Reconstructing thicket clump formation using association rules analysis

Abstract

Aim

Vegetation change such as woody encroachment is characterised by changing species interactions, and processes such as competition and facilitation may be inferred from patterns of association between juveniles and mature individuals of different species. Our aim was to apply and evaluate association rules analysis (ARA), a rule-based data-mining technique more commonly known as market basket analysis, as a novel tool to examine the associations between woody species in different demographic stages along a thicket encroachment gradient.

Location

The research was conducted in a subtropical thicket–savanna mosaic (730 mm mean annual rainfall) in the Eastern Cape, South Africa.

Methods

We used a space-for-time substitution approach and sampled woody plants in different size classes at sites representing early, intermediate and late stages of encroachment. All individuals were recorded as occurring singly or in unique clumps. We used ARA to determine which associations between species in different size classes were common overall, and more common than expected, at each of the three encroachment stages.

Results

The most important association rules indicated that Vachellia karroo recruited singly and in large numbers in open grassland and, once mature, provided nucleation sites for a small suite of species dominated by Scutia myrtina. In the later stages, multiple diverse associations were found in increasingly large clumps.

Conclusions

Our sampling approach and ARA proved useful for characterising common species-size class associations, illuminating changing species interactions and recruitment patterns along a thicket clump formation sequence. In studies of vegetation change, ARA can complement multivariate analyses of species composition to reveal specific associations, and it can provide a less laborious alternative to point-pattern analysis for elucidating spatial associations.