Spatial sampling

How do we begin the analysis of continuous variables?

The first step is to create a spatial sample. For the example in the beginning (rainfall in Switzerland), this would be the meteorological measurement stations. Their positions are fixed and are not freely selectable (unless you use just a subset of all existing stations). However, if you want to analyze e.g. the distribution of contaminants in the soil you first need to define the measuring points. You must be aware of the following characteristics of the sample:

• Representativeness
• Homogeneity
• Spatial distribution of measurements
• Size (i.e. number of measurements)

Representativeness, homogeneity, spatial distribution, and size are related. A size of 5 monitoring stations for an estimation of the total Swiss rainfall would hardly make sense and is neither representative. Nor would the selection of all the Swiss-german monitoring stations be representative for the overall estimation of Swiss rainfall. The size could be sufficient but not the spatial distribution. Selecting all stations below 750 masl the sample could be sufficient according to size and distribution but the phenomenon is not homogeneously represented in the sample. A subsequent estimate would be significantly distorted mainly in the areas above 750 masl.