README

Patch design principles for oyster reef restoration: Lessons from fish habitat use in natural remnant systems

Authors: Christopher Pine, Katherine Erickson, Ana Bugnot, Paul Gribben, & William Figueira

Journal: Ecological Applications

Overview

This repository contains the raw data and code required to replicate the statistical analyses and generate the figures for the above manuscript. The study examines how oyster reef characteristics (size, shape, complexity, and spatial arrangement) influence fish assemblage metrics (abundance, species richness, and diversity) across natural remnant oyster reef systems in southeastern Australia.

Repository Contents

Code and data

Pre-computed model objects

The multivariate analyses use manyglm models fitted via the mvabund R package, which are computationally intensive to run. Pre-computed model objects are provided as .rds files (R serialised data) so users can load results directly without re-running the full analysis.

Figures

Study Design

Surveys were conducted on natural remnant oyster reefs across three sites in southeastern Australia:

• OP-CR – Crookhaven River
• SY-PH – Port Hacking (Sydney)
• SY-TP – Towra Point (Sydney)

At each reef, underwater video surveys were deployed in three zones:

• Centre – On-reef, central area of the reef
• Edge – On-reef, along the reef perimeter
• Off – Off-reef, adjacent bare sediment or seagrass habitat

Surveys were conducted across seasons (Summer and Winter) and years (2019–2021). Fish assemblages were quantified from video footage using the MeanCount method.

Data File Descriptions

1. species.data.csv

This file contains one row per species per survey deployment. Each row records the abundance and occurrence of a single fish species at a specific reef, zone, season, and year, alongside taxonomic information and ecological trait classifications.

Survey identifiers and design

Taxonomic and trait information

Species abundance and occurrence metrics

All metrics below are derived from the MeanCount video analysis method. In this method, the video is divided into equal sample periods, and within each sample period the maximum number (MaxN) of each species visible in any single frame is recorded. The MeanCount is the average of these MaxN values across all sample periods.

Community-level metrics (repeated per survey)

These values are repeated for all species rows sharing the same SurveyID.

Derived grouping variable

2. MeanCount.NO.hyperabundants.dat.csv

This file contains one row per survey deployment. Each row summarises the fish assemblage at a specific reef, zone, season, and year, and includes the morphological and spatial characteristics of the associated reef. Hyperabundant species have been excluded from the assemblage metrics in this dataset.

Survey identifiers and design

Fish assemblage metrics

Reef spatial and morphological characteristics

Landscape metrics (shape, iso.index, circle, frac, para, perimeter, perimeterarea.ratio) were derived using FRAGSTATS or equivalent landscape ecology software from georeferenced reef outlines. These values are constant across all survey deployments from the same reef.

3. MeanCount.with.hyperabundants.dat.xlsx

This file has the same structure and columns as MeanCount.NO.hyperabundants.dat.csv (see above), but retains hyperabundant species in the assemblage metrics. This version of the dataset is used for the species richness analysis, where excluding hyperabundant species is not appropriate. See Section 2 above for full column descriptions.

Figure Descriptions

Figure 1 – Oyster reefs in various locations throughout NSW. (A) Satellite image of Port Hacking oyster reef showing patch configuration (NearMaps). (B) Port Hacking oyster reef at low tide showing the arrangement and height of patches. (C) Centre of an oyster reef during high tide at Port Hacking showing the three-dimensional complexity of the interior patch habitat. (D) Edge of an oyster reef patch illustrating the complexity and diversity of refuge spaces along the edge and the interface between the patch and the sand matrix, highlighting the vertical relief of the reef structure. Photo credits (B, C, D): Christopher Pine.

Figure 2 – Depiction of how Remote Underwater Video (RUV) apparatuses were placed on the three zones of the oyster reef. (A) Centre of the reef. (B) Edge of the reef. (C) Off the reef (from the camera's perspective). (D) Schematic diagram illustrating the placement of the apparatuses on an oyster patch. Photo credits (A, B, C) and illustration credit (D): Christopher Pine.

Figure 3 – Comparison of the mean of each key assemblage metric between seasons (yellow = summer, blue = winter) and reef zones (Centre, Edge, Off) for each study estuary. Panels show: (A–C) Abundance (MeanCount, sampling every minute for 60 minutes); (D–F) Species Richness (per 60 minutes of video); (G–I) Shannon's Diversity.

Figure 4 – Net trend of each patch characteristic on abundance (A), species richness (B), and diversity (C) of fish assemblages. Blue up arrows represent significant net positive trends; red down arrows represent net negative trends. Blank cells indicate non-significant effects or variables removed during model selection. "N.I" indicates variables excluded due to collinearity. Asterisks (*) denote marginally significant effects (0.05 < p < 0.08). Net trends were determined by averaging significant trends across reef zones, seasons, and—where present—season × reef zone pairwise combinations. Detailed tables are available in Appendix 4.

Figure 5 – Net overall relationships between patch characteristics and fish abundance across estuaries (Port Hacking, Towra Point, Crookhaven River), based on estimated marginal means from the generalised linear model. Panels show relationships with: (A) patch area, (B) shape index, (C) circle index, (D) fractal dimension, (E) isolation index, (F) nearest neighbour distance, (G) perimeter-area ratio, (H) rugosity. Colours correspond to site; shaded areas indicate 95% confidence intervals; points represent raw data.

Figure 6 – Net overall relationships between patch characteristics and fish species richness across estuaries, based on estimated marginal means. Panels show relationships with: (A) circle index, (B) fractal dimension. Colours correspond to site; shaded areas indicate 95% confidence intervals; points represent raw data.

Figure 7 – Net overall relationships between patch characteristics and Shannon's H diversity across estuaries, based on estimated marginal means. Panels show relationships with: (A) circle index, (B) fractal dimension, (C) isolation index. Symbols denote season × reef zone combinations; colours correspond to site; shaded areas indicate 95% confidence intervals; points represent raw data.

Figure 8 – nMDS ordination plots of fish abundance data (Bray-Curtis dissimilarity) showing differences in fish assemblage composition by season and zone at each site: (A/B) Port Hacking (stress = 0.2), (C/D) Towra Point (stress = 0.18), (E/F) Crookhaven River (stress = 0.17).

Figure 9 – Conceptual framework illustrating oyster reef design principles in two hypothetical systems. System 1 represents one large continuous reef patch totalling 100 m². System 2 represents a system of smaller connected patches with the same total area budget (100 m²) incorporating the design principles developed in this study, illustrating how ecological understanding can maximise ecological outcomes per unit area restored. Illustration credit: Christopher Pine.

Software and Dependencies

The analysis code (Oyster_Reef_Data_Analysis.qmd) was written in R and rendered using Quarto. Please refer to the Quarto document for specific package dependencies and version information.

Contact

For questions regarding this dataset, please contact Christopher Pine (corresponding author).