Real Time Flood Forecasting for Don River Watershed

CHI developed a PCSWMM model for the Don River watershed in Ontario, Canada, for the primary purpose of predicting peak flows of a storm event at least two hours prior, so flood duty officers can make informed decisions. The model runs almost in real-time using rain gage calibrated RADAR rainfall. The Don River watershed has a drainage area of 360 km2 with 80% of the area is highly urbanized. Approximately 1.2 million people live within the watershed which also includes the largest metropolitan area in Canada, City of Toronto. The watershed had experienced several major flooding events in the past and will greatly benefit from a flood forecasting tool.

Use of RADAR data to generate spatially distributed rainfall within the watershed

RADAR acquisition and processing capabilities in PCSWMM allowed use of spatial variability in rainfall intensity and tracking speed of the storm with calibration using the rain gage records in the watershed. CHI used HEC-RAS transect import tools of PCSWMM to generate the hydraulic components for the model. For the efficiency of the model simulations, the stormwater management ponds in the watershed (there are 81 existing ponds and 41 more are proposed) were “lumped� together on a subcatchment basis. The G. Ross Lord Dam is a flood control facility on the West Don River which operates in conjunction with downstream flood control channels to reduce the risk of flooding to flood prone communities such as Hogg's Hollow. The facility was incorporated in to PCSWMM model with a storage unit representing the reservoir, one outlet representing two mud valves and four orifices representing the two low level gates and the two radial gate spillways. Operation of outlets was simulated with control rules in PCSWMM.

Due to the size of the Don Valley model the calibration effort was divided into submodels selected based on existing four flow gage locations. The Don Valley PCSWMM model, with the RADAR rainfall calibrated to existing rain gages in the area, was able to predict the ten highest peak flows at the flow gaging station near river mouth within a margin of error of ± 20%. Eight of these ten events were predicted to a 10% accuracy even with the coarse subcatchment discretization (average subcatchment was more than 500 ha). Control rules used for the operation of G. Ross Lord Dam were also able to predict the downstream flow accurately. Performance of the PCSWMM model for historical events provided sufficient confidence to Toronto and Region Conservation Authority officials to use the methodology in real-time flood forecasting and warning.

RADAR rainfall calibrated to existing rain gages consistently generated accurate simulations