379 research outputs found
Prioritizing the risk of plant pests by clustering methods : self-organising maps, k-means and hierarchical clustering
No full textFor greater preparedness, pest risk assessors are required to prioritise long lists of pest species with potential to establish and cause significant impact in an endangered area. Such prioritization is often qualitative, subjective, and sometimes biased, relying mostly on expert and stakeholder consultation. In recent years, cluster based analyses have been used to investigate regional pest species assemblages or pest profiles to indicate the risk of new organism establishment. Such an approach is based on the premise that the cooccurrence of well-known global invasive pest species in a region is not random, and that the pest species profile or assemblage integrates complex functional relationships that are difficult to tease apart. In other words, the assemblage can help identify and prioritise species that pose a threat in a target region. A computational intelligence method called a Kohonen self-organizing map (SOM), a type of artificial neural network, was the first clustering method applied to analyse assemblages of invasive pests. The SOM is a well known dimension reduction and visualization method especially useful for high dimensional data that more conventional clustering methods may not analyse suitably. Like all clustering algorithms, the SOM can give details of clusters that identify regions with similar pest assemblages, possible donor and recipient regions. More important, however SOM connection weights that result from the analysis can be used to rank the strength of association of each species within each regional assemblage. Species with high weights that are not already established in the target region are identified as high risk. However, the SOM analysis is only the first step in a process to assess risk to be used alongside or incorporated within other measures. Here we illustrate the application of SOM analyses in a range of contexts in invasive species risk assessment, and discuss other clustering methods such as k-means, hierarchical clustering and the incorporation of the SOM analysis into criteria based approaches to assess pest risk
Host Range Expansion of an Endemic Insect Herbivore is Associated With High Nitrogen and Low Fibre Content in Exotic Pasture Plants
No full text
Rowan Mark Emberson, B.For.Sc [Hons] (<i>Edinburgh)</i>, PhD (<i>McGill</i>), 21July 1941–7 October 2018
No full textRowan Mark Emberson, B.For.Sc [Hons] (Edinburgh), PhD (McGill), 21July 1941–7 October 201
Multi-Scenario Species Distribution Modeling
No full textCorrelative species distribution models (SDMs) are increasingly being used to predict suitable insect habitats. There is also much criticism of prediction discrepancies among different SDMs for the same species and the lack of effective communication about SDM prediction uncertainty. In this paper, we undertook a factorial study to investigate the effects of various modeling components (species-training-datasets, predictor variables, dimension-reduction methods, and model types) on the accuracy of SDM predictions, with the aim of identifying sources of discrepancy and uncertainty. We found that model type was the major factor causing variation in species-distribution predictions among the various modeling components tested. We also found that different combinations of modeling components could significantly increase or decrease the performance of a model. This result indicated the importance of keeping modeling components constant for comparing a given SDM result. With all modeling components, constant, machine-learning models seem to outperform other model types. We also found that, on average, the Hierarchical Non-Linear Principal Components Analysis dimension-reduction method improved model performance more than other methods tested. We also found that the widely used confusion-matrix-based model-performance indices such as the area under the receiving operating characteristic curve (AUC), sensitivity, and Kappa do not necessarily help select the best model from a set of models if variation in performance is not large. To conclude, model result discrepancies do not necessarily suggest lack of robustness in correlative modeling as they can also occur due to inappropriate selection of modeling components. In addition, more research on model performance evaluation is required for developing robust and sensitive model evaluation methods. Undertaking multi-scenario species-distribution modeling, where possible, is likely to mitigate errors arising from inappropriate modeling components selection, and provide end users with better information on the resulting model prediction uncertainty
Improving cluster-based methods for investigating potential for insect pest species establishment: region-specific risk factors
No full textExisting cluster-based methods for investigating insect species assemblages or profiles of a region to indicate the risk of new insect pest invasion have a major limitation in that they assign the same species risk factors to each region in a cluster. Clearly regions assigned to the same cluster have different degrees of similarity with respect to their species profile or assemblage. This study addresses this concern by applying weighting factors to the cluster elements used to calculate regional risk factors, thereby producing region-specific risk factors. Using a database of the global distribution of crop insect pest species, we found that we were able to produce highly differentiated region-specific risk factors for insect pests. We did this by weighting cluster elements by their Euclidean distance from the target region. Using this approach meant that risk weightings were derived that were more realistic, as they were specific to the pest profile or species assemblage of each region. This weighting method provides an improved tool for estimating the potential invasion risk posed by exotic species given that they have an opportunity to establish in a target region
Using artificial neural networks to predict the distribution of bacterial crop diseases from biotic and abiotic factors
No full textConstructing accurate computational global distribution models is an important first step towards the understanding of bacterial crop diseases and can lead to insights into the biology of disease-causing bacteria species. We constructed artificial neural network models of the geographic distribution of six bacterial diseases of crop plants. These ANN modelled the distribution of these species from regional climatic factors and from regional assemblages of host crop plants. Multiple ANN were combined into ensembles using statistical methods. Tandem ANN, where an ANN combined the outputs of individual ANN, were also investigated. We found that for all but one species, superior accuracies were attained by methods that combined biotic and abiotic factors. These combinations were produced by both ensemble and cascaded ANN. This shows that firstly, ANN are able to model the geographic distribution of bacterial crop diseases, and secondly, that combining abiotic and biotic factors is necessary to achieve high modelling accuracies. The work reported in this paper therefore provides a basis for constructing models of the distribution of bacterial crop diseases
Using artificial neural networks to predict the distribution of bacterial crop diseases from biotic and abiotic factors
No full textAbstract Constructing accurate computational global distribution models is an important first step towards the understanding of bacterial crop diseases and can lead to insights into the biology of disease-causing bacteria species. We constructed artificial neural network models of the geographic distribution of six bacterial diseases of crop plants. These ANN modelled the distribution of these species from regional climatic factors and from regional assemblages of host crop plants. Multiple ANN were combined into ensembles using statistical methods. Tandem ANN, where an ANN combined the outputs of individual ANN, were also investigated. We found that for all but one species, superior accuracies were attained by methods that combined biotic and abiotic factors. These combinations were produced by both ensemble and cascaded ANN. This shows that firstly, ANN are able to model the geographic distribution of bacterial crop diseases, and secondly, that combining abiotic and biotic factors is necessary to achieve high modelling accuracies. The work reported in this paper therefore provides a basis for constructing models of the distribution of bacterial crop diseases
- …
