Assumptions and limitations of network models

This section will look in more detail at the modelling assumptions and limitations behind:

• species characteristics; and
• data and scale

Species characteristics and focal species

Different species have different dispersal capabilities. Some species can cover long distances to disperse whereas others may only move a few metres. What works as a habitat network for one species may not be ideal for another, so models often use a 'focal species' approach. Focal species are used in some models to represent the characteristics of a group of species - for example, the corn bunting has been used to represent 'farmland birds' in the Lowland Habitat Network project. Models are then based on the habitat area requirement and dispersal ability of the focal species. However, this information isn't available for many species, so assumptions and estimates have to be made, which may vary in accuracy. It should also be remembered that these characteristics are an 'average' for that species - any one individual plant or animal may behave very differently in reality.

In some cases, there isn't a single species which can be used in the design of a habitat network. In these cases, a 'generic focal species' is sometimes used as an alternative. This isn't a real plant or animal, it's a theoretical creation which represents the characteristics of a range of species. It might be described as an 'ancient woodland specialist with high area requirements and limited dispersal ability' and certain assumptions will be made about how it behaves. These can be used in the modelling process to produce a network design. However, these assumptions may not be reflected in the actual behaviour of the various species which are supposed to be represented by the model.

Data and scale

Data can be both dangerous and useful! Most models require information on the characteristics of the species under consideration. Usually this includes information such as its dispersal ability, the type of habitat it requires, its ability to move through other habitats and so on.  There isn't always sufficient real-world information available for habitat network models to work effectively.  For example, there may be limited data on species characteristics, or we may not know as much as we'd like about the quality of the existing habitat.  The model will only be as good as the data used to construct it and the questions we ask it to answer, so data quality can be a significant limitation.  For example, if a model is constructed with very limited habitat data, we might assume there aren't any patches of good habitat whereas, in reality, there may be several areas of excellent habitat. Many of the subsequent calculations made by the model will then be inaccurate.

When selecting data for modelling there are several things to consider:

• the question you're trying to answer - what is it that you're trying to model?;
• the scale that you want to model on;
• the accuracy of the data you have.

Habitat connectivity is species specific, so any modelling has to begin with species selection as a starting point. Producing network maps can only accurately answer the needs of a single species, or - more generally - a few species or a generic species.  For example, it's not possible to produce a meaningful map of habitat connectivity for both the chequered skipper butterfly and the raspberry plant as the two species have different characteristics.  However, as they're both associated with woodland edge habitats, it may be possible to look at connectivity for woodland edge species in general.

The scale at which you are assessing habitat networks also needs consideration. In general, the more detailed and small scale you want an analysis to be, the more detailed data you'll require.  Suitably detailed habitat data (eg Phase I habitat surveys) may not be available and if it is, it may be many years old and no longer accurate. However, if you're looking at connectivity for a common habitat over a large scale, land cover datasets at a fairly course resolution (such as that contained within OS Mastermap) may be sufficient for your purposes.