Modeling stormwater runoff from an urban park, Singapore

The environmental and societal benefits of urban green space has long been recognized, but such land use often becomes a secondary consideration in urban drainage modeling, in part because it is more difficult to obtain site-specific data for model calibration. Singapore recently has re-branded itself from “garden city” to “city in a garden” and as such parks, urban forests, and even agricultural areas are particularly important in sustaining the liveability of this highly urbanized city-state. The objective of this study was to quantify the rainfall-runoff processes for a park in Singapore and explore the efficacy of PCSWMM in estimating runoff from a nearly 100% pervious area. Admiralty Park consists of two sections, one being a traditional urban grassed area with trees, outdoor exercise areas, and promenades; and the other being a forested nature trail that includes a mangrove habitat opening to the Straits of Johor. Infiltration rates, using a double o-ring infiltrometer, were measured at 5 sites within the traditional grassed area and samples for textural analysis were collected at the same sites. This part of the park is serviced by a tile drain and an ISCO 2150 area-velocity meter was installed near the drain outlet, together with a tipping bucket rain gauge, to monitor rainfall and runoff between 21 December 2013 and 4 May 2014.

The soils were 80-92% sand and classified mainly as sand or sandy loam. Based on the infiltration measurements and fitting the Horton infiltration equation to the data, the initial infiltration capacity (fc) of the soils in the park ranged between 80 and 690 mm/hr. A total of 12 storms were deemed to have a complete data record, with rainfall depths ranging between 9.6 and 99.4 mm and peak intensities between 42 and 144 mm/hr. There was a strong correlation (0.872) between total storm event runoff volume and total rainfall depth, but weaker correlations between peak rainfall intensity and peak runoff rate (0.234) or peak rainfall intensity and total storm event runoff volume (0.382). In particular, two storms measured shortly after the 1 in 140 year drought exhibited unusually high peak runoff rates.

PCSWMM was calibrated for 10 of the 12 events and validated for 2 events. The Horton infiltration equation was used to operationalize PCSWMM and it was found that the unit hydrograph I/I approach also was required to help match the extended falling limb of the observed hydrographs. For the Horton infiltration equation, the calibrated values of fc ranged between 100 and 130 mm/hr, while the values for fo, and α were the same for all events at 20 mm/hr and 10, respectively. The unit hydrograph values for T and K in the I/I calculations were the same for all events at 0.21 and 1.2, respectively, while the value of R (proportion of the rainfall that is translated to I/I) ranged from 1% to 75% (in one case), but typically was around 40%. The calibration efforts prioritized matching peak over total event volume and for the 10 events the Nash-Sutcliffe statistic was an excellent 0.982 for peak flow and 0.728 for event volume. Runoff estimates for the 2 validation events also matched measured flow quite well. PCSWMM appears to be well capable of modeling runoff from urban parks.