Analyzing the occurrence of an invasive aquatic fern in wetland using data-driven and multivariate techniques

In the present study, the data-driven (classification trees and support vector machines) and multivariate techniques (principal component analysis and discriminant analysis) were applied to study the habitat preferences of an invasive aquatic fern (Azolla filiculoides) in the Selkeh Wildlife Refuge (a protected area in Anzali wetland, northern Iran). The applied database consisted of measurements from seven different sampling sites in the protected area over the study period 2007–2008. The cover percentage of the exotic fern was modelled based on various wetland characteristics. The predictive performances of the both data-driven methods were assessed based on the percentage of Correctly Classified Instances and Cohen’s kappa statistics. The results of the Paired Student’s t-test (p < 0.01) showed that SVMs outperformed the CTs and thus yielded more reliable prediction than the CTs. All data mining and multivariate techniques showed that both physical-habitat and water quality variables (in particular some nutrients) might affect the habitat requirements of A. filiculoides in the wetland.