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

Predictive mapping of species and habitats - what can be achieved?

PREHAB used a number of techniques to model and predict abundance and distribution of a large range of vegetation, fish and benthic invertebrates from different parts of the Baltic Sea (modelling methods and assessment criteria are described here). By summarising findings from more than 500 model runs of more than 60 biological variables, this section gives an overall idea about what a manager or researcher can realistically expect from empirical models in terms of predictive power and precision.

Figure 1. Performance (mean±se) of models of fish, invertebrates and vegetation using different criteria. Better performance is characterised by large r2 and AUC, but small nrmse.

Models of distribution were assessed in terms of predictive power (r2) and precision (nrmse). Most models explained significant amounts of variability. Models of invertebrates and vegetation explained more than 30% of the variability while the corresponding value for fish was just above 20%. In terms of precision the average deviation was approximately 15% of the range (e.g. nrmse≈0.15).

Models of distribution were assessed using the AUC-metric. The average performance for all types of benthic organisms can be characterised as ”good” (i.e. AUC>0.8), but for vegetation it can even be considered ”excellent” (AUC>0.9).

In summary, these results suggest that models based on empirical relationsships can be used to predict abundance and distribution of benthic species with variable success. There appear to be some differences among fish, invertebrates and vegetation but the variability among species is sometimes large. While ecological factors may affect the predicatbility of different species (discussed here), one strong message from this evaluation is that the collection of biological data and the choice of predictors is essential for future modelling and mapping. Therfore it is our firm belief that the advice given under these sections may substantially improve models beyound what is observed here using existing data.

Detailed information on models of fish, invertebrates and vegetation are found in the links to the left

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.



Modelled species and groups

Modelled species and groups from PREHAB four case-study areas are listed here!

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

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?