[NARCCAP-discuss] extremely small precipitation value

[Gutowski, William J [GE AT]]

Jianhua:
	I agree with Seth.  The very small negative values, in particular, arise
in some models from truncation error.  These models may accumulate
precipitation over the course of a run, and when they have to put out the
value for a specific period, like the 3-hourly precipitation in NARCCAP,
they subtract the cumulative amount at the beginning of the 3-hr period
from the cumulative amount at the end of the 3-hr period.  This often
means getting small difference of two large numbers.  Even though the
actual precipitation during the period might have been zero, the
truncation error can yield a very small negative (or positive) number.
	Of course, a numerical model's precipitation scheme can produce actual
precipitation that has very small value because, unlike a rain gauge, it
can "observe" very tiny numbers.
	Either way, the very small amounts you mention mean nothing to the
overall accumulated precipitation or time average precipitation in the run.

Bill Gutowski
-- 
William J. Gutowski, Jr.

3021 Agronomy Hall
Dept. of Geological & Atmospheric Sciences
Dept. of Agronomy
Iowa State University
Ames, Iowa  50011-1010

Tel:  1-515-294-5632
Fax: 1-515-294-2619
http://www.ge-at.iastate.edu <http://www.ge-at.iastate.edu/>
http://rcmlab.agron.iastate.edu <http://rcmlab.agron.iastate.edu/>






On 2/12/13 1:01 PM, "Seth McGinnis" <mcginnis at ucar.edu> wrote:

>Hi Jianhua,
>
>All of the RCMs can put out very small precipitation values; those are the
>values in the files, not a problem reading them into R.  The theoretical
>lower
>limit is zero, so you should treat negative values as zero. (We have
>plans to
>set them to zero in the published data at some point in the future.)
>
>Whether you should use a trace threshold and ignore values below that
>level
>depends on how you are using the data.  The conventional trace value used
>in
>meteorological observations is 0.01 inches per time period.
>
>Cheers,
>
>--Seth
>
>
>On Mon, 11 Feb 2013 15:58:14 -0700
> "Jianhua Huang" <jh.eco.cas at gmail.com> wrote:
>>Hi all:
>>
>> 
>>
>>I am using the precipitation (pr) data from rcm3-cgcm3 model. Some values
>>are very close to zero, for example 3*10^(-23),  and there are even some
>>negative values (like -3*10^(-21)). Should I treat these values as zero?
>>I
>>don't know whether this problem is caused by the software I used to
>>process
>>the data (I am using the ncdf4 package in R). Anybody knows the
>>theoretical
>>smallest output from the rcm3-cgcm3 model, so that I can treat all values
>>less than this value as zero?
>>
>> 
>>
>>Thanks
>>
>> 
>>
>>Jianhua 
>>
>>_______________________________________________
>>narccap-discuss mailing list
>>narccap-discuss at mailman.ucar.edu
>>http://mailman.ucar.edu/mailman/listinfo/narccap-discuss
>
>_______________________________________________
>narccap-discuss mailing list
>narccap-discuss at mailman.ucar.edu
>http://mailman.ucar.edu/mailman/listinfo/narccap-discuss