Habitat modelling of 'functional reef' and broad-scale ecosystem components for temperate Australian waters (NESP MaC 2.1)

Brief description

This record describes the outputs of two different modelling exercises that were used to characterise the seafloor habitats for temperate Australian waters. The modelled area includes all shelf waters (<250m depth) in southern Australia south of the Tropic of Capricorn. Bioregional benthic habitat maps were constructed using (1) the Geoscience Australia 250m 2023 grid (Beaman 2023); (2) ground-truthing observations derived from horizontally facing imagery from stereo-BRUV and BOSS camera systems; and (3) several physical datasets as covariates in model development (all oceanographic variables smoothed to 250m resolution). Source data is available from Geoscience Australia's eCat: https://doi.org/10.26186/148758 (bathymetry), Squidle+: (benthic imagery annotations), and (3) AODN Portal: https://portal.aodn.org.au/search (IMOS oceanographic datasets). The specific subset of GA observations used in this modelling exercise is available from https://github.com/UWA-Marine-Ecology-Group-projects/nesp-2.1/blob/main/data/tidy/NESP-2.1_broad-habitat.csv. See the NESP Mac Project 2.1 final report for a description of the sampling design for ground-truthing observations and annotation technique. -----Functional Reef model (binomial)----- This model discriminates ‘functional reef’ from sediment (non-reef) ecosystem types. Functional reef is defined by this project as “any seabed area functioning as a reef, which may include dense beds of sessile invertebrates or molluscs”. This term was chosen because much of the continental shelf is dominated by sediment yet is stable enough to support emergent sessile biota that provide structure and resources for “reef-affiliated” species. The modelling approach uses a Bayesian representation of a Binomial generalised linear model. For ground-truthing benthic annotations, the following benthic categories were collapsed into the ‘functional reef’ classification: sessile invertebrates, bare rocky reef (consolidated), macroalgae, Amphibolis spp. and Thalassodendron spp. All other benthic classifications were assigned to the ‘non-reef’ category. -----Ecosystem Component model (multinomial)----- This model discriminates between five broad habitat types (hereafter ‘ecosystem components’): seagrass, macroalgae, sessile invertebrates, bare consolidated substrata, bare unconsolidated substrata. The modelling approach uses a Bayesian implementation of a Multinomial generalised linear model. For ground-truthing benthic annotations, benthic annotations for mobile species (e.g. echinodermata) were discarded. All remaining annotations were collapsed into the five broad ecosystem components. A selection of mapping (WMS) services are listed in the 'Downloads & Links' section of this record. See the 'Lineage' section for a full description of the data packages available for download, and for more visualisation options.

Lineage

Maintenance and Update Frequency: none-planned

Statement: See NESP Marine and Coastal Project 2.1 final report for a full for a description of the sampling design, annotation technique, and modelling approach. A key objective of this modelling approach was to effectively report the spatial uncertainty associated with model predictions to assist with conservation and management assessments. The outputs from both the Functional Reef and Ecosystem Component models is presented in two alternative ways. VISUALISATION OPTION #1 The first approach presents model outputs as a map of the mean probability of occurrence for reef or each ecosystem class, with uncertainty (inter-quartile range (IQR) error of prediction) shown as a separate map that can be viewed in parallel. The data packages for both the binomial reef and multinomial ecosystem models contain ReadMe files describing the data structure: -----Functional Reef (binomial)----- https://data.imas.utas.edu.au/attachments/seamap_predicted_habitat/temperate_aus/model_outputs/reef_binomial/ProbabilityOfOccurrence -----Ecosystem Components (multinomial)----- https://data.imas.utas.edu.au/attachments/seamap_predicted_habitat/temperate_aus/model_outputs/ecosystemComponents_multinomial/ProbabilityOfOccurrence VISUALISATION OPTION #2 The second uncertainty mapping approach reports a map of the expected probability of occurrence (mean prediction) combined with the IQR error of prediction in a single raster, where each combination of mean x error (10 error quantiles x 10 mean quantiles = 100 unique raster values) is plotted along a bivariate colour gradient. The colour scales are shown on two axes, with y showing the probability of occurrence (or “likelihood”) and x representing the error in those predictions (or “(un)certainty”). The data package for both the binomial reef and multinomial ecosystem models contain ReadMe files describing the data structure: -----Functional Reef (binomial)----- https://data.imas.utas.edu.au/attachments/seamap_predicted_habitat/temperate_aus/model_outputs/reef_binomial/BivariateUncertainty -----Ecosystem Components (multinomial)----- https://data.imas.utas.edu.au/attachments/seamap_predicted_habitat/temperate_aus/model_outputs/ecosystemComponents_multinomial/BivariateUncertainty

Notes

Credit
National Environmental Science Program (NESP) Marine and Coastal Hub

Credit
Department of Climate Change, Energy, the Environment and Water (DCCEEW), Australian Government

Credit
In addition to NESP (DCCEEW) funding, this project is matched by an equivalent amount of in-kind support and co-investment from project partners and collaborators.