Skip to content

Solution and check settings

Solution Settings and Check Settings, under Model Build → General Settings, define settings involved in solving the system of equations in your model. Solution Settings defines parameters involved in solver iterations, and Check Settings lays out the solution accuracy needed before iteration ceases.

Solve Settings

  • Under-relaxation Factor is a parameter that governs how unknown strength parameters are updated at each iteration. If underrelaxation is 1.0, the new strength parameter equals the parameter from the most recent iteration. If this value is 0.7, the new strength parameter is weighted 70% by the parameter at the most recent iteration and 30% by the parameter at the previous iteration. Lower values help damp out oscillations in parameter values from iteration to iteration and may improve convergence in some situations. Higher values close to 1.0 speed convergence when oscillation is not a problem. A good range for most cases is 0.9 to 1.0.
  • Maximum Iterations. When solving, iteration continues until the tolerances specified in Check Settings are met at all boundary condition control points or basis points, or if those criteria are not met, iteration ceases when this maximum number of iterations is reached.
  • Starting Heads Source. For transient runs, the first time step needs initial heads from some source. There are two possibilities here:
    1. Assign a constant value of head for each domain. The starting head at a point is set to the domain's average head, which is specified under Domains input. In the case of a Discharge-Specified (Multi-Domain) well, the starting head is set to the average head of the first domain listed. This option should only be used for very simple transient runs where the initial condition is a uniform, flat, constant head.
    2. Use another solved model in the same project. Select that model in Starting Heads Model. The model must already have a saved solution and may be steady or transient. If it is a transient model, the results from the final timestep are used. The chosen model must extend over the same or a greater area than the transient model, so that initial heads are available at all locations within the transient model.

  • Almost Dry Fraction. This parameter affects the solution in cases with unconfined, confined/unconfined, unconfined interface, and confined interface domains, where the heads can drop to low enough levels that the freshwater saturated thickness of the domain approaches zero. Instead of letting the domain have zero or negative saturated thickness, Anaqsim has a minimum saturated thickness. In an unconfined, confined/unconfined, or unconfined interface domain, the minimum saturated thickness = (average head - base elevation) * Almost Dry Fraction. In a confined interface domain, the minimum saturated thickness = (top elevation - bottom elevation) * Almost Dry Fraction. When heads drop to or below a level that would create this minimum saturated thickness, the aquifer is treated like a confined aquifer with this fixed minimum saturated thickness. This prevents domains from actually going completely dry and helps models converge despite portions of some domains approaching "dry" conditions.

    Setting the Almost Dry Fraction to a low value means that very little simulated horizontal discharge will occur in "'dry" portions of the domain. Heads in these "dry"' areas will drop to unrealistically low levels (below the base of an unconfined domain, for example), and have little real meaning. When contour plots are made of head, these unrealistically low values are neglected. These low heads do appear in other outputs, such as values sampled at the cursor and in profiles. Using higher values of Almost Dry Fraction may help convergence by limiting the magnitude of head gradients in "dry" areas, at the expense of allowing more actual discharge in "dry" areas. You can check the amount of discharge in a "dry" area with Conditions along a Line, under Output Specification, by making a profile of domain discharge along or across a line. Chose a low enough value of Almost Dry Fraction so the discharge in these "dry" areas is acceptably small.

  • Interface Leakage Option. This specifies one of two options for computing vertical leakage at a spatially-variable area sink basis point where a fresh/salt interface is present. If unchecked (default), the vertical leakage is computed just like it is in all other cases: the vertical leakage rate is proportional to the difference in head from one level to the next. If this column is checked and an interface is present in the overlying layer, the head difference is computed by using the freshwater head that is at pressure equilibrium with static salt water at the base of the overlying layer; the head in the overlying layer is not used to compute the head difference. The head that is used to represent the overlying layer is the same as the head at the toe of the interface in the overlying layer. The unchecked mode is appropriate when the resistance to vertical flow between levels is due to the K3 of the domains themselves. The checked mode is appropriate when the resistance to vertical flow between levels is due to an aquitard between the levels that is not explicitly modeled (in this case the resistance of the aquitard must be incorporated into the K3 values specified in the upper and/or lower domain). See Fitts et al (2015) for comparisons of these methods and discussion.

Check Settings

These settings define the accuracy of boundary conditions required of the solution; iteration continues until these conditions are met. If when Solve is pressed the solution converged before reaching the maximum number of iterations, all boundary conditions were met within the tolerances specified here.

The same thresholds are used when a boundary condition is checked. Such conditions include heads at head-specified wells and linesinks, extraction at spatially-variable area sink basis points, and the other conditions listed below. Check boundary conditions, beside the Check Settings table, checks every boundary condition and reports its convergence. Where the discrepancy between the specified condition and the modeled condition is greater than the threshold, it is reported; smaller discrepancies are not. If the solution did not converge to within these tolerances during Solve, the boundary conditions that missed them are listed in the solve log with their accuracy. That list shows which boundary conditions are holding up Solve. The default tolerances are set for length units of feet or metres, and time in days. If those units are changed, the check tolerances should be updated.

Four kinds of boundary condition tolerances are defined as follows.

  • Head Check Tolerance is the threshold for the magnitude of discrepancy between specified and modeled heads, used at head-specified wells and line boundaries.
  • Qn Check Tolerance is the threshold for discrepancies in the computed discharge per length along River line elements. It units are discharge/length [(L3/T)/L = L2/T].
  • Q Check Tolerance is the threshold for discrepancies in discharge, which are used along inter-domain line boundaries and normal-flux specified boundaries. In both cases, the condition being checked is the total discharge across a segment along the line [L3/T]. For inter-domain boundaries, it is the comparison of discharges on opposite sides of the boundary. For normal-flux specified boundaries it is the difference between modeled and specified discharges based on the specified discharge per length times the length of the segment.
  • Extraction Check Tolerance is the threshold for discrepancies in the modeled extraction (Equation 13 of Fitts (2010)) and the extraction computed from head values (Equation 6 of Fitts (2010)). The units of extraction and this threshold are discharge/area [(L3/T)/L2 = L/T].