Multilayer blue-green roofs as nature-based solutions for water and thermal insulation management

Modern society is dealing with several challenges and global changes that are reshaping the urban environment. Population is growing fast and moving from rural to urban areas: projections show that the global population will reach 9.7 billion by 2050 and two-thirds of it will be living in cities (UN 2018), and a consequent increase in impermeabilization of natural surfaces is expected. At the same time, climate changes are expected to increase the global average temperature, with longer dry periods, alternated by short and highly intense rainfall events (IPCC 2007; Masson-Delmotte et al. 2018) also in the Mediterranean region (Treppiedi et al. 2021). The interplay of climate and urbanization changes is projected to facilitate increased pluvial flood generation and intensify urban heat islands (Arnone et al. 2018). Moreover, the COVID-19 pandemic and the consequent socio-economic crisis are strongly influencing and reshaping the way of thinking and living in the urban environment, and, in particular, the need for private green spaces is emerging (Hanzl 2020; Honey-Rosés et al. 2020; Jenkins 2020; Xie et al. 2020; Pelorosso et al. 2021a).

Different solutions have been proposed to mitigate and adapt to climate changes and population densification, creating smart, sustainable, and resilient cities, aiming to achieve the Sustainable Development Goals (SDGs) proposed in the Sustainable Agenda 2030. A solution to cope with urban rainwater excess is represented by rainwater harvesting systems. Such systems have been developed through the centuries in Mediterranean areas to collect and store water during rainfall events, making it available during drought periods (Beckers et al. 2013). Besides this potential usage, rainwater harvesting systems are nowadays mainly used to mitigate extreme rainfall events (Boers & Ben-Asher 1982; Zhang & Hu 2014; Teston et al. 2018; Freni & Liuzzo 2019; Akter et al. 2020; Cristiano et al. 2021b). Thanks to their water storage capacity, rainwater harvesting tanks can reduce the pressure on the drainage systems, detaining a fraction of the rainfall during the events and releasing it in a later moment (Bocanegra-Martínez et al. 2014; Chao-Hsien et al. 2014; Sample & Liu 2014; Akter & Ahmed 2015; Campisano & Modica 2015; Campisano & Lupia 2017; Adugna et al. 2018; Cipolla et al. 2018; Abas & Mahlia 2019; Palermo et al. 2020). If properly stored and treated, this water can be reused for different domestic purposes, such as home-garden irrigation and flushing toilets, potentially reducing potable water consumption.

Besides these traditional solutions, which are often costly and not flexible, other engineering solutions, known as nature-based solutions (NBSs), have also been investigated to mitigate the impact of pluvial floods and to increase urban resilience to climate and demographic changes (Pelorosso et al. 2018; Oral et al. 2020; Recanatesi & Petroselli 2020; Calheiros & Stefanakis 2021). NBSs (e.g., green roofs (GRs), rainwater collection parks, permeable pavements, and wetlands) are solutions inspired and supported by nature that can provide multiple environmental, social, and economic benefits (i.e., ecosystem services) through locally adapted, resource-efficient, and systemic interventions (Faivre et al. 2017; Nesshöver et al. 2017; Oral et al. 2020). Among the different NBSs, GRs are one of the most important technologies to deal with the urban challenges. They have been mainly investigated for their capacity to mitigate pluvial floods, retaining a fraction of the rainfall event (Getter et al. 2007; Brandão et al. 2017; Cristiano et al. 2020, 2021a; Liu et al. 2020), and to thermal insulate the buildings, reducing the energy consumption (Lazzarin et al. 2005; Castleton et al. 2010; Coma et al. 2016). Moreover, GRs present multiple benefits: they guarantee biodiversity, help lowering the urban temperature and reducing the urban heat island (Takebayashi & Moriyama 2007; Susca et al. 2011; Santamouris 2014; Muhammad & Reeho 2017; Solcerova et al. 2017), retain contaminants in the soil layer, contribute to the CO₂ sequestration (Yang et al. 2008; Gregoire & Clausen 2011; Rowe 2011; Speak et al. 2012; Abhijith et al. 2017), and add aesthetic values to urban environments (van den Bosch & Ode Sang 2017; Kumar et al. 2019).

Multilayer blue-green roof (MGR) is an innovative technology that combines all the potential benefits of traditional GRs with the large storage capacity of rainwater harvesting systems (Shafique et al. 2016a, 2016b; Muhammad & Reeho 2017; Andenæs et al. 2018; Cristiano et al. 2021a). Compared to traditional GRs, this technology presents an additional layer below the soil layer, which can store the water leaked from the soil, thus, when the soil moisture reaches the leakage triggering point, water starts to percolate and is collected in the storage layer. The water level within the storage layer can be regulated by means of a remotely controlled valve, which can be also opened or closed manually if necessary.

As part of the Polder Roof field lab, a project supported by the European Climate-KIC programme, four MGR prototypes, called Polder Roofs and developed by the Dutch company Metropolder, have been installed in four Italian cities (Cagliari, Palermo, Perugia, and Viterbo) during Spring 2019, with the aim of exploring the potential benefits of this tool in Mediterranean climate regions. This work has the goal to present some preliminary results obtained during the first period of observation, comparing the potential of the four prototypes, which differ for their setup and installation properties (e.g., soil and vegetation type) and are influenced by different climatological conditions. Two specific benefits of the MGR installation are here investigated: the flood mitigation capacity and the thermal insulation.

The paper is structured as follows. Section 2 provides a general presentation of the MGR system proposed by the Metropolder company, with a detailed description of the four MGRs installed in Italy, focusing on the different characteristics that could influence the MGR retention capacity and temperature mitigation. Moreover, this section illustrates all the installed sensors and the observed data available for each study case. The methodology followed to assess preliminary analysis to evaluate the potential runoff mitigation capacity and thermal insulation potential of this tool are presented in Section 3, while results are illustrated in Section 4. Section 5 presents the potential benefits of MGR, discussing the role of the gate in the outflow generation and how this element can improve the urban drainage management compared to traditional GR. This section also includes an overview of the future research directions. Section 6, finally, summarizes the key points of the project and highlights the main conclusions of this preliminary analysis.

The Polder Roof is an innovative system at a Technical Readiness Level (TRL) of 7, designed by the Metropolder Company, to maximize the water retention capacity of a roof. Currently, there are more than 25 functioning Polder Roof systems in the Netherlands and many more at the engineering stage. The system buffers and holds water while functioning as a foundation of an urban surface most commonly a GR, garden, or park. Though green is the current top layer, there is sustained interest in placing systems on hardened surfaces such as parking areas, streets, and pathways.

Using Metropolder's patent pending Sm^rtMILL (Figure 1(c)), the Polder Roof functions as a dynamic water buffer. The Sm^rtMILL contains sensors which measure rainfall, temperature, water height, and wind speed. Based on the levels of each of these components, the water held on the roof can be maintained or released through a valve connected to the roof drain. The Sm^rtMILL transmits its data to a personal dashboard which displays real-time conditions and allows for remote adjustments. Remotely controlled, fine-grained management, and distribution of stored water is the Polder Roofs contribution to urban resilience. The Polder Roof fits building regulations, policy aims, and is easily applicable and well scalable. The innovations regarding urban flooding, building heating and cooling, insulation, groundwater infiltration, grey water re-use, and urban agriculture are what set the Polder Roof apart from any rooftop system seen before (van der Meulen 2019; Pelorosso et al. 2021b).

In this section, the characteristics of the prototypes located in Cagliari, Palermo, Perugia, and Viterbo are described, with a particular focus on the peculiarities that could influence the water retention capacity and the thermal insulation properties of this system, such as soil and vegetation type. In some locations, the presence of a benchmark roof (either a bare roof or a traditional GR) will allow a comparison with the prototype to highlight the benefits of this innovative system. For each test case, Table 1 resumes the main characteristics of the system, the installed sensors, and the climate of the site where it is installed (e.g., mean annual rainfall depth and temperature reference values, derived from local historical time series).

The four Italian Polder Roofs are equipped with a standard monitoring station, which provides the real-time observation of essential climatic variables, and additional local instrumentation to pursue specific objectives of each case study. In particular, the standard Polder Roof monitoring station includes a tipping-bucket rain gauge, two thermometers for air and water temperature, an anemometer, and a sensor to monitor water level in the storage layer. The Metropolder monitoring station is connected to an online platform, which enables to easily read the different measurements and to regulate the water level in the storage layer, varying the opening degree of the valve.

With the aim to better understand the differences among the case studies, Figure 2(a) shows the empirical complementary cumulative distribution function of the daily rainfall of historical time series in the four locations (circles) and the corresponding fit with a generalized Pareto distribution (thick lines), by the Multiple Threshold Method (Deidda 2010). Historical rainfall time series present different lengths for each case study: Cagliari (1947–1997), Palermo (2002–2020), Perugia (1990–2021), and Viterbo (1916–2015). The shape parameter of the generalized Pareto distribution and the annual average rainfall m are reported in the legend of Figure 2(a). Higher values of the shape parameter k highlight a higher probability of observing extreme values with respect to the local average in the considered location.

The Polder Roof located in Cagliari (Figure 2(b)) is installed 50 cm above the ground on a wooden structure within the garden of the Department of Hydraulics and Hydrology of the Civil Engineering Faculty, University of Cagliari. The MGR has a surface of 16 m² (4 m × 4 m) and it is constituted by an 8 cm storage layer and an 8 cm of soil, classified as sand. The soil is planted with common cactus plants (Cactaceae), which represent one of the local species of the vegetation. These plants are characterized by a Crassulacean Acid Metabolism (CAM), that enables them to close the stoma during the day, avoiding water dispersion. The evapotranspiration efficiency of CAM vegetation is lower with respect to other plants, but it ensures the survival of the species even during very hot and dry periods. Moreover, CAM vegetation requires very low maintenance, with consequently low costs (Cristiano et al. 2020). For these reasons, this type of vegetation is particularly suitable for regions characterized by long hot and dry summers, such as the Mediterranean areas. The MGR outflow is directed into a 350-l rain barrel equipped with a Baro-Diver^®.

In addition to the sensors included in the standard MGR monitoring station provided by Metropolder (see Sections 2.1 and 2.2), four thermometers (MX2203, produced by HOBO) have been installed to analyse the temperature in the soil, and beneath and on the wooden structure, on February 25, 2021. The four thermometers have been set to record the surface temperature with a 30-min resolution. Rainfall data have been provided by the Agency for the Environmental Protection of the Sardinia Region (Agenzia Regionale per la Protezione dell'Ambiente della Sardegna – ARPAS), which manages a dense rain gauge network in the region, with a weather station close to the prototype location. Rainfall data are available for the entire investigated period with a 1-min resolution.

The MGR installed at the University of Palermo (Figure 2(c)) covers a rectangular area of about 32.0 m² (i.e., 4.30 m × 7.45 m) and is characterized by a green upper layer composed by two sub-areas with different depths; a central rectangular area, with an extension of about 18 m², is characterized by a soil thickness of 20 cm, while the remaining border region, whose extension is 14.0 m², has a shallower soil thickness of 10 cm. All sub-areas have been filled with a fertile soil consisting of a mixture of volcanic materials (about 90% of volcanic lapillus and about 10% of pumice with a grain size of 0–10 mm). In order to facilitate vegetation growth, a combination of CAM plants and typical local species were selected. More specifically, the first area was installed with a system of shrub species of garrigue and herbaceous species, such as Rosmarinus officinalis, Teucrium fruticans, Teucrium flavum, Pallenis maritima, Helichrysum rupestre, Trifolium fragiferum, Lavandula spica, Crithmum maritimum, Micromeria graeca, Coridothymus capitatus, Dianthus ruipicola, Origanum heracleoticum, and Convolvulus cneorum. The second area, instead, was realized with a mixture of sedum species and typical Mediterranean meadow species, such as Sedum album, Sedum dasiphyllum, Sedum sediforme, Sedum anopetalum, Stipa capensis, Trifolium angustifolium, Trifolium stellatum, Aegylops geniculata, Silene coeli-rosa. An almost equal size (36.8 m²) bare roof area bordering the MGR installation is also monitored for comparison and it is referred to as ‘unaltered roof’. Analogously to the installation of Cagliari, a system of two 1,000-l rain barrels and two Baro-Diver^® pressure sensors allow to collect and compare the rainwater coming from the MGR and the unaltered roof.

Besides the observations provided by the standard MGR monitoring station (see Sections 2.1 and 2.2), additional data are collected from a weighing rain gauge (OTT Pluvio2-400), a disdrometer (OTT-Parsivel2), and a weather monitoring station equipped with a tipping-bucket rain gauge (DQA130.1 by LSI LASTEM), a thermo-hygrometer sensor (Lambrecht 8096) for the measurement of air temperature and relative humidity, an air pressure sensor (Lambrecht 8128), a global solar radiation sensor (Lambrecht 16131.3), and an ultrasonic wind speed and direction sensor (VENTUS-UBM by Lufft). All the instruments are approximately 65 m from the MGR. Finally, four thermometers (MX2203 HOBO) have been installed, with a couple of sensors placed on both the MGR and the unaltered roof surfaces, and inside the building, in correspondence of the MGR and the unaltered roof.

The MGR in Perugia (Figure 2(d)) is placed on the top of the carpentry of the University for Foreigners of Perugia (Palazzina Orvieto, Perugia). The prototype is 55 m² wide and it is constituted by a 10 cm layer of lightened soil (40% pumice) to reduce the weight on the carpentry attic. Different types of plants are included, such as Sedum, Frankincense (Plectranthus), Aloe Vera, Delospermum. These plants can resist long dry periods without the need of any water supplies for irrigation in the summer months. The roof has a pathway with white pebbles for recreational activities. The installation of a 1,000-l barrel equipped with a pressure measurement to estimate the water level is in progress.

In addition to the measurement recorded by the standard MGR monitoring station (see Sections 2.1 and 2.2), three temperature sensors (MX2203 HOBO) have been placed, two in the soil layer and one under the roof, inside the carpentry. The temporal frequency of the measurements has been set equal to 5 min.

The MGR located in Viterbo (Figure 2(e)) presents a structure similar to the prototype of Cagliari. The MGR is located in the hydrological experimental site (www.mechydrolab.org) of Tuscia University. It is placed on top of a wooden structure, 90 cm above the ground, and has a total surface of 16 m² (4 m × 4 m). It is formed by an 8 cm height storage layer and 10 cm of soil. The vegetation is characterized by perennial camephytic plants adapted to survive in arid climates (S. album and S. acre). The MGR outflow is directed into a 1,000-l rain barrel equipped with a Baro-Diver^®. A second 16 m² bare area, representing a benchmark steel roof, was installed a few meters far from the MGR and is monitored for comparison with another 1,000-l rain barrel and Baro-Diver^® sensor. Two additional 1,000-l rain barrels have been installed in November 2021 to increase the time between two emptying and to reduce the risk of overflow in case of intense precipitation.

Besides the standard MGR monitoring station (see Sections 2.1 and 2.2), the climatic variables are monitored by additional sensors in the nearby Macroraingauge experimental installation (Grimaldi et al. 2018): four additional rain gauges, one thermometer for standard air temperature (placed in a shaded area), one pyranometer, and one additional anemometer are present. Two thermometers (HOBO) measure the temperatures just above the MGR and the benchmark roof. Two additional thermometers (MX2203 HOBO) have been placed on November 9, 2020 to measure the temperature above the MGR and the benchmark roof, with a time step of 5 min.

Among the multiple benefits of the MGR installation on buildings, this study focuses on the potential effects of the MGR prototypes on the runoff generation reduction and their thermal insulation properties. With this aim, this section introduces quantitative indices to evaluate the performances of MGR in terms of water and temperature control.

Thanks to the soil and storage layers, the MGR can retain and store rainwater, thus reducing the runoff generation and consequently contributing to mitigate the pluvial flood risk. The MGR runoff reduction performances have been evaluated through two indices which have been applied at the daily scale, with the aim to evaluate the retention capacity of both the soil layer and the system as a whole.

Rainfall data are available for each location at high temporal resolution (ranging between 1 and 5 min depending on the case study) and for long time windows, covering the period December 2020–December 2021. Water level in the storage layer and outflow measurements, instead, are not available yet in some locations; for this reason, it was not possible to estimate both and for all the four installations.

With the aim to investigate the retention capacity of the MGRs under critical conditions, light rainfall events have been excluded from this study. Following Alpert et al. (2002), that investigated the increase of extreme daily rainfall events in Mediterranean regions and identified rainfall thresholds, we defined as ‘light’ an event with daily rainfall depth lower than 4 mm. Hence, only events with a daily rainfall intensity equal to or higher than 4 mm have been considered in this study to assess and ⁠. The main characteristics of the selected daily events used to estimate the retention capacity of the MGR in each case study are summarized in Table 1.

MGRs help reducing the roof surface temperature contributing to the thermal insulation of the building. To investigate this potential benefit and to quantify the temperature mitigation, four indices, proposed by Bevilacqua et al. (2017), have been here used. As for the runoff mitigation, this investigation focuses on data available during the period from December 2020 to December 2021. Temperature data are recorded with a time step of 5 min (HOBO sensors in Perugia and Viterbo), 10 min (Metropolder sensors), and 30 min (HOBO sensors in Cagliari and Palermo), while the indices are evaluated at the daily scale.

The four indices compare the temperature in a certain gauge G to the temperature in a reference sensor, which is identified as ‘reference’ to represent the unaltered conditions, without the installation of the MGR. Since each prototype presents a unique configuration of the temperature sensor placement, in some cases different sensors have been chosen as a benchmark, depending on the scope of the analyses and on the peculiarities of the study case.

The presents values higher than 1 when the selected location presents an average daily temperature higher than the benchmark and lower than 1 for the opposite situation. values close to 1 highlight that there are no significant differences between the two observations.

This last index evaluates the effects of the MGR installation on both minimum and maximum temperature, in terms of daily temperature excursion.

This section presents and discusses the results obtained from the analysis of the collected data in the four case studies for a 1-year period, from December 2020 to December 2021. The benefits of the MGRs are analysed and evaluated in terms of rainwater retention capacity and the thermal insulation properties, using the indexes introduced in Section 3.

As explained in Section 3.1, the retention capacity is evaluated through two indices, the and ⁠, only for significant daily rainfall events, with a mean intensity higher than 4 mm/day. Depending on the available measurements, for each location it was possible to estimate and in order to investigate the retention capacity of the entire MGR and the soil layer, respectively. Rainfall, outflow, and water level data were recorded at different time resolutions and then aggregated and analysed at the daily scale.

In Cagliari, the periods between June 1, 2021 and August 1, 2021, and between October 1 and 30, 2021, have been excluded for this case study, due to sensors malfunctioning. Although the Baro-Diver installed in the rain barrel in Cagliari was not recording, no outflow has been observed during the investigated period. In Perugia, the rain barrel with the Baro-Diver is not recording data yet, and for this reason, also in this case, only the has been estimated and evaluated for events that did not generate runoff. Finally, for the MGRs located in Palermo and Viterbo it was possible to estimate the for the entire year, while the has been evaluated until September 2021 for the case of Palermo and not analysed for the case of Viterbo, since the sensors that measure the water lever in the storage layer did not work properly over the entire year. The gate has been always kept closed during the investigated period, except for Cagliari, where it has been periodically open for maintenance during the dry periods.

As illustrated in Figure 3(a), Cagliari and Palermo present similar ⁠, with an average retention capacity of the soil layer of 0.46 and 0.42, respectively, suggesting that most of the investigated rainfall events are only partially retained by the soil layer, and hence would have been only partially retained by a traditional GR, whose retention function is exclusively due to the soil layer retention capacity. The MGR of Perugia shows a higher performance in retaining the rainwater, probably due to the fact that, differently from the other two cases, it was not possible to measure the MGR outflow and, consequently, the events generating runoff, which usually are the most intense, has been excluded from the analysis.

The total retention capacity ⁠, ensured by the MGR prototypes, can be observed in Figure 3(b) for the case study of Cagliari, Palermo, and Viterbo. Thanks to the contribution of the additional storage layer, the performance in this case is higher than the corresponding performance of a traditional GR, with an average equal to 1 for Cagliari, to 0.65 for Palermo and to 0.75 for Viterbo. The plot shows that a significant number of rainfall events are completely retained by the systems, especially in Cagliari, where all events were fully retained, and in Viterbo, more than 50% of the events presented a equal to 1.

In Palermo, the MGR retention capacity presents a high variability, with mostly varying between 0.35 and 1. The high variability is determined by the different initial conditions that characterize the rainfall events. The antecedent conditions of soil water content and water level in the storage layer, in fact, strongly influence the retention capacity of a MGR.

The differences between the two indices, and ⁠, and consequently the additional retention capacity provided by the storage layer in a MGR compared to a traditional GR, are well evident in the case of Palermo and Cagliari. In Palermo, when only the soil layer capacity is evaluated, the average is equal to 0.42, while, when the system as a whole is considered, the average index of retention rises up to 0.65 and some of the events that were only partially retained by the soil layer, were instead fully retained by the entire system. In Cagliari, where the average annual rainfall depth is lower than in the other case studies (Table 1 and Figure 2), the MGR enabled to completely retain all the events, highlighting the high potential performance of the MGR installation in Mediterranean cities.

Results presented in this section show the seasonal performance of the MGRs, analysing separately the retention capacity during a warm and dry season, from March 22 to September 21, 2021, and a cold and rainy season, covering the complementary months of the year. The seasonal analysis emphasizes the high impact that antecedent soil moisture in the soil layer and water level in the storage layer have on the runoff mitigation.

The Spring-Summer rainfall events have been almost fully retained by the MGR in Palermo, with an average equal to 0.87, and seven out of nine events completely retained by the structure. The two not fully retained events are consecutive and happened on March 22 and 23, 2021: in this case, the soil was already partially saturated and the storage layer partially full before the events, especially for the second day where conditions close to saturation were reached after the previous rainy day and the dropped to 0.3. In Viterbo, all the 13 Spring-Summer events have been fully retained by the MGR, confirming the importance that this tool can have on the runoff generation reduction.

The retention capacity estimated during the Autumn-Winter period is quite lower compared to the Spring-Summer seasons, due to the higher frequency and intensity of the rainfall events: the average drops to 0.59 in Palermo and to 0.65 in Viterbo. Moreover, the temperature that characterizes this season is consistently lower than in the Spring-Summer period, limiting the evapotranspiration losses from the soil layer.

In Cagliari, temperatures recorded from four gauges have been used to evaluate the potential thermal impacts of the MGR. The sensor position is schematized in Figure 4(a). Two types of sensors have been used for the analysis. The ‘Air’ sensor is provided by the Metropolder monitoring stations and refer to the air temperature above the wooden structure. Three HOBO sensor records have been investigated: one is installed few centimetres beneath the soil surface (‘Soil’), and other two sensors are situated beneath the prototype, at diagonally-opposite corners (‘Under1’ and ‘Under2’). The two sensors situated beneath the prototype are used to help understand the effects on the hypothetical underneath building. An average of the two-time series is plotted in Figure 4(e). The temperature time series for the Cagliari case study are available from February 25th to December 21, 2021 and are plotted in Figure 4(e).

Four HOBO temperature sensors have been installed in Palermo to evaluate the thermal insulation properties of the MGR, as illustrated in Figure 4(b). One sensor has been placed in the soil layer (‘Soil’), a few centimetres below the surface, one has been located on the ‘Unaltered Roof’ surface, and other two sensors have been situated inside the building: one in correspondence of the MGR sensor (‘Room1’), and the other one in correspondence of the ‘Unaltered Roof’ (‘Room2’). Moreover, a Lambrecht 8096 thermo-hygrometer sensor is set to measure the air temperature 2 m above the roof (‘Air’). In Palermo, temperature observations are available from December 2020 to December 2021 and are plotted in Figure 4(f).

In Perugia, temperature measurements from three gages are available: the ‘Air temperature’, provided by the Metropolder monitoring stations and other two HOBO sensors placed at 5 cm under the substrate, under the substrate lean on the plastic crate. In Perugia, the temperature time series are available from August 2021 and are illustrated in Figure 4(g).

Measurements from the two HOBO sensors have been evaluated in Viterbo for the entire 2021 and are presented in Figure 4(h). As schematized in Figure 4(d), one sensor has been set to measure the air temperature (‘Air’), while the other two gauges have been installed in the soil layer of the MGR (‘Soil’) and on the unaltered roof (‘Unaltered Roof’). In this way, it was possible to estimate the impact of the MGR installation on the surface temperature.

Although a direct comparison among all the case studies is not possible due to differences that characterize each location, some representative analyses may illustrate the thermal behaviour of a MGR and its thermal impact.

When comparing the differences in the minimum daily temperatures through the (Figure 5(b)), the MGR temperature is overall higher than the unaltered roof performance, suggesting that the prototype can ensure a good thermal inertia, increasing minimum temperature usually occurring during the night. In Palermo, for example, the MGR temperature is on average 40% higher than the unaltered roof temperature.

Following the same approach with the ⁠, plotted in Figure 5(c), it is clear that, in most of the cases, the presence of the MGR provides a benefit for the peak, usually diurnal, daily surface temperature, since the MGR temperature is lower than the unaltered roof temperature. The effects are stronger in Palermo, where the daily peak air temperature, especially during the Summer, are usually higher than in Viterbo and, in fact, the maximum daily temperature on the MGR is on average reduced by about 40% compared to the corresponding unaltered values, while in Viterbo the average reduction is about 10%.

Finally, the variability of the TER, which describes the thermal excursion of the MGR with respect to that of the Unaltered Roof, is plotted in Figure 5(d). In both locations, the MGR shows a lower daily thermal excursion than the Unaltered Roof, with an average TER equal to 0.15 in Palermo and to 0.85 in Viterbo. Beyond the different climate conditions, the observed marked difference in performance between the two locations can be also addressed to the different materials of the Unaltered Roofs and the different vegetation coverage of the MGRs. Nevertheless, all the indices analysed here in both the case studies are concordant and highlight overall high performances of the MGR in mitigating the surface temperature of the roof, suggesting an increased thermal inertia of the external surfaces, whose temperature results less influenced by the air temperature values and their daily excursions.

The analysis shows that the average MGR daily temperature is always lower than the air temperature, with an average reduction of about 15–20% (Figure 6(a)). The minimum and maximum daily temperatures for the MGR also presents overall values lower than the corresponding air temperature; more specifically, the differences between the minimum daily temperatures are rather low, with an average close to 1 in the Cagliari case study (Figure 6(b)), while a more significant lowering (on average reduction by the 20%) of the MGR temperatures compared to the air temperatures can be noticed for the maximum values (Figure 6(c)). More relevant differences can be observed between the daily thermal excursion of the MGR and the air (Figure 6(d)): the average TER is equal to 0.7 in Cagliari and to 0.55 in Perugia, confirming the high MGR capability to limit the daily temperature variability.

The last analysis aims to directly investigate the thermal insulation properties of a MGR, comparing the temperature underneath the prototype with the soil temperature, used as reference value for the thermal indices. As shown in Figure 4, in the Cagliari case study, two HOBO temperature sensors have been placed beneath the prototype. The average between their measurements has been here used to derive the four thermal indices. In Palermo, a HOBO sensor has been installed inside the building, in correspondence with the prototype. Although the direct comparison between the two case studies is not possible, since in Palermo the roof structure influences the temperature inside the building, it is still possible to derive some conclusions regarding the thermal insulation properties of the MGR.

Different results can be achieved when investigating the minimum and maximum daily temperatures (Figure 7(b) and 7(c), respectively): in these cases, in fact, values prevalently higher than 1 and values lower than 1 over the entire 1-year period confirm that the temperature under the MGR is highly mitigated by the presence of the prototype itself. This can be inferred also by the analysis of the TER (Figure 7(d)), which reports for both location a thermal excursion under the MGR and inside the building constantly lower than that for soil surface of the MGR.

The roof configuration in Palermo enables to investigate the influence of the MGR on the temperature inside the building, comparing the recorded measures in Room 1, in correspondence of the prototype, and in Room 2, below the unaltered roof (see Figure 4(b) for the sensor configuration in Palermo and Figure 4(f) for the average daily time series). Daily temperatures in Room 1 are on average 1.3 °C lower than in Room 2 during the investigated period, while during the dry period, the difference increases up to a maximum of 3.7 °C. Moreover, the average TER between the temperatures recorded in Room 1 and in Room 2 mostly varies between 0.38 (25%-ile) and 0.95 (75%-ile), suggesting that the MGR contributes not only in reducing the average temperature inside the building, but also in limiting the daily excursion.

This study confirms the high potential of MGRs installed in Mediterranean areas in mitigating the runoff generation and in thermal insulating the underneath building. Traditional GRs guarantee an average rainfall reduction generally higher than the 50% of medium and intense rainfall events (Palla et al. 2010; Brandão et al. 2017; Palermo et al. 2019), with a strong variability connected to the structure features (such as soil type and thickness, vegetation type) and weather characteristics. These results have been confirmed by this study, where the analysis of the retention capacity guaranteed by the soil layer, represented by the ⁠, reported an average rainfall reduction between 42 and 70%, for the investigated case studies. Thanks to the additional layer, the MGRs ensure a higher volume to store rainwater, with an average retention index that reaches 0.65 in Palermo, 0.75 in Viterbo, and 1 in Cagliari.

These results support the findings presented by Busker et al. (2022), who investigated the hydrological performance of a MGR in the Netherlands. The different climatological environment leads to a different retention capacity, slightly lower in The Netherlands than in Mediterranean areas. As shown in Section 4, in fact, the retention and storage capacity of the MGRs installed in the Mediterranean areas enable to retain all the rainfall events in Spring-Summer. The high retention capacity in this case is facilitated by the low frequency of the events and by the high evapotranspiration rates, which ensures a higher probability that a rainfall event occurs when the soil is completely dry, and the storage layer is empty.

The retention capacity of the system, especially during the Autumn-Winter season, when rainfall events present higher intensity and frequency in Mediterranean areas, could be increased using a thicker storage layer. In many cases, this solution, however, might not be feasible, especially for already existing building, where the roof might not be able to sustain the weight of a large additional storage layer. An easier solution is given by a functional management of the gate that can optimize the water level in the storage layer, increasing the volume available to store rainwater.

Besides the high retention capacity, the preliminary analysis presented in this work, also emphasizes the additional benefits of a MGR, especially regarding thermal insulation. Compared to a traditional roof, the MGRs have shown lower maximum daily surface temperatures and higher minimum ones, and a lower thermal excursion during the day. These preliminary results suggest a high thermal insulation capacity for the underneath building, which can be translated in a potential save of energy for heating and cooling (Castleton et al. 2010; Coma et al. 2016).

In this preliminary analysis, where the main goal was to present the potential benefits of the MGR prototypes under Mediterranean climate, the influence of the gate management has not been evaluated, since the gate was always closed. However, it cannot be ignored that the retention capacity and the thermal insulation properties of a MGR are strongly affected by the water level in the storage layer.

The retention capacity of a MGR can be largely increased through a wise management of the gate that regulates the water storage layer. As shown in Figure 1, the gate allows to regulate the water level in the blue layer, with the aim to accumulate rainfall and release it as runoff when it is more convenient, depending on the user's needs. If the main goal is to mitigate runoff, for example, the gate should be closed when it is raining, and the harvested water should be released when the rainfall event is over, and the urban drainage system is not stressed anymore. On the other hand, if there is the need to re-use the harvested water for irrigation or other domestic purposes, the gate should be kept closed, to accumulate a larger volume of water. It is therefore clear that, depending on the goals that need to be achieved with the MGR installation, a different management of the gate should be evaluated to optimize the benefits of the system. Busker et al. (2022) recently investigated the potential improvement of the retention capacity of a MGR very similar to those tested in this work, when integrating the gate management with the weather forecasts. This choice enables to empty the storage layer before intense rainfall events and could largely reduce the peak runoff and increases the peak delay. The study calculates that the MGR in the Netherlands can reach a capture rate of 90% for total precipitation when using weather forecast to manage the gate opening. In arid and semi-arid climates, an optimized gate management can allow not only to mitigate the runoff generation, but also ensures water storage, that can be used, especially in Mediterranean cities, to limit additional irrigations and to reduce the pressure on the water supply system. Future analyses will be, hence, dedicated to understanding how the system may alter water quality to evaluate the potential reuses of water collected by the MGR for different purposes.

Recently, Busker et al. (2022) assessed that converting all the potentially suitable roofs in Amsterdam to the MGR would determine 11% of extreme precipitation (>20 mm/hour) to be captured when using proper gate management with the weather forecasts. However, the different climate and typology of Mediterranean cities call for specific and local evaluations. Although the preliminary analysis presented in this work already highlights the potential benefits of MGRs as a sustainable solution for the development of the urban environment, further investigations need to be carried out to fully understand the behaviour of this innovative system in Mediterranean climate and to emphasize the potential impacts and benefits for society. From this point of view, an evaluation of the potential impacts of the MGR at the urban scale would be desirable, i.e., assuming the installation of this system over a great number of suitable roofs in a large neighbourhood or at city scale.

Future applied research will aim at implementing and testing modelling tools to identify best solution and localization of MGR for the mitigation of hydraulic risk and runoff reduction in performance-based planning and design contexts (La Rosa & Pappalardo 2020; Pelorosso 2020). An evaluation at the urban catchment scale will then allow flood mitigation capacity of NBS-based scenarios to be tested considering different costs and levels of interventions. Further analyses will be then dedicated to the definition of the criteria for the identification of potential buildings for the MGR retrofit and the creation of smart and resilient Mediterranean cities. Following this approach, it will also be possible to assess other ecosystem services, as the heat island attenuation or the pollution retention, for a more holistic evaluation of the MGR technology.

This work describes the installation of MGRs prototypes in four Italian cities (Cagliari, Palermo, Perugia, and Viterbo) as part of the Polder Roof field lab project. Compared to traditional GRs, MGRs present an additional water storage layer that increases the rainwater retention capacity of the system. The four installed prototypes are equipped with a standard monitoring station, provided by Metropolder, and differ for their setting, geometry and extension, vegetation and soil types, and some additional monitoring instruments. Moreover, although all cities are located in the Mediterranean area, each site is characterized by different climatological conditions. This work presents preliminary analyses that investigate the retention capacity and the thermal insulation properties of MGRs, comparing measurements and results from the four prototypes during the period from December 2020 to December 2021.

To estimate the rainwater retention capacity of this system, two indices have been proposed, focusing on the performance of the entire structure or only of the soil layer. Results show how the MGR structure ensures a retention capacity higher than a traditional GR and enables to retain most of the high-intensity rainfall events, especially in the warm season. This suggests the need for a weather-based optimization of the MGR gate to control water storage in the wet season for Mediterranean climate.

Four indices have been introduced to evaluate the potential implications of the MGR installation on the temperature of the surrounding environment and in the buildings underneath the multilayer roof. Since each site presents a peculiar configuration of the MGR and different locations of the temperature sensors, it was not possible to make a direct comparison among all study cases. However, in all locations it was possible to observe a positive impact of the MGR, which helps to mitigate the daily thermal excursion in all sites. The high potential of the MGR is clearly visible in the Palermo case study, where the daily temperature excursion of the room directly beneath the MGR is lower than that measured at the room beneath the unaltered roof.

On top of the exploratory hydrological and energetical analysis, this work offers a panoramic view of the future research outlines and directions that needs to be followed in order to fully understand the potential benefits of MGR in Mediterranean Climate. Further analysis will focus on the following:

• the implementation of an ecohydrological model for reproducing the MGR behaviour, from the soil moisture dynamics to the evapotranspiration and outflow;

• the optimal management practices for the gate control, to facilitate the regulation of the water storage, based on the specific purposes;

• the water quality of the MGR outflow, in order to evaluate possible re-uses and plan adequate treatments; and

• the assessment of the potential impact of the MGRs at the urban scale, identifying suitable locations for the NBS installation in each of the four cities.

Funding received from the European Union Climatic KIC programme with grant number EIT SGA 2018 supporting project ‘Polder Roof Fieldlabs (180522)’. This work has also been partially funded by the Florisa Melone Award, promoted by the Italian Hydrological Society (IHS).

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

The authors declare there is no conflict.