The information within this document and associated DVD is intended to assist emergency managers in tsunami planning and preparation activities. The Attorney General's Department (AGD) has supported Geoscience Australia (GA) in developing a range of products to support the understanding of tsunami hazard through the Australian Tsunami Warning System Project. The work reported here is intended to further build the capacity of the QLD State Government in developing inundation models for prioritised locations. Internally stored data /nas/cds/internal/hazard_events/sudden_onset_hazards/tsunami_inundation/gold_coast/gold_coast_tsunami_scenario_2009

In response to the devastating Indian Ocean Tsunami (IOT) that occurred on 26 December 2004, Geoscience Australia developed a framework for tsunami risk modelling. The outputs from this methodology have been used by emergency managers throughout Australia to plan and prepare for future events. For Geoscience Australia to be confident in the information that is provided to the various stakeholders, validation of the model and methodology is required. Tsunami modelling at Geoscience Australia employs a hybrid approach which couples two models at the continental shelf. First we use an elastic dislocation model to simulate the initial sea-floor displacement of an earthquake source. The tsunami is then propagated across the deep ocean using URSGA, a finite difference model that solves the non-linear shallow water wave equation across nested grids. We stop this model at the 100 m water depth contour and couple it to a detailed inundation modelling tool, ANUGA, developed by Geoscience Australia and the Australian National University. ANUGA also solves the non-linear shallow water wave equation and uses a finite volume method. It incorporates bottom friction coefficients and can resolve hydraulic shocks and the wetting and drying process. While the huge loss of life from the 2004 Indian Ocean tsunami was tragic it did provide a unique opportunity to record the impact of a large tsunami event. Information gained from post-tsunami surveys and tide gauge recordings at Patong Bay, Thailand and Geraldton, Western Australia is used to validate our tsunami inundation modelling methodology. By using these two locations we can assess the performance of our models at near-source and distal locations. In addition, wave heights observed in the deep ocean from satellite altimetry are utilised to validate our deep water propagation model.

The information within this document and associated DVD is intended to assist emergency managers in tsunami planning and preparation activities. The Attorney General's Department (AGD) has supported Geoscience Australia (GA) in developing a range of products to support the understanding of tsunami hazard through the Australian Tsunami Warning System Project. The work reported here is intended to further build the capacity of the Tasmanian State Government in developing inundation models for prioritised locations.

The Tropical Cyclone Risk Model (TCRM) is a statistical-parametric model of tropical cyclone behaviour and effects. A statistical model is used to generate synthetic tropical cyclone events. This is then combined with a parametric wind field model to produce estimates of cyclonic wind hazard.

Random forest (RF) is one of the top performed methods in predictive modelling. Because of its high predictive accuracy, we introduced it into spatial statistics by combining it with the existing spatial interpolation methods, resulting a few hybrid methods and improved prediction accuracy when applied to marine environmental datasets (Li et al., 2011). The superior performance of these hybrid methods was partially attributed to the features of RF, one component of the hybrids. One of these features inherited from its trees is to be able to deal with irrelevant inputs. It is also argued that the performance of RF is not much influenced by parameter choices, so the hybrids presumably also share this feature. However, these assumptions have not been tested for the spatial interpolation of environmental variables. In this study, we experimentally examined these assumptions using seabed sand and gravel content datasets on the northwest Australian marine margin. Four sets of input variables and two choices of 'number of variables randomly sampled as candidates at each split' were tested in terms of predictive accuracy. The input variables vary from six predictors only to combinations of these predictors and derived variables including the second and third orders and/or possible two-way interactions of these six predictors. However, these derived predictors were regarded as redundant and irrelevant variables because they are correlated with these six predictors and because RF can do implicit variable selection and can model complex interactions among predictors. The results derived from this experiment are analysed, discussed and compared with previous findings. The outcomes of this study have both practical and theoretical importance for predicting environmental variables.

The 2004 Indian Ocean Tsunami raised the importance of tsunami as a significant emergency management issue in Australia. The Australian government responded by initiating a range of measures to help safeguard Australia from tsunami, in particular the Australian Tsunami Warning System (ATWS). In addition it is supporting fundamental research into understanding the tsunami risk to Australian communities. The Risk and Impact Analysis Group (RIAG) of Geoscience Australia achieves this through the development of computational methods, models and decision support tools for use in assessing the impact and risk posed by hazards. Together with support from Emergency Management Australia, it is developing a national tsunami hazard map based on earthquakes generated from the subduction zones surrounding Australia. These studies have highlighted sections of the coastline that appear vulnerable to events of this type. The risk is determined by the likelihood of the event and the resultant impact. Modelling the impacts from tsunami events is a complex task. The computer model ANUGA is used to simulate the propagation of a tsunami toward the coast and estimate the level of damage. A simplification is obtained by taking a hybrid approach where two models are combined: relatively simple and fast models are used to simulate the tsunami event and wave propagation through open water, while the impact from tsunami inundation is simulated with a more complex model. A critical requirement for reliable modelling is an accurate representation of the earth's surface that extends from the open ocean through the inter-tidal zone into the onshore areas. However, elevation data may come from a number of sources and will have a range of reliability.

The Tropical Cyclone Risk Model (TCRM) is a statistical-parametric model of tropical cyclone behaviour and effects. A statistical model is used to generate synthetic tropical cyclone events. This is then combined with a parametric wind field model to produce estimates of cyclonic wind hazard.

Spatial interpolation methods for generating spatially continuous data from point samples of environmental variables are essential for environmental management and conservation. They may fall into three groups: non-geostatistical methods (e.g., inverse distance weighting), geostatistical methods (e.g., ordinary kriging) and combined/hybrid methods (e.g. regression kriging); and their performance is often data-specific (Li and Heap, 2008). Because of the robustness of machine learning methods, like random forest and support vector machine, in data mining fields, we introduced them into spatial statistics by applying them to the spatial predictions of seabed mud content in combination with existing spatial interpolation methods (Li et al., 2011). This development can be viewed as an extension of the combined methods from statistical methods to machine learning field. These applications have significantly improved the prediction accuracy and opened an alternative source of methods for spatial interpolation. Given that they have only been applied to one variable, several questions remain, namely: are they dataset- specific? How reliable are their predictions for different datasets and variables? Could other machine learning methods (such as boosted regression trees) improve the spatial interpolations? To address these questions, we experimentally compared the predictions of several methods for sand content on the southwest Australian marine margin. We tested a variety of existing spatial interpolation methods, machine learning methods and their combinations. In this study, we discuss the experimental results and the value of this advancement in spatial interpolation, visually examine the spatial predictions, and compare the results with the findings in the previous publications. The outcomes of this study can be applied to the spatial prediction of marine and terrestrial environmental variables.

This final paper for the session presents the results of the new draft earthquake hazard assessment for Australia and compares them to the previous AS1170.4 hazard values. Draft hazard maps will be presented for several spectral periods (0.0, 0.2 and 1.0 s) at multiple return periods (500, 2500 and 10,000 years). These maps will be compared with both the current earthquake hazard used in AS1170.4 and with other assessments of earthquake hazard in Australia. In general the hazard in the draft map is higher in the western cratonic parts of Australia than it is in the eastern non-cratonic parts of Australia. Where regional source zones are included, peaks in hazard values in the map are generally comparable to those in the current AS1170.4 map. When seismicity 'hotspot zones are included, as described in the previous paper, several of them produce much higher hazard peaks than any in the AS1170.4 map. However, such hotspots do not affect as large an area as many of those in the current AS1170.4 map. Finally, hazard curves for different cities will also be presented and compared to those predicted by the method outlined in AS1170.4.

Geoscience Australia is supporting the exploration and development of offshore oil and gas resources and establishment of Australia's national representative system of marine protected areas through provision of spatial information about the physical and biological character of the seabed. Central to this approach is prediction of Australia's seabed biodiversity from spatially continuous data of physical seabed properties. However, information for these properties is usually collected at sparsely-distributed discrete locations, particularly in the deep ocean. Thus, methods for generating spatially continuous information from point samples become essential tools. Such methods are, however, often data- or even variable- specific and it is difficult to select an appropriate method for any given dataset. Improving the accuracy of these physical data for biodiversity prediction, by searching for the most robust spatial interpolation methods to predict physical seabed properties, is essential to better inform resource management practises. In this regard, we conducted a simulation experiment to compare the performance of statistical and mathematical methods for spatial interpolation using samples of seabed mud content across the Australian margin. Five factors that affect the accuracy of spatial interpolation were considered: 1) region; 2) statistical method; 3) sample density; 4) searching neighbourhood; and 5) sample stratification by geomorphic provinces. Bathymetry, distance-to-coast and slope were used as secondary variables. In this study, we only report the results of the comparison of 14 methods (37 sub-methods) using samples of seabed mud content with five levels of sample density across the southwest Australian margin. The results of the simulation experiment can be applied to spatial data modelling of various physical parameters in different disciplines and have application to a variety of resource management applications for Australia's marine region.