next up previous contents
Next: Discussion Up: The Kriging Estimator Previous: Minimum Estimation Variance   Contents

Matrix Form

Both kriging systems can be easier expressed in matrix form. Let denote $K$ the ``kriging matrix''

/begin{displaymath}K = /left(
/begin{array}{cccccc}
/bar{C}(v_{1},v_{1})&/ldots...
.../bar{C}(v_{n},v_{n})&1//
1 && 1 && 1 & 0
/end{array} /right) /end{displaymath}

and $/boma{/lambda}$ and $/boma{c}$, two vectors,

/begin{displaymath}{/boma /lambda} = /left(
/begin{array}{c}
/lambda_1 //
/lamb...
...
/vdots //
/bar{C}(v_{n},V) //
1
/end{array} /right)/ / / , /end{displaymath}

then the kriging system gets the simple form

/begin{displaymath}K /boma{/lambda}= /boma{c}/ / / ,/end{displaymath}

which, in the case of invertibility of the matrix $K$, immediately provides the solution for $/boma{/lambda}$:

/begin{displaymath}/boma{/lambda}=K^{-1} /boma{c}/ / / ./end{displaymath}

The expression of the kriging variance becomes

/begin{displaymath}/sigma_K^2 = /bar{C}(V,V) - /boma{/lambda}^{T} /boma{c}/ / / ,/end{displaymath}

where $/boma{/lambda}^{T}$ denotes the transposed vector of $/boma{/lambda}$.



Rudolf Dutter 2003-03-13