This is an interesting thought. Hydrology and economics both have stock/flow models. I don't know much of anything about hydrology, but you seem to be describi

This is an interesting thought. Hydrology and economics both have stock/flow models. I don't know much of anything about hydrology, but you seem to be describing a directed acyclic graph with multiple sinks (i.e. nodes with no outgoing edges, "drainage basins"). The full water cycle is not explicitly modeled; there is no "edge" from a drainage basin to the atmosphere, and from the atmosphere back to the ground, to create a cycle. In economics, there are many models of cyclical flows of commodities or money (acylic graphs are a special case of limited interest, as far as I know). Historically, the physiocrats were among the first to propose one (the "Tableau Economique"), to support their doctrine that the growth of national wealth was limited by agricultural output (which was, to a first approximation, true in their time). The descendants of the Tableau are Leontief input-output models, which form the basis of national accounts (from which we get concepts like Gross Domestic Product). You can calculate the "leontief inverse" of an input-output matrix that lets you characterize how coupled different firms or sectors are in a given economy. Under assumptions of linear technology, these let you answer questions like: "How much would the total output, in dollars, of the steel sector have to increase to increase the total output of the automobile sector by $10 million dollars, relative to last year?" If a sectors or firm has a relatively large multiplier, it means that it has a greater impact on the whole economic system than those with a smaller multiplier. When you hear claims like "building this factory would indirectly create 10k jobs", it is usually based on a regional input output model commissioned by a lobbying group or government agency. I'm dimly aware these techniques have been used to model the diffusion of inflation from sector to sector, but I believe the state of the art involves convex optimization techniques that are beyond my ken.