Randall et al 1997

Randall, Dean, Leasa Cleland, Catharine S. Kuehne, George W. Link, and Daniel P. Sheer. “Water Supply Planning Simulation Model Using Mixed-Integer Linear Programming ‘Engine’.” Journal of Water Resources Planning and Management. March/April (1997): 116-124.

Summary:
This paper describes the development of a water supply planning simulation model built for Alameda County Water District (southeast of San Francisco Bay area). The goals of the system were to be user-friendly, adaptable so that little coding was necessary for major changes to the system, and be able to be integrated with other models. The developers chose to model the system with a linear program rather than a network formulation because some constraints can be solved directly with an LP while a network formulation would solve these constraints iteratively, taking more time to solve.

The paper describes the different components of the system, including aqueducts, a creek, reservoirs, percolation ponds, water treatment plants, and an aquifer. There are multiple objectives in the system but the modelers chose to assign weights to each of the objectives so that the system could be modeled with a single objective for simplicity in solving. The paper gives many details of how specific operations were modeled and the challenges in doing so.

The model was applied to help the water district plan for its future needs. A range of future demand scenarios were run to determine the potential magnitude and frequency of shortages. The modelers added various new supply-side and demand-side components to the system to see which components would best alleviate their water shortages. In conclusion, the authors say that a LP or network formulation would’ve both been good approaches, but because of a few details, such as a binary constraint and specific operating details in the diversion dam, the LP was the preferred choice.

Discussion:
The paper is an interesting case study in applying linear programming to help find water resources solutions. However, the paper didn’t seem to be using any new methodologies, nor did it give more than a few sentences on why a linear program was better than a network formulation. I’m a little surprised this paper was published in such a major journal since the work seems insignificant, at least to me. I did almost the exact same work for my undergraduate research and I didn't think it was unique enough to be published. However, this is a slightly older publication and perhaps this kind of application of linear programming was a new methodology at that time.

The model was described as being highly simplified in order to save on computing time. If I were to continue research on this topic today, I would make the model more complex for accuracy. Computing time would not be as much of an issue in today’s powerful computing world. Complexities that I would be interested in adding in would include a hydrologic model to assess the impacts of rainwater harvesting.