Environ. Sci. Technol. 2002, 36, 3139-3146
Micrometeorological Measurements of the Urban Heat Budget and CO2 Emissions on a City Scale E I K O N E M I T Z , * ,† KENNETH J. HARGREAVES,† ALAN G. MCDONALD,† JAMES R. DORSEY,‡ AND DAVID FOWLER† Centre for Ecology and Hydrology (CEH), Edinburgh, Bush Estate, Penicuik, Midlothian EH26 0QB, Scotland, and Physics Department, UMIST, P.O. Box 88, Manchester M60 1QD, U.K.
Direct measurements of urban CO2 emissions and heat fluxes are presented, made using the eddy covariance technique. The measurements were made from the top of a tower, approximately 65 m above the street level of Edinburgh, Scotland, and the fluxes are representative of footprint source areas of several square kilometers. The application of a stationarity test and spectral analysis techniques shows that, at this height, the stationarity criterion for eddy covariance is fulfilled for wind directions from the city center for 93% of the time, while for other wind directions this declines to 59%, demonstrating that pollutant fluxes from urban areas can be measured. The average CO2 emission from the city center was 26 µmol m-2 s-1 (10 kt of C km-2 yr-1), with typical daytime peaks of 5075 and nighttime values of 10 µmol m-2 s-1. The correlation between CO2 emission and traffic flow is highly significant, while residential and institutional heating with natural gas are estimated to contribute about 39% to the emissions during the day and 64% at night. An analysis of the energy budget shows that, during the autumn, fossil fuel combustion within the city contributed one-third of the daily anthropogenic energy input of 3.8 MJ m-2 d-1, with the remainder coming from other energy sources, dominated by electricity. Conversely, the total energy input in late spring (May/June) was found to be approximately half this value.
1. Introduction Carbon dioxide (CO2) has long been recognized as the major contributor to radiative forcing of climate (e.g., ref 1), with global emissions of about 7 Gt of C yr-1. Of these emissions, about 6.2 Gt of C yr-1 (88%) is believed to originate from fossil fuel combustion sources, with a significant contribution from traffic (28%) (2). These emissions have been estimated as the product of traffic and production activities with emission factors, e.g., obtained on vehicle test stands. Despite the importance of CO2, the exact magnitude and the spatial distribution of its emissions remain poorly quantified. For example, for 1999 the official U.K. National Atmospheric Emissions Inventory (NAEI) predicted U.K. emissions of 149 * Corresponding author phone: +44 (0)131 445 4343; fax: +44 (0)131 445 3943; e-mail:
[email protected]. † Centre for Ecology and Hydrology (CEH). ‡ UMIST. 10.1021/es010277e CCC: $22.00 Published on Web 06/12/2002
2002 American Chemical Society
Mt of C yr-1, with a contribution from transport of 22% and small combustion sources, including residential, commercial, and institutional sectors, of 21% (3). By contrast, from measurements of the CO2 concentration upwind and downwind of the British Isles during so-called round-Britain flights, we have derived larger CO2 emissions of 260 Mt of C yr-1 in the summer and 307 Mt of C yr-1 in the winter of 1994 (4). Although CO2 surface/atmosphere exchange fluxes are routinely measured above vegetation (e.g., above forest in the major projects EUROFLUX, BOREAS, and AMERIFLUX), there are few (if any) direct validation data of anthropogenic emission inventories, particularly of area sources such as traffic, residential, and institutional heating and other small combustion processes. Here we report the application of conventional micrometeorological techniques to measure CO2 emissions from a substantial fraction of the city center of Edinburgh, Scotland. In urban areas, micrometeorological approaches have previously been applied to measure fluxes of momentum and heat (e.g., ref 5), and we have expanded these techniques to quantify pollutant emissions in the urban environment. The measurements presented here form part of a U.K. initiative to quantify the “Sources and Sinks of Urban Aerosols” (SASUA). The results of the particle flux measurements are reported elsewhere (6-9). This paper reports the magnitude of the CO2 fluxes above the city in relation to wind direction and selected activities within the footprint. The application of micrometeorological techniques to a potentially inhomogeneous area such as a city relies on the assumption that, viewed from a high measurement location, the small scale emission and deposition processes merge together to form a homogeneous net flux. In the analysis, some emphasis is therefore placed on demonstrating that the underlying boundary conditions for flux measurements are satisfied and on the development of appropriate filter criteria. The heat generated by the combustion processes emitting the CO2 is then estimated and put into context of the overall heat budget over the city.
2. Methods 2.1. Measurement Site. Measurements were made during the period October 28-November 30, 2000, in the center of Edinburgh, which accommodates 450 000 inhabitants and covers an area of 261 km2. Eddy covariance equipment was mounted on a 2.5-m mast fixed to the upwind edge (for the preferred westerly airflow) on top of the external wall of Nelson Monument, a 32 m tall stone tower, with an approximate diameter of 2.8 m at the top. The tower is situated at the highest point of Calton Hill, 35 m above the street level of the surrounding city and 102 m above sea level (OS grid coordinate NT 26258 74134; 3°10′52′′ W, 55°57′17′′ N). Situated to the WSW of the tower is Princess Street, the principal shopping street of the Scottish capital, while the city center lies in the wind sector 190-330°. A major natural recreation area (Holyrood Park) lies to the SE of the measurement point, starting at a distance of about 800 m. To the NE and E direction, the immediate surroundings of the tower are dominated by Calton Hill Park itself, a mixture of grass and woodland. Beyond this, at a distance of 200400 m and in the northerly wind sector, the terrain is dominated by residential areas, intercepted by some major commuter roads (cf. Figure 6). 2.2. Theory and Implementation of the Eddy Covariance Technique. Fluxes (Fχ) were measured and calculated according to the eddy covariance technique as the covariance between the deviation of the vertical wind component (w) VOL. 36, NO. 14, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3139
and CO2 concentration (χCO2) from their mean (e.g., ref 10):
Fχ ) w′χ′CO2
(1)
The wind components were measured using an ultrasonic anemometer (asymmetric R1012 Solent Research anemometer, Gill Instruments, Lymington, U.K.) at 20.83 Hz, while the concentrations of CO2 and water vapor (H2O) were measured with a LI-COR 6262 gas analyzer (LI-COR, Lincoln, NE), with a response time of ca. 4 Hz. The gas analyzer was housed in a heated enclosure on top of the tower, and the sample was drawn from 0.4 m below the center of the anemometer at 9 l min-1 through approximately 4 m of 0.25 in. Decabon tubing, resulting in a time-lag of approximately 0.6 s (or 12 data points at 20.83 Hz). The analogue outputs of the LI-COR 6262 were fed into the analogue inputs of the ultrasonic anemometer and recorded at 20.83 Hz, together with the wind data, using a computer program written in LabView 5.1 (National Instruments) (9), which also logged two particle counters, stored the raw data, and performed preliminary online flux calculation. The data were reprocessed, and flux calculations included (i) calculation of the optimum time-lag by calculation of the cross-correlation spectrum over a user-specified window (11); (ii) linear detrending of all time series to account for changes in meteorological conditions and concentrations that are not related to surface emissions (12); and (iii) coordinate rotation, aligning the wind component u with the mean wind direction and setting the mean values of v and w to zero (13). 2.3. Limitations of the Eddy Covariance Technique, Error Analysis, and Correction Procedures. The high measurement location results in a large mean eddy size and therefore relaxes the demands on the sampling frequency. As a result, the vertical spatial separation between the center of the ultrasonic anemometer and the gas inlet (0.4 m) was estimated to lead to flux losses of less than 1% (14). The flux loss due to limitations in the sampling rate of the CO2 analyzer (4 Hz) was typically 1 making a contribution of 0.3, N ) 30) from the expected frequency under the shifted log-normal distribution. The shift may indicate that in the absence of anthropogenic sources the city would be a sink for CO2. (b) The filter criterion removes mainly small flux values, and in particular almost all deposition fluxes, reducing the contribution of deposition observations from 6.1% to 0.9%. This filter criterion removed only 7% of the data for the city center wind sector (190-330°), while it removed 41% of the data from the remaining wind sector in which the fetch and instrument logistics are less ideal for micrometeorological flux measurements. The net effect was a data loss of 18%. An example time-series of concentrations and fluxes is shown in Figure 3 in relation to wind speed and direction. Fluxes are shown both before and after they were filtered for nonstationarities and corrected for storage errors to demonstrate the effect of this procedure. During this period, air concentration and flux show a clear diurnal cycle, which is consistent with the diurnal cycle expected for anthropogenic activities. However, the concentration is additionally governed by dispersion, which is closely linked to wind speed:
FIGURE 3. Example time series of CO2 concentration (χCO2) and flux (Fχ) in relation to wind speed (u) and wind direction (O) for the period Tuesday-Saturday, November 14-18, 2000. The flux graph shows the raw measured flux (thin line) and the flux corrected for storage errors and filtered for stationarity coefficients (ξ) < 0.2 (bold line). at low wind speeds of
