« Save the Statistical Abstract | Main | Small distractions »


Feed You can follow this conversation by subscribing to the comment feed for this post.


Have you read the forecaster discussions from the National Hurricane Center (these are released along with the public advisories)? Two things jump out: (1) as alluded to in the NYT article, there is much uncertainty around measuring intensity, let alone forecasting it - I get a sense that current intensity estimates are built from both intuition and models on top of quite sparse data. (2) The NHC clearly calibrates its intensity forecasts (and estimates for that matter) to a broader goal of communicating the "danger" of the storm.

In this case as I read the discussions, it was clear that Irene was never getting as strong (from a wind-speed perspective) as they expected even prior to landfall, and often was likely not to be even as strong as they were publicly saying. The predictions and advisories followed suit.

I'm not accusing them of hyping the storm - just that they know that the windspeed forecast is what will get the attention, so when they see an enormous storm and dangerous like Irene, they use the intensity forecast to communicate a broader sense of the potential danger of the storm. Archive is at: http://www.nhc.noaa.gov/archive/2011/refresh/IRENE+shtml/120913.shtml - the Friday and Saturday archived discussions are illuminating.

NHC has done a lot in the past couple of years on the track forecasts to emphasize the track cone rather than the line (going so far as taking the trackline out of the default maps). They've also added an advisory that the storm is bigger than the forecast cone. But these advisories expand the apparent risk of the storm. I'll be interested to see what they come up with on the intensity side that doesn't reduce the apparent risk of the storm, which is not what they're in the business to do. Smaller intensity/windspeed estimates with error bands around them may not serve the purpose.


Also, given the data below, does the table report that there was a forecasted 56% probability of winds not exceeding 34kts (39mph) during the time period in question (Sunday presumably), or does it report that there is a 44% chance of between 34kt and 50kt, with a 27% chance of 50-64kt and 5% of >=64kt? That would leave only a 24% forecasted chance of <34kt winds.

I would think for a hurricane the cumulative probability (from now to the period) is more useful as well, so that's a 68% + 29% = 97% forecasted chance of >34kt winds.

NEW YORK CITY 34 1( 1) 23(24) 44(68) 1(69) X(69)
NEW YORK CITY 50 X( X) 2( 2) 27(29) X(29) X(29)
NEW YORK CITY 64 X( X) X( X) 5( 5) X( 5) X( 5)


Gary: Above the table, it said:

...34 KT (39 MPH... 63 KPH)...
...50 KT (58 MPH... 93 KPH)...
...64 KT (74 MPH...119 KPH)...

so I interpret these as cumulative (right-tail) probabilities.


Down on the bottom there are a number of cases where the probability for 50kt is higher than the probability for 34kt, which doesn't make sense if it's cumulative. And none of the probabilities that I could find added to more than 100%, which you'd expect at least one of to do, if they were cumulative. So the data doesn't match the description :)

NORFOLK NAS 34 19 59(78) 13(91)
NORFOLK NAS 50 X 26(26) 28(54)
NORFOLK NAS 64 X 4( 4) 8(12)

NORFOLK VA 34 23 59(82) 11(93)
NORFOLK VA 50 X 32(32) 27(59)
NORFOLK VA 64 X 5( 5) 10(15)

CAPE HATTERAS 34 76 22(98) 1(99)
CAPE HATTERAS 50 8 76(84) 3(87)
CAPE HATTERAS 64 X 46(46) 2(48)


I'm not even sure these numbers are internally consistent - look at the first couple days at Morehead City or Cape Hatteras...


I've always thought that average climate forecasts for a particular time and location should include 95% confidence intervals. If I'm traveling, your average daily temp of 75 tells me little about what clothes I should bring. Maybe, or even likely, it was 90 degrees one year and 50 degrees another.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep