With present day computers, there are few hardware constraints in setting up a 1D model. However, for a 2D model the first step is to decide whether it is feasible and practical to set up a model.
Experienced modelers can usually quickly determine an answer by considering the following:
Clearly understanding/defining the model’s objectives, and if known, the modeling budget.
Determining the minimum cell size required to model the hydraulics accurately enough to meet the study’s objectives. Preferably at least three to four cells across the major flowpaths (depending on the topography). Minor flowpaths may be more coarsely or not represented if they play no significant role hydraulically in regard to meeting the modeling objectives. For example, residual water drains over a floodplain may not affect peak flood levels; in which case, it may not be necessary to model them.
If it is not possible to model a major flowpath with a sufficient cell resolution*, the flowpath can be modeled as a 1D branch cut through the 2D domain. This may allow a larger cell size to be used, and a greater area modeled in 2D, or a faster simulation time. For example, the river may be modeled in 1D and the floodplain in 2D.
Establish possible boundary locations for the model. These are influenced by locations that are well defined hydraulically, and any constraints on the extent of the topographic data (DTM). Dynamically linking with a 1D domain offers significant flexibility in locating the 2D domain.
Determine the number of rows and columns of the grid based on the overall dimensions of the 2D domain and the minimum cell size. Calculate the number of cells (rows by columns), and estimate the average number of cells that would be wet.
In 2001, using a P3 1GHz computer, overnight simulations of models varying in cell size from 5m to 60m for durations of 12 to 120 hours were achieved with several hundred thousand wet cells. For large models, it may be beneficial to start with a coarser cell size to facilitate quick turnover of simulations before proceeding to a finer cell size. This is a relatively easy process as most data input is not cell size dependent. Note that halving the cell size typically corresponds to increasing the simulation time by a factor of eight (four times as many cells and half the timestep).
* Example of a poor representation of a narrow channel in a 2D model