Replication Data for Top-down and bottom-up forcing on coral reefs changes with habitat-specific predator abundance

Empirical field data and derived analysis tables underpinning the manuscript: **Top-down and bottom-up forcing on coral reefs changes with habitat-specific predator abundance** Dedman, S., Farabaugh, N. F., Bond, M. E., Chapman, D., Clua, E., Harborne, A. R., Papastamatiou, Y. P., Kiszka, J. J., Heupel, M., Wirsing, A. J., Heithaus, M. R. (2026, in preparation; *Proceedings of the National Academy of Sciences* or *Science*). ### Purpose The dataset supports a test of whether top-down forcing (sharks driving mesopredators, herbivores, and ultimately benthic cover) or bottom-up forcing (productivity and habitat driving benthic cover, fish, and ultimately sharks) better explains coral reef community structure across French Polynesia, and whether the dominant direction differs between atolls (where reef sharks remain abundant) and high islands (where shark populations are sparser due to greater fishing pressure). ### Nature and scope The archive bundles three layers of data: 1. **Raw observation files**, preprocessed from field surveys to one row per observation unit: - Benthic underwater visual census (UVC) percent cover per transect, by reef and benthic category (hard coral, crustose coralline algae, other algae, fleshy macroalgae, turf algae, soft coral, sand, rubble, pavement, invertebrate cover) - Fish UVC counts and biomass per species per transect - Shark baited remote underwater video (BRUV) MaxN per species per video set - Predator-teleost BRUV MaxN per family per set (lutjanidae, scombridae, megalopidae, carangidae, sphyraenidae, serranidae, lethrinidae) - Reef-to-topology classification, site ordering metadata, functional-group lookup tables, and a derived trophic categorisation following Desbiens et al. 2021 (DOI: 10.1002/ecy.3303). 2. **Reef-aggregated analysis tables** (output of the data-prep stage of the analysis code), provided so reviewers can skip the data-prep step and proceed directly to modelling: 24 reefs total, with separate atoll (n=13) and high-island (n=11) subsets; each row includes shark counts, fish biomass per functional group, environmental covariates (sea surface temperature, net primary productivity, lagoon size, emerged land area, population density, geomorphological relief), and benthic percent cover. 3. **Pre-fitted Bayesian Structural Causal Models** for top-down vs bottom-up scenarios, four lists of 11 fitted brms (Student-t family with cubic shrinkage splines) model objects (Atolls / HighIslands x TopDown / BottomUp), with embedded Stan compiled binaries stripped to reduce file size from ~210 MB to ~17 MB per list. Allows reproducibility of all model-comparison and figure-generation steps without re-running the multi-hour Bayesian fit. ### Methodology Field surveys were conducted as part of the Global FinPrint French Polynesia campaign. At each reef, fish and benthic UVC transects were paired with multiple BRUV deployments to provide concurrent measures of fish biomass per functional group, benthic cover, and shark abundance. Surveys covered the four main French Polynesian archipelagos, sampling reefs spanning the topological gradient from open atoll to near atoll to high barrier reef. Standard transect and BRUV protocols were applied. Full survey methodology is described in the associated manuscript. ### Spatial coverage 24 coral reefs across French Polynesia, comprising: - 13 atolls (open atoll, closed atoll) - 11 high islands (near atoll, high barrier) Reef geomorphological classifications follow standard reef-typology categories; Marquesas-archipelago surveys were excluded from the analysis dataset due to distinct biogeographic context. ### Taxonomic coverage - **Sharks**: pooled reef-resident category (grey reef, blacktip reef, whitetip reef, silvertip, tawny nurse, blacktip), sicklefin lemon shark (*Negaprion acutidens*), and transient pelagic sharks (tigers, hammerheads). - **Reef fish**: aggregated to four functional groups, piscivore, invertivore, herbivore, and planktivore, plus seven predator-teleost families counted at family level. - **Benthos**: hard coral, crustose coralline algae, fleshy macroalgae, turf algae, other algae, soft coral, sand, rubble, pavement, invertebrate cover. ### Files A complete file-level manifest with descriptions, sources, and intended usage is provided in `data/README.md` in the associated code repository. Approximately 150 MB total, mostly in the four pre-fitted brms model-list .Rds files (~80 MB combined after Stan-binary stripping). ### Code availability Analysis code is archived on Zenodo (DOI: 10.5281/zenodo.19835080) and developed openly at https://github.com/SimonDedman/FIU-SharkFishCoral-FrenchPoly. The code expects this data archive to be downloaded into a folder named `NFF_data/` at the project root.