Till startsida
To content Read more about how we use cookies on gu.se

Relationships among variables

Predictive modelling is based on relationsships between biological and physical data. These relationships are usually complicated in many ways. Two aspects are particularly important for the selection of modelling method: the form of relationsships (e.g. linear or non-linear) and the way different environmental factors interact.

Linear or non-linear?
Most species-environment relationsships can are better described as non-linear rather than linear! A good example this is the depth zonation of the main macroalgae groups in the Baltic Sea and elsewhere (Fig. 1). The highest abundance of green algae is typically observed in the shallow littoral zone. The red algae on the other hand is most abundant in the sublittoral zone. This pattern of distribution pattern is explained mainly due to algae competition and adaptation to the light climate. Both types of algae are non-linearly related to depth, i.e. the change in abundance is not constant across the whole depth range. The two types of algae, however, show different relationships to depth; one is monotonically decreasing while the other has maximum abundance at an intermediate depth. The techniques tested within PREHAB (i.e. GAM, MARS, RF and MaxEnt) are all capable of dealing with such patterns in an adequate way.


Figure 1. Abundance and occurrence of benthic species are often non-linearly related to envvironmental factors. Modelling methods need to account for this! The example above is an artificial illustration.


Interactive effects?
Benthic species and habitats are affected by numerous environmental factors and biological processes. All of these factors operate simultaneously and varies among times and places. In order to be sufficiently precise and powerful, the methods used for predictive models need to account for the most important of these factors, often in non-linear ways. One further complication is that the effect of one factor may depend on another factor. For example, the non-linear effect of depth on the cover of green algae as described in figure 1, may depend on whether it is an exposed or sheltered area. In exposed areas the cover of algae may be reduced by wave exposure, and therefore much less algae are found in shallow exposed areas compared to shallow sheltered areas (Fig. 2). These types of effects, often involving more than two factors, are called interactive and they are often important for the spatial distribution of many benthic species. Consequently, the techniques used for modelling need to be able to handle also these relationsships. This is true for the methods tested within PREHAB (i.e. GAM, MARS, RF and MaxEnt).


Figure 2. The importance of one environmental factor depends on other factors. For example the relationship between the cover of green algae is different between sheltered and exposed areas. Modelling methods need to account for this! The example above is an artificial illustration.



© University of Gothenburg, Sweden Box 100, S-405 30 Gothenburg
Phone +46 31-786 0000, About the website

| Map

The University of Gothenburg uses cookies to provide you with the best possible user experience. By continuing on this website, you approve of our use of cookies.  What are cookies?