DRAFT: Create a 'cut down' model

Why ‘Cut Down’ A Model?

Cutting down a model refers to extracting a subarea of a larger model and configuring it to satisfactorily reproduce the results of the larger model. This is commonly desired when a regional model is available (often through a government agency or authority) and it is to be used for a localised development impact assessment. The modeller may wish to cut down a model to make it fit for purpose or to facilitate further model refinement for an area of interest through, for example:

• reducing simulation time;
• reducing CPU RAM requirements during initialisation; and/or
• reducing size of outputs.

Should a TUFLOW HPC model not be suitable to be cut down, and/or otherwise has fit for purpose simulation times and memory requirements, then introduction of nested Quadtree resolution refinement may provide a suitable solution. It should be noted that for a Single Level Quadtree model of the same cell size and orientation, Quadtree and HPC produce near though not identical results, due to differences in the solvers, as discussed [[1]]. Incorporating areas of smaller cell size will also reduce the model’s timestep, lengthening the simulation time.

For refinement of outputs only, then use of Model Output Zones can be considered.

How To Cut Down A Model?

Generally, upstream and downstream boundaries need to be extracted from the original model and applied to the cut down model to reproduce the same flow conditions. The cut down model’s boundary locations are ideally chosen where the model configuration and hydraulic behaviour is simple, for example in areas of constrained flow, and avoiding potentially more complicated model elements or interactions such as 1D networks.

It is important to ensure that the selection of the cut down model’s boundary locations consider their zone of influence, ensuring the area of interest is outside of this zone and sufficiently distance to avoid impacting results there.

Consideration should also be given to the spatial extents of any internal inflow boundaries such as direct rainfall (2d_rf) or source area (2d_sa) flows. External inflow and outflow boundaries to the cut down model will ideally align with these, to ensure no double counting or missing inflow volume, and to ensure internal flows are distributed across the same areas as the original model.

Not all models are suitable to be cut down. This can be due to potentially complex flowpaths and hydraulic controls. Examples where models may not be suitable include, but are not limited to:

• Where there is insufficient or unsuitable model data to inform new boundary locations.
• Cutting a model across a floodplain with significant storage, upstream breakout flowpaths or varying flow directions.
• Where the area of interest for a cut down model is hydraulically controlled by a downstream interaction such as backwater effects that may depend on relative flows between a tributary and main river, which may present challenges in design event modelling and critical storm selection.
  

Configure inflow boundaries

• Extract flow from the original model via a cross section perpendicular to the flow direction, digitised left to right looking downstream.
• Ideally this is sourced from a 2d_po plot output line already configured in the original model. Post-processed flow can also be extracted from the temporal .xmdf map output in the QGIS TUFLOW plugin, however this will be limited to the map output interval set in the original model, which is typically less than the timeseries output interval that is applied to 2d_PO plot outputs.
• A constrained location (i.e. without significant flood storage) free of any circulation or break out flows is best to ensure all flow is captured.
• Reproduction of flood velocity is important. As such, if terrain grade down the catchment is variable, choosing a flatter boundary location separated from the area of interest by a steeper section can help to reproduce the incoming flow velocities more closely.
• Apply the extracted flows as a 2d_bc flow-time (QT) boundary to the cut down model.
• Enforcing an upstream water level (2d_bc HT) is also possible, extracted from the original model, however this approach is more likely to introduce changes in inflow volume and is generally not recommended.

Configure outflow boundaries

• Enforcing a 2d_bc head-time (HT) water level boundary is typical. 
• Extract water levels from the original model. Ideally this is sourced from a 2d_po plot output point already configured in the original model. Post-processed water levels can also be extracted from the temporal .xmdf map output in the QGIS TUFLOW plugin, however this will be limited to the map output interval set in the original model, which is typically less than the timeseries output interval that is applied to 2d_PO plot outputs.
• If terrain grade is variable down the catchment, choosing a flatter outflow location separated from the area of interest by a steeper section (such that tailwater level remains below the area of interest) can help to reproduce flow behaviour and limit boundary effects, with the steeper terrain becoming the hydraulic control.
• Enforcing an automatic generated 2d_bc height-flow (HQ) water level boundary (by specifying a slope in the "b" attribute) may also be appropriate if the area of interest is not in the zone of influence of the reconfigured downstream boundary. 

Reduce model extent

Active code boundary

• Trim the extent of the 2d_bc active code boundary to remove the excess model area, snapping it to the new external inflow and outflow boundaries locations of the cut down model.
• For rain on grid models, ensure that the cut down code boundary still follows subcatchment delineation as necessary, to ensure internal inflow volumes are replicated.

Further updates and improvements

Up
Back to TUFLOW Modelling Guidance