Species distribution modelling – an increasingly powerful tool

species distribution modelling

GIS can be used to express the results of models as habitat suitability maps

 

Distribution modelling is increasingly being used in conservation science, bringing together field data, maps of environmental variables and GIS to create a powerful tool.

Linking records and maps

Species distribution modelling is a technique for linking spatially referenced records of species occurrence – for example collected during appropriately designed field-based biodiversity monitoring programmes – with maps of environmental variables such as elevation, climate, vegetation or human disturbance, in order to create a statistical model of the relationship between a species and its environment, ie the species realised ecological niche. GIS can then be used to express the results of models as habitat suitability maps across a desired spatial extent.

The output habitat suitability maps are also a powerful tool to communicate conservation messages to non-scientists

The species records may be a set of presences only or a set of presence and absence records, depending on the detectability of the species and sampling method. Almost any environmental variables can be used in a distribution model, although it is normal to select a restricted set of variables at a particular spatial scale based on a working hypothesis about the aspects of the environment which may be important to the focal species.

Refining models

A wide range of statistical approaches have been developed for fitting distribution models including various types of regression models, machine learning and classification methods. Regression approaches such as GLM are simple to implement; however, more complex information theoretic approaches, especially maximum entropy, have proved to be very powerful. Once a model has been built and refined, it is critically important to validate the model on a subset of data which was not used in model construction, in order to objectively assess how well the model performs.

Applications

Distribution models have a wide range of applications in conservation science including predicting potential impacts of climate change, supporting conservation prioritisation and reserve selection, predicting the spread of invasive species and guiding field surveys to discover new species.

The output habitat suitability maps are also a powerful tool to communicate conservation messages to non-scientists.

Written by

Twitter Digg Delicious Stumbleupon Technorati Facebook Email

3 Responses to “Species distribution modelling – an increasingly powerful tool”

  1. A gem of a paper on the application of GIS-based ecological niche modelling (ENM) and predictive models of species distribution (and perhaps the most interesting paper I read in 2010) is:

    Lozier, J.D., Aniello, P., and Hickerson, M.J., 2009. Predicting the distribution of Sasquatch in western North America: anything goes with ecological niche modelling. Journal of Biogeography 36: 1623-1627.

    Dr. Long notes, “The species records may be a set of presences only or a set of presence and absence records, depending on the detectability of the species and sampling method.” The Lozier et al. (2009) paper explores extensively the relationship between the quality of the species records one uses in building a distribution model, and what that model ends up looking like. The Lozier et al. (2009) paper is a wonderful cautionary tale, effectively demonstrating how what seem to be sound predictive models of species distribution can be produced from poor-quality, or questionable, species records. When it comes to species distribution modelling, a version of an old adage would appear to apply — what comes out in one’s species distribution model is only as good as what goes in. Dr. Long’s point on the need to refine and assess models (i.e., “Once a model has been built and refined, it is critically important to validate the model on a subset of data which was not used in model construction, in order to objectively assess how well the model performs.”) is, thus, well-taken.

  2. Colin Maycock Reply May 2011 at 12:41 am

    When it comes to modeling tropical plant distribution a factor often not considered is the quality of the environmental layers used to develop the enms. There is a large body of research that shows that many tropical plants are habitat/soil specific (See Clark et al. 1998, Paoli et al. 2006, Swaine 1996), and yet when it comes using soil information in the ecological niche modeling many studies are using very low resolution soil layers i.e. grid cells of 10000 hectares. Even the HWSD (grid cells of 100 ha) is probably of not a fine enough resolution, as it is derived largely from the 1:5,000,000 scale FAO-UNESCO Digital Soil Map of the World. For our niche modeling work in Sabah, we are using digitized version of the 1:250,000 The Soils of Sabah maps.
    My only addition to Dr. Long’s point that “Once a model has been built and refined, it is critically important to validate the model on a subset of data which was not used in model construction, in order to objectively assess how well the model performs.” is that this testing and model validation should be done on new GPSed locality data and not on post-collection georeferenced data. Although it has to be acknowledged that this will not always possible.

  3. You can earn some extra $$$ from your website, i see several opportunities
    here. You should search in google for:
    Yoogurn’s money making

Leave a Reply