@phdthesis{fdi:010092494,
  title = {{E}ffects of global changes on microbioeroding communities living in massive corals from the {W}estern {I}ndian {O}cean over long term},
  author = {{A}laguarda, {D}.},
  editor = {},
  language = {{ENG}},
  abstract = {{C}oral reefs are increasingly threatened by global changes as they affect both accretion and erosion processes. {A}mong these processes, reef bioerosion is a major natural process of degradation resulting from the action of various organisms on and in carbonate substrates. {R}ecently, a particular attention has been given to the roles played by bioeroding (or perforating) microflora,which include

cyanobacteria, microalgae, and fungi, in the functioning of coral reefs, especially in the carbonate budget, because of their important role in the dissolution of dead carbonates over short term (day, month, year). {T}he thesis's main objective was thus to study the effects of global change factors such as ocean warming and acidification, on the composition, distribution, and abundance of reef microbioeroding

communities over long term. {S}ince long-term experiments with dead corals are difficult to carry out, several coral cores from two slow-growing massive coral genera ({D}iploastrea sp. and {P}orites sp.) were collected along the {M}ozambique {C}hannel, particularly in {M}ayotte. {T}hose cores covered the last decades (30 to 50 years). {S}uch massive corals are known to be natural geological archives largely colonized by

microbioeroding communities which leave traces while dissolving {C}a{CO}3. {T}o study the dynamics of microbioeroding communities in the two targeted coral genera, two innovative methods were developed: a machine learning approach to quickly and accurately analyze thousands of {S}canning {E}lectron {M}icroscope pictures of microbioeroding traces along three vertical transects parallel to the main coral growth axis,

and a lipid biomarkers approach along a coral core of {D}iploastrea sp.. {T}he machine learning method based on a {CNN} model was first developed on the coral {D}iploastrea sp. with an accuracy of 93%. {I}t was then adapted to {P}orites sp. by modifying a hyperparameter (95% accuracy). {T}he geochemical approach tried identifying specific lipid markers of the boring microalga {O}streobium sp. and the coral {D}iploastrea sp. during the last decades.

{T}he results showed that the abundance of microbioeroding traces is 3 to 4 times higher in the coral {D}iploastrea sp. than in {P}orites sp and has decreased in both coral genera over the last decades. {I}n {D}iploastrea sp., the decrease was 90% over the last 54 years and was coupled with a very important change in community composition between 1985-1986. {T}he density (bulk) of {D}iploastrea sp. has also dropped significantly over the last 5 decades.

{L}ogistic regressions showed that temperature, wind speed, and internal p{H} of the coral, more or less coupled, are correlated to the abundance of microbioeroding traces. {T}he geochemical approach also highlighted a significant decrease of a lipid biomarker group, the amides, over the last decades. {A}lthough it is difficult to attribute amides to a specific taxon or species in the coral skeleton, {I} hypothesize that they could potentially reflect the presence of microbioeroding communities. {T}o confirm or refute the observed trends, there is a need to study more coral cores, from different areas, and over a longer period. {I}n addition, other factors should be studied to understand better the decrease in the abundance of microbioeroding communities and its implication in coral health and resilience, such as trace metals and other variables of the carbonate system.},
  keywords = {{OCEAN} {INDIEN} ; {MAYOTTE} ; {MOZAMBIQUE} {CANAL}},
  address = {{P}aris},
  publisher = {{S}orbonne {U}niversit{\'e}},
  pages = {343  multigr.},
  year = {2023},
  URL = {https://www.documentation.ird.fr/hor/fdi:010092494},
}