Challenges in field approximations of regional scale hydrology

S.S. Wanniarachchi, N.T.S. Wijesekera
2020 Journal of Hydrology: Regional Studies  
A B S T R A C T Study region: Considering the availability of gauged data, the Karasnagala watershed of Attanagalu Oya located in the Gampaha district in the Western province of Sri Lanka was modeled with EPA SWMM 5. Study focus: This study analyses the effect of the catchment field approximations for accurate flood hydrograph prediction. Following an event based approach, 3 days Minimum Inter event Time (MIT) and 0 mm/day Minimum inter Event Depth (MED) were used as the threshold. Fifty events
more » ... were separated from 1971 to 1982 period. Four major field approximation types were identified: stream geometrical parameters approximations, soil infiltration parameter approximations, approximation of watershed intermittent storages, and subcatchment delineation approximation. Soil parameter approximations and the stream network geometry parameter approximations were verified by the field observations. New hydrological insights for the region: Model calibration and verification revealed that EPA SWMM5 can be successfully used to develop regional Karasnagala watershed model with mean ratio of absolute error (MRAE) 0.289 for calibration, and 0.375 MRAE for verification. Incorporation of intermittent storages with optimized model layout obtained the best fitting of hydrograph recession MRAE 0.167. Subcatchment lumping with a 16 sub basin configuration showed the marginal increment of modeling error when compared with distributed modeling. Stream parameter approximations revealed that the head water streams/lesser order streams parameters sensitivity is higher than that of the higher order streams. In soil parameter approximations, saturated hydraulic conductivity of the soil was the most influencing parameter. Most hydrological simulations are capable of identifying the time to peak and the peak flow, but less good in the recession. Mandeville (2016), emphasises the key role played by the slow flow component of storm runoff, which is obviously the recession that has not been given appropriate attention in the existing event-based studies in literature. However, the flood peak and the flood residence time depending on the recession behaviour are important for flood plain management. According to McIntyre et al. (2012) , the impact of land use change on the shape of a hydrograph depends strongly on the properties of the channel network and hence an accurate representation of channel routing processes is critical to the assessment. For the convenience of modeling, most models approximate complex sub stream networks into simple networks. The shape, roughness, and side slopes of the modeled stream segments and the model layout are changed to represent the routing effect from the intermittent storages (Rouhani et al., 2009) . Most hydrological models are characterized by a set of parameters, which are uncertain, and not directly measurable, thus obtaining an optimal model becomes a complex task (Zhan et al., 2013) . When satisfying the conditions such as time efficiency and the cost efficiency in most of engineering applications, approximated values are used for both field measured and derived parameters. Generally, models solving the shallow water wave equations, are used to design storm water infrastructure (Zoppou, 2001) . The cross-sectional information, roughness coefficients, boundary conditions, and the details of internal structures are the information required to solve shallow water wave equations. For some catchments, these information are not available. Generally, field parameter approximations are required in application of shallow water wave equations for the models (Zoppou, 2001) . The effects of these approximations on the model output are rarely discussed in literature. Therefore, the objectives of this study strive to analyse the applicability of an urban stormwater model (EPA SWMM 5) for a regional watershed; to analyse the effect of catchment field approximations for accurate flood hydrograph prediction; and to identify the deficiencies and possible solutions for the regional watershed modeling. Storm water management model (EPA SWMM 5) SWMM is a fully dynamic, deterministic, distributed, conceptual watershed simulation model for single-event or long-term simulation of runoff quantity and quality, primarily from urban areas (Gülbaz and Kazezyılmaz-Alhan, 2017). It has the capability to include urban stormwater controlling elements such as pumps, orifices, manholes, weirs, flow dividers, and conduits. Version 5 running with Microsoft Windows provides an integrated environment for editing data, running hydrologic, hydraulic and water quality simulations, and viewing of the model results (Gülbaz et al., 2019; Rossman, 2009 ). It conceptualizes a drainage system as a series of water and material flows between four major environmental compartments. 1 Atmosphere compartment, from which precipitation falls and generates surface and subsurface hydrographs based upon antecedent subsurface moisture conditions, soil characteristics, detention storages, land use characteristics, drainage area and topography, is known as physiographic characteristics. 2 Land surface compartment, where the pollutants are deposited is represented by subcatchment objects. 3 Groundwater compartment, which receives infiltration from the land surface compartment is based on either Green-Ampt or Horton infiltration models and transfers a portion of this inflow to the transport compartment. 4 Transport compartment, contains a network of conveyance, storage, regulation and treatment elements. All compartments do not need to appear in a particular SWMM model (Rossman, 2009) . SWMM runs according to any given time frame. There are no intrinsic limitations regulating the duration of a "continuous" simulation in the model except the time required by the computer to perform a large number of calculations. Compromising the time steps during wet and dry periods assists minimizing computing time without significant loss of information in the model output. When rainfall input is introduced to the model, it begins a time step by the time step accounting of water movement and losses by doing various processes as runoff, infiltration, evaporation, and surface storage. In this situation, each subcatchment surface is treated as nonlinear reservoir. The capacity of the reservoir denotes the maximum depression storage consisting surface ponding, surface wetting, and interception. Surface runoff occurs when the depth of water in the reservoir exceeds the maximum depression storage
doi:10.1016/j.ejrh.2019.100647 fatcat:il7d7ehfv5ajniynlu3ptutmwi