A random forest for lung nodule identification

A method is presented for identification of lung nodules. It includes three stages: image acquisition, background removal, and nodule detection. The first stage improves image quality. The second stage extracts long lobe regions. The third stage detects lung nodules. The method is based on the random forest learner. Training set contains nodule, non-nodule, and false-positive patterns. Test set contains randomly selected images. The developed method is compared against the support vector machine. True-positives of 100% and 85.9%, and false-positives of 1.27 and 1.33 per image were achieved by the developed method and the support vector machine, respectively.

Market makers v's the general public : a first look at S & P500 futures trade data

This paper considers 15 minute records of trading volume and traded prices coinciding with the reporting intervals required by the Commodity Futures Trading Commission. Records are extracted from trade records for market trade and also two way trade between market makers (CT1) and the general public (CT4) from January 1994 to June 2004. Futures price records are matched with S&P500 cash index price records. Simultaneous volatility models are specified and estimated to test trading volume to futures volatility lead/lag effects and also futures volatility to cash index volatility lead/lag effects. As we disaggregate from the market records to CT1 and CT4 records and further into year to year samples volume to futures volatility leading effects and also futures volatility to cash volatility leading effects dominate. The results raise important issues for risk management and dynamic hedging models employing intra-day trader data. A number of important issues for further analysis are also raised in this paper.

The Samuelson hypothesis in futures markets : an analysis using intraday data

This paper considers the Samuelson hypothesis, which argues that the futures price volatility increases as the futures contract approaches its expiration. Utilizing intraday data from 20 futures markets in six futures exchanges, we find strong support for the Samuelson hypothesis in agricultural futures. However, the Samuelson hypothesis does not hold for other futures contracts. We also provide supporting evidence that the ‘negative covariance’ hypothesis is the key factor for the empirical support of the Samuelson hypothesis. In addition, our findings remain largely unaltered even after we control for seasonality and liquidity effects.

Realized volatility and correlation in grain futures markets: testing for spillover effects

Fluctuations in commodity prices are a major concern to many market participants. This paper uses realized volatility methods to calculate daily volatility and correlation estimates for three grain futures prices (corn, soybean, and wheat). The realized volatility estimates exhibit properties consistent with the stylized facts observed in earlier studies. According to daily realized correlations and regression coefficients, the spot returns from the three grain futures are positively related. The realized estimates are then used to evaluate the degree of volatility transmission across grain futures prices. The impulse response analysis is conducted by fitting the vector autoregressive model to realized volatility and correlation estimates, using the bootstrap method for statistical inference. The results indicate that rich dynamic interactions exist among the volatilities and correlations across the grain futures markets.

Dispersion of information or market behaviour : general public trading in S & P 500 index futures

This paper considers 15 minute records of trading volume and traded prices coinciding with the reporting intervals required by the Commodity Futures Trading Commission. Records are extracted from trade records for two way trade between market makers (CTI1) and the general public (CTI4) from January 1994 to June 2004. Futures price records are matched with S&P500 cash index price records. Simultaneous volatility models are specified and estimated to test trading volume to futures volatility lead/lag effects and also futures volatility to cash index volatility lead/lag effects. There is evidence that existing theoretical models of the general public trading behaviour do not explain such behaviour in these very actively traded markets. These effects can depend more on market conditions than what is suggested in theoretical models.

The trading dynamics of close-substitute futures markets : evidence of margin policy spillover effects

Volatility spillover is well documented among closely related securities. I investigate the relationship between margin policy and trading dynamics of the Nikkei 225 index futures markets of Osaka Securities Exchange (OSE) and Singapore Exchange (SGX). I find that OSE’s margin policy influences trading dynamics across both markets, although it is the less liquid SGX market that performs price discovery. This suggests that policy markers of close substitute markets should coordinate, or at least communicate policy intentions due to policy spillover. SGX’s market design facilitates price discovery, suggesting that a microstructure framework capable of overcoming the liquidity entry barrier is of interest to any futures exchange contemplating contract proliferation.

Economic linkages across commodity futures : hedging and trading implications

We investigate cross-market trading dynamics in futures contracts written on seemingly unrelated commodities that are consumed by a common industry. On the Tokyo Commodity Exchange, we find such evidence in natural rubber (NR), palladium (PA) and gasoline (GA) futures markets. The automobile industry is responsible for more than 50% of global demand for each of these commodities. VAR estimation reveals short-run cross-market interaction between NR and GA, and from NR to PA. Cross-market influence exerted by PA is felt in longer dynamics, with PA volatility (volume) affecting NR (GA) volume (volatility). Our findings are robust to lag-specification, volatility measure, and consistent with full BEKK-GARCH estimation results. Further analysis, which benchmarks against silver futures market, TOCOM index and TOPIX transportation index, confirms that our results are driven by a common industry exposure, and not a commodity market factor. A simple trading rule that incorporates short-run GA and long-run PA dynamics to predict NR return yields positive economic profit. Our study offers new insights into how commodity and equity markets relate at an industry level, and implications for multi-commodity hedging.

Differing responses to landscape change : implications for small mammal assemblages in forest fragments

Human modification of landscapes typically results in many species being confined to small, isolated and degraded habitat fragments. While fragment size and isolation underpin many studies of modified landscapes, vegetation characteristics are less frequently incorporated. The relative influence of biogeographic (e.g. size, isolation) and vegetation parameters on assemblages is poorly understood, but critical for conservation management. In this study, a multiple hypothesis testing framework was used to determine the relative importance of biogeographic and vegetation parameters in explaining the occurrence of an assemblage of small mammals in 48 forest fragments in an agricultural landscape in south-eastern Australia. Fragment size and vegetation characteristics were consistently important predictors of occurrence across species. In contrast, fragment isolation was important for just one native species. Differing abilities of species to move through the landscape provide a reasonable explanation for these results. We conclude that for effective conservation of assemblages, it is important to: (1) consider differing responses of species to landscape change, and (2) move beyond a focus primarily on spatial attributes (size, isolation) to recognise that landscape change also has profound effects on habitat composition and quality.

Ecological value of riparian zones to birds in forest landscapes

Riparian zones are a characteristic component of many landscapes throughout the world and increasingly are valued as key areas for biodiversity conservation. Their importance for bird communities has been well recognised in semi-arid environments and in modified landscapes where there is a marked contrast between riparian and adjacent non-riparian vegetation. The value of riparian zones in largely intact landscapes with continuous vegetation cover is less well understood. This research examined the importance of riparian habitats for avifauna conservation by investigating the ecological interactions contributing to the pattern of bird assemblages in riparian and adjacent non-riparian habitats. Specifically, the focus is on the bird assemblages of riparian zones and those of adjacent non-riparian vegetation types and the influence that associated differences in resource availabilities, habitat structure and conditions have on observed patterns. This study was conducted in the foothill forests of the Victorian Highlands, south-east Australia. Mixed-species eucalypt (genus Eucalyptus) forests dominate the vegetation of this region. Site selection was based on the occurrence of suitable riparian habitat interspersed within extensive, relatively undisturbed (i.e. no recent timber harvesting or fire events) forest mosaics. A series of 30 paired riparian and non-riparian sites were established among six stream systems in three forest areas (Bunyip State Park, Kinglake National Park and Marysville State Forest). Riparian sites were positioned alongside the stream and the non-riparian partner site was positioned on a facing slope at a distance of approximately 750 m. Bird surveys were carried out during 29 visits to each site between July 2001 and December 2002. Riparian sites were floristically distinct from non-riparian sites and had a more complex vegetation structure, including a mid-storey tree layer mostly absent from non-riparian sites, extensive fine litter and coarse woody debris, and dense ground-layer vegetation (e.g. sedges and ground ferns). The characteristic features of non-riparian habitats included a relatively dense canopy cover, a ground layer dominated by grasses and fine litter, and a high density of canopy-forming trees in the smaller size-classes. Riparian zones supported a significantly greater species richness, abundance and diversity of birds when compared to non-riparian habitats. The composition of bird assemblages differed significantly between riparian and non-riparian habitats, with riparian assemblages displaying a higher level of similarity among sites. The strongest contributors to observed dissimilarities between habitat types included species that occurred exclusively in either habitat type or species with large contrasts in abundance between habitat types. Much of the avifauna (36%) of the study area is composed of species that are common and widespread in south-east Australia (i.e. forest generalists). Riparian habitats were characterised by a suite of species more typical of wetter forest types in south-east Australia and many of these species had a restricted distribution in the forest mosaic. Some species (7%) occurred exclusively in riparian habitats (i.e. riparian selective species) while others (43%) were strongly linked to these habitats (i.e. riparian associated species). A smaller proportion of species occurred exclusively (2%) in non-riparian habitats (i.e. non-riparian selective species) or were strongly linked to these habitats (10%; i.e. non-riparian associated species). To examine the seasonal dynamics of assemblages, the variation through time in species richness, abundance and composition was compared between riparian and non-riparian sites. Riparian assemblages supported greater richness and abundance, and displayed less variation in these parameters, than non-riparian assemblages at all times. The species composition of riparian assemblages was distinct from non-riparian assemblages throughout the annual cycle. An influx of seasonal migrants elevated species richness and abundance in the forest landscape during spring and summer. The large-scale movement pattern (e.g. coastal migrant, inland migrant) adopted by migrating species was associated with their preference for riparian or non-riparian habitats in the landscape. Species which migrate north-south along the east coast of mainland Australia (i.e. coastal migrants) used riparian zones disproportionately; eight of eleven species were riparian associated species. Species which migrate north-south through inland Australia (i.e. inland migrants) were mostly associated with non-riparian habitats. The significant differences in the dynamics of community structure between riparian and non-riparian assemblages shows that there is a disproportionate use of riparian zones across the landscape and that they provide higher quality habitat for birds throughout the annual cycle. To examine the ecological mechanisms by which riparian assemblages are richer and support more individual birds, the number of ecological groups (foraging, nest-type and body mass groups) represented, and the species richness of these groups, was compared between riparian and non-riparian assemblages. The structurally complex vegetation and distinctive habitat features (e.g. aquatic environments, damp sheltered litter) provided in the riparian zone, resulted in the consistent addition of ecological groups to riparian assemblages (e.g. sheltered ground – invertebrates foraging group) compared with non-riparian assemblages. Greater species richness was accommodated in most foraging, nest-type and body mass groups in riparian than non-riparian assemblages. Riparian zones facilitated greater richness within ecological groups by providing conditions (i.e. more types of resources and greater abundance of resources) that promoted ecological segregation between ecologically similar species. For a set of commonly observed species, significant differences in their use of structural features, substrates and heights were registered between riparian and non-riparian habitats. The availability and dynamics of resources in riparian and non-riparian habitats were examined to determine if there is differential availability of particular resources, or in their temporal availability, throughout the annual cycle. Riparian zones supported more abundant and temporally reliable eucalypt flowering (i.e. nectar) than non-riparian habitats throughout the annual cycle. Riparian zones also supported an extensive loose bark resource (an important microhabitat for invertebrates) including more peeling bark and hanging bark throughout the year than at non-riparian sites. The productivity of eucalypts differed between habitat types, being higher in riparian zones at most times for all eucalypts combined, and for some species (e.g. Narrow-leaved Peppermint Eucalyptus radiata). Non-riparian habitats provided an abundant nectar resource (i.e. shrub flowering) at particular periods in the annual cycle. Birds showed clear relationships with the availability of specific food (i.e. nectar) and foraging resources (i.e. loose bark). The demonstration of a greater abundance of resources and higher primary productivity in riparian zones is consistent with the hypothesis that these linear strips that occupy only a small proportion of the landscape have a disproportionately high value for birds. Riparian zones in continuous eucalypt forest provide high quality habitats that contribute to the diversity of habitats and resources available to birds in the forest mosaic, with positive benefits for the landscape-level species pool. Despite riparian and non-riparian habitat supporting distinct assemblages of birds, strong linkages are maintained along the riparian-upslope gradient. Clearly, the maintenance of diverse and sustainable assemblages of birds in forest landscapes depends on complementary management of both riparian and non-riparian vegetation.

Effect of Phytophthora cinnamomi on the habitat utilization of Antechinus stuartii in a Victorian forest

Phytophthora cinnamomi (Cinnamon fungus) is a pathogenic soil fungus which infects plant communities along the south-eastern coast of Australia, and the south-western corner of Western Australia. The symptoms of this disease include chlorosis, death of branches (ie. ‘dieback’), retarded growth and the eventual death of infected plants. This leads to devastating effects upon plant communities by altering both the structural and floristic characteristics of these communities. Small mammal species are dependent on specific features of their habitat such as vegetation structure and floristics. This thesis investigated alterations to the habitat of the insectivorous marsupial mouse, Antechinus stuartii, due to the presence of P. cinnamomi. The study was undertaken in an area of an open forest in the Brisbane Ranges, Victoria. Significant changes were found in both the floristic composition and structure of the vegetation at study sites infected with P, cinnamomi, compared to uninfected sites. The habitat utilization by A. stuartii of uninfected and infected vegetation was investigated using live trapping and radio-telemetric techniques. Capture rates were higher at sites uninfected by P. cinnamomi, and both male and females selected areas free from infection. Home range areas of males were significantly larger than those of females as assessed by telemetry. Both sexes spent a high proportion of time in areas dominated by Xanthorrhoea australis (Austral grass tree). There were significant relationships between the abundance of A. stuartii and the denseness of vegetation above 1 metre in height, and in particular, the proportion of cover afforded by X. australis. There were no significant differences in the cover of Eucalyptus spp. between uninfected and infected sites, but there were significantly more nest hollows in infected areas. The abundance of invertebrates was examined using pitfall traps. There were no significant differences in the abundance of the larger invertebrate taxa at infected and uninfected sites, but higher abundances of some micro-invertebrate groups in infected areas were recorded. The most likely factors considered to be influential in the habitat selection of A. stuartii were vegetation structure, and the presence of X. australis. To assess whether these factors were important the leaves of X. australis were removed with a brushcutter, to mimic the early effects of infection with P. cinnamomi. Animals did not respond to the alteration of vegetation structure in the short term (3-4 days). Longer-term experiments are required to assess the habitat utilization of A. stuartii at different periods following habitat manipulation. The implications of the presence of P. cinnamomi on the conservation of fauna are discussed. The destructive nature of the pathogen, and the slow rate of recovery from the disease, means that P. cinnamomi can be considered a threatening process to plant communities and the fauna that reside within that habitat. Future management of this disease within natural areas must therefore be cognisant of the potential of P. cinnamomi to significantly affect faunal as well as vegetative communities.

Gap-crossing decisions of forest birds in a fragmented landscape

Habitat loss and fragmentation are recognized as primary drivers of biodiversity loss worldwide. To understand the functional effects of habitat fragmentation on bird populations, data on movement across gaps in habitat cover are necessary, although rarely available. In this study, we used call playback to simulate a conspecific territorial intruder to entice birds to move through the landscape in a predictable and directional manner. We then quantified the probability of movement in continuous forest and across cleared gaps for two forest-dependent species, the grey shrike-thrush (Colluricincla harmonica) and the white-throated treecreeper (Cormobates leucophaeus). Fifty-four playback trials were conducted for each species across distances ranging from 25 to 480 m in continuous forest and 15-260 m across gaps in a forest-agricultural landscape in southern Victoria, Australia. The probability of movement was significantly reduced by gaps in forest cover for both species. Shrike-thrushes were six times more likely to move 170 m in continuous forest than to cross 170-m gaps. The mean probability that treecreepers would cross any gap at all was less than 0.5, and they were three times less likely to move 50 m across a gap than through continuous forest. Both species displayed non-linear responses to increasing gap distance: we identified a gap-tolerance threshold of 85 m for the shrike-thrush and 65 m for the treecreeper beyond which individuals were most unlikely to cross. The presence of scattered paddock trees increased functional connectivity for the shrike-thrush, with individuals crossing up to 260 m when scattered trees were present. We conclude that gaps in habitat cover are barriers to movement, and that characteristics of the intervening matrix influence landscape permeability.