Summarizing the GRF’s correlation structure
The semivariogram can be computed with geoR::variog()
.
Some arguments:
option
:"bin"
(default) returns values of binned semivariogram"cloud"
returns the semivariogram cloud- i.e. points \((h_{ij},v_{ij})\) with:
\(h_{ij}=||s_i−s_j||\)
\(v_{ij} = \frac{1}{2} \cdot (Z(s_i) - Z(s_j))^2\)
- i.e. points \((h_{ij},v_{ij})\) with:
"smooth"
returns the kernel smoothed semivariogram
uvec
: to define the variogram binning (option = "bin"
).- if a scalar is provided, this is the default number of bins
- if not provided: 13 bins are calculated
max.dist
: maximum distance for the semivariogramtrend
: specifies the mean part of the model. By default,trend = "cte"
so the mean is assumed constant over the region.
A trend can be fitted using ordinary least squares: then variograms are computed using the residuals.