Ecosystem functions of sea cucumbers in coastal habitats

Abstract

Holothurians (sea cucumbers) play a key role in the biogeochemical processes of many soft-bottomed ecosystems, as they remove organic matter in benthic sediment through deposit-feeding and can possibly ameliorate organic enrichment. Concurrently, inorganic nitrogen is released to the water column as a metabolic byproduct that supports primary productivity. These ecological benefits of holothurians maintain key ecosystem functions in coastal habitats under constant anthropogenic impacts. Meanwhile, ocean warming could expose sea cucumbers to temperatures beyond their optimal thermal range, leading to a direct alteration of their physiological performance that could consequentially transform ecological functions. In this thesis, I used a combination of field and laboratory experiments to examine the effects of anthropogenic eutrophication and ocean warming on Holothuria leucospilota in coastal habitats of a megalopolis to review their ecological importance under physically challenging environments. I also established an automatic holothurian detection model with deep learning to support efficient quantification of sea cucumber populations and the evaluation of their ecological roles on a regional scale.

Field experiments revealed that holothurians enhanced the concentration of inorganic nitrogen in the water column, which might intensify primary production in eutrophic environments, while their efficiency on removing organic load could be positively related to the seasonal bioavailability of sedimentary organic matter in enriched systems. This illustrated that the net effects of holothurians on ameliorating eutrophication could be underlined by the interactions of environmental characteristics with their biological activities in enriched systems. Predicted future ocean temperatures will, however, alter the ecological functions of tropical sea cucumbers. Although H. leucospilota can sustain rates of nutrient remineralization under moderate temperature increase (+ 3°C above current baseline), higher predicted temperatures (+ 5°C) caused mass mortality which possibly incapacitate ecological functions in tropical habitats. To further understand the ecological values of sea cucumbers on a system level, I developed an automatic holothurian detection model that supports holothurian density computations over large spatial scale using convolutional neural networks (YOLO). A commercial grade waterproof camera was utilized to collect images for model training while data augmentation was optimized to improve model performance. The output model has a mean average precision (mAP) score of 0.93, which is the best among holothurian detection models of similar kind. Making use of this model, I then extensively surveyed H. leucospilota populations to estimate their total contribution on nutrient mineralization in coastal habitats of Hong Kong. The density of H. leucospilota was approximated at 0.126 m-2 while their projected annual contribution on NH4+ production was about 1.40 mg m-2 l-1. This exemplifies the practicality of utilizing machine learning models to facilitate the evaluation of ecological functions at a system level.

Overall, holothurians aid the mineralization of nitrogen nutrients but ocean warming could debilitate their ecological functions. This thesis also demonstrated the potential of projecting ecological contributions from an organismal level to broader scales using deep learning. These combined approaches have enhanced the current understanding of the ecological roles of holothurians and highlighted the advantages of incorporating machine learning in ecological studies.