Tuesday, June 26, 2012

More animals in the streets/The cone of uncertainty

Following on the recent post about a seal running around in Duluth, a twitterer tweeted this photo of an alligator using a crosswalk recently during Tropical Storm Debby in Florida.

Croc2

Like that other post, there are lots of photos of submerged cars in Florida:

  Croc

Perhaps the most interesting to me though was to see the hurricane path forecasts of Debby. Here’s South Florida Water Management District’s plot of the computer model guidance from 3 days ago (22 of June 2012):

storm_96

Each line is the possible track of the eye of the hurricane, as predicted by 15 different forecasting models run by various agencies. Things seem sure for the first 2 days that the storm will head north, but then after that it was anyone’s guess if it would head west to Texas, north or east across Florida.

Here’s the actual path (in purple) with another way of displaying the most recent forecasts (i.e. as a cone of uncertainty, than a spaghetti of possible scenarios).

track

And then here’s the equivalent spaghetti plot for the most recent model runs:

storm_04

Notice a tighter clustering of the model runs now, they all head off in the same general path. The important thing to take away from this is that sometimes weather models give confident results with high certainty, but in some situations the results can be all over the map (literally). In theory, sometimes the cone of uncertainty should be narrow, sometimes fat.

Do notice, however that the legend for the graph labels the orange cone “historical std dev [standard deviation]”, a measure of the typical error versus leadtime for all the historical predictions, averaged over a number of years. The width of this cone then, in practice, doesn’t change, even if the forecasters know that a particular situation is more uncertain than another. That said, requiring the forecaster to predict both the center of the cone and its width every time is an extra operational workload, and it’s something that takes extra training to do accurately.

2 comments:

  1. Tom - Those early diverging spaghetti plots were fascinating. I first started looking at them after a New Mexico water manager pointed out that the early western version of the tracks could bring moisture here (drought relief!). Two things worth noting. 1) Forecasters, take heart! A water manager was paying attention! and 2) This water manager was smart enough to make jokes about the uncertainty.

    ReplyDelete
  2. John, thanks for the comment...

    It's amazing how much difference a few months make, with this winter starting off fine and then things taking a hard right towards drought. Hard to put a number on it, not having looked at it closely, but my impression is that this spring was a 1 in 20 year event in the Southwest...This graph puts the Gunnison in 2012 (the red number) on par with 2002!
    http://www.wcc.nrcs.usda.gov/ftpref/data/water/wcs/daily_forecast/charts/Chart09114500_scatteryr.gif

    On the other hand too, maybe Phil Pasteris has some stories about how big hurricanes crossing Texas can turn around their whole water situation in a few days...

    ReplyDelete