A four-year record of rainfall and runoff data from nine different extensive (80mm substrate) green roof test beds has been analysed to establish the extent to which the substrate composition and vegetation treatment affect hydrological performance. The test beds incorporated three different substrate components with different porosity and moisture retention characteristics, and three different vegetation treatments (Sedum, Meadow Flower and unvegetated).Consistent differences were observed, with the vegetated beds showing higher levels of rainfall retention and better detention compared with unvegetated beds. The seasonal Meadow Flower beds had similar hydrological performance to Sedum-vegetated beds. There was a 27% performance reduction in annual volumetric retention attributable to differences in substrate and vegetation. The beds with the most porous/permeable substrates showed the lowest levels of both retention and detention.As with previous studies, retention efficiency in all nine beds showed a strong dependency on rainfall depth (P), with retention typically >80% for events where P10mm. The effects of vegetation and substrate were most evident for rainfall events where P>10mm, with the mean per-event retention varying between beds from 26.8% to 61.8%. On average, the test beds were able to retain the first 5mm of rainfall in 65% of events where P>5mm, although this ranged from 29.4% to 70.6% of events depending on configuration. In terms of detention, all but one of the test beds could achieve runoff control to a green field runoff equivalent of 2l/s/ha for more than 75% of events.Detention was also characterised via the calibration of a reservoir-routing model that linked net rainfall to the measured runoff response. The parameter values identified here – when combined with a suitable evapotranspiration/retention model – provide a generic mechanism for predicting the runoff response to a time-series or design rainfall for any unmonitored system with comparable components, permitting comparison against local regulatory requirements

Although it is widely accepted that the detention performance of green roofs is of interest to stormwater engineers and planners, no single metric allows detention to be unambiguously defined. Detention effects are highly sensitive to rainfall characteristics and antecedent conditions, and individual roofs typically exhibit wide variations in detention performance between storm events. This paper uses a straightforward hydrological model to explore two alternative approaches to describing detention performance: a probabilistic approach based on long time-series simulations; and a design storm approach. It is argued that the non-linear reservoir routing parameters (scale, k and exponent, n) provide fundamental descriptors of the detention process, with modelling enabling performance to be determined for specific rainfall inputs. The study utilises 30-year rainfall time-series predictions for four contrasting UK locations to demonstrate the utility of the two proposed design approaches and to comment on locational variations in detention performance.

Green roofs are increasingly popular measures to permanently reduce or delay storm-water runoff. The main objective of the study was to examine the potential of using machine learning (ML) to simulate runoff from green roofs to estimate their hydrological performance. Four machine learning methods, artificial neural network (ANN), M5 model tree, long short-term memory (LSTM) and k nearest neighbour (kNN), were applied to simulate storm-water runoff from 16 extensive green roofs located in four Norwegian cities across different climatic zones. The potential of these ML methods for estimating green roof retention was assessed by comparing their simulations with a proven conceptual retention model. Furthermore, the transferability of ML models between the different green roofs in the study was tested to investigate the potential of using ML models as a tool for planning and design purposes. The ML models yielded low volumetric errors that were comparable with the conceptual retention models, which indicates good performance in estimating annual retention. The ML models yielded satisfactory modelling results (NSE >0.5) in most of the roofs, which indicates an ability to estimate green roof detention. The variations in ML models' performance between the cities was larger than between the different configurations, which was attributed to the different climatic characteristics between the four cities. Transferred ML models between cities with similar rainfall events characteristics (Bergen–Sandnes, Trondheim–Oslo) could yield satisfactory modelling performance (Nash–Sutcliffe efficiency NSE >0.5 and percentage bias |PBIAS| <25 %) in most cases. However, we recommend the use of the conceptual retention model over the transferred ML models, to estimate the retention of new green roofs, as it gives more accurate volume estimates. Follow-up studies are needed to explore the potential of ML models in estimating detention from higher temporal resolution datasets.

Evapotranspiration is a key hydrological process for reducing stormwater runoff in bioretention systems, regardless of their physical configuration. Understanding the volumes of stormwater that can be returned to the atmosphere via evapotranspiration is, therefore, a key consideration in the design of any bioretention system. This study establishes the evapotranspiration dynamics of three common, structurally different, bioretention vegetation treatments (an Amenity Grass mix, and mono-cultures of Deschampsia cespitosa and Iris sibirica) compared with an un-vegetated control using lab-scale column experiments. Via continuous mass and moisture loss data, observed evapotranspiration rates were compared with those predicted by the FAO-56 Penman-Monteith model for five 14-day dry periods during Spring 2021, Summer 2021, and Spring 2022. Soil moisture reductions over the 14-day trials led to reduced rates of evapotranspiration. This necessitated the use of a soil moisture extraction function alongside a crop coefficient to represent actual evapotranspiration from FAO-56 Penman-Monteith reference evapotranspiration estimates. Crop coefficients (K_c) varied between 0.65 and 2.91, with a value of 1.0 identified as a recommended default value in the absence of treatment-specific empirical data. A continuous hydrological model with K_c=1.0 and a loading ratio of 10:1 showed that evapotranspiration could account for between 1 and 12% of the annual water budget for a bioretention system located in the UK and Ireland, increasing to a maximum of 35% when using the highest K_c observed (2.91).