This research seeks to explain the creation and perpetuation of flood hazards in Peninsular Malaysia in terms of a 'hazard response-in-context' model. Socio-political (socio-cultural and political economy) and institutional contexts are found to be central to understanding hazards as essentially socially-created phenomena superimposed onto a physical process system through which hazards are transmitted. Malaysia is an ex-colonial, newly-industrialising country. The pace of social, economic and political change is fast, as is the pace of technological change. Other things being equal, these are the contexts in which flood hazards are magnified. Contexts are changing, and changing physical systems have given rise to increased flood risk, exposure and vulnerability. Other contexts, largely structural, such as persistent poverty, low residential and occupational mobility, landlessness, and ethnic culture have also contributed to increased vulnerability to flood hazards. The situation, behaviour and response of individual floodplain occupants in Peninsular Malaysia are found to be heavily influenced by macro socio-political contexts. These are also termed contextual forces and they are fundamentally 'structural'. Macro contexts also 'condition' institutions (meso context) and influence their approach to hazard management including their effectiveness. Institutions (including organisations) were found to be largely inadequate in their management and reduction of flood hazards, and can be improved to create positive influences on flood hazard reduction as well as help individuals (micro context) cope more effectively. Both socio-political and institutional contexts were found to be important as they amplify hazards or fail to adequately address and reduce them. The pioneering of what is termed 'segment analysis' to analyse links between contexts at various levels is an important contribution in this research. The research concludes that the hazard response-in-context model is appropriately applied to Peninsular Malaysia as it handles both structural and institutional contexts and individual management of flood hazards effectively.

By providing a rapid and cost effective means of generating sequencing resources for almost any organism, ‘Next generation sequencing technologies’ (NGS) have great potential to help address numerous gene and genome level questions in molecular biology. Progress in NGS is exponentially increasing sequence throughput and large scale studies in the genomics/transcriptomics of non-model organisms are becoming a reality. Therefore the main focus of the work presented in this thesis is on the analysis of the large scale non-model insect datasets generated by NGS technologies and their potential to develop functional genomics tools for these species. Four different NGS datasets from four very different insects the Greenhouse whitefly (Trialeurodes vaporariorum) the Passionvine butterfly (Heliconius melopmene), the blowfly (Lucilia sericata) and the Green Dock beetle (Gastrophysa viridula) were analysed and annotated. Molecular research in these insects has been hindered in the past due to limited nucleotide sequence information. Transcriptome data generated by 454 pyrosequencing was used as a starting point to study the genomics of these ecologically and economically important non-model insect species. The resulting transcriptomes were annotated for gene families involved in xenobiotic metabolism, namely the glutathione-S-transferases (GSTs), cytochrome P450s (P450s) and the carboxylesterases (CCEs). In each case the number and diversity of gene family members is discussed with those documented in other insects. In the case of H. melpomone, the transcriptome data was also used to complement the genomic research by identifying and validating cytochrome P450 gene models in the recently sequenced genome. Furthermore, Illumina generated RNA-seq data was used for SNP characterisation in L. sericata. Transcriptome sequencing is shown to be a useful and cost effective technique to enhance the resources available for non-model organisms as well as for gene discovery in the absence of the reference genomic resources. By focusing on genes involved in xenobiotic metabolism this thesis has isolated numerous candidate genes potentially involved in important processes such as insecticide resistance (Lucilia and Trialeurodes) and host plant exploitation (Gastrophysa and Heliconius). NGS technologies and bioinformatics can thus open up avenues to develop functional genomics resources for diverse species of interest to ecologists and evolutionary biologists.

Biological systems are usually complex nonlinear systems of which we only have a limited understanding. Here we show three different aspects of investigating such systems. We present a method to extract detailed knowledge from typical biological trajectory data, which have randomness as a main characteristic. The migration of immune cells, such as leukocytes, are a key example of our study. The application of our methodology leads to the discovery of novel random walk behaviour of leukocyte migration. Furthermore we use the gathered knowledge to construct the under- lying mathematical model that captures the behaviour of leukocytes, or more precisely macrophages and neutrophils, under acute injury. Any model of a biological system has little predictive power if it is not compared to collected data. We present a pipeline of how complex spatio- temporal trajectory data can be used to calibrate our model of leukocyte migration. The pipeline employs approximate methods in a Bayesian framework. Using the same approach we are able to learn additional information about the underlying signalling network, which is not directly apparent in the cell migration data. While these two methods can be seen as data processing and analysis, we show in the last part of this work how to assess the information content of experiments. The choice of an experiment with the highest information content out of a set of possible experiments leads us to the problem of optimal experimental design. We develop and implement an algorithm for simulation based Bayesian experimental design in order to learn parameters of a given model. We validate our algorithm with the help of toy examples and apply it to examples in immunology (Hes1 transcription regulation) and signal transduction (growth factor induced MAPK pathway).

The existence of `dark matter' throughout the universe is now well established but its form and origin remain one of the greatest problems for modern cosmology. Particle physics posits a solution to this problem in the form of `Weakly Interacting Massive Particles' (WIMPs) which are predicted to exist by many theories extending physics beyond the standard model. The discovery of such particles would consequently have profound implications for both disciplines. Many experiments around the world are now endeavouring to achieve this goal but currently the most successful are those using scintillator detectors. This thesis describes a study of the use of Pulse Shape Discrimination (PSD) techniques to reduce the rate of electron recoil background events in scintillator dark matter experiments. The development of new classes of detector with novel pulse shape properties is described and the results of tests using elastic scattering of monoenergetic neutrons to simulate nuclear recoil signal events are presented. Monte Carlo simulations have been used to assess the performance of CASPAR, a particularly promising new technique, and the results presented here indicate that this has the potential to considerably improve dark matter sensitivity, particularly for spin dependent WIMP interactions. An analysis of 867 kg. days of data from an operational NaI(Tl) detector is described and the resulting evidence for a small population of events with anomalous pulse shape properties discussed.

Stimulated Raman scattering is a convenient way to extend the spectral coverage of well-established solid-state laser sources. In an intracavity Raman laser, a Raman crystal is placed inside the cavity of the fundamental laser, and the output frequency is red-shifted by an amount corresponding to a vibrational energy level of the Raman crystal. Despite the physical simplicity of these lasers, the interactions between the various optical fields are complex, and must be understood in order to realise efficient, high power operation. This thesis presents a detailed investigation of thermal and spectral effects in CW intracavity Raman lasers. A disk geometry was used to reduce the thermal lens in the laser gain crystal, thereby permitting more flexible cavity design. This facilitated experiments to probe and control the thermal and spectral effects. Diamond was assessed as a potential Raman crystal with weak thermal lensing. The optical losses in several crystals were measured and while some low loss material was identified, the supply of such material is not yet reliable. The thermal lens in a common Raman crystal, BaWO4, was measured and found to be negative and astigmatic. Using a coupled cavity configuration, experiments were performed to disentangle the effects of the thermal lenses in the laser gain and Raman crystals. This information was used to re ne the cavity design and improve the performance of the laser in a more systematic way than would otherwise have been possible. It was shown that Raman lasers using laser gain disks can provide comparable performance to rod-based systems. The first ever detailed investigation into spectral broadening in CW crystalline intracavity Raman lasers was undertaken using a combination of theory and experiment. The use of etalons to limit the broadening was investigated and it was found that these could improve the spectral brightness of the laser.

Digital forensics is an emerging field that has uniquely brought together academics, practitioners and law enforcement. Research in this area was inspired by the numerous challenges posed by the increased sophistication of criminal tools. Traditionally, digital forensics has been confined to the extraction of digital evidence from electronic devices. This direct extraction of digital evidence, however, no longer suffices. Indeed, extracting completely raw data without further processing and/or filtering is, in some cases, useless. These problems can be tackled by the so-called "computational forensics" where the reconstructs evidence are undertaken further processing. One important application of computational forensics is criminal tracking, which we collectively call "forensic tracking" and is the main subject of this thesis. This thesis adopts an algorithmic approach to investigate the feasibility of conducting forensic tracking in various environments and settings. Unlike conventional tracking, forensic tracking has to be passive such that the target (who is usually a suspect) should not be aware of the tracking process. We begin by adopting pedestrian setting and propose several online (real-time) forensic tracking algorithms to track a single or multiple targets passively. Beside the core tracking algorithms, we also propose other auxiliary algorithms to improve the robustness and resilience of tracking. We then extend the scope and consider vehicular forensic tracking, where we investigate both online and offline tracking. In online vehicular tracking, we also propose algorithms for motion prediction to estimate the near future movement of target vehicles. Offline vehicular tracking, on the other hand, entails the post-hoc extraction and probabilistic reconstruction of vehicular traces, which we adopt Bayesian approach for. Finally, the contributions of the thesis concludes with building an algorithmic solution for multi-modal tracking, which is a mixed environment combining both pedestrian and vehicular settings.

No description available.

A large collection of digital images of natural scenes provides a database for analyzing and modeling small scene patches (e.g., 2 x 2) referred to as natural microimages. A pivotal ¯nding is the stability of the empirical microimage distribution across scene samples and with respect to scaling. With a view toward potential applications (e.g. classi¯cation, clutter modeling, segmentation), we present a hierarchy of microimage probability models which capture essential local image statistics. Tools from information theory, algebraic geometry and of course statistical hypothesis testing are employed to assess the "match" between candidate models and the empirical distribution. Geometric symmetries play a key role in the model selection process. One central result is that the microimage distribution exhibits reflection and rotation symmetry and is well-represented by a Gibbs law with only pairwise interactions. However, the acceptance of the up-down reflection symmetry hypothesis is borderline and intensity inversion symmetry is rejected. Finally, possible extensions to larger patches via entropy maximization and to patch classification via vector quantization are briefly discussed.

No description available.

Microorganisms were discovered in clouds over 100 years ago but detailed information on community structure and function is severely limited. Clouds may be a niche within which bacteria could thrive and influence dynamic cloud processes using ice nucleating and cloud condensing abilities. Gaining an understanding of the bacterial communities and their possible role in these processes might introduce another discipline into meteorology and climate modelling. Cloud and rain samples were collected in 2003 from Bowbeat Windfarm in the Scottish Moorfoot Hills and two mountains in the Outer Hebrides. Community composition was determined using a combination of amplified 16S ribosomal DNA restriction analysis and sequencing. 100 clones from the Bowbeat sample revealed ten OTUs of which three contained more than two clones. 256 clones from the Hebrides samples revealed 111 OTUs of which 33 contained two or more clones. In all the cloud samples the largest OTUs were identified as fluorescent Pseudomonas sp. To investigate bacterial metabolic activity in clouds a further four cloud samples were collected from Bowbeat in 2006. Reverse transcriptase and quantitative PCR did not definitively reveal metabolic activity in cloud bacteria, however the methodology requires further testing. Heterogeneous nucleation is central to the Bergeron-Findeisen process of raindrop formation. Several bacterial species act as heterogeneous nuclei by producing an ice nucleation (IN) protein. PCR targeting the IN gene of Pseudomonas fluorescens (InaW) in Pseudomonas isolates and cloud DNA did not amplify the IN gene. Freezing cultures using differential scanning calorimetry also failed to reveal the IN phenotype. A finding which evolved from the research was all the fluorescent Pseudomonas cloud isolates displayed biosurfactant activity. Surfactants are very important in the process of activating aerosols into cloud condensation nuclei (CCN). It is also known that surfactants influence cloud droplet size and increase cloud lifetime and albedo. Some bacteria are known to act as CCN and so it is conceivable that these fluorescent pseudomonads could be using surfactants to facilitate their activation from aerosols into CCN. This might allow water scavenging, counter desiccation and aid their dispersal.