A PCA-based modelling technique for predicting environmental suitability for organisms from presence records

摘要: . We present a correlative modelling technique that uses locality records (associated with species presence) and set of predictor variables to produce statistically justifiable probability response surface for target species. The indicates the suitability each grid cell in map terms suite variables. constructs hyperspace using principal component axes derived from components analysis performed on training dataset. dataset comprises values associated localities where has been recorded as present. origin this is taken characterize centre niche organism. All (grid-cells) region are then fitted into at these (the prediction dataset). Euclidean distance any gives measure ‘centrality’ hyperspace. These distances used derive region. was applied bioclimatic data predict three alien invasive plant (Lantana camara L., Ricinus communis L. Solanum mauritianum Scop.) South Africa, Lesotho Swaziland. models were tested against independent test by calculating area under curve (AUC) receiver operator characteristic (ROC) curves kappa statistics. There good agreement between records. pre-processing climatic variable reduce deleterious effects multicollinearity, use stopping rules prevent overfitting important aspects process.