Spinup of the 3D groundwater model ParFlow can be computationally expensive, and yet the convergence is dependent on many factors.
Here, we simply examine a spinup simulation using ncl script
parflow_terrsympHydro_pulse.ncl. This script co-plots the time-series of number of function evaluations needed for model convergence, and the offline atmospheric domain average rain rate. Here the atmospheric domain is relatively larger than the model computational domain.
The increase in number of function evaluations results in slowing of the model run speed, i.e. larger wall-clock time.
Qualitatively, the above figure shows that the ParFlow Pulse (here number of Function evaluations) is correlated to domain precipitation.
So, ParFlow computation is relatively faster during recession period than during precipitation period in above simulation.
