The implementation of biofiltration systems, rainwater tanks and urban irrigation in a single-layer urban canopy model

Urban Climate 10 (2014) 148–170

Contents lists available at ScienceDirect

Urban Climate journal homepage: www.elsevier.com/locate/uclim

The implementation of biofiltration systems, rainwater tanks and urban irrigation in a single-layer urban canopy model M. Demuzere a,c,⇑, A.M. Coutts b,c, M. Göhler d, A.M. Broadbent b,c, H. Wouters a,e, N.P.M. van Lipzig a, L. Gerbet b,c a

KU Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, 3001 Leuven, Belgium School of Earth, Atmosphere and Environment, Monash University, Melbourne, Victoria, Australia c CRC for Water Sensitive Cities, Australia d Department of Computational Hydrosystems, UFZ Helmholtz Centre for Environmental Research, Leipzig, Germany e Department of Environmental and Atmospheric Modelling, Flemish Institute for Technological Research (VITO), Mol, Belgium b

a r t i c l e

i n f o

Article history: Received 9 April 2014 Revised 24 October 2014 Accepted 29 October 2014

Keywords: Climate sensitive urban design Water sensitive urban design Urban evapotranspiration Biofiltration systems Urban irrigation Community Land Model-Urban (CLMU)

a b s t r a c t Urban vegetation is generally considered as a key tool to modify the urban energy balance through enhanced evapotranspiration (ET). Given that vegetation is most effective when it is healthy, stormwater harvesting and retention strategies (such as water sensitive urban design) could be used to support vegetation and promote ET. This study presents the implementation of a vegetated lined bio-filtration system (BFS) combined with a rainwater tank (RWT) and urban irrigation system in the single-layer urban canopy model Community Land Model-Urban. Runoff from roof and impervious road surface fractions is harvested and used to support an adequate soil moisture level for vegetation in the BFS. In a first stage, modelled soil moisture dynamics are evaluated and found reliable compared to observed soil moisture levels from biofiltration pits in Smith Street, Melbourne (Australia). Secondly, the impact of BFS, RWT and urban irrigation on ET is illustrated for a two-month period in 2012 using varying characteristics for all components. Results indicate that (i) a large amount of stormwater is potentially available for indoor and outdoor water demands, including irrigation of urban vegetation, (ii) ET from the BFS is an order of magnitude larger compared to the contributions from the impervious surfaces, even though the former only covers 10%

⇑ Corresponding author at: KU Leuven, Department of Earth and Environmental Sciences, Celestijnenlaan 200E, 3001 Leuven, Belgium. E-mail address: [email protected] (M. Demuzere). http://dx.doi.org/10.1016/j.uclim.2014.10.012 2212-0955/Ó 2014 Elsevier B.V. All rights reserved.

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

149

of the surface fraction and (iii) attention should be paid to the cover fraction and soil texture of the BFS, size of the RWT and the surface fractions contributing to the collection of water in the RWT. Overall, this study reveals that this model development can effectuate future research with state-of-the-art urban climate models to further explore the benefits of vegetated biofiltration systems as a water sensitive urban design tool optimised with an urban irrigation system to maintain healthy vegetation. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction Significant research has been undertaken on biophysical processes governing the urban climate, but often this research does not take into account the knowledge and needs of urban planners and policy makers. Urban planning strategies target issues such as housing, transport and infrastructure but very few strategies comprehensively consider the urban climate and its interaction with the built environment (Coutts et al., 2010). As a way forward, Climate-Sensitive Urban Design (CSUD) is often mentioned as a comprehensive way of designing cities by skilfully combining measures in the areas of urban form, ventilation, solar radiation, natural water cycle and vegetation to mitigate urban heat and create thermally comfortable, attractive and more sustainable urban environments (e.g. Emmanuel, 2005). Within the framework of CSUD, vegetation is regarded as a key tool as it can provide shade and evaporative cooling, enhance thermal comfort, improve air and runoff water quality and reduce storm water runoff intensity (Grimmond et al., 2010; Bowler et al., 2010; Coutts et al., 2013; Demuzere et al., 2014). However, for vegetation to provide these benefits it must be healthy and well supplied with water. As such, water sensitive urban design (WSUD) should be considered to help support vegetation in urban areas. Impervious land cover in built-up areas typically prevents infiltration as stormwater drainage networks are designed to rapidly remove runoff away from the city. This results in a soil moisture deficit that is, especially in dry environments, balanced by imported potable water for irrigation. If potable water restrictions are enforced during periods of drought, vegetation health can be compromised. Implementing WSUD can support the restoration of the natural water balance promoting more healthy urban vegetation and purposefully modify the urban energy balance to support CSUD through enhanced evapotranspiration (Coutts et al., 2013). In this respect, biofiltration systems (also commonly referred to as biofilters, bioretention systems, tree-pits and rain gardens) are often mentioned as a WSUD strategy suited to improve water quality by filtering water through biologically influenced media and to reduce stormwater runoff flow rates and volumes aiming to restore a more natural water balance, protect downstream receiving waters and encourage water loss via subsurface flows and evapotranspiration (Walsh et al., 2005; Fletcher et al., 2013; Hamel and Fletcher, 2014). In vegetated biofiltration systems, plants are included to enhance the removal of moisture and pollutants from the soil to further improve water quality. It is anticipated that the increase in infiltration and evapotranspiration originating from biofiltration systems could have positive benefits for urban climate, due to an optimal soil moisture content for evapotranspiration as a function of vegetation type. Although the evaporative term is key to this, coupling both the energy and water balances (Hamdi et al., 2009), the PILPS-urban international comparison of urban models (Grimmond et al., 2010, 2011) revealed that even complex models very poorly model or even neglect the evaporative term and generally model vegetated or pervious fractions as independent natural fractions outside the street canyon that do not interact with street canyon properties. In addition, they rarely take into account detailed urban hydrological processes such as runoff, infiltration, interception or irrigation of which the latter is stated as a critical process that needs to be taken into account by urban models (Best and Grimmond, 2014).

150

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Until now only few studies quantified the role of water-use practises and vegetation as cooling strategies. Mitchell et al. (2007) used the water balance model Aquacycle (Mitchell et al., 2001) in conjunction with the energy balance model SUES (Grimmond and Oke, 1991) to study the impact of a number of different WSUD and water-use scenarios on the urban water balance and the micro-climate in a suburb in Canberra (Australia). Compared to a desert landscape (no vegetation present), the full vegetated WSUD treatment train was able to increase summer evaporation E by 1.44 to 1.76 mm day1. Gober et al. (2009) used the local-scale urban meteorological parametrisation scheme (LUMPS) (Grimmond and Oke, 2002) to examine the variation in temperature and evaporation for ten census tracts in Phoenix‘s urban core (USA). For each of three water use scenarios they found a change in night-time cooling through a more efficient use of outdoor water compared to the base-case scenario. A number of studies also used the more sophisticated Weather Research and Forecasting (WRF) model, the Noah LSM (Chen et al., 1997) and the urban canopy model of Kusaka and Kimura (2004) to investigate the effect of urban landscape irrigation on near-surface air temperatures and energy fluxes. Using this framework, Grossman-Clarke et al. (2010) reported for the Phoenix area a maximum daytime latent heat flux of 500 Wm2 for irrigated agricultural land compared to about 20 and 75 Wm2 for commercial/industrial and xeric residential areas. For the same area of interest, Georgescu et al. (2011) indicated a shift toward greater latent heating for regions undergoing a change to irrigated agriculture but their results do not indicate a significant impact on near-surface temperatures because of irrigation within urban areas. Finally, Coleman et al. (2010) showed that for the Los Angeles basin (USA), the addition of anthropogenic soil moisture was able to reduce the modelled daytime surface temperature bias, although it did worsen the cold bias and the model’s skill during the nighttime hours. Nevertheless they state that anthropogenic moisture is an important source of enthalpy and humidity that it should be represented in models, especially for dry areas characterized by a natural precipitation amount comparable to or less than anthropogenic moisture. The aforementioned studies not only provide preliminary insights into the role of (irrigated) vegetation with respect to changes in the partitioning of the surface energy balance and urban cooling, their limited number and approaches also depict current limitations in modelling urban irrigation and stormwater reintegration in complex street canyon geometries. From a hydrological point of view, more advanced hydrological models such as Aquacycle are available with a sophisticated representation of the urban hydrological cycle (including irrigation) but do not resolve the complexity of the urban canopy layer. The Surface Urban Energy and Water Balance Scheme (SUEWS) also has a sophisticated parametrisation of the urban water balance (including irrigation) but uses a relatively simplistic bulk approach for the urban canopy layer which represents the urban form as a flat horizontal surface (Järvi et al., 2011; Järvi et al., 2014). On the other hand, the majority of more complex urban canopy models currently do not represent urban irrigation although this has been pointed out by Loridan et al. (2010) as a deficiency of urban canopy models in highly urbanised areas. The Noah LSM – WRF based studies described above did not explicitly take into account irrigation, but adjusted soil moisture contents for urban vegetation and agricultural land to a reference soil moisture for one or more soil layers (Grossman-Clarke et al., 2010; Georgescu et al., 2011) or applied a pre-processed spatially-distributed anthropogenic moisture dataset at the surface on various time steps (Coleman et al., 2010). In recent developments, a study by Vahmani and Hogue (2014) presents a new irrigation module in SLUCM that updates the top soil layer moisture content to the irrigated soil moisture content at a selected interval. Based on MODIS-Landsat ET observations they report that the incorporation of the irrigation module enhances the Noah SLUCM performance in terms of ET over the Los Angeles region compared to a simulation without irrigation. A similar conclusion is drawn for the land surface temperature biases for parks and heavily vegetated areas, while the simulation still showed negative biases for highly urbanised areas. They also conclude that further research is needed to quantify the impacts of irrigation on energy and water cycles over broader urban domains. As an ongoing effort to bridge the gap between the urban hydrological and energy balance models, the state-of-the-art single-layer urban canopy model Community Land Model-Urban (CLMU) (Oleson et al., 2008a; Oleson et al., 2008b) is extended with a vegetated lined biofiltration system (hereafter referred to as BFS) combined with a rainwater storage tank (RWT) and a simple but flexible sub-surface irrigation system (Section 2). In order to evaluate and test the implementations provided in Section 2.2, CLMU was simulated driven by observed meteorological forcing (Section 3.1) for a two-month period in

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

151

2012 during which an intensive observation campaign was performed in the City of Yarra, Melbourne (Australia) (Section 3). This site has been selected because of its availability of BFS soil moisture data (Section 3.2) and the relevance of the Australian city Melbourne itself, that has been challenged with extreme weather events in the past and likely to experience an increase of such events in the future (Coutts et al., 2013). In a first phase, the uncertainty of the modelled soil moisture with respect to the soil hydraulic properties is evaluated (Section 4.1). A second phase addresses the results of the impact of the BFS, RWT and urban irrigation on urban evapotranspiration for various scenarios (Section 4.2). General findings of this study are discussed and summarised in Section 5. 2. Model description and simulation strategy The Community Land Model Urban is part of the Community Land Model version 4 (CLM4) which is the land-surface scheme embedded in the Community Earth System Model (CESM) (Oleson et al., 2010b; Lawrence and Chase, 2010). CLM4 applies a nested sub-grid hierarchy in which each grid cell represents up to five land units: wetlands, glaciers, vegetation, water and urban (Oleson et al., 2010a, their Fig. 1.2). At first some relevant default characteristics of CLMU are provided (see Section 2.1), while afterwards changes to these default parametrisations are described in more detail (see Section 2.2). For further information on the full model description, the reader is referred to Oleson et al. (2008b, 2010a). 2.1. Default CLMU characteristics The representation of the urban land unit follows the concept of Oke (1987) in which the considerable complexity of an urban environment is reduced to a single-layer urban canyon that consists of a canyon floor and a roof (with width of the canyon W and height of the buildings H, respectively) (Fig. 1A). The urban land unit consists of five columns: roof, sunlit and shaded wall, impervious and pervious road. The walls are hydrologically inactive, while liquid and solid precipitation (ponding and dew) can be intercepted, stored and evaporated from the roof and canyon floor (both impervious and pervious road). By default, the maximum amount of ponding (also referred to as interception storage, Wpond;max ) for the roof and impervious road columns is set to 1 mm, while this is 10 mm of liquid and solid precipitation for the original pervious road fraction that represents surfaces such as residential lawns and parks which may have active hydrology (Fig. 1A) (Oleson et al., 2008b; Demuzere et al., 2013; Wouters et al., 2014). The pervious fraction acts as a bare soil with roots and allows for infiltration, sub-surface drainage, redistribution of water within the soil column, interaction with an unconfined aquifer and evaporation from all soil layers (10 soil layers and five hydrologically inactive bedrock layers). The advantage of such a modelling framework is that the pervious fraction is an integral part of the urban canyon and will thus interact with urban canyon air properties such as humidity and temperature. Because of the additional computational and data requirements, vegetation is currently not explicitly represented in the street canyon and processes such as interception and shadowing by this vegetation are not yet integrated (Oleson et al., 2010a; Demuzere et al., 2013). 2.2. Modifications to CLMU In this study, the pervious road fraction is redesigned to represent a lined BFS (Figs. 1B and 2) (Section 2.2.1). In addition, a rainwater tank (RWT) approach and an urban irrigation scheme are implemented to collect and reintegrate surface runoff from the rooftops and the impervious road to maintain an optimal soil moisture content for the BFS vegetation (Section 2.2.2). During rain events stormwater will enter the BFS while irrigation is needed during dry episodes to allow for actively transpiring vegetation (Figs. 1B and 2). 2.2.1. Implementation of a lined biofiltration system Biofiltration systems can be lined or unlined depending on the way exfiltration is dealt with. The unlined system allows for exfiltration on all sides of the system. The lined system does not allow for

152

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Fig. 1. Plan and cross-sectional view of the urban land unit in CLMU: (A) a default configuration with a pervious road fraction that interacts with an unconfined aquifer and (B) a configuration with a biofiltration system, rainwater tank and urban irrigation. H refers to roof height, W to canyon width and W p to the fraction of the canyon road that contains the pervious road fraction/biofiltration system. The orange lines refer to runoff from the roof and impervious road, dark red the excess runoff from the rainwater tank to the biofiltration system and from the BFS detention pond to the subsurface drainage, green the irrigation trajectories from both the rainwater tank and piped water supply (referred to as external water) and purple (light yellow) the drainage of (un-)filtered water at the bottom of the biofiltration system. The full arrows in the lower left panel refer to the (direction of) transfer of water in excess of the maximum ponding limit of the roof and impervious road. The dashed arrows refer to the water in excess of the maximum water storage of the impervious surfaces and the detention pond of the BFS. The building (facets) are indicated in the lower left panel in their corresponding colour code. See Fig. 2 for more detail on the rainwater tank, the biofiltration system and the direction of the water transfers. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

exfiltration and is generally used at sites where there is a need to protect build infrastructure or where one would like to avoid interactions with ground water (FAWB, 2009). This study opts for a lined biofiltration system with an underdrain (Fig. 2), as it aims to quantify the amount of water that is needed to maintain healthy vegetation and the gain in evaporation rate in a densely built-up street canyon. The hydrology of the lined biofiltration filter initially follows the hydrological component of the pervious road (Oleson et al., 2010a) in which changes are implemented with respect to the drainage term, water uptake, evaporation and soil hydraulic properties. The original pervious fraction is characterised by the following water balance:

DW sno þ

NX lev soi

ðDW liq;i þ DW ice;i Þ ¼ ðqrain þ qs7no  Eperv ious  qov er  qdrain ÞDt

ð1Þ

i¼1

where DW sno ; DW liq and DW ice are changes in snow water, soil water and ice respectively (all in mm). The index i is used for the vertical discretization of the soil (based on a number of hydrologically active soil layers (Nlev soi ¼ 10). Dt is the time step (in seconds), qrain and qsno are the liquid and solid precipitation respectively, qov er the surface runoff, qdrain the drainage and Eperv ious the total evaporation from the pervious fraction of the canyon (all in mm s1). Compared to the original Eq. (5.1) in Oleson et al.

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

153

Fig. 2. Detailed conceptual overview of the lined biofiltration system as implemented in CLMU. The full arrows denote the direction of water transfers while the dashed arrow refers to drainage of both the filtered and unfiltered water. The dotted arrows present the depth of the three main zones of the BFS: detention pond, filter media and drainage layer. Colour codes are the same as in Fig. 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(2010a), Eq. (1) is simplified as there is no interaction with water in an unconfined aquifer and liquid and solid runoff due to snowcapping are not present in urban areas. Following Eq. (3.77) in Oleson et al. (2010a) Eperv ious is defined as:

Eperv ious ¼ qatm ðqac  qg;perv ious Þ=r s;perv ious

ð2Þ

in which qac and qg;perv ious are the urban canopy layer air and pervious ground specific humidity respectively (kg kg1), qatm the density of moist air (kg m3) and r s;perv ious is the surface resistance of the pervious canyon floor fraction (s m1). The surface runoff qov er consists of overland flow due to saturation excess and infiltration excess so that

  W pond;max qov er ¼ max 0; qliq;0   qinfl;max Dt

ð3Þ

where qliq;0 is the liquid water reaching the surface from rain and/or snow melt (mm s1) and W pond;max (mm) a maximum ponding limit of the detention pond (Fig. 2). The variable qinfl;max is the maximum soil infiltration capacity (mm s1) and is a function of soil texture and soil moisture of the top soil layer. Based on the above equations, the actual infiltration qinfl into the biofiltration systems is defined as the residual of the water balance (when no snow is present) so that

qinfl ¼ qliq;0  qov er  qsev a with qsev a the evaporation of liquid water from the top layer (mm s (2010a).

ð4Þ 1

) (see Eq. (3.111) in Oleson et al.

154

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

In CLMU, each soil layer is characterised by a root area fraction that allows water uptake by roots and hence evapotranspiration from the soil. The amount of water that is removed by evapotranspiration (ei ) from each layer i is a function of the total evaporation Eperv ious and the effective root fraction re;i following the original Eq. (5.78) from Oleson et al. (2010a). The effective root fraction re;i is a function of the root area fraction (r i in %), a soil wetness factor per layer (wi , unit less) and a wetness factor for the total soil column (a, unit-less). For more information on the exact definitions of these terms, the reader is referred to Eq. (3.86) and 3.87 in Oleson et al. (2010a). In the default CLMU model, the soil water interacts with a groundwater component in the form of an unconfined aquifer. As the lined BFS does not interact with a water table, free drainage out of the bottom of the soil column is implemented that partly follows Eq. (7.116) in Oleson et al. (2004) resulting in a drainage flux qdrain :

qdrain ¼

wexcess liq

Dt



wdeficit liq

Dt

þ k½zh;10  þ

dk½zh;10  Dhliq;10 : dhliq;10

ð5Þ

wexcess is the amount of liquid water (mm) in excess of saturation in all soil layers, wdeficit the amount of liq liq water required to keep all soil levels above zero liquid water content, k½zh;10  the hydraulic conductivity for layer i ¼ 10 and thus the drainage out of the bottom of biofiltration system (mm s1) and

dk½zh;10  dhliq;10

D

hliq;10 the change in hydraulic conductivity due to the change in liquid water content of layer i ¼ 10 (mm s1). As the drainage term reflects the amount of water that infiltrates through the soil medium, it is susceptible to pollutant removal. At present, these biochemical processes are not yet integrated in CLMU, but this nevertheless provides insight into the amount of runoff water that is filtered by the BFS. Finally, soil hydraulic properties (SHPs) play an important role in land surface models as they control how the water moves through the soil column via relationships between the soil moisture, soil water potential and hydraulic conductivity (Gutmann and Small, 2005). The standard CLMU model uses the Campbell (1974) approach modified by Clapp and Hornberger (1978) and Cosby et al. (1984) in which the SHPs of every soil layer varies with soil texture defined as a percentage of sand and clay. This use of soil texture alone has been shown to be an inadequate method to determine SHPs by Gutmann and Small (2005). They found that the soil texture classes account for only 5% of the variance in energy and moisture fluxes, while the van Genuchten n and saturated hydraulic conductivity parameters alone explain approximately 60% of this variability. Thus we decided to modify the default scheme and implement the Brooks and Corey (1964) model in which the water content is expressed as a power function of the soil water potential W (or soil water pressure head h in the original notations)

( Se;i ¼

ðai Wi Þki 1

Wi < 1=ai Wi P 1=ai

ð6Þ

with Se;i ¼ ðhi  hr;i Þ=ðhs;i  hr;i Þ in which hi ; hr;i and hs;i are the volumetric soil, residual and saturated water contents for layer i respectively (mm3 mm3). Combining the Brooks and Corey (1964) with the capillary model of Mualem (1974), the hydraulic conductivity can be written as 2=ki þlþ2

K i ¼ K s;i Se;i

ð7Þ 1

with K s;i the saturated hydraulic conductivity (mm s ). Parameters ai (air-entry value in mm1, Eq. (6)) and ki (the pore-size distribution index) relate respectively to the inverse of the saturated soil water potential Wsat (mm) and the inverse of the exponent Bi of the Campbell (1974) SHP model. Parameter l is a pore-connectivity parameter assumed to be 2.0 following the original study of Brooks and Corey (1964). Thus the above equations contain five independent parameters a; k; hr ; hs and K s that define the SHPs of the biofiltration system. 2.2.2. Implementation of an urban irrigation system In the street canyon environment, all the runoff from the impervious fraction of the canyon floor (Fig. 1) is directed to the BFS. We also adopt the hypothetical strategy proposed by Burns et al. (2012) whereby the overflow from rainwater tanks is directed to the vegetated infiltration system providing both retention, infiltration and adequate water supply to maintain healthy vegetation (Fig. 2).

155

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

The rainwater tanks collect the stormwater runoff from the roof in which the percentage of roofs contributing to stormwater collection is defined by RWT prop , the proportion of households that actually have a rainwater tank (in %). The actual joint volume of the rainwater tanks is denoted by RWT v ol (mm). Water in excess of this volume is directed to the biofiltration system. Hence, the water balance described in Eq. (1) is extended by two additional sources of water: qRWT;excess and qov er;imperv ious , being respectively the amount of water in excess of the rainwater tank and runoff from the impervious fraction of the canyon floor respectively (in mm s1). In addition, an optimised sub-surface irrigation system is added to maintain a target soil moisture content, following the irrigation module presented in Oleson et al. (2013). Once a day (time step It ) it is checked whether the actual soil moisture content falls below a target soil moisture wtarget;i (mm). The latter is a weighted average of the minimum soil moisture that results in no water stress (wo;i , mm) and the soil moisture at saturation (wsat;i , mm) in that layer:

wtarget;i ¼ ð1  airr Þwo;i þ airr wsat;i

ð8Þ

airr is an irrigation factor set to 0.7 (unitless), an empirically-determined value matching global irrigation amounts to global water use records (Oleson et al., 2013). This value also resembles the irrigation demand factor from Vahmani and Hogue (2014) (chosen to be 0.65) and allows for a flexible adjustment of the irrigation system in case outdoor water use estimates are available. wo;i is determined by inverting Eq. (6) and solving it for hi ¼ ho;i by replacing Wi with Wo and then converting it to mm. Herein, Wi is replaced with Wo first. Afterwards, ho;i is integrated over the depth of layer i to obtain its soil moisture content wo;i . Wo represents the soil water potential when stomata are fully open and can vary according to the type of vegetation used in the BFS system. Global values for the plant functional types available in CLM can be retrieved from Table 8.1 in Oleson et al. (2013), but here we use a specific value for Melaleuca Argentea of 33 m (O’Grady et al., 2005). As such, the deficit Id (mm) is computed over the first seven soil layers as they are characterised by the densest root system: Id ¼

7 X ðmaxðwo;i  wi Þ; 0Þ

ð9Þ

i¼1

Taking the maximum in Eq. (9) means that a surplus in one layer cannot make up for a deficit in another layer. As such, the original water balance (Eq. (1)) can be further extended to include the irrigation term Id :

DW sno þ

NX lev soi

ðDW liq;i þ DW ice;i Þ ¼ ðqrain þ qsno þ qRWT;excess þ qov er;imperv ious  Eperv ious  qov er  qdrain ÞDt þ Id

i¼1

ð10Þ In our approach, the irrigation system is connected to the rainwater tank (Fig. 2). In case Id 6 RWT v ol , then Id is completely taken from the rainwater tank. In case Id > RWT v ol then at first the remaining water from the rainwater tank is consumed (after irrigation RWT v ol ¼ 0) while no irrigation will be applied when RWT v ol ¼ 0 (e.g. water restriction condition). In case no water restrictions would apply during times when RWT v ol ¼ 0, the system allows for water sourced from external sources such as stormwater drainage pipes. Moreover, the timing of the day at which the irrigation starts can be defined via It . Thus this irrigation system allows for a flexible way of irrigating the BFS in which both its water source, volume and timing can be regulated (see also Section 2.3.5). 2.3. Simulation strategy and scenarios The offline CLMU simulations for Smith street (Melbourne) were used in two distinct phases:  Phase I: evaluation of the soil moisture dynamics and uncertainty associated with SHPs. The modelled soil moisture dynamics were evaluated against observed soil moisture data (Section 3.2) and discussed with respect to its sensitivity to the prescribed soil hydraulic parameters. Here, the BFS

156

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

geometry, soil texture and vegetation characteristics are based on the existing BFS in Smith street (Gerbet, 2012) (Table 1). For this phase, the RWT and urban irrigation system were not activated and the results are discussed in Section 4.1.  Phase II: A range of scenarios illustrate the impact of BFS characteristics and irrigation regimes on evapotranspiration. The default values for the various scenarios are mainly derived from the official guidelines on stormwater BFS (Knights, 2009) and previous research (see Sections 2.3.1–2.3.5 for more details). Results from this phase are described in Section 4.2. For the second phase of the simulation strategy, changes in total latent heat flux for various scenarios of the lined BFS and urban irrigation scheme were tested and compared to a base-case scenario. Each scenario is described below in more detail while the scenarios’ respective abbreviations and parameter values are summarised in Tables 1 and 2. 2.3.1. Base-case simulation (‘BC’ scenario) Originally, Smith street had almost no vegetation prior to the implementation of the BFS, hence the choice of the local council to retrofit the street in order to provide some vegetation and a form of stormwater treatment. As such, the base-case scenario is defined as the original street canyon state and is characterised by a completely impervious street canyon floor (Wperv ious in Fig. 1 is 0). In this configuration, evaporation is only originating from water intercepted on the roof and impervious road surfaces. Runoff generated above an excess of the 1.0 mm ponding limit is removed from the model as exported stormwater. 2.3.2. Non-vegetated versus vegetated (‘V’ scenarios) Bratieres et al. (2008) stated that the selection of vegetation in a biofiltration system should not only depend on their efficiency in removing nutrients but also their capacity in surviving potential stressful growth conditions, such as dry environments. In their study they found that e.g. Carex Appressa (a sedge) and Melaleuca Ericifolia (swamp paperbark, a small tree) performed significantly better than other species. As both species are listed as an indicative species for biofiltration systems (Knights, 2009), we tested their influence on the evaporation term originating from the BFS. The V0 scenario presents the case of a non-vegetated BFS, while scenarios V1 and V2 respectively have the Melaleuca Argentea (Abernethy and Rutherfurd, 2001) and Carex Appressa (Emily Payne, Pers. Communication) root characteristics. Scenario V3 combines the experiments V1 and V2 resulting in a BFS filter media with the dense and shallow root system of the Carex Appressa and the deeper but sparsely distributed root system of Melaleuca Argentea (Table 1). For these scenarios runoff from the impervious road fraction flows into the BFS while the RWT and urban irrigation scheme are not yet activated. The latent heat flux thus not only originates from the roof and impervious road fraction but also from the BFS itself, while a drainage term is added that exports the filtered water from the base of the BFS out of the model domain. Unless indicated otherwise, the scenario simulations with BFS use a fraction cover (referring to the percentage of the road fraction taken by the pit) of 10%. 2.3.3. Soil texture (‘T’ scenarios) Bratieres et al. (2008) suggested to use a sandy loam filter media as it provides adequate support for plant growth and minimal leaching (Tsl scenario, see Table 1). In contrast, the adoption guidelines on stormwater biofiltration systems suggested the use of a loamy sand soil texture (experiment V3), which has higher hydraulic conductivity values than a sandy loam filter media (FAWB, 2009). The saturated hydraulic conductivity for loamy sand (0.04 mm s1) agrees well with the desired saturated hydraulic conductivity in the range of 0.02 to 0.05 mm s1 (FAWB, 2009) and corresponds to the lined biofiltration system characteristics modelled by Burns et al. (2012). As these soil types are specifically suggested providing stormwater benefits and not directly urban climate benefits, a loamy soil is also tested (Tsl scenario, see Table 1). The largest differences between the three textures are related to lower values for the Brooks-Corey air-entry pressure and pore size distribution parameters (a and k respectively) and a decreasing saturated hydraulic conductivity (K s ) for an increasing loam content (Table 1). For the three tested soil textures, the recommended mean soil hydraulic parameter values are taken from Meyer et al. (1997).

BFS geometry & characteristics Connected to impervious canyon floor % Pervious of canyon floor Depth detention pond [m] Depth Filter Media [m] Depth Drainage Layer [m] Soil texture characteristics Soil texture 3

3

hr [mm mm ] hs [mm3 mm3] a [mm1] k [mm1] Ks [mm s1] Vegetation characteristics Species % Roots

Evaluation

BC

V0

V1

Yes

No

Yes

Yes

2 0.16 0.7 0.3

0 – – –

10 0.2 0.7 0.3

10 0.2 0.7 0.3

Sand

–

Loam

– – – – –

Sandy loam 0.064 0.41 0.0076 0.89 0.017

Loam

0.047 0.43 0.015 1.67 0.08

Loamy sand 0.057 0.41 0.013 1.27 0.04

0.078 0.43 0.0037 0.56 0.0029

0.078 0.43 0.0037 0.56 0.0029

Eucalyptus Olivacea/Olea Europaea 2

–

–

V1 + V2

V3

V3

V3

–

–

0.1 to 37

0.1 to 37

0.1 to 37

0.1 to 37

Melaleuca Argentea 0.1 to 0.8(a)

V2

Carex Appressa 0 to 37(b)

V3

Tsl

Tl

W p;x

1–100

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Table 1 Parameter settings for the Smith Street evaluation, base-case, vegetation, soil texture and BFS fraction scenarios. Generally, only the parameter values that are changed in between sensitivity experiments are reported. Experiments’ abbreviations are the following: BC = Base case scenario, V0 = no vegetation, V1 = Melaleuca Argentea vegetation, V2 = Carex Appressa vegetation, V3 = V1 + V2 and Tsl & Tl = sandy loam and loam soil texture scenarios respectively. Wp;x refers to the scenarios with a varying fraction of the urban canyon covered by the BFS, where subscript x refers to the % of canyon floor that is pervious. For both the evaluation and the experiments, the drainage layer consists out of gravel. The dimensions of the lined BFS for the scenarios were taken from FAWB (2009), in which a lined BFS typical of practice in Australia has a ponding depth of 0.2 m, a filter media depth of 0.7 m (model layers 1 to 9) and a gravel drainage layer of 0.3 m (lowest hydrological active soil layer) (Table 1).

157

158

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Table 2 Irrigation scenarios matrix. By default, the rainwater tank is empty at the start of each simulation (RWT v ol;ini ), except for scenario I3hf. For scenario I6, irrigation is applied at every model time step in order to continuously reach ET pot . Number

RWT v ol (mm)

RWT v ol;ini (mm)

RWT prop (%)

External water (mm)

It

ID I1 I2 I3 I3hf I4 I5 I6

25 100 25 100 100 25 25 25

0 0 0 0 50 0 0 0

22 22 100 100 100 22 22 22

– – – – – – – Unlimited

8 AM 8 AM 8 AM 8 AM 8 AM 2 PM 8 PM Every model time step

2.3.4. Fraction cover (‘Wp;x ’ scenarios) From historical climate data in Melbourne, Bratieres et al. (2008) derived that biofiltration systems should at least be 2% of their catchment area if one would like to treat (from a chemical point of view) at least 90% of the runoff data. As no details are provided for an optimal pit size in terms of urban climate, a full range of BFS fraction covers (Wp in Fig. 1) is chosen with x varying from 1 to 100%, with intermediate steps of 5, 10, 20, 35, 50, 60, 70, 80 and 90% (Table 1). The F10 scenario is identical to the Tl scenario. 2.3.5. Irrigation regimes (‘I’ scenarios) As the stormwater biofiltration guidelines FAWB (2009) currently only advise on the BFS fraction from a stormwater management perspective but not in terms of their potential role as urban climate mitigation tools, the urban irrigation scenarios were by default performed with a 10% fraction cover, similar to the V and T scenarios and the vegetated BFS filter media was initialised at residual volumetric soil water content. The default irrigation experiment (ID) is characterised by irrigation applied at 8 AM (It ), the use of a rainwater tank volume (RWT v ol ) of 25 mm (Burns et al., 2012) and proportion of rainwater tanks (RWT prop ) of 22%, corresponding to the estimated number of Melbourne households that currently have a rainwater tank (ABS, 2011). Besides the default irrigation scenario ID, a number of additional irrigation scenarios were performed in which the irrigation parameters were changed independently as described in Table 2. In a first scenario I1, the volume of the RWT quadruples to 100 mm. In scenario I2, the number of households contributing to RWT increases to 100% while scenario I3 combines the features of scenarios I1 and I2. Scenario I3hf is similar to I3 but is initialised with a half-full rainwater tank at the beginning of the simulation period (RWT v ol;ini ¼ 50 mm). Scenarios I4 and I5 change the time of irrigation to 2 and 8 PM respectively. Applying irrigation at 8 AM and PM corresponds to morning and evening times where watering systems are generally allowed under permanent water saving rules and Stage 1 & 2 water restrictions in Melbourne (Athuraliya et al., 2012). The 2 PM case is introduced to check whether irrigation during highest solar radiation loads has an effect on the amount of evapotranspiration. For all the above scenarios, irrigation is applied only when water is available in the rainwater tank, which is initially empty (RWT v ol;ini ¼ 0 mm). Since irrigation water is drawn from household tanks only, the frequency of irrigation does not adhere to any water saving restrictions. As such, a daily check for irrigation (as opposed to e.g. alternate days in the stage I water restrictions (Athuraliya et al., 2012)) is allowed to illustrate the impact of this system on evapotranspiration. Scenario I6 represents a case for which irrigation is applied at every model time step to demonstrate the maximum potential ET from the biofiltration system. Here, no water restrictions apply so that there is an unlimited availability of external water (e.g. recycled water) (Table 2). The latter resembles passive irrigation for which harvested water is released at a low flow rate onto the nearby BFS, in support of the stormwater management objectives aim to restore low flow regimes (Hamel and Fletcher, 2014).

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

159

3. Observations and evaluation metrics 3.1. Observed meteorology driving CLMU All evaluation and scenario simulations with CLMU were performed in off-line mode, meaning that the model was forced with local atmospheric data. Recently, Grimmond et al. (2011) and Demuzere et al., 2013 showed that the default CLMU model provided satisfactory results with respect to the modelled radiative and turbulent heat fluxes for the nearby urban flux tower site in Preston, Australia. As such, the model will no longer be evaluated in terms of radiative and turbulent heat fluxes for Smith Street, but attention will be given to the hydrological performance of the biofiltration systems combined with the urban irrigation approach. The atmospheric forcing parameters to run the model are surface pressure [Pa], incident short- and longwave radiation [Wm2], atmospheric temperature [K], wind speed [m s1], precipitation [mm s1] and relative humidity [%]. Meteorological data for Smith Street were available at 30 minute intervals and sourced from the nearby Bureau of Meteorology (BoM) Melbourne Regional Office weather station located one kilometre South-West of Smith Street (37° 480 2700 S, 144° 580 1200 E, station ID 086071). The incoming short- and longwave radiation which was sourced from Monash University Clayton Campus, located 15 km South-East of Smith Street (37° 540 4500 S, 145° 70 5800 E). Simulations were done for a two month period between February 9 and April 9 2012, characterised by a total precipitation sum of 94 mm, a maximum precipitation peak of 4.6 mm in half an hour on the 16th of February, three dry periods with limited or no precipitation persisting for at least ten days and a maximum temperature of 36.3 °C on the 25th of February (not shown). In addition to the meteorological forcing data, CLMU used the thermal and radiative characteristics of building materials from the nearby neighbourhood of Preston as previously used and listed in Tables 1 and 2 of Demuzere et al. (2013). 3.2. Observed soil moisture from BFS systems As part of a study on tree response to urban environmental conditions and water availability within an urban canyon, soil moisture measurements were taken from four biofiltration systems located in a North-South oriented Smith Street canyon (Collingwood 3°470 4300 S, 144°590 0500 E) (Gerbet, 2012). The systems were approximately 1.05 m by 1.05 m in size, with a detention pond of 0.16 m, a soil depth of 0.7 m overlying a 0.3 m gravel drainage layer below (Table 1). As shown in Fig. 3, the stormwater treepits are connected to the gutter and thus have the impervious road fraction as part of their catchment. Trees were planted in the biofiltration systems, alternating between two species: Olea Europaea and Eucalyptus Olivacea (Table 1). Aside from the biofiltration systems, the urban canyon was completely impervious and had a height (H = 7.7 m) to width (W = 19.8 m) ratio of around 0.39. The filtration media (soil) was sand which promoted a high rate of infiltration (Gerbet, 2012). Two soil moisture probes (CS616, Campbell Scientific) were placed vertically in the four biofiltration system to measure the volumetric soil water content integrated over the top 0.3 m of the sandy soil. 3.3. Evaluation metrics The performance of the Brooks and Corey (1964) model is evaluated according to the coefficient of determination (r 2 ), its weighted version (wr2 ) and the modified Nash Sutcliffe efficiency Ej suggested by Krause and Boyle (2005):

( 2

wr ¼

br

2 1

ðbÞ r 2

b61 bP1

ð11Þ

and

Pn ðOi  Pi Þj Ej ¼ 1  Pi¼1 j n i¼1 ðOi  OÞ

ð12Þ

160

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Fig. 3. Photo of a vegetated biofiltration system installed in Smith Street, Collingwood, Melbourne (Australia).

In Eq. (11), b is the gradient of the regression on which r2 is based. For a good agreement between measured Oi and modelled Pi soil water content, the intercept a should be close to zero and the gradient b close to 1. By weighing r2 with the gradient b, the under- or over-predictions are quantified together with the dynamics which results in a more comprehensive reflection of model results. In Eq. (12), j (2 N) is set to 1 in order to provide a better overall evaluation compared to the general Nash Sutcliffe square forms that are more sensitive to significant over- or under-predictions (Krause and Boyle, 2005). The range of Ej lies between 1.0 (perfect fit) and 1. 4. Results Firstly, the soil moisture dynamics from the altered soil hydrology approach in the BFS are evaluated in Section 4.1. Secondly, the modifications of CLMU (viz. the BFS, RWT and urban irrigation) and their impact on the urban hydrology and total urban canopy water vapour flux are illustrated in Sections 4.2.1 & 4.2.2.

161

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

4.1. Phase I: Evaluation and uncertainty of modelled soil moisture

40

0.5

30 20

0.3

0

0.0

0.1

10

0.2

θ [mm³/mm³]

0.4

0.79 0.03 0.89 0.7 0.39

Precipitation [mm / 1800s]

r² = a= b= wr²= Ej =

Precipitation Mean of measured θ [mm³/mm³] 95% CI and mean

50

The performance in modelled soil moisture dynamics resulting from changing the five soil hydraulic model parameters a; k; hr ; hs and K s within their inherent uncertainties is tested against the observed soil moisture data set for Smith Street (Gerbet, 2012). More details on the Smith street BFS evaluation settings are provided in Table 1. In a first step, the SHP probability density functions are defined to assess the modelled soil moisture uncertainty with respect to these parameters. Next, Latin Hypercube Sampling (LHS) (McKay et al., 1979) was used to randomly sample one hundred parameter sets from the probability distributions which are taken from Meyer et al. (1997). They also provide a parameter correlation matrix (their Table B-1) which is implemented into the random sampling procedure LHS. The statistics of the 100 random parameter sets as well as the correlation matrix are close to the ones provided in Meyer et al. (1997) although not identical since we use only one hundred random samples (not shown). The 95% confidence interval of the 100 simulations covers an average soil water content of h = 0.07 mm3 mm3 and generally follows the main dynamics of the mean measurements. At the end of the period, some rainfall event signatures are missed in the modelled soil moisture which could be due to the fact that the observed precipitation time series are not taken at the site itself but originate from the nearby BoM weather station (see Section 3). The root-mean-square-error of the one hundred runs with respect to its mean is 0.0323 mm3 mm3. The coefficient of determination is r 2 = 0.79 while the slope and gradient of the regression on which r2 is based are respectively a = 0.03 and b = 0.89 (Fig. 4). This reflects the fact that the minimum soil moisture contents are well represented while over the whole time series, low (high) soil moisture contents are generally overestimated (underestimated). Only during the heavy precipitation events in the beginning of March, CLMU overestimates peak soil moisture. Overall, this results in a weighted coefficient of determination wr2 of 0.7 and an E1 value of 0.39. Hence it can be concluded that the model is sensitive to the values of the five parameters within their uncertainty bounds but – even though it is not optimised given the observed soil moisture – it catches the desired soil water dynamics in the biofiltration system.

2012/02/9

2012/02/19

2012/03/1

2012/03/11

2012/03/21

2012/03/31

Date

Fig. 4. Observed precipitation (blue) and mean of observed (red) and modelled (black) soil moisture contents for the sandy soil fraction for the period 09-02-2012 00h until 09-04-2012 0h. The uncertainty of the one hundred model runs are expressed as the 95% confidence interval (grey shaded area). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

162

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

4.2. Phase II: Impact of the CLMU modifications on ET 4.2.1. Impact of the biofiltration system CLMU uses a tile-approach so that the contribution of each hydrological active urban surface (roof, impervious road and BFS) was weighted by their fractional surface cover. Unless indicated otherwise, the results below represent the total urban canopy ET weighted by their fraction cover and accumulated over the two month period of interest between February 9th and April 9th 2012 [in mm]. This period was characterised by a total precipitation sum of 94.4 mm of which, for the BC scenario, 11.7 mm evapotranspires and 82.7 mm is removed via runoff. Taking into account the surface fractions of the roof/impervious road, these individual surfaces contribute with 5/6.7 mm (42.7/57.3%) and 37/45.7 mm (44.7/55.3%) to ET and Q ov er respectively. These numbers already indicate two important features: (1) interception of water on impervious surfaces substantially contributes to ET, a process that should be included in urban climate models (Wouters et al., 2014) and (2) a large amount of stormwater is potentially available to service both indoor and outdoor water demands, including irrigation of urban vegetation. Fig. 5 summarises the variables that are of direct interest to illustrate the functioning of this biofiltration design: the potential gain in evapotranspiration, the amount of water available/needed from the watertank and external water sources (scenario I6 only) and the drainage flux at the bottom of the BFS (see also Tables 1 and 2). For the non-vegetated bare soil scenario (V0) no roots are present and ET is originating from water intercepted by the urban surfaces and evaporating directly after a rainfall event. A decrease in ET (1.2 mm) is found when a non-vegetated BFS is implemented (V0). Here, the runoff from the impervious road fraction quickly drains through the filter media while the Melaleuca Argentea sparse root system (V1) is not able to compensate for the loss in ET originating from the impervious fraction. The shallow but dense root system of Carex Appressa in V2 or the combination of the two types of selected vegetation species into one biofiltration system (V3) results in a small ET increase of approximately 1 mm. This limited increase in ET is due to the high hydraulic conductivity of the loamy sand filter media that allows for quick drainage throughout the soil column, indicated by a similar magnitude of drainage for all ‘V’ scenarios (Fig. 5). Replacing loamy sand with sandy loam (Tsl) or loam (Tl) increases ET by 2.2 (17%) and 4.6 (37%) mm respectively (compared to the V3 loamy sand scenario), indicating that a biofiltration system with a loamy soil filter appears to be most effective in terms of ET. The increase of the latter is related to the lower hydraulic conductivities of both the (sandy) loam filter media in experiments Tsl and Tl which also results in a decrease of drainage at the bottom (37.6 mm for Tl versus 46.4 mm for V3, see Fig. 5). As surface runoff is not occurring in these scenarios, this results in an increased soil moisture content of the filter media (not shown), supporting plant health and a higher accumulated ET of 17.2 mm for Tl compared to 12.6 mm for scenario V3. Based on these results, all fraction cover and irrigation sensitivity experiments as described in Tables 1 and 2 use the loamy soil hydraulic parameters for the filter media together with the V3 vegetation characteristics. The results of the fraction cover experiments (Wp;x ) reveal first of all that a BFS fraction cover of 1% is not sufficiently large to store the runoff water in its detention pond, resulting in 12.4 mm of water being exported from the street canyon as unfiltered drainage water (see Fig. 2). For all other fraction cover scenarios, the BFS detention pond volume is sufficiently large to store any runoff water from the roof and impervious road surface (not shown). Furthermore, the F experiments reveal that the increase in ET relates to the fraction cover according to the quartic function ET ¼ 0:037W 4p þ 0:00018W 3p  0:014W 2p þ 0:62W p þ 11:8 (Fig. 6). This results in an increase of ET to approximately 25.3 mm for a 35% BFS fraction cover, which is a 116% increase of ET compared to the BC scenario without biofiltration system. Simultaneously, the drainage term at the bottom of the BFS shows a similar decrease from 43.4 mm for Wp;1 to 26.3 mm for Wp;35 (Fig. 5). Further increasing the BFS fraction cover reveals that ET further increases up to 29 mm when Wp ¼ 60% and levels off afterwards as shown by the full black line in Fig. 6. For these large fraction covers, not enough water is available from precipitation and impervious road runoff to maintain an increase in ET with increasing BFS cover fraction. Subsequently, soil moisture levels can drop to wilting point during dry episodes (Fig. 7).

163

25

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

I6

I7

I5

I4

I2

I3

I1

ID

Wp,35

Wp,5

Wp,20

Wp,1

Tsl

Tl Wp,10

V3

V2

V0

V1

BC

0

5

10

Water [mm]

15

20

ET BFS Drainage x 0.1 RWT water External water

Scenarios

Fig. 5. Accumulated ET (red) and drainage at the bottom of the BFS (orange) for the base-case (BC), (non-)vegetation (V), soil texture (T), fraction cover (Wp;x ) and irrigation (I) scenarios [in mm] for the full two-month period (characterised by a total of 94 mm of precipitation, see Section 3.1). The numbers in the F scenario notation refer to the % of the road fraction covered by BFS. For clarity, only the results of the scenarios Wp;1 to Wp;35 are shown. The scenarios with larger BFS fraction covers are shown in Fig. 6. Accumulated drainage should be multiplied by 10 to get the actual values. As the Tl scenario also uses 10% BFS surface cover, scenarios Tl and F10 are identical. The dotted lined refers to the ET level for scenario Tl=Wp;10 without irrigation applied. The blue (light green) bars represent the total amount of RWT (external) water [in mm] applied during the irrigation scenarios. More details on the abbreviations and scenarios are provided in Tables 1 and 2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4.2.2. Impact of urban irrigation The default irrigation scenario ID is characterised by a total accumulated ET of approximately 18.8 mm, an increase of 1.6 mm compared to the Tl/Wp;10 scenarios without irrigation (Fig. 5). For this increase in ET, 6.6 mm of irrigation water is sourced from the rainwater tank. Irrigation scenarios I1, I4 and I5 have a similar total accumulated evapotranspiration and water available from the RWT as the default scenario ID. Only scenario I1 has more water available due to an increase of RWT v ol to 100 mm. Scenarios I2 and I3 are characterised by larger ET values up to 21.6 mm (25% increase compared to the Tl/F10 scenario), due to 100% of roofs contributing to the RWT (against 22% in the previous scenarios) and thus an increased amount of water available for irrigation. Scenario I6 is significantly different as it applies irrigation at every model time step with an unlimited amount of available water. As such, soil moisture is at ho levels (Eq. (8)) throughout the simulation period. This results in a potential ET of 25.4 mm but also the highest total amount of irrigation water applied (6.5 and 22 mm from RWT and an external water source respectively)(Fig. 5). Further increasing the BFS fraction under these irrigation conditions reveals that maximum potential ET can

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

120

164

●

ET without irrigation ET according to I6 ET according to I3

● ●

● ●

●

80

●

60

●

●

40

Accumulated ET [mm]

100

●

●

20

I6

● ● I3

● ● ● ● Tl/Wp10 ● ● Wp5

● ●

●

●

●

●

●

●

● ●

●

Wp50

● Wp60

●

●

●

●

Wp70

Wp80

Wp90

Wp99

Wp35

Wp20

0

Wp1

0

20

40

60

80

100

Fraction cover [% of road fraction] Fig. 6. Accumulated ET [in mm] for the Wp;x scenarios without irrigation (black), according to irrigation scenarios I6 (blue) and I3 (red). The default fraction cover and irrigation scenarios as listed in Tables 1 and 2 are shown by the full circles while additional Wp;x scenarios following I3 and I6 characteristics are depicted by open circles. The shaded area refers to a potential gain in ET for various BFS fraction covers. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

be linearly approximated by ET ¼ 14:5 þ 1:1W p resulting in a maximum potential ET of 125 mm for the unrealistic scenario that the entire street canyon would be represented by a biofiltration system (Fig. 6). As scenario I6 is highly unlikely due to a limitation in water availability, the fraction cover experiments were repeated following the I3 scenario. Here, the gain in ET follows a cubic function (ET ¼ 0:00013W 3p  0:51W 2p þ 1:1W p þ 11:8) and levels off when the BFS fraction cover exceeds 60% of the street canyon surface, similar to the fraction cover scenarios without irrigation (Fig. 6). This shows that, for the two-month period considered here, the configuration of 100% of the households contributing to a 100 mm tank is not sufficient to maintain an ET increase for BFS fraction covers exceeding 60% of the street canyon surface. The RWT water available in scenario ID, based on a rainwater tank volume proposed by Burns et al. (2012) and the actual number of households having a RWT (ABS, 2011), is not sufficient to irrigate the BFS over the whole period as additional external water is needed during dry periods (red bars in Fig. 7). For scenarios ID, I4, I5 (the latter is not shown) and I1 the water in RWT (6.6 and 8.1 mm respectively) is not sufficient to follow the irrigation water demand. Scenarios I2 and I3 behave differently as they use 16.8 mm of RWT water for irrigation, providing sufficient water to cover the dry periods during this limited 2-month period (Fig. 7). In this case, rainwater tanks with a volume of 100 mm are large enough to store all water collected from all roofs (I3) while the use of RWT v ol = 25 mm results in an overflow during strong rainfall events (depicted by the light blue line in 7). Starting with a halffull RWT (I3hf) results in an increase of ET up to 22.6 mm due to an additional increase in ET at the beginning of the simulation period supported by an increased water availability from RWT (Fig. 5). Over the whole period, sufficient water is available to meet the irrigation requirements as shown in Fig. 7. The timing of the day for irrigation does not significantly alter the total amount of accumulated ET (I4 and I5).

165

ID

5 0

1

2

3

4

Overflow from RWT Days fullfulling irrigation requirements Accumulated water in RWT (x 0.1)

5 0

1

2

3

4

0.0 0.4 0.8 1.2

I1

5 0

1

2

3

4

0.0 0.4 0.8 1.2

I2

2 1

Dry Period 2

Dry Period 3

5 0

Dry Period 1

0

1

2

3

4

0.0 0.4 0.8 1.2

I4

0.0 0.4 0.8 1.2

I3hf

3

4

5 0

1

2

3

4

0.0 0.4 0.8 1.2

I3

2012/02/9

0.0 0.4 0.8 1.2

5

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

2012/02/19

2012/03/1

2012/03/11

2012/03/21

2012/03/31

Time Fig. 7. Time series of the irrigation scenarios ID to I4. Scenarios I5 and I6 are not shown since the former shows the exact same hydrological behaviour as I4 and the latter is rather unrealistic in terms of irrigation scenario. The grey shaded areas depict the three dry periods, while orange and light blue lines show the accumulated water in RWT (y-axis right-hand side) and overflow from RWT into the BFS (y-axis left-hand side, in mm) respectively. The black dotted lines refer to the maximum storage content of the rainwater tanks taking into account the modelled roof fraction, RWT v ol and RWT prop . The accumulated RWT and maximum RWT storage content values should be multiplied by 10 to obtain the actual amount in mm. The red bars refer to days full-filling the irrigation requirements without water being available from the RWT. The extent of these bars reveal the amount of external water that would be needed to bring soil moisture levels up to the target soil moisture wtarget;i (Eq. (8)).

5. Discussion and conclusion This study presents the implementation of a vegetated lined biofiltration pit combined with a rainwater tank and urban irrigation system in the single-layer urban canopy model Community Land Model-Urban. Runoff from roofs and other impervious fractions (e.g. roads) is harvested and used to support an adequate soil moisture level for vegetation in a lined biofiltration system. BFS characteristics such as vegetation type, soil texture, fraction cover and various irrigation scenarios were tested with respect to their effect on total urban canyon evapotranspiration.

166

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

The implementation of the Brooks and Corey model (Brooks and Corey, 1964) together with the sensitivity of modelled soil moisture dynamics to the soil hydraulic parameters was first evaluated against soil moisture observations sampled at biofiltration systems in Smith Street (Melbourne, Australia) (Gerbet, 2012). Our analysis shows that, although the model is sensitive to random changes of the model parameters within their uncertainty bounds, the main observed soil moisture dynamics are captured within the 95% confidence interval of modelled moisture results. This suggests that the modification of the default CLMU soil moisture scheme with the Brooks and Corey model is an appropriate choice for modelling soil moisture. This modelling framework is then applied to illustrate the effect of biofiltration systems and urban irrigation on urban canopy evapotranspiration. The main findings of the various scenario simulations for a two-month period in 2012 can be summarised as follows:  There is a non-negligible impact of water intercepted on impervious surfaces and its effect on ET, as was previously addressed in more detail by Demuzere et al. (2013). As such, future urban canopy layer model developments should take into account the water retention capacity of built surfaces and their role in the urban energy balance (see e.g. Wouters et al., 2014).  A large amount of stormwater is potentially available to service both indoor and outdoor water demands, including irrigation of urban vegetation.  The choice of vegetation (represented here by % of root fraction) is based on biofiltration guidelines that are in turn derived from a hydraulic and chemical point of view which does not necessarily reflect the type of vegetation species known for improving the urban climate (e.g. improved thermal comfort). In the case that biofiltration systems are designed in the framework of CSUD through WSUD, species with higher evapotranspiration rates are more beneficial but will probably require more water and thus higher irrigation rates and frequencies. As such, proper attention should be paid to the fraction cover and available water for each specific environment.  The argumentation above is also valid for the selected types of soil textures tested in this study. For example, biofiltration system guidelines advise on the use of a loamy sand texture characterised by a high hydraulic conductivity beneficial for rapid infiltration, stormwater management and pollutant removal. For example, Bratieres et al. (2008) suggest that, from a biochemical point of view, the optimal design of a biofilter should use a sandy loam filter media, as this texture possesses the largest capacity to remove both nutrients (up to 70% for nitrogen and 85% for phosphorus) and suspended solids (consistently over 95%). But these types of soils simultaneously reduce the amount of water available for vegetation. Our soil texture experiments show that, from a hydrological and biophysical point-of-view, loamy soils provide a better water retention capacity compared to sandy loam and loamy sand soils resulting in a higher evapotranspiration rate and more support for healthy vegetation. As such, the benefits of WSUD strategies should be integrally assessed in order to determine an optimal strategy from a hydraulic, biochemical -and physical point of view.  There is a large impact of the BFS fraction cover on ET. As an example, roof, impervious road and BFS surface fractions for the Tl=Wp;10 scenario represent respectively 44.5, 50 and 5.5% of the urban canopy, resulting in contribution to the total ET (17.2 mm) of 30, 34.6 and 35.4% respectively. If one would scale the contribution of the various urban surfaces to unity [e.g. 1 m2], the Tl=Wp;10 scenario results in 11.2, 12.1 and 111 mm of ET originating from these respective surfaces. This indicates that, although the BFS only covers 10% of the road fraction, ET is an order of magnitude larger than the contributions of the impervious surfaces when scaled to unity and as such a controlling factor in the evaporative cooling.  In the case of 22% of the households contribute to a rainwater tank of 25 mm, not enough rainfall can be harvested to support a continued optimal soil moisture for vegetation. For the limited simulation period considered in this study, an increase of RWT volume to 100 mm and an increased proportion of households contributing to the RWT up to 100% provides sufficient water for irrigation throughout the whole period when the RWT is initially half-full.  Finally, this study reveals that having a larger fraction of roof tops contributing to rainwater tanks is only relevant when the tanks are scaled accordingly. If not, water overflows into the BFS and is lost for irrigation during dry periods.

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

167

The variable nature of precipitation (Gallant et al., 2013) and the limited extent of the observation period makes our results susceptible to the specific meteorological conditions recorded during this observation campaign. As an example, the accumulated precipitation during our period of interest compares to the 70th percentile of the accumulated precipitation amount distribution for the same months of the historical period 1982–2011 sourced from the same BoM weather station (not shown). In addition, the amount of water harvested from the roof tops during our observation period originates to a large extent from the 5-day period between dry periods 1 and 2, characterised by continuous moderate precipitation events with a total 5-day running sum of 46.8 mm. This value is highly exceptional as it compares to the 97th percentile of the 5-day running sum distribution for the summer half-year (October to March) precipitation record from the same historical 30-year daily accumulated precipitation record. Nevertheless, these first results show the potential of the system to increase urban ET and reveal that, with an initially half-full RWT, enough water is available, however (large) enough RWTs are necessary to harvest water in support of keeping vegetation healthy. Moreover, the number of roof-tops contributing to the tank and the tank size itself should be balanced: a small number of rooftops are not able to fill large tanks while a large fraction of roofs contributing to small tanks will result in overflow, both situations leading to insufficient water being available to irrigate during dry periods. In addition, this study applies CLMU in an offline mode so that changes at the land-surface due to the biofiltration systems do not affect the urban canopy layer characteristics aloft and thus neglect potential feedback mechanisms in the model outcome. For example, the modelled increase in evaporation could result in a temperature drop and thus a drop in potential ET, causing the actual evaporation to be lower. Introducing an urban boundary layer model (see e.g. Bueno et al., 2013) or performing online simulations (CLMU coupled to a full-scale atmospheric model) will allow the meteorological forcing to adjust to changes at the surface and as such, stimulate future research on vegetation as a potential CSUD tool in creating thermally comfortable cities. Before scientific and professional communities can effectively address the full potential of green urban infrastructure and compare its co-benefits/trade-offs with other strategies (Demuzere et al., 2014), progress should be made towards the explicit representation of street canyon vegetation including the effects of e.g. shading and thus additional cooling, similar to the concepts as described in Lee and Park (2008), Lee (2011) and Lemonsu et al. (2012). Nevertheless, the methodology presented in this study can effectuate future research with state-of-the-art urban climate models, eventually coupled to a fully-resolved atmospheric model, to further explore the benefits of vegetated biofiltration systems as a climate and water sensitive urban design tool, combined with an urban irrigation system to maintain healthy vegetation. Acknowledgements This work is funded by the Flemish regional government through a contract as a FWO (Fund for Scientific Research) post-doctoral position. The CLMU simulations are supported by the Australian National Computing Infrastructure (NCI) National Facility at the ANU and the computational resources and services provided by the Hercules Foundation and the Flemish Government – department EWI. The authors would like to thank Jan Diels (KU Leuven) for his useful comments on the soil hydraulic properties and Keith Oleson (NCAR) for his suggestions on the implementation of the biofiltration system. We would also like to thank Yarra City Council for access to the tree-pits and acknowledge the contribution from the CRC for Water Sensitive Cities. Monash University provides research into the CRC for Water Sensitive Cities through the Monash Water for Liveability Centre. Finally we would like to thank the two anonymous reviewers for their insightful and useful comments on an earlier version of this paper. References Abernethy, B., Rutherfurd, I.D., 2001. The distribution and strength of riparian tree roots in relation to riverbank reinforcement, 79, 63–79 ABS, 2011. Household Water and Energy Use, Technical Report October 2011, Australian Bureau of Statistics, Victoria, Australia.

168

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Athuraliya, A., Roberts, P., Brown, A., 2012. Yarra Valley Future Water: Residential Water Use Study Volume 2 Summer 2012, Technical Report August. Best, M.J., Grimmond, C., 2014. How does the representation of vegetation within a urban land surface model influence the distribution of energy between the sensible and latent heat fluxes over a number of sites with varying land cover? In: the AMS 11th Symposium on the Urban Environment. URL: https://ams.confex.com/ams/94Annual/webprogram/Paper234291.html. Bowler, D.E., Buyung-Ali, L., Knight, T.M., Pullin, A.S., 2010. Urban greening to cool towns and cities: a systematic review of the empirical evidence. Landscape Urban Plan. 97, 147–155. http://dx.doi.org/10.1016/j.landurbplan.2010.05.006, URL: . Bratieres, K., Fletcher, T.D., Deletic, a., Zinger, Y., 2008. Nutrient and sediment removal by stormwater biofilters: a large-scale design optimisation study. Water Res. 42, 3930–3940. http://dx.doi.org/10.1016/j.watres.2008.06.009, URL: . Brooks, R.H., Corey, A.T., 1964. Hydraulic properties of porous media. Hydrology 3. Bueno, B., Hidalgo, J., Pigeon, G., Norford, L., Masson, V., 2013. Calculation of air temperatures above the urban canopy layer from measurements at a rural operational weather station. J. Appl. Meteorol. Climatol. 52, 472–483. http://dx.doi.org/10.1175/ JAMC-D-12-083.1, URL: . Burns, M.J., Fletcher, T.D., Walsh, C.J., Ladson, A.R., Hatt, B.E., 2012. Hydrologic shortcomings of conventional urban stormwater management and opportunities for reform. Landscape Urban Plan. 105, 230–240. http://dx.doi.org/10.1016/j.landurbplan. 2011.12.012, URL: . Campbell, G.S., 1974. A simple method for determining unsaturated conductivity from moisture retention data. Soil Sci. 117, URL: . Chen, F., Janjic´, Z., Mitchell, K., 1997. Impact of atmospheric surface-layer parameterizations in the new land-surface scheme of the NCEP mesoscale Eta model. Boundary-Layer Meteorol., 391–421, URL: . Clapp, R.B., Hornberger, G.M., 1978. Empirical equations for some soil hydraulic properties. Water Resour. Res. 14, 601–604. Coleman, R.F., Drake, J.F., McAtee, M.D., Belsma, L.O., 2010. Anthropogenic moisture effects on WRF summertime surface temperature and mixing ratio forecast skill in Southern California. Weather Forecast. 25, 1522–1535. http://dx.doi.org/ 10.1175/2010WAF2222384.1, URL: . Cosby, B.J., Hornberger, G.M., Clapp, R.B., Ginn, T.R., 1984. A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res. 20, 682. http://dx.doi.org/10.1029/WR020i006p00682, URL: . Coutts, A., Beringer, J., Tapper, N., 2010. Changing urban climate and CO2 emissions: implications for the development of policies for sustainable cities. Urban Policy Res. 28, 27–47. http://dx.doi.org/10.1080/08111140903437716, URL: . Coutts, a.M., Tapper, N.J., Beringer, J., Loughnan, M., Demuzere, M., 2013. 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. Geogr. 37, 2–28. http://dx.doi.org/10.1177/0309133312461032, URL: . Demuzere, M., Oleson, K., Coutts, A.M., Pigeon, G., van Lipzig, N.P.M., 2013. Simulating the surface energy balance over two contrasting urban environments using the Community Land Model Urban. Int. J. Climatol. 33, 3182–3205. http://dx.doi.org/ 10.1002/joc.3656, URL: . Demuzere, M., Faehnle, M., Orru, K., Olazabal, E., Geneletti, D., Heidrich, O., Orru, H., Feliu, E., Bhave, A., Mittal, N., 2014. Mitigating and adapting to climate change: multi-functional and multi-scale assessment of green urban infrastructure. J. Environ. Manage. 146, 107–115. Emmanuel, R., 2005. Thermal comfort implications of urbanization in a warm-humid city: the Colombo Metropolitan Region (CMR), Sri Lanka. Building Environ. 40, 1591–1601. http://dx.doi.org/10.1016/j.buildenv.2004.12.004, URL: http:// linkinghub.elsevier.com/retrieve/pii/S0360132304003750. FAWB, 2009. Adoption Guidelines for Stormwater Biofiltration Systems, Technical Report. Facility for Advancing Water Biofiltration, Monash University, Melbourne, Australia. Fletcher, T.D., Andrieu, H., Hamel, P., 2013. Understanding, management and modelling of urban hydrology and its consequences for receiving waters: a state of the art. Adv. Water Resour. 51, 261–279, URL: , doi: http://dx.doi.org/10.1016/j.advwatres.2012.09.001. Gallant, A.J.E., Reeder, M.J., Risbey, J.S., Hennessy, K.J., 2013. The characteristics of seasonal-scale droughts in Australia, 1911– 2009. Int. J. Climatol. 33, 1658–1672. http://dx.doi.org/10.1002/joc.3540, URL: http://doi.wiley.com/10.1002/joc.3540. Georgescu, M., Moustaoui, M., Mahalov, A., Dudhia, J., 2011. An alternative explanation of the semiarid urban area oasis effect. J. Geophys. Res.: Atmos. 116, n/a–n/a. http://dx.doi.org/10.1029/2011JD016720, URL: http://doi.wiley.com/10.1029/ 2011JD016720. Gerbet, L., 2012. Response of Trees to the Urban Environment. Master’s Thesis. Monash University (unpublished). Gober, P., Brazel, A., Quay, R., Myint, S., Grossman-Clarke, S., Miller, A., Rossi, S., 2009. Using watered landscapes to manipulate urban heat island effects: how much water will it take to cool phoenix? J. Am. Plan. Assoc. 76, 109–121. http://dx.doi.org/ 10.1080/01944360903433113, URL: http://www.tandfonline.com/doi/abs/10.1080/01944360903433113. Grimmond, C.B.S., Oke, T.R., 1991. An evaporation–interception model for urban areas. Water Resour. Res. 27, 1739–1755. Grimmond, C.S.B., Oke, T.R., 2002. Turbulent heat fluxes in urban areas: observations and a local-scale urban meteorological parameterization scheme (LUMPS). J. Appl. Meteorol. 41, 792–810. Grimmond, C.S.B., Blackett, M., Best, M.J., Barlow, J., Baik, J.J., Belcher, S.E., Bohnenstengel, S.I., Calmet, I., Chen, F., Dandou, A., Fortuniak, K., Gouvea, M.L., Hamdi, R., Hendry, M., Kawai, T., Kawamoto, Y., Kondo, H., Krayenhoff, E.S., Lee, S.H., Loridan, T., Martilli, A., Masson, V., Miao, S., Oleson, K., Pigeon, G., Porson, A., Ryu, Y.H., Salamanca, F., Shashua-Bar, L., Steeneveld, G.J., Tombrou, M., Voogt, J., Young, D., Zhang, N., 2010. The international urban energy balance models comparison project: first results from phase 1. J. Appl. Meteorol. Climatol. 49, 1268–1292. http://dx.doi.org/10.1175/2010JAMC2354.1, URL: http:// journals.ametsoc.org/doi/abs/10.1175/2010JAMC2354.1. Grimmond, C.S.B., Blackett, M., Best, M.J., Baik, J.J., Belcher, S.E., Beringer, J., Bohnenstengel, S.I., Calmet, I., Chen, F., Coutts, a., Dandou, a., Fortuniak, K., Gouvea, M.L., Hamdi, R., Hendry, M., Kanda, M., Kawai, T., Kawamoto, Y., Kondo, H., Krayenhoff,

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

169

E.S., Lee, S.H., Loridan, T., Martilli, a., Masson, V., Miao, S., Oleson, K., Ooka, R., Pigeon, G., Porson, a., Ryu, Y.H., Salamanca, F., Steeneveld, G.J., Tombrou, M., Voogt, J.a., Young, D.T., Zhang, N., 2011. Initial results from Phase 2 of the international urban energy balance model comparison. Int. J. Climatol. 31, 244–272. http://dx.doi.org/10.1002/joc.2227, URL: http:// doi.wiley.com/10.1002/joc.2227. Grossman-Clarke, S., Zehnder, J.a., Loridan, T., Grimmond, C.S.B., 2010. Contribution of land use changes to near-surface air temperatures during recent summer extreme heat events in the phoenix metropolitan area. J. Appl. Meteorol. Climatol. 49, 1649–1664. http://dx.doi.org/10.1175/2010JAMC2362.1, URL: http://journals.ametsoc.org/doi/abs/10.1175/2010JAMC2362.1. Gutmann, E.D., Small, E.E., 2005. The effect of soil hydraulic properties vs. soil texture in land surface models, 32, 2–5, doi: http://dx.doi.org/10.1029/2004GL021843. Hamdi, R., Deckmyn, a., Termonia, P., Demarée, G.R., Baguis, P., Vanhuysse, S., Wolff, E., 2009. Effects of historical urbanization in the Brussels Capital Region on surface air temperature time series: a model study. J. Appl. Meteorol. Climatol. 48, 2181– 2196. http://dx.doi.org/10.1175/2009JAMC2140.1, URL: http://journals.ametsoc.org/doi/abs/10.1175/2009JAMC2140.1. Hamel, P., Fletcher, T.D., 2014. The impact of stormwater source-control strategies on the (low) flow regime of urban catchments. Water Sci. Technol. 69, 739–745. http://dx.doi.org/10.2166/wst.2013.772, URL: http://www.ncbi.nlm.nih.gov/ pubmed/24569271. Järvi, L., Grimmond, C.S.B., Christen, a., 2011. The Surface Urban Energy and Water Balance Scheme (SUEWS): evaluation in Los Angeles and Vancouver. J. Hydrol. 411, 219–237. http://dx.doi.org/10.1016/j.jhydrol.2011.10.001, URL: . Järvi, L., Grimmond, C.S.B., Taka, M., Nordbo, A., Setälä, H., Strachan, I.B., 2014. Development of the Surface Urban Energy and Water balance Scheme (SUEWS) for cold climate cities. Geoscientific Model Dev. Discuss. 7, 1063–1114. http://dx.doi.org/ 10.5194/gmdd-7-1063-2014, URL: . Knights, D., 2009. Water Sensitive Urban Design Vegetation Selection Guide, Technical Report, Northern Territory Department of Planning and Infrastructure, Sydney, Australia. Krause, P., Boyle, D.P., 2005. Comparison of different efficiency criteria for hydrological model assessment. Adv. Geosci. 5, 89–97. Kusaka, H., Kimura, F., 2004. Coupling a single-layer urban canopy model with a simple atmospheric model: impact on urban heat island simulation for an idealized case. Society 82. Lawrence, P.J., Chase, T.N., 2010. Investigating the climate impacts of global land cover change in the community climate system model. Int. J. Climatol. 30, 2066–2087. http://dx.doi.org/10.1002/joc.2061, URL: http://doi.wiley.com/10.1002/joc.2061. Lee, S.H., 2011. Further development of the vegetated urban canopy model including a grass-covered surface parametrization and photosynthesis effects. Boundary-Layer Meteorol. 140, 315–342. http://dx.doi.org/10.1007/s10546-011-9603-7, URL: . Lee, S.H., Park, S.U., 2008. A vegetated urban canopy model for meteorological and environmental modelling. Boundary-Layer Meteorol. 126, 73–102. http://dx.doi.org/10.1007/s10546-007-9221-6, URL: . Lemonsu, a., Masson, V., Erell, E., Pearlmutter, D., Shashua-Bar, L., 2012. Inclusion of vegetation in the Town Energy Balance model for modeling urban green areas. Geoscientific Model Dev. Discuss. 5, 1295–1340. http://dx.doi.org/10.5194/gmdd-51295-2012, URL: . Loridan, T., Grimmond, C.S.B., Grossman-Clarke, S., Chen, F., Tewari, M., Manning, K., Martilli, A., Kusaka, H., Best, M., 2010. Trade-offs and responsiveness of the single-layer urban canopy parametrization in WRF: an offline evaluation using the MOSCEM optimization algorithm and field observations. Q. J. Roy. Meteorol. Soc. 136, 997–1019. http://dx.doi.org/10.1002/ qj.614, URL: http://doi.wiley.com/10.1002/qj.614. McKay, M.D., Beckman, R.J., Conover, W.J., 1979. A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21, 239–245. http://dx.doi.org/10.2307/1268522, URL: http:// dx.doi.org/10.2307/1268522. Meyer, P.D., Rockhold, M.L., Gee, G.W., 1997. Uncertainty Analyses of Infiltration and Subsurface Flow and Transport for SDMP Sites, Technical Report NUREG/CR-6565, Office of Nuclear Regulatory Research, U.S. Nuclear Regulatory Commission, Washington, DC. Mitchell, V., Mein, R., McMahon, T., 2001. Modelling the urban water cycle. Environ. Model. Softw. 16, 615–629. http:// dx.doi.org/10.1016/S1364-8152(01)00029-9, URL: . Mitchell, V.G., Cleugh, H.A., Grimmond, C.S.B., Xu, J., 2007. Linking urban water balance and energy balance models to analyse urban design options. Hydrol. Process. 22, 2891–2900. http://dx.doi.org/10.1002/hyp.6868. Mualem, Y., 1974. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resour. Res. 12, 513–522. O’Grady, a.P., Eamus, D., Cook, P.G., Lamontagne, S., 2005. Comparative water use by the riparian trees Melaleuca Argentea and Corymbia bella in the wet-dry tropics of northern Australia. Tree Physiol. 26, 219–228, URL: . Oke, T.R., 1987. Boundary Layer Climates, 2nd editio ed. Routledge. Oleson, K.W., Dai, Y., Bonan, G.B., Bosilovich, M., Dickinson, R., Dirmeyer, P., Hoffman, F., Houser, P., Levis, S., Niu, G.Y., Thornton, P.V.M., Yang, Z.L., Zeng, X., 2004. Technical Description of the Community Land Model (CLM), Technical Report NCAR/TN461+STR, National Center for Atmospheric Research, Boulder, CO. Oleson, K.W., Bonan, G.B., Feddema, J., Vertenstein, M., 2008a. An urban parameterization for a global climate model. Part II: sensitivity to input parameters and the simulated urban heat island in offline simulations. J. Appl. Meteorol. Climatol. 47, 1061– 1076. http://dx.doi.org/10.1175/2007JAMC1598.1, URL: . Oleson, K.W., Bonan, G.B., Feddema, J., Vertenstein, M., Grimmond, C.S.B., 2008b. An urban parameterization for a global climate model. Part I: formulation and evaluation for two cities. J. Appl. Meteorol. Climatol. 47, 1038–1060. http://dx.doi.org/ 10.1175/2007JAMC1597.1, URL: http://journals.ametsoc.org/doi/abs/10.1175/2007JAMC1597.1. Oleson, K.W., Bonan, G.B., Feddema, J., Vertenstein, M., Kluzek, E., 2010a. Technical Description of an Urban Parameterization for the Community Land Model (CLMU), Technical Report NCAR/TN-480+STR, National Center for Atmospheric Research, Boulder, CO.

170

M. Demuzere et al. / Urban Climate 10 (2014) 148–170

Oleson, K.W., Lawrence, D.M., Bonan, G.B., Flanner, M.G., Kluzek, E., Lawrence, P.J., Levis, S., Swenson, S.C., Thornton, P.E., Dai, A., Decker, M., Dickinson, R., Feddema, J., Heald, C.L., Hoffman, F., Lamarque, J.F., Mahowald, N., Niu, G.Y., Qian, T., Randerson, J., Running, S., Sakaguchi, K., Slater, A., Stockli, R., Wang, A., Yang, Z.L., Zeng, X., Zeng, X., 2010b. Technical Description of version 4.0 of the Community Land Model (CLM), Technical Report NCAR/TN-478+STR, National Center for Atmospheric Research, Boulder, CO. Oleson, K.W., Lawrence, D.M., Bonan, G.B., Drewniak, B., Huang, M., Koven, C.D., Levis, S., Li, F., Riley, W.J., Subin, Z.M., Swenson, S.C., Thornton, P.E., Bozbiyik, A., Fisher, R., Heald, C.L., Kluzek, E., Lamarque, J.F., Lawrence, P.J., Leung, L.R., Lipscomb, W., Muszala, S., Ricciuto, D.M., Sacks, W., Sun, Y., Tang, J., Yang, Z.L., 2013. Technical Description of version 4.5 of the Community Land Model (CLM), Technical Report NCAR/TN-503+STR, National Center for Atmospheric Research, Boulder, CO. Vahmani, P., Hogue, T.S., 2014. Incorporating an urban irrigation module into the noah land surface model coupled with an urban canopy model. J. Hydrometeorol. 15, 1440–1456. http://dx.doi.org/10.1175/JHM-D-13-0121.1, URL: http:// journals.ametsoc.org/doi/abs/10.1175/JHM-D-13-0121.1. Walsh, C.J., Fletcher, T.D., Ladson, A.R., 2005. Stream restoration in urban catchments through redesigning stormwater systems: looking to the catchment to save the stream. J. North Am. Benthol. Soc. 24, 690–705. http://dx.doi.org/10.1899/04-020.1, URL: http://dx.doi.org/10.1899/04-020.1. Wouters, H., Demuzere, M., De Ridder, K., van Lipzig, N.P.M., 2014. The Impact of Impervious Water-storage Parameterization on Urban Climate Modelling, Re-submitted to Urban Climate.