31 Mar 2009

Self-Fulfilled Prophesies

As I noted here and here, simulations are only as good as the parameters specified, but this paper [PDF] takes that obvious truth a step further. In it, Kidson et al., estimate the reliability of Metropolitan's water supply -- given different shares of imported water in Met's total portfolio. (Local water may come from recycling, recycling, desalination, etc.)

Of course, you wouldn't know that this result was pre-determined by reading the abstract!
This study applies Monte Carlo stochastic inflow generation and a deterministic sequential approach to capture MWD’s harvesting and storage dynamics. Long-run reliability of MWD’s Year 2020 target portfolio is estimated, accounting for hydrological correlation between source regions and entitlement constraints. The results indicate that the target portfolio delivers high reliability (>98%), but that reliability was sensitive to portfolio composition. As the proportion of the hydrological sources increases within the portfolio, reliability declines markedly given 2020 parameters. This outcome implies that it is essential for Southern California to implement MWD’s diversification plan (and continue the development of alternative supplies over the next 10 years) in order to maintain future water source reliability.
"hydrological sources" refers to imported water, but let me spell it out for you: As the share of water coming from distant places (as opposed to local places) rises, water supply reliability falls. That's pretty obvious -- if you believe that water coming from a distance is less-reliable than water under your feet.

Here's a different way of explaining the "result": You have two sources of pizza -- your freezer and a delivery service. Say that is takes 20 min to prepare a frozen pizza and 30 min to get a pizza delivered. Now, if you get half your pizza from the freezer and half from delivery, your average wait time for any given pizza will be 25 minutes (50% * 20 minutes from the freezer + 50% * 30 minutes for delivery). Taking that as a baseline, it should be pretty obvious that increasing your consumption of delivered pizza will increase your average wait time while decreasing your delivered pizza consumption will lower the average wait time.

[Note: If you've run across delivered pizza that's better than frozen pizza, you may be willing to wait a little longer. The same applies to imported water -- it may be better (cheaper) than local water, even if it's less reliable.]

From this example, you can see why the paper's main result -- MET can increase water reliability if it uses more local water and less (unreliable) imported water -- is so obvious that the simulations are unnecessary.

So why did three people take all that time to write the paper?
  1. Academics call this "research."
  2. MET's managers want $millions for desalination and recycling plants, and they can use this "academic study" to convince MET's directors ("Southern California") to approve the expenditure.
BTW, I doubt if more than 10 percent of the directors who might be asked to approve such an expenditure (based on arguments "supported" by this paper) would even know what "Monte Carlo stochastic inflow generation and a deterministic sequential approach" means. (The first bit means that the inflow is estimated as the average of many draws from a probability distribution; the second bit means -- I think -- that water was allocated after it was received according to a known demand.)

Bottom Line: It's ALWAYS better to have local supplies of water (or pizza), but sometimes it's not worth the cost. (I coulda told you that without even going to Monte Carlo!)