Natural hazards such as floods, dam breaks, storm surges and tsunamis impact communities around the world every year. To reduce the impact, accurate modelling is required to predict where water will go, and at what speed, before the event has taken place. ANUGA is free, open source, software created to model water flow arising from these events. The resulting knowledge can be used to reduce loss of life and damage to property in communities affected by such disasters by providing vital input to evacuation plans, structural mitigation options and planning. The software was developed collaboratively by the Australian National University (ANU) and Geoscience Australia (GA) and is available at http://sourceforge.net/projects/anuga. ANUGA solves the non-linear shallow water wave equations using the finite volume method with dynamic time stepping. A major capability of ANUGA is that it can model the process of wetting and drying as water enters and leaves an area. This means it is suitable for simulating water flow onto a beach or dry land and around structures such as buildings. ANUGA is also capable of modelling complex flows involving shock waves and rapidly changing flow speeds (transitions from sub critical to super critical flows). ANUGA is a robust software package that contains over 800 unit tests. It has been validated against wave tank experiments [1] and model outputs from the 2004 Indian Ocean tsunami have compared very well with a run-up survey at Patong Beach, Thailand. This particular activity has also underpinned the results provided to Australian emergency managers managing tsunami risk. This presentation will outline the key components of ANUGA, examples for a range of hydrodynamic hazards as well as a sample of validation outputs.

Geoscience Australia often produces spatially continuous marine environmental information products using spatial interpolation methods. The accuracy of such information is critical for well-informed decisions for marine environmental management and conservation. Improving the accuracy of these data products by searching for robust methods is essential, but it is a vexed task since no method is best for all variables. Therefore, we experimentally compared the performance of 32 methods/sub-methods using seabed gravel content data from the Australian continental EEZ. In this study, we have identified and developed several novel and robust methods that significantly increase the accuracy of interpolated spatial information. Moreover, these methods can be applied to various environmental properties in both marine and terrestrial disciplines.

Keynote presentation to cover * the background to tsunami modelling in Australia * what the modelling showed * why the modelling is important to emergency managers * the importance of partnerships * future challenges

In this study, we aim to identify the most appropriate methods for spatial interpolation of seabed sand content for the AEEZ using samples extracted on August 2010 from Geoscience Australia's Marine Samples Database. The predictive accuracy changes with methods, input secondary variables, model averaging, search window size and the study region but the choice of mtry. No single method performs best for all the tested scenarios. Of the 18 compared methods, RFIDS and RFOK are the most accurate methods in all three regions. Overall, of the 36 combinations of input secondary variables, methods and regions, RFIDS, 6RFIDS and RFOK were among the most accurate methods in all three regions. Model averaging further improved the prediction accuracy. The most accurate methods reduced the prediction error by up to 7%. RFOKRFIDS, with a search window size of 5, an mtry of 4 and more realistic predictions in comparison with the control, is recommended for predicting sand content across the AEEZ if a single method is required. This study provides suggestions and guidelines for improving the spatial interpolations of marine environmental data.

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.

Following the tragic events of the Indian Ocean tsunami on 26 December 2004 it became obvious there were shortcomings in the response and alert systems for the threat of tsunami to Western Australia's (WA) coastal communities. The relative risk of a tsunami event to the towns, remote indigenous communities, and infrastructure for the oil, gas and mining industries was not clearly understood in 2004. Consequently, no current detailed response plans for a tsunami event in WA coastal areas existed. The Boxing Day event affected the WA coastline from Bremer Bay on the south coast, to areas north of Exmouth on the north-west coast, with a number of people requiring rescue from abnormally strong currents and rips. There were also reports of personal belongings at some beaches inundated by wave activity. More than 30 cm of water flowed down a coast-side road in Geraldton on the mid-west coast, and Geordie Bay at Rottnest Island (19 km of the coast of Fremantle) experienced five 'tides' in three hours, resulting in boats hitting the ocean bed a number of times. The vivid images of the devastation caused by the 2004 event across a wide geographical area changed the perception of tsunami and achieved an appreciation of the potential enormity of impact from this low frequency but high consequence natural hazard. With WA's proximity to the Sunda Arc, which is widely recognised as a high probability area for intra-plate earthquakes, the need to develop a better understanding of tsunami risk and model the potential social and economic impacts on communities and critical infrastructure along the Western Australian coast, became a high priority. Under WA's emergency management arrangements, the Fire and Emergency Services Authority (FESA) has responsibility for ensuring effective emergency management is in place for tsunami events across the PPRR framework.

Within the general trend of post-Eocene cooling, the largest and oldest outlet of the East Antarctic Ice Sheet underwent a change from ice-cliff to ice-stream and/or ice-shelf dynamics, with an associated switch from line-source to fan sedimentation. Available geological data reveal little about the causes of these changes in ice dynamics during the Miocene Epoch, or the subsequent effects on Pliocene-Pleistocene ice-sheet history. Ice-sheet numerical modeling reveals that bed morphology was probably responsible for driving changes in both ice-sheet extent and dynamics in the Lambert-Amery system at Prydz Bay. The modeling shows how the topography and bathymetry of the Lambert graben and Prydz Bay control ice-sheet extent and flow. The changes in bathymetric volume required for shelf-edge glaciation correlate well with the Prydz Channel fan sedimentation history. This suggests a negative feedback between erosion and glaciation, whereby the current graben is overdeepened to such an extent that shelf-edge glaciation is now not possible, even if a Last Glacial Maximum environment recurs. We conclude that the erosional history of the Lambert graben and Prydz Bay in combination with the uplift histories of the surrounding mountains are responsible for the evolution of this section of the East Antarctic Ice Sheet, once the necessary initial climatic conditions for glaciation were achieved at the start of the Oligocene Epoch.

Geoscience Australia has collaboratively developed a number of open source software models and tools to estimate hazard, impact and risk to communicaties for a range of natural hazard to support disaster risk reduction in Australia and the region. These models and tools include: * ANUGA * EQRM * TCRM * TsuDAT * RICS * FiDAT This presentation will discuss the drivers for developing these models and tools using open source software and the benefits to the end-users in the emergency management and planning community as well as the broader research community. Progress and plans for these models and tools will also be outlined in particular those that take advantage of the availability of high performance computing, cloud computing, webservices and global initiatives such as the Global Earthquake Model.

Machine learning methods, like random forest (RF), have shown their superior performance in various disciplines, but have not been previously applied to the spatial interpolation of environmental variables. In this study, we compared the performance of 23 methods, including RF, support vector machine (SVM), ordinary kriging (OK), inverse distance squared (IDS), and their combinations (i.e., RFOK, RFIDS, SVMOK and SVMIDS), using mud content samples in the southwest Australian margin. We also tested the sensitivity of the combined methods to input variables and the accuracy of averaging predictions of the most accurate methods. The accuracy of the methods was assessed using a 10-fold cross-validation. The spatial patterns of the predictions of the most accurate methods were also visually examined for their validity. This study confirmed the effectiveness of RF, especially its combination with OK or IDS, and also confirmed the sensitivity of RF and its combined methods to the input variables. Averaging the predictions of the most accurate methods showed no significant improvement in the predictive accuracy. Visual examination proved to be an essential step in assessing the spatial predictions. This study has opened an alternative source of methods for spatial interpolation of environmental properties.

In this study, we conducted a simulation experiment to identify robust spatial interpolation methods using samples of seabed mud content in the Geoscience Australian Marine Samples database. Due to data noise associated with the samples, criteria are developed and applied for data quality control. Five factors that affect the accuracy of spatial interpolation were considered: 1) regions; 2) statistical methods; 3) sample densities; 4) searching neighbourhoods; and 5) sample stratification. Bathymetry, distance-to-coast and slope were used as secondary variables. Ten-fold cross-validation was used to assess the prediction accuracy measured using mean absolute error, root mean square error, relative mean absolute error (RMAE) and relative root mean square error. The effects of these factors on the prediction accuracy were analysed using generalised linear models. The prediction accuracy depends on the methods, sample density, sample stratification, search window size, data variation and the study region. No single method performed always superior in all scenarios. Three sub-methods were more accurate than the control (inverse distance squared) in the north and northeast regions respectively; and 12 sub-methods in the southwest region. A combined method, random forest and ordinary kriging (RKrf), is the most robust method based on the accuracy and the visual examination of prediction maps. This method is novel, with a relative mean absolute error (RMAE) up to 17% less than that of the control. The RMAE of the best method is 15% lower in two regions and 30% lower in the remaining region than that of the best methods in the previously published studies, further highlighting the robustness of the methods developed. The outcomes of this study can be applied to the modelling of a wide range of physical properties for improved marine biodiversity prediction. The limitations of this study are discussed. A number of suggestions are provided for further studies.