[R-sig-Geo] number points interpolation

[Paul Hiemstra]

Hi Alexandru,

Depends...if you want to make an interpolated map on 8 points, I would 
say that 8 points is quite a small amount. A rule of thumb would be 
around a minimum of 30 observations. On the upside, I think you have 
more than just 8 points because of the temporal dimension. So for 
fitting the trend between satellite and ground observations you can use 
n_times * 8 observations, assuming that you fit one set of trend 
coefficients to the entire dataset. When the coefficients become 
variable in time, this number decreases again. When predicting maps per 
time step you do only have 8 residuals to base the map on, which is on 
the low side.

You could also look at a space-time kriging approach, this allows you to 
use information from other time steps to predict the map for the current 
timestep. See Heuvelink 2010 for an example of this approach [1].

So lot's of thoughts from my side, but it depends really on your approach.

cheers,
Paul

[1] http://dx.doi.org/10.1111/j.1538-4632.2010.00788.x

On 09/01/2010 07:39 AM, Alexandru Dumitrescu wrote:
> Hello everybody,
>
> Is 8 spatial points sufficient to perform a multivariate spatial
> interpolation,  using as independent variable satellite data, on a monthly
> temporal scale? Sample points consist in meteorological data which seams to
> be highly correlated with satellite images.  The geographical area is the
> territory of Romania.
>
>
> Thank you,
>
> Alexandru
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>    


-- 
Drs. Paul Hiemstra
Department of Physical Geography
Faculty of Geosciences
University of Utrecht
Heidelberglaan 2
P.O. Box 80.115
3508 TC Utrecht
Phone:  +3130 253 5773
http://intamap.geo.uu.nl/~paul
http://nl.linkedin.com/pub/paul-hiemstra/20/30b/770