Dear Prakash,

A stochastic weather generator can only reproduce the statistics of the
sample of data used to generate the parameters, and cannot compensate
for a small sample of observed data.  If you use a small number of
years, you should have less confidence in the generated sequences.  If
the years that you select are unusually warm, cool, dry or wet, the
generated sequence will be biased.

I've seen a couple of rules-of-thumb supported by analyses:

Richardson, C. W., 2000. Data requirements for estimation of weather
generation parameters, Trans. Am. Soc. Agric. Eng. 43: 877-882.  This
paper recommended at least 20 years for precipitation and 10 years for
temperature and solar radiation.

Soltani A, Hoogenboom G. 2005. Minimum data requirements for parameter
estimation of stochastic weather generators. Climate Research 25:
109-119.  Recommended at least 15 years for WGEN.

Because the weather generator in WeatherMan uses more parameters than
WGEN, it likely needs more observed data particularly to capture
year-to-year variability, which it models explicitly.

On 6/29/2010 7:11 AM, Prakash Dixit wrote:
> Hi All,Could anyone please let me know that to generate longterm (say 50 yrs) daily weather data from WEATHERMAN how many minimum number of years of daily data should be available. Can 50 yrs of data be reliably generated using say 3 or 5 years of daily data?ThanksPrakash Dixit
>

--
James W. Hansen, Ph.D.
Research Scientist - Agricultural Systems, CCAFS Theme 4 Leader
The International Research Institute for Climate and Society
The Earth Institute, Columbia University, Lamont Campus
61 Route 9W, Monell Building
Palisades, NY 10964-8000
USA

voice:  +1-845-680-4410
fax:    +1-845-680-4864
e-mail: [log in to unmask]
WWW:    http://iri.columbia.edu/cgi-bin/staff?jhansen
Skype:  j.w.hansen