Quantification of turbulent heat fluxes for adaptation strategies within urban planning


  • Anja Goldbach,

    Corresponding author
    1. Applied Climatology and Landscape Ecology, Faculty of Biology, University of Duisburg-Essen, Campus Essen, D-45127 Essen, Germany
    • Faculty of Biology, Applied Climatology and Landscape Ecology, University of Duisburg-Essen, Campus Essen, D-45127 Essen, Germany.
    Search for more papers by this author
  • Wilhelm Kuttler

    1. Applied Climatology and Landscape Ecology, Faculty of Biology, University of Duisburg-Essen, Campus Essen, D-45127 Essen, Germany
    Search for more papers by this author


With the objective of quantifying turbulent heat fluxes in areas with various types of urban land use to improve living and environmental conditions through better urban planning, comparative energy balance measurements using the eddy-covariance (EC) technique were conducted in Oberhausen (Germany) between 15 August 2010 and 15 April 2011. The results of this paper show that the sensible heat flux (QH) is 20% higher and that the latent heat flux (QE) is 90% lower at an urban site (URB) compared to a suburban site (SUB). Green spaces in cities counteract the growing thermal stresses on city dwellers and therefore represent a possible urban climate mitigation measure. This positive effect may be reduced or lost entirely during long periods of drought. The Bowen ratio (QH/QE) increased from 0.65 to 2.8 with decreasing soil moisture at the SUB, and the monthly average QE value decreased from 94 W m−2 (August 2010, heavy precipitation) to only 47 W m−2 (March 2011, dry period), providing impressive evidence of this relationship. Considering the aspects of urban climate, the creation of urban green spaces can only be effective if an optimum water supply is provided. Furthermore, additional planning recommendations are given for urban planners within cities located at mid-latitudes, as derived from the measured results. Copyright © 2012 Royal Meteorological Society

1. Introduction

Despite the small share of total land surface on earth that is occupied by urban areas, > 70% of the earth's population is expected to live in these areas by the end of the 21st century (United Nations, 2011). The morphological and thermophysical structures of towns and anthropogenic behaviours affect radiation and energy fluxes in urban areas, creating well-known urban climate effects, such as the urban heat island (Oke, 1982). In the future, increasing amounts of people will be exposed to the negative effects of urban climate, which will be exacerbated by predicted climate change (Schär and Jendritzky, 2004; Stott et al., 2004; Kuttler, 2011a). In regard to urban climate studies, developing adaptation and mitigation strategies tailored to the problem areas and including these strategies in the local planning process is necessary (Georgi and Dimitriou, 2010; Hassaan and Mahmound, 2011; Kuttler, 2011b; Schwarz et al., 2011). Concerning urban climatology, evaporation areas (i.e. green spaces and water bodies) can serve as key role in mitigating summer thermal loads. The positive effects of non-sealed surfaces and urban parks and gardens, especially in reducing temperatures through shading, evapotranspiration and cool-air production, are as well known as the cooling effects of urban water bodies (Saito et al., 1990/1991; Kuttler, 1991; Spronken-Smith and Oke, 1998; Upmanis, 1999; Saaroni and Ziv, 2003; Bongardt, 2006; Robitu et al., 2006; Pearlmutter et al., 2009; Lin et al., 2010; Hassaan and Mahmound, 2011; Peters et al., 2011; Shashua-Bar et al., 2011). In addition, positive effects may be observed in neighbouring developed areas, provided that the green space is of sufficient size, the wind speed is adequate and the buildings on the fringe of the area are appropriate (Eliasson, 1996; Bongardt, 2006; Ng et al., 2012). However, quantifying the effects of such areas is difficult.

A technique for directly measuring atmospheric water vapour transport in the boundary layer is the eddy covariance (EC) method, which has been used for approximately two decades and is well-established (Lee et al., 2004). The first recorded evapotranspiration measurements in urban areas based on this method were performed by Grimmond et al. (1993) in Sacramento (USA) and Grimmond and Oke (1995) as comparative energy balance studies in four North American cities. Many other studies followed, performed especially in cities located at mid-latitudes, such as Christchurch (Spronken-Smith, 2002), Basel (Christen and Vogt, 2004; Rotach et al., 2005) and Łódź (Offerle et al. 2006a, 2006b). Meanwhile, a large number of investigations has been conducted in subtropical cities [e.g., Los Angeles (Grimmond et al., 1996), Mexico City (Barradas et al., 1999), Marseille (Grimmond et al., 2004), Tokyo (Moriwaki and Kanda, 2004), Ouagadougou (Burkina Faso; Offerle et al., 2005), Miami (Newton et al., 2007) and Cairo (Frey et al., 2011)] and in cities located at high latitudes [e.g., Edinburgh (Nemitz et al., 2002) and Helsinki (Vesala et al., 2008)]. Table I provides an overview of more recent EC studies, including measurements of turbulent sensible (QH) and latent (QE) heat fluxes. Most of these studies have focussed on the urban variability of turbulent fluxes during clear and calm summer weather conditions. To investigate seasonal variations, measurements over a period of at least one year, such as those performed in Łódź (Offerle et al., 2006a, 2006b), are necessary.

Table I. Urban/suburban energy balance studies of EC measurement examples (in situ) with respect to QH- and QE-fluxesThumbnail image of
  • QH: turbulent sensible heat flux density; QE: turbulent latent heat flux density; QH maximum: maximum turbulent sensible heat flux density; QE maximim: maximum turbulent latent heat flux density; QH/QE (β): Bowen ratio; Su: Summer; Wi: Winter; Sep: September; Jan: January; n.a.: not available

  • a

    Maximum mean hourly value within the study period.

  • b

    Maximum mean hourly value during clear and calm weather conditions (all days).

  • c

    Daytime mean value during clear and calm weather conditions (all days).

  • d

    Maximum mean hourly value from June to August within an urban (vegetation) sector.

  • e

    Maximum mean hourly value from May to June (during wintertime).

  • f

    Daytime mean value from 1100 to 1500 CET.

  • g

    Daily mean value in July (February).

  • h

    Maximum mean hourly value during clear and calm weather conditions in the summer months (in the winter months).

  • The objective of this study is to quantify the evapotranspiration in areas with various types of urban land use in respect to seasonal and diurnal variations, aiming to use enhanced adaptation strategies within urban planning for predicted climate changes to improve the environmental conditions of urban areas located at mid-latitudes.

    2. Material and methods

    In the city area of Oberhausen (North Rhine-Westphalia (NRW), Germany, 77 km2, 213 000 inhabitants; City of Oberhausen, 2011), comparative energy balance measurements (QH, QE) were recorded between 15 August 2010 and 15 April 2011 at an impervious urban site and at a suburban site in a green area near a river (cf. Figures 1 and 2).

    Figure 1.

    Classification of land use types in the city of Oberhausen, NRW, Germany. The urban (URB) and suburban (SUB) measurement sites are marked with black crosses (data basis: RVR, 1998, modified). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

    Figure 2.

    Land use classification within a radius of 1,000 m (top, data basis: RVR, 1998, modified), aerial photographs of the surrounding area at the measured sites (centre, Microsoft Corporation, 2011) and measurement setups (bottom) at the URB (left) and SUB (right). The crosses indicate the locations of the measured sites

    2.1. Study site characteristics

    At the urban station in the city centre of Oberhausen (URB), a 10-m-high telescopic mast was installed on the roof of a hospital (22 m above ground level (agl)) (Figure 2, left). The mean building height zH is 14 m in the surrounding of the station, the plan area density λP = 0.82 and the vegetation area density λV is 0.18, respectively (cf. Table II). A 1,000-m radius around each measurement station served as the reference area (cf. footprint; Section 2.2), because neutral stratification conditions were applied for > 50% of the measurement period (data not shown).

    Table II. Measurement heights and surface characteristics of both sites (radius of 1,000 m)
    Surface characteristicsURBSUB
    • z: measuring height agl; zH: mean building height; z0: aerodynamic roughness length (using the log law); zd: displacement height (zd = 2/3 × zH; according to Grimmond and Oke, 1999); λP: plan area density for all wind directions; λPφ: wind directions, 191° < φ < 326° and 348° < φ < 56°; λV: vegetation area density for all wind directions; λVφ: wind directions, 191° < φ < 326° and 348° < ϕ < 56°; λP: ratio between the horizontal area of buildings and the total horizontal surface area λV: ratio of the horizontal area covered by vegetation to the total horizontal surface area.

    • a

      Only wind directions according to the land use and spectral analyses (191° < φ < 326° and 348° < φ < 56°, cf. Sections 2.1 and 2.4).

    z (m)3230
    zH (m)14.06.2
    z0 (m)1.41.0
    zd (m)9.34.1

    At the SUB, a freestanding telescopic mast with a height of 30 m was installed (Figure 2, right). The flat surrounding area was mainly characterized with evaporation areas (green spaces, water bodies), and the residential buildings (one to three floors) were located in the E to S wind sector (zH = 6.2 m; λP = 0.31; λV = 0.69; Figure 2, right; Table II). A more detailed comparison of the land use structures in the surroundings of the two sites is given in Table III. On the basis of the ‘thermal climate zone’ classification of Stewart and Oke (2009), the sites were assigned to the ‘Old Core’ (URB) and ‘Regular Housing (treed/open)’ (SUB) classes.

    Table III. Fraction of the surface coverage within a radius of 1,000 m at both study sites for all wind directions and also only for wind directions according to the land use and spectral analyses (191° < φ < 326° and 348° < φ < 56°, cf. Sections 2.1 and 2.4)
    Surface coverAll wind directions191.25° < φ < 326.25° and 348.75° < φ < 56.25°
     Traffic areas0.8727.750.5933.53
     Green and agricultural areas0.5417.040.2916.38
     Open/non-built space0.020.680.010.31
     Traffic areas0.5918.900.3318.64
     Green and agricultural areas1.3944.270.9955.90
     Open/non-built space0.051.630.031.86

    A comprehensive analysis of the land use in the surroundings of each measurement station shows that there are only slight differences between λP and λV in specific wind sectors (Figure 2, top). For each 22.5° wind sector in the source area, λP and λV were determined by GIS analyses. To compare the turbulent heat fluxes of the impervious and vegetated areas, only the sectors with a λP of ≥ 0.6 (for URB) and λV of ≥ 0.5 (for SUB) were considered (cf. Table VI; Section 3.3). On this basis, the data for the wind sectors ESE, SE, SSE and S (101–191°) were excluded from the analyses of both sites.

    2.2. Footprint analyses

    In view of the small-scale heterogeneity of the land use in the studied urban area, considering the topographic effects at the measurement sites and the distribution of potential sources and sink areas in the surroundings to obtain representative results is necessary. Footprints were calculated using an analytic model (FSAM; Schmid, 1994), representing the 80% flux source areas for the turbulent heat fluxes recorded at the measurement site under unstable (ζ < − 0.05), neutral (−0.05 < ζ < + 0.05) and stable (ζ > + 0.05) atmospheric conditions (Figure 3). The dimensionless stability parameter, ζ, was calculated using equation ζ = (zzd)/L (L is the Obukhov length in m). Concentric source area demarcations were determined for the different stability conditions (unstable: 500 m, neutral: 1000 m, stable: 2500 m) and used as the basis for further analyses.

    Figure 3.

    Flux source areas (80%, shaded ovals) for instable, neutral and stable atmospheric conditions calculated for 16 wind directions by using the flux source area model (FSAM) of Schmid (1994) for both sites

    The source areas of radiation fluxes (K↓, K↑, L↓, L↑; Section 2.3) are considerably smaller and located around the measurement station with radial symmetry (Schmid, 1997) in comparison to the turbulent fluxes. The two source areas do not cover the same area; however, the average land use distribution is comparable (data not shown).

    2.3. Measurement setup and data processing

    Both turbulent measurement sites were equipped with sonic anemometers (USA-1, Metek, Germany) to sample the horizontal and vertical components of the wind vector (u, v, w) and the acoustic temperature, TS. The molar densities of H2O und CO2 were measured with a non-dispersive infrared open-path analyser (LI 7500, Li-Cor, Biosciences, USA) (Figure 2, bottom). The two instruments were installed 32 m agl (URB), 30 m agl (SUB), respectively, and had a sampling frequency of 10 Hz. The components of the radiation balance were measured at the same height with a radiation balance instrument (CN4, Kipp & Zonen, The Netherlands) with a temporal resolution of 3 min (Figure 2, bottom).

    Thirty-minute averages and covariances were calculated from the raw data as a basis for further data analyses (Aubinet et al., 2000). The data were processed using the TK2 programme of Bayreuth University (Mauder and Foken, 2004). Post-processing included a spike test (Vickers and Mahrt, 1997), double rotation (Kaimal and Finnigan, 1994), sonic temperature correction (Schotanus et al., 1983), density and humidity fluctuation corrections (Webb et al., 1980), spectral correction (Moore, 1986) and crosswind correction (Liu et al., 2001).

    The turbulent heat fluxes are calculated as follows:

    equation image(1)
    equation image(2)

    where QH and QE are the turbulent sensible and latent heat fluxes in W m−2, ρ is the air density in kg m−3, cP represents the specific heat capacity of the air at constant pressure in J (kg K)−1, w′Ts is the kinematic heat flux in (K m) s−1, qmath image represents the specific evaporation heat of water at 20 °C in J kg−1 and w′q′ is the kinematic water vapour flow in (kg m) (kg s)−1.

    The radiation balance, Q*, is defined as:

    equation image(3)

    where K↓ is the global radiation, K↑ represents the short wave reflex radiation (dependent on albedo, α), L↓ is the long wave atmospheric counter radiation and L↑ represents the long wave radiation from the earth's surface (all parameters in W m−2). As the earth's surface is the reference surface for all radiation fluxes, all of the fluxes directed toward the earth's surface are given with a positive sign (energy gain), and all of the fluxes directed towards the atmosphere are denoted with a negative sign (energy loss).

    The (urban) energy balance may therefore be described by

    equation image(4)

    where QF is the anthropogenic heat production and ΔQS is the heat storage flux (all parameters in W m−2). All of the heat fluxes directed from the earth's surface to the atmosphere or into the soil receive a positive sign, and all the fluxes directed towards the earth's surface have a negative sign.

    2.4. Spectral analyses

    The turbulent transport of substances and momentum in the atmospheric boundary layer occurs via turbulence elements of different sizes, each of which makes a contribution to the total turbulence (Stull, 1988; Foken, 2008). Against the backdrop of Taylor's frozen turbulence hypothesis (Kaimal and Finnigan, 1994), the energy spectrum, Sx(f), of an atmospheric time series, x(f), may be determined after applying a fast Fourier transform (FFT) to its autocorrelation function. This application allows the contribution of the individual turbulence elements (frequency ranges) to the variance of the original time series to be identified. The co-spectrum [determined from two different atmospheric time series, x(t) and y(t)] therefore indicates the contribution of a certain frequency range to turbulent flux. As the turbulent energy spectra possess a characteristic structure, deviation from the typical pattern may be seen as a disturbance in the turbulent structure, for example, as a result of the site characteristics (i.e. flow distortion, etc.). Comprehensive analyses of the power spectra (w, T, q) and co-spectra (wTS; wq) of the 12 investigated wind sectors provided a detailed picture. Wind sectors ENE, E and NNW must therefore be excluded because of their disturbed turbulence spectra (in this article, only excerpts for the standardized co-spectrum of wq are presented in Figure 4; further details are to be published elsewhere).

    Figure 4.

    Normalized turbulence co-spectra of w′q (averaged over all of the atmospheric conditions). The predicted power laws (f−4/3) by Kolmogorov (1941) for the co-spectra were also plotted. Due to the land use analyses (cf. Section 2.1), wind directions ESE, SE, SSE and S were rejected prior the turbulence characteristic analyses

    Furthermore, the turbulence structures within the urban boundary layer at both stations are analysed using the drag coefficient and also wind vector standard deviation ratios (σi (i = u, v, w) in m s−1) divided by the friction velocity, u* in m s−1, resp. σi ratio (i = u, v, w) divided by the mean horizontal wind speed, U in m s−1, versus stability ζ (Roth, 2000). The values (data not shown) are comparable to the turbulence measurement results within the neighbouring urban area of Essen (Weber and Kordowski, 2010).

    2.5. Evaluation basis

    Following the covariance and 30-min average determinations, a comprehensive dataset that includes 11,712 values was available for the measurement period stated above. Significant data losses occurred as a result of the increased data quality requirements according to the basic theoretical assumptions for the EC method (quality flags after Foken and Wichura (1996); φ- and u-criterion; cf. Table IV) and of largely unavoidable technical reasons (lens disturbances or fouling of the open path sensors with dust, pollen and/or precipitation (AGC; Active Gain Control from LI-7500), and measurement system failures). To obtain comparable datasets, the data records were filtered using various quality criteria (Table IV, shown on a monthly basis). Only approximately 40% of the data in the two data records, each covering 244 days, were available for evaluation. Following the land use and spectral analyses (φ-criterion; cf. Sections 2.1 and 2.4), 31.9% (URB) and 34.2% (SUB) of the data were rejected. Additional quality testing criteria and occasional measurement sensor or data recording equipment failures resulted in further data losses (cf. Table IV).

    Table IV. Number of eliminated 30-min time blocks within the EC datasets according to their causes
    Month/yearφ-criterionAGCu*-criterionFlag ≥ 7InstrumentationNumber of available 30-min block averagesNumber of available days
    1. φ-criterion: excluded wind directions according to the land use and spectral analyses (191° < φ < 326° and 348° < φ < 56°, cf. Sections 2.1 and 2.4); AGC: sensor fault or pollution of the non-dispersive infrared open path analyser resulting from dust, precipitation and/or pollen; u*-criterion: friction velocity, u* < 0.15 m s−1 (according to Gu et al., 2005); Flag: quality flags according to Foken and Wichura (1996); Instrumentation: technically induced malfunction of the instrumentation. Note: The sum of the single causes is still larger than the total number of causes.

    (a) URB
    Percentage on whole data set (%)31.910.35.614.212.740.0 
    (b) SUB
    Percentage on whole data set (%)34.214.814.926.73.238.5 

    Data losses occurred irrespective of the time of the day (data not shown). In contrast, the adequately developed turbulence test (u*< 0.15 m s−1; Gu et al., 2005) indicates a diurnal or seasonal pattern (also not presented here). The number of rejected data records is highest during low wind weather conditions, meaning that the data recorded during the night are underrepresented.

    3. Results and discussion

    In this section, the differences of radiation and energy budget between the impervious inner city area and the suburban green space are described. Furthermore, additional planning recommendations via enhanced adaptation strategies within cities located at mid-latitudes in regard to climate change predictions are given.

    3.1. Meteorological conditions during the study period

    The representativeness of the 8-month investigation period in climatic terms was verified on the basis of the long-term temperature and precipitation averages (1961–1990) (Figure 5). Only slight deviations between the mean monthly air temperatures and the long-term average occurred. However, December 2010 (approx. − 5 K when compared to the record ranging from 1961 to 1990) and April 2011 (+5 K) were extremely cold and warm months, respectively. In addition, August 2010 was extremely wet (166 mm, 267% of the average long-term monthly precipitation), and notably dry spells occurred in March 2011 (21 mm, 34%) and April 2011 (19 mm, 36%).

    Figure 5.

    Comparison of monthly mean air temperatures and monthly precipitation sums during the measured long time period (reference period: 1961–1990). Air temperatures were measured in the city centre of Oberhausen, and the long-term data were received from meteorological station Duisburg–Laar (DWD, 2011). Precipitation data were obtained from Oberhausen–Buschhausen (Emschergenossenschaft/Lippeverband, 2011). Data basis: 01 August 2010–30 April 2011. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

    During the measured period, the wind direction distributions at the two sites were similar with a maximum from S to SW directions (Figure 6). At the URB, an easterly wind sector was underrepresented, probably due to the shading effects of a 3-m-high lift shaft on the hospital roof in the direct vicinity of the station, among other reasons.

    Figure 6.

    Wind frequency distribution classified according to different wind speeds at the URB (45 m asl, left) and SUB (33 m asl, right). All of the wind data of the measured period are used and binned into 22.5° classes

    The distribution of wind speeds at the two sites was also similar with slightly higher values at the SUB. During the measured period, average wind speeds of approximately 3.4 m s−1 with peak values of 10 m s−1 were measured (data not shown) at both sites in the winter months (December and January).

    3.2. Surface radiation budget

    The mean diurnal course of the radiation balance components varied with the season; however, differences between the sites are minimal (Figure 7). The seasonal variations (magnitude) of radiation balance components are caused by the change of radiation angle and surface coverage throughout the year. The global radiation, K↓, at both sites was nearly equal with slightly higher values at the URB on average (approximately 4 W m−2). These differences are comparable with those in Basel (Christen and Vogt, 2004). Furthermore, the radiation balance components were analysed for days under clear and calm weather conditions (data not shown). The differences between the URB and SUB sites are more significant for all components during these days. For instance, then lower K↓ values occur at the URB; probably due to the air pollution that mainly covers the city centre (Kuttler and Schaefers, 2001).

    Figure 7.

    Average diurnal and seasonal courses of radiation balance fluxes (K↓, K↑, L↓, L↑, Q*, cf. Equation (3), Section 2.3) of the measured period (15 August 2010–15 April 2011) at both sites (cf. Section 2.3 for sign convention)

    Concerning the mean diurnal course of the entire investigation period, the short wave reflex radiation produced lower values at the URB (α, 0.11) than at the SUB (α, 0.18). In contrast to the results received in Basel (Christen and Vogt, 2004), the comparable differences of K↑ in Oberhausen were mostly pronounced in March and April 2011 (as is the opposite in Basel). This result is probably due to the large number of days with snow cover in winter (December 2010) when the α-values of the two stations were similar. In March and April, high K↓ values during long periods of high pressure systems strongly influence the K↑ values because of the differences in the surface coverage of the sites at this time.

    The long wave atmospheric counter radiation, L↓, measured at the two sites was almost identical (maximum difference of ± 9 W m−2, ± 2%) in contrast to the L↑, which was higher at the URB (maximum difference of 32 W m−2, 9%). The mean radiation temperature (calculated using the Stefan–Boltzmann law with an ε of 1, mean L↑ at the URB and SUB), was 9.5 °C and 8.9 °C at the URB and SUB, respectively. Especially for the days with clear and calm weather conditions, greater differences appeared between the sites (max. difference of 58 W m−2, 16%).

    Changes in the radiation budget components are therefore due to the different shares of sealed materials in the surroundings of the stations. Higher L↑ and lower K↑ values at the URB result in a slightly lower radiation balance, Q*, than at the SUB site (differences: maximum − 35 W m−2, − 10%; cf. Figures 8 and 9). These findings were also indicated by the investigations in Lódź (Offerle et al., 2006b), whereas Christen and Vogt (2004) found slightly higher urban Q* values in Basel.

    Figure 8.

    Isoflux diagrams of the daily (x-axis) and annual (y-axis) variations of radiation balance (Q*), turbulent sensible heat flux density (QH) and turbulent latent heat flux desity (QE) at the URB (left) and SUB (right). Note the different parameter scales

    Figure 9.

    Average diurnal courses of turbulent heat flux densities (QH, QE) during days with clear and calm weather conditions (20 August 2010; 20 September 2010; 25 December 2010; 10 January 2011; 22 March 2011; 08 April 2011) at both sites. The days with clear and calm weather conditions are defined according to Dütemeyer (2000)

    3.3. Turbulent heat fluxes

    3.3.1. Diurnal and seasonal variability

    In contrast to the short-term energy balance measurements, the longer-term measurements can indicate seasonal variation in turbulent heat fluxes (Figure 8).

    At both sites, the hourly average QH values reached a daily maximum in the early afternoon (at approximately 1300 CET) with 159 W m−2 (URB) and 127 W m−2 (SUB), respectively; the value at URB was approximately 20% above the value at the SUB. For the city of Essen, Weber and Kordowski (2010) reported QH differences of approximately 30% within the city core (one year of measurements). Contrary to expectations (high K↓ in summer months), the absolute QH maxima at the two sites in Oberhausen were in April 2011 and not in August 2010. This finding was due to the fact that the first half of April was largely dominated by high-radiation stagnant weather conditions, whereas August was characterized by cyclonal weather conditions with low solar radiation and high total precipitation (cf. Section 3.1; DWD, 2010–2011). The maximum values are lower than those reported on the basis of energy balance measurements in other cities (cf. Table I, range: 180–390 W m−2 for urban and 160–275 W m−2 for suburban stations); this difference is due to the fact that August 2010 (lower K↓) and the measured period (Kmax in July) were not representative. As expected, the QH values measured in winter at both stations are lower than those measured in August, with higher QH values being observed at the URB.

    The SUB is characterized by a higher incidence of stable atmospheric conditions (data not shown). As a result of stronger surface cooling, the QH to the earth's surface was intensified (negative sign: 14 h in August, 20 h in December), while only a slightly negative value was observed at the URB, due to the uncoupled surface warming of sealed materials and therefore increasingly unstable conditions of the nocturnal urban atmosphere (negative sign: 8 h (August); 0 h (December); cf. Figures 8 and 9). While Spronken-Smith (2002) and Frey et al. (2011) reported negative nocturnal QH values in urban areas, Christen and Vogt (2004) and Oke et al. (1999) solely found positive nocturnal QH values for such locations.

    The turbulent latent heat flux, QE, is subject to higher diurnal and seasonal variations than the QH (Figure 8). Both sites were subjected to similar diurnal and seasonal courses, however, with considerable differences in amplitude. The absolute daily maximum QE values at both locations occurred at the time of highest radiation (1300 CET), with the value at the URB reaching 123 W m−2 (100%, August 2010), which is approximately 90% lower than the QE value at the SUB (232 W m−2, 188%). Grimmond et al. (2002) compared summer QE daily maximum values for ten North American cities, reporting a range of 20–250 W m−2 between the cities. The measured values reported here are comparable with those of other studies (cf. Table I; value range: 60–110 W m−2 for urban and 100–280 W m−2 for suburban locations), although relatively high notable QE values were recorded at the URB. These values are probably due to the severe rainfall in August 2010, with temperatures deviating slightly from the long-term averages. Evapotranspiration was probably promoted by the high water vapour deficit in the atmosphere. Furthermore, a higher vegetation density within the city centre of Oberhausen compared to the North American cities induced a higher QE value.

    Regarding seasonal variation, the QE values at both stations in the winter were lower than in the summer. The winter daily maxima of 55 W m−2 (URB) and 76 W m−2 (SUB) are comparable to values reported in other studies (cf. Table I). During March and April 2011, the QE values were below the expected levels as a result of inadequate water availability at the two sites. At the SUB, Bowen ratios > 1 actually occurred (cf. Table V; Figure 9).

    Table V. Monthly means of turbulent heat flux densities and Bowen ratios at both study sites
    Month/yearQH/Q*QE/Q*β = QH/QE

    3.3.2. Surface energy partitioning

    The partitioning of turbulent heat fluxes is useful for assessing the type of energy conversion under different radiation conditions. On this basis, it is possible to determine how much energy from the radiation balance, Q*, is used for QH and QE at both sites. Table V displays the monthly values of energy partitioning (QH/Q*, QE/Q*) for the daytime hours (Q*> 0 W m−2) and the Bowen ratio (β) at the two stations.

    At the URB, the QH/Q* quotient is considerably higher for all months and also shows higher variation. The differences between the two sites are less pronounced in August (URB 0.4; SUB 0.26). On this basis, a considerably larger share of solar radiation is transferred from the surface to the atmosphere via QH at the URB (increase in ambient temperature). The large differences in the winter (URB 1.21; SUB 0.30) probably result from the QF. As it is not possible to measure the QF directly, this additional heat production, which is mainly induced by space heating systems, is included in the QH value. A QH/Q* quotient of > 1 at the URB (December 2010 and January 2011) clearly reflects the influence of the QF. An assessment of the spatial and temporal variability of the QF conducted by Ferreira et al. (2011), Pigeon et al. (2007), Sailor and Lu (2004), Ichinose et al. (1999) or Kłysik (1996) has not been performed at present.

    The behaviour of the QE/Q* ratio is inversely proportional to that of the QH/Q* ratio. At the SUB, the QE/Q* ratios for each month are higher than that of the QH/Q* ratio; however, the variation is also wider. This variation most probably depends upon the seasonal changes in surface characteristics (vegetation cycle) and the soil moisture. Because of the limited availability of water at the URB (e.g. no water stored in the soil or use of sewers) the QE value shares to Q* reaches only 35% and 24% in summer and winter, respectively. At the SUB, this component is considerably higher (86%, summer; 27%, winter), especially during the summer. During the winter, lower radiation values, a low saturation deficit in the atmosphere, limited transpiration of vegetation and availability of water (e.g. snow and ground frost) lead to low QE values at both stations. In contrast, the evaporation potential at the SUB is so high in August and September that it cannot be met by the Q* in some cases. To compensate for this deficit in Q*, energy is taken from the surroundings (i.e. the ‘oasis effect’; data not shown; cf., e.g. Taha et al., 1991). Remarkably low QE/Q* values were found at the two sites in March 2011, probably due to extended dry spells.

    During the investigation period, QE contributed an average of 58% to the turbulent atmospheric interchange at the SUB. On the basis of a mean Q* value of 200 W m−2, an average of 116 W m−2 is therefore transferred to the atmosphere via evaporation processes. On the basis of an eight-month measurement period (244 days), a constant specific water evaporation heat at 20 °C (qmath image, 2.4 MJ kg−1) and 12 h of daylight (Q*, > 0 W m−2), the evaporation equivalent was approximately 510 mm/8 months, corresponding to approximately 80% of the precipitation during the measured period (640 mm/8 months). At the URB, only 30% of the Q* value was transferred to the atmosphere by the QE; the evaporation equivalent based on the same assumptions was approximately 260 mm/8 months, corresponding to only 40% of total precipitation.

    When compared to other studies (Table I), a lower QH/Q* ratio and a higher QE/Q* ratio at the SUB was observed. The results presented here are most comparable with those of Balogun et al. (2009) for Kansas City. The QH/Q* ratio at the URB is within the range of results reported from other studies (cf. Table I), while the QE/Q* quotient is significantly higher. Among other factors, these differences are due to the land use structures of the areas surrounding the stations (different λP and λV values) and therefore are also due to the different source areas for turbulent fluxes (cf. Figure 11).

    The mean Bowen ratios (β = QH/QE) are > 1 at the URB and < 1 at the SUB, with the March 2011 values serving as an exception (cf. Figure 9). During spring, the β values do not only rise as a result of increasing solar radiation (K↓), but also because of the fact that vegetation is not yet active, an effect that is especially marked in pronounced dry spells (maximum values in March). High β values in the winter months (December 2010 and January 2011) occur as a consequence of the increase in the QF at constant soil humidity (i.e., snow cover and ground frost). In August and September 2010, lower β values were observed at both stations as a result of the transpiration activity of the vegetation; the β values at the SUB were generally lower (Table V).

    Grimmond and Oke (1995) reported typical β values (Q*, > 0 W m−2) between 0.88 and 1.54 for four inner North American city areas. The β values at the SUB are comparable to those of Balogun et al. (2009) and at the URB to those of Offerle et al. (2006b). Other studies have reported significantly higher β values; this finding might be due to the different λP and λV values in the vicinity of the stations, as well as the differences in measurement durations and meteorological characteristics (global circulation regime).

    3.3.3. Days with clear and calm weather conditions

    With respect to predicting climate change, the number of summer heat waves in Central Europe will most likely increase (Beniston, 2004; Schär et al., 2004; IPCC, 2007). Such conditions as high maximum temperatures and radiation values and long durations have a negative impact on the thermal comfort of city dwellers in the form of thermal stress (Mayer, 1996, 2006). The energy balances of the two sites during low-exchange high-pressure conditions were therefore compared to identify the effects of the different land use structures on the energy budget and on the temperature levels.

    The clear and calm days within the measured period were determined according to Dütemeyer (2000). Only the days with a full data record for both stations after quality filtering were considered. As a result, six days were available. A diurnal plot of the QH and QE values is shown in Figure 9.

    In terms of their development over time, the diurnal plots of turbulent heat fluxes at both stations are comparable. However, differences in the order of a magnitude were observed. Especially around noon and in the afternoon, the QE values at the SUB are considerably higher (MaxSUB, 300 W m−2; MaxURB, 80 W m−2, ≈ 27%). In the winter months, similar relationships were observed but with lower turbulent fluxes. The values measured at the SUB confirm the high evapotranspiration effect of green spaces, thereby leading to a more balanced thermal environment than that at the URB. During all months, the hourly average QH values between sunrise and sunset were significantly higher at the URB (MaxURB, 245 W m−2 (April 2011); MaxSUB, 155 W m−2, ≈ 63% (April 2011)).

    A comparison of turbulent fluxes at the two stations show Bowen ratios when β is > 1 and < 1 at the URB and SUB, respectively, during clear and calm days. However, deviations were observed in March (β, > 1 at the SUB; Figure 9). As a result of below-average precipitation (P ≈ 20 mm month−1; cf. Section 3.1), evaporation was limited, despite a sustained period of high pressure as a result of the lack of humidity in the soil and the seasonally low vegetation transpiration. These data show that green spaces only have high evapotranspiration potential if sufficient water is available; the green spaces can subsequently contribute to thermal comfort within the urban area (Figures 10 and 11).

    Figure 10.

    Mean Bowen ratio (β), which is dependent on the soil moisture [phrased as numbers of days after rainfall (p > 0.1 mm d−1)], over the measuring period at both sites

    Figure 11.

    Mean Bowen ratio β (top) and partition of the turbulent heat flux densities (QH/Q*, QE/Q*; bottom) as a function of the plan area density (λP) during the measured period (daytime: Q*> 0 W m−2). The error bars indicate ± 1σ

    3.3.4. Effect of soil moisture on the Bowen ratio (β)

    Figure 10 compares the relationship between the Bowen ratio (β) and the soil moisture (measured by the number of days after a precipitation event) for the two sites. Within the investigation period, the β values rose sharply (on average from 2.56 to 7.73) with decreasing surface humidity at the URB, as a large part of the energy input was increasingly discharged via QH due to the lack of water (i.e. the increase in ambient temperatures). However, a similar behaviour was observed at the SUB. As the β value rose from 0.65 to 2.8 on average, the evaporation performance of the location was severely reduced by the limited availability of water. The measured values show a reduction in evaporation performance of 50% at the SUB during a sustained dry spell in the investigation period (monthly mean QE value, 94 W m−2 (August 2010, high precipitation); monthly mean QE value, 47 W m−2 (March 2011, dry spell); not considered here). The decrease in soil moisture (confirmed by weekly soil measurements), which could not be compensated by the water transfer from deeper soil horizons (via capillary rise) during the longer dry spells, led to a reduction in the QE and therefore to a rise of β to > 1 as a result of high K↓ values. This result had a detrimental impact on the thermal effectiveness of the location (air temperature increase), as a large part of the energy input was subsequently transferred to the atmosphere via QH. Offerle et al. (2006b) also reported an increased β during longer dry spells in Basel.

    3.3.5. Variability of turbulent fluxes by surface coverage

    The differences in β between urban and rural (suburban) locations were already recognized as a consequence of urbanization at an early stage (Oke, 1987; Kuttler, 1988). The following section therefore considers the effects of the (anthropogenic) land use structure in the station surroundings on the partitioning of turbulent heat fluxes (QH/Q* and QE/Q*) and β values, thereby describing the urbanization effect.

    As already mentioned in Section 2.1, the plan area density (λP) was determined for each wind sector (Table VI). The turbulent heat fluxes were subsequently analysed as a function of wind direction. Figure 11 shows the β values and the QH/Q* and QE/Q* ratios (for Q*> 0 W m−2) as a function of λP.

    Table VI. Mean plan area density (λP: ratio between the horizontal area of buildings and the total horizontal surface area), mean turbulent heat flux density partition (QH/Q* and QE/Q*) and mean Bowen ratio (β), which is dependent on wind direction (binned into 22.5°-classes), within a radius of 1,000 m at both study sites in the evaluation period
     Wind directionλPQH/Q*QE/Q*β = QH/QE
     NE, WSW0.850.600.343.36
     N, WNW0.890.690.304.85

    With rising λP, the β value increases exponentially from 0.42 (λP, 0.06) to 2.23 (λP, 0.96) (Figure 11, top). This behaviour is because of the fact that a positive QE value over sealed surfaces is only possible if moisture has collected on the surface (i.e. precipitation, dew and irrigation); surface water is severely restricted by the drainage of precipitation through a sewage system. In contrast, for surfaces with vegetation or without sealing, water is also supplied from deeper soil horizons (via capillary rise) when there is sufficient precipitation, generally leading to larger positive QE values over longer periods (cf. Figures 8 and 9) and therefore resulting in a lower β value. In addition, the fluctuation range of β at the URB (0.64≤λP≤0.96) is significantly larger than at the SUB (0.06≤λP≤0.50) (Figure 11, top). This result is probably due to the large number of anthropogenic sources in the direct surroundings of the URB, which influences the measurements of turbulent fluxes, especially QH (e.g. emissions from domestic heating systems, road traffic and power stations). A β value of > 1 at a λP of 0.41 (SW wind direction at SUB, cf. Table VI) is probably caused by the influence of the urban air plume, as the centre of Oberhausen is located to the southwest of the SUB (cf. Figure 1).

    The partitioning of turbulent heat fluxes (QH/Q* and QE/Q*) during the daytime hours (Q*, > 0 W m−2) is clearly dependent on the λP (Figure 11, bottom). While the QH/Q* ratio increases exponentially from 0.21 to 0.74 with rising λP, a clear linear decrease in the QE/Q* quotient from 0.91 to 0.27 is observed. The differences between the QH/Q* and QE/Q* quotients rise especially steeply from a λP of > 0.8. This striking result suggests that a reduction in the sealed surface share from approximately 96% to approximately 83% leads to a significant fall in QH (and therefore in air temperature), indicating that thermal conditions in the inner city areas can be improved by creating green spaces (increase in QE/Q*). Assuming an average Q* of 200 W m−2 in the investigation period (Q*, > 0 W m−2), a reduction in plan area density, λP, from 96% (QE/Q*, 0.27) to 83% (QE/Q*, 0.36) in the Oberhausen city area would lead to an increase of approximately 18 W m−2 in QE. This increase would lead to the additional evaporation of approximately 80 mm of water (e.g. precipitation, dew and transpiration) in the investigation period, corresponding to approximately 13% of the precipitation. If this energy were not needed for the evaporation process, approximately 1.5 × 105 m3 of air could be heated by 1 K. This result clearly shows the positive effect of inner city vegetation on thermal conditions as a result of its high evapotranspiration rate with adequate water supply.

    Comparable statements concerning β, QH/Q* and QE/Q* are made in a large number of studies (Grimmond et al., 1996), which conducted energy balance measurements in city districts with different λV values.

    3.4. Consequences for urban planning

    Regarding climate change prediction, greater attention should be paid to the integration of climatology as a component in urban planning processes (Erell et al., 2011). Urban climatologists develop guidelines for adaptation and mitigation strategies, focusing on improving urban climates and intending to facilitate the incorporation of scientific results in regional and local planning practices (MUNLV, 2010). The following planning recommendations for cities located at mid-latitudes that were derived from the results of Chapter 3.3 would be useful for urban planning processes.

    Evaporation areas in cities counteract the growing thermal stresses on urban populations and therefore represent a possible urban climate mitigation measure (Kuttler, 2011b). Evaporation is a notably energy-intensive process, leading to an improvement of the urban area environment through an air temperature reduction (Figures 8, 9, 11; Table V). The high potential of evaporation areas is explained with simple calculated examples in Chapter 3.3. In this context, not only evaporation but also shading has a positive impact on thermal conditions. To consider all of the positive effects, the ‘principle of savannah’ (cf. Spronken-Smith, 1994) should be considered while planning or redesigning of an urban green space. With this consideration, a park should consist of distinct turf, shrubs and trees, as in a savannah with high and shady trees, which are distributed such that a shadow area maximum results throughout the day. However at night (Q*, < 0 W m−2), cooling is not limited by shading from the trees.

    The results also show that longer dry spells have a detrimental impact on the thermal effectiveness of the location (Figure 10). In addition to developing urban green spaces, urban planners must therefore ensure that these areas receive an adequate water supply (Cleugh et al., 2005).

    Furthermore, even small urban green spaces and/or planting of street canyon trees can improve thermal comfort because of their cumulative effects. These facts are demonstrated by an exponential decrease of the QH/Q* ratio after decreasing the plan area density λB (Figure 11, bottom).

    In Germany, the observed shrinking-city phenomenon (Oswalt and Rieniets, 2006) might play a significant role in developing adaptation and mitigation strategies through effective changes in non-sealed surface space regarding thermal comfort.

    The reduction of thermal stress has been investigated only for the green spaces and not for the positive effects in neighbouring built up areas. This fact is dependent on the size of the green area, wind speed and buildings on the fringe of the area (Eliasson, 1996; Bongardt, 2006). Therefore, models and simulations should be integrated into urban planning processes to simulate the effect of land use changes to urban climate. In this context, limitations of building development can also be defined to keep these areas free from urban development, as well as to provide a network of green spaces for ventilation.

    4. Summary and conclusions

    Comparative energy balance measurements using the EC-method were performed in Oberhausen (Germany). Quantifying the evaporation in areas with different urban land use types to improve the living and environmental conditions of urban areas through better adaptation strategies within urban planning for predicted climate change was considered.

    Following comprehensive analyses of the EC-measured data records, only approximately 40% of the data collected during the investigation period (15 August 2010 to 15 April 2011) were available for evaluation (cf. Table IV). The diurnal courses of the radiation balance at the urban (URB) and suburban (SUB) stations were similar (ΔQ*max. = Q*URBQ*SUB = − 35 W m−2, − 10%). At the URB, the sensible heat fluxes (QH) were 20% higher, and the latent heat fluxes (QE) were approximately 90% lower than at the SUB. As a result of global circulation regimes during the investigation period, the maximum QH values at the two sites were measured during a sustained, high-radiation dry spell in April 2011 and not in mid-summer. In contrast, the highest QE values were recorded in August 2010, as a result of high precipitation.

    Inner-city evaporation surfaces (green spaces) counteract thermal stress by shading and evapotranspiration. These surfaces therefore represent a possible urban climate mitigation measure (Kuttler, 2011b). However, the positive effects of evapotranspiration may be reduced or even eliminated during long dry spells. This effect is clearly demonstrated by the increase of QH/QE quotients from 0.65 to 2.8 at the SUB with decreasing soil moisture (Figure 10). For this location, a 50% reduction in evaporation performance during dry spells was also recorded [reduction in the monthly mean QE from 94 W m−2 (August 2010, high precipitation) to only 47 W m−2 (March 2011, dry spell)].

    From an urban climate point of view, aside from the positive effect of shading, inner-city green spaces can only be effective as a measure to improve the urban thermal climate if an adequate water supply is ensured. Especially in dry periods when there is no link to the ground water level; a reduction in soil moisture cannot be offset by water transport from deeper soil layers, so adequate irrigation must be provided for green spaces. Otherwise, the positive thermal effects of green spaces resulting from transpiration will be reduced to a minimum or eliminated entirely, which is confirmed by the measured values.

    Furthermore, additional planning recommendations for urban planners derived from the measured results are given (e.g., the ‘principle of savannah’). Notably, these planning recommendations are mainly helpful for cities located at mid-latitudes.


    The dynaklim project, Dynamic adaptation to the effects of climate change in the Emscher-Lippe region (Ruhr area; www.dynaklim.de), on which this report is based, was funded by the Federal Ministry of Education and Research (BMBF) under grant number 01LR0804G. The authors are responsible for the content of this publication.