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

Predictive mapping of species - which methods are useful?

Predictive mapping is based on empirical relationsships between environmental factors – the predictors – and the biological variables of interest. These relationsships are quantified into mathematical models, which we use to predict and map biodiversity in areas not sampled for biological data. There are several methods for building the predicting models.

In principle we can distinguish between two types of situations; we either want to predict the distribution (presence vs absence) for example of vegetation, or we want to predict the abundance.

In the former situation we use classification models for the predictions, and in the latter regression models. Each of these situations require specific considerations with respect to data requirements but also to methods for model building, testing and prediction.

Methods for complex relationships
There exist a large (and growing) number of methods for building classification and regression models. Compared to traditional methods such as linear or logistic regression, most of the modern methods are highly flexible in the sense that they can fit data to non-linear relationships and they can deal with interactions among predictors (i.e. when the effect of one factor depends on another factor). In PREHAB we conclude that several of the tested techniques manage to do this in an adequate way.

Ensemble methods recommended
In PREHAB we have evaluated a wide (but not exhaustive) range of modelling techniques, based on different philosophies, in order to assess the general applicability for different types of response variables, sampling methods and geographical areas. This was done with the aim to give a general assessment of the practical use of different methods. However, it is important to stress that there are pros and cons of different techniques, and often the most sensible approach is to use ensemble methods (i.e. a range of methods) to evaluate the robustness of predictions.

Going from models to maps
The purpose of predictive models as evaluated within PREHAB is to provide a tool for mapping the distribution and abundance of benthic biodiversity which can be used for different purposes in coastal management and planning. Despite the sometimes static and seemingly deterministic impression given by such maps, it is important to remember that any map is a graphical representation of a model with a certain degree of uncertainty. Transferring models to maps require specific methods and maps need to be interpreted sensibly.

Description of methods

There are several statistical methods for determining the predictive model of the relationship between environmental predictor and the biodiversity you want to map. PREHAB has evaluated four of them.

Relationships among variables

Usually, biological data is complex. Relationships are often strongly nonlinear between the environmental variables we use as predictors and the biodiversity we want to map.

Measures of performance

The performance of models were tested using AUC, R2 and NMRSE.


PREHAB quality procedures

The modelling in PREHAB was done by different people using data from several different case-study areas but with a common set of routines.


© 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?