IMIS | Flanders Marine Institute

Flanders Marine Institute

Platform for marine research

In:

IMIS

Publications | Institutes | Persons | Datasets | Projects | Maps
[ report an error in this record ]basket (1): add | show Printer-friendly version

one publication added to basket [124196]
Where is the worm? Predictive modelling of the habitat preferences of the tube-building polychaete Lanice conchilega
Willems, W.; Goethals, P.; Van den Eynde, D.; Van Hoey, G.; Van Lancker, V.; Verfaillie, E.; Vincx, M.; Degraer, S. (2008). Where is the worm? Predictive modelling of the habitat preferences of the tube-building polychaete Lanice conchilega. Ecol. Model. 212(1-2): 74-79. dx.doi.org/10.1016/j.ecolmodel.2007.10.017
In: Ecological Modelling. Elsevier: Amsterdam. ISSN 0304-3800, more
Peer reviewed article

Available in Authors 

Keywords
    Habitat selection; Lanice conchilega (Pallas, 1766) [WoRMS]; Lanice conchilega (Pallas, 1766) [WoRMS]; Polychaeta [WoRMS]; Marine
Author keywords
    Lanice conchilega; Polychaeta; Habitat preference; Generalized linear models (GLM); Artificial neural networks (ANN)

Authors  Top 
  • Willems, W., more
  • Goethals, P., more
  • Van den Eynde, D., more
  • Van Hoey, G., more

Abstract
    Grab samples to monitor the distribution of marine macrobenthic species (animals >1 mm, living in the sand) are time consuming and give only point based information. If the habitat preference of a species can be modelled, the spatial distribution can be predicted on a full coverage scale from the environmental variables. The modelling techniques Generalized Linear Models (GLM) and Artificial Neural Networks (ANN) were compared in their ability to predict the occurrence of Lanice conchilega, a common tube-building polychaete along the North-western European coastline. Although several types of environmental variables were in the data set (granulometric, currents, nutrients) only three granulometric variables were used in the final models (median grain-size, % mud and % coarse fraction). ANN slightly outperformed GLM for a number of performance indicators (% correct predictions, specificity and sensitivity), but the GLM were more robust in the crossvalidation procedure.

 Top | Authors