Comparative efficiency and accuracy of point intercept and exhaustive count scoring methods for the extraction of ecological data on mobile species from underwater imagery in western Antarctica

Abstract

In recent years, underwater imagery has become a popular tool to extract ecologically relevant information from benthos in order to monitor and predict future trends in marine ecosystems. However, while images are collected quickly, their processing is laborious and guidelines are sometimes not well defined to provide ecological accuracy. Therefore, it is necessary to find a trade-off between efficiency in image scoring and accuracy in the extracted ecological data. Thus, using the right annotation protocol is critical to meet accuracy and efficiency constraints. Several annotation methods have been developed but their reliability is mostly assessed for sessile benthic species. However, mobile benthic species play a key role in community pattern and should be also be considered in monitoring programs. The aim of this study was to compare exhaustive count and point intercept methods based on their accuracy and efficiency to estimate mobile species abundance and detect biodiversity patterns. Firstly we looked at abundance at broad taxonomic level: (i) we investigated the relationship between the abundance estimates extracted from each method, then (ii) we looked into abundance threshold needed to be reached to detect the presence of mobile species with point intercept method, and (iii) the influence of size and occurrence of studied organisms on the relationship between the two abundance estimates was analyzed. Then, we compared biodiversity pattern detected by each method: (i) taxonomic richness and (ii) community structure. Results emphasized the high level of accuracy of the exhaustive count method to estimate species abundance as well as detecting the biodiversity pattern. Whereas, point intercept method was unreliable to provide accurate abundance estimate for most of the mobile species, to detect the presence of species without a very high level of sampling and to capture the whole species richness. However, both methods showed common community patterns at larger scale (100s \si{m^2}). Occurrence and size of the species has an influence on the detection by the point intercept method: small and rare organisms are hardly detected. An effective mean to improve the accuracy of the point intercept method would be to increase the point density, however it would reduce significantly the method efficiency. The sampling rate would tend towards the one of the exhaustive count method, diminishing the interest of the point intercept method. Therefore, point intercept method is not an effective strategy for sampling small mobile taxa, a presence/absence sampling could be a solution to increase efficiency of the exhaustive count method,