From seafloor geomorphology to predictive habitat mapping: progress in applications of biophysical data to ocean management. Peter T. Harris Geoscience Australia, Canberra ACT, Australia Currently seconded to: UNEP/GRID Arendal, Norway
Acknowledgements Thanks to Thaiënne van Dijk and conference organisers for invitation to attend Hydro12 Geoscience Australia travel support Research products of Geoscience Australia and many other scientists/institutions involved www.geohab.org
Uses of hydrographic data: Safe Navigation (nautical charts) Offshore oil and gas exploration and development Fisheries management Offshore minerals and aggregates Determining national marine jurisdiction (ECS) Port development and maintenance Computer models of tides, waves, ocean currents Communication (Google Oceans) Naval operations Marine zone management
Outline of talk: Introduction - Spatial Marine Management Approaches to seafloor characterisation and applications to decision-making Geomorphic features Unsupervised seascapes classification Supervised predictive habitat models Conclusions
Spatial marine management Flaws of the sector-based approach Fishing, oil and gas, shipping etc. managed separately Inconsistent with broader goals of biodiversity and ecosystem conservation Ecosystem based management (EBM) Addresses multiple pressures Acknowledges ecosystem services How to apply EBM? Need a spatial frame of reference (bioregions, ecosystem components, inventory of assets).
Mapping approaches that help to define bioregions and habitats: Seafloor geomorphic features Multivariate seascapes Predictive habitat modelling
Approaches to seafloor characterisation: Approach How generated? Advantages Disadvantages Geomorphic features Biophysical interpolations unsupervised classifications (eg. seascapes) Predictive habitat maps supervised classifications (maximum entropy, decision-trees, etc.) interpreted from bathymetric data apply classification scheme multivariate analysis to spatially combine several biophysical data layers include direct observations of marine life with biophysical data to predict the potential distribution of species and benthic communities. + simple to communicate, technically easy to generate - limited predictive power + simple to generate with spatial data - limited predictive power, difficult to communicate + good predictive power, performance indicators - Difficult to generate (data hungry), relate to single species or group
Applications to decision-making (Australia case study) 1. Geomorphic features used to define Australian marine bioregions
Geomorphic features (IHO classes) mapped based on 250 m bathymetric grid of Australia Heap and Harris (2008) Australian Journal of Earth Sciences, 55:555-585.
Heap and Harris (2008) Geomorphic features map of Australia
The correlation between geomorphic features and benthic habitats is the focus of recent GeoHab book: www.geohab.org
Marine management based on IMCRA 2006 41 provincial bioregions Many boundaries based on geomorphology IMCRA = Integrated Marine and Coastal Regionalisation of Australia
Heap and Harris (2008) Biophysical model - Geomorphology
Example of application of geomorphic features to assessment of industrial use Petroleum titles cover an area of about 620,000 km 2 or about 8.7% of Australia s EEZ (excluding offshore territories)
Harris et al. (2007) APPEA Journal, 48:327-343
Applications to decision-making (Australia case study) 2. Seascapes used to identify biodiversity hotspots and priorities for Marine Protected Areas
Integration of ecologically-significant biophysical variables to create a single map (Seascapes) Not scale dependant Integrated product Input physical data (Seascapes) = (e.g., tidal currents) + (e.g., bathymetry) + (e.g., slope) + (e.g., % sand)
Seven variables derived from interpolation of bathymetry, samples & modelled data Water Depth Slope %Gravel %Mud Effective Disturbance Seafloor Temperature Primary Productivity Completed using ERMapper ISOClass facility (Iterative Self Organising Classification)
Depth Slope %Gravel Grid resolution %Mud 0.01 o, ~5 km Seafloor Effective Primary Grid resolution Temperature 0.01 Productivity Disturbance o, ~5 km Grid resolution 0.01 o, ~5 km Grid resolution 0.01 o, ~5 km Grid Grid resolution 0.01 0.01 o, ~5 o ~5 km km Grid resolution 0.01 o, ~5 km
Australia Shelf Seascapes 13 Ecologically unique Seascapes (Unsupervised Classification) 1. Moderate depth, flat, slightly gravelly, cold, low disturbance, moderate primary productivity
How relevant to Biodiversity? Seascape heterogeneity based on Focal Variety Analysis Used to identify hotspots of seascape heterogeneity (surrogate for biodiversity) 20 x 20 cell analysis area
Australian Shelf Seascapes - Heterogeneity Harris et al. (2008) Ocean Coastal Management, 51:701-711.
SEWPaC Proposal June 2012, 60 reserves covering 3.1 million square kilometres, largest system of marine reserves in the world. Some MPAs suggested by seascape analysis, others by geomorphology
Applications to decision-making (Australia case study) 3. Predictive habitat model of coral habitat distribution in the Great Barrier Reef to assess marine park management scheme
Physical measurements: - Depth - Slope - Temperature - Sediment size - Current speed - etc. + = Biological observations Predicted habitat for species or community
Predictive Habitat Modeling Techniques (Huang et al., Ecological Informatics, 2011) BIOCLIMatic (BIOCLIM) (Nix, 1986) DOMAIN (Carpenter et al., 1993) Logistic Regression (LoR) (Peeters and Gardeniers, 1998; Ozesmi and Ozesmi, 1999; Felicisimo et al., 2002) Decision Trees (DT) (Zacharias et al., 1999; Pitcher et al., 2007) Genetic Algorithm for Rule-set Production (GARP) (Stockwell and Peters, 1999) Ecological Niche Factor Analysis (ENFA) (Hirzel et al., 2002) Generalised Additive Model (GAM) (Zaniewski et al., 2002) Artificial Neural Networks (ANN) (Joy and Death, 2004) Generalised Linear Model (GLM) (Brotons et al., 2004; Hirzel et al., 2006) Multivariate Adaptive Regression Spline (MARS) (Leathwick et al., 2005) Maximum Entropy (MAXENT) (Phillips et al., 2006) Support Vector Machine (SVM) (Drake et al., 2006; Guo et al., 2005,) Generalised Dissimilar Model (GDM) (Ferrier et al., 2007) Limiting Variable and Environmental Suitability (LIVES) (Li and Hilbert, 2008)
Maps of reef distribution based on satellite images and air photographs
Reef geology Most Holocene reefs in the GBR have Pleistocene reef limestone foundations. = BANK Most reefs are multi-generation limestone geomorphic banks (but not all banks support reefs) Some reefs were unable to keep pace with post-glacial sea level rise (submerged reefs)
Used new 100 m bathymetry grid (Beaman, 2010) - data contoured at 5 m http://www.deepreef.org/bathymetry/65-3dgbr-bathy.html
Digitising geomorphic banks: Banks exceed 15 m in elevation and have at least one steep side Digitised by hand NSS coral reef Example from 11 o S
Statistics of geomorphic banks in the GBR: Overall mean depth of banks = 27.3 m Total area of all bank types (less NSS reefs) = 25,599 km 2
How much of the 25,599 km 2 of bank area actually supports living coral communities? Use small, high-resolution data set to predict area of potential coral habitat Hydrographer s Passage
Occurrence records derived from optical images taken by autonomous underwater vehicle, together with: Area of deep coral habitat estimated using Maximum Entropy (MaxEnt) on a data set from Hydrographer s Passage - depth - slope - aspect - rugosity - acoustic backscatter -geomorphic zone (slope, crest, flat or depression) 5 m resolution 70% data used as training set Coral coverage on nine banks = 55 +/- 23%.
Interpretation: 1. 55 +/- 23% of all banks = 14,000 +/- 6000 km 2 deep water coral communities. 2. NSS coral reefs area = 20,679 km 2 3. Therefore the area of coral habitat in the GBR is at least 50% larger and perhaps double the size previously believed to exist. 1 2
Protection of banks (deep reefs): Not protected from trawling ZONE TYPE Area (km 2 ) of banks included Number of banks included* Percent of banks by area Preservation Zone 190 35 0.7 Marine National Park Zone 7,301 602 28.5 Conservation Park Zone 654 42 2.6 Habitat Protection Zone 12,983 889 50.7 General Use Zone 3,157 370 12.3 Banks beyond GBR Marine Park 1,315 166 5.1 *Parts of banks may occur in more than one zone. Harris et al. (in press) ICES Journal of Marine Science
Concluding remarks
Concluding remarks Bathymetry underpins all seafloor characterisation maps Geomorphic features and seascapes useful for government decision-making and management Predictive habitat modelling the future GeoHab 2013 will be held in Rome, Italy (6-10 May)
Thank You!