Scan of parameter space and model calibration.

A: Scan of parameter space over the learning rate α and the movement rate m. The darker red colours indicate higher sum of squares between empirical and simulated diffusion curves. The black cross indicate the parameter values generating the lowest sum of squares between empirical and simulated diffusion curves, thus indicating the model that is best able to replicate these diffusion curves (i.e. best-fit model). B: Empirical and simulated (for the best-fit model) diffusion curves over 20 days for the sub-populations in which solvers using solution s2 where initially released; C: Empirical and simulated (for the best-fit model) diffusion curves over 20 days for the sub-populations in which solvers using solution s1 where initially released; D: Empirical and simulated (for the best-fit model) diffusion curves over 20 days for the control sub-populations in which no trained solvers where initially released.