This is a nice clear piece on system re-op. Whenever you hear someone say ‘system re-op’, this is basically what they mean. After you read that, come back to talk about water modeling.
Andrew Sullivan thought Dr. Lund’s analysis of a dry climate hydrology stood out of the longer article on California water.
Even with the worst conceivable climate change, the kind of global warming that brings 70-year droughts to California, the state might do okay.
That seems counterintuitive, but that’s what Jay Lund, who heads the UC Davis Center for Watershed Sciences, loves about his model of the state water system, CALVIN. He and his colleagues ran a range of climate scenarios through CALVIN, asking for a look at what very dry, very warm scenarios might do to the state’s water system out to the year 2100. The results were shocking.Basically, in CALVIN’s rendering of the future, the state’s economy is fine. “It was amazing how little the damage was to the state’s economy,” Lund said. …
Agriculture does not fare quite as well, but the state’s agricultural production only falls 6 percent. That’s despite increasing urbanization of agricultural land and, in the driest scenario, a 40 percent reduction in water deliveries to the Central Valley. “The farmers are all smart people and they’ll cut back the least profitable stuff,” Lund said. They’ll also fallow land, according to CALVIN—roughly 15 percent of the irrigated parcels currently farmed today, or 1.35 million acres.
I am sure that’s what CALVIN said, but you have to understand what CALVIN does to interpret it. (Dr. Lund, I know you stop by. Apologies if I don’t get this right.) CALVIN is an optimization model; when it does a run, it is hunting for the best possible outcome it can get by following the rules of its reservoirs and canals and fields and cities. Even more important, CALVIN knows the entire what, ninety? years of its hydrology all at once. It has perfect foresight. So basically, when it starts optimizing, it can look forward in 2012, see that the next two years (2013, 2014) are dry, and stash all the water it can. It also doesn’t have to hedge on whether to empty reservoirs before a flood; it knows whether a flood is coming. If there are no floods coming, it doesn’t have to empty reservoirs and can hold tight to stored water. When you know your hydrologic future, you can do a lot to store or release water smoothly.
I am not telling you anything shocking about CALVIN; all the modelers know this. But the best outcomes out of CALVIN include something we don’t have, which is specific, detailed knowledge about the hydrology in the future. When I see a surprisingly good outcome from CALVIN, I generally take that as the most optimistic case. Sure, we could do that well if we are perfect optimizers who know the future. Then I mentally adjust for being not-perfect optimizers who don’t know the future and figure our real life results will be somewhat worse.
On the public record 
Calvin has lots of issues (it does engineering-economics without friction, esp. of the political kind), but this is a real problem. Surely Jay can re-run CALVIN with a myopic storage decision? That would be more useful.
For all I know, he does that now. My understanding isn’t necessarily current.
Another thoughtful post.
Like all models (and estimations in general), CALVIN has imperfections. Perfect hydrologic foresight is one of them – something also partially implicit in many simulation models (since their rules were calibrated for the same historical hydrology that we run them for). For the application discussed above, most optimization benefits come from more economical adjustments to conditions by water trading and shifting drought storage from reservoirs to aquifers. Most of these benefits are immune to hydrologic foresight. We have run CALVIN with less foresight – see Andy Draper’s fascinating dissertation (2001). California’s large groundwater storage capacity really helps dampen the effects of imperfect hydrologic foresight.
My biggest concerns for all water system modeling in California (including CALVIN) is just getting the mass balances right. Fortunately, recent groundwater modeling efforts by DWR and USGS have really helped here.
I’m delighted to see a nice discussion of modeling – although any modeling discussion is probably painful for others. Just to clarify, the purpose of CALVIN is not to simulate political and legal reality, but rather to quantitatively explore California’s water supply system and see what might might be possible to improve its economic performance within environmental constraints. (Imperfect foresight creates acute problems for flood management, for which CALVIN is nearly useless.) Sorry for the long reply.
In the original post I think that it would have been more correct to say that CALVIN has perfect hindsight, rather than foresight. But using ninety years of observed record is a good start. There is a lot of variability, including effectively two six-year droughts, in those ninety years of record. However, that is still a relatively short record, so that if we had 1000 years of record during the present high sea level stand, a period of relatively stability in terms of global weather, notwithstanding the Little Ice Age 300 years ago, there would be both more extreme droughts and more extreme floods. And, since weather appears to be more random than say earthquakes, those more extreme floods or earthquakes could occur at anytime. Jay’s discussion was not at all too long for the importance of the subject. There is a real need for better tools than CALVIN or CALSIM as well as better education of the masses who tend to think that anything that comes out of a computer “model” must be right. Actually, that’s not fair to the real masses who tend to have more commonsense that the politicians and other players who overemphasize the importance of imperfect models and think they represent “good science” when in fact they are just crystal ball gazing of varying degrees of expertise.
Not only is CALVIN imperfect, I have some problems with Mr. Lund’s basic assumptions:
“The farmers are all smart people…” Well, here in the Northstate, yes. In the Central Valley, where farmers plant thousands of acres of permanent crops KNOWING dry years and shortages are inevitable and KNOWING their water rights are junior to the north’s water rights, not so much.
Perhaps CALVIN’s model parameters can be amended to include bad business decisions.
I appreciate Jay putting this into the big picture perspective. I would add that the result is not only because of optimizing behavior in the future built into Calvin, but is also because the importance of water to today’s economy is much lower than many people think. This years drought and its large reductions in water consumption will probably amount to about a $2 billion loss in a $2 trillion CA economy. The overall sensitivity to water will be even smaller in the future as the economy becomes more urban and service oriented and technology continues to advance.
The urban areas will have an adequate water supply for their economies, although important future choices will determine the composition of that future portfolio will look like and what it will cost. Jay is right, the issue is that the state is unlikely to be able to sustain about 10% of its irrigated cropland over the next 50-100 years.