Water, Air, & Soil Pollution

, Volume 216, Issue 1, pp 83–92

A Pilot-Scale Evaluation of Greenroof Runoff Retention, Detention, and Quality

Authors

  • Norman Buccola
    • Civil and Environmental Engineering DepartmentPortland State University
    • Mechanical and Materials Engineering DepartmentPortland State University
Article

DOI: 10.1007/s11270-010-0516-8

Cite this article as:
Buccola, N. & Spolek, G. Water Air Soil Pollut (2011) 216: 83. doi:10.1007/s11270-010-0516-8

Abstract

Vegetated roofs are becoming more commonly deployed as a means of mitigating stormflow in urban areas. A greenroof performance comparison of stormwater runoff has yet to be conducted with controlled rain events and quantifiable antecedent soil moisture. This study aimed to investigate the rainwater management provided by various greenroof design schemes. Runoff retention, peak flow lagtime, conductivity, and pH of ten different small-scale greenroof schemes were observed and analyzed under repeatable rain simulations in a pilot-scale study. Sedum rupestre Angelina, Sedum hispanicum, Trifolium repens (white clover), Trifolium pratense (red clover), Vinca major (Big-Leaf Periwinkle), and Lolium multiflorum (ryegrass) were grown in the same type of soil media but separate 2′ × 2′ trays at depths of 5 cm and 14 cm to observe how soil depth and root zone development affects stormwater flow through for each plant type. Results showed that increased green roof soil depth improved water retention and runoff lagtime; the effect of plant type was mixed and inconclusive. Runoff conductivity test results depended primarily on soil depth and the existence or absence of vegetation. Testing results show that pH normalization provided by a greenroof does not depend significantly with the substrate depth.

Keywords

Best management practicesLow impact developmentStormwater managementRunoffNonpoint source pollutionGreenroofsEcoroofs

1 Introduction

Increasingly, building owners and city planners are turning to vegetated roofing systems, commonly known as greenroofs, as a means to mitigate stormflow in urban areas. As stormwater retention increases for a given management scheme, the amount of stormwater load on municipal sewer systems decreases. With many older cities that have combined sewers, a stormwater surge during an intense rain event could lead to an overflow of untreated sewage into receiving water bodies. While conventional roof runoff may contain heavy metals (e.g., lead, copper, or zinc) or toxics (such as polycyclic aromatic hydrocarbons) that may be detrimental to water quality, widespread usage of retention devices such as greenroofs can reduce the quantity of water unleashed during a storm, thus reducing the mass loading of any contaminants contained in the stormwater surge. At the same time, runoff detention structures such as subterranean large-diameter pipes or above ground ponds can hold large amounts of water temporarily to reduce the wave of runoff that occurs immediately after a storm event. Greenroofs also detain stormwater by delaying the runoff peak from a storm. By using greenroofs to help lower the peak runoff flowrate for a given drainage area, sewer network design and construction may be more efficient with respect to materials (smaller diameter pipes needed), maintenance (less burden on wastewater treatment plants), and cost.

Many researchers have monitored the rainfall–runoff behavior of greenroofs (DeNardo et al. 2003; Liptan and Strecker 2003; Moran et al. 2004; Whitlow 2006; Carter and Rasmussen 2006). Although the large variation in local climatic conditions, testing period/conditions, growing medium depth/composition, and vegetation coverage/composition make it difficult to compare all of these results side by side, the variation in water retention from these independent greenroof runoff studies mentioned can be compiled and averaged to yield an annual retention rate over wet and dry seasons of roughly 54%. Furthermore, there is a general agreement from the greenroof monitoring literature that greenroofs have a greater ability to retain stormwater during drier conditions and/or lower intensity storm events. Studies conducted by Kohler et al. (2002) and Hutchinson et al. (2003) have also shown that increased substrate depth will improve greenroof runoff retention rates. Teemusk and Mander (2007) also presented evidence that vegetation density can affect the ability of a greenroof to retain rainwater thereby reducing stormwater runoff. Antecedent soil moisture appears to greatly affect the ability of a soil layer to absorb precipitation and delay or reduce storm runoff. Although soil moisture has been monitored by some greenroof monitoring studies (Kohler 2004; Van Seters et al. 2007), it has not been measured explicitly in repeatable experimental hydrologic greenroof monitoring as it has in this current work.

Greenroofs also appear to impact the quality of runoff. Many researchers have found that the phosphorous loadings in greenroof runoff are higher than normal flat roof runoff loadings (Forster and Knoche 1999; Clark et al. 2001; Hutchinson et al. 2003; Van Metre and Mahler 2003; Moran et al. 2004; Hunt et al. 2006; Van Seters et al. 2007). While phosphorus is in an indication of solid particles in solution, electrical conductivity (EC) measures the amount of free ions (salts) in solution and is easily measured in the field. A study by Berghage et al. (2007) showed that greenroof runoff resulted in higher EC than did nongreened roof runoff. High runoff conductivity can limit the amount of habitat-forming plant species in a water body and may indicate that higher concentrations of more toxic trace metals may be present. High free ion concentration in water can also be a nuisance to household plumbing and lead to calcification of pipes in boilers. In areas with acidic rainwater (pH between 4 and 6), greenroofs have been shown to normalize and raise the pH of runoff (pH between 6 and 8; Berghage et al. 2007; Teemusk and Mander 2007). The buffering of roof runoff in urban environments could be an important improvement as many aquatic species are intolerant of a pH of 5.7 or less. With an increasing number of municipalities looking to reuse water, opportunities for on-site water pretreatment such as greenroofs are becoming more popular by necessity.

One of the primary reasons that many designers, architects, engineers, and planners build greenroofs is to decrease storm flow, thereby improving water quality in receiving waters. To achieve these design goals, the optimum greenroof soil depth and plant type(s) should maximize retention without sacrificing water quality. Thus, the need to better understand the behavior of greenroof runoff quantity and water quality as a function of plant type and media depth is the premise of this body of work.

2 Materials and Methods

A range of greenroof plant types, soil depths, and storm conditions were tested and analyzed in order to investigate their potential to reduce (retention) or delay (detention) stormwater flows and nutrient/contaminant loadings from urban areas. This study focused on varying plant type, growing media depth, and storm intensity. Antecedent soil moisture, air temperature, relative humidity, rain water temperature, rainstorm flowrate, runoff temperature, and runoff flowrate were monitored for each plant type.

A specially designed greenroof wind tunnel was constructed in the Greenroof Design and Test Laboratory at Portland State University (see Fig. 1) with two experimental goals: (1) to measure the heat transfer through a greenroof and calculate the effective R value and (2) to measure the water retention and detention characteristics and runoff water quality of a greenroof. The focus of this paper is on the latter experimental goal. Recreating a roof-top environment, an upper air deck had the capability of circulating hot air (simulating a hot roof; 20–60°C) while cooled air (simulating a cool building 10–24°C) could be circulated through the lower deck. Two full-spectrum lights with roughly one tenth the energy of natural sunlight (approximately 100 W/m2) simulated sunlight. Air temperature and relative humidity were measured in the upper air deck, while cold air temperature was measured in the lower deck. Variable flow shower heads (0.0012 to 0.014 L/s) were plumbed into the upper air deck to create various storm conditions on small-scale greenroofs. Ambient air relative humidity remained fairly constant within a range of 20–30% for all tests. Wind speed was controllable up to 5 mph. To limit the entrainment of simulated rain water by wind, wind speed was minimized for all experiments. Overall, the majority of testing was done under low airflow with constant air temperatures of roughly 28°C in the upper air deck and 23°C in the lower air deck.
https://static-content.springer.com/image/art%3A10.1007%2Fs11270-010-0516-8/MediaObjects/11270_2010_516_Fig1_HTML.gif
Fig. 1

Testing apparatus schematic and photo

The greenroof wind tunnel was designed to accommodate two, 0.6 m × 0.6 m square trays side by side at a 0.02° slope (1 cm in 55.4 cm) to facilitate drainage. The trays were constructed of steel sheet metal with 9.2 cm-high sides, expandable to 18.5 cm high during deep-soil experiments. Tray expansions were attached to the bottom tray using fasteners, silicone sealant, and a layer of rubber to limit leakage around the extension interface.

A drainage layer was placed in the bottom of each plant tray (Henry DB-50 greenroof drainage layer material) followed by the growing media (Pro-Gro Mixes’ Intensive Greenroof Blend consisting of pumice, sand, and two types of compost). The first phase of experiments employed a growing media depth of 5 cm while the second phase was conducted using a soil depth of 14 cm. A particle size analysis was conducted to reveal that the growing medium contained on average 11% gravel, 78% sand, and 11% silt and clay by weight.

The plants chosen for the experiments were as follows: Sedum rupestre Angelina, Sedum hispanicum, Lolium multiflorum (perennial ryegrass), Trifolium repens (White Clover), Vinca major (Big-Leaf Periwinkle), and a proprietary wildflower seed mix (Ed Hume brand: “Wildflower Meadow Mix”) that consisted primarily of Trifolium pratense (Red Clover) at the time of testing. All plants were grown and maintained in a commercial greenhouse on campus, and temporarily moved to the test lab for no longer than 1 week at a time. T. repens, L. multiflorum, and T. pratense were planted from seed while Catharanthus roseus, S. rupestre Angelina, and S. hispanicum were planted as starts from a local nursery. Upon the completion of the 5 cm soil depth tests, V. major, L. multiflorum, and S. hispanicum were transplanted and laid atop an additional 9 cm of growing media for a total soil thickness of roughly 14 cm per tray. T. pratense was reseeded at a 14 cm depth while S. rupestre Angelina remained at the original 5 cm depth due to its lack of development and vegetative coverage.

Establishing design storms for the greenroof system experiments had multiple constraints at the time of construction: materials and irrigation emitters available, peak hydraulic conductivity of the soil, and individual experiment time (in order to conduct all necessary experiments within an allotted time frame). The two design storms used in this study are summarized in Table 1. This simulated rainfall served as conservative estimates of the extreme conditions to which a greenroof would be exposed. Incoming flowrate of the experimental precipitation was calculated from using the slope of the test bed weight change history from the beginning of the storm to the point that runoff first began. Intensity (I) was related to average flow (Q) and effective area (A) by I = Q/A.
Table 1

Experimental storm sizes

 

Heavy (n = 36)

Medium (n = 15)

Average flow (L/s)

0.014

0.0012

Standard deviation (L/s)

0.004

0.0008

Duration (min)

15.0

28.0

Total volume (L)

12.5

2.0

Effective area (m2)

0.15

0.15

Intensity (mm/h)

340.0

30.0

The heavy and medium storm was made possible by two, coupled sets of overhead spray emitters; one roughly centered over each plant tray (see Fig. 2). Although the spray emitters did not cover the total area for each plant tray, the shower location for both heavy and medium storms stayed consistent through all experiments.
https://static-content.springer.com/image/art%3A10.1007%2Fs11270-010-0516-8/MediaObjects/11270_2010_516_Fig2_HTML.gif
Fig. 2

Photo of medium storm setup

Prior to testing, each tray of plants was weighed in order to determine its relative soil moisture in relation to the residual soil moisture. It was assumed that the weight of the dry plant material (woody matter) was negligible compared to the water held in them. This follows a general assumption used in gardening literature that plants contain roughly 75–90% water (Tenenbaum et al. 2001; Champagne et al. 2003). A digital scale was modified to interface with an Agilent 34970A data logger. The weight of the water in the collection vessel could then later be post-processed by taking the difference of each set of adjacent weight recordings to yield a mass flowrate for each time increment.

Water conductivity was measured with an Orion conductivity/salinity meter Model 140. Water pH was measured with an Orion Model EA 920 m and Orion 9157BN probe which was calibrated before each day of testing. For both conductivity and pH, tap water was measured as the control specimen on each testing day and compared to greenroof stormwater runoff.

3 Results and Discussion

A total of 51 runoff quantity tests were completed (27 at 5 cm soil depth and 24 at 14 cm soil depth) yielding measurable water retention characteristics for each greenroof. For the majority of these water quantity tests, runoff conductivity and pH were also measured immediately afterward. As measured by the soil supplier, the porosity of the soil used in all experiments was 0.73, the dry bulk density was 0.74 g/cm3, the saturated density was 1.28 g/cm3, and the density at field capacity was 1.22 g/cm3.

3.1 Runoff Detention/Retention

The test-averaged runoff hydrograph from 5 cm depth tests (heavy and medium intensities) is compared to the averaged runoff from 14 cm depth tests (heavy and medium intensities) in Fig. 3. In each case, the precipitation was constant for the prescribed duration, then ceased. Discharge would typically increase until the peak runoff was achieved, then would decrease. Retention for each test is the ratio of the integrated discharge curve divided by the integrated rainfall curve. Detention is the time between the area centroid of the rainfall curve to the peak of the discharge curve. As can be seen in Fig. 3, the overall averaged runoff from all 14 cm soil depth, heavy flow tests had a peak that was smaller in magnitude and occurred later than that of the averaged 5 cm depth tests. A unit hydrograph (UH) based on heavy flow data averaged across 5 and 14 cm depths was created to re-interpret lower flow runoff data. While the UH is a rainfall–runoff model used to predict the direct runoff hydrograph from a particular drainage basin given a unit of rainfall (Gupta 2001), in this case, it helped to visualize the rainfall–runoff relationships under the medium flow scenario.
https://static-content.springer.com/image/art%3A10.1007%2Fs11270-010-0516-8/MediaObjects/11270_2010_516_Fig3_HTML.gif
Fig. 3

Average runoff profile for 5 and 14-cm soil depth flow tests; n number of sample tests, avg runoff runoff hydrograph averaged over all tests of given storm type and soil depth, unit hydrograph model based on test-averaged heavy storm runoff

Also illustrated by Fig. 3, retention and detention were both affected by the intensity of the rain event and by the depth of the soil. The summary results are given in Table 2, where it can be clearly seen that the retention increased with decreasing rainfall and that retention also increased with an increase in depth.
Table 2

Overall average runoff retention and lagtime values

Intensity

Overall average retention

Overall average lagtime (min)

5 cm depth

14 cm depth

5 cm depth

14 cm depth

mean

err

mean

err

mean

err

mean

err

Heavy (34 cm/h)

20%

16%

56%

12%

3.7

3.0

4.9

1.1

Medium (3.0 cm/h)

36%

18%

64%

30%

5.3

2.3

8.1

2.2

err one standard deviation

Unexpectedly, the test-averaged runoff data from the heavy storms displayed a clear peak in flowrate before the end of the applied rain for both 5 and 14 cm depth tests. This phenomena is not fully understood, but may be explained as follows: As the experimental rain covered approximately 40% of the greenroof tray area, the runoff flowrate typically increased gradually as the soil under the shower became saturated first. Upon the saturation of this area, there may have been lateral water movement into unsaturated areas may have resulted in lower runoff flowrates before the end of the experimental storm.

Table 3 displays a correlation matrix relating preexisting input variables of soil depth, storm intensity (“intensity”), and antecedent soil moisture (“anticedent.s.m”), with output variables of detention, retention, and conductivity. For this analysis, bare soil tests were excluded. Notably, retention had a positive correlation with soil depth (0.68) and negative correlation with storm intensity and antecedent soil moisture (−0.40 and −0.39, respectively). This indicates that retention increased under drier antecedent soil conditions, lower intensity rainfall, and a deeper greenroof depth. Detention, or lag time, showed some of the same trends as retention: increased detention for decreasing rainfall intensity (−0.82 correlation) and increased detention with an increase in soil depth (0.37 correlation).
Table 3

Correlation matrix for measured variables

 

Soil depth

Intensity

Anticedent s.m.

Detention

Retention

Conductivity

Soil depth

1.00

−0.32

−0.18

0.37

0.68

0.20

Intensity

−0.32

1.00

0.04

−0.82

−0.40

−0.30

Anticedent s.m.

−0.18

0.04

1.00

0.07

−0.39

−0.10

Detention

0.37

−0.82

0.07

1.00

0.46

0.25

Retention

0.68

−0.40

−0.39

0.46

1.00

0.25

Conductivity

0.20

−0.30

−0.10

0.25

0.25

1.00

antecedent s.m. antecedent soil moisture

Extrapolation of retention and detention results presented for this pilot-scale study to a full-scale greenroof requires examination. Retention measured at the pilot-scale, as given in Table 2, was comparable to what would be expected for an actual greenroof because the soil depth tested is typical of most greenroof installations. Retention was normalized by the rain flux, in volume per unit area, so it is reasonable to expect a 5-cm soil depth to show 20–50% retention and a 14-cm soil depth to show 55–65% retention. Detention, on the other hand, as indicated by the lag time between the rainfall and the subsequent discharge of that water influx, depends both on soil depth through which the water drains and upon the transverse dimension of the greenroof, the distance that the discharge must travel to the drain. In pilot-scale tests reported here that transverse characteristic length was about 0.5 m substantially less than the 10–20 m for a full-scale greenroof. Published field studies of full-scale greenroofs have shown lag time to be on the order of an hour or more (Spolek 2008) rather than on the order of minutes as given in Table 2. The detention results reported here, therefore, indicate the relative change rather than the absolute value.

Different plants were selected and tested based on a variety of criteria. Two trays of each plant were grown and tested with the results averaged. Plant trays were tested multiple times at different rain intensity and initial soil moisture. Results varied widely, and no statistical differences could be substantiated due to that variance. Some trends relating to plant type can be identified, though, comparing average retention rates based on plant type listed in Table 4. Under heavy storms (340 mm/h), all plants retained more rain than bare soil, while under medium storms (29 mm/h) this same trend did not exist. All plants showed greater retention when planted in deeper soil and most showed lower retention for higher intensity rainfall. Beyond those general trends, no single plant species stood out as the best option to optimize rainwater retention. With regards to detention, Table 4 shows T. pratense to have the highest average lagtime when compared to other plant types under equivalent soil depth and storm intensity.
Table 4

Average retention for different plant types

 

Retention

Detention

Rain intensity

340 mm/h

29 mm/h

340 mm/h

29 mm/h

Soil depth

5 cm

14 cm

5 cm

14 cm

5 cm

14 cm

5 cm

14 cm

Plant

 Bare soil

13%

45%

46%

76%

0

2.7

4.5

6.3

 T. pratense

36%

60%

N.D.

73%

5

4.5

N.D.

10

 V. major

24%

51%

N.D.

34%

4.8

2.5

N.D.

6.5

 L. multiflorum (ryegrass)

14%

51%

27%

71%

2.5

3.5

4.5

7.8

 S. hispanicum

24%

59%

17%

71%

3.2

3.5

6.5

6.8

 S. rupestre Angelina

20%

N.D.

56%

N.D.

3.2

N.D.

2.5

N.D.

N.D. not determined

3.2 Water Quality

Both 5-cm soil depth and 14-cm soil depth plant trays were tested for conductivity (indicative of the amount of dissolved ions in solution) and pH. In total, 42 conductivity tests were completed (18 at 5 cm, 24 at 14 cm depths, respectively). Figure 4 shows how the mean conductivity of the runoff varied with the greenroof factors that were hypothesized to have an effect: soil depth, rainfall intensity, and whether the greenroof was vegetated or bare. Over the total of 42 test runs, the average runoff conductivity was 253 μS/cm, as compared to 37 μS/cm for the incoming rain water. So the conductivity of the discharge is substantially increased as it passes through the greenroof.
https://static-content.springer.com/image/art%3A10.1007%2Fs11270-010-0516-8/MediaObjects/11270_2010_516_Fig4_HTML.gif
Fig. 4

Effect of vegetation, soil depth, and rainfall intensity on discharge conductivity

Examination of Fig. 4 reveals that for all rainfall intensities and soil thicknesses, the conductivity was lower for vegetated roofs when compared to bare soil roofs. For control of runoff conductivity, and its impact on receiving water bodies, vegetation had a noticeable beneficial effect. Figure 4 also shows that vegetated roofs with thicker soil, an increase from 5 to 14 cm in this study, induced a commensurate increase in conductivity. Table 5 shows the results of a multiple linear regression with conductivity as a function of plant type, soil depth, storm intensity, and antecedent soil moisture (adjusted R2 value of 0.54) showed this observed difference between substrate depths to be statistically significant (p value = 0.0087) in determining runoff conductivity. This effect is probably due to a greater residence time or surface contact between the water and soil. In any case, thinner soil vegetated roofs had lower conductivity.
Table 5

Multiple linear regression with conductivity as the dependent variable

 

Estimate

Std. Error

t value

p value

Significance

(Intercept)

340.391

124.959

2.724

0.01117

*

Plant type

−22.435

11.597

−1.935

0.06358

.

Soil depth

37.283

9.96

3.743

0.00087

***

Intensity

−435.973

244.439

−1.784

0.08574

.

Antecedent s.m.

−97.963

108.985

−0.899

0.37667

 

Detention

−1.755

4.206

−0.417

0.67974

 

Retention

−80.859

84.791

−0.954

0.34873

 

Antecedent s.m. antecedent soil moisture

Significant codes: 0 “***”; 0.001 “**”; 0.01 “*”; 0.05 “.”; 0.1 “ ” 1

Residual standard error: 76.7 on 27° of freedom

Multiple R2, 0.6256; Adjusted R2, 0.5425

F statistic, 7.521 on 6 and 27 DF

p value 8.286e-05

Although all of the pH measurements of water runoff from both the 5 and 14 cm depth tests resulted in the normal range between 6 and 8, almost every instance also showed an increase in pH of the runoff when compared to the incoming rainwater. Overall, pH was increased by an average of 0.43 for all tests. Seemingly, Fig. 5 shows that an increase in soil depth by 9 cm (from 5 to 14 cm) did not affect the ability of the greenroof to increase pH of the stormwater.
https://static-content.springer.com/image/art%3A10.1007%2Fs11270-010-0516-8/MediaObjects/11270_2010_516_Fig5_HTML.gif
Fig. 5

Change in pH from rainfall to discharge as affected by greenroof soil thickness, rainfall intensity, and presence of vegetation

4 Conclusions

What might be hoped for, a single combination of soil depth and plant type that yields the greatest runoff retention and detention while simultaneously maximally reducing conductivity and increasing pH, was not discovered. As with most engineering designs, a compromise solution must be reached. Since plant type does not seem to have a significant effect on either discharge quantity or quality, plant selection can be based on other factors, such as energy transfer, plant hardiness, and aesthetics. For many of the plant types tested, summertime irrigation would be necessary for sustenance, which adds to both first and operating costs for the greenroof.

Soil depth is a design decision that will affect overall greenroof performance. While additional soil depth will yield greater retention, it will also yield a higher conductivity (a potential indication of suspended solids) in solution. This is detrimental to water quality. As plants become mature and roots take hold of the soil, the amount of solids washed away may be lower with older plants. As different plants age, their root structures may age differently. Due to their continued growth every year, it is reasonable to assume that the roots of perennials would provide more sediment containment than annuals.

These pilot-scale greenroof tests have shed light on the potential water retention and runoff water quality differences between plant type in both 5 and 14 cm depth vegetated roofs. Plant types tested in this study displayed positive correlations with substrate depth, rainwater detention, and rainwater retention. This agrees with previous research: an increase in greenroof growing media depth leads to higher runoff retention rates and extended runoff lagtimes. A deeper soil depth resulted in a greater increase in greenroof runoff lagtimes during medium storms (30 mm/h) than during heavy test storms (340 mm/h). T. pratense tests resulted in the highest average runoff lagtime (detention) when compared with other plant types under similar conditions. The key water quality findings from these experiments have led to important conclusions: Measured pH increases in vegetated roof water runoff had no dependence on plant type or soil depth. Further, a deeper vegetated roof growing media depth displayed higher runoff conductivity. Finally, these results confirm that a vegetated roof will have lower runoff conductivity when compared to an unvegetated roof of bare soil.

Copyright information

© Springer Science+Business Media B.V. 2010