The California Water & Environmental Modeling Forum (CWEMF) has proposed to update its water modeling protocol guidance, last issued in 2000. This modeling protocol applies to many other settings, including electricity production and planning (which I am familiar with). I led the review of electricity system simulation models for the California Energy Commission, and asked many of these questions then.
Questions that should be addressed in water system modeling include:
- Models can be used for either short-term operational or long term planning purposes—models rarely can serve both masters. The model should be chosen for its analytic focus is on predicting with accuracy and/or precision a particular outcome (usually for short term operations) or identifying resilience and sustainability.
- There can be a trade off between accuracy and precision. And focusing overly so on precision in one aspect of a model is unlikely to improve the overall accuracy of the model due to the lack of precision elsewhere. In addition, increased precision also increases processing time, thus slowing output and flexibility.
- A model should be able to produce multiple outcomes quickly as a “scenario generator” for analyzing uncertainty, risk and vulnerability. The model should be tested for accuracy when relaxing key constraints that increase processing time. For example, in an electricity production model, relaxing the unit commitment algorithm increased processing speed twelve fold while losing only 7 percent in accuracy, mostly in the extreme tail cases.
- Water models should be able to use different water condition sequences rather than relying on historic traces. In the latter case, models may operate as though the future is known with certainty.
- Water management models should include the full set of opportunity costs for water supply, power generation, flood protection and groundwater pumping. This implies that some type of linkage should exist between these types of models.