Assessment of finely discretized urban drainage models derived from WorldView-2 satellite imagery

Moh Moh Lin Khin, Darko Joksimovic and Ahmed Shaker, Ryerson University, Toronto, ON, Canada


Development of urban drainage models has been facilitated over the years by increased availability of GIS datasets and tools that utilize the digital data to rapidly develop models of higher complexity.  In urban areas where such data is not readily available, it can be derived from other sources, including satellite remote sensing images. Distributed models with fine discretization of drainage areas are frequently used to evaluate the retrofit potential of Low Impact Development (LID) techniques without the ability to calibrate them due to the absence of monitored data.  Since the modelling approaches for LIDs are numerous and still very much in development, a variety of approaches are used in practice and further guidance is needed, particularly in the absence of appropriate digital datasets.

The objectives of the research presented in this paper were to: 1) investigate the use of remote sensing images in development of detailed, distributed urban drainage models and 2) assess different modelling approaches for conventional (curb and gutter, roadside ditches) and LID (bioretention, porous pavement) drainage systems.  High resolution WorldView-2 satellite image of a residential study area that features these drainage systems was processed to develop finely discretized SWMM models, in addition to lumped and discretized models built using the GIS datasets available from the municipality.  A performance assessment of all developed models, which were not calibrated using the observed data, involved comparing the model outputs to the observed flow timeseries.  The results indicate the ability of models derived using different data sources and of different complexity to capture the catchment runoff, providing some guidance to practitioners. 

The modelling results were mixed, with the lumped model capturing the runoff volume closely, but not performing well with respect to peak flows.  The finely discretized models, derived using both data sources, performed well for the curb and gutter system.  They overestimated the runoff from smaller events for the ditched road, the cause of which is suspected to be the poor condition of culverts leading to increased ponding and infiltration.  All models overestimated the runoff from the street in which LIDs were implemented, although the observed runoff was very small during the monitored period.

Permanent link: