You can’t size your infrastructure for the outliers

Morganza spillway, Mississippi River

Morganza spillway, Mississippi River

Yesterday’s post about recurring suggestion that we build giant water infrastructure was premised in part on this year’s extremely wet year on the Missouri-Mississippi system, the “Why not build a big pipe from there to the desert” argument. As I mention, I hear this idea a lot, but especially this year, when it seems like there’s so much extra water over there that those folks don’t need.

OtPR nailed the problem with this argument in a post yesterday: “If you build for the peak, most of the time you’ll have excess capacity.” The case study at hand in OtPR’s post is a farmer who expanded to take advantage of this year’s extremely wet year, buying a bunch of extra equipment to put 500 marginal acres into production growing tomatoes:

But if Mr. Coburn bought a new tractor and harvester to support his most marginal 500 acres of land, he is so fucking stupid he deserves to lose his farm. Mr. Coburn knows from the past two years that he doesn’t get water for that acreage every year. How many wet years does he need to amortize $2M worth of equipment on 500 acres? Right now he’s burdened those 500 acres with $4,000 per acre worth of machinery, because in the wettest year in recent memory, he got some surplus water?

That argument applies across scales. If you build dams and canals capable of handling the big floods on the Mississippi, saving it and sending it to the arid West, most of the time they won’t be useful, because most of the time they’ll be empty. There are a whole lot of reasons I think the idea is a non-starter. This is one of them.


  1. Not that I think a national (really, continental) water grid is a good idea, but your argument doesn’t really prove the case.

    Consider the electric grid, which is designed for peak load (for which, this unbelievably hot week, I am immensely grateful).

    Consider bridges, which are designed for a load of large cargo trucks essentially parked on them. Most of the time, that is overdesign, but once in a while, it matters.

    Yes, in many systems there is a question of just HOW big an outlier you should prepare for. In the case of Japanese nuclear plants, obviously a five hundred year tsunami should have been well within the design spec and wasn’t, or at least not successfully so. A million year tsunami? Maybe not. You have to draw the line somewhere.

    But extreme events happen. Designing a system to function most of the time is sometimes a reasonable compromise. It depends on the importance of the system.

    If agriculture can be sustained with upper-midwestern water and southwestern sunshine, I say bring it on. Food is, in fact, sort of important in the grand scheme of things.

  2. Designing for peak is What We Do.

    Looked under the hood of a Mustang GT?
    or just a Mustang? 300hp V6, you need that for what? To rob a bank?

    Checked your computer monitor lately? 10% cpu utilization.

    Extreme events are now happening because we overbuilt Carbon Producing coal and oil plants, 100 year events now happen EVERY Year.

  3. The two Michaels are clearly right here, but in a way that requires refining the argument, not throwing it out.

    The question in each case – electric grid, Mustang, CPU and water infrastructure – is what is the cost of each additional increment of bigness at the margin versus what is the benefit for having reliability that much farther out on the risk curve. For each of these examples, a decision has been made about where the cutoff is, and the users of the system are bearing the cost of the reliability premium.

    Water storage and transportation infrastructure is staggeringly expensive given the volume of the stuff that you have to collect and move to be useful. Building such infrastructure large enough to catch the big events, like this year’s Missouri-Mississippi flood, and have it sit around unused a large fraction of the time – simply isn’t affordable. (In fact, there’s good evidence that water’s equivalent of the “base load” isn’t affordable for end users here in the West.) The affordability calculations in the cases y’all have offered – for Mustangs, computer chips, bridges, etc. – seem different to me.

    Thanks for pushing me to refine this argument.

  4. Pingback: You can’t size your infrastructure for the outliers on Ecocentric Blog | Food, Water and Energy Issues

Comments are closed.