Dear Ilan,

Thanks for your interest in the connectivity tool. The underlying connectivity model, implemented in Matlab, has been applied at global scale; see here. Figure 2 of that paper might be similar to what you are looking for, or inspire some thinking.

The main scalability constraints on the tool are memory and run time. The version of it implemented in MGET is restricted to a 32-bit address space, by virtue of MGET running inside of ArcGIS Desktop which is a 32-bit program. This means it is effectively restricted to about 1-2 GB of memory (MS Windows allows up to 2-3 GB for a 32-bit process but the practicalities of memory fragmentation effectively limit MGET to 1-2 GB.)

Speaking very roughly, once your simulation exceeds about 500x1000 cells and perhaps 30-60 days, the memory limit will be exceeded. This is not fully independent of the spatial size of the cells; the relationship between cell size and current speed affects how small of a time step is necessary for the simulation to remain numerically stable, which ultimately affects the number of time slices needed.

If one is skilled in Matlab and as a Matlab license, it is possible to run the Matlab code outside of MGET in a 64-bit address space. In this case, you are only constrained by the available RAM of your machine, which is likely to be much higher than 1-2 GB.

I would suggest you consult with Eric Treml, who developed the underlying model and has continued to innovate in methods for modeling connectivity. He may be able to provide specific advice on the best way forward. Also I recommend this article for a general overview on the state of modeling.

Best,

Jason