8.2 Solver Specific Considerations
8.2.1 Classic Specific Boundaries / Options
There are a number of boundary types that are considered legacy types, are not commonly used, nor recommended to be used. These boundary types are not supported in HPC/Quadtree and are not currently planned to be supported. These include:
- ST - Flow vs Time. For an alternative, use the 2d_sa layer (Section 8.4.2).
- QH - Flow vs Water Level.
- VG - Variable Geometry. For an alternative, use Variable Z Shapes (Section 7.2.5.3).
- WT - Wind Stresses. For an alternative, use either an external stress boundary (Section 8.7) or cyclone boundary (Section 8.6).
- HS - Sinusoidal Water Level. In TUFLOW Classic these were initially a fixed field boundary where a number of tidal constituents (phase, amplitude and period) could be specified instead of a time-series. This was done initially in TUFLOW Classic in th 1990’s to be more memory efficient, however, memory (in particular time-series data) is typically no longer a limiting factor, so this approach has not been adopted for HPC. In TUFLOW Classic, HS and HT boundaries can be applied simultaneously to represent the different components of the boundary - for example a sinusoidal tide (HS) and a storm surge component (applied as HT), with the summation giving the overall storm tide. If multiple boundaries are to be applied to the same cells (e.g. to apply a storm surge on top of an ocean tide), these must be in separate GIS layers. In contrast, TUFLOW HPC does not support HS boundaries, or the combining of multiple water level boundaries on the same cells. Instead the tidal and surge components must be combined into a single HT time series.
8.2.2 HPC Specific Boundaries / Options
8.2.2.1 HPC Energy Options for 2D HT, HQ and QT Boundaries
Enforcing a uniform water level along the length of a boundary is not always realistic, and in some situations can lead to recirculating flow around the boundary (high velocity flow into the model towards one end of the boundary only for much of this flow to leave the model near the other end of the boundary). Historically, this could sometimes be remedied by increasing the viscosity applied along the boundary. For the TUFLOW HPC scheme we have however found a better approach (drawn from CFD modelling), to utilise a uniform total energy level along the boundary.
Consider the height vs time (HT) boundary, the approach is to utilise the user defined elevation as a total energy, and then apply a negative elevation correction to elevation on in-flowing cells (it is also permissible to apply an energy correction as a positive elevation adjustment on out-flowing cells, but generally this is not necessary for stabilising the boundary). This approach works well, however, it also means that for fast flowing inflow boundaries, a significant discrepancy will occur between the user-specified water elevation and the realised elevation in the model. This discrepancy is addressed by redefining the user-defined elevation as the water surface elevation corresponding to the Root Mean Squared (RMS) inflow velocity:
\[\begin{equation} h_{i} = H(t) + \frac{\bar{v^{2}} - v_{i}^{2}}{2g} \tag{8.1} \end{equation}\]
Where:
- \(H(t)\) is the user define height vs time boundary data
- \(\bar{v^{2}}\) is the mean square flow velocity averaged over boundary cells that are inflowing
- \(v_{i}^{2}\) is the magnitude squared of flow velocity at the boundary cells
- \(h_{i}\) is the applied water level at the boundary cells
Note, that this energy correction is only applied on in-flowing cells. In practice this approach means that for an out-flowing boundary the water level is constant over the length of the boundary line, but for an inflowing boundary some variation in water level will be observed along the length of the boundary with slower flowing regions being slightly above the defined elevation and faster flowing regions being slightly below the defined elevation.
For a QT boundary, the user specifies the flow rate as a function of time, not an elevation. However, internal to the software a fixed elevation boundary is utilised (it is a HX boundary connected to 1D storage node – refer Section 10.2.1 for a more detailed description of HX boundaries). Recirculation may also occur on the QT boundary (more so when it is digitised oblique to the flow path or digitised over highly irregular terrain), which again is addressed using the same energy correction described above.
Two new approaches have been introduced in the 2023-03 release and are available for HPC only (including Quadtree). The original approach (Method A) is the default for TUFLOW Classic, while Method B and Method C are new options for HPC. The extent of the energy corrections for the different approaches are detailed in Table 8.2. Note that Method C, which does not apply the correction to HX channel end boundaries is the default.
| Method | Description |
|---|---|
| Method A | No energy corrections applied (backward compatible) |
| Method B |
Energy correction applied for:
|
| Method C (default) |
Energy correction applied for: |
8.2.2.2 HPC HQ Boundary Approach
The HQ boundary computes the flux across the entire boundary line and uses a rating curve to apply a water level to the model. Each line is treated as a separate boundary and has a stage-discharge relationship, this is consistent with the TUFLOW Classic approach to HQ boundaries. Earlier versions of TUFLOW HPC applied the water surface slope on a cell by cell basis. It is possible to revert to all model HQ boundaries using a cell by cell approach by specifying ‘Cell’ within the .tcf command:
8.2.2.3 HPC HQ Boundary Stability
For cross-sections with a very low longitudinal slope, the stage vs flow (HQ) relationship can become unstable - a small change in flow causes a change in level that causes a similar but opposite change in flow. This manifests as high frequency noise on the HQ boundary levels and flows, but unless the model has fine temporal output resolution it can go unnoticed. Starting with release 2023-03-AD, the flow at all HQ boundaries is monitored for stability during the simulation. A stability warning message will be printed if at any timestep the flow at a given HQ boundary is (1) greater than 0.1% of the maximum flow defined in the HQ table for that boundary, and (2) less than 80% of the flow at that boundary in the previous timestep. The instabilities are easily addressed by calculating time-filtered flows and using these instead for the HQ table lookup and interpolation:
\[\begin{equation} Q_{f,i} = \frac{(F - 1)Q_{f,i-1} + Q_i}{F} \end{equation}\]
Where:
- \(Q_{f,i}\) is the filtered boundary flow at timestep \(i\) (initialised as zero)
- \(Q_i\) is the instantaneous boundary flow
- \(F\) is the filtering factor (must be greater than or equal to 1.)
Starting with release 2023-03-AD, the user may specify the global HQ boundary filter constant \(F\) using the command below. The default value from the 2025.0.0 release is 5, prior to this it was 1 (meaning no filtering).
8.2.2.4 HPC Additional Boundary Options
A number of other commands are available for controlling the behaviour of boundaries which are specific to the HPC solver. In most instances, these can be left at the default value. Refer to the following Appendix links for further details on each of these commands:
HPC Infiltration Drying Approach == Method A | Method B | Method C
HPC SX Momentum Approach == Method A | {Method B}
8.2.2.5 Quadtree BC Parallel Inertia Approach
Inflow boundaries should be perpendicular to inflow direction, but sometimes this is not possible and a slightly oblique boundary is required. For a momentum conserving scheme is it important that the incoming flow brings with it the correct amount of transverse momentum. With the 2023-03-AA release, the handling of transverse momentum, or parallel inertia, for Quadtree simulations (“QT”, “HT” and “HX” type 2d_bc layers) has been changed to “Method B” to provide better consistency with TUFLOW Classic and HPC. To revert to the 2020 release method, use “Method A” with the following command: