Is there a way to quantify the numerical error in a trainsient simulation?
I know in steady state run, you can actually see the error and convergence chart. 
And in MODFLOW, people like to use the discrepancy error curve to justify the accuray of a model run?