1 Introduction

From previous studies about urban vegetation and in particular urban trees, we know that trees produce distinct meteorological and climatological effects at all scales in urban environments (Heisler 1986; Oke 1989; Voogt and Oke 1997). This includes the influence of tree cover on the local-scale surface energy balance (Grimmond et al. 1996) and the reduction of canopy air temperature in comparison to the built environment by evapotranspiration and shadowing. However, the cooling effect differs from site to site and during the course of a day (Souch and Souch 1993; Spronken-Smith and Oke 1998; Upmanis et al. 1998; Shashua-Bar and Hoffman 2000; Potchter et al. 2006; Bowler et al. 2010). Trees reduce surface temperatures of the near surroundings (Hoyano 1988; Robitu et al. 2006; Shashua-Bar et al. 2009; Meier et al. 2010) and the building energy use (Rosenfeld et al. 1995; Akbari et al. 2001), and help to improve human thermal comfort by modifying the human energy balance (Mayer and Höppe 1987; Brown and Gillespie 1990; Streiling and Matzarakis 2003; Thorsson et al. 2004; Gulyás et al. 2006). Trees also modify the wind field in street canyons. This is important in relation to dispersion processes of pollutants (Gromke and Ruck 2007; Litschke and Kuttler 2008).

However, there are only few studies regarding the spatial and temporal variation of canopy temperature of urban trees and their response to the urban environment (Montague and Kjelgren 2004; Mueller and Day 2005). According to previous field studies, urban vegetative evapotranspiration is affected by advection and edge effects, i.e. the oasis or clothesline effect because of strong surface heterogeneity in urban environments (Oke 1979; Hagishima et al. 2007). The experiment of Heilman et al. (1989) shows that presence of walls changes the water consumption of adjacent shrubs. Kjelgren and Montague (1998) reveal that trees over asphalt have higher leaf temperatures than those over natural surfaces. This results in higher water loss or the closing of stomata depending on tree species. A study from Basel, Switzerland, based on thermal-infrared (TIR) imagery obtained during a helicopter flight shows that tree canopy temperatures at noon are species-specific and depend on location, leaf size and stomatal conductance (Leuzinger et al. 2010). There have been no studies which quantify canopy temperature over a wide range of species, sites and atmospheric conditions in an urban environment, beyond those that report temporal dynamics of several individual trees at different sites (e.g. Montague and Kjelgren 2004) or one-time canopy temperature of more than 400 trees (Leuzinger et al. 2010).

The objective of this study is to estimate spatio-temporal variability of canopy temperature (T C) and canopy-to-air temperature difference (∆T C = T C − T a) of individual trees in an urban environment. We aim at quantifying the spatio-temporal patterns of T C and ∆T C in relation to meteorological conditions and tree-specific (physiological) and site-specific characteristics. State-of-the-art TIR cameras mounted on towers or high-rise buildings allow a simultaneous sampling of spatial and temporal changes of canopy temperatures by recording time series of thermal images. We will refer to this as time-sequential thermography (TST, Hoyano et al. 1999). The obtained TST data cover a period of 62 days from 1st July to 31st August 2010. In the long-term perspective, we will collect climate data of species-specific canopy temperature and ∆T C in an urban environment in order to compare different years. In detail, we address the following questions:

  1. 1)

    Are there differences in terms of ∆T C between and within the observed crowns in our study area, and if so, how is the magnitude during a diurnal cycle?

  2. 2)

    To what extent ∆T C varies due to different atmospheric and moisture conditions, species-specific variables as well as site-specific conditions?

In the next section, we describe the study site, our experimental setup, the pre-processing steps of TST data and the tree survey. Section 3 presents the observed spatio-temporal patterns of ∆T C and their relation to meteorological conditions, species-specific variables and site-specific conditions. The “Discussion” section gives answers to the above-mentioned questions under consideration of previous studies.

2 Materials and methods

2.1 Study site and experimental setup

The study was conducted in the city of Berlin, Germany (52°27′N, 13°19′E) in the south-western Steglitz-Zehlendorf district. During the experiment, the site was characterised by a high amount of vegetation, five to six-storey block development with courtyards and mature trees, low-rise residential houses, street trees, one park with mature trees and a playground, villas with gardens and one high-rise building. The study was conducted under the umbrella of the urban climate research program ‘Energy eXchange and Climates of Urban Structures and Environments’ (EXCUSE) that focuses on quantification of energy, momentum and mass exchange processes in the urban boundary layer. The EXCUSE research infrastructure includes a measurement platform on top of the high-rise building and comprises a TIR camera system (7.5–14.0 μm, 320 × 240 pixels, VarioCAM® head, InfraTec GmbH). The fixed field of view (FOV) of the TIR camera covers an area of approximately 0.3 km2. Within the FOV, a meteorological station measures air temperature T a, relative humidity RH (HMP45A, Vaisala) and downward short-wave radiation ↓ E SW (Starpyranometer 8101, Schenk GmbH) at a height of 22.5 m above ground, whereas wind velocity and wind direction measurements are conducted at a height of 23 m above ground using a cup anemometer and vane (Lambrecht GmbH). A tipping-bucket rain gauge measures precipitation at 0.1-mm resolution. The position of the meteorological station is depicted in Fig. 1.

Fig. 1
figure 1

The field of view (FOV) and viewing direction (northwest) of the TIR camera. The symbols indicate tree genus. Numerals are used to indicate individual trees in the text

Full size image

2.2 Thermal remote sensing data

The TIR camera was oriented towards northwest (325°) and inclined by 59° from the nadir angle. Therefore, main parts of the observed crowns were oriented towards southeast. We expect different T C for unseen areas of the crown resulting from effective thermal anisotropy (Lagouarde et al. 2000; Voogt and Oke 2003). Hence, we analysed a directional canopy temperature and not the complete canopy temperature, which represents the three-dimensional (3-D) crown surface (Voogt and Oke 1997).

TIR images were recorded at 1-min intervals. The pre-processing of TIR images comprised four steps. At first, radiance was converted into brightness temperature T b using firmware calibration parameters and case temperature under the assumption that the surface is a Lambertian blackbody. The wide-angle lens produced radiometric and geometric distortions, which were corrected in steps two and three. Furthermore, an atmospheric correction was applied to account for spatial variability of line-of-sight (LOS) geometry due to the oblique view and the 3-D character of the examined urban environment. The reader is referred to the work of Meier et al. (2011) for further details on radiometric, geometric and atmospheric corrections. We utilised data of the WMO station 10381 Berlin-Dahlem in order to make an atmospheric correction for cloudy conditions. All TIR data collected under rainy conditions and cloud heights below 330 m were not used. Finally, 63180 TIR images (2106 half-hourly averages) were used for the analysis. This corresponded to 71 % of the two-month measurement period.

We applied the Stefan–Boltzmann law (δ = Stefan–Boltzmann constant) assuming an emissivity (ε) of 0.97 for the canopy surface (Campbell and Norman 1998) in order to compute T C. Further, we account for reflected downward long-wave radiation (↓ E LW) measured on top of the high-rise building (Eq. 1). A detailed consideration of reflected long-wave radiation originated from other parts than the sky (walls, roofs and ground surfaces) was not yet possible due to unknown view factors of tree pixels and not TIR-scanned surrounding surfaces.

$$ T_{\rm{b}}^4 \cdot \delta = T_{\rm{C}}^4 \cdot \varepsilon \cdot \delta + (1 - \varepsilon ) \cdot {}^{ \downarrow }{E_{\rm{LW}}} $$
(1)

The sensor-target distances were between 150 and 550 m. Therefore, the spatial resolutions of tree pixels were between 0.5 and 2 m. This spatial resolution exceeds typical leaf size values, and in fact T C is a spatial integral over a lot of leaves and partly branches.

We calculated the spatial median of all T C corresponding to an individual tree crown in order to quantify the canopy temperature of a single tree. The temperature range within the crown was defined as the difference between the 95th and 5th percentiles of T C. Although T a was measured 22.5 m above ground, we expect spatial variability of T a at this height within the FOV. This aspect was beyond the scope of the study and was not considered in the calculation of ∆T C.

2.3 Tree data

We selected 67 individual trees and compiled a crown mask based on photographs from the high-rise building and TIR images. Each crown could be clearly separated from other surrounding surfaces. Figure 1 shows the FOV of the TIR camera and all selected crowns, which represent approximately 40 % of all trees inside the FOV. Among these, we found Acer platanoides, A. pseudoplatanus, A. campestre, Fagus sylvatica, Populus nigra, Platanus acerifolia, Quercus robur, Tilia spp. and various conifers (e.g. Taxus baccata, Pinus sylvestris, Larix decidua, Abies procera). For 41 broadleaf trees, we measured the leaf size of 20–50 leaves per tree using a leaf area meter (Li-Cor Model 3100). In order to estimate the influence of tree-specific location within the city, we computed the mean fraction of impervious surfaces (FIS) around the trunk in a square of 50 × 50 m. This area was oriented towards southeast taking into account the viewing direction of the camera, i.e. 10 m towards north and west and 40 m towards south and east, respectively. For this purpose, we used the ‘Impervious Soil Coverage’ map of Berlin that is available in a 2.5-m raster format (Senate Department for Urban Development 2007). One important driver of T C is the stomatal conductance (g s) (Jones 1992). The tree-specific stomatal conductance was not measured in our study. However, there are differences between species regarding the maximum g s (Körner et al. 1979). We made a literature survey in order to have an estimation of the maximum g s values for the investigated species. Species-specific tree data are summarised in Table 1.

Table 1 Number of replicates and in parenthesis number of individuals, where leaf size was measured, number of thermal-infrared (TIR) pixels (i) and species-specific values of leaf size and standard deviation (where applicable), fraction of impervious surfaces (FIS) and standard deviation (where applicable)
Full size table

3 Results

This section starts with an overview of spatio-temporal patterns showing the spatial distribution of minimum, mean and maximum ∆T C. Then we present the influence of atmospheric conditions and tree-specific and site-specific parameters on the spatio-temporal variability of ∆T C. All time specifications refer to Central European Time (CET).

3.1 Spatio-temporal patterns

In addition to the ∆T C patterns, we present spatio-temporal patterns of the surface-to-air temperature difference (∆T) for every pixel of the whole FOV based on Eq. 1 to facilitate interpretability. Trees on the city square showed the highest mean ∆T C. They reached values up to 0.9 K (Fig. 2b, lower right corner). The average canopy temperatures of park trees were close to air temperature. Some individuals, for instance P. nigra (No. 43 in Fig. 1), reached values up to −0.5 K. Only trees and lawn surfaces showed negative ∆T values in the mean pattern. All other urban surfaces showed positive ∆T values.

Fig. 2
figure 2

Spatio-temporal patterns of a minimum, b mean and c maximum canopy-to-air temperature difference (∆T C), as well as d minimum, e mean and f maximum surface-to-air temperature difference (∆T). The x- and y-axes values are the column and row index of each pixel

Full size image

The minimum ∆T C ranged from −3.0 to −1.6 K (Fig. 2a). Trees along the road were relatively warm in comparison to park trees. Roofs of buildings and lawns showed the lowest minimum ∆T values. They were up to 5–6 K cooler than air temperature (Fig. 2d).

The maximum ∆T C ranged from 1.6 to 6.4 K (Fig. 2c). Trees No. 2 (A. campestre) and No. 10 (A. platanoides) reached the highest maximum ∆T C values. Two-thirds of the 67 trees reached their maximum ∆T C between 0900 and 1100 hours. Roofs of buildings showed maximum surface-to-air temperature differences of more than 30 K (Fig. 2f).

3.2 Influence of weather conditions

3.2.1 Meteorological conditions during the study period

Table 2 summarises meteorological conditions during June, July and August 2010 and the corresponding long-term values (standard period 1961–1990, 2 m above ground) recorded at the climate station Berlin-Dahlem, Albrecht-Thaer-Weg (Chmielewski and Köhn 1999), which is situated at a distance of 1.5 km to the northwest of the study site. The June values are included in order to show the very low amount of precipitation (6.4 mm) before the study period. During July, T a and VPD exceeded the long-term means by about 4.6 K and 7.4 hPa respectively.

Table 2 Monthly means of air temperature (T a) and vapour pressure deficit (VPD), monthly sums of downward short-wave radiation (↓ E SW) and precipitation in 2010 in comparison to climate data for the 1961–1990 period measured at the climate station Berlin-Dahlem, Albrecht-Thaer-Weg (Chmielewski and Köhn 1999)
Full size table

Figure 3 shows daily means of T a, VPD and daily sums of precipitation. The daily minimum, maximum and mean values of ∆T C are only depicted for days with more than 75 % data availability. Especially July was very hot and dry leading to persistent high values of T a and VPD. On July 11, the average T a was higher than 30 °C, and VPD reached 27 hPa. The daily means of ∆T C ranged between −0.4 and 0.9 K. The maximum ∆T C was higher than 5 K for 14 days in July and only for 5 days in August.

Fig. 3
figure 3

Temporal variation of daily air temperature (T a), vapour pressure deficit (VPD), precipitation and canopy-to-air temperature difference (∆T C). The grey bands indicate the hot and dry period A as well as the warm and wet period B

Full size image

Due to the different weather conditions, we decided to analyse two representative periods in detail. We made a distinction between the hot period with dry soils (period A, 10th July 2010) and the warm period with replenished soils (period B, 29th July 2010). The grey bands in Fig. 3 show these 2 days. The sum of downward short-wave radiation was similar on both days, i.e. 26.5 MJ m−2 during period A and 26.1 MJ m−2 during period B.

3.2.2 Diurnal course and histogram of canopy temperatures

The maximum range (maximum T C − minimum T C) was 5.2 K for period A and 5.4 K for period B. Although maximum variability was equal in both cases, it was observed at different times, i.e. at 1300 hours for period A and at 0930 hours for period B. In both periods, maximum canopy temperatures were reached before maximum air temperatures.

The grey box plots (Fig. 4) of the hot and dry period (A) reveal that nearly all tree crowns were warmer than air temperature during daytime. Mean ∆T C was positive between 0630 and 2030 hours. The white box plots (Fig. 4) of the warm and wet period (B) show that several tree crowns were cooler than air temperature during the day. In both periods, the mean ∆T C values were negative during the night.

Fig. 4
figure 4

Diurnal courses of mean canopy-to-air temperature difference (∆T C), canopy temperature (T C) of 67 trees (box plots), air temperature (T a) and vapour pressure deficit (VPD) for period A and B. Grey box plot represents data from period A

Full size image

The difference between both periods in terms of ∆T C is clearly visible in the histograms presented in Fig. 5. The histogram of period A is shifted to higher values and shows a bimodal distribution. Average ∆T C was higher during the hot and dry period (mean 0.9 K, median 0.7 K) than during the warm and wet period (mean 0.1 K, median −0.3 K). Both distributions had a positive skewness.

Fig. 5
figure 5

Histogram of canopy-to-air temperature difference (∆T C) during periods A and B

Full size image

3.3 Species-specific temperature patterns

Table 2 summarises the observed temperature patterns of all species. The average daily mean ∆T C of all species was slightly above air temperature (0.2 K). Only P. nigra was cooler than air temperature. All other species were on average warmer or equal to air temperature. A. campestre and T. baccata reached the highest average values of 0.7 K. Species-specific maximum ∆T C varied between 1.9 and 5.6 K. The mean canopy temperature was between 22.3 and 23.4 °C. A. platanoides, P. acerifolia and T. baccata reached maximum T C values of 40 °C. The average daily range within the crown was 0.8 K. Species-specific maximum temperature range within the crown was between 1.6 K (P. sylvestris) and 4.4 K (A. pseudoplatanus, A. campestre).

3.3.1 Comparison between period A and period B

For period B the magnitude and ascending order of daily mean ΔT C corresponds to Table 3. During period A, all species had significantly higher daily mean and noon ΔT C values (Fig. 6a, d). The largest temperature difference (period A − period B) of 0.9 K was observed for L. decidua, Tilia x europaea, T. baccata and P. wallichiana.

Table 3 Species-specific values ± standard deviation (where applicable) of canopy temperature (T C), canopy-to-air temperature difference (ΔT C) and temperature range within the crown, sorted by increasing average daily mean of ΔT C (only days with more than 75 % data availability)
Full size table
Fig. 6
figure 6

Species-specific values ± standard deviation (where applicable) of a daily mean canopy-to-air temperature difference (ΔT C), b temperature range within the crown at noon, c ΔT C at midnight and d ΔT C at noon. Grey symbols indicate data of period A, and black symbols indicate period B. The species were sorted using the values of period B. p-Values refer to t-tests of period A versus period B temperatures

Full size image

At night, species-specific range (maximum ∆T C – minimum ∆T C) was very low, i.e. 1.2 K for period A and 0.6 K for period B, but interestingly A. platanoides and A. pseudoplatanus showed the lowest values in contrast to the daytime situation (Fig. 6c). The largest differences (period A − period B, nighttime ∆T C) were observed for T. baccata (0.5 K), A. campestre (0.4 K) and L. decidua (0.4 K).

Furthermore, data from the noon situation was analysed, because we expect minimum influence of shadowing due to buildings and other objects at this time. While downward short-wave radiation was similar at noon (830 W m−2), a higher T a was observed during period A (34.4 °C vs. 21.1 °C). The species-specific range at noon was 4.4 K for period A and 4.2 K for period B. P. nigra showed the lowest values during both periods. A. campestre showed the highest ∆T C, but only a moderate temperature difference of 0.9 K between both periods at noon. The largest temperature differences were observed for P. wallichiana (1.9 K), Tilia x europaea (1.8 K) and A. negundo (1.5 K).

At noon the range of T C within the crown was higher for deciduous trees than for conifers (Fig. 6b). All species except P. nigra (−0.1 K) showed an increase in the range of T C within the crown of up to 0.8 K for F. sylvatica. The mean range of T C within the crown at noon was 1.1 K for the warm and wet period and 1.6 K for the hot and dry period.

3.3.2 Relation between leaf size and canopy temperature

The size of the examined tree-specific leaves was between 19 cm2 (P. nigra, No. 43) and 111 cm2 (A. pseudoplatanus, No. 15). A. platanoides and A. pseudoplatanus showed the highest standard deviation (see also Table 1). Figure 7 shows that trees with smaller leaves had lower ∆T C during daytime for both periods. The leaf size of A. campestre (No. 2) is small in comparison to the other Acer trees, but this individual showed the highest ∆T C during daytime in both periods. During the night, small leaves were slightly warmer than big leaves, but regression analysis yielded a very low correlation between these two parameters (not shown, period A: r 2 = 0.11, p = 0.04, period B: r 2 = 0.26, p < 0.01).

Fig. 7
figure 7

Relation between leaf size and canopy-to-air temperature difference (∆T C) averaged for the daytime situation of period A (left) and period B (right). Horizontal error bars indicate ± standard deviation of leaf size. Vertical error bars indicate ± temporal standard deviation of ∆T C between 0800 and 1600 hours

Full size image

3.4 Influence of site-specific characteristics

The spatio-temporal patterns of ∆T C presented in Fig. 2 already show a relation between the specific urban location of the tree and its thermal behaviour. To investigate this relation, we used the species-specific FIS around each tree as a quantitative descriptor representing the influence of urban structures on its thermal behaviour. The FIS values were between 11 % (Thuja occidentales) and 95 % (A. campestre). Figure 8 shows that species surrounded by a higher fraction of impervious surfaces had higher daily mean ∆T C. This relation was slightly stronger for the hot and dry period regarding the linear correlation coefficients r 2 (A: 0.38 vs. B: 0.36).

Fig. 8
figure 8

Relation between fraction of impervious surfaces (FIS) and daily mean canopy-to-air temperature difference (∆T C) for period A (left) and period B (right). Vertical error bars indicate ± standard deviation of ∆T C, and horizontal error bars indicate ± standard deviation of FIS within the species (where applicable)

Full size image

4 Discussion

The spatial and temporal resolution of the presented urban surface temperature in combination with meteorological measurements allows analysing canopy-to-air temperature differences of individual trees belonging to 18 species. For the first time, half-hourly data spanning a period of 2 summer months could be presented. Average canopy temperature was nearly equal to air temperature. Therefore, it is possible to determine the average canopy temperature based on air temperature. This is helpful if canopy temperature is not explicitly measured or calculated by an energy balance approach. For instance, the SOLWEIG model (Lindberg and Grimmond 2011) uses this assumption for the simulation of spatial variations of 3-D radiation fluxes and mean radiant temperature. However, the assumption is questionable for a full description of a diurnal cycle and species-specific differences.

4.1 Variability of T C within the crown

The maximum observed variability of T C within the crown varied between 1.6 and 4.4 K. These differences could have an impact on the diversity of animal and plant communities in tree crowns. Broad-leaved deciduous trees showed higher values than conifers (Fig. 6b). Leuzinger and Körner (2007) report higher values for mature forest trees, i.e. between 6 and 11.8 K, although deciduous trees show also higher values than conifers. The different experimental set-up could have been the reason for these differences in magnitude. They mounted the TIR camera directly above the canopy resulting in a higher geometrical resolution, which provides more details of canopy structure like branching, leaf area density or sunlit and shadowed areas. The range of species-specific daily average ∆T C corresponds to the magnitude of the mean temperature variability within the crowns (Table 2, 1.1 vs. 0.8 K). However, at noon the temperature variability among species was greater than variability within the crown (Fig. 6b, d).

It has been proposed that temperature variability within a canopy is a better indicator of plant water stress than the average canopy temperature (Fuchs 1990). The observed increase of temperature variability within the crowns for period A (Fig. 6b) could be an indicator of plant water stress. However, buildings and isolated trees produce shadow patterns. We cannot neglect this influence, even if we have analysed temperature patterns at high sun elevations.

4.2 Variability of species-specific ∆T C in the daytime

The observed differences between species can be explained with the help of species-specific stomatal conductance and leaf size (Table 1). P. nigra and Q. robur reach high stomatal conductance values. We expect high transpiration rates and lower ∆T C for these tree species during ample water supply. The observed ∆T C for these species (Table 2 and Fig. 6) confirm this. For the hot and dry period as well as during noon, we expect stomatal regulation of transpiration via stomatal closure. Leuzinger et al. (2010) present species-specific variations of ∆T C based on calculations of the leaf energy balance under the assumption of decreasing g s up to 20 % of maximum g s. A comparison of these results with our observations, especially maximum ∆T C, suggests that g s did not decline below 50 %. This could indicate the presence of sufficient soil moisture and no severe water stress even during the hot and dry period. However, in order to confirm differences in stomatal regulation of transpiration, we need more information on the variation of tree-specific stomatal conductance.

The size and shape of leaves are important for heat loss by convection (Vogel 2009). These species-specific properties influence the heat and mass exchange between leaf and atmosphere (Schuepp 1993). As expected and previously shown (Leuzinger and Körner 2007; Leuzinger et al. 2010), trees with small and lobed leaves (e.g. Q. robur) as well as conifers (e.g. L. decidua, P. sylvestris, P. wallichiana) tend to reach lower temperatures. The thin flattened petioles of P. nigra cause the flutter of leaves even during low wind speed. This behaviour additionally reduces the boundary layer resistance and therefore leaf temperature (Grace 1978). Contrary to that, trees with large leaves tend to reach higher temperatures (e.g. A. platanoides, A. pseudoplatanus, P. acerifolia). In the study of Leuzinger et al. (2010), individuals of A. platanoides showed maximum ∆T C of >5 K. This corresponds to our results (Fig. 2c and Table 2).

4.3 Variability of species-specific ∆T C at night

Due to the reduced boundary layer resistance for smaller leaves, they tend to show higher temperatures than larger leaves in case of ∆T C < 0 as described in the example in Campbell and Norman (1998). This might be the reason for the observed temperature pattern during the night (Fig. 6c). Furthermore, if the TIR camera observes a relatively open crown, then we also measure radiation emitted from surfaces within the crown. These surfaces should have higher temperatures during the night because of the different long-wave radiation balance.

The crowns of trees No. 4 and No. 9 reach exactly the roof edges of an adjacent building. This might be influencing the cooling process of the crown, because roofs have the lowest surface temperature during nighttime (Fig. 2). Spronken-Smith and Oke (1998) analysed thermal imagery of urban parks and their surroundings, where roofs were also cooler than tree canopies during nighttime.

4.4 Spatial variability of ∆T C

The influence of the tree-specific location within the city is superimposed onto the above-stated relation between leaf size and ∆T C. The leaves of tree No. 2 are rather small (33 cm2), but this tree is surrounded by a high fraction of impervious surfaces and reached the maximum ∆T C during the study. Previous studies point out the effect of urban ground cover on tree microclimate (Whitlow et al. 1992) and leaf temperatures (Kjelgren and Montague 1998; Montague and Kjelgren 2004; Mueller and Day 2005). Our study confirms that trees over impervious surfaces exhibit consistently higher canopy temperatures due to interception of upward long-wave radiation from fabricated surfaces having higher surface temperatures. Higher leaf temperature enhances the leaf-to-air vapour pressure deficit and the relevance of stomatal control on transpiration (Kjelgren and Montague 1998; Celestian and Martin 2005). However, for short periods building shadows can reduce canopy temperature (Meier et al. 2010).

One frequently discussed way to reduce the radiative input in urban areas is the use of high albedo materials. This measure would decrease the emission of long-wave radiation, but would also increase the trees’ absorption of reflected short-wave radiation. A deeper understanding requires information on the complete 3-D radiation exchange between trees and fabricated structures. In this context, the combination of TST, meteorological observations and energy balance models like DART EB (Gastellu-Etchegorry 2008) can be helpful.

Spatial variability of soil moisture and atmospheric turbulence within the FOV can produce temperature differences because these factors are important drivers of the leaf energy balance. The first aspect is beyond the scope of the study. Regarding the second aspect, Christen et al. (2011) showed that temperature fluctuations caused by atmospheric turbulence are spatially coherent and reveal the form of individual crowns. Conifers show a slightly lower spatial coherence than deciduous trees.

5 Conclusions

Our results illustrate that canopy temperature of urban trees depends on species-specific properties and the location of the tree. Moreover, the study shows that despite a distinct hot and dry period, the observed mature trees remain relatively cool in contrast to impervious surfaces. Knowledge of species-specific canopy temperature and the impacts of urban structures on the energy and water balance of trees is essential to optimise urban climates, e.g. to tackle adverse effects of the urban heat island. However, any advice for urban planning purposes depends on the climatological impacts you would like to have or not. Species with a lower canopy temperature like P. nigra, Q. robur or Tilia cordata are suitable for the reduction of air temperature. However, Q. robur and several species of Populus as well as P. acerifolia have a high ozone formation potential because of high emissions of volatile organic compounds (VOC) (Benjamin and Winer 1998). On the other hand, species with higher canopy temperatures like A. platanoides emit less VOC (Kesselmeier and Staudt 1999). A comprehensive evaluation and optimisation of benefits from urban trees should take the full range of climatological effects into account. These include the modification of surface temperature underneath trees, air quality, human thermal comfort, the ventilation of street canyons and the energy consumption of buildings.