@article{fdi:010094296,
  title = {{T}he {F}ast-{G}reedy algorithm reveals hourly fluctuations and associated risks of shark communities in a {S}outh {P}acific city},
  author = {{C}hafia, {I}. and {Z}ahir, {J}. and {L}ett, {C}hristophe and {A}gouti, {T}. and {M}ousannif, {H}. and {V}igliola, {L}aurent},
  editor = {},
  language = {{ENG}},
  abstract = {{U}nprovoked shark bites are increasing globally. {R}egional hotspots like {N}oum{\'e}a show rising incidents involving bull sharks ({C}archarhinus leucas) and tiger sharks ({G}aleocerdo cuvier), leading to the culling of these protected species. {I}dentifying high-risk areas and times is key to balancing human safety and shark conservation. {H}ere, we collected five years of acoustic telemetry data for both shark species in the lagoon of {N}oum{\'e}a. {T}he data were categorized by species, divided into 24 hourly subsets, and modeled as bipartite graphs. {T}he {F}ast-{G}reedy algorithm was applied to identify distinct communities of sharks and stations. {N}ormalized mutual information was used to cluster communities and detect spatiotemporal patterns. {T}he study revealed up to 9 hourly communities for bull sharks and 21 for tiger sharks, each grouping into 3 clusters. {S}everal high-risk areas and times were identified. {B}ull sharks formed schools, and a cluster was observed in the harbor between 6:00 and 13:00, increasing bite risk on nearby beaches in the morning. {T}iger sharks were more solitary and were present day and night at most stations except those in relatively turbid areas. {B}oth species showed fission-fusion dynamics, with communities merging at dusk, indicating increased movement and a higher risk during this low-light period. {A} key innovation of our modeling framework was its ability to handle temporal variability in community detection algorithms applied to bipartite networks. {T}he model identified key overlap periods of shark-human activity, highlighting the need for real-time monitoring, safety measures, and public awareness to reduce bite risk and promote coexistence.},
  keywords = {{B}ipartite graphs ; {C}ommunity detection algorithms ; {D}ynamic network analysis ; {A}coustic telemetry data ; {R}isk management ; {NOUVELLE} {CALEDONIE} ; {NOUMEA}},
  booktitle = {},
  journal = {{E}cological {I}nformatics},
  volume = {90},
  numero = {},
  pages = {103263 [13 p.]},
  ISSN = {1574-9541},
  year = {2025},
  DOI = {10.1016/j.ecoinf.2025.103263},
  URL = {https://www.documentation.ird.fr/hor/fdi:010094296},
}