Files associated to The Handbook of Herbivory quantification

Our research, based on an extensive dataset of tropical plants, provides a field guide for herbivory estimation that balances the advantages and shortfalls of these approaches. We evaluated the accuracy of different methods for measuring herbivory across a large dataset of phylogenetically diverse plants (71 plant species) with varying levels of natural and artificial leaf damage. Our study suggests a handbook and methodological guidelines to support quantitative estimates of herbivory, considering accuracy, precision, and time efficiency. We found that visual estimation methods tend to overestimate herbivory levels compared to digital methods. Conversely, deep-learning algorithms underestimated herbivory on leaf margins with natural damage but showed similar accuracy to ImageJ® and R software for artificially damaged leaves. Training in herbivory estimation significantly improved accuracy and reduced the time required for visual assessments. While image quality did not significantly affect the accuracy of herbivory estimates, factors such as the number of leaves per image, herbivory level, and leaf size significantly impacted measurement time. Our study highlights methodological biases and issues in herbivory quantification and suggests that these biases can be mitigated through standardized protocols, proper training, and the use of appropriate tools.