Assessment of evaporative water loss from Dutch cities

Building and Environment 83 (2015) 27e38

Contents lists available at ScienceDirect

Building and Environment journal homepage: www.elsevier.com/locate/buildenv

Assessment of evaporative water loss from Dutch cities Cor Jacobs a, *, Jan Elbers a, Reinder Brolsma b, Oscar Hartogensis c, Eddy Moors a,
rquez a, Bert van Hove c, d María Teresa Rodríguez-Carretero Ma a

Wageningen UR, Alterra, Climate Change and Adaptive Land and Water Management, PO Box 47, 6700 AA Wageningen, The Netherlands Deltares, PO Box 177, 2600 MH Delft, The Netherlands c Wageningen University, Meteorology and Air Quality Group, PO Box 47, 6700 AA Wageningen, The Netherlands d Wageningen University, Earth System Science Group, PO Box 47, 6700 AA Wageningen, The Netherlands b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 6 May 2014 Received in revised form 25 June 2014 Accepted 2 July 2014 Available online 10 July 2014

Reliable estimates of evaporative water loss are required to assess the urban water budget in support of division of water resources among various needs, including heat mitigation measures in cities relying on evaporative cooling. We report on urban evaporative water loss from Arnhem and Rotterdam in the Netherlands, using eddy covariance, scintillometer and sapflow observations. Evaporation is assessed at daily to seasonal and annual timescale. For the summer half-year (AprileSeptember), observations from Arnhem and Rotterdam are consistent regarding magnitude and variability of evaporation that typically varies between 0.5 and 1.0 mm of evaporation per day. The mean daily evaporative cooling rate was 20 e25 Wm
2, 11e14% of the average incoming solar radiation. Evaporation by trees related to sapflow was found to be a small term on the water budget at the city or neighbourhood scale. However, locally the contribution may be significant, given observed maxima of daily sap flows up to 170 l per tree. In Arnhem, evaporation is strongly linked with precipitation, possibly owing to building style. During the summer season, 60% of the precipitation evaporated again. In Rotterdam, the link between evaporation and precipitation is much weaker. An analysis of meteorological observations shows that estimation of urban evaporation from routine weather data using the concept of reference evaporation would be a particularly challenging task. City-scale evaporation may not scale with reference evaporation and the urban fabric results in strong microweather variability. Observations like the ones presented here can be used to evaluate and improve methods for routine urban evaporation estimates. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Urban evaporation Water budget Evaporative cooling Reference evaporation

1. Introduction Evaporation links the energy budget to the water budget of urban regions. This link implies that evaporation can help to prevent heating of cities, since the energy used to evaporate water is not available anymore for warming the urban fabric or atmosphere. Lack of evaporation in cities due to replacement of natural soil and green cover by impervious structures like buildings and streets has since long been known to contribute to the Urban Heat Island (UHI) effect [1]. Inversely, many surveys have demonstrated cooling effects of vegetation in the urban microclimate [2], even in the temperate maritime climate of the Netherlands [3e5]. Improving green infrastructure by planting or maintaining vegetation, including application of green roofs, has become a popular measure

* Corresponding author. Tel.: þ31 317486460; fax: þ31 317419000. E-mail address: [email protected] (C. Jacobs). http://dx.doi.org/10.1016/j.buildenv.2014.07.005 0360-1323/© 2014 Elsevier Ltd. All rights reserved.

to mitigate heat in cities and increase human thermal comfort [2,6,7]. Trees may be particularly effective in this respect because they not only provide cooling by evaporation, but also by shading [2,8]. Although open water in the city also contributes to urban evaporation, the effect of water bodies on mitigating heat in the city is less clear [3,5,9]. Mitigating heat in cities and at the same time reducing urban water consumption is an extremely challenging task. Cooling by evaporation obviously requires ample water supply, especially if vegetation is involved. At the same time urban water use may have to be reduced in the near future because more drought events are expected under climate change [10]. Reliable estimates of evaporative water loss are required in support of appropriate water management, especially during hot conditions when water has to be divided among various needs, including general water supply to citizens, evaporative cooling of cities and survival of urban vegetation. Monitoring evaporation is expected to become more urgent in the future because of urbanisation, climate change and their

28

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

impact on outdoor thermal comfort [11]. In the Netherlands, evaporation is at present estimated mainly in support of regulation of groundwater levels. Unfortunately, hardly any information is available on evaporation from Dutch cities, in spite of its importance regarding specific Dutch problems linked to the urban water budget, including salt water intrusion in coastal areas and decay of wooden foundations upon lowering the groundwater level [12]. The objective of this paper is to assess evaporation from urban areas in The Netherlands. We analyse observations carried out in two cities in the framework of “Climate Proof Cities” (CPC) [13]. In Arnhem, evaporation was measured using the eddy covariance technique. In Rotterdam, evaporation was derived from scintillometer observations and sapflow observations were carried out to examine water use of individual urban trees. These various estimates of evaporation are complemented by meteorological observations that enable us to compute the so-called reference evaporation (see Section 2.5) in both cities. In agriculture, this quantity has often been used as a starting point for estimates of crop water requirements when direct observations of evaporation are not feasible [14]. Given the lack of information on evaporation from Dutch cities, the first research question underlying the present study is: how much water evaporates from Dutch cities? Quantification of urban evaporative water loss helps to assess avoided heating of cities and is interesting for water managers, notably the Dutch Water Boards and larger municipalities that are responsible for water management in the Netherlands. Our analyses of the data will mainly focus on observations from the summer half-year (AprileSeptember) because of the link with heat in the city. The second research question is: what fraction of precipitation received in the urban areas evaporates again? Estimating the relation between evaporation and precipitation including the amount of recycled water supports water management issues, while also improving understanding of the role of urban design in the water budget. To overcome water shortage during hot periods harvesting as much water from precipitation as possible may become crucial in order to refill depleted water reservoirs [10]. Buildings and sealed soils may render cities quite efficient in recycling water to the atmosphere since they can act as an interception reservoir from which water easily evaporates. However, urban characteristics generally prevent infiltration and promote rapid transport of water away from cities via stormwater networks and sewage systems. The third research question is: to what extent can the concept of reference evaporation be used to derive city evaporation on a routine basis? Direct and indirect specialized routine observations of evaporation from cities are scarce in spite of the expanding network of urban flux observations [11,15]. There is a need for relatively simple methods that allow routine monitoring of evaporation in support of water management. In agricultural practice estimations of crop water requirements have been derived from the reference evaporation with reasonable success [14]. In particular on dry days, urban evaporation is largely due to city vegetation [16]. Therefore, and because the concept of reference evaporation can be linked with remote sensing observations quite easily [17], applying the method used in agriculture but adjusted to the urban setting is sometimes considered a practical way to estimate urban evaporation, at least from vegetated parts [11], with promising results at specific urban green spots [18,19]. However, given the agricultural background of the methodology, application of the concept in the urban environment is far from trivial because of the impact of the urban fabric on local weather conditions. Thus, it is not clear if the concept allows reliable routine evaporation estimates at the neighbourhood to city scale, which is the relevant scale for many Dutch water managers.

2. Site description and methods 2.1. City characteristics Observations were performed in Arnhem and Rotterdam (see Ref. [13] for a map showing their location). Characteristics of these cities relevant to urban evaporation are provided first. More specific characteristics of the measurement locations are given in the subsections that follow, or described elsewhere in this issue [3]. Arnhem is located near a forested area in the Netherlands. The total area of the municipality of Arnhem is 102 km2 and the number of residents is just over 150,000 [20]. The share of impervious surface, including streets and buildings, is 44% [21]. In 2009, the municipality had a relatively large share of green space, covering 77 km2 or over 75% of the city area. However, in the city centre the green fraction is much less (~12%, see Section 2.2). Almost one million (±983,500) trees are present in Arnhem, of which 56,500 in parks and estates and 47,000 distributed as city trees. The total area of surface water within the municipality is 4.5 km2, of which 1.9 km2 is occupied by the River Rhine [22]. The average annual temperature (1981e2010) measured at a weather station operated by the Royal Netherlands Meteorological Institute (KNMI) about 8.5 km North of Arnhem is 9.8  C. The average maximum temperature in the summer months (June, July, August) is 21.9  C, the average minimum temperature 11.5  C. The mean annual precipitation is 861 mm, of which on average 224 mm is received during the summer months [23]. The city of Rotterdam covers a total area of 319 km2 and hosts nearly 615,000 residents. The share of impervious surface is 45%. Over 30% of the Rotterdam area, 114 km2, is surface water [20]. In 2003, the total area covered with vegetation was estimated to be about 52 km2, or about 16% of the city area (25% of the land area) [21]. Of course, these numbers vary widely across the city [3]. The number of trees within the municipality is almost 600,000, of which 450,000 are located in parks and about 150,000 in streets [24]. The mean annual temperature measured at the KNMI airport weather station near the city is 10.4  C. During the summer months, the average maximum temperature is 21.5  C, the average minimum temperature 12.2  C. The mean annual precipitation sum is 856 mm, of which on average 220 mm is received during the summer months [23]. 2.2. Eddy covariance measurements In Arnhem, actual evaporation was determined using the eddy covariance (EC) method. An EC station was set up in the heart of the city (N51.9847, E5.9183), on the roof of a 6-storey building. The height of the roof is 36 m ASL, the building height is about 18 m. The EC equipment was mounted on top of a mast extending 5 m above the flat roof so that the EC measurements are carried out at 23 m above ground level. The impact of spurious eddies created by wind flow past this building on the EC measurements is assumed to be negligible. The EC technique yields direct evaporation estimates at the scale of a hectare to a few square kilometres, depending on the observation height, the atmospheric conditions and the surface characteristics [25]. A first-order analysis of the footprint of the flux measurements [26] shows that in our case city parts within 1 km from the tower usually contribute to the measured flux well over 80%. A map of the building height within a circular zone with radius 1 km around the site is shown in Fig. 1. The average building height in this footprint of the flux measurements, zH (m), is nearly 11 m. Thus, the measurement height is 2.1 times zH, implying that effects of individual buildings have largely been blended [27]. A few isolated tall buildings in the area are located at least 800 m away from

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

29

evaporation, the observations from Arnhem are compared with similar observations from the Loobos flux site, located 23.5 km NNW of the site in Arnhem, in a homogeneous pine forest area in the same water catchment. The Loobos site is equipped with similar instrumentation, installed on top of a 26 m high mast and EC data processing follows the same data correction and QC procedures as previously referred to [32,33]. 2.3. Scintillometry

Fig. 1. Building height in a circular area (radius ¼ 1 km) in Arnhem, representing the main footprint of the EC flux measurements. The location of the EC site is highlighted by means of the green dot. Only buildings higher than 5 m are shown. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the site and will not affect the EC measurements noticeably (see Fig. 1). Table 1 provides an overview of the land use within 1 km from the tower. The EC measurements were performed using an R3-50 sonic anemometer (Gill, Lymington, UK) and an LI-7500 gas analyser (LI-COR Biosciences, Lincoln, USA). The data processing, quality assessment and control and data selection procedures applied here correspond with international recommendations on EC measurements in general [25,28] and the more specific ones for EC flux measurements over urban areas [27]. After data selection, gaps in the time series were filled [29,30] and daily sums of evaporation were computed. Following the recommended procedures, results in an uncertainty in EC based estimates of evaporation of less than 15% [31]. Meteorological observations at the EC site were performed using instruments and a data logger mounted on a tripod, secured close to the edge of the roof. Incoming and outgoing shortwave and longwave radiation components were measured with the instrument extending 1.5 m outside the edge of the roof (NR01, Hukseflux, the Netherlands). In our analyses of the Arnhem data we only use downwelling radiation that is not affected by walls and other objects below the sensor. The location is sufficiently high to avoid shading effects on downwelling radiation. Air temperature was measured at 1.5 m and 5 m above the roof using a shielded sensor (107, Campbell Scientific, USA). Wind speed and direction were determined at 2 m above the roof using a 2D ultrasonic anemometer (Gill, Lymington, UK). At 0.4 m above the roof, precipitation was measured using a tipping bucket rain gauge (ARG100, EM, UK). The measurements started on 20 June 2012. Here, we analyse data for the period 21 June 2012e30 September 2013 (486 days). In order to allow a better evaluation of the impact of urban land use on

Table 1 Surface cover in the footprint area (see Fig. 1) of the EC measurements in Arnhem. Surface cover (e)

Fraction (%)

Buildings Sealed surface Railroad Water Vegetation: forest Vegetation: grass

35 49 4 1 3 9

In the city of Rotterdam, evaporation was estimated using scintillometry. The scintillation technique is based on the propagation of electromagnetic radiation through the turbulent atmosphere between the instrument transmitter and receiver. The energy of the electromagnetic radiation exhibits fluctuations known as scintillations. For optical wavelengths, temperature fluctuations cause scintillations that can be related to sensible heat flux, H in Wm
2, via micrometeorological theory [34]. An optical Large Aperture Scintillometer (LAS) allows determining the area-averaged H at a scale of several km2. An increasing number of studies use a LAS in urban areas, e.g. Refs. [35e38], because it has some advantages over the more traditional measurement techniques, like EC. Most importantly, the footprint is much larger, up to the size of an entire city centre (several km2). More practically, the measurement is insensitive to turbulence produced locally by obstacles near the transmitter and receiver. Here, the latent heat flux (LvE in Wm
2, where Lv is the latent heat of vaporization) or evaporation (mm) is estimated indirectly from daytime observations, as a residual term of the energy budget:

Lv E ¼ Rn
H
G

(1)

where Rn (Wm
2) is net radiation. Furthermore, G (Wm
2) is soil heat flux that in the case of urban environments includes heat used to warm or cool the urban fabric. This term can be quite large in the urban environment, but is difficult to estimate. We installed a LAS in a North-South orientation over a 3451 m path covering the city centre of Rotterdam (Fig. 2). At the North end of the path (N51.5648, E4.2775) the transmitter was installed 51 m above ground level. At the South end, the receiver was installed 77 m above the ground level (N51.5463, E4.2813). The effective measurement height of the scintillometer path was 60 m [39]. Buildings elevate the level from which the turbulence that passes through the scintillometer path originates. This so-called displacement height, d (m), was estimated as d ¼ 0.7zH. Using a scintillometer footprint model [40] we first calculated prototype footprints for 8 wind sectors (based on H ¼ 150 Wm
2, a wind speed of 10 ms
1, an aerodynamic roughness of 1 m and assuming the standard deviation of the lateral wind speed to be 0.4 ms
1). These footprints were then used as weighing function to determine zH from a digital elevation map of the city centre, ignoring obstacles smaller than 2 m tall and averaging building heights quadratically. Depending on the wind direction we found 10m < d < 20 m. Fig. 2 shows an example of the footprint for westerly wind. The scintillometer data were used to determine 30 min averaged H, ignoring humidity effects on the optical scintillation measurements. This is a reasonable assumption over urban areas because the Bowen ratio (b ≡ H/LvE) is larger than 0.8 most of the time. Then, the contribution of humidity to the scintillation statistic is less than 2%. To moderate the accumulation of the energy balance closure problem [41] in the evaporation estimate, LvE was set to zero during night-time, defined as zero incoming shortwave radiation. For the daytime estimates we use G ¼ 0.3Rn [42]. Small data gaps (<20% on a daily basis) were filled using a spline interpolation technique to ensure reasonable estimates of daily evaporation. Additional data

30

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

Fig. 2. Left: scintillometer path displayed on top of a satellite image of the city centre of Rotterdam (source: Google Maps). Right: scintillometer path displayed on top of a digital elevation map (DEM). The colour bar indicates the DEM heights (m). Also, a prototype flux footprint is shown for westerly wind. Each contour represents a cumulative 10% contribution up to 90%. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

needed to process the fluxes were obtained from the urban meteorological station “Centre” in Rotterdam [3]. The net radiation used to obtain G and needed to arrive at LvE was taken as a city average from three rooftop stations (“Centre”, “East” and “Ommoord”) and three ground-based stations (“Rijnhaven”, “South” and “Vlaardingen”) [3].

2.4. Sapflow measurements Evaporation from plants through stomata in the leaves (often called transpiration) of individual trees was determined at two locations in Rotterdam using sapflow observations (Table 2). Two sample trees are growing in a typical street environment (Berkelselaan, henceforth denoted as “Berk”). The other three trees are Table 2 Properties of trees selected for sapflow measurements.

2.5. Reference evaporation

Treea

Species

Year of planting

Stem diameterb (cm)

Sapwood areac (cm2)

Berk 1

Common lime (Tilia  europaea) Common lime (Tilia  europaea) Ash (Fraxinus excelsior L.) Common lime (Tilia  europaea) Common lime (Tilia  europaea)

1960

29

528

7

1960

31

578

8

1979

45

930

12

1960

38

754

10

1947

44

905

12

Berk 2 Park 1 Park 2 Park 3 a b c

growing at the edge of a park (henceforth denoted as “Park”). All trees are growing in sandy soil. Sapflow was determined using a radial flowmeter (UP Umweltanalytische Produkte GmbH) based on diffusion of heat in the xylem of trees [43]. In each tree, two shielded temperature probes were installed at a height of about 3.5 m, 10 cm above each other. The upper probe was continuously heated using a temperature stabilized constant current source (84 mA). To compute the sapflow from the difference in temperature between the two probes an estimate is required of the active sapwood area [43], which is given in Table 2 along with other relevant properties of the trees. The temperature difference during night-time conditions, used as a baseline for the daytime readings, was determined from the sensor readings obtained between 0 and 4 LT. Data were acquired at a 10-min temporal resolution and subsequently accumulated to daily sums of sapflow (l day
1).

Crown diameter (m)

Berk ¼ Location “Berkelselaan”, Park ¼ location “Park”. Measured at breast height (dbh). Estimated from the diameter and assuming a sapwood depth of 8 cm.

Estimating evaporation using the concept of reference evaporation comprises at least two steps [14]. In the first one, the reference evaporation is computed, which describes the impact of weather on evaporation [14,44]. It is defined as the evaporation from an extensive healthy and productive, well-managed grassland or alfalfa field with specific characteristics (see Ref. [14]) and ample water supply. It is ideally computed from meteorological observations over such a surface. In the second step, reference evaporation is multiplied with a so-called crop coefficient, KC (
), to account for various biophysical effects of crop characteristics like roughness, rooting depth, leaf area and development stage on evaporation. For a specific crop type, KC may vary among climate zones and seasons, but otherwise the impact of weather on KC is assumed negligible. Multiplying the reference evaporation with KC yields the

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

evaporation from “excellently managed, large, well-watered fields that achieve full production under the given climatic conditions”. A third step may be applied to account for the impact of water, management and other environmental stresses on the crop [14]. Application of this concept to urban environments implies that the biophysical principles governing evaporation of the reference grassland and of a city are similar [14], which is unlikely to be the case at the neighbourhood to city scale but could be true for urban green spots. Also, the strong impact of urban structures on the local weather and the energy balance and the huge heterogeneity encountered in urban environments raise the question as to how representative the obtained reference evaporation actually is and if the implicit assumption that the impact of weather on evaporation is accounted for using measurements over a reference grassland is valid. Here, we address both issues regarding the use of reference evaporation in urban environments. In this paper, reference evaporation is computed in two ways. First, we apply the widely accepted methodology recommended by the United Nations Food and Agricultural Organization (FAO) [14]:

EFAO ¼

0:408DA þ g Ta900 þ273 u2 D D þ gð1 þ 0:34u2 Þ

(2)

where A ¼ Rn
G (Wm
2) is the available energy. Following the standard methodology we take G ¼ 0 for these computations, since we evaluate EFAO at a daily timescale [14]. Further, u2 (ms
1) is the wind speed measured at a height of 2 m, Ta ( C) the air temperature, D (kPa) the vapour pressure deficit, D [kPaK
1] the slope of the saturation vapour pressure deficit versus temperature function and g (kPaK
1) the psychrometric constant. Second, we use the modified Makkink equation [17], used by the Royal Netherlands Meteorological Institute to assess climatological drought conditions:

EMAK ¼ 0:408  0:65

D S D þ g in

(3)

EMAK is akin to EFAO but only needs incoming solar radiation (Sin in MJ day
1) and air temperature (implicit in D/(D þ g)) as input, making it insensitive to the longwave radiation exchange, the reflection of sunlight from the surface and wind speed. Furthermore, it does not require specification of G. It is easy to link EMAK with remote sensing data [17]. More importantly in the present context, Equation (3) presumably is less sensitive to modifications of microweather conditions by buildings and sealed surfaces. Because of the latter characteristic, we also use EMAK to scale and further analyse the behaviour of directly observed actual evaporation versus reference evaporation in Rotterdam and Arnhem. To assess the representativeness of EFAO and EMAK in urban environments we computed these quantities from the meteorological observations in Rotterdam [3], using data from the summer halfyear (AprileSeptember) of 2012. For this period, data coverage is excellent and observations from sapflow and scintillometry are available as well. Data from 12 urban stations are analysed, of which seven are “rooftop stations” and five are “ground stations” [3]. Data from one station in the network had to be excluded because technical problems prevented proper evaluation of reference evaporation. The reference evaporation from the meteorological station in the open grassland area outside the city (reference station) is considered to represent the reference grassland according to the computing guidelines [14]. EFAO requires observations at a level of 2 m. This matches the level of the measurements at the ground stations, but not at the rooftop stations in which case the measurement height is 4e6 m. The height difference may have an effect on wind speed in particular.

31

The guidelines on conversion of wind speed from one height to the other when computing EFAO [14] are usually invalid in the urban environment. Therefore we convert the wind speed observations from the rooftop stations to a level of 2 m using the ratios predicted for an urban environment by an exponential wind profile scaling [45]. This correction results in a 20e30% lower wind speed at 2 m than at the actual observation height, depending on the measurement height and the average frontal area density in the neighbourhood around the station [3,45]. Other variables obtained at the station locations are considered representative for a height of 2 m without such a correction. 3. Results and discussion 3.1. Evaporation from EC in Arnhem The daily evaporation measured at the EC site in Arnhem is shown in Fig. 3, along with the daily average air temperature and the sums of precipitation and global radiation. Also shown is the Bowen ratio (b ≡ H/LvE) computed from daily averages of H and LvE, respectively. The urban data are compared with the observations at the nearby Loobos forest site. Although radiation, air temperature and even precipitation are quite similar at the sites, the seasonal cycle of the urban evaporation is markedly different. It is much less pronounced over the city than over the forest. Notably the summer evaporation in the city remains much lower on average, but tends to increase upon precipitation events. As expected, b is higher and generally varies between 1 and 10, with a tendency to become less upon precipitation events. By contrast, b at the forest typically varies around one during the summer season, with a tendency towards higher values during dry spells. The high Bowen ratios confirm that during the summer months the energy used by evaporation is generally much less than the energy used for direct heating of the atmosphere [1], indicating that enhancing evaporation can further mitigate urban heating. In the period considered here, the total evaporation amounted to 374 mm or 36% of the precipitation (1040 mm) in the city. The forest evaporated 792 mm, which is equivalent to 88% of the precipitation (896 mm). The lower evaporation and higher b in the city demonstrates the well-known effects of the large fraction of sealed surface and the lack of vegetation in the city [16]. Next, we concentrate on the summer period (AprileSeptember) only. The missing data gap fraction on a particular day is demanded to be less than 25%, which minimizes the possibly artificial effect of gapfilling. Table 3 shows the average evaporation from the urban and the forest site along with the precipitation (P), Ta and EMAK, stratified into three precipitation classes: dry days, moderately wet days with P < 3 mm and very wet days with P  3 mm. The average summer day evaporation in Arnhem amounts to 0.86 mm day
1. This amounts to 60% of the precipitation in Arnhem (0.86/1.44). The fraction is almost equal for both summer seasons (2012 and 2013) and is in reasonable agreement with previous observations and model-based estimates [46,47]. The mean evaporation amounts to a daily mean (24 h) cooling rate of about 25 Wm
2, which is 14% of the average incoming solar radiation (183 Wm
2) and 5% of the incoming total shortwave plus longwave radiation (535 Wm
2) observed in 2012 and 2013. The evaporation from the city is on average almost a factor of three higher on days with P  3 mm than on dry days (P ¼ 0 mm). This is in sharp contrast with the forest, where evaporation is only about 25% higher in the wettest precipitation class. Thus, interception characteristics of the city have an even stronger impact on the evaporation than in the case of the forest of which the impact of interception is known to be large [33]. The importance of interception is further demonstrated by the fact that the observed

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

E (mm day -1)

32

5 4 3 2 1

P (mm day -1)

0 40 32

24

Date

16 8

Sin (MJ m-2 day-1)

0

35 28 21

Titel

14 7

Ta (°C)

0 30

24 18

12

Titel

6

Bowen ratio (-)

0 -6 100 10 1

Date

0.1

0.01

Date Fig. 3. Observed evaporation (E), precipitation (P), incoming solar radiation (Sin), air temperature (Ta) and Bowen ratio at the urban site in Arnhem (continuous black lines) compared with the Loobos forest site (grey dotted lines) in the period 21 June 2012e30 September 2013). Note that the Bowen ratio is plotted logarithmically in order to better reveal the differences between Arnhem and Loobos.

C. Jacobs et al. / Building and Environment 83 (2015) 27e38 Table 3 Observed precipitation (P), evaporation (E), reference evaporation EMAK and air temperature Ta at the EC sites in Arnhem and Loobos, respectively, stratified according to the amount of precipitation on a particular day. The number of days in each class is n. Only days in the period AprileSeptember with sufficient data coverage (>75%) were included in the analysis.

Arnhem Dry days 0 < P < 3 mm P  3 mm All days Loobos Dry Days 0.1 < P < 3 P>3 All days

n

P (mm day
1) E (mm day
1) EMAK (mm day
1) Ta ( C)

149 66 39 254

0 1.03 7.66 1.44

0.57 1.07 1.61 0.86

3.17 2.28 2.07 2.77

16.71 16.25 15.15 16.35

106 62 27 195

0 1.12 8.16 1.48

1.85 2.24 2.31 2.04

2.98 2.25 1.88 2.59

14.35 14.63 13.37 14.30

Mean evaporation (mm day-1)

relationship between urban evaporation and P is opposite to the relationship between EMAK and P. Reference evaporation is intended to describe the basic evaporation physics of dry surfaces and may fail when the surface becomes wet [48]. Fig. 4 depicts the average evaporation for dry days with different time lags since the last wet day and compares it with the average evaporation rate on wet days. The results suggest a gradual drying in Arnhem after a precipitation event. The drying typically extends over a few (at least 2e3) days, with evaporation dropping from 1.24 mm day
1 on wet days to 0.42 mm day
1 on dry days with at least two preceding other dry days. Assuming that the evaporation on the third and subsequent dry days after a precipitation event is exclusively due to dry vegetation, and using a vegetation fraction of 0.12 (Table 1) we obtain an average evaporation rate of 3.5 mm day
1 per square metre of vegetation. This estimate seems reasonable since it is only 6% higher than the mean EMAK on those days (3.3 mm day
1). For this reason, and because of a regular occurrence of precipitation events the gradual decrease of city evaporation is unlikely to be caused by drought stress of vegetation

1.50 1.25 1.00

0.75 0.50 0.25 0.00

Fig. 4. Measured evaporation on wet days and dry days, stratified according to the time lag since the last preceding wet day. Days with lag 1 are the first dry day after a wet day, days with lag 2 the second one and days with lag 3 or more have at least two preceding dry days. The error bars indicate the 95% confidence range around the bin average.

33

(cf. [49]). A possible explanation that warrants further investigation may be the enhanced storage capacity of flat roofs in the footprint area of the EC tower in Arnhem. More and prolonged storage of water on roofs implies less drainage to sewage systems and infiltration in city soil, while more water remains available for evaporation from wet spots during several days. This feature would be an important consideration in neighbourhood and city design targeted at heat mitigation. It would also be interesting to evaluate its effects against the ones of the green roof concept. 3.2. Evaporation from scintillometry in Rotterdam Fig. 5 depicts daily evaporation and b for the year 2012 in Rotterdam, obtained using the LAS. The average evaporation of 0.48 mm day
1 over the whole period (261 days) corresponds to 15% of the precipitation (3.25 mm day
1) received on the days with valid evaporation measurements. Evaporation shows a clear seasonality. The maximum daily evaporation in the summer period is 1.74 mm; the average is 0.68 mm or about 21% of the precipitation (3.20 mm day
1). We note that this ratio may be exceptionally low because of the anomalously high amount of summer precipitation in 2012. The mean summer-season daily evaporation rate corresponds to a cooling rate of about 20 Wm
2, or 11% of the mean incoming solar radiation measured at the reference station outside the city (188 Wm
2) and 4% of the incoming all-wave radiation (542 Wm
2). Like in Arnhem, b shows the high values typical for a city, varying between 1 and 10, except shortly after a rain event. The relation between evaporation and precipitation was investigated here as well. Table 4 lists results from this analysis. In Rotterdam evaporation also increased with precipitation, but the response was much weaker than in Arnhem. Again, the trend is opposite to the one in EMAK obtained from the observations at reference station. In contrast with the observations, EMAK shows a strong decrease with increasing precipitation. In Rotterdam, no statistically significant differences were found between evaporation rates on days with an increasing time lag since the last preceding wet day. This lack of a gradual drying suggests that the impact of interception on the evaporation is less pronounced than in Arnhem. This could be partly caused by differences in building style, and partly by the fact that the footprint of the LAS represents a much larger area in Rotterdam, representing more vegetation as well as open water surfaces (Section 2.3 and Fig. 2). If the evaporation on dry days with at least two preceding other dry days is assumed to represent evaporation from dry vegetation and taking the vegetation cover in the footprint area to be 0.16 (see Section 2.1) the daily evaporation rate from vegetation is estimated to be 3.9 mm day
1 per square metre of vegetation. This very rough estimate is 22% larger than EMAK for those days (3.2 mm day
1) and could be affected by evaporation from open water. 3.3. Sapflow observations in Rotterdam Fig. 6 depicts observed sapflow in Rotterdam, showing averages for the two locations along with the observations at the individual trees. The leafless period in April can clearly be distinguished. In May, there is a quick budding of leaves and the water consumption of the trees rises to levels varying around 80 l day
1 in the park (Park), and about 40 l day
1 in the street environment (Berk). From June to September, the average daily (24 h) water loss per m2 of crown area was 0.72 mm at Park and 0.98 mm at Berk, corresponding to cooling rates of 21 and 28 Wm
2, respectively. This is amounts to 11e15% of the incoming solar radiation in that period (190 Wm
2) and 4e5% of the incoming all-wave radiation (554 Wm
2). In the framework of heat mitigation options it is also instructive to consider evaporational cooling per tree [50]. On a per-

34

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

Evapotranspiration [mm/(day)]

2

1.5

1

0.5

0 01Jan

01Feb

01Mar

01Apr

01May

01Jun

01May

01Jun

01Jul

01Aug

01Sep

01Oct

01Nov

01Dec

01Jul

01Aug

01Sep

01Oct

01Nov

01Dec

Date

2

Bowen Ratio [−]

10

1

10

0

10

−1

10 01Jan

01Feb

01Mar

01Apr

Date

Fig. 5. Scintillometer-based estimate of daily summed evapotranspiration rates (mm day
1, upper panel) and Bowen Ratio b ¼ LvE/H (-, lower panel) for 2012 in Rotterdam. The grey bars indicate days with rainfall exceeding 1 mm day
1.

tree basis, daily mean (24 h) cooling rates in JuneeSeptember were on average 2.2 kW and 1.1 kW at Park and Berk, respectively. Daily maximum flows of up to 170 l were observed, equivalent to a daily mean cooling rate of 4.8 kW. On some of the fair weather days we observed an hourly water use of up to about 12e16 l, corresponding to hourly cooling rates of 8.2e10.9 kW per tree. This confirms the possibly important role of trees in urban heat mitigation. The sapflow and therefore the cooling rates depend strongly on growing conditions and tree properties like species and tree age [50e52]. Furthermore, drying of soils in the absence of precipitation and irrigation may ultimately affect cooling rates, depending on the ability and strategy of plant species to deal with drought stress [49,53]. The water consumption averaged over all five trees at the two sites and over all available observations from the summer half-year is 50 l day
1. Taking into account the crown areas computed from the crown diameters given in Table 2, the average water use is found to be 0.64 mm day
1, and the estimated total water use including all days in the period is about 116 mm. This in turn corresponds to only 22% of the summer period precipitation (520 mm

[23]). Excluding the virtually leafless period in April and initial leaf growth stages in May we obtain a water use of 61 l day
1 or 0.83 mm day
1, resulting in an estimated total water use of 101 mm over 122 days, or 26% of the precipitation (386 mm [23]). We recall that these ratios may be exceptionally small because of the anomalously high summer precipitation in Rotterdam in 2012. Extrapolating the summer season water use to 600,000 trees in the core city of Rotterdam (319 km2 [24]) we find an area-average of only 14 mm of tree evaporation. Although this is a crude estimate, it suggests a rather limited contribution from trees to the urban water budget, even when considering that evaporation of intercepted water has not been included, the observations were performed in a wet summer and the representativeness of the sample trees may be limited. At neighbourhood to city scale secondary effects related to groundwater extraction by trees, like decay of wooden foundations and salt water intrusion in coastal areas [12], are therefore expected to be relatively small as well. However, we recall that during daytime individual trees can extract significant amounts of water, notably around noon, with possibly significant effects on the groundwater level locally.

Table 4 Observed precipitation (P), evaporation (E), reference evaporation EMAK sensible heat flux H and net radiation Rn from the scintillometer measurements in Rotterdam, stratified according to amount of precipitation on a particular day. The number of days in each class is n. Only days in the period AprileSeptember with sufficient data coverage (>80%) were included in the analysis. EMAK and Rn are averages obtained from measurements at the urban meteorological stations (see Section 3.4). Rotterdam

n

P (mm day
1)

E (mm day
1)

EMAK (mm day
1)

H (MJ m
2 day
1)

Rn (MJ m
2 day
1)

Dry days 0 < P < 3 mm P  3 mm All days

56 43 45 144

0 1.29 9.02 3.20

0.63 0.64 0.79 0.68

2.84 2.07 1.80 2.29

6.72 4.74 3.52 5.13

8.52 6.99 6.56 7.45

Sap flow (l day-1)

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

35

180 150

Berk

120 90 60

30

Sap flow (l day-1)

0 180 150

Park Date

120 90 60

30 0

Date Fig. 6. Observed seasonal course of the sapflow at Berkelselaan (Berk, upper) and in the park (Park, lower). Site averages are shown (solid lines) as well as individual observations (squares: ash; circles, triangles and diamonds: common lime).

3.4. Reference evaporation in Rotterdam

EFAO (mm day-1)

Fig. 7 summarizes the results for EFAO (Equation (2)) and EMAK (Equation (3)) from the observations in Rotterdam. It depicts the difference between the daily maximum and minimum reference evaporation from the urban stations as the shaded area; results

from the reference station are plotted as the dashed line. While the seasonal trend is similar among all stations, the reference evaporation outside the city usually is higher than or nearly equal to the maximum from the urban stations. The average EFAO in the urban area is almost equal to the average EMAK (2.38 and 2.36 mm day
1, respectively). They are 15% and 11% lower,

7 6 5 4 3 2 1

EMAK (mm day-1)

0 7 6 5 4 3 2 1 0

Date Fig. 7. Seasonal course of reference evaporation computed from meteorological observations in Rotterdam. Upper panel: EFAO (Equation (2)); lower panel: EMAK (Equation (3)). The shaded area gives the range between the maximum and minimum value from the urban stations. The dashed line is the reference evaporation at the rural site. See [3] for information on the urban meteorological stations.

36

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

respectively, than their rural counterparts (2.79 and 2.66 mm day
1, respectively). Since EMAK is independent of surface characteristics and wind speed it is expected to be less sensitive to heterogeneity of the urban environment. The difference in the range of daily urban estimates (shaded area in Fig. 7) indeed is somewhat smaller for EMAK than for EFAO. The average difference between the daily minimum and maximum value in EMAK from the urban sites is 0.97 mm day
1 or 41% of the mean. For EFAO it is 1.22 mm day
1 or 51% of its mean. However, the intra urban differences in EFAO and the difference with the rural reference tend to be larger when EFAO is large. This will be the case on sunny days when shading effects may be expected along with larger differences in urban surface temperatures and therefore in longwave radiation exchange. To analyse the aforementioned differences it is instructive to split Equation (2) into a radiation component Erad and an aerodynamic component Ewind such that:

g Ta900 0:408DA þ273 u2 D þ D þ gð1 þ 0:34u2 Þ D þ gð1 þ 0:34u2 Þ (4)

3.5. Synthesis of results The evaporation estimates from EC in Arnhem and scintillometry in Rotterdam can be further assessed in an international context within the framework described by Loridan and Grimmond [16]. The average ratios of midday evaporation to all-wave downwelling radiation from various cities appear to collapse into a simple function of the so-called active vegetation index. We assume that the vegetation fractions in the flux footprint of Arnhem (0.12) and Rotterdam (0.16) are a reasonable first estimate of the active vegetation index, at least during the summer months with fully developed leaves (JuneeSeptember). It is shown in Fig. 9 that midday evaporation during these months scaled to downward allwave radiation is reasonably consistent with the relationship given in Ref. [16], with the wetter summer of 2012 leading to somewhat higher ratios. The analysis also suggest that the part of the framework used here predicts the impact of urban vegetation on neighbourhood-scale evaporation reasonably well on average,

3.0 2.5 2.0 1.5 1.0 0.5

<----- Rooftop stations----->

Vlaardingen

Ridderkerk

Hoogvliet

South

Rijnhaven

East

Centre

Ommoord

Bernisse

Capelle

Spaansepolder

0.0

Reference

Reference evaporation (mm day-1)

Fig. 8 shows the computed average EFAO for the urban stations and the reference station split into Erad and Ewind. In addition, EMAK has been included by means of the dashes. The figure clearly reveals the generally higher reference evaporation at the rural site. An exception is the higher EMAK at the Bernisse station. Here, the average incoming solar radiation was comparable to the one at the reference station (192 versus 188 Wm
2), but the mean air temperature was somewhat higher, in particular the minimum temperature (11.4 versus 9.8  C). The difference in reference evaporation between the city and the rural site tends to be larger for the ground stations than for the rooftop stations. The lower wind speed in the city causes a lower Ewind at the urban sites, which explains 76% of the mean difference with the rural EFAO reference. In addition, Ewind explains more of the variance in EFAO than does ERAD (69% versus 11%), even if station Rijnhaven (located in the relatively open harbour space near a water surface) is disregarded (58% versus 24%). This suggests the wind speed to be

Lansingerland

EFAO ¼ Erad þ Ewind ¼

an important cause of the intra-urban differences in EFAO. Unfortunately, local wind fields in the urban environment are hard to assess on a routine basis, rendering attempts to correct for differences in Ewind between the reference grassland and specific locations in the urban environment prone to errors. EMAK is insensitive to wind speed, reflected shortwave radiation and net longwave radiation. However, owing to the reduction of incoming shortwave radiation (26.2 Wm
2 on average) the main differences in EMAK are not compensated by reduced reflection (20.4 Wm
2 on average), like is the case in EFAO via ERAD. In spite of some compensation by the higher urban temperatures, the total average reduction of 0.30 mm day
1 is almost equal to the average effect of Ewind in EFAO (0.31 mm day
1) and is much larger than the effect of ERAD (0.10 mm day
1). Even using EMAK as the reference evaporation would not solve the heterogeneity issue, mainly because of differences in incoming radiation. However, because shading effects of urban structures can be readily described for specific locations [54] corrections might be attempted in this case.

<-Ground stations->

Fig. 8. Average reference evaporation from meteorological observations in Rotterdam. The bars are the sum of Ewind (dark lower parts) and Erad (light upper parts) (see Equation (4)), thus representing EFAO (Equation (2)). The dashes indicate EMAK (Equation (3)). A distinction has been made between rooftop stations (lighter green shading) and ground stations (darker green shading). Within each station class, the results have been ordered by decreasing Ewind. See Ref. [3] for information on the urban meteorological stations. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Ratio of evaporation to downwelling all-wave radiation

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

0.12 0.10 0.08 0.06 0.04 0.02 0.00 0.00

0.10

0.20

0.30

0.40

Active vegetation index Fig. 9. Observed midday evaporation (
/þ three hours around solar noon) for dry days, scaled to downward all-wave radiation, as proposed in Ref. [16]. Shown are median values for Arnhem (squares) and Rotterdam (circle) for the summer half-year of 2012 (dark grey) and 2013 (light grey). The error bars denote the 95% confidence interval. The dashed line represents the first part of Equation (27) from Ref. [16], which summarizes the ratio of evaporation to downward all-wave radiation at low (up to 0.43) active vegetation index (y ¼ 0.11e0.2$MAX(0.43-x,0)).

3.0

90

2.5

75

2.0

60

1.5

45

1.0

30

0.5

15

0.0

Sapflow (l day-1)

Eact (mm day-1)

suggesting it may have some potential to be used in urban planning. Fig. 10 compares the measured evaporation from all techniques and all sites to EMAK, after binning the data from each technique into six classes of EMAK of about equal sample size. While the sapflow increases linearly with EMAK, the urban evaporation from Arnhem tends to decrease. There is hardly any relation between EMAK and the evaporation measured in Rotterdam. The forest largely reveals a relation between evaporation and EMAK that is consistent with the one from the sapflow observations in Rotterdam, in spite of the possible influence of interception in the case of the forest [48]. The results suggest that the concept of reference evaporation will fail to describe actual urban evaporation at neighbourhood to city scale and over an entire season with wet and dry periods and as such will not be a useful concept in the framework of Dutch water

0 0

2

4

6

37

management. Like for forests, urban evaporation is probably strongly influenced by surface wetness or interception, at the expense of atmospheric influence on evaporation [48]. In cities, such effects may be present on dry days as well, because of a large share of open water (like in Rotterdam) or a slowly drying interception reservoir (like in Arnhem). At neighbourhood to city scale, the contribution of dry, unstressed and well-watered urban vegetation to evaporation will often be small. Nevertheless, the response of the sapflow of trees to EMAK would suggest that the reference evaporation concept is a useful starting point to estimate actual water requirements of urban vegetation spots, like suggested in Refs. [11,18,19]. However, reference evaporation is defined for extensive grassland, while the urban landscape is typically a small-scale, artificial heterogeneous landscape, creating variability in local conditions and strong horizontal advection effects [55,56] that violate the assumptions underlying the concept. Our results on EFAO and EMAK challenge the implicit assumption that reference evaporation accounts for the effect of weather on evaporation in the city [44]. They are consistent with those from recent work showing that, the other way around, urban weather data cannot be used to estimate reference evaporation for agricultural applications [57]. Since effects of weather are implicit in crop factor KC related to EFAO [58], our results also imply significant variation of KC within one city even for a specified vegetation type. Application of corrections factors to account for microclimatic effects has been proposed [11,56]. Although defining microclimatic factors and crop factors for urban vegetation is difficult, encouraging results have been reported [11,18,19]. We conclude that estimation of urban evaporation from routine weather data using the concept of reference evaporation is a particularly challenging task. In view of conceptual as well as practical difficulties, further research is required to establish whether operational evaporation estimates based on the concept of reference evaporation are sufficiently robust in the urban context. Operationalization of frameworks specifically linking urban characteristics to surface-independent meteorological observations, like in Ref. [16], and possibly connected to remote sensing data, may ultimately prove to yield more robust estimates of urban evaporation at the neighbourhood to city scale and at seasonal timescales. However, such frameworks obviously require testing and fine-tuning during development. The present work shows that the micrometeorological techniques applied here, scintillometry and eddy covariance, can be applied for that purpose, in particular if linked with observations at the point scale, like sapflow measurements. Acknowledgements This research was part of “Climate Proof Cities”, carried out in the second phase of the Knowledge for Climate Program, cofinanced by the Dutch Ministry of Infrastructure and the Environment. It is also part of the strategic research program KBIV ‘Sustainable spatial development of ecosystems, landscapes, seas and regions’, funded by the Dutch Ministry of Economic Affairs, Agriculture and Innovation, and carried out by Wageningen University and Research Centre (Project KB-14-002-005). We thank two anonymous reviewers for their useful comments that helped us to improve the manuscript.

Emak (mm day-1) References Fig. 10. Actual evaporation (Eact, left axis) or sapflow (right axis, circles) as a function of EMAK computed from local observations at or near the observation sites. Data are binned in classes of EMAK of about equal sample size (n) per site. Eact was derived from EC measurements in Arnhem (squares, n ¼ 42 or 43) and Loobos (triangles, n ¼ 32 or 33), respectively, and from scintillometer data obtained in Rotterdam (diamonds, n ¼ 24). Bin sample size n of sapflow data is 20 or 21. Vertical error bars denote ±1 SE.

[1] Oke TR. The energetic basis of the urban heat-island. Q J Roy Meteor Soc 1982;108:1e24. [2] Bowler DE, Buyung-Ali L, Knight TM, Pullin AS. Urban greening to cool towns and cities: a systematic review of the empirical evidence. Landsc Urban Plan 2010;97:147e55.

38

C. Jacobs et al. / Building and Environment 83 (2015) 27e38

[3] Van Hove L, Jacobs CMJ, Heusinkveld BG, Elbers JA, Van Driel B, Holtslag AAM. Temporal and spatial variability of urban heat island and thermal comfort within the Rotterdam agglomeration. Build Environ 2015;83:91e103. [4] Steeneveld GJ, Koopmans S, Heusinkveld BG, van Hove LWA, Holtslag AAM. Quantifying urban heat island effects and human comfort for cities of variable size and urban morphology in the Netherlands. J Geophys Res-Atmos 2011;116:D20129. [5] Heusinkveld BG, Steeneveld GJ, van Hove LWA, Jacobs CMJ, Holtslag AAM. Spatial variability of the Rotterdam urban heat island as influenced by urban land use. J Geophys Res-Atmos 2014;119:677e92. [6] Georgi B, Swart R, Marinova N, Van Hove B, Jacobs C, Klostermann J, et al. Urban adaptation to climate change in Europe. Challenges and opportunities for cities together with supportive national and European policies. Copenhagen: European Environmental Agency; 2012. EEA Report. [7] Gill SE, Handley JF, Ennos AR, Pauleit S. Adapting cities for climate change: the role of the green infrastructure. Built Environ (1978-) 2007;33:115e33. [8] Armson D, Rahman MA, Ennos AR. A comparison of the shading effectiveness of five different street tree species in Manchester, UK. Arboric Urban 2013;39: 157e64. [9] Steeneveld GJ, Koopmans S, Heusinkveld BG, Theeuwes NE. Refreshing the role of open water surfaces on mitigating the maximum urban heat island effect. Landsc Urban Plan 2014;121:92e6. [10] Coutts AM, Tapper NJ, Beringer J, Loughnan M, Demuzere M. Watering our cities: the capacity for water sensitive urban design to support urban cooling and improve human thermal comfort in the Australian context. Prog Phys Geog 2013;37:2e28. [11] Nouri H, Beecham S, Kazemi F, Hassanli AM. A review of ET measurement techniques for estimating the water requirements of urban landscape vegetation. Urban Water J 2012;10:247e59. [12] Brolsma R, Buma J, Van Meerten H, Dionisio M, Elbers J. Effect van droogte op stedelijk gebied. Kennisinventarisatie [Effect of drought in urban areas. Knowledge inventory]. Utrecht: Knowledge for Climate; 2012. [13] Albers RAW, Blocken B, Bosch PR, Van Den Dobbelsteen AAJF, Van Hove LWA, Spit TJM, et al. Overview of challenges and achievements in the Climate Proof Cities program. Build Environ 2015;83:1e10. [14] Allen RG, Pereira LS, Raes D, Smith M. Crop evapotranspiration e guidelines for computing crop water requirements. Rome: FAO; 1998. FAO Irrigation and Drainage Paper No 56. [15] Grimmond C, Christen A. Flux measurements in urban ecosystems. Flux-LettereNewsetter Fluxnet 2012;5:1e7. [16] Loridan T, Grimmond CSB. Characterization of Energy flux partitioning in urban environments: links with surface seasonal properties. J Appl Meteorol Clim 2012;51:219e41. [17] De Bruin HAR, Lablans WN. Reference crop evapotranspiration determined with a modified Makkink equation. Hydrol Process 1998;12:1053e62. [18] DiGiovanni K, Montalto F, Gaffin S, Rosenzweig C. Applicability of classical predictive equations for the estimation of evapotranspiration from urban green spaces: green roof results. J Hydrol Eng 2013;18:99e107. [19] Nouri H, Beecham S, Hassanli AM, Kazemi F. Water requirements of urban landscape plants: a comparison of three factor-based approaches. Ecol Eng 2013;57:276e84. [20] Statistics Netherlands. CBS Statline; 2013. www.cbs.nl [accessed 27.11.13]. [21] Eurostat. City statistics e urban audit; 2013. epp.eurostat.ec.europa.eu/portal/ page/portal/region_cities/city_urban [accessed 27.11.13]. [22] Municipality of Arnhem. Groen Green. Made in [Arnhem] (Mixed dutch and English); 2009. Available from: www.arnhem.nl [accessed 27.11.13]. [23] KNMI (Royal Netherlands Meteorological Institute). 2013: www.knmi.nl/ klimatologie [accessed 05.12.13]. [24] Municipality of Rotterdam. Economische kerngegevens rotterdam 2012 (in Dutch); 2013. Available from: www.rotterdam.nl [accessed 27.11.13]. [25] Aubinet M, Vesala T, Papale D, editors. Eddy covariance. A practical guide to measurement and data analysis. Dordrecht: Springer Netherlands; 2012. [26] Schuepp PH, Leclerc MY, MacPherson JI, Desjardins RL. Footprint prediction of scalar fluxes from analytical solutions of the diffusion equation. Bound-Lay Meteorol 1990;50:355e73. [27] Feigenwinter C, Vogt R, Christen A. Eddy covariance measurements over urban areas. In: Aubinet M, Vesala T, Papale D, editors. Eddy covariance a practical guide to measurement and data analysis. Dordrecht: Springer Netherlands; 2012. pp. 377e97. [28] Aubinet M, Grelle A, Ibrom A, Rannik U, Moncrieff J, Foken T, et al. Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology. Adv Ecol Res 2000;30:113e75. [29] Reichstein M, Falge E, Baldocchi D, Papale D, Aubinet M, Berbigier P, et al. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Glob Change Biol 2005;11: 1424e39. [30] MPI Jena. Eddy covariance gap-filling & flux-partitioning tool; 2013.. In: www.bgc-jena.mpg.de/~MDIwork/eddyproc/index.php [accessed 05.12.13].

[31] Allen RG, Pereira LS, Howell TA, Jensen ME. Evapotranspiration information reporting: I. Factors governing measurement accuracy. Agr Water Manage 2011;98:899e920. [32] Elbers JA, Jacobs CMJ, Kruijt B, Jans WWP, Moors EJ. Assessing the uncertainty of estimated annual totals of net ecosystem productivity: a practical approach applied to a mid latitude temperate pine forest. Agr For Meteorol 2011;151: 1823e30. [33] Moors EJ. Water use of forests in The Netherlands. Wageningen: Wageningen UR Alterra; 2012. [34] Bruin HAR, Hurk BJJM, Kohsiek W. The scintillation method tested over a dry vineyard area. Bound-Lay Meteorol 1995;76:25e40.
ski M, Fortuniak K, Pawlak W. Turbulent sensible heat flux in Ło
d
[35] Zielin z obtained from scintillometer measurements e comparison of free and mix algorithms. Contemp Trends Geoscience 2012;1:109e17. [36] Wood C, Kouznetsov R, Gierens R, Nordbo A, J€ arvi L, Kallistratova M, et al. On the temperature structure parameter and sensible heat flux over Helsinki from sonic anemometry and scintillometry. J Atmospheric Ocean Technol 2013;30. [37] Lagouarde J-P, Irvine M, Bonnefond J-M, Grimmond C, Long N, Oke T, et al. Monitoring the sensible heat flux over urban areas using large aperture scintillometry: case study of Marseille city during the ESCOMPTE experiment. Boundary-Layer Meteorol 2006;118:449e76. [38] Gouvea M. Spatially integrated measurements of sensible heat flux using scintillometry. In: Nineth symposium on the urban environment; 2010. [39] Hartogensis O, Watts C, Rodriguez J, De Bruin H. Derivation of an effective height for scintillometers: La Poza experiment in Northwest Mexico. J Hydrometeorol 2003;4:915e28. [40] Meijninger WML, Green AE, Hartogensis OK, Kohsiek W, Hoedjes JCB, Zuurbier RM, et al. Determination of area-averaged water vapour fluxes with large aperture and radio wave scintillometers over a heterogeneous surface e Flevoland field experiment. Bound-Lay Meteorol 2002;105:63e83. [41] Foken T. The energy balance closure problem: an overview. Ecol Appl 2008;18:1351e67. [42] Grimmond C, Oke TR. Heat storage in urban areas: local-scale observations and evaluation of a simple model. J Appl Meteorol 1999;38:922e40. [43] Granier A. Evaluation of transpiration in a Douglas-fir stand by means of sap flow measurements. Tree Physiol 1987;3:309e20. [44] Katerji N, Rana G. FAO-56 methodology for determining water requirement of irrigated crops: critical examination of the concepts, alternative proposals and validation in Mediterranean region. Theor Appl Climatol 2013:1e22. [45] Macdonald RW. Modelling the mean velocity profile in the urban canopy layer. Bound-Lay Meteorol 2000;97:25e45. [46] Grimmond C, Oke T. Evapotranspiration rates in urban areas. IAHS-AISH P 1999;259:235e44. [47] Berthier E, Dupont S, Mestayer P, Andrieu H. Comparison of two evapotranspiration schemes on a sub-urban site. J Hydrol 2006;328:635e46. [48] Shuttleworth WJ, Calder IR. Has the Priestley-Taylor equation any relevance to forest evaporation? J Appl Meteorol 1979;18:639e46. [49] Gill SE, Rahman MA, Handley JF, Ennos AR. Modelling water stress to urban amenity grass in Manchester UK under climate change and its potential impacts in reducing urban cooling. Urban For Urban Green 2013;12:350e8. [50] Rahman M, Smith J, Stringer P, Ennos A. Effect of rooting conditions on the growth and cooling ability of Pyrus calleryana. Urban Urban Gree 2011;10: 185e92. [51] Peters EB, Hiller RV, McFadden JP. Seasonal contributions of vegetation types to suburban evapotranspiration. J Geophys Res 2011;116:G01003. [52] Peters EB, McFadden JP, Montgomery RA. Biological and environmental controls on tree transpiration in a suburban landscape. J Geophys Res-Biogeo 2010:115. [53] Teuling AJ, Seneviratne SI, Stockli R, Reichstein M, Moors E, Ciais P, et al. Contrasting response of European forest and grassland energy exchange to heatwaves. Nat Geosci 2010;3:722e7. [54] Matzarakis A, Rutz F, Mayer H. Modelling radiation fluxes in simple and complex environmentsdapplication of the RayMan model. Int J Biometeorology 2007;51:323e34. [55] Spronken-Smith RA, Oke TR, Lowry WP. Advection and the surface energy balance across an irrigated urban park. Int J Climatol 2000;20:1033e47. [56] Bieringer PE, Annunzio AJ, Platt N, Bieberbach G, Hannan J. Contrasting the use of single-realization versus ensemble-average atmospheric dispersion solutions for chemical and biological defense analyses. J Appl Meteorology Climatol 2014;53:1399e415. [57] Luo Y, Jiang Y, Peng S, Khan S, Cai X, Wang W, et al. Urban weather data to estimate reference evapotranspiration for rural irrigation management. J Irrig Drain E-ASCE 2012;138:837e42. [58] Shuttleworth W, Wallace J. Calculating the water requirements of irrigated crops in Australia using the Matt-Shuttleworth approach. T ASABE 2009;52: 1895e906.