Impact of design variables on hydrologic and thermal performance of green, blue-green and blue roofs

Climate change and urbanization are worldwide challenges that result in increasing adverse impacts on ecosystems. Current studies show that climate change elevates air and surface temperatures and magnifies storm events in both magnitudes and duration (Kendon et al. 2014; Westra et al. 2014). Urbanization results in fewer green spaces, leading to increased surface runoff, greater absorption of solar radiation and lower evaporation rates (Gunawardena et al. 2017). A growing interest exists in implementing green, blue-green, and blue infrastructure on the ground and on building rooftops to mitigate climate change's and urban sprawl's adverse impacts by enhancing multiple ecosystem services (Voskamp & Van de Ven 2015). Green, blue-green and blue roofs are among these infrastructures that do not only control stormwater runoff but can also enhance microclimate conditions (Almaaitah et al. 2021). Consequently, these environmental benefits aid in developing sustainable, adaptive, resilient cities (Brears 2019).

Green roofs contain vegetation and soil medium planted over a waterproofing membrane (Shafique et al. 2016). Green roofs are typically classified as extensive when growing medium <150 mm and intensive when growing medium >150 mm. They have been broadly studied in the literature for their hydrologic performance and thermal performance. In Canada, a cross-province study of identical green roof setups found cumulative annual retention of 67% in Calgary, Alberta (semi-arid, continental), 40% in London, Ontario (humid, continental) and 34% in Halifax, Nova Scotia (humid, maritime) (Sims et al. 2016). Hill et al. (2017) assessed the relative impact of irrigation, substrate type and planting on extensive green roof modules in Toronto, Ontario. The study found that smaller volumetric runoff coefficients were associated with modules that did not receive irrigation and had biologically derived substrates with a high quantity of organic matter. In contrast, the more significant volumetric runoff coefficients were associated with modules that received daily irrigation and had mineral substrates with low organic matter. There are two schools of thought on the organic matter content in green roof media. The first view claims that although higher organic matter favours plant growth, it magnifies the risk of a decrease in substrate depth due to the decomposition of organic matter, which can compromise drainage (Buist & Friedrich 2008; Snodgrass & McIntyre 2010). The second view argues that the substrate depth decrease does not always occur and is more influenced by other factors such as wind scour. Building on this research debate, Hill et al. (2016) surveyed thirty-three extensive green roofs aged 0–33 years in southern Ontario and found no evidence supporting the hypothesis that suggests higher organic matter results in a reduction in substrate depth.

High organic matter in soil has been found to elevate evaporative cooling (Lawrence & Slater 2007; Tuffour et al. 2014). However, little work has been done to quantify its thermal effect in green roof substrates. Several studies sought to examine the thermal performance of green roofs, including the impact on energy savings and indoor environments (Yaghoobian & Srebric 2015; Cai et al. 2011; Bevilacqua 2021), but a few studies looked at the outdoors microclimate benefits. Solcerova et al. (2017) found that air temperature above well-watered extensive sedum-covered green roofs in the Netherlands was colder at night, when the UHI is most apparent, but warmer during the day. When comparing irrigated sedum with non-irrigated grasses and wildflowers, irrigated sedum-covered extensive green roofs were 2 °C lower at the substrate's surface and 1.5 °C lower above the substrate layer (MacIvor et al. 2016).

A blue-green roof is similar to a green roof in terms of layer design but has an expanded drainage layer, providing a greater chance for evapotranspiration and gradual release of collected stormwater (Shafique et al. 2016; Cirkel et al. 2018). In Seoul, South Korea, the surface temperature of a blue-green roof module was 4 °C less than a control roof (Shafique et al. 2016). Another study found reduced temperatures of a blue-green roof due to maximized evapotranspiration in Amsterdam, Netherlands (Cirkel et al. 2018). Almaaitah & Joksimovic (2022) assessed the hydrologic and outdoor microclimate benefits of a full-scale productive blue-green roof (rooftop farm) in Toronto, Canada. The study reported air-cooling of 1.4–2.5 °C, stormwater retention of 85–88% and peak attenuation of 82–85%. These recent studies suggest promising environmental benefits of blue-green roofs but did not explore the influence of design variables (e.g., substrate type and thickness and drainage layer size) on their performance.

Blue roofs are emerging technologies that are constructed as tray-based and check-dam systems. A tray-based blue roof consists of trays filled with coarse aggregates designed to temporarily detain water during rainfall events (Campisano et al. 2018). A check-dam blue roof consists of a series of dams/weirs incorporated with orifices drilled at the bottom of each dam to prevent permanent ponding and allow rainwater to slowly drain towards the roof's outlet (Philadelphia Water Department 2020). A full-scale pilot installation of modular tray-based blue roofs was implemented using trays incorporated with orifices and filled with aggregates in Catania, Italy (Campisano et al. 2020). The study found that the blue roof had runoff volume and peak flow reductions of 34 and 60%, respectively. A pilot study of a check-dam blue roof installed in a building in New York, United States, showed that these systems could provide 20–80% detention of stormwater runoff (NYC DEP 2012). To the date of writing, no peer-reviewed published study has explicitly investigated the combined hydrologic or thermal performance of check-dam blue roofs.

Another emerging approach to achieving cooling load savings is covering buildings’ rooftops with white and high solar reflective coatings, shaping the so-called ‘cool roofs.’ Many studies found that cool roofs reduce surface temperature, enhancing indoor cooling efficiency (Xue et al. 2015; Meenakshi & Selvaraj 2018; Rawat & Singh 2022). Nevertheless, previous research has been restricted to individual cool roof systems without considering combining the reflective properties of cool roofs with the water detention properties of blue roofs.

The hydrologic and thermal functionalities of green, blue-green and blue roofs are governed by their design factors, including substrate type and thickness, drainage layer thickness and outlet size. There is limited research comparing the performance of these rooftop systems, hindering practitioners’ understanding of how they perform when subjected to the same climatic zones and weather conditions. The objective of this study is to assess the impact of design variables on the potential of green, blue-green, and blue roofs to control stormwater runoff and improve the outdoor microclimate. The assessment is based on monitoring precipitation, drainage, near-surface air temperatures, in-substrate temperatures and below-membrane temperatures of differently designed, size-identical modules over the monitoring period from June to November 2021.

Six experimental modules were constructed at the green roof innovation testing laboratory (GRITLab) on the fifth-story roof of a building at the St. George Campus of the University of Toronto in Ontario, Canada. Each module has a drainage area of 2.88 m² (2.4 m × 1.2 m) and is elevated from the roof deck to a height of 0.8 m. Design characteristics of the six modules monitored and analyzed in the study are depicted in Table 1.

The two substrate types were selected to represent the popular commercially available engineering growing media. Mineral-based substrates had properties that aligned with the German Green Roof Guidelines – FLL standards (FLL 2008). Samples from both substrate types were collected before installation for soil laboratory testing. The physical properties of the growing media (Table 2) were obtained through grain size distribution, standard compaction test, falling head test, and loss on ignition according to the procedure described in the ASTM standards ASTM D422-63; D698-78; D2434-68 and Test Methods for Examination of Composting and Compost (TMECC). Unsaturated hydraulic conductivity for each growing media was measured in the field using a mini-desk infiltrometer (METER Group, Pullman, WA). The standard sedum blend contained a variety of cultivars, including Goldmoss Stonecrop (sedum acre L.), Aizoon Stonecrop (sedum aizoon L.), English Stonecrop (sedum anglicum L.) and Dragon's Blood Stonecrop (sedum spurium ‘summer glory’ M. Bieb). Irrigation of vegetated modules was performed occasionally during dry periods, depending on preceding rainfall events, and not on a fixed schedule.

Precipitation was measured using a tipping bucket rain gauge (resolution: 0.1 mm; accuracy: 1%) (TE525M, Texas Electronics, Texas). Meteorological parameters were measured using an on-site weather station consisting of a wind monitor (resolution: 0.0980 m/s; accuracy: ± 0.3 °C) (05103, RMYoung, Traverse City, Michigan), pyranometer (resolution: 7–14 W/m²; accuracy: 0.2%) (CMP 11, Kipp & Zonen, Delft, The Netherlands), and relative humidity and temperature probe (resolution: ± 0.2% for relative humidity and ± 0.1 °C for temperature; accuracy: ± 1% for relative humidity and ± 0.2–0.3 °C for temperature) (HMP45C, Campbell Scientific, Edmonton, Alberta, Canada).

The modules’ thermal performance and impact on the rooftop microclimate were assessed by measuring each sensor's mean daily air temperatures at two different elevations, 60 cm (T_A) and 15 cm (T_B) above the surface. The air cooling effect of each module was determined through the mean daily temperature difference between the air above the module and the control roof at the lower elevation (TR₁₅). Diurnal air temperature cooling was investigated above each module using the collected hourly data in the summer. In-substrate temperatures of the vegetated modules and surface temperatures of the blue roof modules were determined using the collected T_C­ data. Temperature fluctuations of the substrates were quantified through temperature amplitudes of the substrate (TAS) determined from the substrate temperature (T_C) (maximum daily substrate temperature – minimum daily substrate temperature). Below-membrane temperatures in all modules were determined using the collected T_D. The monitoring interval of the thermistors was 5 minutes and was aggregated to hourly and daily intervals for data analysis. All the mean daily thermal parameters (T_A, T_B, TR₁₅, Tc, TAS, and T_D) were averaged across each week of the summer (June, July, August) until mid-September. Spearman's rank coefficient was used to assess the correlation between thermal performance indicators and weather characteristics (ambient temperature, relative humidity and wind speed).

During the monitoring period, forty rainfall events were observed and used in the hydrologic analysis. Compared to historical averages, August was dry and had 80% less rainfall, but September and October were wet and had 50% higher rainfall depth. The total depth of all the monitored events was over 368 mm, with 125 mm in the summer months of June, July and August and 243 mm in the fall months of September, October and November. The maximum and median rainfall depths were 42 mm and 4.1 mm, respectively. Monthly rainfall characteristics are presented in Table 3.

The data suggests that the retention in some modules was influenced by season. GR-FLL and GR-organic had a similar retention performance in the summer with minimal differences between individual events (p > 0.05). However, in the fall season, the retention performance of GR-organic was reduced, and GR-FLL retained 5–7% more stormwater than GR-organic, generating a significant difference in retention performance (p < 0.01) and resulting in an overall p < 0.05. Nevertheless, both green roof modules captured storms smaller than 4 mm in both seasons. Retention performance in organic substrates is more vulnerable to reduced weather temperatures despite having higher holding water capacities because storage capacity is primarily restored through evapotranspiration which is a temperature-dependent process. Additionally, reduced solar radiation in the fall season limits the energy available for latent heat. In contrast, FLL substrates drain stormwater rapidly even for relatively smaller events, offering maximized storage for larger subsequent rainfall events and positively impacting cumulative retention.

The effect of substrate type on retention performance was more apparent in blue-green roof modules. BGR-organic retained the maximum stormwater in all of the months; its cumulative retention was slightly higher than local studies by Sims et al. (2016) and Liu & Minor (2005). BGR-organic had RET values 10–33% higher than those achieved by GR-organic, and the difference was statistically significant (p < 0.01). This observation implies that the attenuation feature and slower drainage of organic substrate complement the storage functionality of the expanded drainage layer, thereby increasing retention. BGR-FLL did not achieve significantly higher retention than GR-FLL (p > 0.05), suggesting that when FLL substrates are used, the thickness of the substrate is more influential on stormwater retention than the storage capacity in the drainage layer. Gong et al. (2019) explained that increased retention in larger substrates increases pore volumes, allowing more water to adhere to soil particles through surface tension. The present study provides evidence in favour of this hypothesis when green and blue-green roofs are compared but only on FLL substrates, not organic ones. Both BGR-organic and BGR-FLL showed seasonal variation. Median RET for BGR-organic and BGR-FLL were 100 and 95% in the summer and reduced to 83 and 78% in the fall, respectively. The seasonal variation in blue-green roofs is attributed to decreased weather temperatures, which reduce evapotranspiration and extend the time water stays in the drainage layer, thereby reducing the capacity for capturing rainfall from subsequent events.

The retention achieved in BR-2.4 was always less and statistically different from that achieved by BR-4.8 (p < 0.01). The analysis of results revealed that check-dam blue roofs equipped with smaller orifices could provide cumulative retention similar to vegetated modules but drastically varies on an event-based basis. BR-2.4 had a median RET of 87% in the summer, slightly lower than the vegetated modules. Except for BGR-organic, retention of BR-2.4 was higher than all vegetated modules in the fall season, despite reduced temperatures and potential for evaporation. Weekly site visits and visual inspections revealed that stormwater retained in the fall season behind the dams evaporated too slowly. This observation suggests that although smaller orifices with check-dam blue roofs may provide comparable retention to vegetated modules, they may create a standing water problem. The challenge of designing optimal orifice size in blue roofs that is sufficiently small to allow detention and large enough to prevent detention time from exceeding the maximum drain down time allowed by building codes was first pointed out by Campisano et al. (2018). BR-4.8 achieved the lowest retention of all modules, with a median RET of 66 and 53% in the summer and fall seasons. Visual inspections did not determine extended standing water in this module since larger orifices allowed for faster drainage even when there was no potential for evaporation.

Peak attenuation was slightly lower in FLL substrates than organic substrates in the green roof modules (p < 0.01). Since permeability is higher in mineral substrates, stormwater passes rapidly and drains faster than in organic substrates, where permeability is lower (Stovin et al. 2015). However, peak attenuation was not statistically different between FLL and organic substrates on blue-green roof modules (p > 0.05), suggesting that the expanded drainage is more influential on peak attenuation than substrate type. When FLL substrates were used, blue-green roof modules achieved slightly higher peak attenuation than green roof modules (p < 0.01). However, when organic substrates were used, there was no statistically significant difference between green and blue-green roofs (p > 0.05). Analysis of the interquartile range revealed that both pairs of vegetated modules had PA of 100% in thirty events. However, in the remaining ten events, both blue-green roof modules had a PA range of 90–95%, whereas green roof modules had a PA of 80–85%. Improved peak attenuations by the blue-green roof modules, despite being constructed with smaller substrate thicknesses, are attributed to the larger drainage layer, as found in other studies (Shafique et al. 2016; Almaaitah & Joksimovic 2022). Retained stormwater in blue-green roofs flows from the substrate into the drainage layer and only drains out of the system when storage capacity is at its maximum. When this process occurs, the discharge rate is lower than that from the green roof due to the longer path the captured water needs to travel through the drainage layer. These findings demonstrate that blue-green roofs effectively enhance peak attenuation of stormwater by overcoming the small storage capacity in the drainage layers of conventional green roofs, as explained by Martin & Kaye (2020). There were significant differences in peak attenuation performance between the blue roof modules (p < 0.01), confirming the impact of orifice size in check-dam blue roofs, similar to observations made previously on tray-based blue roofs by Campisano et al. (2018). High peak attenuation in check-dam blue roofs is attributed to flow routing through the orifices in each dam until the captured rainwater reaches the main drainage outlet. A slight drop in peak attenuation performance of vegetated modules from summer to fall was observed. Median PA for both pairs of green and blue-green roof modules dropped by 9–11% and 2–4%, respectively.

Blue roof modules had the shortest T_delay, with BR-2.4 and BR-4.8 having median values of 15 and 12 minutes, respectively. There was no statistical difference between both modules (p > 0.05). Seasonal variation in T_delay performance was modest and only observed on vegetated modules, similar to PA performance. Median T_delay dropped from summer to fall by 29% (GR-FLL), 10% (GR-organic), 11% (BGR-FLL) and 17% (BGR-organic).

Interestingly, rainfall intensity only correlated significantly with retention of GR-FLL (rs = −0.51) and to a lesser degree with GR-organic (rs = −0.41). No significant correlation was found between rainfall intensity and retention of BGR-organic (p > 0.05). These findings prove that using organic substrates and larger drainage layers in vegetated roofs decreases the impacts of intense storms on retention, creating an opportunity to enhance adaptation to climate change. All vegetated modules had negative correlations with rainfall duration (rs = −0.63 to −0.75, p < 0.01). Longer-duration storms tend to produce larger discharge, resulting in reduced retention, as Sims et al. (2019) explain. Rainfall depth moderately correlated with retention of BR-2.4 (rs = −0.55, p < 0.01) and strongly with BR-4.8 (rs = −0.81, p < 0.01). Similarly, rainfall intensity had a less influence on the retention achieved by BR-2.4 (rs = −0.45, p < 0.01) than that achieved by BR-4.8 (rs = 0.66, p < 0.01). Retention by BR-2.4 had a lower correlation with rainfall duration than BR-4.8, likely due to the water standing issue explained previously. It is crucial to note that the standing water issue may have occurred due to the unique design of the small-scale test modules in this study and may not be present on full-scale roofs.

Similar to retention performance, peak attenuation in all modules correlated with rainfall depth (rs = −0.78 to −0.65, p < 0.01) and rainfall duration (rs = −0.62 to −0.77, p < 0.01). BR-4.8 was the only module with a slight correlation between peak attenuation and rainfall intensity (rs = −0.39, p < 0.01), suggesting that larger orifices may also play a role in peak attenuation due to more significant conveyance capacity. T_delay in all vegetated modules had negative correlations with peak rainfall intensity (rs = −0.41 to −0.56, p < 0.05). Similarly, significant correlations were found between T_delay in vegetated modules, except GR-FLL (p > 0.05), and rainfall depth (rs = −0.34 to −0.56, p < 0.05). In blue roofs, T_delay did not significantly correlate with rainfall depth (p > 0.05) and only moderately with peak rainfall intensity (rs = −0.35 to −0.42, p < 0.05). No significant correlations were found between ADP and hydrologic performance indicators of all modules.

Air temperatures consistently increased from June to August during the summer and slightly decreased at the beginning of September. Mean monthly temperatures were typical for Toronto's temperate climate. Relative humidity elevated during the night and dropped during midday. The wind speed and the relative humidity had similar averages in each summer month, but the wind speed remarkably rose in September. Table 4 summarizes the results of monitoring the meteorological parameters and their statistics over the summer season until the beginning of the fall.

Organic-based modules (i.e., GR-organic and BGR-organic) achieved the highest near-surface air cooling with a mean cooling effect of 2.92 °C (±0.2 °C). This finding confirms the hypothesis that organic soil offers a greater potential for air cooling (Hill et al. 2016; MacIvor et al. 2016). The mean near-surface air temperature above GR-FLL was 1.58 °C (±0.1 °C) lower than above the control surface, comparable to temperature differences observed by MacIvor et al. (2016) on extensive FLL green roofs. Interestingly, BGR-FLL had a mean near-surface air cooling of 2.83 °C (±0.2 °C), providing a cooling effect similar to that of organic-based modules. Therefore, if FLL substrates are used in blue-green roofs, they may achieve equivalent cooling effects to organic-based substrates in green roofs. This observation aligns with prior studies, which found a higher cooling potential in blue-green roofs than in green roofs (Shafique et al. 2016; Cirkel et al. 2018). On average, both blue roof modules cooled the above near-surface air by 2.2 °C (±0.1 °C), with an alternating performance observed between mid-August to early September. This finding suggests that check-dam blue roofs coated with reflective materials may provide higher air cooling than typical FLL green roofs but significantly lower than blue-green and organic-based green roofs. Therefore, it is evident that substrate type and drainage layer size substantially influence outdoor microclimates.

Examining T_A above the modules indicates slightly higher air temperatures, with air temperatures at 60 cm above the modules being 0.25–0.35 °C higher than those at 15 cm for all modules. While this study did not measure air temperature at 60 cm above the control surface, data analyses reveal that T_A above vegetated modules was still lower than T_B above the control surface. Evapotranspiration impacts the variation of thermal responses across vertical profiles in green roofs (Kumar & Kaushik 2005; Jim 2012). Solcerova et al. (2017) attributed lower cooling at distant points of green roofs’ vertical profiles to the air layer that is well mixed and more influenced by the air advected from the surroundings than by the roof cover type.

While maximum hourly cooling was observed at noon for all vegetated modules, the trend was the opposite for blue roof modules. BR-2.4 and BR-4.8 exhibited peak cooling effects after midnight and in the evening. However, at the start of the day, blue roofs’ air cooling dramatically dropped until BR-2.4 and BR-4.8 reached their minimum cooling at noon, creating an inverse bell-shaped pattern. No past studies assessed the outdoor microclimate effects of blue roofs. Nevertheless, general studies on non-engineered blue infrastructure (i.e., lakes and ponds) proved that time of the day (i.e., daytime, nighttime) and water temperature are influential factors in the cooling effect (Žuvela-Aloise et al. 2016). For instance, a study investigated the cooling effect of a pond on adjacent lawn microclimate and found hotter mean daytime conditions despite having lower air temperatures than a concrete control roof (Fung & Jim 2020).

Similarly, stronger negative correlations were found between relative humidity and TR₁₅ of organic-based modules (rs = −0.47 to −0.48, p < 0.05) than FLL modules (r = −0.31 to −0.36, p < 0.05). Elevated air humidity compromises evapotranspiration and thus results in lower cooling. A similar pattern was found on an extensive green roof (Jim & Peng 2012). Unsurprisingly, significant correlations were found between in-substrate temperatures and ambient temperature in vegetated and non-vegetated modules (rs = 0.60 to 0.78, p < 0.01). There were no significant correlations between in-substrate temperatures in vegetated modules and relative humidity (p > 0.05). However, surface temperatures of both blue roof modules exhibited significant correlations with relative humidity (rs = −0.38 to −0.41, p < 0.05). A possible explanation for this observation is that most sudden increase in relative humidity during the day is due to precipitation. Rainfall that falls on the hard blue roof (unvegetated) surfaces reduces its temperature. The strong correlations between T_D data and ambient temperature and relative humidity of the roof suggest that these sensors were affected by the radiated heat from the ground. Therefore, it is crucial to not over-interpret the below-membrane temperature data in this study.

The maximized hydrologic and thermal benefits of blue and blue-green roofs provide an opportunity to adapt to climate change impacts. Although blue roofs are unvegetated and cannot contribute to biodiversity, they deliver hydrologic and thermal benefits while producing less loading on buildings. Air cooling of blue and blue-green roofs increases thermal comfort and alleviates high urban heat caused by climate change and UHI. The increased storage of drainage layers in blue-green roofs can be a water source for plants during droughts and a stormwater control solution during frequent and intense rainfall events.

The purpose of the current study was to determine the impact of design variables on the hydrologic and thermal performance of green, blue-green and blue roofs. The novel blue-green and check-dam blue roofs were found to be promising technologies to adapt to climate change impacts. Substrate type and thickness, drainage layer thickness, and orifice size influenced both environmental services. However, several trade-offs would emerge which need to be carefully evaluated based on the design needs. From this study, the following conclusions are drawn:

• 1.

  The blue-green roof with organic substrate had the highest stormwater retention, peak attenuation, and longer peak delay.

• 2.

  Despite improved detention performance, the blue-green roof with FLL substrate did not achieve higher retention than the green roof with FLL substrate, indicating that the substrate thickness is more important in promoting retention than the drainage layer in FLL vegetated roofs.

• 3.

  The check-dam blue roof with smaller orifices provided retention comparable to vegetated roofs but at the expense of the drain down time allowed by building codes. However, both blue roofs in this study had lower detention performance than vegetated roofs.

• 4.

  Organic substrates in green and blue-green roofs resulted in the highest air cooling. However, they had higher in-substrate temperatures and thermal fluctuations than their counterparts with FLL substrates.

• 5.

  The blue-green roof with FLL substrate had a significantly higher near-surface air cooling than green roofs with FLL substrate, revealing the influence of the drainage layer in improving outdoor microclimates.

• 6.

  Diurnal analysis showed that maximum hourly cooling occurs in the noontime for vegetated roofs and the evening for blue roofs.

• 7.

  Below-membrane temperatures were highest in blue roofs and blue-green roofs and lowest in green roofs, suggesting that increased wet conditions elevate membrane temperatures.

Therefore, this study recommends using organic growing media in blue-green roofs for maximum stormwater runoff control and outdoor microclimate improvement. However, high moisture conditions in organic blue-green roofs may not create a thriving environment for drought-tolerant plants, requiring reconsidering plant selection. If the plant selection is restricted to drought-tolerant species, the FLL growing media could be a better option. FLL substrates can still promote outdoor microclimate benefits when installed over larger drainage layers in blue-green roofs. Other practical implications to consider are the effects of these design variables on membrane temperatures and increased loads on the building. High and persistent moisture conditions in organic growing media and blue-green roofs may adversely affect the thermal insulation and result in higher membrane temperatures, negatively influencing indoor building microclimate.

Enlarging drainage layers in the blue-green roofs may add an increased load on buildings, creating a structural challenge. However, depending on substrate type and thickness, the stored water in the drainage layer may be lighter than some types of growing media. Additionally, thermal fluctuations within substrates resulting from wet conditions may decrease the longevity of roof membranes. Future research will have to investigate further the performance of blue and blue-green roofs over a long-term period (i.e., > one year) and explore possible operational limitations during the Winter season (e.g., orifice blockage and damage to drainage layers). Altogether, the optimal selection of green, blue-green, and blue roofs would require a holistic approach considering environmental, social and economic aspects.

The authors would like to acknowledge the following industrial partners who provided in-kind materials: ABT, Inc. (Permavoid materials) and Bioroof Systems Inc. (green roof materials). Technician support was provided by the Daniels Faculty of Architecture, Landscape and Design. The authors would also like to thank Tony Ung for making helpful arrangements and Sahildeep Panesar for assisting in field and lab work.

This research was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) Collaborative Research and Training Experience Program (CREATE).

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

The authors declare there is no conflict.