Frequency maps can be constructed for class variables. An example of a class variable is soil moisture where each plot is assigned a soil moisture class. The plots of the Swedish Survey of Forest Soils and Vegetation are objectively distributed, and thus, can be regarded as surface representative, the share of the classes of each variable can be calculated within a geographical area.
Sweden is described in a coordinate system called RT 90, the Swedish National Grid. The southern border of the coordinate system is set at 6100000 m and the northern border at 7700000 m north of the equator. The western border is set at 1200000 m and the eastern border at 1900000 m east of Greenwich. The area was covered by a 25x25 km grid. For each studied variable, frequencies of classes at each intersection were estimated. At each intersection, all plots within a radius of 40 km were included. The actual distance of the plot to the intersection was considered following the principle that the significance of the plot decreases with distance to the intersection squared. Further the so-called area-factor was considered because it varies, partly as an effect of to which region the plot belongs, and partly because plots sometimes are split.
The area-factor is the area that the plot represents, for example, when the area covered by forest is summed. In region 1, the interior parts of northern Sweden, a plot represents a much larger area than a corresponding plot in region 5, along the coastal areas of southern Sweden. When a plot is split the area-factor must be reduced in proportion to the share of the split surface. When this was the case, based on its area-factor, the significance of the surface was linearly reduced in the estimation.
Since the distance between intersections was 25 km and the radii of the area was 40 km, each plot often influenced several frequency distribution estimations, in each case with different degrees of weighting. This resulted in an equalisation and improbable peaks in the final presentation were, thus, avoided.
For each intersection and studied variable, in the accordingly produced grid, is found the share of each class always summing up to 100%, and the number of observations.
The material was divided into five classes with equal class width and each class is assigned a colour. A linear interpolation to a 5x5 km grid was produced applying the SAS routine, G3GRID. This grid was plotted using the SAS routine GPLOT, where each intersection (x is east coordinate and y is north coordinate) was represented by its class's colour. At the borderland between mountain and water, as well as in areas with a small share of forest land, plot frequency is sparse and, thus, values at the intersections were discarded. A lower limit for the number of plots representing each value was introduced. The level of this limit varied depending on to which region the value belonged. This is because the number of plots is sparser in the north of Sweden than in the southern parts. The levels for the different regions received similar relations as the relations between numbers of plots for the different regions.