An automated approach to characterizing Guam’s reef communities

Matthew Mills

doi:10.25907/00761

The need to further the understanding of tropical reefs and their many associated complex processes is as pressing as it has ever been. Tropical reefs, which are synonymous with biodiversity, are being threatened by an number of chronic and acute stressors worldwide, and the many reefs surrounding Guam are no different. This study took a multidisciplinary approach to characterizing CCRA diversity, structure, and benthic composition of Guam’s reefs. This study provided the first report of the crustose calcifying red algal (CCRA) genus Ramicrusta for Guam and the Mariana Archipelago. Four Ramicrusta species were described and an additional two were reported from Guam using a combination of thorough anatomical observation and DNA-based identification. In addition, the CCRA diversity of Guam was revised following a large DNA barcoding effort. DNA sequences obtained from 492 CCRA specimens revealed 154 putative CCRA species, more than six times what had been previously reported for Guam, emphasizing the value of small tropical islands as hotspots of marine biodiversity. This study was also the first to quantify the structural complexity of eight reefs around Guam using Structure-from-Motion photogrammetry. The reefs were found to possess high structural complexity (average Rugosity = 3.95; slope = 45.23 degrees), which could be of great interest to reef managers in the future. Further analysis of transects revealed benthic diversity to be negatively correlated with slope (Pr(>Chi) = 0.039) and standard curvature (Pr(>Chi) = 0.017), while coral cover was negatively correlated with depth standard deviation (Pr(>|t| = 0.023) and positively correlated with slope (Pr(>|t| = <0.001). The shift in benthic community at a long-term monitoring site was also reported. Photoquadrat surveys revealed the almost complete mortality of the habitat structuring staghorn Acropora abrotanoides (from 10.1 ± 3.7% to 0.2 ± 0.3% cover) and other scleractinian corals (from 28.7 ± 6.5% to 6.9 ± 2.6% cover), along with a substantial increase in turf algae/cyanobacteria (from 52.1 ± 3.8% to 81.1 ± 2.4% cover). The community estimates generated from 2019 surveys were also predicted using hyperspectral survey data, suggesting that this methodology could be a viable alternative to traditional reef monitoring. This was further explored in the final part of this study, which compared the community estimates generated from hyperspectral and photoquadrat surveys at eight sites on western Guam across two levels of ID resolution. The automated predictions based on hyperspectral categories could not reliably predict benthic composition using high taxonomic resolution ID categories, but showed promise when predicting and mapping broad ID categories. This was particularly true when predicting scleractinian coral cover. The annotation libraries generated for this study could not yet sufficiently train the model to account for the high biodiversity on Guam’s reefs, but this would likely improve with increased sampling effort. This study is the first to directly compare the results using this methodology with those generated via traditional photoquadrat survey techniques across multiple sites, levels of ID category, and degrees of certainty. The results reported in this thesis contribute a wealth of new information regarding Guam’s reefs, which can be used to inform a number of potential future studies on the island and beyond.

An automated approach to characterizing Guam’s reef communities

Files and links (1)

Abstract

Details

Metrics