As suggested in the previous sections, there are different approaches that have been used to estimate anthropogenic emissions in the urban environment. These include inventory approaches, micrometeorologically based energy budget closure methods, and building energy modelling approaches. As discussed below, each approach has its advantages and limitations.
3.1. Inventory approaches
An inventory approach is one in which energy consumption data are gathered, typically at coarse resolution in time and space. These approaches may rely on utility scale energy consumption data, or data resulting from energy consumption surveys. The vast majority of inventory applications make the assumption that consumption is equal to anthropogenic sensible heat emissions, with no time lag, no accounting for environmental loads, and no partitioning into sensible and latent components. Further, they generally must use some mechanism to map annual or monthly consumption data onto diurnal profiles. They must also map city-wide or regional emissions estimates onto a finer scale grid to represent spatial variability within the city. Taha (1997) gathered data from a number of early inventory-based approaches and found that the annual city-wide estimates of anthropogenic heating typically ranged from 15 to150 W/m2. One of the earliest of these inventory studies was that of Torrance and Shum (1975) who related annual energy consumption data to population density to arrive at annual estimates of 83.7 W/m2 for a densely populated city. In this work, Torrance and Shum represented space heating through a simplified empirical model that scaled space heating energy use with heating degree days. Their industrial component of anthropogenic heat was simply assumed to be constant in time. They represented the transportation sector by using hourly traffic count data averaged over the city. Their model represented the spatial variation of anthropogenic heating only in terms of differentiating ‘urban’ from ‘rural’ components. Despite its limitations, this study was perhaps the first to estimate anthropogenic heat emissions (up to 135 W/m2 for a summer day in an unspecified dense city) and the resulting impact on urban air temperatures. Specifically, through use of a transient one-dimensional atmospheric model, Torrance and Shum estimated that anthropogenic heating could result in up to a 4 °C warming in summer. Subsequent studies improved on both the spatial and temporal disaggregation of inventoried energy consumption data. Kimura and Takahashi (1991), for example, used detailed land use data in combination with energy consumption data to estimate anthropogenic heating and its impact on urban air temperatures in Tokyo. While they used a fairly simple approach to estimate the diurnal profile (early morning minimum equal to half the midday value), they were able to take advantage of detailed (2 km by 2 km) gridded land use data to create a somewhat realistic spatial map of anthropogenic heat emissions. Peak anthropogenic heating from their analysis was on the order of 100 W/m2. By feeding these sensible heat emissions into a hydrostatic three-dimensional numerical model, they then estimated that most of the nocturnal summertime heat island in Tokyo (2–3 °C) is due to anthropogenic heating. This was a particularly significant finding, as it clearly illustrated the important role that anthropogenic heating plays in the formation of the urban heat island.
Similar inventory approaches have been applied in a range of urban regions around the world (e.g. in Korea by Lee et al. (2009) and in London by Harrison et al. (1984)). One such example is the work of Klysik, (1996), who used an inventory approach to estimate anthropogenic heat emissions in Lodz, Poland. In his work, the annual heating energy consumption data were mapped to monthly profiles using variability in monthly air temperatures. Population density was used as a metric for discerning spatial variation in electricity use. Anthropogenic heat emissions from the vehicle sector were estimated on the basis of annual fuel sales. While limited to coarse resolution in space and time, Klysik's analysis suggests monthly averaged anthropogenic heat flux in Lodz ranging from 12 W/m2 in summer to 54 W/m2 in winter.
Ichinose et al. (1999) took a major step forward in spatially resolving anthropogenic heating. They linked energy statistics for various building types and end uses in Tokyo [obtained from Ichinose et al. (1994)] with detailed digital geographic land use data (12 land use categories) to estimate hourly anthropogenic heating at a 25 m by 25 m grid resolution. As expected, the high spatial resolution led to some rather large estimates of heating associated with grids that were centred on large high-rise office buildings. Specifically, they found that hourly energy consumption at this scale (assumed to be equal to anthropogenic sensible heating) could be as large as 1590 W/m2 in the winter and was more than 400 W/m2 during summer daytime hours. Their analysis also integrated these estimates of anthropogenic heat into a numerical simulation (using the Colorado State University Mesoscale Model). They found that anthropogenic heating resulted in summer air temperature elevations of less than 1 °C during the day, but as much as 1.5 °C at night. In winter, when the magnitude of anthropogenic heating was much higher (and mixing heights presumably lower), they found that the nocturnal warming reached 2 to 3 °C.
Fan and Sailor (2005) implemented the inventory-based approach of Sailor and Lu (2004) in a study of anthropogenic heating in Philadelphia, Pennsylvania. That study found peak anthropogenic heating at the city scale to be 60 W/m2 in summer and approximately 90 W/m2 in winter. The impacts of this anthropogenic heating on urban air temperatures ranged from 1 °C in summer to 3 °C in winter. One of the unique characteristics of the method presented in Sailor and Lu (2004) is that it emphasised use of data that are widely available for most large US cities. The modelling approach was inventory based and diurnally distributed based on widely applicable profile functions. As with most inventory-based approaches, the work of Sailor and Lu implemented the assumptions that energy consumption equals anthropogenic heat and that all of this heating is sensible. Another limitation of this work was that it provided no mechanism for mapping city-wide anthropogenic heating data onto finer spatial scales. Nevertheless, its wide applicability led to its application in Sailor and Hart (2006) to 50 US cities.
In a study of London, Hamilton et al. (2009) used a similar approach to estimate inventories of building energy consumption across the city, and then mapped this consumption onto 1-km grids based on detailed land use and building characteristic data. They found spatially and temporally averaged anthropogenic heating in London to be 9 W/m2, but noted that the spatial and diurnal variations were significant.
Recently, there has been an increased interest in implementing anthropogenic heating in large-scale general circulation models. Thus far, these efforts have also used an inventory-based approach with necessarily crude representations of spatial and temporal variability. While observing anthropogenic heating impacts on regional climates in central Europe, for example, Block et al. (2004) used regional estimates of energy consumption and population density as a proxy for anthropogenic heat emissions. Their work simply assumed this heating to be constant in time, and used such a large estimation area that the magnitude of anthropogenic heating was 14.5 W/m2 or less throughout their domain. Flanner (2009) also used an inventory approach to estimate anthropogenic heating for use in global climate models. His work mapped country-wide energy consumption data into 2.5 by 2.5 min latitude/longitude grids using population density. On the basis of the work of Sailor and Lu (2004), Flanner used weighting functions to create seasonal and diurnal profiles. Owing to the relatively coarse scale of the analysis, the resulting magnitude of anthropogenic heat was relatively low.
These inventory approaches all suffer from two main limitations: first, they can only resolve energy consumption at relatively coarse spatial and temporal scales and must use some technique to create estimates at finer scales and, second, they make the assumption that energy consumption is equivalent to anthropogenic sensible heat emissions.
3.2. Estimates based on energy budget closure
In contrast to inventory approaches that are generally based on relatively coarse data, direct measurement of the urban energy budget enables estimation of anthropogenic heating over targeted neighbourhoods within a city. In theory, if one draws a control volume around a portion of a city as in Figure 3, it is possible to track all energy movement into and out of this control volume. In this figure, Q* is the net radiation, QE and QH are latent and sensible heat transported across the top of the control volume, ΔQA is the total heat advected out of the sides of the control volume, ΔQs is the storage of thermal energy within the control volume, and QF is the total source of anthropogenic heat within the control volume. From the micrometeorological standpoint, one can use net radiometers to track net radiation (Q*), and eddy covariance techniques to estimate sensible and latent heat fluxes (QH and QE). If the net advection out of the control volume is assumed to be negligible and the total energy stored in the control volume is either estimated through surface measurements or assumed to be stationary from one day to the next, the energy balance can be simplified and solved for QF:
This equation is applicable for small time steps (say 1 h), or if one integrates the energy balance over the course of a 24-h period, the net storage term may be assumed small, and ignored.
Figure 3. A control volume analysis of the energy budget of a portion of a city in close proximity to a micrometeorological tower (a). As illustrated, the nominal control volume (b), radiation measurement footprint (c), and eddy flux footprint (d) may not coincide
Download figure to PowerPoint
One particularly attractive yet limiting aspect of this method is that it inherently includes anthropogenic heating from all sources. This sort of approach has been applied by a relatively small number of investigators with differing levels of success. Offerle et al. (2005) applied this concept to data gathered in a downtown region of Lodz, Poland. These data included surface temperature and heat flux measurements, measurements of net radiation, and an eddy covariance approach to estimate sensible and latent heat flux. The challenge with using an energy balance residual approach is that any errors within the measurement system are simply incorporated into the estimate of the residual. Specifically, due to significant challenges in using eddy flux methods in a non-homogeneous urban system, the magnitude of the errors in estimating sensible and latent fluxes may be large. Also, as illustrated in Figure 3, the source area for radiation measurements can be thought of as being a circular area with the centre at the instrument tower. The flux measurements, however, essentially represent a source footprint area dictated by upwind flow and dispersion characteristics. As a result, the nominal source area for the radiation measurements may not coincide with or be similar to the source area for the eddy flux measurements. The assumptions regarding thermal storage and advection add uncertainty to estimates of anthropogenic heat flux. Therefore, it is not surprising that the results of Offerle et al. include unrealistically negative numbers for anthropogenic heating in summer months. Their estimates for wintertime anthropogenic heating, however, were more consistent with what is expected from inventory methods—magnitudes around 50 W/m2.
In a contrasting approach, Kato and Yamaguchi (2005) employed an energy budget residual approach in which remote-sensed satellite data were used in conjunction with surface meteorological data to estimate anthropogenic heating in Nagoya Japan. Owing to the nature of their measurements (surface radiation), Kato and Yamaguchi's results do not allow estimation of anthropogenic heat directly, but rather, the increase in surface sensible heat flux resulting from anthropogenic heating and the associated warming of the near-surface air.
Pigeon et al. (2007) present one of the most recent and comprehensive energy-budget-based estimates of anthropogenic heat release. Their research focused on Toulouse, France as part of the CAPITOUL experiment. This study included both an inventory-based approach with some similarities to Sailor and Lu (2004) as well as an energy budget residual approach similar to Offerle et al. (2005). The particularly useful aspect of their study was that it enabled direct comparison of the two approaches. They found the two approaches to be in general agreement with wintertime anthropogenic heating as high as 70 W/m2 and summertime values closer to 15 W/m2.
Overall, the energy budget micrometeorological approach to estimating sensible anthropogenic heating has value as a local validation for other estimation approaches. The key limitations of this approach include a high level of uncertainty associated with some of the assumptions required to deduce anthropogenic heating as a residual in an energy budget. Further, these approaches cannot differentiate among the different source terms and often cannot reliably produce diurnal profiles of heat release.
3.3. Estimates based on building energy models
While the inventory and energy-balance-based approaches are readily capable of estimating anthropogenic heating from all sources in the urban environment, anthropogenic heating from the building sector is particularly complicated and has been the subject of a number of focused studies. These studies generally involve explicit modelling of energy consumption within buildings and careful evaluation of heat rejection. The town energy budget (TEB) model of Masson (2000) was an early attempt to model the anthropogenic heat from buildings into the urban environment. While TEB represented a significant advancement with respect to modelling turbulent processes and energy budgets in urban canyons, its representation of buildings was rather simplified. It treated buildings as envelope structures with conduction through the envelope dictating the need for heating or air conditioning, which is then emitted into the environment. This formulation is insufficient as it ignores internal loads in buildings, which can lead to increased cooling demand and reduced heating demand. It also ignores shortwave heating of internal spaces through glazing (windows), which typically accounts for 20 to 40% of the wall surface area of buildings and air exchange between the building and the outdoor environment (infiltration and ventilation). Nevertheless, the TEB model has undergone additional development and improvement over the years as well as validation efforts that verify its ability to accurately represent the urban surface energy balance (e.g. Pigeon et al., 2008).
Kikegawa et al. (2003) presented one of the first studies that integrated building energy simulation results with an urban canopy meteorological model. Their building energy submodel explicitly accounted for building occupants, radiative transfer through windows, type of air-conditioning heat exchanger (air-cooled vs evaporatively cooled), and performance of air-conditioning systems. As a result of using such a detailed building energy submodel, they were able to partition HVAC heat rejection between latent and sensible components and provide a more realistic estimate of anthropogenic emissions. In a follow-on study, Kondo and Kikegawa (2003) used this modelling approach to explore the relationship between anthropogenic energy use and air temperature within the urban canopy. In this study, they simply assumed a one-to-one ratio between latent and sensible emissions from air-conditioning equipment. In 2006, the same research team published a study using this modelling approach to explore impacts of urban heat island countermeasures on building sector air-conditioning energy use (Kikegawa et al., 2006). This work was followed by an application where they applied their building energy modelling approach to explore the impact of anthropogenic heat on summertime urban air temperatures in Tokyo (Ohashi et al., 2007). In that study, they found that air-conditioning heat rejection resulted in summertime air temperature elevation of 1 to 2 °C in Tokyo. It should be emphasised that, as with most building energy focused modelling approaches, this work ignored anthropogenic heat from other sources (vehicles and industry). In a similar set of studies, Dhakal and colleagues (Dhakal et al., 2003, 2004) used building energy simulation to estimate the magnitude and timing of anthropogenic heat emissions from residential and commercial buildings in Tokyo. They were among the first to document the time lag and magnitude difference between energy consumption and heat emissions.
In another category of modelling studies, researchers have integrated detailed building energy simulations for prototypical buildings with geographical information system data bases containing characteristics of building types and sizes. Such studies allow researchers to model a representative building stock and then for each land use parcel (or taxlot) within a domain associate a floor-space scaled anthropogenic heating and moisture emissions from the corresponding building prototype. As illustrated in Figure 4, the anthropogenic emissions data can then be aggregated from the taxlot scale to the scale of the atmospheric model grid cell. Hsieh et al. (2007) conducted such a study for Taipei, Taiwan using the EnergyPlus building energy simulation tool and a total of 10 building prototypes (4 residential and 6 commercial). Heiple and Sailor (2008) conducted a very similar study in which they used a total of 30 building prototypes (8 residential and 22 commercial) to represent the building stock in Houston, Texas. Initially they simply applied this approach to estimate hourly energy consumption at grid resolutions of 100 m. This approach was later implemented in a study that extracted building sector anthropogenic sensible heat and moisture and linked these data to an inventory-based transportation sector model (Sailor et al., 2007). This study estimated that, due to significant environmental loads, heat rejection from buildings can be 50 to 100% greater than the energy consumption of the building. It also found that 50 to 80% of the heat rejected from buildings in the commercial core of a city such as Houston (with many large buildings with evaporative cooling towers) is rejected as latent heat. While the generation of building prototypes may appear to be a significant obstacle to implementing this method, it should be noted that there are many efforts in the building energy research community that have addressed this need, beginning with the work of Huang et al. (1991) and culminating with the recent release of the US Department of Energy's commercial building benchmarks—a compilation of prototypes for 16 building types across 16 climate zones (Torcellini et al., 2008). Such prototype models are readily linked at the taxlot level to GIS data bases containing building size and categorisation data.
Figure 4. A series of prototypical buildings such as (a) are modelled and then mapped to individual tax lots wihin a GIS (b), and then aggregated up to an atmospheric model grid cell (c)
Download figure to PowerPoint
In these sorts of building energy modelling approaches, it is important to note that the building energy simulation tool (e.g. EnergyPlus) requires input meteorological data. The simulations are usually driven by typical meteorological year (TMY) data files that are a climatological representation of typical weather at the nearest airport weather station. Such representations have the obvious drawbacks of possible differences between the urban location under study and the airport site that was used to generate the TMY file. Of course, it is also possible to use actual measurements from a local weather station. For the most part, however, these building sector prototype approaches have focused on generating representative profiles of anthropogenic emissions of heat and moisture rather than estimates for any specific episode. The latter would require a dynamic linking of an urban atmospheric model and an anthropogenic heating model.