One of the primary ecosystem services associated with greenroofs is stormwater retention. Unfortunately, studies of the cumulative effects of greenroof implementation at the watershed scale have only been published for a few cities globally. Our study asks 1) How can the potential for greenroof retrofit be quantified, and how might this potential vary across an urban landscape? and 2) How might these greenroof retrofits affect stormwater retention at the watershed scale? The total roof area suitable for extensive greenroof installation (greenroof potential) was evaluated for the entire Gwynns Falls in Baltimore Maryland (MD) using GIS building and lidar data. Roofs were selected primarily based on their size and slope; structural concerns were not considered in this study though recognized as important. More detailed analyses were conducted for areas of two subwatersheds: the Glyndon and Dead Run subwatersheds, which have been characterized as suburban and densely urban respectively. Estimates of greenroof potential for each area were then used to simulate the effect of greenroof addition on stormwater retention using a stormwater management model (SWMM) model. This model was developed by the US Environmental Protection Agency and adapted for the Gwynns Falls watershed by the Baltimore County’s Department of Environmental Protection and Resource Management. Greenroofs were represented as storage facilities in the model whereby a uniform maximum storage depth of 3.175 cm was assumed.
Results/Conclusions
Preliminary findings suggest that 35.4 km2 of roof area may be available for greening in the entire Gwynns Falls watershed. For the study area in the Glyndon subwatershed, greenroof potential encompassed 38,450 m2 representing 5.35% of the total study area. In contrast, 136,559 m2 of the Dead Run subwatershed study area was found suitable for greening, or 11.33% of the total study area. This variation in greenroof potential did not translate to large differences in stormwater retention. A SWMM model of greenroof addition to both subwatershed areas predicted reductions in total outflow of 10,167 m3 and 11,329 m3 for the Glyndon and Dead Run study areas respectively for storms occurring between mid October and November 2007. Stormwater retention also varied with storm frequency in our models for both subwatersheds, and this variation may be attributed to our treatment of evapotranspiration. The influence of uncertainty about greenroof evapotranspiration rates and other aspects of the project on model predictions as well as opportunities for future research will be discussed.
