PARENT SESSION
Poster Session 34: Herbivory
Thursday, August 11, 5:00 PM - 6:30 PM, Exhibit Hall 220 A-E, Level 2, Palais des congrès de Montréal
Estimating herbivory: A comparison of digital quantitative estimation and subjective classification of leaf area and how best to deal with zero inflated data.
Cicconetti, Gregory*,1, James, Kendra1, Niesenbaum, Richard1, 1 Muhlenberg College, Allentown, PA, USA
ABSTRACT- Although the focus of many ecologists on herbivory necessitates estimation of leaf area removed and analyses of this trait with respect to other variables, methods employed to do this have varied greatly. In recent years, quantifying leaf area lost to herbivores via computers employing supervised classification algorithms on digital images has become more commonplace. While the added precision offered is attractive, operator subjectivity is not eliminated and time-intensive digital quantification requires training individuals in specialized software. Subjectively classifying the lost leaf area into categories (0%, 0-5%, 5-10%, etc.) does not offer the same precision, but there are advantages. Leaves can be quickly classified in the field, allowing for a larger sample size without the need for sophisticated software or a camera. We present results of a comparison of approaches using leaf area removal in Lindera benzoin as a model system. We show that subjective classification can yield statistical conclusions similar to the added precision of digital quantification without the cost and labor associated with the former. We also offer statistical remedies for zero-inflated data that are typical of studies on relatively rare events such has damage to an individual leaf. For the subjective classification, categorical data analysis is employed. For quantitative measures of leaf area, we compare incident rates of herbivory and area lost among those that experienced herbivory following a log transformation.
Key words: herbivory, estimation, supervised classification, statistics