A two-stage runoff detention model for a green roof

Urbanization has caused an increase in per-event stormwater runoff volumes. Existing combined \nsewer systems are becoming less able to take in storm runoff without overflowing, which may \ncause flooding and water quality issues. Sustainable drainage systems (SUDS) are structures and \npractices intended to reduce the volume and rate of a site’s runoff to pre-development levels. \nGreen roofs, not requiring exclusive land use, can be easily integrated into dense urban areas. \nHowever, their hydrological behaviour requires further understanding. \nA generic tool was created for routing detained rainwater through separately-modelled substrate \nand drainage layer components of a green roof. Components were monitored in isolation, in \npurpose-built rainfall simulators, under laboratory conditions. Configuration variables (e.g. roof \nslope) were varied and their effects on runoff response assessed. Nonlinear storage routing \nmethods were used to fit modelled to monitored runoff profiles, by optimizing routing \nparameters. The sensitivity of these parameters to test variables was assessed, greatly reducing \nthe number of individual values required for modelling either layer. \nThe runoff response of a two-layered green roof system at field capacity was tested under \nlaboratory conditions. The substrate model, in series with the drainage layer model, was \nparameterized for the two-layered system, and time-series runoff predictions and observations \nwere compared. The model produced consistently accurate results. This model was reparameterized \nfor three monitored test beds in Sheffield, UK, using estimated parameter values \nfor the three untested system configurations. The model was found to be fit for purpose, \napproaching laboratory accuracy in the best cases. Peak flow predictions were improved by \nallowing limited runoff to occur before a roof’s water content completely reached field capacity. \nFurther work should extend the model’s applicability to long time-series, through improved \nevapotranspiration modelling. Further laboratory observations of individual roof components are \ndesirable, to increase the range of modellable green roof configurations.