Vertical Relationships

Chi-Square Analysis

Landscapes are often defined by their vertical relationships. In this section, the covariance of soil texture (clay, loam or sandy loam) and land use category (grassland, cropland or forest) is analysed. This is done through a chi-square analysis.

From the original grid of 107 sample points, 74 points remain for the chi-square analysis after excluding points that fall in the Zenne, on infrastructure, or in smaller categories outside the selected soil texture and land use classes. This gives the following table of observed values:

GrasslandCroplandForestRow total
Clay22610
Sandy Loam1714940
Loam146424
Column total33221974

The expected values can then be calculated, resulting in the following values:

GrasslandCroplandForestRow total
Clay4.462.972.5710
Sandy Loam17.8411.8910.2740
Loam10.707.146.1624
Column total33221974

Unfortunately, the requirements for a chi-square analysis are not met, as three of the nine expected values are below 5. This means the analysis cannot proceed, and no conclusion can be drawn about the vertical relations between soil texture and land use category. A denser grid of sample points might help bring the expected values above 5, but that falls outside the scope of this landscape analysis.

Jaccard Similarity Measure

Given that no conclusion can be drawn about the covariance of land use and soil texture, a different metric might help explain land use. Since one of the defined landscape types is residential (the Hills), this analysis examines whether residential land use covaries with altitude.

To perform a Jaccard similarity measure, the sample points can be measured in a 2x2 grid. Residential use is defined as the “Private Property” category of the land use thematic map, and altitude is divided at 40 m above sea level. For this analysis, 93 of the original 107 sample points remain after excluding points that fall in the Zenne or on infrastructure.

ResidentialNon-residentialTotal
< 40 m56166
> 40 m101727
Total157893

Calculating the Jaccard score for each cell gives the following values:

ResidentialNon-residential
< 40 m0.070.73
> 40 m0.310.19

Two things can be said about this similarity measure: there is no association between residential land use and low altitude, and there is a strong association between non-residential land use and lower altitudes. At higher altitudes, the associations are very mild and not very meaningful.