Need to improve SWMM's subsurface flow routing algorithm for green infrastructure modeling

Joong Gwang Lee, William Shuster and Sudhir Kshirsagar


SWMM can simulate various subsurface flows, including groundwater (GW) release from a subcatchment to a node, percolation out of storage units and low impact development (LID) controls, and rainfall derived inflow and infiltration (RDII) at a node. Originally, the subsurface flow of infiltrated water was considered an insignificant process in SWMM, and particularly for simulations of combined sewer overflows in urban catchments. However, subsurface flow can be significant where the drainage area has high proportional area in pervious surfaces, high groundwater tables, or a combination of these and other factors that facilitate infiltration, percolation, and subsurface runoff. In this case, considerable amount of infiltrated stormwater could be slowly released to the receiving water or wastewater conveyance infrastructure through subsurface flow processes. Implementing infiltration LID or green infrastructure (GI) practices would also increase volume inputs to the local water cycle, with enhancement of subsurface flows.

The present GW module for SWMM represents these flow processes, though from a highly empirical standpoint. In this conception, GW flow is modeled from each subcatchment based on the two-zone configuration: an unsaturated upper zone and a saturated lower zone. The main driving force of the GW flow is the difference in the hydraulic heads between the top elevation of the saturated zone and the elevation of the channel bottom (i.e., the invert elevation of a node). However, the algorithms/parameters of the GW routing are lumped, non-specific, and thereby largely implicit. Percolation out of storage units and LID controls is simulated using either a fixed rate or the Green-Ampt method. Each method is applied to simulate only 1-dimensional water flow in the vertical direction. One immediate process gap is that lateral exfiltration through vertical walls from storage or LID units cannot be modeled by this approach. This is despite the fact that lateral exfiltration can be considerable in many cases where the storage or LID unit is designed to be narrow and deep, or the bottom of the unit is prone to be clogged by fine sediments. RDII is the increased portion of water flow in a sewer system that occurs during and after a rainfall. The amount of RDII could be significant where the sewer system is aged, damaged, or otherwise prone to exchange into and out of the pipe via leaks. SWMM simulates RDII using the RTK method, which is based on three synthetic triangular unit hydrographs to estimate the fast, medium, and slow RDII responses. Observed flow data must be applied to estimate the RTK parameters by fitting the modeled and the observed hydrographs. The RTK parameters should be changed with any retrofitting the sewer system or implementing onsite GI/LID practices. Because of this, the RDII simulation by the estimated RTK parameters from the existing condition cannot be used for GI/LID modeling, which is based on the future condition. On the other hand, both GW and RDII are modeled at a node (i.e., a point outlet) in SWMM, but physically routed to the channelized receiving water system (i.e., a linear outlet).

This study is aiming to identify the need of improving the existing algorithms of subsurface flow routing in SWMM, in particular for GI modeling. Given these gaps in SWMM GW, exfiltration, and RDII process representations, we propose an innovative approach that can make use of field hydrologic data, rather than relying on empirical approaches. Potential directions for these improvements are examined using Darcy’s equation: Q = K A dH/dL, where Q = rate of water flow [L3T-1], K = hydraulic conductivity [LT-1], A = cross-sectional area in flow [L2], dH/dL = hydraulic gradient [dimensionless], H = hydraulic head [L], and L = flow length between the points of interest [L]. An appropriate site-specific value for K can be estimated using field data. Suitable approximation must be applied in estimating a dynamic cross-sectional area, A (i.e., the variations of A are dependent on time and space) as a function of H and the size of drainage area. The appropriate value of L from a drainage area to a channel system can be estimated using geographic information system (GIS). Geometric approximation would also need to be applied when we estimate A, H, and L between the points of interest. Some alternatives are examined using the custom lateral or groundwater flow equation editor in SWMM.


Permanent link: