# [ncl-talk] Using bootstrap resampling

Dennis Shea shea at ucar.edu
Tue Oct 31 19:15:26 MDT 2017

```re: "cone of uncertainty"

A snippet from:

https://www.washingtonpost.com/news/capital-weather-gang/
wp/2017/09/05/understanding-hurricane-forecasts-making-
sense-of-spaghetti-cones-and-categories/

"So what does the width of the cone represent? Contrary to popular belief,
it’s not a storm-specific indicator of confidence levels or a direct
reflection of model forecasts. Instead, the width of the cone at a given
time corresponds to the Hurricane Center’s average track error that far in
advance, based on the accuracy of past forecasts (calculated by a five-year
running mean of NHC forecast accuracy). The width of the cone at a given
point in time is the same for every storm and every forecast all year long.

The cone’s radius is designed to be 67 percent (two-thirds) the width of
the historical track error at a given point in time. In other words, if the
Hurricane Center is typically “off” by 300 miles five days out, then the
cone will extend outward 200 miles on either side."

--------

Tha above was a bit surprising. I always thought the "cone of uncertainty"
was based on an ensemble of forecasts related to the 'current' storm ...
not historical accuracies superimposed on a single current hurricane
forecast track.

------

Oh, your question ....   Yes!   :-)

On Tue, Oct 31, 2017 at 3:41 PM, Andrew Kren - NOAA Affiliate <
andrew.kren at noaa.gov> wrote:

> Dear ncl-talk,
>
> I'm trying to determine whether hurricane forecasts (track, wind, and
> intensity errors) are statistically different from each other given two
> model experiments in which one is a control and second contains added
> observations. The idea is to examine what impact the observations have on
> hurricane track.
>
> A number of forecasts are included in the analysis to where the arrays are
> a function of (storm name, forecast cycle, and forecast hour). I used the
> test to determine if the means were significantly different but I'm
> wondering if using bootstrap reampling is a more robust measure (
> https://www.ncl.ucar.edu/Document/Functions/Bootstrap/bootstrap_diff.shtml
> )?
>
> I have not used this function before however I've looked over the
> examples, specifically this one (https://www.ncl.ucar.edu/
> Applications/Scripts/bootstrap_diff_1.ncl) but it only includes
> 1-dimensional data. If I want to use bootstrap on my data, would nDim be
> (/0,1/) in order to get results as function of forecast hour?
>
> Thanks much!
>
> --
> Andrew Kren
> Research Scientist I, Global Observing Systems Analysis (GOSA) Group
> NOAA ESRL Global Systems Division (Rm 3C515)
> 325 Broadway, Boulder, CO 80305
> (303) 497-5418
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20171031/950e7818/attachment.html>
```