Step 23: Loop!
Experiment with the input parameters, and the boundary conditions, to see if you can improve the model scores (reduce prediction errors). At this point, you have done all the hard work of setting up the software and learning how to use it. For very little extra effort, you can make additional simulation runs to perform comparison experiments, and see how the quality of the simulation is affected by boundary conditions, fault friction coefficient, mantle flow pattern, fault dips and interconnections, etc., etc., etc.! Use the prediction errors obtained from OrbScore2 (in the previous step) as an objective guide as to which model is more realistic, and try to minimize the sizes of these various prediction errors. Remember, every model (and every parameter set) contains its own systematic errors with respect to a real planet. However, we may reasonably hope that when we compare two simulations to see the differential effect of one parameter, the effect(s) of these systematic errors will largely cancel. In this way, we hope to get reliable insights about real planets from these imperfect models.
|
In order to try changing the... |
Return to Step... |
|
plate rheology, or basal drag (Note: I always suggest trying different values of fault friction, since this has such large effects on the solutions.) |
|
|
side (or subduction-zone) boundary conditions |
|
|
structure of the lithosphere |
|
|
map of heat-flow |
|
|
dip of faults (Note: Changing a vertical strike-slip fault to a dip-slip fault allows extra degrees of freedom in the solution; sometimes the effects are dramatic!) |
|
|
elimination of doubtful faults |
|
|
spatial resolution of the model |
|
|
addition of new faults |