Interpolate a scattered surface onto a grid by GP regression

Rasmussen & Williams (2006), eq. (2.23)-(2.24). Simple kriging with a known covariance and zero mean is algebraically the same predictor as the GP posterior mean, so method selects the reported label, not a different computation.

Usage

Srfintp(coords, values, grid, method = "gp", kernel = NULL, noise = 0)

Arguments

coords

Observation locations, n x d.

values

Observed values.

grid

Target locations, m x d.

method

"gp" or "kriging"; identical computation.

kernel

c(sf, l) or a function k(x1, x2).

noise

Nugget standard deviation.

Value

Named list with estimate, variance, method_used, n, method.

References

Rasmussen & Williams (2006), eq. (2.23)-(2.24).

Examples

Srfintp(coords = c(1, 2, 3, 4, 5, 6, 7, 8), values = c(1, 2, 3, 4, 5, 6, 7, 8),
  grid = c(1, 2, 3, 4, 5, 6, 7, 8))
#> $estimate
#> [1] 1 2 3 4 5 6 7 8
#> 
#> $variance
#> [1] 0.000000e+00 0.000000e+00 0.000000e+00 2.220446e-16 1.110223e-16
#> [6] 1.110223e-16 0.000000e+00 0.000000e+00
#> 
#> $method_used
#> [1] "gp"
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "Surface interpolation by GP posterior mean -- Rasmussen & Williams (2006) eq. (2.23)"
#>