[R-sig-Geo] spatio temporal anisotropy: help with stAni argument

[Dr. Benedikt Gräler]

Dear Bruno,

stAni relates spatial and temporal distances to each other in the model 
for the data set under study. A value of 100 km/day reflects that two 
stations 100 km apart have about the same strength of dependence as the 
values one day apart at the same location (This is in general only true 
for at most one pair of spatial and temporal distances and a 'global' 
compromise is sought). The function gstat::estiStAni provides a few 
heuristics to estimate the spatio-temporal anisotropy.

When a spatio-temporal vgm including a metric component is fitted, the 
stAni parameter is optimized under the objective to find the theoretical 
vgm surface that best fits the empirical one.

In krigeST, an additional and approximate stAni parameter can be 
provided to ease the (pre-)selection of the most relevant observations 
in the space-time cube for the prediction when local kriging is carried 
out. Note that the most relevant (i.e. strongest correlated) 
observations need not to form a cube, but depend on the actual vgm-model 
in use (see also the vignette "Spatio-Temporal Interpolation using 
gstat" at [1]).

The unit of stAni depends on the spatial and temporal units of your data 
set. Check the units of the empirical variogram to find out which units 
R/gstat uses for your data set.

Miss-specifying the stAni parameter will put too large or too small 
weights on the observations with a temporal difference resulting in bad 
predictions (this also holds vise-versa for spatial distances).

HTH,

  Ben

[1] 
https://cran.r-project.org/web/packages/gstat/vignettes/spatio-temporal-kriging.pdf


On 21.05.2018 15:16, Bruno Sesti wrote:
> Hi,
> I have some problems in understanding the usage of argument 'stAni' of
> vgmST and krigeST functions, for.the setting of spatio temporal anisotropy.
> How have I to interpret this argument, that is the integer value that I saw
> I have to assign to it, what does it means?
> How can this argument influence the result of prediction?
> Are units of spatial and temporal ranges and lags related to this argument?
> How can I study the spatio temporal anisotropy of my data?
> Kind regards.
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

-- 
Dr. Benedikt Gräler
52°North Initiative for Geospatial Open Source Software GmbH
Martin-Luther-King-Weg 24
48155 Muenster, Germany

E-Mail: b.graeler at 52north.org
Fon: +49-(0)-251/396371-39
Fax: +49-(0)-251/396371-11

http://52north.org/
Twitter: @FiveTwoN

General Managers: Dr. Albert Remke, Dr. Andreas Wytzisk
Local Court Muenster HRB 10849