A Statistician Goes to the XXIX Olympics
Sujit K. Ghosh, North Carolina State University
Sujit Ghosh (fourth from right) and Lian Xie (second from right), both 2008 Beijing Olympics sailing boat event held at Qingdao, China
from North Carolina State University, stand with fellow researchers in
front of the Qingdao Metrological Bureau building in Qingdao, China,
during the 2008 Beijing Olympics.
I
t was exciting to be a part of the Olympics in Qingdao, China, absolute errors. In other words, MOS compares the output from
this August. You must be wondering what a statistician was the operational wind forecast model (i.e., MM5, a fifth-generation
doing at the Olympics. Well, boats don’t sail without our wind Mesoscale model) with past observations and quantifies the model
forecasts. I was part of a team from North Carolina State University trend and its error in space and time. Then, the MOS system finds
tasked with predicting if the wind speed at a given hour of a given an optimal way to remove as much error from the existing forecasts
day would be at least 3 m/s—a required threshold value of the as possible, thus producing an improved real-time wind forecast. It
wind speed that is critical for deciding whether to hold the sailing was exciting work!
boat event.
There were other groups from outside of China that assisted the
My team—Huiping Miao, an NCSU PhD statistics student;
Chinese weather forecasters, including a group from the United
Lian Xie, professor of the NCSU Department of Marine, Earth, States National Center for Atmospheric Research (NCAR) at
and Atmospheric Sciences; Bin Liu, a postdoctoral fellow from Boulder, Colorado. Our MOS system, which is based on a dynami-
Qingdao; and Zengan Deng, a visiting student from Qingdao— cally weighted AR model, performed very well in forecasting the
worked in collaboration with the Qingdao Metrological Bureau wind speed as compared to some of the other forecast models. Our
(QMB) to develop a Model Output Statistics (MOS) system to system was able to reduce the mean absolute error for wind speed
improve surface wind forecasts by reducing biases in an existing and direction by as much as 30%–50% when it was tested for
(numerical) weather forecast model currently used by QMB. August during 2005–2007. The NCSU MOS system also per-
The new MOS system was used to forecast the wind field (speed formed well in forecasting the wind speed during the 2008 Beijing
and direction) during the sailing boat event at the Olympics. Olympics sailing boat event.
Data from the past three years indicated there were more weak Our trip to China was supported in part by the QMB, which is
wind (2–5 m/s) days than favorable wind days in August. Also happy with our MOS system and is now using it. Of course, it was
in Qingdao during August, land-sea breezes occur frequently, and fun to meet a number of Chinese scientists at QMB, and I loved
their impact on the sailing boat event was of concern. So, it was the food and hospitality. I never imagined that when I became a
critical to determine whether our MOS could correctly forecast the statistician, I would one day be at the Olympics.
threshold value of 3 m/s wind speed. A news article related to this project, titled “Scholars Get the
The MOS system developed by the NC State team is based Olympic Call,” appeared in The News & Observer, a local news-
on an autoregressive (AR) time series model. Forecasts from the paper in North Carolina. Its online edition can be found at
time series model are combined with the existing (numerical
www.newsobserver.com/front/story/1163385.html. Also, a slide show
model) forecasts as a weighted sum of both forecasts. Optimal taken during our China visit can be viewed at http://picasaweb.
dynamic weights for each day are obtained by minimizing mean google.com/sujitkg/QMBChina#slideshow. n
NOVEMBER 2008 AMSTAT NEWS 29
AMSTAT November 08.indd 29 10/24/08 2:28:24 PM
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84