|
Spatially continuous phenomena such as rainfall or sea level cannot simply be described by a mathematical function. To analyze these variables, a spatial sample, i.e. a certain number of measurement points, is set up. To visualize a continuous spatial variable, the values between the measurement points need to be interpolated. First, the spatial sample or the set-up of the measurement points must be defined according to these characteristics: representativeness, homogeneity, a spatially optimal distribution, and sufficient number of points. Depending on the phenomenon and the measurement method, the design type of the spatial sample can vary (e.g. random sample, systematic sample, stratified sample, or clustered sample). Before the interpolation, we need to check if there is a dependency between the spatial data. Two methods are suited for this purpose: variography or the “moving windows”-method. Variography shows spatial dependency of the samples but not whether or not this dependency is equally distributed over the whole study area. For this, the “moving window”-method is applied. For the interpolation itself, several approaches with different consequences exist. Two ways to interpolate are presented here: the distance-based interpolation IDW (inverse distance weighting), and the geostatistical interpolation. With IDW, different distances are incorporated differently into the estimation. The influence of the distance weighting can be controlled by choosing the distance-weighting exponent. The higher the exponent, the more influence the measurement values of the adjacent points have on the result. However, it is not possible to have a direction-dependent weighting. With the geostatistic interpolation, the variography as the basis is derived from statistically distributed parameters. From the variography, the similarity of adjacent data points as a function of their distance from each other is indicated. The most important geostatistic interpolation methods are the Kriging methods.