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

Factors affecting predictability

Generally, the success of predictive modelling depends on several interconnected factors. The "predictability" of a species or habitat depends to a large extent on the characteristiques of the species in question, and the quality and quantity of available data.

Seasonal species are less reliable
Species that stay in one place from year to year, e.g. blue mussels and bladder wrack, produce more consistent information over time. This reduces the stochastic variability in the dataset used to build models. Seasonal and irregular patterns in the species presence or detactability at a site often result in unreliable models. Annual filamentous algae, for example have seasonal life cycles with changes in community throughout the year (Kraufvelin et al). The datasets used for modelling are usually a compilation from many sources collected in a number of years and different times of year, which introduces measurement based error into the data.

Choosy species are more predictable
The number of observations largely define how well a model can distinguish places that are well suited for a species or group from those that are not. The success of models hence strongly depends on the ability of the environmental factors we have available to represent the distribution of our species of interest. Some species are very tightly coupled to different environmental factors such as salinity, exposure or light regime, while others are not. Therefore, models predicting, for example, the occurence of eelgrass (Zostera marina), usually have a high accuracy in the Northern Baltic Sea, due to the very specific environmental requirements of this species.

Stochastic factors is a limit
A source of error is introduced when random unpredictable factors have a strong effect on the species of interest, which we are not able to map and include in models. As an example, perennial slow growing brown algae such as Furcellaria and Fucus, are frequently missing from environmentally suitable sites due to their susceptible to random disturbances and overgrowth by faster growing algae, and they do not recolonise habitats rapidly. In general, species whos distributions are strongly affected by factors that cannot easily be included in models, such as grazing, predation, recruitment and stochastic factors, will produce poorer performing models. Note however, that the importance of such factors are likely to differ among spatial and temporal scales.

Patchy habitat requires data with higher resolution
In many coastal areas the seafloor consists of a fine-scale mosaic of naturallypatchy habitats with highly variable topography and substrate (Kaskela et al., (Accepted); Kautsky and Maarel, 1990), which will affect distribution in a stochastic manner making it more difficult to predict species presence, unless the patchiness and stochastic processes can be included in the model. The way models are able to describe and predict distributions of species in such a variable ecosystem, and the type of generalisation that can be made, is very much dependent on the scale of the study and the resolution of available data (Austin, 2007; Elith and Leathwick, 2009; Guisan and Thuiller, 2005). The availability of the relevant environmental predictors as GIS layers at the appropriate scale of variability is especially important.

Broad taxonomic groups are harder to predict
Often species are grouped into larger groups (such as filamentous algae, red algae or aquatic plants), due to either the difficulty of identifying individual species or the lack of observations on individual species. The tradeoff is the specificity of needs the individual species had. By combining e.g. all red algae together, the group becomes very heterogenous with regards to its ecological attributes and therefore, environmental requirements. Consequently, on theoretical grounds, we can expect models of broader groups to be less precise than those of individual species.

Filamentous algae - an example
As an example, in the Nothern Baltic the filamentous green algae in the shallow sublittoral consists mostly of the proliferate and easy to identifyCladophora glomerata, which is present throughout the open water season (Kiirikki and Lehvo, 1997). In contrast, filamentous red algae as a group consist of many different species, with different seasonal cycles and also show more interannual variability within sites (Kiirikki and Lehvo, 1997). As a result, filamentous red algae as a group yields poorer models than filamentous green algae. The pros and cons of modelling groups or species have to be weighed up against each other, and one should always bear in mind the ecological similarities and differences of species included in such groups.




© 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?