Aim: Kelp forests throughout temperate regions of the world serve as foundation species that play a critical role in sustaining the health and function of marine ecosystems but are experiencing declines in abundance due to loss in resilience as the ocean climate changes. Ocean warming along southeast Australia has already been linked to dramatic losses of kelp species and is contributing to the range expansion and population increases of two species of sea urchin. The purpose of this research is to understand the impact of multiple stressors on the decline in kelps in this region.

Location: Coastal waters off Victoria, Australia

Methods: In this study, we use long-term (> 20 years) datasets on biological observations across Victorian waters to determine trends in coverage and the impact of multiple environmental variables (temperature, habitat, currents, waves, connectivity, urchin abundances) on two important kelps that serve as foundation species (Phyllospora comosa and Ecklonia radiata) using boosted regression trees. These models were then used to develop predictive distribution models for each species and also to predict areas of future risk.

Results: We found that both kelp species are decreasing in percent coverage over time and multiple environmental variables, including increasing temperatures, intensifying wave energy, changes in currents and recruitment patterns, and increases in urchin populations are all contributing to the declines of kelps. Additionally, future projections of temperature and wave energy show that these species will likely continue to decrease across 71% of Victorian waters.

Main Conclusions: This information can help to better manage these important foundation species by providing maps of their current and past distributions, along with projections of climate change, to target different areas for urchin culling or macroalgae restoration to reduce future losses.

This dataset contains all the spatial data used in the model creation and predictions across Victoria for kelps and urchins. 

Ecklonia radiata and Pyllospora comosa, and estimates of abundance of H. erythrogramma and C. rodgersii were recorded annually by SCUBA at >200 locations over 22 years by three different sampling programs: Parks Victoria Subtidal Reef Monitoring Program (SRMP), Reef Life Survey (RLS), and Victorian Fisheries Authority (VFA) fishery independent abalone surveys. To combine all these datasets together for analyses, all datasets were processed and standardised to matching units. Below gives a brief overview of each of the methods used.

Subtidal Reef Monitoring Program

Parks Victoria established the SRMP for the purpose of monitoring the status of reef-associated kelp, macroinvertebrates, and fish within the coastal waters of Victoria. Through long-term surveys, the SRMP tracks the trends in species abundances, species diversity, and reef community structure through time. The sites surveyed for the SRMP include sites inside Victorian marine protected areas (MPAs) and in areas adjacent to them. These datasets allow for comparison between highly protected areas (e.g., Marine National Parks and Marine Sanctuaries, IUCN Category II reserves) and unprotected reefs (Edgar and Barrett 1997, Edgar and Barrett 1999). The visual census methods used in these surveys were developed by Edgar and Barrett (Edgar and Barrett 1997, Edgar et al. 1997, Edgar and Barrett 1999). Each site is located using a global positioning system (GPS) unit and a 200 m transect line is laid along a shallow (<10 m) contour with its centre on the GPS point. Three different methods are used for each transect to collect information on 1) mobile fishes and cephalopods, 2) benthic invertebrates and cryptic fishes, and 3) kelp and sessile invertebrates. We are using the data from the invertebrate and kelp surveys in this study.

The methods used to capture the abundances of invertebrates are restricted to within one metre of one side of the transect line, and all invertebrates within that 1 m band are counted across four 50 m transects. We extracted all observations of C. rodgersii and H. erythrogramma from these surveys and then calculated their densities by dividing the counts by the area surveyed. To quantify kelp, 0.25 m² quadrats are placed at 10 m intervals along the transect line. These quadrats are divided into a grid of 7 x 7 perpendicular wires to provide 49 points plus one corner, making 50. The cover of kelp is then estimated by counting the number of times a species intersects with one of the points.

Reef Life Survey

Reef Life Survey (RLS) is citizen-science program that trains volunteer SCUBA divers to conduct visual surveys of underwater reefs. RLS started surveying along the Victorian coast in 2008 and continues to extend the time series of observations of fish, invertebrates, and kelps in the region. Detailed methods for how RLS data are acquired are described in Edgar and Stuart-Smith (2009). Briefly, each site is surveyed using a 50 m line transect along a depth contour with two or more depth contours targeted at each site. Invertebrates are counted within 1 m of either side of the transect. Densities for C. rodgersii and H. erythrogramma were calculated for the RLS sites by dividing the counts by the area surveyed. Additionally, at 2.5 m intervals along the transect line, digital photo-quadrats are acquired. These photos are taken downward from a 50 cm height above the seabed. These photo quadrats are then classified into 16 functional or morphological categories of algae using CATAMI (Althaus et al. 2015), which is a standard image classification system for Australia. Percent coverage of E. radiata and P. comosa were calculated based on these photo quadrats.

Victorian Fisheries Authority

To supplement the kelp data from the RLS and SRMP datasets, we used estimates of E. radiata and P. comosa percent cover and densities of H. erythrogramma and C. rodgersii from fishery independent diver surveys conducted annually by the VFA at 195 monitoring sites along the coast of Victoria from 2002 to 2016. At each of these sites, the divers estimate percent coverage of E. radiata and P. comosa and count the urchins within six 30 m by 1 m transects at random cardinal directions from the central site coordinates. The estimates for percent cover are based on visual approximation within 30 sections along each transect and then averaged to the nearest 10 % across all transects for each site. We divided the urchin counts by the area surveyed to provide density estimates per metre squared for both species of urchin at each site.

Data Collation

Across all datasets, we calculated the density of urchins per 2500 m² to standardise all datasets to the total number of urchins within a 50 m resolution grid. Percent cover was left as percent cover. Once all the datasets were standardised, they were combined into a single dataset for analyses. This final dataset includes a time series from 1998 to 2020.

We calculated variables describing the structure of the seafloor using the Benthic Terrain Mapper toolbar (Wright et al. 2005) and Spatial Analyst within ArcGIS (ESRI 2011). All derivatives extracted for each site and their descriptions are shown in Table 1. Those variables with a temporal component (wave orbital velocity, current speed, SST, connectivity parameters) were extracted for each year in the biological sampling while accounting for a 1-year lag in recruitment for kelp and a 2-year for urchins when applying the connectivity variables. These lags accounted for the lag in time before these species were counted across the different sampling programs due to recruitment and size. The resolution of all environmental variables was converted to 50 m resolution for analysis. The seafloor variables were at a finer resolution (2.5 m) and zonal statistics was used to find the mean and standard deviation of each of them in a 50 m resolution grid. The oceanographic variables ranged from 500 m resolution (the hydrodynamic variables) to 2000 m resolution (SST). These variables were converted to 50 m resolution using Empirical Bayesian Kriging to interpolate between cell centres. Further information on the sources and methods to create the environmental parameters used in this study can be found in the Supplementary Information or in Ierodiaconou et al. (2018).

The Biological data has a separate file for each species. These each contain the observation location (lat/long), the year, and the percent cover (both decimal and integer) of kelps or density of urchins at both 1 m² and 2500 m².

The Environmental data contains all the seafloor and environmental data for the modelling of species/habitat associations. The naming conventions are below (see manuscript for more details on how each of these layers was calculated.

• AverageAnnualCurrentSpeeds
  • a_ACS_[year].tif - average annual current speed and the year associated with that data layer (m/s)
• Connectivity
  • InDegree_BU_[year].tif - average in degree for black urchins (C. rodgersii) and the year associated with that data layer
  • InDegree_Kelp_[year].tif - average in degree for kelps (both species) and the year associated with that data layer
  • InDegree_PurpleUrchin_[year].tif - average in degree for purple urchins (H. erythrogramma) and the year associated with that data layer
  • LocRet_BU_[year].tif - average local retention for black urchins (C. rodgersii) and the year associated with that data layer
  • LocRet_Kelp_[year].tif - average local retention for kelps (both species) and the year associated with that data layer
  • LocRet_PurpleUrchin_[year].tif - average local retention for purple urchins (H. erythrogramma) and the year associated with that data layer
  • SlfRec_BU_[year].tif - average self-recruitment for black urchins (C. rodgersii) and the year associated with that data layer
  • SelfRec_Kelp_[year].tif - average self-recruitment for kelps (both species) and the year associated with that data layer
  • SelfRec_PurpleUrchin_[year].tif - average self-recruitment for purple urchins (H. erythrogramma) and the year associated with that data layer
  • TtlInfl_BU_[year].tif - average total inflow for black urchins (C. rodgersii) and the year associated with that data layer
• MaxWinterWaveOrbitalVelocities
  • w_MOV_[year].tif - average winter wave orbital velocities and the year associated with that data layer (m/s)
• Seafloor
  • BioUnt_W2E.tif - the biounits for Victoria numbered from west to east
  • bpi125_fs.tif - the bathymetric position index at a 125 m scale calculated at 2.5 m resolution and then averaged within a 50m moving window
  • btystd_fs.tif - the bathymetric standard deviation calculated at 2.5 m resolution and then averaged within a 50m moving window
  • rug_fs.tif - the rugosity calculated at 2.5 m resolution and then averaged within a 50m moving window
  • slope_fs.tif - the slope calculated at 2.5 m resolution and then averaged within a 50m moving window
  • Victoria_MBES_LiDAR_50m_bathy_AHD_VG94_fs50m_m.tif - the 50 m bathy data for Victoria
• Summer_SST
  • s_SST_C_[year].tif - average summer SST and the year associated with that data layer (degrees Celsius)

The Future predictions downloaded from CSIRO are included in the "FuturePredictions" folder. "Hs" is wave height and "tos" is sea surface temperature. See manuscript for more detail and to access the locations where these data were downloaded from.