The example dataset provided is for demonstration purposes only and should not be used for other purposes. Some of the data within the dataset may have been altered specifically for the purpose of testing and demonstration and do not reflect actual conditions at the location. The example model has been simplified where possible for storage size and computation runtime purposes, and is not reflective of detailed study models. 

Introduction

This is a collection of example datasets that serve as a comparison of results between EPA SWMM and HEC-RAS. EPA-SWMM's Extran test suite is a collection of EPA-SWMM datasets that are used in the EPA's quality assurance testing for their dynamic wave routing model. The EPA-SWMM Extran tests are described in the Quality Assurance Report for Dynamic Wave Routing on the EPA SWMM website and the datasets are also available for download there. In addition, brief descriptions of the datasets and and a SWMM *.inp file for them can be found on OpenSWMM. Corresponding HEC-RAS models were built for each test and simulation results were compared between the two models. All HEC-RAS and EPA-SWMM models used in this comparison can be found in the zip file below.

Download: SWMM_Extran_Examples.zip

Geometry Setup

The corresponding HEC-RAS datasets were created by importing the SWMM Extran test models into HEC-RAS via the SWMM Import Tool (described here). Since the EPA-SWMM models are not georeferenced, and HEC-RAS relies on the georeferenced lengths of conduits in the computations, the XY point data for the nodes were manipulated to ensure the conduit lengths were correct in the coordinate reference system. After importing the pipe network geometry, a terrain was created using terrain modifications to match the SWMM Max Depth attributes for each node, and a downstream channel was burned in to model the open channel where the pipe network discharges into a 2D area.
An additional small length of conduit and another node was added to the HEC-RAS geometry at the outfall of the pipe network. This was done because the node where the two pipe branches come together is a Junction type node and cannot also serve as an outfall to an open channel. The extra conduit and additional node (Culvert Opening type) allows the pipe network to outfall into the receiving 2D channel. This was preferred over having the conduits outfall individually to the 2D channel for comparison purposes.

Surface flooding is handled differently between HEC-RAS and EPA-SWMM. In HEC-RAS a surface model (1D or 2D geometry) must be connected to the pipe network to allow water to surcharge out of a pipe network node. EPA-SWMM allows water to leave the model domain all together when nodes are surcharging. Since some of the test datasets surcharge, simple Top Inlets and 2D Areas were added to allow water to leave the pipe network and flow onto the surface as shown below.

One major difference between HEC-RAS and EPA-SWMM computations are how minor losses are handled. While EPA-SWMM only computes entry and exit losses, HEC-RAS computes minor losses due to angled flows at junctions, plunges and pipe bends as well, following FHWA's HEC-22 guidance. In order to make model results more comparable, entry and exit losses were set to 0 in both datasets which zeros out all types of minor losses at junctions in HEC-RAS.

HEC-RAS allows the volume of each junction to be accounted for in computations by setting the Base Areas on nodes which adds volume to the node cell's elevation-volume property table. To make results comparable to EPA-SWMM, the nodes were given 0 Base Area so that no additional storage volume was added to the system. 

Boundary Conditions and Observed Data

The boundary conditions for the HEC-RAS simulation were extracted from the SWWM model and added to the appropriate nodes as Flow Hydrographs. In addition, the SWMM model results at nodes were brought into HEC-RAS as observed data for easy results comparison using the Import EPA SWMM Output as Observed Data... tool (described here) in the Unsteady Flow Data Editor.

Computation Settings

For most datasets, the pipe network computation setting in HEC-RAS were kept at the defaults: running the diffusion wave solver with adaptive timestep targeting a Courant of 0.9. For some of the datasets with weir and orifice structures the computation settings were modified to better account for rapidly varying flow conditions by using the full shallow water equations and a lower target courant. 

SWMM Extran1 Test - Base Case

The profile plot below shows the results for the upper pipe run of the Extran 1 test, comparing EPA-SWMM (in pink) stages to HEC-RAS stages (in blue) at the peak of the event. Results are very comparable at the peak of the event when most of the system is in pressure flow. HEC-RAS stages are slightly higher than EPA-SWMM results primarily due to how surcharging flows are handled in each model. 

Stage and flow results at each node were also found to be very comparable to the EPA-SWMM results for the Extran 1 test. The pipe network outlet, Node 16009, shows good agreement for both stages and flow between the two models. HEC-RAS has approximately 2 cfs higher flow out at the peak of the event which is due to the difference in surcharging methods discussed below. The pipe network outlet, Node 16009, results are shown below.Flows for the terminal conduits on the upper branch (Conduit 1600) and lower branch (Conduit 1570) are in agreement between both models. Note that because HEC-RAS conduits are discretized with cells, flows and velocities can be found at any face in the conduit. These plots are showing the results for the most upstream and most downstream faces.  Conduit 1600 shows the slight difference in peak flows due to surcharging at Node 80608, as expected. Conduit 1570 of the lower branch shows  matching peak flows between the two models. The HEC-RAS results show some backflow early in the simulations as flow from the upper branch arrives and the junction and backwaters up the conduit of the lower branch.

One difference found in the model results are the stages and flows around node 80608 which is surcharging flow out of the pipe network in both models. In HEC-RAS the surcharging flow is computed using the dimensions and coefficients of the Top and Side Inlets connected to the node. This surcharging requires the stages to be above the top inlet elevation in HEC-RAS, in contrast to EPA-SWMM where the pipe network stage is pinned at the Max Depth for the node. 

This difference in surcharging method results in a slightly higher stage in HEC-RAS (~0.4 ft) at that node, and less surcharge flow (~2 cfs) out of the pipe network.An animation of the results profile in HEC-RAS shows stages, velocities and flows along the profile for the upper branch of the pipe network. 

Extran 2 Test - Higher Tailwater

SWMM Extran 2 test is identical in geometry to Extran 1, but changes the downstream boundary condition of the open channel from a free outfall to a higher set stage of 94.4 ft. The backwater from the higher downstream change does not impact the outfall of the pipe network, so the results are identical to Extran 1.

Extran 3 Test - Bottom Orifice 

The Extran 3 test utilizes bottom orifice at the 80608 node. Though HEC-RAS uses a bottom orifice computations for Top Inlets connected to the surface, only side orifices can be represented inside the pipe network.

Extran 4 Test - Weir

The Extran 4 test adds a weir to the pipe network, connecting the upper branch to the lower branch between nodes 82309 and 15009. This weir diverts flow from the upper branch to the lower branch, reducing stages and eliminating the flooding occurring. In HEC-RAS weirs are not their own geometry element - instead a conduit can be designated as a weir. To add the weir in the HEC-RAS geometry a new conduit was drawn to between the two nodes and the conduit dimensions were set to the size of the weir opening. US and DS Offsets were set to raise the conduit to the crest height, then the conduit was set to a Structure Type of Weir. Finally, the minor loss coefficients were adjusted to achieve an effective weir coefficient of  3.0. See the Hydraulic Reference Manual for more detail on pipe network structure coefficients. 

Replicating this EPA-SWMM dataset in HEC-RAS is a little awkward because pipe networks in HEC-RAS are purely geospatial while EPA-SWMM is not and allows zero-length features. In addition, the weir type is set to Transverse in the EPA-SWMM model, but it behaves more like a lateral weir as flow over the weir is actually perpendicular to the flow direction of the upper branch of the network. Nevertheless, the diffusion wave solver, which removes the acceleration terms, compared well to the EPA-SWMM results. 

Stages at node 82309 (the weir headwater) and flow through the weir are compared in the plots below for HEC-RAS and EPA-SWMM. HCE-RAS computes a slightly lower headwater stage at the weir and computes about 10% less flow over the weir when compared to EPA-SWMM. Comparing the HEC-RAS weir rating curve for this simulation to the standard weir flow equation with C=3.0 shows that the HEC-RAS still matches the weir equation well, but deviates slightly on the upper end.

Extran 5 Test - Side Orifice

The Extran 5 test starts with the Extran 1 base geometry and then adds storage to node 82309 and connects an orifice just downstream of the node on the upper branch. This change adds more storage to the upper pipe branch but the orifice chokes the flow enough such that more surcharge flooding occurs at the nodes on the upper branch. To setup this geometry in HEC-RAS, node 82309 was given a base area equivalent to that of the storage node in the SWMM model, and conduit 1602 was split by inserting a new node, 82308. The new conduit that was created was set to the weir/orifice Control Structure, set to 2 ft circular shape, and given US and DS offsets of 0.5 ft to meet the requirements for control structures Finally, the entrance and exit losses coefficients were set to achieve an orifice coefficient of 0.85. See the Hydraulic Reference Manual for more detail on pipe network structure coefficients. 

A plan view of the new upper branch geometry is shown below depicting the new base area at 82309 and the new orifice conduit. Stages at node 82309 (the orifice headwater) and flow through the orifice are compared in the plots below for HEC-RAS and EPA-SWMM. HEC-RAS maintains about 0.5 ft higher stage at the orifice opening which is driving about 5 cfs (8%) more flow through the orifice. This difference in stage and orifice flow once again can be attributed to the difference in surcharge calculations between HEC-RAS and EPA-SWMM; HEC-RAS computes flow through the top inlet with the HGL while EPA-SWMM surcharges any volume beyond the rim elevation. The result of this is EPA-SWMM loses 4.6 ac-ft (~14% of inflow) to flooding at node 80608 while HEC-RAS loses~3.6 ac-ft, and HEC-RAS delivers ~5% more volume to the outfall.

The plots below show the stage and surcharge flooding occurring at node 80608 due to the orifice restricting flow in the upper branch. HEC-RAS maintains a higher stage at the node and the head difference is used to compute flow through the top inlet as opposed to EPA-SWMM allowing all excess volume beyond the rim elevation to exit the pipe network.

When looking at the downstream boundary, HEC-RAS shows higher flows and volumes (~5%) reaching the model boundary outlet due to the difference in surcharge calculations shown above.

Extran 6 and 7 Tests - Pump

The Extran 6 and 7 tests are modifications of the Extran1 dataset to include a pumping stations at that pumps water from the upper branch to the lower. The pump draws water from the surcharging branch to the branch with additional capacity, preventing surface flooding. This comparison uses the EPA-SWMM Extran 6 geometry as defined in the Extran Examples .inp file provided on OpenSWMM. The HEC-RAS dataset was built using the Extran1 dataset as a starting point. Then nodes 82309 and 15009 were connected via a pumping station. The pump station is controlled by a set of Rule Operations defined in the Unsteady Flow Data Editor that aim to mimic the step function pump curve from the EPA SWMM model. 

The Rule Operations for the pump in the HEC-RAS model and the step-function EPA-SWMM pump curve are shown below:

On the whole, flow results between the two models agreed well. The figure below shows flow results at the pipe network outfall node 16009. Similar to the other Extran tests, the EPA-SWMM results show a slower recession on the falling limb of the hydrograph at most nodes which is likely due to how each model is discretizing the pipe network.A few differences were noted between EPA SWMM results and HEC-RAS results for this dataset. First, the timing of the pump was slightly off between the two datasets which is likely a result in differences between how each software package handles pump operations (on/off, transitions etc.) and slightly different hydraulic conditions in the system. Another difference was that at peak flows the HEC-RAS results show a near steady state in the system whereas the EPA-SWMM model shows flow and stage increasing slightly on the upper branch until the inflow hydrographs recede. This difference can be seen in the hydrograph plot for node 82309 (pump headwater) shown below.

Examining the results for conduit 1602 on the upper branch just downstream of the pump, it appears that this conduit is more restrictive in EPA-SWMM than HEC-RAS, causing the higher stages at node 82309. In other words, a higher head is required to pass a given flow through the conduit in EPA-SWMM as compared to HEC-RAS. This is likely due to the fact that the conduit is partially pressurized during the simulation as shown in the profile plot below. Since HEC-RAS uses a finer discretization of the pipe network (conduit cells and faces instead of node to node) it can resolve the moving pressure front within a conduit as opposed the EPA-SWMM's Extran surcharge method which is setting the cross-sectional area to full for the whole conduit, increasing the friction losses.