<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hi all,<br>
<br>
We are opening up our servers to make available the recently
computed 2nd-generation NCEP Global Ensemble Forecast System (GEFS)
reforecast data. For the currently operational version of the GEFS
(implemented Feb 2012), we now have, for every 00Z initial condition
from 1985 to 2010, 11-member, 16-day ensemble forecast data
available. <br>
<br>
Details are available at <a class="moz-txt-link-freetext"
href="http://esrl.noaa.gov/psd/forecasts/reforecast2/README.GEFS_Reforecast2.pdf">http://esrl.noaa.gov/psd/forecasts/reforecast2/README.GEFS_Reforecast2.pdf</a>.
<br>
<br>
The web page is at <a class="moz-txt-link-freetext"
href="http://esrl.noaa.gov/psd/forecasts/reforecast2/">http://esrl.noaa.gov/psd/forecasts/reforecast2/</a>.
<br>
<br>
We will fill in 2011 and 2012 data in the coming months. A BAMS
article is in preparation providing a more complete description of
the data set and potential applications.<br>
<br>
We hope and expect that this data set will facilitate both the
development of improved forecast products and facilitate many
interesting research projects. Anticipated applications include:<br>
<br>
(1) <b>Statistical post-processing</b>. With 25+ years of
reforecasts, even with rare events such as heavy precipitation or
tropical cyclones, we hope to have provided enough samples to allow
a useful calibration of the forecasts. The relative merits of
addressing model error via statistical post-processing vs. via
stochastic parameterization and/or multi-model approaches may be of
interest to this community.<br>
<br>
(2) <b>Understanding the forecastability of common and uncommon
phenomena</b>. Say you are interested in a particular phenomena
that doesn't happen very often, perhaps major lake effect snowstorms
in Buffalo. How well does the model do in forecasting such
phenomena? A multi-decadal data set will provide a sufficiently
large sample size to explore such a question and be able to provide
statistically meaningful results.<br>
<br>
(3) <b>Initialization of regional reforecasts</b>. Perhaps there
is an old case that you would like to run with WRF at 3-km grid
spacing. You need lateral boundary conditions and initial
conditions. The full reforecast data set archive on our DOE site
(again, see README file) can provide this.<br>
<br>
We don't have a large staff dedicated to making this all work. We
don't know what will happen should a dozen people try to download a
lot of data at the same time. Let us know if you have any
problems. We promise to do our best to sort through any issues.<br>
<br>
Our online data set is massive, nearing 150 TB in size (the full
model data set, about 1 PB, is also archived on tape at DOE).
Should you be interested, say, in accessing precipitation forecasts
over a particular country, every member, every day, every year, this
will involve reading and subsetting data from hundreds of thousands
of files. Expect a turnaround time for such requests of a day or
two. For smaller amounts, the turnaround time will be
commensurately faster. Also, please note that we have designed the
web-based interface to limit how much data you can download in one
slug. <br>
<br>
Thanks in advance should you choose to try using this data set. We
hope this data set is of use to you and your organizations.<br>
<br>
Tom Hamill, Jeff Whitaker, Gary Bates, and Don Murray <br>
(with ample assistance from Tom Galarneau, Mike Fiorino, and many
others)<br>
NOAA Earth System Research Lab<br>
Boulder, Colorado, USA<br>
</body>
</html>