https://orcid.org/0000-0002-5605-6500
ABSTRACT
?
?ABSTRACT
We examined the performance of ecoroofs with different substrate depths (75 mm, 125 mm) as well as a conventional roof using rainfall and runoff data collected between 2014 and 2017 in Portland, Oregon, USA. The two ecoroof plots exhibited similar effectiveness in retaining storm runoff across all storm size categories, and both ecoroofs retained the most stormwater during small storms. Retention ratio (RR) was highest during the dry season and the lowest in the early and middle of the wet season. RR was also positively associated with antecedent dry weather period (ADWP) with significantly higher retention when ADWP was longer than 3 days. Together, event storm size and ADWP explain approximately 50% of RR variation in middle and late-wet seasons. Overall, these findings indicate that environmental parameters such as storm frequency and size had a greater influence on greenroof performance than an additional 50 mm of substrate.
Introduction
Ecoroofs are vegetated roof systems that are designed to offer multiple benefits. Such benefits include the aesthetic value of the property (Berto, Stival, and Rosato Citation2018), thermal regulation (Nardini, Andri, and Crasso Citation2012), air quality improvement (Moradpour, Afshin, and Farhanieh Citation2018), and increased habitat for birds and insects (MacIvor and Ksiazek Citation2015). Ecoroofs also offer multiple hydrological benefits, including a reduction in storm runoff and increased evapotranspiration (Mentens, Raes, and Hermy Citation2006; De-Ville, Menon, and Stovin Citation2018; Shafique, Kim, and Kwon Citation2018), which lessen the stress on sewer systems (Berto, Stival, and Rosato Citation2018). As rooftops account for substantial urban impervious surface area, sometimes up to 40–50% (Mentens, Raes, and Hermy Citation2006; Zhang and Guo Citation2013), the potential for urban space to benefit from ecoroofs as an alternative is large. With anticipated climate change and increasing impervious surface areas in most urban areas, the importance of using ecoroofs to control storm runoff and ecosystem services has increased in recent years (Pradhan, Al-Ghamdi, and Mackey Citation2019).
The ability of ecoroofs to serve as a control measure for the reduction of stormwater runoff is well known. In the Pacific Northwest of the USA, ecoroofs exhibit the potential to retain anywhere from 12% to 17% (Spolek Citation2008) or 23.2% to 32.9% (Schultz, Sailor, and Starry Citation2018) of precipitation annually. Previous studies show that rainfall retention (RR) of ecoroofs may vary by season in temperate climates, sometimes much higher in the summer season than in the winter (De-Ville et al. Citation2018), with reported average seasonal values ranging from 20% to 48% in mostly German locations (Mentens, Raes, and Hermy Citation2006), and 12% to 42% in Portland, Oregon, USA (Spolek Citation2008). It is also typical for the reported runoff reduction per event to have a broad range, such as the 6.4% to 100% reported by Cipolla, Maglionico, and Stojkov (Citation2016). This variation is mostly attributed to differences in antecedent dry weather period (ADWP) before events, as well as differences in seasonal evapotranspiration and rainfall amount.
Due to the relatively new impetus toward incorporating ecoroofs in city plans, many focus on the roof design characteristics that provide the most functionality. The design determines whether the roof is intensive or extensive, a designation related to substrate depth, substrate composition, vegetation, as well as the maintenance requirements (Palla, Gnecco, and Lanza Citation2010). As extensive roofs are thinner (less than 150 mm of substrate), they tend to be a preferred choice for builders with their lighter and cheaper installation (Soulis et al. Citation2017) and lower maintenance requirements. Thus, understanding the relationship between substrate depth and RR of ecoroofs is necessary to ensure that locally adopted standards are economically efficient. Substrate depth has been shown numerous times to affect plant species composition in the short (Durhman, Rowe, and Rugh Citation2007; Thuring, Berghage, and Beattie Citation2010; Liu et al. Citation2019) and long term (Rowe, Getter, and Durhman Citation2012), but less is known about how depth and associated features can then affect stormwater management. A recent 1-year study of a commercial ecoroof site in Portland found that a 50 mm increase in depth only resulted in a 10% increase in annual retention (Schultz, Sailor, and Starry Citation2018), indicating that the additional benefits of higher depth might be marginal. Likewise, Nardini, Andri, and Crasso (Citation2012) did not see a difference in stormwater retention between 200 mm and 120 mm deep ecoroof modules, albeit planted with different species, over the course of six months.
Antecedent dry weather periods (ADWP) and storm size are also useful in predicting how depth might affect green roof performance (Fassman-Beck et al. Citation2013; Stovin, Vesuviano, and De-Ville Citation2017). As antecedent moisture content affects the inter-event recovery of storage capacity(Feitosa and Wilkinson Citation2016), it is critical to understand how RR might shift over different seasons with different storm frequency, duration, and amount. The performance of different ecoroofs with different depths under a range of rainfall events with different sizes and ADWP has not been investigated thoroughly. While the current study builds upon a previous study by Schultz, Sailor, and Starry (Citation2018), there are three major differences in our approach. First, we used three years of records, compared to one year of record in the previous study, allowing us to investigate the effectiveness of ecoroofs under a wide range of events. Second, our storm separation method and minimum storm amount are different to allow for better runoff separation. Third, we refined our seasons to better reflect the climate of the region. In other words, instead of using conventional months for each season, we broadly categorized the wet season from October to April and the remaining months for the dry season. Finally, we quantified the effects of storm amount and ADWP individually and jointly using multiple regression analysis. With this case study, we seek to fill the gap on factors explaining differences in ecoroof performance by depth in a winter rainfall climate by answering the following questions.
How do seasonal rainfall retention ratios (RR) in different ecoroof plots change over multiple seasons? We hypothesize retention ratios will be higher in the late wet and dry season compared to early and middle wet season.
How do RR in different plots relate to event precipitation amount, ADWP, and year? We hypothesize precipitation amounts will be negatively associated with RR while ADWP is positively associated with RR; the strength of the relation should vary by season; ecoroof performance changes from one year to the text.
Case study: Portland ecoroof
Portland’s ecoroof initiative is well documented in various city’s programs and plans. The 2008–2012 incentive program promoted the private and public installation of ecoroofs, resulting in a higher concentration of facilities. Portland’s Central City 2035 comprehensive plan, passed in 2018, requires large new construction projects within the plan boundaries to incorporate, at minimum, an extensive ecoroof (City of Portland Citation2018). With cost being a significant consideration in the benefit of ecoroof installation, the determination of a minimum ecoroof depth capable of providing the intended stormwater control benefits during frequent small rainfall events must occur, securing both functionality and feasibility. As the city continues to engage with designs intended to promote future climate resilience, they have been monitoring selected ecoroof facilities for performance. The ecoroof highlighted in the current study was constructed in 2013, containing 3,441 m2 of vegetated surface (45°35ʹ44” N and 122°40ʹ47”) and faces 303° azimuth. It has three segments, the first consisting of a control section made of traditional roofing design (1468 m2), a second which has a depth of 75 mm, and a third which has 125 mm substrate (Figure 1). The substrate is composed of compost, pumice, and sandy loam. The plant species in the ecoroofs are primarily evergreen sedum, native grasses, and flowers, with the most dominant species being Sedum Takesimense (Ecoroofs Oregon Citation2020). The profile has a one-inch coarse aggregate layer (1/4-3/8” granules), overlain by a typical semi-intensive media. A very dense fabric with a high capillary potential underlies the media (see Figure 3 in Schultz, Sailor, and Starry Citation2018). The fabric is designed to slow runoff and disperse moisture evenly across the roof, maximizing and equalizing moisture uptake. Although there is no synthetic drainage layer, the coarse aggregate layer serves the purpose.
Published online:
30 November 2020
Figure 1. Ecoroof study site located in Portland, OR, USA
Study area
The 375 km2 city has approximately 653,000 residents. It has a Csb climate designation according to Köppen classification, a Mediterranean type regime characterized by drier summers and wetter winter months. The average temperature is mild for years 1981–2010, at 12.5°C, and precipitation averages 913.9 mm annually, with 667.8 mm of that occurring between October and March (U.S.climatedata Citation2019). With urban growth boundaries that limit the outer expansion of urban growth, the city has promoted in-fill development to accommodate the need of the increasing population.
The greater Portland area has been subject to many stormwater concerns, with a reduction in combined sewer overflows a noted priority in management. The city also aims to reduce pluvial flooding, which causes damage through events such as basement sewer backups (City of Portland Citation2016; Michelson and Chang Citation2019). Not all stormwater is treated before discharging into the Willamette River, which transects the more urbanized city center, influencing stormwater management requirements as well. To reduce storm runoff, the city installed various green stormwater infrastructure throughout the city (Baker et al. Citation2019). Such stormwater infrastructure is crucial in local stormwater management, as Portland is projected to see a combination of higher intensity precipitation events (Cooley and Chang Citation2017, Citation2020), as well as a higher volume of precipitation in the winter months (Rupp, Abatzoglou, and Mote Citation2017). Combined with steady population growth (Oregon Metro Citation2016) and associated land development, impervious surface areas are likely to increase in the future, stressing the need for localized longer-term studies.
Data
Precipitation and outflow data from 1 October 2014 to 30 September 2017 were collected by the City of Portland at the Walmart ecoroof in Portland, OR (Figure 1). Flow from the different roof sections was measured with Plasti-fab extra-large 60-degree trapezoidal flumes with Hach US9001 down-looking ultrasonic depth sensors and recorded every five minutes from the section outlets draining both the 125 mm and 75 mm ecoroof sections. This same collection setup was used for the larger conventional roof section, serving as the control case. Hourly precipitation data were derived from the USGS HYDRA network (USGS Citation2018), with the data collected from on-site rain gauges located atop the Walmart roof (station # 220). In case the equipment failed to record runoff data or the recorded runoff data were suspicious (see the assumptions below), we removed such cases for consistency in our statistical analysis. A total of 38 storms (19.5%) were removed, resulting in 157 complete records for the three-year period.
Ecoroof outflow values for the conventional, 75 mm, and 125 mm plots were aggregated to an hourly timescale. Precipitation events were defined as any period when there was more than 2 mm of precipitation following an antecedent dry weather period (ADWP) of at least 12 hours. This is more substantial than the 6-hour ADWP used to distinguish between precipitation events in other literature (Stovin, Poë, and Berretta Citation2013; Schultz, Sailor, and Starry Citation2018). The longer ADWP was used as a means to account for an overall gap in practical knowledge regarding the actual duration of ecoroof discharge following a precipitation event in the wet winter climate. The longer ADWP also ensures that the substrate has adequate time to recover storage capacity between the events, assuming the different depths might require varied dry periods to accomplish this. summarizes mean, minimum, and maximum values of event storm amount, ADWP, and runoff from the three roofs.
Table 1. Descriptive statistics of precipitation, antecedent dry weather period, event storm runoff of Walmart roof in Portland in Water year 2015–2017
Assumptions
We assumed that all the ecoroof plots had identical design characteristics, varying only in their relative substrate depth. The substrate used for both plots was an industry standard mixture of pumice, sandy loam soil, and composted organic materials. However, it is possible that the exact characteristics could have changed over time. Weathering, the addition of organic material (De-Ville, Menon, and Stovin Citation2018), and other pedological processes could indicate that the plots are no longer identical to their design specifications (Bouzouidja et al. Citation2018). Because the precipitation gauge was located on the conventional roof section, the climatic and meteorological conditions were assumed to be consistent across plots. This eliminates the need to incorporate other possible contributing meteorological factors of RR. We also removed suspicious cases when reported runoff values from conventional roofs did not generate more than 60% precipitation (27 events or runoff values from ecoroofs were 20% higher than those from conventional roofs during individual storm events (8 events) or when the runoff from the two ecoroofs are substantially different from each other during the middle wet period (2 events in December 2014). Many of the events that were removed occurred during periods of freezing rain or snow, which make linking inputs to outputs for specific storm events challenging.
Data analysis and statistical methods
After precipitation events and their corresponding outflow were determined, the proportion of discharge to precipitation (runoff ratio) was calculated. We subtracted runoff ratio from 1 and multiplied by 100 to derive event-level RR (Equation (1)), where Q = event total discharge and P = event total precipitation. 
(1)
(1)
To compare RR across the three types of roofs during storm events, we calculated event runoff for each roof from the beginning of the storm until the roof completed generating event runoff. Because the storm event data are not normally distributed even after the application of transformations, we used a non-parametric test. The Kruskal-Wallis Rank Sum Test was chosen to determine whether there are differences among groups; in this case, the retention values recorded for each of the three roof plots. A Dunn’s Test of Multiple Comparisons was applied post hoc to identify groups contributing to model significance.
To visualize the potential differences in RR of the roofs by season, the data were divided into the wet season and the dry season following the categorization specified by Chen and Chang (Citation2019), with the wet season subdivided further into beginning (October-November), middle (December-February), and end (March-April) of season. A box-and-whisker plot was created to compare the distribution of RR values among the different plots between seasons. Precipitation events were divided into four groups – low (less than 5 mm), medium low (5–9.9 mm), medium high (10–14.9 mm), and high (over 15 mm) events. Additionally, ADWP was divided into three categories – low ADWP (less than one day), medium ADWP (1–3 days), high ADWP (over 3 days).
To investigate the relationship between RR and precipitation amount and ADWP, Spearman’s rank correlation coefficients were used by each season. The non-parametric test was selected because precipitation and ADWP data are not normally distributed. Additionally, to determine the response of individual plots to precipitation amount and ADWP groups, a Kruskal-Wallis rank sum test was run for the retention values of all roofs. A Dunn’s Test of Multiple Comparisons was performed on the model to derive additional information. Finally, multiple regression was conducted to identify how much variations of RR are explained by ADWP, storm amount and year. All data preparation and statistical analysis were performed in Microsoft excel SPSS 27, and GraphPat Prism to clean, analyze, and visualize data.
Results
Retention in three roofs by season
A significant relationship exists between roof type and retention performance, as indicated by the Kruskal-Wallis rank sum test (, p < 0.01). Average retention by the ecoroofs ranged from double to nearly three times that of the conventional roof, depending on season (Figure 2). However, there was no significant difference in retention between the two ecoroof plots ().
Table 2. P values of statistical tests comparing differences in rainfall retention ratio by season and different plots of ecoroof
Published online:
30 November 2020Figure 2. Rainfall retention ratios during dry season (n = 34) and early (n = 34), middle (n = 48), and late (n = 41) wet season in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
Figure 2. Rainfall retention ratios during dry season (n = 34) and early (n = 34), middle (n = 48), and late (n = 41) wet season in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
The seasonal and sub-seasonal temporal component appears to impact RR of all the three roof sections (Figure 2). This variation in RR values remains consistent between the 125 mm and 75 mm ecoroof plots within each category. The greatest RR for ecoroof plots was observed during the dry season, particularly for the 125 mm plot (Over 90%, except a few outliers). The opposite was observed for the early and middle wet seasons, where the median RR is low (around 50%) for both ecoroofs.
Retention in the three roofs by precipitation amount and ADWP
As reported in , there are statistically significant positive correlations between ADWP and RR in all wet-seasons with the highest correlation in the late-wet season. However, ADWP and RR are not significantly correlated with each other in the dry season. Event precipitation amount is negatively associated with RR in all seasons. When RR values are compared by different precipitation groups, RR are always significantly higher in the ecoroofs than the conventional roof (p < 0.001). The differences in RR between the conventional roof and ecoroof decrease as precipitation totals increase (Figure 3).
Published online:
30 November 2020Figure 3. Rainfall retention ratio under different storm sizes in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
Figure 3. Rainfall retention ratio under different storm sizes in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
Table 3. Spearman rank correlation coefficients between retention rate and ADWP and precipitation by ecoroof
When the two ecoroofs are compared to each other in each precipitation group, there is no significant difference in RR between the two (p > 0.05). RR values are always significantly higher in the ecoroofs than the conventional roofs in all ADWP groups (Figure 4). The difference between the conventional roof and ecoroofs is greatest for the longer ADWP group. Similar to the case by precipitation totals, there is no statistically significant difference in RR between the two ecoroofs by ADWP groups.
Published online:
30 November 2020Figure 4. Rainfall retention ratio under different antecedent dry weather period in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
Figure 4. Rainfall retention ratio under different antecedent dry weather period in a conventional and 75 mm, and 125 mm ecoroofs in Portland, OR, USA
Regression analysis showed that precipitation and ADWP explain approximately half of the RR variation in both ecoroofs in the mid- and late-wet seasons, while both variables are not significant for explaining RR variation in the dry season (). ADWP is marginally significantly related to RR in the early wet season. ADWP has a slightly higher effect on increasing RR in the 125 mm ecoroof than the 75 mm ecoroof, as indicated by higher coefficient values. One 1 mm increase in storm precipitation amount would result in 0.9% reduction in retention ratio in the late-wet season. Year is statistically significant only for the 125 mm ecoroof in wet seasons.
Table 4. Results of regression analysis for the two ecoroofs by season. Numbers are unstandardized regression coefficients
Figure 5 illustrates how the three roofs responded to a storm event that has a higher than 5-day ADWP. While the conventional roof generated runoff immediately after rainfall, the ecoroofs generated runoff much slower with dampened peak flows, particularly for the 125 mm ecoroof.
Published online:
30 November 2020Figure 5. Storm rainfall and runoff amounts during a representative winter storm in a conventional roof, 75 mm, 125 mm ecoroofs in Portland, OR, USA
Figure 5. Storm rainfall and runoff amounts during a representative winter storm in a conventional roof, 75 mm, 125 mm ecoroofs in Portland, OR, USA
Discussion
Effects of different seasons with different precipitation amount and ADWP
Our findings show that the effectiveness of ecoroofs in retaining storm runoff can be substantially different by sub-seasons. While ecoroofs retain approximately 50% of stormwater during the early and mid-wet seasons, RR values are much higher for the late wet (75%) and dry seasons (close to 100%). Such different seasonal effects are likely associated with different ADWP and storm size in different seasons. The lowest RR in the middle of the wet seasons is likely to be associated with the fact that soils might be saturated after a prolonged wet period with continuous rainfall events (short ADWP). The highest RR of the ecoroofs for the low precipitation events and longer ADWP is expected, given that soil is relatively dry after a prolonged dry period. Hence, the volume of water can be held longer in the soil pore during these lower volume events than high volume events. These findings are consistent with others from a Mediterranean climate (Brandao et al. Citation2017).
Compared to Schultz, Sailor, and Starry (Citation2018)’s study, RR values in our study are higher, which is likely attributed to the two facts. First, we used longer ADWP to separate storms; thus, a shorter duration of ADWP, which could generate more runoff, are excluded in our analysis. Second, since we used the shorter runoff duration compared to Schultz et al.’s study, storm runoff generated from our ecoroofs is likely to be underestimated, resulting in higher RR. However, there are no significant differences between the two ecoroofs in the same season, storm amount, and ADWP group. We suspect that the additional storage of the deeper roof may not be available for retention, as suggested by Talebi et al. (Citation2019), who noted that changes in substrate depth are only impactful on retention capacity if that moisture is available for evapotranspiration. Additionally, the deeper ecoroof plot could take longer to dry out thoroughly, and when ADWP is short, the actual available storage volume for both roofs could be similar at storm onset.
Effects of vegetation and ET
RR is consistently higher during the dry summer period likely attributed to longer ADWP and higher temperatures, which will lead to higher ET. When using local AgriMet daily potential ET data, the dry period has approximately 3.5 times higher ET than the winter wet period. One reason to revisit the Schultz, Sailor, and Starry (Citation2018) paper, which did not observe a large effect of substrate depth on stormwater retention, was to determine whether years later, a more mature plant community might facilitate a depth effect through increased rooting and plant biomass that would be positively associated with ET and RR. Ongoing work does not indicate any substantial effects of depth on species composition or plant height (Starry et al. Citation2017), despite a well- established plant community on both ecoroofs (Figure 1). Seasonal changes in community composition and the noticeable reduction in herbaceous species that go dormant in winter months can also partially explain the reduced performance during the Portland mid-wet-season. That said, the dominant roof species, Sedum takesimense is known for retaining some leaves year-round (Stephenson Citation1994), hence without this species, retention may have been even lower in the mid-to-late wet season.
Implications for management
The overall indication of the study findings is that the 75 mm and 125 mm ecoroof sections are both capable of managing a wide range of precipitation events. This finding could serve local planning, as the structural load requirements of an ecoroof can vary greatly with substrate depth, with requirements ranging from 100 kg/m2 to 3200 kg/m2 by increasing depth of a saturated sandy loam soil by 150 cm (Feitosa and Wilkinson Citation2016). This implies that there will be a structural and financial benefit to installing the slightly thinner ecoroof, without losing the benefit of stormwater runoff mitigation. However, ecoroofs might not be adequate for handling large storm events, when such events are combined with short ADWP, especially in winter when many ecoroof plant species are dormant. This finding implies that a combination of green and grey stormwater infrastructure types (Zhang, Fong, and Chui Citation2018) could be a better approach to controlling stormwater than merely relying on one type of infrastructure during periods of higher flood risk. Given that winter precipitation is likely to intensify under climate change scenarios in the study region (Cooley and Chang Citation2020), urban storm managers should take such information for future planning. Additionally, the potentially changing performance of the deeper ecoroof needs further investigation since year is a significant explanatory variable for explaining RR variation in the 125 mm ecoroof, but not in the 75 mm roof.
Conclusions
The results of this study imply that both 75 mm and 125 mm ecoroofs provide a similar level of effectiveness at controlling stormwater runoff throughout the year. Both ecoroofs have substantially higher retention ratios compared to the conventional roof. When comparing the two ecoroofs’ performance under various storm amount and ADWP, the performance is significantly improved under small rainfall events and longer ADWP as opposed to larger rainfall events and shorter ADWP. Both ADWP and storm size explain approximately 50% of RR variations in mid- and late-wet seasons in both ecoroofs. The early wet season retention was explained in part by year, at least for the 125 mm ecoroof section. Ecoroofs’ RR rates were highest during the dry season and the lowest during the early and middle of the wet season when soils were saturated.
While the extra substrate depth did not result in increased RR, more work is needed to determine whether this could have been due to erosion or whether a 50 mm increase in depth was not ecologically meaningful. Different depth ranges need to be explored, possibly through modeling work, and plant selection needs to be evaluated in the context of seasonal precipitation patterns. Furthermore, because storm size and timing, as well as year, were key predictors of RR, more work is needed to understand how both local weather and a potentially changing climate may affect ecoroof performance. Long-term monitoring of soil moisture as well as runoff in conjunction with ecoroof maintenance can reveal some of the nuanced differences in runoff retention ratios between the two ecoroofs.
| Descriptive statistics | Precipitation (mm) | Antecedent dry weather period (day) | Conventional roof runoff (mm) | 75 mm ecoroof runoff (mm) | 125 mm ecoroof runoff (mm) |
|---|---|---|---|---|---|
| Mean | 14.64 | 4.67 | 11.85 | 6.65 | 6.34 |
| Min | 2.03 | 0.5 | 1.33 | 0 | 0 |
| Maximum | 112.3 | 57.38 | 78.98 | 74.06 | 75.68 |
| Test | Early-wet (n = 34) | Mid-wet (n = 48) | Late-wet (n = 41) | Dry (n = 34) |
|---|---|---|---|---|
| Kruskall-Wallis | < 0.001 | < 0.001 | < 0.001 | < 0.001 |
| Pairwise comparison | ||||
| Conventional vs. 75 mm | < 0.001 | < 0.001 | < 0.001 | < 0.001 |
| Conventional vs. 125 mm | < 0.001 | < 0.001 | < 0.001 | < 0.001 |
| 75 mm vs. 125 mm | 0.922 | 0.967 | 0.698 | 0.625 |
For Pairwise comparison, significance values have been adjusted by the Bonferroni correction for multiple tests.
| Ecoroof | Early-wet (n = 34) | Mid-wet (n = 48) | Late-wet (n = 41) | Dry (n = 34) |
|---|---|---|---|---|
| ADWP vs. RR | ||||
| 75 mm | 0.501** | 0.365* | 0.697** | 0.176 |
| 125 mm | 0.510** | 0.439** | 0.717** | 0.123 |
| Precipitation vs. RR | ||||
| 75 mm | −0.211 | −0.617** | −0.760** | −0.623** |
| 125 mm | −0.103 | −0.638** | −0.729** | −0.474** |
** statistically significant at the 0.01 level.
| 75 mm ecoroof | Early wet (n = 34) | Mid-wet (n = 48) | Late-wet (n = 41) | Dry (n = 34) |
|---|---|---|---|---|
| Precipitation (mm) | −0.4* | −0.6*** | −0.9*** | −0.6 |
| ADWP (day) | 0.044* | 0.028*** | 0.031*** | 0.003 |
| R2 | 0.156* | 0.457*** | 0.620*** | 0.092 |
| 125 mm ecoroof | ||||
| Precipitation (mm) | −0.003* | −0.7*** | −0.009*** | −0.002 |
| ADWP (day) | 0.045* | 0.033*** | 0.040*** | 0.002 |
| Year (2017) | −0.196* | 0.150* | ||
| Year (2015) | −0.211*** | |||
| R2 | 0.258* | 0.546*** | 0.478*** | 0.296* |
*statistically significant at the 0.1 level; **statistically significant at the 0.05 level; ***statistically significant at the 0.01 level.
Acknowledgements
This material is based upon work supported by the National Science Foundation under Grant Number SES-1444755, which established the Urban Resilience to Extremes Sustainability Research Network. Views expressed are our own and do not necessarily reflect the views of the National Science Foundation. We would like to extend a special thanks to Adrienne Aiona and Peter Abrams at the City of Portland’s Bureau of Environmental Services for providing data and sharing ongoing knowledge of the site. We appreciate two anonymous reviewers who helped clarify some points of the manuscript. Martin Lafrenz at Portland State University also offered some methodological insights during the initial phase of the project.
Disclosure statement
No potential conflict of interest was reported by the author(s).