Classifying 3-D Models of Coral Reefs Using Structure-From-Motion and Multi-View Semantic Segmentation

Benthic quadrat surveys using 2-D images are one of the most common methods of quantifying the composition of coral reef communities, but they and other methods fail to assess changes in species composition as a 3-dimensional system, arguably one of the most important attributes in foundational systems. Structure-from-motion (SfM) algorithms that utilize images collected from various viewpoints to form an accurate 3-D model have become more common among ecologists in recent years. However, there exist few efficient methods that can classify portions of the 3-D model to specific ecological functional groups. This lack of granularity makes it more difficult to identify the class category responsible for changes in the structure of coral reef communities. We present a novel method that can efficiently provide semantic labels of functional groups to 3-D reconstructed models created from commonly used SfM software (i.e., Agisoft Metashape) using fully convolutional networks (FCNs). Unlike other methods, ours involves creating dense labels for each of the images used in the 3-D reconstruction and then reusing the projection matrices created during the SfM process to project semantic labels onto either the point cloud or mesh to create fully classified versions. When quantitatively validating the classification results we found that this method is capable of accurately projecting semantic labels from image-space to model-space with scores as high as 91% pixel accuracy. Furthermore, because each image only needs to be provided with a single set of dense labels this method scales linearly making it useful for large areas or high resolution-models. Although SfM has become widely adopted by ecologists, deep learning presents a steep learning curve for many. To ensure repeatability and ease-of-use, we provide a comprehensive workflow with detailed instructions and open-sourced the programming code to assist others in replicating our methodology. Our method will allow researchers to assess precise changes in 3-D community composition of reef habitats in an entirely novel way, providing more insight into changes in ecological paradigms, such as those that occur during coral-algae shifts.

Implications of 2D versus 3D surveys to measure the abundance and composition of benthic coral reef communities

A paramount challenge in coral reef ecology is to estimate the abundance and composition of the communities residing in such complex ecosystems. Traditional 2D projected surface cover estimates neglect the 3D structure of reefs and reef organisms, overlook communities residing in cryptic reef habitats (e.g., overhangs, cavities), and thus may fail to represent biomass estimates needed to assess trophic ecology and reef function. Here, we surveyed the 3D surface cover, biovolume, and biomass (i.e., ash-free dry weight) of all major benthic taxa on 12 coral reef stations on the island of Curaçao (Southern Caribbean) using structure-from-motion photogrammetry, coral point counts, in situ measurements, and elemental analysis. We then compared our 3D benthic community estimates to corresponding estimates of traditional 2D projected surface cover to explore the differences in benthic community composition using different metrics. Overall, 2D cover was dominated (52 ± 2%, mean ± SE) by non-calcifying phototrophs (macroalgae, turf algae, benthic cyanobacterial mats), but their contribution to total reef biomass was minor (3.2 ± 0.6%). In contrast, coral cover (32 ± 2%) more closely resembled coral biomass (27 ± 6%). The relative contribution of erect organisms, such as gorgonians and massive sponges, to 2D cover was twofold and 11-fold lower, respectively, than their contribution to reef biomass. Cryptic surface area (3.3 ± 0.2 m2 m−2planar reef) comprised half of the total reef substrate, rendering two thirds of coralline algae and almost all encrusting sponges (99.8%) undetected in traditional assessments. Yet, encrusting sponges dominated reef biomass (35 ± 18%). Based on our quantification of exposed and cryptic reef communities using different metrics, we suggest adjustments to current monitoring approaches and highlight ramifications for evaluating the ecological contributions of different taxa to overall reef function. To this end, our metric conversions can complement other benthic assessments to generate non-invasive estimates of the biovolume, biomass, and elemental composition (i.e., standing stocks of organic carbon and nitrogen) of Caribbean coral reef communities.

Beyond coral-algal regimes: high taxonomic resolution surveys and trait-based analyses reveal multiple benthic regimes

As coral reef communities change and reorganise in response to anthropogenic and climate disturbances, there is a growing need of detecting and understanding the different emerging species regimes and their contribution to key ecosystem processes. Using a case study on coral reefs at the epicentre of tropical marine biodiversity (North Sulawesi), we explored how application of different biodiversity approaches (i.e. use of major taxonomic categories, high taxonomic resolution categories and trait-based approaches) affects the detection of distinct fish and benthic community assemblages. Our results show that using major categories (family level or above) to study coral reef communities fails to identify distinct regimes. We also show that for detection of different benthic regimes, especially communities dominated by non-coral organisms, monitoring of only scleractinian coral communities is insufficient, and that all types of benthic organisms (e.g. sponges, ascidians, soft corals, algae etc.) need to be considered. We have implemented for the first time, the use of a trait-based approach to study the functional diversity of whole coral reef benthic assemblages, which allowed us to detect five different community regimes, only one of which was dominated by scleractinian corals. We circumvented the challenge that for some benthic groups (e.g. sponges, ascidians or some soft corals) visual identification up to the species level is not possible, by identifying and categorising traits that can be applied to groups of similar organisms instead of specific species. Furthermore, by the parallel study of benthic and fish communities we provide new insights into key processes and functions that might dominate or be compromised in the different community regimes.

Early trajectories of benthic coral reef communities following the 2015/16 coral bleaching event at remote Aldabra Atoll, Seychelles

Documenting post-bleaching trajectories of coral reef communities is crucial to understand their resilience to climate change. We investigated reef community changes following the 2015/16 bleaching event at Aldabra Atoll, where direct human impact is minimal. We combined benthic data collected pre- (2014) and post-bleaching (2016–2019) at 12 sites across three locations (lagoon, 2 m depth; seaward west and east, 5 and 15 m depth) with water temperature measurements. While seaward reefs experienced relative hard coral reductions of 51–62%, lagoonal coral loss was lower (− 34%), probably due to three-fold higher daily water temperature variability there. Between 2016 and 2019, hard coral cover did not change on deep reefs which remained dominated by turf algae and Halimeda, but absolute cover on shallow reefs increased annually by 1.3% (east), 2.3% (west) and 3.0% (lagoon), reaching, respectively, 54%, 68% and 93% of the pre-bleaching cover in 2019. Full recovery at the shallow seaward locations may take at least five more years, but remains uncertain for the deeper reefs. The expected increase in frequency and severity of coral bleaching events is likely to make even rapid recovery as observed in Aldabra’s lagoon too slow to prevent long-term reef degradation, even at remote sites.