There is a persistent assumption that free water modelling software is what you use until you can afford the real thing. It has the relationship backwards. EPANET, SWMM and MODFLOW are public domain codes from the US EPA and USGS, they are the reference implementations in their fields, and several commercial packages embed or extend them rather than competing with them.
So the question is not whether the free solver is good enough. Usually it is. The question is what the licence fee buys.
What you are actually paying for
- The interface and the data management. Building a large network in a bare EPANET or MODFLOW interface is slow. Commercial products put a proper GIS editor, a scenario manager and a results database around the same solver, and on a big model that is most of the project time.
- Support with a name on it. When a model will not converge two days before a deadline, a support contract is someone obliged to help. A mailing list is someone who might.
- Capability the free code does not have. This is real and it is narrower than the sales material suggests. Integrated 1D and 2D, density dependent flow, certain water quality kinetics, unstructured meshes: these are genuine reasons to buy.
- Acceptability. Some regulators and some clients specify a package. Arguing that your open source result is equivalent is a fight you may not want during a submission.
What the free packages buy you back
Three things that matter more on funded and international work than they do on a domestic commercial job.
- The deliverable outlives the project. A study handed to a ministry or a basin authority in a licensed format is readable for as long as someone renews the licence. In EPANET or MODFLOW it is readable indefinitely, by anyone, on any machine. For a donor funded programme, that is the difference between capacity built and capacity rented.
- The method is inspectable. Public domain source means a disputed result can be traced to the code that produced it. On work that will be challenged, that is worth something.
- Scale costs nothing. Running ten thousand scenarios on a free solver costs compute. On a seat licensed product it may cost seats. This is the practical enabler for the kind of scripted analysis that makes a study reproducible.
The honest position
Most organisations should be running both, and most already are without having decided to. The pattern that works is a commercial package where the interface and the support are earning their fee on a large recurring model, and the free stack for everything that needs to be automated, handed over, or defended in public.
What does not work is choosing on price alone in either direction. A free solver with no one who can drive it is more expensive than a licence. A licence for a capability you never use is a subscription to a logo.
We cover the specific trade offs by domain in the flood, water network and groundwater comparisons.
