heat release due to AC systems improved mean 2 m air temperature prediction reducing the root-mean-square error in ~0.5°C (AC65% simulation) during the night. WRF simulation (AC_NoOut) underestimated nighttime temperatures, and the extra heat added by AC systems reduced this bias. During the day, the improvement was minor because of the small effect of AC systems on near-surface air temperature.
Our work demonstrates 1°C local heating of urban atmospheres in hot/dry cities due to air conditioning use at nighttime. This increase in outside air temperature in turn results in additional electricity demands for air conditioning. Sustainable development and optimization of electricity consumption in cities would require turning “wasted heat” from AC into “useful energy” which can be utilized inside houses for various purposes including, for example, in water heaters. Implementing this mitigation strategy would achieve several objectives: successfully reducing the urban heat island temperature by 1°C at night, reducing AC electricity consumption on a city scale, and providing a real example of urban climate mitigation. With regard to economic impacts, it is estimated for Phoenix metropolitan area that successfully reducing the urban heat island temperature with this strategy would result in at least 1200–1300 MWh of direct energy savings per day alone.
Climate change projections predict significant increases in the frequency, intensity, and duration of summertime extreme heat events presenting significant challenges for the energy sector and electric grid [U.S. Department of Energy, 2013]. Reliable methods are needed for forecasting energy demands that can help to inform and assist in the future planning of sustainable energy needs of rapidly growing urban areas. Comparison with observational data has demonstrated that the presented physics-based modeling system is an effective tool for assessing urban cooling requirements in semiarid environments. In ongoing work, we are evaluating AC electricity consumption for different scenarios of urban expansion and under summertime extreme weather conditions. These studies are needed to develop reliable projections on the future cooling needs of rapidly urbanizing regions.
Acknowledgments
This work has been funded by National Science Foundation grant ATM-0934592. We thank the Global Institute of Sustainability at Arizona State University and the reviewers for their valuable comments, which helped to improve this paper.
