Effect of aggregation of on-site storm-water control devices in an urban catchment model

Abstract

Spatially distributed on-site devices such as detention tanks and bioretention are becoming more common as a means of controlling urban storm-water quantity and quality. One approach to modeling the cumulative catchment-scale effects of such devices is to resolve the catchment down to the scale of a land parcel or finer, and then to model each device separately. This involves computational and input data demands that may be impracticable, especially in planning or preliminary design stages of storm-water system design. To reduce these demands, the spatial resolution can be coarsened by aggregating land parcels and devices, but this may compromise model accuracy. The focus of this study was examination of the effects of aggregation on predictions of water quantity and quality (for a representative contaminant, total suspended solids) for detention, infiltration, and bioretention devices. A detailed model for urban storm water improvement conceptualization simulation was set up for a 0.83 km2 catchment with 810 source areas and associated devices, and the model was then reduced to three aggregation levels (55 devices, seven devices, and one device). The influence of aggregation was assessed by comparing the predictions of the aggregated models against the predictions of the detailed model. Aggregation had little effect on the predictions of maximum concentration (<2% difference compared with the detailed model), load (<4%), and baseflow (<5%) when the devices were sized in proportion to the impervious area and when there was high soil permeability. Aggregation to a single device increased peak flow compared with the detailed model, by up to 38.1% for bioretention and less for other devices. The peak flow increase was a consequence of reducing the range of travel times in the aggregated drainage network. Aggregation to seven devices had considerably less effect on peak flow (up to 8.7% increase). Addition of variability to the size of the devices introduced further aggregation effects. Methods to extend the aggregation approach to cater for variability in device sizing are proposed in the paper. The results of the study suggest that aggregation can be used to reduce computational and input data demands, with little penalty in terms of prediction accuracy.