Broadscale reconnaissance of coral reefs from citizen science and deep learning

Abstract

Coral reef managers require various forms of data. While monitoring is typically the preserve of scientists, larger scale reconnaissance data that can be used to inform spatial decisions does not usually require such precise measurement. There is an increasing need to collect such broadscale, up-to-date environmental data at massive scale to prioritise limited conservation resources in the face of global disturbances. Citizen science combined with novel technology presents an opportunity to achieve data collection at the required scale, but the accuracy and feasibility of new tools must be assessed. Here we show that a citizen science program that collects seascape images and analyses them using a combination of deep learning and online citizen scientists can produce accurate benthic cover estimates of key coral groups. The deep learning and citizen scientist analysis methods had different but complementary strengths depending on coral category. When the best performing analysis method was used for each category in all images, mean estimates from 8086 images of percent benthic cover of branching Acropora, plating Acropora, and massive-form coral were [~]99% accurate compared to expert assessment of the same images, and >95% accurate at all coral cover ranges tested. The effort to achieve 95% accuracy at a site - our ecologically relevant target based on the accuracy of other tools - was attainable based on citizen scientist involvement in pilot years of the program, with 18-80 images needed depending on coral type and reef state. Power analyses showed that sampling up to 114 images per site was needed to detect a 10% absolute difference in coral cover per category (power = 0.8), accounting for natural heterogeneity. However, the benthic cover of all other coral groups as a single category could only be estimated with 95% accuracy at 60% of survey sites and for images with 10-30% coral cover. Disaggregating this other coral group into more distinct coral categories may improve accuracy. Overall, citizen science can provide an accuracy that is acceptable for many end-users for select coral morphologies. Such a combination of emerging technology and citizen science presents an attainable tool for collecting inexpensive, widespread reconnaissance data of coral reefs that can complement higher resolution survey programs or be an accessible tool for resource-poor locations.