9.3 The Modelling Process

9.3.1 Data Input Requirements

The minimum requirements for setting up a TUFLOW AD model are:

1. A properly constructed and stable TUFLOW hydraulic model; and
2. Boundary conditions for constituent concentrations (e.g. ocean salinities, catchment inflow pollutant concentrations etc.).

Initial conditions, dispersion coefficients, settling and decay rates will all be set to zero if not specified to be otherwise.

Preferable (and recommended) data requirements include:

1. Water quality calibration information as time-series data at points. This is particularly important for dispersion coefficient calibration;
   
2. Spatially variant initial conditions;
   
3. Particulate matter settling rates (if any); and
   
4. Dissolved species decay/transformation rates (if any).

9.3.2 Calibration and Sensitivity

Advection dispersion models are usually calibrated against water quality observations. For example, salinity recovery data can be used to calibrate and validate models, with longitudinal and transverse dispersion coefficients being the primary free variables. Dissolved and/or particulate constituents can then be simulated using the derived dispersion coefficients, and can include use of settling and/or decay rates as needed.

Ideally, models should be calibrated for conditions similar to those under investigation (e.g. a catchment inflow to an estuary) although this is not always possible, particularly when data is limited. In these situations, sensitivity analyses could be carried out by increasing and decreasing calibration variables, though this not a preferred approach due to the large variability in the literature with respect to acceptable dispersion coefficients.

9.3.3 Model Resolution

The 2D domain cell size needs to be sufficiently small to reproduce advection dispersion behaviour. It is worth noting that, in general, the larger the cell size is with respect to the scale of mixing processes, the greater potential there is for numerical dispersion to play a role in the model execution process. Even though TUFLOW AD has in-built measures to reduce these effects, it is advisable to keep this in mind, and perform at least some limited testing with your model to determine its sensitivity to cell size.

9.3.4 Computational Timestep

With TUFLOW Classic, the selection of the timestep is important in that the run time is directly proportional to the number of timesteps required to calculate model behaviour for the required time period. Notwithstanding this, with TUFLOW Classic the AD module will automatically substep with respect to the main hydraulic engine on the basis of maintaining both advective and dispersive stability (see Section 9.1.2) so the selection of timestep should be focused on ensuring hydraulic stability, as AD stability should follow, providing reasonable dispersion coefficients are set.

TUFLOW HPC, being an explicit method already, uses a self-adaptive timestep which is much smaller than Classic. Further, it uses the fourth order Runge-Kutta temporal integration scheme, and the advection dispersion calculations are performed on the same timestep and with the same scheme. When running AD models in HPC There is generally no need for the user to change the timestepping from the default method. Using high dispersion coefficients may cause TUFLOW HPC to use a smaller timestep than otherwise required for the hydraulic calculations due to the diffusion (Peclet) number control.

9.3.5 Example AD Models

A step-by-step description outlining how to set up an AD model is provided on the TUFLOW Wiki Advection Dispersion Modelling page. Example models are also available in the Advection Dispersion Example Model Dataset on the TUFLOW Wiki.