Errors in prediction

[Juha Uotila <uotila@fimr.fi>]

Hi all,

I have a drifter data set of latitude and longitude time series. In addition, I
know the variance due to the measurement errors in each point. I have
interpolated the original positions at the constant intervals by applying the
GSTAT program using ordinary Kriging as in example below.

>#
> 
>data(lon): 'pos54.kmp2', x=1, v=2, min=3, max=10, radius=2, average, V=4;
>data(lat): 'pos54.kmp2', x=1, v=3, min=3, max=10, radius=2, average, V=4;
> 
>variogram(lat):  0.04 Nug(0) + 51.2041 Pow( 1.54327 );
>variogram(lon):  0.04 Nug(0) + 13.1435 Pow( 1.3052 );
>data(): './date_1.98', x=1;
> 
>set output = '/tmp/gstat.int';     

Here I defined the separate variograms for the both latitude and longitude and
the Nugget was defined as an average error in position. Then I had
variances due to the position errors of each observation in 4th column of the
position file 'pos54.kmp2'. 

Now, I would like to know if this is a correct way to interpolate the position
component with respect of time? Especially, I wonder if there is any conflict
between including the Nugget value as a measurement error and also measurement
errors in column 4 from where it is added to the diagonal of the variogram
matrix (according to GSTAT manual). I think I am now confusing the observation
variogram (incl. measurement errors) with the variogram of the true data?

regards, Juha Uotila