David Rohde

Holocene Demographic Changes and the Emergence of Complex Societies in Prehistoric Australia

A continental-scale model of Holocene Australian hunter-gatherer demography and mobility is generated using radiocarbon data and geospatial techniques. Results show a delayed expansion and settlement of much of Australia following the termination of the late Pleistocene until after 9,000 years ago (or 9ka). The onset of the Holocene climatic optimum (9-6ka) coincides with rapid expansion, growth and establishment of regional populations across ~75% of Australia, including much of the arid zone. This diffusion from isolated Pleistocene refugia provides a mechanism for the synchronous spread of pan-continental archaeological and linguistic attributes at this time (e.g. Pama-Nyungan language, Panaramitee art style, backed artefacts). We argue longer patch residence times were possible at the end of the optimum, resulting in a shift to more sedentary lifestyles and establishment of low-level food production in some parts of the continent. The onset of El Niño - Southern Oscillation (ENSO; 4.5-2ka) restricted low-level food production, and resulted in population fragmentation, abandonment of marginal areas, and reduction in ranging territory of ~26%. Importantly, climate amelioration brought about by more pervasive La Niña conditions (post-2ka), resulted in an intensification of the mobility strategies and technological innovations that were developed in the early- to mid-Holocene. These changes resulted in population expansion and utilization of the entire continent. We propose that it was under these demographically packed conditions that the complex social and religious societies observed at colonial contact were formed.

Investigating the association between weather conditions, calendar events and socio-economic patterns with trends in fire incidence: an Australian case study

Fires in urban areas can cause significant economic, physical and psychological damage. Despite this, there has been a comparative lack of research into the spatial and temporal analysis of fire incidence in urban contexts. In this paper, we redress this gap through an exploration of the association of fire incidence to weather, calendar events and socio-economic characteristics in South-East Queensland, Australia using innovative technique termed the quad plot. Analysing trends in five fire incident types, including malicious false alarms (hoax calls), residential buildings, secondary (outdoor), vehicle and suspicious fires, results suggest that risk associated with all is greatly increased during school holidays and during long weekends. For all fire types the lowest risk of incidence was found to occur between one and six a.m. It was also found that there was a higher fire incidence in socially disadvantaged neighbourhoods and there was some evidence to suggest that there may be a compounding impact of high temperatures in such areas. We suggest that these findings may be used to guide the operations of fire services through spatial and temporal targeting to better utilise finite resources, help mitigate risk and reduce casualties.

Matching of catalogues by probabilistic pattern classification

We consider the statistical problem of catalogue matching from a machine learning perspective with the goal of producing probabilistic outputs, and using all available information. A framework is provided that unifies two existing approaches to producing probabilistic outputs in the literature, one based on combining distribution estimates and the other based on combining probabilistic classifiers. We apply both of these to the problem of matching the HI Parkes All Sky Survey radio catalogue with large positional uncertainties to the much denser SuperCOSMOS catalogue with much smaller positional uncertainties. We demonstrate the utility of probabilistic outputs by a controllable completeness and efficiency trade-off and by identifying objects that have high probability of being rare. Finally, possible biasing effects in the output of these classifiers are also highlighted and discussed.

Applying machine learning to catalogue matching in astrophysics

We present the results of applying automated machine learning techniques to the problem of matching different object catalogues in astrophysics. In this study, we take two partially matched catalogues where one of the two catalogues has a large positional uncertainty. The two catalogues we used here were taken from the H I Parkes All Sky Survey (HIPASS) and SuperCOSMOS optical survey. Previous work had matched 44 per cent (1887 objects) of HIPASS to the SuperCOSMOS catalogue. A supervised learning algorithm was then applied to construct a model of the matched portion of our catalogue. Validation of the model shows that we achieved a good classification performance (99.12 per cent correct). Applying this model to the unmatched portion of the catalogue found 1209 new matches. This increases the catalogue size from 1887 matched objects to 3096. The combination of these procedures yields a catalogue that is 72 per cent matched.

Spatial forecasting of residential urban fires: A Bayesian approach

The application of GIS-based techniques to analyse incident data such as crime has received a relatively large amount of research interest, the analysis of disaggregate fire incident data, in comparison has been the focus of much less attention. This paper, for the first time applies a Bayesian methodology to generate disaggregate spatial forecasts of residential household fires across metropolitan South-East Queensland (SEQ), Australia. Expected counts of fire for a one year period are calculated for each suburbs across the SEQ region where the expected risk varies from 2 or less fires per year up to 25 fires per year. The application of the Bayesian forecast methodology has the potential to inform policy decisions both from a reactive, resource allocation perspective and a more proactive perspective, such as through spatial targeting to implement preventative measures to reduce fire risk.

MCMC methods for univariate exponential family models with intractable normalization constants

The exchange algorithm for handling models with intractable partition functions is combined with new methods for adaptive rejection sampling in order to allow Markov chain Monte Carlo methods to sample from the posterior of a new class of exponential family models: exponential of even degree polynomials. It is demonstrated that these models have intuitive properties and can be fit to multimodal univariate datasets. Possible computational benefits of the new approach are contrasted with latent variable methods.

The Sensitivity of the Number of Clusters in a Gaussian Mixture Model to Prior Distributions

One of the main advantages of Bayesian approaches is that they offer principled methods of inference in models of varying dimensionality and of models of infinite dimensionality. What is less widely appreciated is how the model inference is sensitive to prior distributions and therefore how priors should be set for real problems. In this paper prior sensitivity is considered with respect to the problem of inference in Gaussian mixture models. Two distinct Bayesian approaches have been proposed. The first is to use Bayesian model selection based upon the marginal likelihood; the second is to use an infinite mixture model which ‘side steps’ model selection. Explanations for the prior sensitivity are given in order to give practitioners guidance in setting prior distributions. In particular the use of conditionally conjugate prior distributions instead of purely conjugate prior distributions are advocated as a method for investigating prior sensitivity of the mean and variance individually.

Astronomical catalogue matching as a mixture model problem

Astronomical telescopes increasingly operate in surveymode sweeping the sky systematically and producing highly processed data products such as astronomical catalogues which are lists of objects with positional information and other measurements usually including flux in a particular band. An important problem in electronic astronomy is the appropriate way to combine information from different catalogues produced by different telescopes. A key problem in combining this information is to establish different observations of the same object in the two catalogues i.e. the problem of catalogue matching. Positional information is not always sufficient in establishing matches reliably in these cases additional information from the non-positional measurements may also be used. This non-positional information is often scientifically interesting and its inter-catalogue properties may be the main object of study. In previous studies it is argued that while models of non-positional properties may assist in catalogue matching if these properties are scientifically interesting then the conclusions drawn from the analysis may be distorted by using this non-positional information. In this paper it is demonstrated that by employing a predictive Bayesian formalism it is possible to use all available information to assist in obtaining the most reliable matches and still obtain undistorted conclusions. Distortions are avoided because predictive distributions are computed where all the configurations of matches are marginalized over, rather than other approaches which choose a single most likely configuration of matches.

Graphical tools for conditional probabilistic exploration of multivariate spatial datasets

Two exploratory data analysis techniques the comap and the quad plot are shown to have both strengths and shortcomings when analysing spatial multivariate datasets. A hybrid of these two techniques is proposed: the quad map which is shown to overcome the outlined shortcomings when applied to a dataset containing weather information for disaggregate incidents of urban fires. Common to the quad plot, the quad map uses Polya models in order to articulate the underlying assumptions behind histograms. The Polya model formalises the situation in which past fire incident counts are computed and displayed in (multidimensional) histograms as appropriate assessments of conditional probability providing valuable diagnostics such as posterior variance i.e. sensitivity to new information. Finally we discuss how new technology in particular Online Analytics Processing (OLAP) and Geographical Information Systems (GISs) offer potential in automating exploratory spatial data analyses techniques, such as the quad map.

The evidence for smoke alarm type: photoelectric vs ionisation.

The basis of this study emerges from discussions with the Queensland Fire and Emergency Services (QFES) on the results of an extensive network meta-analysis published in Epidemiologic Reviews, which concluded that: “... the ‘best’ intervention for increasing functional smoke alarms is one that provides education, provides and fits low-cost or free equipment, and provides a home safety inspection; and that ionization alarms with lithium batteries are the ‘best’ type in terms of increasing the prevalence of functional smoke alarms.” 1(p42) These recommendations are, in one small but important aspect, out of step with the recommendations of important fire safety bodies such as Australasian Fire Authorities Council2 and The International Association of Fire Fighters;3 the difference being that these organisations advocate photoelectric smoke alarms rather than ionisation smoke alarms.