ENERGY AND RADIATION BALANCE OF A CENTRAL EUROPEAN CITY

INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 24: 1395 1421 (2004) Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/joc.1074 ENERGY AND RADIATION BALANCE OF A CENTRAL EUROPEAN CITY ANDREAS CHRISTEN* and ROLAND VOGT Institute of Meteorology, Climatology and Remote Sensing, Department of Geosciences, University of Basel, CH-4055 Basel, Switzerland Received 23 February 2004 Revised 26 May 2004 Accepted 26 May 2004 ABSTRACT Results from an experimental network of seven energy balance stations in and around a European city are presented. The network of micrometeorological stations was part of the Basel Urban Boundary Layer Experiment (BUBBLE) carried out in the city of Basel, Switzerland. Three urban sites provided turbulent flux densities and radiation data over dense urban surfaces. Together with a suburban site and three rural reference sites, this network allowed the simultaneous comparison of urban, suburban, and rural energy balance partitioning during one month of summertime measurements. The partitioning is analysed together with long-term data to evaluate the magnitude of the urban flux density modification, and to document characteristic values in their diurnal and yearly course. Simple empirical relations between flux densities and surface characteristics are presented. The energy balance partitioning is addressed separately for daytime and nocturnal situations. All four components of the surface radiation budget are analysed. Moreover, the vertical flux density divergences within the urban canopy layer are discussed. Copyright 2004 Royal Meteorological Society. KEY WORDS: urban energy balance; urban radiation balance; turbulent flux densities; eddy correlation; storage heat flux; albedo, urban rural differences; vertical flux density divergence 1. INTRODUCTION Understanding the surface energy balance of built-up areas is important for forecasting the urban boundary layer. The modified partitioning of the urban energy balance affects the whole boundary layer, its stability, thermodynamic properties, and the mixing layer height. And finally, the modified urban energy balance results in typical urban climate phenomena like the urban heat island. Thus, a detailed understanding of the surface energy balance is essential for accurate meteorological and air pollution forecasting in cities, as well as in urban dispersion models. Moreover, current meso-scale models are run with increasingly higher spatial resolution, resulting in more grid points that will be attributed to true urban. Exchange parameterizations for these surfaces must take into account roughness characteristics and urban-specific surface exchange processes. In recent years, a number of sophisticated urban parameterizations have been developed for meteorological models that partially solve and partially parameterize effects of complex three-dimensional urban surfaces (Masson, 2000; Martilli et al., 2002). Also, experimental results help us to determine the relative importance of different processes, and to verify and improve parameterizations of the energy exchange over such a complex surface. In the 1990s a number of full-scale experimental studies significantly increased the knowledge in this field (Arnfield, 2003). Most of the studies investigated the energy balance over suburban residential areas, because they cover the largest areas of today s cities. An impressive series of experiments conducted at the beginning of the 1990s over North American suburban surfaces yielded insight into various specific features of urban surface energy partitioning (Grimmond and Oke, 1995, 1999a, 2002). Owing to logistical * Correspondence to: Andreas Christen, Institute of Meteorology, Climatology and Remote Sensing, Department of Geosciences, University of Basel, Klingelbergstrasse 27, CH-4055 Basel, Switzerland; e-mail: andreas.christen@unibas.ch Copyright 2004 Royal Meteorological Society

1396 A. CHRISTEN AND R. VOGT and methodological problems, only a few energy balance campaigns addressed dense city centres. In recent years, a stronger effort has been made to investigate the energy exchange and partitioning over such highly populated central-city surfaces in different campaigns (Oke et al., 1999; Feigenwinter et al., 1999; Kanda et al., 2002). Surface characteristics like albedo α, complete aspect ratio λ C, roughness length z 0, or moisture availability significantly control the energy balance partitioning of any surface. Therefore, it is not only important to have accurate and representative measurements, but also detailed information on the characteristics of the surface and their spatial variability, i.e. on the urban three-dimensional structure, land cover, and typical materials. This is especially true for heterogeneous surfaces such as cities. There are several methods to classify urban surfaces into different categories and then attribute typical values. The three-dimensional structure is usually described by appropriate morphometric parameters (Grimmond and Oke, 1999b). Classifications according to morphometric parameters help to generalize results from micrometeorological studies and to relate findings to universal surface characteristics which are applicable for any city surface. Besides the morphology of a city, the climatic setting, local meteorological effects, and the weather conditions during the measurement campaign also significantly affect the surface energy balance partitioning. By simultaneously measuring the urban and rural partitioning at many locations, some of these influences can be reduced and eliminated. However, the characteristics of the rural surroundings (e.g. surface cover and water availability) vary widely and, therefore, can dramatically affect the urban modification of the surface energy balance, as shown by Oke and Grimmond (2002). In this paper the results from a study with an experimental network of seven energy balance stations in and around a central European city are presented. The network of micrometeorological stations was part of the Basel Urban Boundary Layer Experiment (BUBBLE) carried out in the city of Basel, Switzerland. Data are analysed in order to evaluate the magnitude of the energy balance modification due to urbanization. Key features and specific effects of the urban energy balance are described, rather than the details of the mechanisms of the radiative and turbulent transfer. Simple empirical relations between flux densities and surface characteristics, such as land use and morphometric parameters, are presented. Further, the aim is to give typical values for European cities for use in simple parameterizations of urban climates. Section 2 summarizes the methods and documents the sites and instrumentation. The four components of the surface radiation budget are discussed in Section 3, followed by the two turbulent flux densities and the storage heat flux densities in Section 4. 2. METHODS The general approach to estimate the urban energy balance is to measure high above the mean building height z H in order to avoid local effects of single roughness elements, i.e. the measurement height z m has to be above the roughness sublayer height z* (Rotach, 1999, 2002). Therefore, sensors are mounted on high towers to measure vertical energy flux densities Q which refer to the upper margin z m of an imaginary box enclosing all buildings and vegetation of the urban surface. The energy balance at the upper margin of such an urban box can then be written as Q + Q H + Q E + Q S + Q F = 0 (1) Flux densities of net all-wave radiation Q, sensible heat Q H and latent heat Q E can all be measured directly at z m, the latter two by the eddy correlation technique. In this paper, a simple budget view is applied with the following sign convention: all positive terms transport energy towards the surface, and negative terms indicate an energy loss from the surface. Note that this is in contrast to the conventional sign convention in flux gradient relationships. In this paper, upward-directed turbulent flux densities are negative because they represent an energy loss from the surface. This has the advantage that all terms have a consistent sign convention, which defines whether a term is currently an energy gain (+) to or loss ( ) from the surface. The measured turbulent flux densities are an area-averaged response of the surface, where the flux source areas depend on wind direction and stability (Schmid and Oke, 1990). The instruments at any urban

URBAN ENERGY AND RADIATION BALANCE 1397 site measure an integrated flux from an array of buildings, streets, backyards, and vegetation, which are representative of the local scale ( urban neighbourhood, 10 4 10 6 m 2 ). The source areas of the downwardlooking radiation instruments and the variable-source areas of the eddy correlation instrumentation usually do not refer to the same area. Therefore, such a setup has to assume horizontal homogeneity of the surface properties and flux densities at ground level, which is especially difficult to find over complex urban surfaces. Above z*, flux densities and scalars are horizontally homogeneously distributed, with the consequence that horizontal energy advection is negligible most of the time. A problem arises from the determination of the urban storage heat flux density into the ground and buildings, Q S. In contrast to rural surfaces, Q S cannot be measured easily. The large number of surface materials, orientations, and their interaction make direct measurements very laborious and nearly impossible. Therefore, Q S is usually modelled or determined as the residual term of the energy balance equation, assuming complete closure of the energy balance. Finally, there is another urban-specific input term. The anthropogenic heat flux density Q F includes all additional energy input produced by human activities, such as the energy released by combustion of fuels and electric heat, e.g. in industry, traffic, through heating and air conditioning. Q F must be either converted to radiation, sensible or latent heat, or be stored. Net all-wave radiation Q is partitioned into shortwave (K) and longwave (L) components. The arrows denote the directions of the radiation flux densities, and the same sign convention as above is applied: components transporting energy to the surface are positive, and those removing energy are negative: Q = K + K + L + L (2) 2.1. Measurement sites The dataset presented was sampled and analysed during BUBBLE, 2001 03. BUBBLE (Rotach et al., accepted for publication) is an effort within the framework of the European COST 715 action, Meteorology applied to urban air pollution problems (Fisher et al., 2002) which has been developed to increase understanding of urban surface energy exchange, turbulent exchange mechanisms, dispersion processes, and their link to the whole urban boundary layer. An extensive measurement campaign was carried out in and around the city of Basel, Switzerland, in 2001 02. Basel is a mid-size town with a built-up area of approximately 130 km 2 (30 km 2 dense urban, 80 km 2 suburban and 20 km 2 industrial areas) and a population of 400 000. A network of seven energy balance sites with eddy correlation instrumentation was operated in and around the city (Figure 1). The simultaneous operation of this experimental network provides the basis for a detailed investigation of turbulent flux densities over different urban and rural surfaces under the same synoptic forcing. The energy balance sites were located in flat areas with homogeneous surface properties in the prevailing wind directions. Two urban towers provide measurements from densely built-up surfaces in the city centre (U1, U2). Source areas of the eddy correlation instrumentation at U1 and U2 have been calculated according to Schmid et al. (1991). The 90% source areas reach approximately 300 to 500 m upwind under typical summertime convective situations, when flux densities are significant. Those areas are characterized by z H > 10 m, a high plan aspect ratio of buildings λ P, a low vegetation cover, and therefore low moisture availability (Table I). Most buildings in this area are residential multi-storey row houses, enclosing a large backyard. The backyards are either open (green spaces) or built up by one-storey garages, car parking or flat commercial industrial buildings. This city structure has a high population density, typically between 200 and 300 ha 1, and is typical of that found in many European cities constructed before the 1950s. A third urban site (U3) was placed near U1, on a large, 170 80 m 2, concrete roof in a recently developed city centre area characterized by big building blocks. The suburban tower (S1) was set up in a vegetated backyard of a uniform residential neighbourhood, consisting of two to three storey, single and semi-detached houses constructed at the beginning of the 20th century. The rural sites are all located in non-irrigated agricultural grassland (R1, R3) or crops (R2). The surface characteristics of all sites are summarized in Table I. Plan aspect ratios are retrieved from aerial photographs and describe the two-dimensional plan area of a particular surface type per total area. In

1398 A. CHRISTEN AND R. VOGT Figure 1. Map of all surface sites during the IOP in June July 2002. Squares indicate energy balance sites; circles are standard meteorological or profiler sites. The black line denotes the River Rhine. The denser built-up areas are shown in a dotted pattern. Topography is indicated by the contour lines in metres a.s.l this study, all surfaces and related characteristics are classified into three categories: the plan aspect ratio of buildings λ P, the vegetation plan aspect ratio λ V (gardens, lawns, trees, and also bare soil), and the plan aspect ratio of impervious and paved surfaces λ I (streets, pavements, car parks, but not the buildings). For the builtup surfaces, a number of more sophisticated three-dimensional morphometric parameters have been calculated according to Grimmond and Oke (1999b). Calculations were done with a detailed three-dimensional raster model of the roof and building geometry provided by the city authorities (Grundbuch- und Vermessungsamt Basel-Stadt). The model does not include vegetation. Mean building height z H, complete aspect ratio λ C (Voogt and Oke, 1997) which describes the surface enlargement due to the three-dimensional surface compared with a plane and frontal aspect ratio λ F of the buildings (averaged over all wind directions) are calculated. ψ s0 is the average sky view factor at ground level, including all open spaces (i.e. streets, parks, backyards), and was derived for each raster element by taking local horizon angles into account. For an idealized infinitely long canyon, the average ψ s0 in the canyon is related to the commonly used canyon height-to-width ratio λ s by λ s = 0.5 tan[cos 1 (ψ s0 )] (Oke, 1981). λ s values are given for comparison with previous studies. They were retrieved using the above relation and, therefore, do not represent the aspects of a single street canyon but incorporate all open spaces. 2.2. Schedule of observations Data from three observation periods are presented. First, the summertime Intensive Operation Period (IOP) includes data from all seven sites (two urban, one suburban, three rural). This dataset allows for the detailed comparison of the diurnal variation of urban rural differences during a period of high energy availability. It covers 1 month, between 10 June and 10 July 2002, when all sites (except U3) operated continuously.

URBAN ENERGY AND RADIATION BALANCE 1399 Table I. Surface characteristics of the energy balance sites Site code U1 U2 U3 a S1 R1 R2 R3 Name Basel Sperrstrasse Basel Spalenring Basel Messe Allschwil Grenzach Village Neuf Lange Erlen Land use (UTZ) b Urban Urban Urban Suburban Rural Rural Rural Grassland Agricultural land Grassland Residential 3 4 storey buildings (A1/A2) Tower 32 m tower within and above street canyon Coordinates (WGS-84) Residential/commercial 3 5 storey buildings (A1) 38 m tower on pike roof Newly developed commercial city centre area (Dc1). EC system 2 m over a parking lot on top of a large 26 m building Residential single and row houses, 2 3 storey (Dc3) 16 m tower in vegetated backyard Instrumented radio tower 5 m mast 10 m mast 47 33 57.2 N 47 33 17.6 N 47 33 47.6 N 47 33 19.0 N 47 32 12.0 N 47 37 7.6 N 47 35 32.3 N 07 35 48.8 E 07 34 34.6 E 07 36 2.6 E 07 33 41.5 E 07 40 31.5 E 07 33 27.1 E 07 38 56.9 E Height a.s.l. (m) 255 278 255 277 265 240 275 Plan aspect ratios c = 0.54 = 0.37 = 1.00 e = 0.28 = 0.02 = 0.00 = 0.01 λp λp λploc λp λp λp λp λv = 0.1.6 λv = 0.31 λvloc = 0.00 e λv = 0.53 λv = 0.91 λv = 0.98 λv = 0.94 λi = 0.30 λi = 0.32 λiloc = 0.00 e λi = 0.19 λi = 0.07 λi = 0.02 λi = 0.05 3D morphometric parameters d zh = 14.6 m zh = 12.5 m zh = 18.8 m zh = 7.5 m λf = 0.37 λf = 0.31 λf = 0.25 λf = 0.12 λc = 1.92 λc = 1.75 λc = 1.64 λc = 1.31 ψs0 = 0.36 ψs0 = 0.51 ψs0 = 0.57 ψs0 = 0.62 λs = 1.30 λs = 0.84 λs = 0.72 λs = 0.63 z0ra = 1.73 m z0ra = 1.59 m z0ra = 2.57 m z0ra = 1.28 m Roof materials f 45% tiles, 50% gravel, 5% corrugated iron Building materials Plaster, concrete, brick 70% tiles, 30% gravel 100% concrete e 95% tiles, 5% gravel Plaster, concrete Concrete, glass Plaster, brick a Operated from 24 June to 12 July 2002 only. b Urban terrain zone according to Ellefsen (1991). c λ P (plan aspect ratio of buildings), λv (plan aspect ratio of vegetated surfaces), λi (plan aspect ratio of impervious surfaces) are derived from aerial photographs for a circle of 250 m around the site. d z H (average building/canopy height), λf (frontal aspect ratio averaged over all wind directions), λc (complete aspect ratio), ψs0 (average sky view factor at ground level) and λs (mean street canyon height-to-width ratio). z0ra (roughness length) calculated from the 3D model for a circle of 250 m around the site. z0ra was calculated according to Raupach (1994). e Local values of the roof (25 m radius) are shown. Because of the low sensor height (2 m), measurements do not represent the whole neighbourhood. f Area fractions of roof materials derived from aerial photographs for a circle of 250 m around the site.

1400 A. CHRISTEN AND R. VOGT During this period, the mean solar radiation was 23 MJ day 1 m 2, the mean air temperature was 20 C, and the rainfall was 65 mm (mostly from thunderstorms). The IOP includes 10 clear-sky days and is significantly warmer and slightly drier than the 30 year average (16.8 C and 89 mm respectively). Winds 10 m above z H were on average 2.0 ms 1, and due to a thermal circulation in the Rhine Valley mainly from the sectors west to north (51%, day) and east to south (36%, night). Second, measurements taken during 1 year (September 2001 August 2002) are presented. Data are available for U1, U2 and R3. This period is characterized by an annual mean temperature of 10.7 C and 826 mm of precipitation (30 year averages: 9.0 C and 791 mm). Winds 10 m above z H were 2.1 ms 1 on average, and sectors southwest to northwest (44%) and east to southeast (34%) dominated. Finally, at U2 and R3, several parameters like radiation and temperature humidity profiles are available since 1994. This long-term urban dataset 1994 2002 is used to compare the findings from the IOP and year-long data with climatological values. 2.3. Instrumentation and data processing Table II lists the instrumentation used in this study. All sites provide continuously measured standard meteorological parameters, directly measured turbulent energy flux densities (except R3), radiation, and storage heat flux densities (either directly measured or calculated as the residual term). Some sites are equipped with micrometeorological profiles of temperature humidity sensors and cup anemometers. Temperatures were monitored with ventilated psychrometers mounted 2 m above ground at the rural stations and approximately 5 m above z H at the built-up sites. In the city centre, additional temperature sensors were deployed in street canyons (U1, U2) and backyards (U2). 2.3.1. Radiation measurements. The four components of the radiation balance were measured at all sites. Radiation measurements over the rural surfaces were carried out 2 m above ground, and observations at the built-up sites were at approximately two times z H (Table II). At this height, the urban shortwave reflection and the longwave emission represent a mixed signal of roofs, walls, streets, backyards, and gardens. An exception is the radiation setup at site U3 with the sensor installed at 2 m above a parking lot and which represents a rather extreme surface of 100% concrete. All radiation instruments were checked and recalibrated against reference instruments after the experiment during two field intercomparisons. The Eppley PIR pyrgeometers were corrected according to Philipona et al. (1995) and the longwave signals from the Kipp & Zonen CNR1 were corrected for shortwave sensitivity. Periods with dew on the sensors a phenomenon uniquely observed at the rural sites were detected using the differences between dew point and case temperature and masked out by visual inspection. The resulting missing data were linearly interpolated for gaps shorter than 2 h. Gaps with a longer duration are homogenized with the nearest site of similar land use. 2.3.2. Turbulent energy flux density measurements. Sensible heat flux density Q H and latent heat flux density Q E are directly derived from eddy correlation measurements using three-dimensional ultrasonic anemometer thermometers combined with humidity fluctuation measurements (Table II). All instruments were mounted vertically and no run-to-run streamline rotation was applied. Q H and Q E are calculated from block averages of 20 Hz raw data averaged over 1 h. Flow distortion effects of the three-dimensional ultrasonic anemometer thermometers were minimized by applying a matrix derived from wind tunnel investigations for each individual instrument. All instruments were checked and outputs compared with each other in the wind tunnel (except the instrument at U3). Q H was calculated from the covariance of acoustic temperature and vertical wind w T s, which was corrected for crosswind either internally by the sensor electronics or during post-processing and for spectral loss (Moore, 1986). Additionally, Q H is corrected for humidity effects (Schotanus et al., 1983). This humidity correction reduces the magnitude of the raw w T s by 3% at the urban sites and by 13% at the rural sites, because the rural sites have a higher evapotranspiration. Both the crosswind correction and the spectral correction are of minor importance, with less than 1% influence at all sites. The box view of the urban surface simplifies the storage by enclosing all surface elements (ground, buildings, and vegetation), but it also incorporates the air volume between the ground and the measurement

URBAN ENERGY AND RADIATION BALANCE 1401 Table II. Instrumentation of the energy balance sites Site U1 U2 U3 S1 R1 R2 R3 Net radiation Q 1 Net radiometer 2 K&Z CM11 1 Net radiometer 1 Net radiometer 2 K&Z CM11 1 Net radiometer 1 Net radiometer K&Z a CNR1 2 Eppley PIR K&Z CNR1 K&Z CNR1 2 Eppley PIR K&Z CNR1 K&Z CNR1 z = 31.7 m z = 32.9 m z = 2 m above roof z = 15.1 m z = 1.4 m z = 2m z = 2m z/zh = 2.2 z/zh = 2.6 z/zh = 2.0 Sensible heat flux QH Sonic Sonic Sonic Sonic Sonic Sonic Gill HS Metek USA-1 Campbell CSAT 3 Metek USA-1 Metek USA-1 Campbell CSAT 3 100/20 Hz b 40/20 Hz 60/20 Hz 40/20 Hz 20/20 Hz 60/10 Hz z = 31.7 m z = 29.9 m z = 2.2 m above roof z = 15.8 m z = 28 m z = 3.3 m z/zh = 2.2 z/zh = 2.4 z/zh = 2.1 Latent heat flux QE Campbell KH2O Campbell KH2O Storage heat flux QS Vertical divergence of sensible heat flux QH/ z Campbell KH2O Campbell KH2O Campbell KH2O Campbell KH2O 20 Hz 20 Hz 10 Hz 20 Hz 20 Hz 10 Hz z = 31.7 m z = 29.9 m z = 2.2 m above roof z = 15.8 m z = 28 m z = 3.3 m z/zh = 2.2 z/zh = 2.4 z/zh = 2.1 Residual term Residual term Residual term and 2 experimental heat flux plates 6 levels of sonics at z/zh = 0.25, 0.77, 1.01, 1.23, 1.53, 2.17 6 levels of sonics at z/zh = 0.45, 1.11, 1.33, 1.74, 2.39, 3.01 Residual term 3 heat flux plates (depth: 5 cm), 4 soil themistors 3 levels of sonics at z/zh = 1.11, 1.61, 2.11 3 heat flux plates (depth: 4 cm), 4 soil themistors 3 heat flux plates (depth: 7 cm), 4 soil themistors a Kipp & Zonen. b Internal sampling rate/output rate.

1402 A. CHRISTEN AND R. VOGT level, within which a small amount of energy, Q T, can be stored (removed) by warming (cooling) the air. This energy change is not part of Q S and not a component of the surface energy balance. It is a consequence of our concept of an elevated surface when measuring flux densities at z m and not at ground level. It is assumed that Q T is mainly driven by sensible heat flux density divergence Q H / z from ground up to z m rather than other effects (see Section 4.5). Therefore, in order to reduce the surface down to a theoretically flat ground level, Q T is incorporated into Q H. Most of the time, Q T is below 10% of Q S and its magnitude is always less than 20 W m 2. Moreover, Q T nearly vanishes when calculating daily totals and, therefore, does not affect the long-term energy partitioning. Q E is calculated from the covariance of absolute humidity and vertical wind w ρ v including a correction for O 2 -sensitivity (Tanner et al., 1993) and a small vertical wind component (WPL-correction; Webb et al., 1980). Furthermore, a spectral correction was calculated taking into account sensor separation (Moore, 1986). The O 2 -correction is most pronounced at the urban sites, increasing the magnitude of w ρ v by 12%, compared with the much smaller influence at the rural sites (+1%). The large urban correction term is because of extremely high Bowen ratios β = Q H /Q E. Similarly, the WPL-correction shows a higher relative impact on Q E at the urban sites (increase of Q E by 25%) compared with rural sites (+2%). Finally, the spectral correction is largest when sensor separation is large and measurement height is low. The average impact on Q E ranges between +2% (U1, U2) and +7% (R2). 2.3.3. Storage and anthropogenic heat flux densities. At the rural sites, the ground heat flux density was measured directly by means of three heat flux plates inserted between 3 and 5 cm depth. It was corrected for flux density divergence in the soil layer above the plates using measured soil temperatures. This storage heat flux density is denoted Q S but neglects storage in the vegetation. Q S of an urban surface incorporates all storage into artificial surfaces (streets, buildings), into the urban vegetation, and into the ground. A residual term of the energy balance can be calculated for each averaging interval. Note, that these values of Q S must be treated carefully because the residual term incorporates all instrumental and methodological uncertainties of the other flux densities. At the parking-lot site (U3), two experimental heat flux plates were placed directly on the horizontal concrete surface. The anthropogenic heat flux density Q F is estimated to be 5 W m 2 in the suburban neighbourhood and 20 W m 2 in the city centre without taking variation in time into account. See Section 4.4 for details on the determination and discussion of Q F. Most energy balance measurements over natural and agricultural surfaces that directly measure all flux densities show that Q + Q H + Q E + Q S is not zero. The remaining gap in the closure is of the order of 20% of Q (Wilson et al., 2002). In the present study, the rural sites with direct measurement of Q S show an average daytime closure gap of 17% (R1) and 18% (R2). During nighttime, the gap is typically around 30%. It is most likely that the two turbulent flux densities are underestimated. To calculate budgets, Q H and Q E at the rural sites R1 and R2 have, therefore, been increased with a constant β in order to close the energy balance and to be able to relate the results to R3 where classical profile methods are used. It is obvious that whenever Q S is determined as a residual term (U1, U2, S1) there is no closure gap by definition. However, there is no reason why this closure gap (underestimation of turbulent fluxes) would not also be present over urban areas, if we were able to measure Q S directly. If Q S is determined as a residual term, then a possible underestimation of Q H and Q E enlarges the residual term, and hence Q S. Any Q S calculated as a residual term must be interpreted as an upper limit. 3. SURFACE RADIATION BUDGET 3.1. Shortwave irradiance Table III summarizes the daily total of all radiation flux densities for the summertime IOP period. The solar irradiance K during this period is nearly of similar magnitude at all urban and rural sites. This is in contrast to earlier urban climate studies, which concluded that K is significantly attenuated in the city due

URBAN ENERGY AND RADIATION BALANCE 1403 Table III. Average daily totals of all radiation fluxes for the summertime IOP (10 June to 10 July 2002) Site U1 U2 U3 a S1 R1 R2 R3 Surface (materials) within the FOV of downward looking sensors and approx. radius of the FOV (m) Urban mix (roofs, walls, street) 100 m Urban mix (roofs, walls backyard) 100 m Concrete roof 5 m Suburban (roofs, walls, gardens) 50 m Grassland (30 cm, 100% coverage) 5 m Agriculture (20 cm, 50% coverage) 5 m Grassland (15 cm, 100% coverage) 5 m Shortwave K (MJ m 2 day 1 ) +22.8 +22.7 +22.7 +22.6 +22.9 +22.8 +22.9 K (MJ m 2 day 1 ) 2.4 2.6 7.1 3.0 4.8 4.5 4.5 K (MJ m 2 day 1 ) +20.4 +20.1 +15.6 +19.6 +18.1 +18.4 +18.4 Longwave L (MJ m 2 day 1 ) +30.7 +30.6 +30.7 +30.5 +30.6 +30.8 +30.9 L (MJ m 2 day 1 ) 38.4 37.7 39.8 37.9 37.9 36.8 36.7 L (MJ m 2 day 1 ) 7.7 7.1 9.1 7.4 7.3 6.0 5.8 Net radiation Q (MJ m 2 day 1 ) +12.6 +13.0 +6.4 +12.3 +10.8 +12.3 +12.6 Albedo b α (%) 10.4 10.9 31.7 13.1 21.9 20.2 19.5 a Only 24 June to 10 July 2002. b Median albedo for K > 50 W m 2.

1404 A. CHRISTEN AND R. VOGT z/z H = 0 Sun Elevation ( ) Sun Elevation ( ) Figure 2. (a) Relative amount of the shortwave input radiation into different height layers of the urban canopy at U1 as a function of sun elevation. Values are normalized to the shortwave input on a horizontal plane assuming similar sun elevation. The partitioning was calculated from the three-dimensional model with 1 m resolution and for a box of 500 500 m 2 around the tower location. (b) Angular variation of the average measured albedo for one urban (U1), the suburban (S1) and a rural surface (R1, grassland) to higher aerosol concentrations (Landsberg, 1981). This effect, however, is not found in the mid-size city of Basel. It may be masked by instrument resolution and the fact that all rural reference sites are located close to the urban core. At the built-up sites, K is not equally distributed within individual layers of the urban canopy. Shading and exposure of the buildings and structures greatly influence its vertical distribution and, therefore, a vertical divergence of the shortwave radiation flux density K / z is observed. Figure 2(a) illustrates K / z within the urban canopy based on calculations with the 1 m digital building model of the urban canopy around U1. The curves show the shortwave irradiance reaching different layers of the canopy under different sun elevation angles relative to the irradiance reaching a horizontal plane under the same sun elevation angle. The calculation neglects reflections and diffuse radiation. Under realistic sun elevations around 30, only about 15% of the incoming solar radiation reaches the ground unmodified as direct K. Some 50% is absorbed and reflected by building parts above z H. With increasing sun elevation, more radiance penetrates to the ground level directly, e.g. 30% at 60. The vertical divergence K / z within the urban canopy determines the locations where energy is available and exchanged, and therefore the magnitude and the partitioning of the urban energy flux densities (also see Section 4.5). 3.2. Albedo Most urban surfaces have a significantly lower magnitude of K compared with rural sites and, therefore, a higher shortwave input K. The observed mean albedo values in the city centre (U1, U2) are around 10%; rural values are all around 20% (Table III). The parking lot (U3) is an exception. The very high value of 32% is not representative of large urban areas because its spatial distribution is limited, as aerial photographs and satellite images show. The field of view of the downward-facing sensor at U3 includes only a concrete surface, whereas the other urban measurements integrate over larger areas (Table III). At U1, U2, and S1 the large field of view includes different surface materials, and the complex morphometric configuration (orientations, density, height) results in multiple reflections and shading, which all lower the reflectivity of the surface. The albedo measured in the city centre is significantly lower than values applied in numerical models, which are typically of the order of 15 to 20% for residential neighbourhoods (Sailor and Fan, 2002), suggesting that dense European city centres are better absorbers of K than most North American city surfaces. Comparably low values of 8% were recently reported from a dense urban canopy in Lodz, Poland (Offerle et al., 2003a) and from the city centre of Marseilles, France (Lemonsu et al., 2004). In the long-term mean, the lower albedo of the city leads to a surplus in K of the order of 0.44 GJ year 1 m 2 compared with rural surfaces. K is the most strongly modified term of all four radiation components (Table IV).

URBAN ENERGY AND RADIATION BALANCE 1405 Table IV. Yearly totals of the radiation balance components for the full-year period (September 2001 August 2002) Urban (U1, U2) Rural (R3) Difference (U1,U2) (R3) GJ year 1 m 2 GJ year 1 m 2 GJ year 1 m 2 Wm 2 K +4.17 +4.17 +0.00 +0.1 K 0.47 0.91 +0.44 +14.0 K +3.70 +3.25 +0.44 +14.1 L +10.05 +10.16 0.12 3.7 L 11.94 11.61 0.33 10.5 L 1.89 1.44 0.45 14.2 Q +1.81 +1.81 0.00 0.1 α a (%) 11.4 21.7 α (snow-free conditions) a (%) 10.7 19.7 a Situations with K > 50 W m 2 only. The present study suggests that the albedo decreases with increasing height and density of buildings. The mean summertime albedo is highest at the suburban site (13.1%), lower at U2 (10.9%) and again slightly lower at the most compact urban canopy U1 (10.4%) (Table III). 3.2.1. Daily variations. Similar to most natural surfaces, the urban albedo also shows a dependence on sun elevation. Figure 2(b) illustrates this angular dependence for the city centre (U1), suburban (S1) and one of the rural (R1) surfaces. For the urban and suburban surfaces the angular dependence becomes important when sun elevation is below 20, due to the highly directional reflectance of horizontal surfaces. In contrast to physical models (Li et al., 1995) and observations over vegetated surfaces (R1), the albedo of both urban and suburban surfaces is fairly constant between 20 and 65 sun elevation under both clear-sky and overcast conditions. No difference between clear and overcast conditions is observed for the urban surfaces (Figure 2(b)). This is in contrast to plant canopies, where clear-sky situations, in general, increase albedo (e.g. R3 at low sun elevation angles). 3.2.2. Annual variations. The long-term measurements at U2 (1996 2002) show little monthly variation of urban albedo values during snow-free conditions. In the long-term average, the albedo at U2 is 10.7% (snow excluded). All monthly averages lie within 2%. The urban albedo increases slightly throughout the summer, from an average value of 10.2% in March to 11.8% in October (not shown). This small increase can be mainly attributed to urban trees, since similar increases (but higher in magnitude) are observed over forests (Lehn, 1991). In general, changes in the urban albedo due to different sun elevation angles are more important than monthly variations due to phenological influences. The most dramatic urban rural difference K U R is observed in winter during periods with snow cover. Figure 3(a) shows data from one clear day with a 20 cm snow cover in the rural area. The midday albedo values in the city centre are around 17% (U1) and 15% (U2, not shown). These values are surprisingly low compared with the simultaneously measured rural value of 70% (R3). The snow-free vertical walls, shading, faster snow melt, and removal (road maintenance) all work to reduce the impact of snow on the urban albedo. 3.3. Longwave flux densities Urban L values are higher in magnitude than L measured over rural surfaces most of the time (Table III). This implies a higher urban radiation surface temperature T s and/or a different emissivity. Moreover, radiation trapping in street canyons affects L significantly. Urban L values are slightly lower than rural values all through the day (Table III). Air masses close to the sensor at the urban sites are drier than the air masses measured over rural surfaces, resulting in a reduction of L. However, L is also affected by aerosol content and boundary layer temperatures, which are

1406 A. CHRISTEN AND R. VOGT Figure 3. (a) Net radiation Q (thick lines) and albedo (thin lines) during a clear winter day with a 20 cm snow cover. Solid lines are measurements over the urban surface (U1); dashed dotted lines are simultaneously measured rural values (R3). (b) Urban rural differences in air temperature T U R and absolute humidity a U R. Temperature and humidity values are an average of U1 and U2 (urban) and R1, R2 and R3 (rural) over the summertime IOP from 10 June to 10 July 2002. Solid lines are measurements over the urban surface (U1); dashed dotted lines are simultaneously measured rural values (R3). Roof-level temperatures are measured 5 m above z H ; street-level temperatures are from instruments operated inside street canyons at U1 and U2 (2 3 m above ground) both supposed to enhance L in urban areas. Note that L also differs between rural sites, and the observed urban rural differences are in the order of the instrumental errors. Measurements of L are surely affected by larger uncertainties than the shortwave irradiance. 3.3.1. Daily variations. In the diurnal course, urban rural differences L U R are strongest in the evening, when the city emits much more longwave radiation than the rural surroundings, and L U R reaches values of about 20Wm 2 (Figure 4(c)). L U R decreases during night to about 10 W m 2 in the early morning, primarily an effect of the different cooling rates due to radiation trapping in street canyons. The intensity of the nocturnal heat island displayed as air temperature (Figure 3(b)) shows a close relationship with L, i.e. the highest urban heat island intensity ( +3 K) is observed just after sunset, with continuously decreasing values during the night. The lower daytime urban rural differences in air temperature again correspond well to the observed daytime L U R, which reaches its smallest values with only 5 Wm 2 around noon (Figure 4(c)). Note that the measurement height of urban temperatures significantly influences urban rural temperature differences. At street level, the heat island is found all through the day, but at roof level during midday slightly cooler air temperatures are measured compared with the rural surroundings (Figure 3(b)). This (roof-level) urban cool island is especially prominent during summer ( 0.5 K) and is also reported from other studies (e.g. Unwin, 1980; Jauregui et al., 1992). The summertime absolute humidity content in the city centre (U1, U2) is around 0.6 gm 3 less than the humidity at the rural sites (R1, R2, R3). This urban dry island is most pronounced in the evening and almost disappears in the early morning (Figure 3(b)). The influence of the drier urban atmosphere on L is not negligible. This humidity difference corresponds to a reduction of the total water vapour content by about 10%. The urban dry island is also shown in the long-term 1994 2003 data, where the average daytime humidity difference U1 R3 is 0.5 gm 3 ( 6% of the total water content). 3.3.2. Annual variations. In the yearly total, the city centre at U1 and U2 loses 0.33 GJ m 2 year 1 more energy through L than the rural reference R3. The energy loss through L is slightly higher in summer (Figure 4(d)). The observed urban reduction of L is of the order of 0.12 GJ m 2 year 1 (Table IV).

URBAN ENERGY AND RADIATION BALANCE 1407 Figure 4. Daily variation (a) and annual variation (b) of the radiation balance components for the period September 2001 to August 2002 at U2. Differences of radiation components between the dense urban surfaces (U1, U2) and one rural reference site (R3) in the daily variation (c) and annual variation (d) for the same period. Negative values indicate components with a relative energy loss from the city; positive terms are components where the city can achieve an energy surplus compared with the rural reference 3.4. Net all-wave radiation Q The larger shortwave energy input K of the urban surfaces due to the lower albedo is mostly offset by larger L loss. This results in a more or less equal daily total of the net all-wave radiation Q over the urban, suburban and rural surfaces, except at the parking lot U3 (Table III). In the yearly average (day and night), Q U R even vanishes (Table IV). A satellite image analysis of the city of Basel supports the current observations that Q of rural and urban surfaces are very similar (Parlow, 2003). 3.4.1. Daily variations. During daytime, Q U R is greater and positive (Figure 4(c)), i.e. the city centre gains more energy compared with rural sites, an effect mainly controlled by the low urban albedo (Section 3.2). The midday Q U R is typically around +40Wm 2. Throughout the night, on the other hand, the city loses more energy through Q than any of the rural sites ( Q U R 15 W m 2 ). During nighttime, the higher longwave emission and the partly drier atmosphere in the city enhance the loss (Section 3.3). 3.4.2. Annual variations. The yearly variations show little difference between urban and rural sites. On average, daily totals of Q lie between ±0.5 MJday 1 m 2 (Figure 4(d)). Neglecting snow conditions,

1408 A. CHRISTEN AND R. VOGT wintertime daily Q totals are slightly more negative than rural ones, i.e. the long nights cause the city to lose more radiation than it can gain due to its lower albedo during the shorter daylight period. The highest daytime differences Q U R are observed in winter, when the contrasting albedo of snowcovered rural surfaces and the darker urban surfaces dramatically increase K U R, and hence Q U R up to 200 W m 2 can be measured (Figure 3(a)). The sample day with snow cover illustrates that Q at the urban site is positive most of the day, compared with the values of the snow covered rural surface, which is either negative or around zero. The enhanced Q accelerates urban snowmelt (Todhunter et al., 1992; Semadeni-Davies et al., 2001). 4. SURFACE ENERGY BUDGET 4.1. Latent heat flux densities Mid-latitude cities with negligible irrigation show less evapotranspiration than their rural surroundings, since Q E is mainly driven by vegetation, and in cities vegetation covers only a small fraction of the surface. Additionally, a faster run-off in the built-up areas lowers the water availability. Therefore, it is not surprising that the city centre (U1, U2) with its low vegetation aspect ratio λ V, its large impervious surfaces, and its negligible irrigation shows small Q E values (Table V). 4.1.1. Daily variations. Figure 5 shows the average diurnal course of all energy balance components for each site measured during the summertime IOP. All weather conditions from clear to completely overcast and rainy days are included. The average daytime partitioning during the IOP is illustrated in Figure 6. The triangle shows the partitioning of Q into Q H, Q E and Q S. The ratios Q H /Q, Q E /Q, and Q S /Q are useful parameters for the detection of diurnal trends in the partitioning of Q and for comparing situations with different magnitudes of Q forcing. During daytime, the magnitude of Q E in the city centre is around 20% of Q (IOP). With increasing green space, Q E becomes more important. The magnitude of the simultaneously measured Q E at the suburban site is 30% of Q and about 60% of Q at the rural sites (Table V). The relationship between daytime Q E /Q and the vegetation aspect ratio λ V is illustrated in Figure 7(a). It is not surprising that vegetation in the urban environment significantly enhances daytime Q E and reduces Q H, a result also supported by other studies where simultaneous measurements were carried out in neighbourhoods with different λ V (Grimmond et al., 1996). λ V can be easily retrieved from aerial photographs and satellite pictures and is, therefore, a useful surface parameter to estimate the daytime Q E. Figure 7(b) illustrates the average daytime Bowen ratio β = Q H /Q E during the summertime IOP as a function of the vegetation aspect ratio λ V. Typical daytime values of β are around 2.5 at urban sites and 0.5 over rural surfaces. The measurement values also suggest that the Bowen ratio can be approximated as a function of λ V. This has been formulated in Equation (3). The relationship is shown in Figure 7(b) (dashed curve). It simplifies the response of a patchy urban surface to a linear superposition of the rural Bowen ratio β rural weighted by the fraction of vegetated surfaces λ V and a hypothetical Bowen ratio for a completely impervious surface β imp weighted by the fraction of impervious surfaces (1 λ V = λ I + λ P ): β(λ v ) = 1 λ v λ v k + k + β 1 rural 1 with k = β imp β rural + 1 and β imp >β rural (3) β(λ V ) is the average daytime Bowen ratio for a given vegetation aspect ratio λ V, i.e. for a particular urban or suburban neighbourhood. β rural is the (known) Bowen ratio over grassland in the rural surroundings of the city (λ V = 1). Because β imp is usually not available, a global parameter k between zero and one is introduced, which is valid for the whole rural urban region. k may depend on various factors, like climatic setting of the city, precipitation, phenology, and the difference between rural and urban discharge coefficients. This empirical relationship works well for Basel, where the long-term value of k is surprisingly constant around 0.2 (see below). This would certainly cause problems in cities with extensive irrigation.

URBAN ENERGY AND RADIATION BALANCE 1409 Table V. Average values of the daytime (nocturnal) hours during the IOP. Daytime values are averaged from 1100 to 1500 CET, nocturnal values from 2200 to 0400 CET. Sites are sorted according to increasing plan aspect ratio of buildings λp. Positive (negative) fluxes are directed towards (away) from the surface. At the sites where all components of the energy balance are measured directly, any missing energy (i.e. the gap to close the balance (Section 2.3.3)) was used to increase the two turbulent flux densities slightly, and force the closure without changing β Q (Wm 2 ) QH (W m 2 ) QE (W m 2 ) QS (W m 2 ) QF (W m 2 ) QH/Q QE/Q QS/Q β R1 +423 ( 57) 101 (+18) 260 (+13) 62 (+26) 0.24 ( 0.31) 0.61 ( 0.23) 0.15 ( 0.46) +0.39 (+1.35) R2 +443 ( 45) 123 (+4) 251 (+9) 69 (+32) 0.28 ( 0.09) 0.57 ( 0.20) 0.15 ( 0.71) +0.49 (+0.45) R3 +455 ( 41) 122 (+12) b 282 (+8) b 51 (+21) 0.27 ( 0.29) 0.62 ( 0.20) 0.11 ( 0.50) +0.43 (+1.43) S1 +453 ( 56) 168 (+7) 134 ( 9) 153 (+53) e + 5 d 0.37 ( 0.13) 0.30 (+0.16) 0.34 ( 0.95) +1.23 ( 0.78) U2 +481 ( 62) 228 ( 8) 100 ( 4) 163 (+64) e +10 c 0.47 (+0.13) 0.21 (+0.06) 0.34 ( 1.03) +2.28 (+1.94) U1 +482 ( 65) 230 ( 23) 88 ( 13) 184 (+80) e +20 c 0.48 (+0.35) 0.18 (+0.19) 0.38 ( 1.23) +2.62 (+1.80) U3 a +322 ( 81) 193 ( 10) 45 ( 5) 104 (+76) +20 d 0.60 (+0.13) 0.14 (+0.06) 0.32 ( 0.94) +4.27 (+2.08) a Only operated from 24 June to 10 July 2002. b Values determined by profile method. c Constant anthropogenic heat flux Q F determined according to Section 4.4. d Constant anthropogenic heat flux Q F estimated. e Values determined as residual term of the energy balance equation.

1410 A. CHRISTEN AND R. VOGT Figure 5. Ensemble diurnal course of energy balance at three urban (U1 U3), one suburban (S1) and two rural (R1, R2) sites during the IOP (10 June to 10 July 2002). U3 has only been operated for part of the IOP (24 July to 12 June 2002) Site U2 is drier than the vegetation aspect ratio suggests (Figure 7(b)). This is because the relationship does not accurately represent the forcing or because either β or λ V are erroneous. Indeed, a large city park can be found 150 m to the south and southwest of the tower and increases λ V of that particular neighbourhood by nearly 10%. This wind direction is rarely observed (<7%) and, therefore, does not affect the overall moisture availability. If the flow is from that particular wind direction, then β = 1.5 is measured compared with the average value of β = 2.3 when taking all wind directions into account. The rural surfaces show increasing evapotranspiration towards the evening and, therefore, decreasing β throughout the afternoon. In the early morning of the IOP, rural values start at around β 0.5. β decreases continuously throughout the day to β 0.2 at 1700 (Figure 6). This decrease of β is most pronounced in summer (March to October). An increasing vapour pressure deficit, which is caused by the diurnal course of temperature and the growth of the mixing layer and the associated entrainment of drier air from the free troposphere, enhances Q E. At the built-up sites, there is also a small decrease of β, but the overall partitioning is less affected. In the city centre, β stays close to β = 2 all the day (also see Section 4.2.1). Nocturnal values of Q E in the city centre are generally low. Q E is directed upward most of the time (Figure 8(c)). Exceptions are the early morning hours, when Q E can be directed downward. 4.1.2. Annual variations. Figure 9(b) illustrates the monthly averages over 6 years of daytime (1100 1500) urban and rural β. Here, β is calculated from vertical gradients of potential temperature, θ/ z, and specific humidity, q/ z. The temperature and humidity gradients were continuously measured at an urban (U2) and a rural (R3) site. Both sites show a nearly similar yearly course, but with distinctly different magnitudes. The higher precipitation during summer, together with an increased transpiration activity, lower the summertime β of both urban and rural surfaces. On the other hand, when vegetation activity is low in winter and spring, values are generally higher at both sites. The highest values are measured in March, when available energy is large (6 MJ day 1 m 2 ) but precipitation is moderate and vegetation activity still low. A similar annual