I'm an archaeologist developing a simulation model of the flow of water through a prehistoric canal system in Arizona. The system, built by a people called the Hohokam, consisted of a large number of canals extending over 20 km along the Salt River in the area that is now covered by the city of Phoenix; another one similar in scale was built along the Gila River just to the south. Archaeologists have a number of questions about this system and, in particular, the organization and coordination it took to build, maintain, and manage such a system.

For a variety of technical reasons- most significantly the fact that my simulation will incorporate a number of other elements such as plant growth and even agent-based decision making on the landscape- I'm choosing not to make direct use of an existing flow simulation, but instead to reproduce the algorithm of one within my own simulation environment. I've been using the HEC-RAS version of the implicit finite difference implementation of the St. Venant equations, somewhat simplified to match some easier constraints in my own project's context. I'm not quite finished with this, and may be forced to plead for help on this in the near future.

For the moment I'm wondering about one issue in particular: evaporation. The online FEQ documentation claims that FEQ can account for evaporation, but I could find no information other than this one mention of the feature. My understanding of what is encompassed by the St. Venant equations may be incomplete (and if so I will very much appreciate clarification), but I'm assuming that simply removing some small fraction of water after one time step and before the next will throw the momentum and continuity equations somewhat out of balance. I'd very much like to see documentation about how FEQ handles this, or any papers or references that can discuss acceptable mathematical treatments. Incidentally, the issue is of interest because there has been a suggestion that the concentration of salts in the canal water increased significantly downstream as the water moved through the system, so that users downstream were more seriously and adversely impacted. I'd like to be able to incorporate this into my simulation.