Hi all,
I agree with Mario for the case where data are abundant. For cases where
data are more sparse, or for cases where the trend is obvious, strong
and important, ignoring the trend is a waste.
A completely different case is that where the trend is not described by
coordinate polynomials, but by external variables known for each
location, think of the variable altitude for interpolating annual mean
temperature in a mountainous area. In this case, ignoring the trend is a
waste in general, imo.
Best regards,
--
Edzer
Mario Rossi wrote:
Rajni,
you generally do not need to detrend if using ordinary kriging (or
other linear weighted average estimator) in the interpolation case,
i.e., when the points you're estimating are surrounded by data in
(generally) all sides, and particularly in the direction(s) of the
trend. Also, a relatively small neighborhood should be used (generally
less than the variogram ranges; as opposed to large neighborhoods or
on at all). In such case, universal kriging or IRF-k would not give
you any additional advantage.
One possibel reference is Journel, A.G., and Rossi, M.E., "When Do We
Need A Trend Model?", Journal of Mathematical Geology, Vol. 22, No. 8,
1989.
Another could be Rossi, M.E., and Posa, D., "Stationary and
Non‑stationary Kriging Applied to Dissolved Oxygen Data",
Metron‑International Journal of Statistics, Vol. 47, No 1‑4, 1989, Rome.
Cheers,
Mario
+
+ To post a message to the list, send it to [email protected]
+ To unsubscribe, send email to majordomo@ jrc.it with no subject and "unsubscribe
ai-geostats" in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the
list
+ As a general service to list users, please remember to post a summary of any
useful responses to your questions.
+ Support to the forum can be found at http://www.ai-geostats.org/
