Evaluation of Bayesian networks for modelling habitat suitability and management of a protected area. Douglas, S. J. and Newton, A. C. Journal for Nature Conservation, 22(3):235--246, June, 2014.
Evaluation of Bayesian networks for modelling habitat suitability and management of a protected area [link]Paper  doi  abstract   bibtex   
To be effective, management of protected areas should be based on the best available evidence, includingthe scientific literature and expert knowledge. However, lack of such evidence in a suitable form to support decision-making may hinder effective management. Here we examine the use of Bayesian networksto support the management of protected areas, through the development of habitat suitability modelsfor eight species of conservation concern. Bayesian networks were constructed on the basis of the sci-entific literature and expert knowledge, and were then tested using results from a field survey. Modelsof all species demonstrated very high discrimination between presence and absence sites, as indicatedby AUC values \textgreater0.8, with values \textgreater0.9 obtained for four species, and Kappa values in the range of 0.4–0.9.The Bayesian networks were then used to examine the impact of different management interventionson habitat suitability of each species, including tree cutting, grazing and burning. Species differed interms of their sensitivity to different management interventions, and model output provided evidence ofboth negative and positive interactions between types of intervention. These results highlight the trade-offs that must often be made when undertaking conservation management, and demonstrate the value ofBayesian networks in helping to make such trade-offs explicit. The identification of management impactsthrough analysis of available evidence also demonstrates the value of Bayesian networks for supportingevidence-based approaches to protected area management.
@article{douglas_evaluation_2014,
	title = {Evaluation of {Bayesian} networks for modelling habitat suitability and management of a protected area},
	volume = {22},
	issn = {16171381},
	url = {http://linkinghub.elsevier.com/retrieve/pii/S1617138114000053},
	doi = {10.1016/j.jnc.2014.01.004},
	abstract = {To be effective, management of protected areas should be based on the best available evidence, includingthe scientific literature and expert knowledge. However, lack of such evidence in a suitable form to support decision-making may hinder effective management. Here we examine the use of Bayesian networksto support the management of protected areas, through the development of habitat suitability modelsfor eight species of conservation concern. Bayesian networks were constructed on the basis of the sci-entific literature and expert knowledge, and were then tested using results from a field survey. Modelsof all species demonstrated very high discrimination between presence and absence sites, as indicatedby AUC values {\textgreater}0.8, with values {\textgreater}0.9 obtained for four species, and Kappa values in the range of 0.4–0.9.The Bayesian networks were then used to examine the impact of different management interventionson habitat suitability of each species, including tree cutting, grazing and burning. Species differed interms of their sensitivity to different management interventions, and model output provided evidence ofboth negative and positive interactions between types of intervention. These results highlight the trade-offs that must often be made when undertaking conservation management, and demonstrate the value ofBayesian networks in helping to make such trade-offs explicit. The identification of management impactsthrough analysis of available evidence also demonstrates the value of Bayesian networks for supportingevidence-based approaches to protected area management.},
	language = {en},
	number = {3},
	urldate = {2015-04-06TZ},
	journal = {Journal for Nature Conservation},
	author = {Douglas, Sarah J. and Newton, Adrian C.},
	month = jun,
	year = {2014},
	pages = {235--246}
}
Downloads: 0