The influence of a green roof configuration’s moisture balance on hydrological performance - an abbreviated version for World Green Roof Day 2020

Dr Simon Poe

It has been over 3 years since I finished a PhD in Civil Engineering with the above title and I guess that the first World Green Roof Day is as good a time as any to share some of the findings with a broader audience than the world of academia.

A number of findings from the research have already been disseminated through publication in academic journal papers (listed below). However, this article contains a number of findings that have not been published and I hope that it will be of some interest to the green roof community.

The overall aim of my thesis was to improve the understanding of the physical controls that affect a green roof’s hydrological response, leading to the development of a model that accounts for the relevant physical parameters and processes. This model can then be used to predict response and, potentially, optimise system design.

Background

Green roofs are a viable SuDS component because they can partly offset the loss of urban terrestrial landscape and help to maintain the site’s pre-development hydrology through the effects of retention and detention of rainwater.

In this context, retention refers to the volume of rainfall that is retained within the green roof system and does not leave the roof as runoff. The retention response is affected by the configuration, antecedent conditions and rainfall characteristics.

Detention refers to the temporal delay that occurs between rain (that is not retained) falling on to the roof and emerging as runoff. Detention combines the effects of delays in runoff due to plant cover, vertical movement through the substrate, interactions between the plant roots and the substrate, horizontal transfer across the drainage layer and, on a full-scale roof, the subsequent route into the collection system that is downstream of the roof.

Experimental Setup

The research was carried out at the University of Sheffield (Department of Civil & Structural Engineering) under the tutorship of Dr Virginia Stovin and my employers, Alumasc, sponsored my involvement.

The hydrological responses of nine small (3 m2) test beds (TB), combining three substrates and three vegetation treatments, were monitored on a roof at the University in Sheffield (UK) over a continuous 4 year period between February 2010 and February 2014.

The configurations were designed to replicate typical extensive green roof systems, but also to identify the impact of different physical characteristics of the component parts. Two Alumasc brick-based substrates (HLS and SCS) were used. A third “experimental” mix comprising LECA was also trialled. Each substrate was combined with one of three vegetation treatments:

Sedum vegetation comprises hardy, succulent plants that have the greatest moisture storage potential. ET was slower with Sedum than with Meadow Flower. However, Sedum had greater tolerance to drought and extreme temperatures.

Meadow Flower contains some succulent species, but is mainly comprised of wild flowers and thirsty grasses. A greater density and variety of foliage was observed in summer but greater variability in coverage was witnessed during other seasons. ET rates were higher in spring and summer with Meadow Flower. However, this is exacerbated drought conditions.

Non-vegetated surface that provides a reference roof to identify the impact of vegetative cover. This was maintained to be free from vegetation. The lack of vegetative cover led to a lower maximum storage capacity, but benefitted from high initial evaporative losses.

The references to each of the nine configurations are shown in Table 1, together with the configuration’s maximum moisture storage capacity (SMAX) in mm/m2.

Table 1: References and storage capacities of the configurations

Rainfall, runoff, climate (air temperature, solar radiation, relative humidity and wind speed) and moisture content were monitored. The test setup was financed by the EU’s European Regional Development Fund and designed by Kasmin et al. (2010).

Over the four year period, 323 events were recorded with rainfall greater than 2 mm. Valid runoff responses were measured for between 164 (TB9) and 257 (TB6) events. A full record was available for all nine test beds (AE9) for a sub-set of 48 events. It is this sub-set that is analysed here.

Results & Analysis

Predictably, the data highlighted large variations in the hydrological performance of green roofs. Across the 48 events, per-event retention ranged between 7% and 100%. The range of peak attenuation results was similarly wide (9-100%).

However, runoff generally reflected the differential between (i) rainfall depth and (ii) the available moisture capacity, or soil moisture deficit (SMD) at the start of the rainfall event. These two parameters will now be looked at in further detail.

(i)                 Rainfall

An inverse relationship between retention and rainfall depth was evident in this study. The lowest per-event retention performance was typically associated with larger, lower probability rainfall events (see Figure 1).

Figure 1: Per-event retention for AE9 events, categorised by rainfall depth

When rainfall was less than 5 mm, retention was very high, ranging between 93.7% (TB9) and 99.1% (TB7). This high retention was expected as such rainfall depths would be lower than the maximum storage capacity of all test beds (TBs). As rainfall increased, mean per-event retention levels had a greater range across the different configurations. For example, when rainfall was greater than 20 mm, retention ranged between 14.9% [TB9] and 54.8% [TB1].

Observations of peak attenuation broadly mirror the trends identified for retention. The highest mean peak attenuation was recorded during events with the lowest rainfall depth (see Figure 2).

Figure 2: Peak attenuation for AE9 events, categorised by rainfall depth

When rainfall was less than 5 mm, mean peak attenuation ranged between 96.1% (TB9) and 99.0% (TB7). However, when rainfall was in excess of 20 mm, a wider range of attenuation was observed, with a low of 25.2% (TB9) and a high of 64.4% (TB1).

(i)                 Soil Moisture Deficit

Green roofs have a finite retention capacity that is governed by the vegetation and substrate configuration. The extent to which this maximum capacity (SMAX) is available for retention at the onset of any individual rainfall event (i.e. SMD) will depend on the removal of moisture through ET during the antecedent dry weather period (ADWP).

Seasonal differences in the inter-event regeneration of available moisture capacity therefore impacted retention responses. However, as can be seen in Figure 3, actual seasonal-mean retention was also influenced by the incidence of rainfall.

Figure 3: Seasonal mean rainfall and runoff

A clear trend in the data was evident in winter, where retention of between 32-38% was associated with the highest mean rainfall and the lowest pre-event moisture deficits. However, excepting winter, seasonal differences in configuration median per-event retention percentages were minor due to low mean rainfall.

The influence of seasonal climate upon retention potential (rather than actual retention) was therefore considered through analysis of the seasonal mean SMD prevailing at the start of a rainfall event. This analysis relies upon the CS616 moisture content data and therefore considers the regeneration of available capacity (or SMD) within TB1, TB2, TB3 and TB7 (see Figure 4).

Figure 4: Configuration mean soil moisture deficit after the ADWP

Predictably, the greatest pre-event SMD was measured in summer conditions, followed by spring (vegetated configurations only), autumn and winter. The LECA-based TB3 had the greatest theoretical mean pre-event SMD. However, with LECA-based beds, runoff was frequently observed before θFC. For this reason, the analysis focussed on the SMD statistics for TB1 and TB2 (for vegetated configurations) and TB7 (for non-vegetated configurations).

Based on mean pre-event SMD, in summer, maximum potential retention would have been between 16.2 mm (TB7) and 17.7 mm (TB2). This would have reduced to between 10.9 mm and 13.0 mm in spring (for TB1 and TB7 respectively), 9.3 mm to 13.6 mm in autumn (for TB1 and TB7 respectively) and 0.1 mm to 4.1 mm in winter (for TB1 and TB2 respectively).

These soil moisture deficit trends broadly mirror the seasonal variations in retention with lower retention efficiency in winter than in summer. This is consistent with the trends published in Poe et al (2015) where there was (i) a statistically significant difference in seasonal evapotranspiration and (ii) differences due to configuration. With regard to the latter, there is a trade-off between high initial rates of ET from a non-vegetated bed and the greater moisture storage capacity of a vegetated bed. There was a similar trade-off when comparing Sedum with Meadow Flower. Meadow Flower will typically regenerate the SMD more quickly (due to high ET). However, this can exacerbate the risk of drought stress. Unless irrigation measures are included, the vegetation is at risk of permanently wilting.

 Model Development

A moisture balance model was developed with the benefit of the empirical data generated by the research. The model predicts the response of the green roof to simulated rainfall, accounting for the influences of configuration and climate. The soil moisture deficit is modelled. ET is estimated using a plant- and seasonally-influenced soil moisture extraction function (SMEF). Runoff is modelled with high temporal resolution by adopting a reservoir routing approach and a detention parameter.

The predictive accuracy of the model was validated using the Nash-Sutcliffe Model Efficiency (NSME) approach, scoring a “very good” rating.

Model Application

Firstly, the model was applied to simulate hydrological responses to a continuous 30 year period. Here, the model predicted the responses of three configurations: TB2 (Sedum mat on SCS) – a configuration that is typical of a Sedum roof in the UK; TB4 (Meadow Flower on HLS) – a good representation of a planted biodiverse roof in the UK; and TB7 (non-vegetated HLS) – a configuration that resembles a brown or unplanted biodiverse roof.

Subsequently, the model simulation of the response of TB2 to design rainfall events with return periods of 1, 30 and 100 years is presented.

30 Year Simulation

A 30 year synthetic time series of weather variables was generated using the UK Climate Projections (UKCP09) Weather Generator tool (https://ukclimateprojections-ui.metoffice.gov.uk/ accessed on 23 March 2016). The tool simulated climatic conditions for a 25 km grid covering Sheffield (WXGen25km grid reference: 1275).

Synthetic hourly time series of rainfall and temperature were created for a 30 year period between 2010 and 2039. A daily time series of potential evapotranspiration using the Penman-Monteith method (PETFAO56) was also generated for the same timescale.

Over a 30 year period, as simulated using the UKCP09 Weather Generator tool, the mean annual rainfall depth was 774 mm – lower than the 30-year historical average of 834.6 mm (1981-2010) for Sheffield (Grid Ref. 53.383, -1.483). With 149 events per annum, the mean per-event rainfall depth was 5.2 mm. The mean rainfall depth during ‘significant events’ (i.e. return period greater than 1 year) was higher, at 41.2 mm.

Differences in the hydrological performance of the three configurations were minor but systematic – retention was highest with Meadow Flower and lowest from the non-vegetated configuration. Mean annual retention of the 774 mm of rainfall was between 47.2-51.4%. Mean per-event retention of significant events (mean rainfall of 41.2 mm) ranged between 17-22%.

The better retention performance of Meadow Flower was associated with a greater risk of drought stress. Irrigation to mitigate this risk reduced mean annual retention by 1.4% and per significant retention by 2.6%. However, retention performance was still higher for Meadow Flower with irrigation than for Sedum without it.

Through long-term simulations, designers can evaluate the performance of alternative green roof configurations and/or refine a configuration (e.g. to include irrigation or to increase soil depth). As such, a long-term simulation represents a valuable preliminary feasibility assessment tool for SuDS design practitioners.

Design Storm Events

Whilst mean annual retention and per-event retention statistics over a 30 year simulation provide a good indication of a green roof’s SuDS long-term potential, response of green roofs to design rainfall events are also important to design engineers.

Responses to design rainfall events are simulated here for TB2 in response to summer storm events with return periods of 1, 30 and 100 years (see Figure 5). The broader research also simulated responses to a number of alternative scenarios (i.e. configuration, season, initial moisture content, length of ADWP and rainfall characteristics).

Figure 5: Response of TB2 to summer events with return periods of 1, 30 & 100 years

Generally, retention and peak attenuation were inversely related to rainfall depth. Here, the response to each event was limited by the SMD at the start of rainfall. As such, when rainfall exceeded SMD, retention was typically equal to the SMD and runoff was the difference. As larger rainfall depths were considered, so the retention percentage reduced.

The response of TB2 to a 1 in 1 year design rainfall event of six hours duration was analysed. Rain depth was 14.3 mm. Peak rainfall was 0.26 mm/minute. Following a 7 day ADWP in summer conditions, the SMD of 19.7 mm was sufficient to wholly retain this event. Runoff would have been predicted if the ADWP had been less than 5 days.

Predictably, as higher rainfall depths were considered, the relative hydrological performance fell.

For a 1 in 30 year storm event of 6 hours duration, rainfall depth was 62.6 mm and peak rainfall was 1.14 mm/minute. Retention was 32.1% as 42.5 mm of runoff occurred with a peak intensity reaching 0.57 mm/min (i.e. peak attenuation was 49.8%).

The rainfall event with a 100 year return period produced a rain depth of 85.2 mm and a peak rainfall rate of 1.56 mm/minute. The greater rainfall depth predictably resulted in a lower per-event retention of 23.7% (20.2 mm). This compares against a maximum permissible retention (i.e. if SMD were equal to SMAX) of 38.1%. Peak runoff was reduced by 38.1% to 0.97 mm/min.

Conclusions

Green roofs can provide valuable additional drainage capacity to complement traditional drainage systems and other SuDS components. However, the hydrological response will vary. Retention depths can range between zero and the green roof’s maximum moisture storage capacity (SMAX) depending on:

1.      Green roof configuration;

2.      Rainfall characteristics; and

3.      Antecedent moisture conditions.

When expressed as a percentage, retention is inversely related to rainfall depth, with low probability, high volume events typically leading to low retention percentages. Predictably, the detention effect of extensive green roofs is limited by their shallow substrate depth. However, peak rates of runoff are often reduced as a result of retention and detention combined.

Seasonal climate has a significant impact on ET and retention. From a hydraulic perspective, the SuDS contribution of green roofs in winter is minimal, but is typically much greater in summer.

As such, designers tasked with mitigating year-round flood risk must combine extensive green roofs with additional downstream retention and detention measures (e.g. blue roofs) to form more holistic SuDS networks.

Papers

Po?, S., Stovin, V., Dunsiger, Z. (2011). The Impact of Green Roof Configuration on Hydrological Performance. Proceedings of the 12th International Conference on Urban Drainage. Porto Allegre, Brazil, 11-16 September.

Po?, S., Stovin, V. (2012). Advocating a physically-based hydrological model for green roofs: Evapotranspiration during the drying cycle, Proceedings of the World Green Roof Congress, 18-21 September 2012, Copenhagen, Denmark.

Stovin, V., Po?, S. and Berretta, C. (2013). A modelling study of long term green roof retention performance, Journal of Environmental Management, 131, 206-215.

Berretta, C., Po?, S., Stovin, V. (2014a). Moisture content behaviour in extensive green roofs during dry periods: The influence of vegetation and substrate characteristics, Journal of Hydrology, 511, 374-386. 

Berretta, C., Po?, S., Stovin, V. (2014b). The Influence of Substrate and Vegetation on Extensive Green Roof Hydrological Performance. Proceedings of the 13th International Conference on Urban Drainage. Sarawak, Malaysia, 7-12 September.

Po?, S., Stovin, V. and Berretta, C. (2015). Parameters influencing the regeneration of a green roof’s retention capacity via evapotranspiration, Journal of Hydrology, 523, 356-367.

Stovin, V., Po?, S., De-Ville, S. and Berretta, C. (2015a). The influence of substrate and vegetation configuration on green roof hydrological performance, Ecological Engineering, 85, 159–172.