Into the air: a freestanding vertical greenery system (VGS) for evapotranspiration (ET) of roof runoff

Nature-based solutions (NBS) for stormwater management such as rain gardens, swales, and permeable pavements are increasingly used as a supplement to conventional pipe-based urban drainage infrastructure which is under pressure from climate change, urban densification, and urban expansion (Fletcher et al. 2015; Leng et al. 2020; Oral et al. 2020). Stormwater-NBS designs have evolved during the past decades, since at least the 1970s, when where German frontrunners developed the modern green roof and swale-trench infiltration systems (Grotehusmann et al. 1994) to enhance evapotranspiration (ET) and soil infiltration thereby reducing the direct runoff, overall aiming to re-establish the pre-urban water balance (Henrichs et al. 2016; Uhl et al. 2016; Berland et al. 2017; Zölch et al. 2017). Although green roofs play a positive role in urban stormwater management (Francis & Jensen 2017), they rarely provide a standalone solution due to limited capacity of their most popular form – the extensive green roof, e.g. in the case of Denmark a volume reduction of only 18–28% for 0.1–1 y return periods has been found (Locatelli et al. 2014). Accordingly, the majority of stormwater-NBS are based on terrain depressions and cavities, since infiltration seems to be the only mechanism that can match the discharge capacity of the conventional grey solutions, typically targeting 2–10 y events, in this way reducing ET to a curiosity in stormwater management. Acknowledging the original ambition of improving the ET-component of the urban water balance and realizing that infiltration-based stormwater-NBS have downsides in terms of land area requirement and soil excavation works and further may be obstructed by low hydraulic conductivity, high-standing groundwater, or soil contamination, our goal was to develop an ET-based stormwater-NBS with a small environmental footprint, capable of managing at least the 5 y events. Our idea was to modify the concept of vertical greenery systems (VGS) to provide sufficient storage and ET capacity to qualify as a standalone stormwater-NBS, referred to as a stormwater-VGS.

In recent decades, numerous examples of VGS have emerged (Manso & Castro-Gomes 2015; Bustami et al. 2018; Brković Dodig et al. 2019; Prodanovic et al. 2019b; Ascione et al. 2020; Manso et al. 2021), however, none with the primary goal of managing stormwater, but rather with a range of goals including urban cooling, greywater treatment, insulation of buildings, biodiversity support, and beautification through greenery, where contribution to stormwater management may be a part (Bustami et al. 2018; Perini & Roccotiello 2018; Eriksen 2019; Prodanovic et al. 2019a; Ascione et al. 2020). The problem with using VGS as a stormwater management element is the same as that of extensive green roofs mentioned above: the ET rate will in many regions be too low to ensure that storage capacity is regained before the next significant storm event may occur (Locatelli et al. 2014; Stovin Dunnett & Yuan 2017; Pearlmutter et al. 2021). According to the Food and Agriculture Organization of the United Nations (FAO), the maximum daily ET-rates from a lush short-cut well-watered grass field, referred to as a reference crop against which other crops may be compared, range from 1 mm in cold humid climates to 2–4 mm in temperate humid climates and 9 mm in warm arid climates (Allen et al. 1998).

In a temperate humid climate such as that of Northern Europe, the depths of critical rainstorms range from 40–70 mm for the 5–10 y events (Madsen et al. 2009b; Sørup et al. 2016) illustrating that many square meters of ET-surfaces are needed if stormwater runoff is to be removed in a reasonable time (1–3 days) (Jensen et al. 2020). Thus to maximize ET in VGS, a more promising approach may be to go from one-sided systems adjacent to buildings to two-sided freestanding systems, allowing for ET to take place unhindered from both sides, thereby doubling the ET surface area, and at the same time activating the ‘clothesline’ effect, where protruding vegetation due to more turbulence and edge effects has a higher ET rate than that of a same-height closed population, e.g. for single stands of trees ET-rates 1.2 to 1.4 times higher than that of a forest stand have been observed (Allen et al. 1998). Inspiration for design of two-sided freestanding VGS can be found in vegetated noise reduction barriers, including systems made from planter boxes as well as systems with ground-rooted vegetation, the latter typically being the narrower, and thus the more space efficient (Bendtsen 2010).

Another challenge in converting VGS into freestanding stormwater-VGS is the conveyance of the stormwater from the surface where it is generated to the freestanding VGS. Since pumping of stormwater should generally be avoided due to the increase in CO₂ footprint and system complexity, the more obvious approach to conveyance is to use the potential energy of stormwater originating from roof tops. This points to a preferential link between ET-based stormwater-NBS and runoff from buildings, while ET of road runoff is more challenging if pumps are not to be used. Gravity-driven conveyance is in principle simple, but in a climate with winter temperatures below zero the risk of freezing water in the conveyance tubes must be addressed.

Apart from stormwater-related factors, a number of additional aspects need consideration for a freestanding stormwater-VGS to be integrated into the urban spatial context, including safety, maintenance conditions, costs, CO₂, and resource footprint, as well as aspects related to neighborhood attractiveness and local stakeholders’ perception of the new structure which inevitably will change the local landscape (Lausen et al. 2022). By aiming at providing more services to society, in the form of site-relevant environmental co-benefits, e.g. biodiversity support, health, and amenity values, the chances of providing an acceptable stormwater-VGS increases (Backhaus & Fryd 2013; Frantzeskaki 2019; Sørup et al. 2019).

The objective of this study was to document the key experiences and understandings gained through the process of developing and implementing a new form of VGS with specified stormwater performance goals, site-relevant co-benefits and considerations of safety, lifetime of operation, CO₂, and resource footprints.

With the GCS now in operation in Copenhagen, it is demonstrated that stormwater management can be based on ET from freestanding VGS without the use of pumps. To our knowledge, this is the first VGS construction designed to perform at a specified storm return period. Furthermore, by describing the implementation challenges, this study advances the knowledge on the development of well-integrated urban stormwater solutions that provide multiple benefits with focus on regaining and enhancing the ET-component of the water cycle in urban areas.

Performance goals specification and identification of possible corresponding stormwater-VGS designs followed an iterative process, involving researchers and experts from the fields of landscape architecture, hydraulic and construction engineering, as well as companies providing solutions for stormwater management, urban plantings, and noise reduction. Also, stakeholders, e.g. the Danish Road Directorate, the Municipality of Copenhagen, the water utility company of Copenhagen (HOFOR), the social housing association KAB (Københavns Almindelige Boligselskab), and its residents who took part. The design process consisted of brainstorms, workshops, literature studies, surveys, physical tests, and mock-ups. From project initiation to the construction of the Copenhagen stormwater-VGS, named the ‘Green Climate Screen’ (GCS), it took 6 y, including an almost 2 y long negotiation and design adaptation period to obtain the construction permit from Municipality of Copenhagen, and a 3-month construction period.

In the iterative design process, many ideas appeared to be dead-ends and will not be reported here, however, a summary can be found in Supplementary Material 1, Additional Material 1: Rejected Ideas and Designs (AM 1). A full project description is available at: https://www2.mst.dk/Udgiv/publikationer/2020/gentagelse.pdf (in Danish).

The technical issues and possible design solutions that were considered in the GCS design process are presented in Table 1 together with reasons for adoption or rejection, as further elaborated in Supplementary Material 1, AM 1.

The associated assessment was guided by values relating to NBS research (Raymond et al. 2017; Frantzeskaki 2019; Sørup et al. 2019; Venkataramanan et al. 2020; Pearlmutter et al. 2021; Sowińska-Świerkosz & García 2021) and included CO₂ and resource footprint (e.g. low energy use in material production and VGS operation, minimum area footprint, minimum soil excavation), economic efficiency (e.g. long lifetime of operation), multifunctionality, local laws and regulations, and aesthetics and recreation (e.g. involvement of inhabitants to create socially acceptable solution). Those values were considered in addition to the primary criteria of stormwater management to a specified return period in a safe way. The definition of co-benefits was based on the framework presented by Raymond et al. (2017) and on the classification of services that stormwater-NBS projects can provide by Sørup et al. (2019).

This paper presents the final design of a multi-service freestanding VGS for ET and infiltration of roof runoff, referred to as the GCS. An 80 m long version of this stormwater-VGS design has been in operation in Copenhagen since 2019 (Figure 1). The GCS dimensioning is elaborated in Supplementary Material 2, Additional Material 2: Stormwater Capacity (AM 2), while in Supplementary Material 3, Additional Material 3: Co-benefits (AM 3), the possible co-benefits that were considered in the GCS design process are described.

Relying on the principle of communicating vessels, the roof runoff is conveyed to the GCS through pressure tubings (ID 50 mm) mounted 4 m above ground level inside of the four existing downpipes; the tubings run below ground and up to the top of the screen (3 m above ground level), where the water is released into a perforated distribution gutter mounted above the screen (Figure 2(b)). The water percolates through the perforations and is absorbed into the vertical mineral wool wall beneath.

The GCS manages the stormwater in a mineral wool wall allowing for storage and evaporation, supported by a planter box for further storage, infiltration, and ET.

The initially targeted stormwater management return period, T, for the GCS was T = 5 y (Table 1), however, the final design is likely to exceed this service level significantly. Based on Danish rainfall statistics (Madsen et al. 2009a) with critical storms mainly occurring in the summer period, a maximum emptying time of 72 h, and a number of assumptions as explained below, it is expected that the GCS in Folehaven can manage all runoff from the 240 m² large roof area it is serving up to the T = 15 y 24 h event (corresponding to 62.4 mm), with the mineral wool wall storing and evaporating the first 12.5 mm rain depth (T = 0.1 y), and the planter box to additionally evapotranspire 1.4 mm and infiltrate 48.5 mm (details in Supplementary Material 2, AM 2). In this calculation, the following is assumed (as further explained in Supplementary Material 2, AM 2): a water storage capacity in the mineral wool of 0.13 (m³/m³), a summer evaporation rate from the mineral wool wall of 10 L/m² per day (5 L/m² per day from each side), a water storage capacity of 0.40 (m³/m³) in the planter box, an ET rate from the planter box of 3 L/m² per day, and an infiltration rate into the soil beneath the planter box of 10⁻⁶ m/s.

Although the constructed GCS meets the water storage requirement, the capacity of the mineral wool part in stormwater-VGS may be of concern, especially when stacked. In case of the GCS in Copenhagen, the main storage capacity is provided by the planter box (11.6 m³) with only 3 m³ provided by the 2 m high mineral wool wall. As explained in Supplementary Material 2, AM 2, the mineral wool seems to be susceptible to what may be referred to as a ‘stacking effect’, causing the storage capacity of mineral wool slabs to decrease with stack height. According to laboratory tests, this effect is responsible for the water holding capacity to drop from 6.4 g of water per g of dry wool in non-stacked wool to 2.9 g of water per g of dry wool when stacked to a height of 2 m, in both cases with high-density wool facing downwards and after 24 h of drainage (Supplementary Material 2, AM 2). Materials that can outcompete mineral wool in terms of water holding capacity, structural features, and durability, especially resistance to microbial mineralization in the moist aerobe environment of a stormwater-VGS, are not easy to identify. Côrtes et al. (2019) evaluated shredded cork insulation boards for use in VGS systems but found a water holding capacity of 20 kg/m³ (Côrtes et al. 2019, 2021), which is much less than the 130 kg/m³ observed for the mineral wool used in the GCS, and gave no estimates of lifetime.

Granting that most of the water holding capacity of the GCS is provided by the planter box, the impact of the ET mineral wool wall is still significant, both in terms of capacity (12.5 mm), and in terms of improvement of water balance since 24 h rainfall less than this depth comprises 90% of the annual precipitation, based on recorded daily rainfall from 1983 to 2012 (NOAA 2022). Thus, most of the roof runoff will evaporate, and infiltration beneath the planter box will only occur a few times per year. Furthermore, compared with the extensive green roof stormwater storage mentioned in the Introduction (max. 28% of 24 h duration event with 1 y return period, corresponding to 9 mm), the GCS can store six times more.

The capacity of the conveyance system was verified in a full mock-up prior to construction, mimicking the real heights and distances, and using a flow rate of 3.5 L/s, which exceeds the highest 10-min duration rain intensities in Denmark of 1.92–3.12 mm/min (corresponding to flow rates from 60 m² of roof of 1.92–3.12 L/s) observed in the period 1999–2019 (Cappelen 2020). To avoid frost damage of the conveyance system, each pressure tube is equipped with a valve (not shown in Figure 2) that allows for emptying of the tube at a slow rate (approximately 1/10 of the flow rate in the tube) into the gravel-filled infiltration trench (Figure 4).

The removal of stormwater in the presented stormwater-VGS system is based partly on ET, which is a hydraulic mechanism not previously relied upon in stormwater management (Prenner et al. 2021). ET capacity is correlated with the size of the evapotranspiring surfaces. Compared with green roofs and VGS systems on building facades, a freestanding VGS allows for ET from both sides and thus to allow for clothesline effect to become active, as explained below. Surfaces for ET were incorporated in the GCS as the double-sided mineral wool wall and the surface of the planter box. Compared with the roof area that the GCS is serving (240 m²), the ET surface area of the GCS is estimated to be 362 m² (the two sides of the mineral wool wall and the vegetated surface area of the planter box), while the GCS ground area footprint is 73.6 m². Thus, the ET surface area has been increased 4.9 times via use of the vertical dimension. In general, this surface-area-to-ground-area ratio of almost five in the present case may represent a useful indicator of ET efficiency in stormwater-VGS – the larger this ratio the better.

The mineral wool modules (Figure 3(b) and 3(c)) allowed not only for profiting of the vertical dimension but also double-sided ET from the freestanding element, which accommodates the ‘clothesline’ effect. An indication of the potential for the clothesline effect in case of stormwater management has earlier been mentioned by Brix & Arias (2011) in their study on wetlands, and by Stovin et al. (2017) in case of rain gardens, yet none of the studies provides observed values for the ‘clothesline’ effect. The estimations of ET potential of GCS in the current study used an ET rate of 5 mm/d (Supplementary Material 2, AM 2). This rate might seem high compared with the Danish yearly average of actual ET of 582 mm, or 1.6 mm/d, recorded during the years 1990–2000 (DMI). However, the highest ET-rates reported for Danish vegetation are 1,200–1,500 mm per year as documented for wind-exposed willow stands used for treatment of domestic wastewater, with maximum rates between August and October. Excluding the winter period, these values correspond to daily ET-rates of 5.6–7 mm, or 2–3 times as high as the FAO turf grass reference (Allen et al. 1998). Therefore, the ET rate used in calculations appears valid when acknowledging the clothesline effect. Furthermore, in case of stormwater-VGS, also the ‘oasis effect’ (Allen et al. 1998) may enhance ET compared with ET of neighboring surfaces. This effect is observed where vegetation has higher substrate moisture than surrounding vegetation, reflecting a better water supply, which will in general be the case of stormwater-VGS. The ‘oasis effect’ also can contribute to the larger consideration on the microclimatic changes that the vertical element can induce in the local environment including a cooling effect provided by ET and shading (Zölch et al. 2016).

According to the estimates presented in Supplementary Material 2, AM 2, the 72 h evaporation volume of the mineral wool wall exceeds its storage capacity, i.e. more could be evaporated if the storage was increased. For this reason, the process of infiltration into groundwaters remains the main hydraulic mechanism for higher return periods. The planter box contributes little to the ET but has a high storage and infiltration capacity as its calculated 72 h infiltration capacity exceeds its storage capacity (Supplementary Material 2, AM 2 paragraph 1.4). Also, there is uncertainty regarding this evaporation volume estimate since the ET rate depends on the amount of available water which will gradually be depleted as the wool dries up (Bougoul et al. 2005; van de Wouw et al. 2017; Pérez-Urrestarazu et al. 2019).

A freestanding stormwater-VGS placed in a streetscape must meet a high standard of safety to avoid accidents if collapsing. Multiphysics analysis using the Comsol Multiphysics^® version 5.6 from Comsol AB simulation program was performed to ensure appropriate statics of the GCS assuming fully water saturated conditions and a maximum wind speed of 23 m/s.

To ensure a long operating lifetime, mineral wool was chosen as a structural storage material. It is mainly made from volcanic rock and practically non-degradable. The steel footings used to mount the wooden beams on concrete pillars (Figure 3) and other metal parts were galvanized (hot-dip zinc 60 my-m) preventing corrosion for at least 10 y. Wood decay was controlled by using hardwood Robinia pseudoacacia for beams and a willow (Salix viminalis × Salix schwerinii) panel construction method allowing the panel material made from 5 to 7 cm diameter poles to dry up after rain (no substrate contact) thereby preventing fungi attack, resulting in a lifetime of at least 37 y according to the provider.

As for many of the existing VGS (Manso et al. 2021), there are still considerations regarding cost of implementation and lifetime of operation, especially considering the use of heavy CO₂ footprint materials. The life cycle analysis (LCA) for VGS systems that were studied so far indicate that stainless steel used in support systems of VGS constructions have 10 times higher environmental burden than supporting structures constructed of recycled plastic or hardwood (Pearlmutter et al. 2021). Considering this, the GCS used steel elements only when necessary for safety reasons to provide enough static strength. Since the freestanding VGS element, especially when located next to the road, needs more support than building-attached VGS, this was considered as a compromise for providing more ET potential. The use of mineral wool as a storage component can be controversial from the sustainability standpoint as this material has been produced with high energy consumption (Beccali et al. 2003), hence having a high CO₂ footprint in the production process. Although organic materials like coconut fibers, seaweed, wooden chips, etc., would have the advantage of being renewable, their gradual degradation, and subsequent compaction would require regular refilling, making the maintenance challenging and costly (Pérez-Urrestarazu et al. 2019). A green-roof type substrate consisting of crushed tiles mixed with compost was tested in two types of living wall-VGS, made from either gabions or an A-shaped wooden structure, lined with geotextile for holding the substrate in place, and cut open to allow for greenery in the form of plug perennials (Supplementary Material 3, AM 3; Figure 2). In both configurations, significant compaction of substrate was observed. Compared with organic materials, mineral wool has a longer lifetime and a higher porosity, minimizing the required size of the storage compartment.

The GCS now in operation in Copenhagen is one of a number of possible outcomes of the listed project criteria, local characteristics, and local requirements (Supplementary Material 3, AM 3), therefore with other pre-conditions the GCS could have reached another form. As such, some of the design choices that were made are not general but specific to local conditions, e.g. the 3 m height, or the length and shape of the screen, as well as the sizing of stormwater volume based on the disconnected rooftop size (Table 1). However, a strength of the presented solution lies precisely in its adaptability to the local conditions. The final construction is based on the modular technique of fixed-in-place panels that are mounted in the ground with minimal attachment points, resulting in minimal earth work. This allows for variability both in the vertical surface or lengthwise, e.g. the height of the panels can be lowered or heightened to required parameters, and modules can be easily reproduced and modified according to changing uses and functions. A modular approach is also an often-chosen solution within the research on innovations in living wall systems (Serra et al. 2017; Castellar et al. 2018). The adaptability of the design indicates a potential for stormwater-VGS based on ET to be placed not only in streetscapes but as stormwater elements next to buildings, in suburban gardens or in industrial sites.

To compensate for the loss of existing benefits such as degradation of habitats and spaces for recreation and to provide benefits that may easily be obtained in conjunction with the new structure, the GCS design targeted co-benefits related to socio-cultural values, enhanced biodiversity, and improved health through noise reduction, local cooling, and air quality improvement (Supplementary Material 3, AM 3). As detailed below, some of the co-benefits influenced the stormwater management component.

With a position next to a busy road (Figures 1 and 5(c)), the design target second to stormwater management was noise abatement. Simulations of noise reduction conducted with software SoundPlan, version 6.5 by Delta (Supplementary Material 3, AM 3) following a 3 or 4 m high VGS indicated that a 3 m high screen would be sufficient if equipped with a small green roof formed as a cap facing the noise source (Figure 5(b)) and serving purposes of aesthetic too. To ensure noise reduction, the VGS should have a bulk weight of at least 15–20 kg/m² (Miljøstyrelsen 2007), be free of voids or gaps, and preferably be made from sound absorptive rather than reflective materials which were the case (Supplementary Material 3, AM 3).

Another important influence on the hydraulic design of the VGS was the addition of the window modules (Figure 3) as a part of the cultural co-benefit (Supplementary Material 3, AM 3 paragraph 1.1). This modification reduced the total water storage capacity of the screen by 1.5 m³, corresponding to 10% (Supplementary Material 2, AM 2).

By introducing the planter boxes, also as a cultural benefit increasing the aesthetic appearance of the screen, the overall stormwater storage capacity was tripled, from 3 to 14.6 m³ (Supplementary Material 2, AM 2). The planter boxes further enabled a more diverse vegetation scheme with focus on native species to support biodiversity (Supplementary Material 3, AM 3). As the vegetation cover increases, ET and subsequently local cooling is expected to improve, as is air quality by deposition of particulate matter on the leaves (Supplementary Material 3, AM 3). Vegetation will also protect the willow construction from degradation due to blocking of direct sunlight.

The importance of this study lies in the fact that it describes not only the technical components of a multifunctional stormwater-VGS but focuses also on analyzing the real-life consideration on the integration of stormwater management into VGS and considers how the co-benefits influence the technical aspects and each other (Supplementary Material 3, AM 3). This is a novel approach that is in line with the current NBS development guidelines (Commission, E. et al. 2020). Despite that, the idea of using vertical surfaces for stormwater management is not new (Ascione et al. 2020), yet according to Prenner et al. (2021) even though the introductory concept of stormwater-VGS is often mentioned in the literature, only a handful of studies presents details on their practical implementation (Bustami et al. 2018; Brković Dodig et al. 2019; Prenner et al. 2021). In contrast to those cases, our study focused on the fact that the VGS are part of the urban structure and can be developed to service multiple benefits (Lähde et al. 2019; Venkataramanan et al. 2020; Pineda-Pinto et al. 2022).

The performance goals and the design options for the freestanding stormwater-VGS (Table 1 and Supplementary Material 3, AM 3) show how influential the human-related factors were on the final design and outlook. While the freestanding GCS avoided issues with fear of building damage caused by vegetation or water often mentioned in studies on VGS (Magliocco 2018), especially influential on the final design were the concerns regarding the height of the screen and its reference to the feeling of personal safety as well as shading of the surrounding area (Supplementary Material 3, AM 3). These concerns were not earlier addressed in VGS, mainly because they are usually considered as an attachment to the building and not a separate structure.

By reversing the stormwater storage orientation from horizontal to vertical, the spatial footprint and earth work of stormwater-NBS can be minimized, and similarly to other VGS, stormwater-VGS can offer a range of additional ecosystem services. In contrast to conventional NBS for stormwater management, the GCS presents the advantage of being applicable in cases where infiltration-based stormwater-NBS is not an option, e.g. due to low hydraulic conductivity, soil contamination, or shallow groundwater, and with the gravity-driven conveyance system, it operates at zero energy cost.

For practitioners, the GCS holds the potential of being modified to existing site conditions and a variety of design expressions can be developed depending on the local conditions, which makes the design a viable alternative to conventional NBS.

Further research should include an LCA focusing on the materials used for GCS construction, as well as an investigation of more climate and resource-friendly alternatives to concrete, steel, and mineral wool. Research on minimizing the ‘stacking effect’ of the mineral wool or identification of alternative materials for water storage is needed too. Additionally, the use of vegetation to add transpiration as a mechanism on the evaporative module should be further elaborated as should studies on the public perception of freestanding stormwater-VGS structures in urban landscapes.

This research was funded by the Danish Ministry of Environment under the MUDP-program (NST-404-00177), and by the Municipality of Copenhagen. The social housing association KAB-Folehaven supported the realization of the GCS by providing land, roof runoff, engaged residents, and maintenance of the GCS. Tim Larsen Engineering, Danish Technological Institute, and Pilebyg A/S are acknowledged for their significant contributions to the development of the GCS.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.