**** NOTE on using edotmax ****

During iterations when the frictional sediment near the surface becomes unstable and starts to slump, the model viscosity in this region is progressively reduced during iterations in an attempt to put the material on plastic yield. This numerical softening leads to run-away in the viscosities in the subroutine viscosities1, where viscosity is set to sleast/edot, where sleast is either the yield stress for a plastic material or the viscous stress.

Increasing velocities (e.g. slumping) lead to increases in edot and then to decreases in viscosity which then results in a further increase in velocity and the cycle spirals down out of control with decreasing viscosities. This does not occur for viscous materials because they always are assigned their true material viscosity, not an artificial one that puts them on plastic yield.

The use of edotmax limits the maximum strain rate and controls the velocity of slumping. It does not guarantee good convergence during iterations but it allows controlled slumping so that the model can be continued through a slumping event. All other parts of the model will be nicely converged.

The way to run this is to choose an edotmax that is larger than any of the normal strain rates in the model, so it won't affect them. Set the max number of iterations per timestep to a relatively low value so you don't have a large number of timesteps with many iterations, and set force_continue = 1 so that the model will continue to timestep even though the convergence is not as good as the criterion you have set during slumping events.