Field collections and environmental DNA surveys reveal topographic complexity of coral reefs as a predictor of cryptobenthic biodiversity across small spatial scales

Coral reefs represent some of the most biodiverse ecosystems in the world but are currently undergoing large-scale degradation due to anthropogenic stressors. Such degradation usually begins with coral bleaching, and if the stress condition is inflicted for too long may eventually result in loss of structural complexity (or “flattening”) of the reef, dramatically changing habitat availability for reef-associated fauna. Despite having been linked to important ecosystem functions, cryptobenthic organisms are often overlooked in ecological monitoring programs, and their microhabitat dependencies are poorly understood. Here, we combined collection-based biodiversity monitoring techniques with five different environmental DNA (eDNA) sampling techniques (Reef water, sediment, crevice water, slurp gun, and bulk sediment) to survey cryptobenthic fishes and annelids on a Maldivian fringing coral reef. Collectively, 176 fish and 140 annelid taxa were detected with eDNA across 14 surveyed sites, more than doubling the reported annelid taxa in the region with 88 new occurrences. Water filtered near the reef structure revealed the highest species richness out of the five eDNA sampling techniques tested. Furthermore, we found correlations between fish species richness and topographic complexity for both collection- and eDNA-based techniques. This suggests that detection by eDNA may be linked to site-specific predictors and reveal community differences across small spatial scales (tens of meters). We also report that reef flattening (going from structural complex to less complex sites) can cause a 50% reduction in fish diversity and that cryptobenthic fish species richness was highly associated with branching corals. In contrast, annelid communities showed no clear correlations with environmental predictors, but co-amplification of non-target, non-annelid taxa may have distorted such correlations if present. This suggest that the predictive powers of eDNA for environmental gradients may be dependent on the targeted taxa.