Reef Life Survey is a citizen science foundation that is producing large amounts of seafloor image data. These images are collected alongside standardised observations of fish and invertebrates for over 15,000 surveys on shallow coral and rocky reefs globally. The imagery is publicly accessible through UMI (an IMOS sub-facility) via Squidle+, which provides tools to annotate and process images by geographically dispersed users and algorithms.
Advances in statistical modelling and improved computation resources are enabling the use of larger and more complex datasets to answer critical questions about marine ecosystems. This presents a unique challenge of how to efficiently produce consistent and quality annotations, particularly when utilizing volunteers.
This talk will showcase how these datasets are influencing data-driven management, including State of Environment reporting and other high-profile research. While we explore some of the challenges arising from variable quality of images, we highlight unique opportunities through coverage of huge spatial and increasing temporal scales, and co-located fish and invertebrate observations. Results from preliminary work on how RLS and SQ+ are improving data quality through detecting annotator bias for the current RLS classification scheme are presented. Further opportunities to optimise use of the annotation data will also be discussed.