Introduction
Our overall research aim is to develop accurate and useful hydrological modelling tools to predict the runoff response from a green roof for any arbitrary rainfall time series. Our modelling work has been supported by the development of several experimental facilities, which are described below.

In 2006 we established our first instrumented test plot, located on the Mappin Building roof, University of Sheffield. This test bed has yielded high quality continuous record of rainfall and runoff and forms the basis of most of our published outcomes to date (Hartini Kasmin, PhD student) [see Stovin et al. (2012)]. This test bed has subsequently (2009) been duplicated 10 times on the roof of the Hadfield Building. The Hadfield test beds incorporate moisture content measurements. These beds underpin a comparative study of how vegetation and substrate affect runoff retention and attenuation (Simon Poe, PhD student), and an investigation of how substrate ageing impacts upon hydrological performance (Simon De-Ville, PhD student).

In addition to the external test beds, we have also established/utilised a wide range of other experimental facilities, including:
  • A 5 x 1 m rainfall simulator located in ZinCo's laboratory in Stuttgart, used extensively by Gianni Vesuviano to quantify detention effects of green roof drainage board components [see Vesuviano and Stovin (2013) and Vesuviano et al. (2014)].
  • A custom-built mini rainfall simulator at the University of Sheffield, used to quantify detention effects of green roof substrate components [see Yio et al., (2013)].
  • Controlled climate tests facilities at the University of Sheffield, used to quantify evapotranspiration rates from planted green roof microcosms.
Proper understanding of the dependency of green roof evapotranspiration on substrate moisture content (via a SMEF (or Soil Moisture Extraction Function)) underpins the development of our continuous simulation rainfall-runoff hydrological model (Stovin, Poe and Berretta (2013) and Berretta et al. (2014)).


Mappin Building Test Bed Set-up
The test bed (3 x 1 m) (Figure 1) uses a standard commercial (Alumasc/Zinco) extensive green roof system, comprising a sedum vegetation layer growing in 80 mm of substrate. The base of the rig is laid at a slope of 1.5 °. The substrate is composed of a mixture of crushed brick and fines. A fine particle filter membrane separates the substrate from the underlying FloraDrain FD25 ‘egg box’ drainage layer. The drainage layer alone has a nominal retention capacity of 3 l/m 2 (i.e. 3 mm rainfall).

Runoff from the roof is collected in a tank via a gutter at its downstream end. The collection tank is tapered to provide greatest sensitivity for low levels of runoff. Depth in the collection tank is monitored using a pressure transducer to provide a continuous record of runoff. Rainfall is monitored using a standard rain gauge sited adjacent to the test bed. Data from the pressure transducer and the rain gauge is logged using a Campbell Scientific data logger (CR1000) at 1-min interval.

Figure 1 Mapping Building Hydrological Monitoring Test Bed

Selected Observations
Preliminary data on the test bed’s hydrological performance was presented by Stovin et al. (2007). Figure 2 presents sample monitored data from a storm that occurred on 14–15 February 2006. In this event 9.2 mm rain fell, resulting in only 3.55 mm runoff, i.e. 61% stormwater volume retention. The peak reduction was 61%. Over the whole of the (wet English) spring 2006, 11 events were monitored, and the average volume retention was found to be 34%, with an average peak reduction of 57% compared with the monitored rainfall. These figures are comparable with other European data reviewed by Mentens et al. (2003). Based on an average volume retention of 34%, it may be inferred that the roof could reduce annual runoff in many parts of the United Kingdom by 300 mm when compared with a conventional roof. 

Figure 2 Rainfall-Runoff response of the green roof test plot 14-15 February 2006

Data from summer 2007
Figure 3 shows the daily rainfall and runoff totals for the Sheffield test bed during June and July 2007. This period includes the period 24–25 June, when Sheffield was affected by serious flooding, largely as a result of the River Don bursting its banks in the city centre. One contributory factor in the flooding event was the heavy rainfall that occurred in the previous fortnight, leaving the ground saturated, and with little capacity to absorb further rainfall. The local records presented in Figure 3 show that nearly twice as much rain fell during 13-15 June (115.8 mm) compared with 24-25 June (61.8 mm).

The green roof temporal runoff response was as might be expected. The event on 11 June occurred after a long dry spell and all of the 12.8 mm rainfall was retained within the roof. At the start of the 13-15 June event some retention was observed, with 16.1 of 24.8mm rainfall retained (65%) on 13 June. The following day the roof was evidently saturated, with only 5% retention (of 74.4 mm) being observed. On 15 June, slightly negative retention was observed as some of the accumulated rainfall drained out of the roof. The roof had limited opportunity to regain its moisture-holding capacity over the following days, and only 27% of the 24.8 mm rainfall on 24 June was retained. The following day the roof was unable to offer any retention, with virtually 100% of the 37 mm rainfall becoming runoff.

Overall retention in June still amounted to 33%, whereas the less extreme rainfall conditions in July enabled the roof to achieve 45% retention. In May and August 2007 the roof retention was 79 and 100%, based on 80.2 and 23.6 mm rainfall, respectively.

The performance data presented above suggests that green roofs alone cannot provide complete protection against very large (extreme) events. However, this does not diminish their value as source control elements within more comprehensive SUDS treatment train approaches. The data provided here does demonstrate that they can play a significant role in reducing the total volume of runoff, with potential benefits to runoff quality. The CIRIA SUDS Manual (C697, 2007) highlights the importance of using source controls that mimic natural interception storage to retain small events (< 5 mm) to reduce the flashiness of urban runoff and the pollutant loads conveyed to urban watercourses.

Figure 3 Rainfall and Runoff totals for June and July 2007