Grant Hamilton

An Integrated Bayesian Network approach to Lyngbya majuscula bloom initiation.

Blooms of the cyanobacteria Lyngbya majuscula have occurred for decades around the world. However, with the increase in size and frequency of these blooms, coupled with the toxicity of such algae and their increased biomass, they have become substantial environmental and health issues. It is therefore imperative to develop a better understanding of the scientific and management factors impacting on Lyngbya bloom initiation. This paper suggests an Integrated Bayesian Network (IBN) approach that facilitates the merger of the research being conducted by various parties on Lyngbya. Pivotal to this approach are two Bayesian networks modelling the management and scientific factors of bloom initiation. The research found that Bayesian Networks (BN) and specifically Object Oriented BNs (OOBN) and Dynamic OOBNs facilitate an integrated approach to modelling ecological issues of concern. The merger of multiple models which explore different aspects of the problem through an IBN approach can apply to many multi-faceted environmental problems.

Investigating the Use of a Bayesian Network to Model the Risk of Lyngbya majuscula Bloom Initiation in Deception Bay, Queensland, Australia

ABSTRACT Modelling the risk factors driving an environmental problem can be problematic when published data describing variables and their interactions are sparse. In such cases, expert opinion forms a vital source of information. Here we demonstrate the utility of a Bayesian Net (BN) model to integrate available information in a risk analysis setting. As an example, we use this methodology to explore the major factors influencing initiation of Lyngbya majuscula blooms in Deception Bay, Queensland, Australia. Over the past decade Lyngbya blooms have increased in both frequency and extent on seagrass beds in Deception Bay, with a range of adverse effects. This model was used to identify the main factors that could trigger a Lyngbya bloom. The five factors found to have the greatest effect on Lyngbya bloom initiation were: the available nutrient pool, water temperature, redox state of the sediments, current velocity, and light. Scenario analysis was also conducted to determine the sensitivity of the model to different combinations of variable states. The model has been used to identify knowledge gaps and therefore to direct additional research efforts in Deception Bay. With minor changes the model can be used to better understand the factors triggering Lyngbya blooms in other coastal regions.

Bayesian model averaging for harmful algal bloom prediction.

Harmful algal blooms (HABs) are a worldwide problem that have been increasing in frequency and extent over the past several decades. HABs severely damage aquatic ecosystems by destroying benthic habitat, reducing invertebrate and fish populations, and affecting larger species such as dugong that rely on seagrasses for food. Few statistical models for predicting HAB occurrences have been developed, and in common with most predictive models in ecology, those that have been developed do not fully account for uncertainties in parameters and model structure. This makes management decisions based on these predictions more risky than might be supposed. We used a probit time series model and Bayesian model averaging (BMA) to predict occurrences of blooms of Lyngbya majuscula, a toxic cyanophyte, in Deception Bay, Queensland, Australia. We found a suite of useful predictors for HAB occurrence, with temperature figuring prominently in models with the majority of posterior support, and a model consisting of the single covariate, average monthly minimum temperature, showed by far the greatest posterior support. A comparison of alternative model averaging strategies was made with one strategy using the full posterior distribution and a simpler approach that utilized the majority of the posterior distribution for predictions but with vastly fewer models. Both BMA approaches showed excellent predictive performance with little difference in their predictive capacity. Applications of BMA are still rare in ecology, particularly in management settings. This study demonstrates the power of BMA as an important management tool that is capable of high predictive performance while fully accounting for both parameter and model uncertainty.

An approximate Bayesian computation approach for estimating parameters of complex environmental processes in a cellular automata

Modelling an environmental process involves creating a model structure and parameterising the model with appropriate values to accurately represent the process. Determining accurate parameter values for environmental systems can be challenging. Existing methods for parameter estimation typically make assumptions regarding the form of the Likelihood, and will often ignore any uncertainty around estimated values. This can be problematic, however, particularly in complex problems where Likelihoods may be intractable. In this paper we demonstrate an Approximate Bayesian Computational method for the estimation of parameters of a stochastic CA. We use as an example a CA constructed to simulate a range expansion such as might occur after a biological invasion, making parameter estimates using only count data such as could be gathered from field observations. We demonstrate ABC is a highly useful method for parameter estimation, with accurate estimates of parameters that are important for the management of invasive species such as the intrinsic rate of increase and the point in a landscape where a species has invaded. We also show that the method is capable of estimating the probability of long distance dispersal, a characteristic of biological invasions that is very influential in determining spread rates but has until now proved difficult to estimate accurately.

Improving detection probabilities for pests in stored grain.

BACKGROUND The presence of insects in stored grain is a significant problem for grain farmers, bulk grain handlers and distributors worldwide. Inspection of bulk grain commodities is essential to detect pests and thereby to reduce the risk of their presence in exported goods. It has been well documented that insect pests cluster in response to factors such as microclimatic conditions within bulk grain. Statistical sampling methodologies for grain, however, have typically considered pests and pathogens to be homogeneously distributed throughout grain commodities. In this paper, a sampling methodology is demonstrated that accounts for the heterogeneous distribution of insects in bulk grain. RESULTS It is shown that failure to account for the heterogeneous distribution of pests may lead to overestimates of the capacity for a sampling programme to detect insects in bulk grain. The results indicate the importance of the proportion of grain that is infested in addition to the density of pests within the infested grain. It is also demonstrated that the probability of detecting pests in bulk grain increases as the number of subsamples increases, even when the total volume or mass of grain sampled remains constant. CONCLUSION This study underlines the importance of considering an appropriate biological model when developing sampling methodologies for insect pests. Accounting for a heterogeneous distribution of pests leads to a considerable improvement in the detection of pests over traditional sampling models.

From Science to Management: Using Bayesian Networks to Learn about "Lyngbya"

Toxic blooms of Lyngbya majuscula occur in coastal areas worldwide and have major ecological, health and economic consequences. The exact causes and combinations of factors which lead to these blooms are not clearly understood. Lyngbya experts and stakeholders are a particularly diverse group, including ecologists, scientists, state and local government representatives, community organisations, catchment industry groups and local fishermen. An integrated Bayesian network approach was developed to better understand and model this complex environmental problem, identify knowledge gaps, prioritise future research and evaluate management options.

Sampling stored‐product insect pests: a comparison of four statistical sampling models for probability of pest detection

BACKGROUND Developing sampling strategies to target biological pests such as insects in stored grain is inherently difficult owing to species biology and behavioural characteristics. The design of robust sampling programmes should be based on an underlying statistical distribution that is sufficiently flexible to capture variations in the spatial distribution of the target species. RESULTS Comparisons are made of the accuracy of four probability-of-detection sampling models - the negative binomial model,(1) the Poisson model,(1) the double logarithmic model(2) and the compound model(3) - for detection of insects over a broad range of insect densities. Although the double log and negative binomial models performed well under specific conditions, it is shown that, of the four models examined, the compound model performed the best over a broad range of insect spatial distributions and densities. In particular, this model predicted well the number of samples required when insect density was high and clumped within experimental storages. CONCLUSIONS This paper reinforces the need for effective sampling programs designed to detect insects over a broad range of spatial distributions. The compound model is robust over a broad range of insect densities and leads to substantial improvement in detection probabilities within highly variable systems such as grain storage.

Giant coral reef fishes display markedly different susceptibility to night spearfishing

Abstract The humphead wrasse (Cheilinus undulatus) and bumphead parrotfish (Bolbometopon muricatum) are two of the largest, most iconic fishes of Indo‐Pacific coral reefs. Both species form prized components of subsistence and commercial fisheries and are vulnerable to overfishing. C. undulatus is listed as Endangered and B. muricatum as Vulnerable on the IUCN Red List of Threatened Species. We investigated how night spearfishing pressure and habitat associations affected both species in a relatively lightly exploited setting; the Kia fishing grounds, Isabel Province, Solomon Islands. We used fisheries‐independent data from underwater visual census surveys and negative binomial models to estimate abundances of adult C. undulatus and B. muricatum as a function of spearfishing pressure and reef strata. Our results showed that, in Kia, night spearfishing pressure from free divers had no measurable effect on C. undulatus abundances, but abundances of B. muricatum were 3.6 times lower in areas of high spearfishing pressure, after accounting for natural variations due to habitat preferences. It is likely the species’ different nocturnal aggregation behaviors, combined with the fishers’ use of night spearfishing by spot‐checking underpin these species’ varying susceptibility. Our study highlights that B. muricatum is extremely susceptible to night spearfishing; however, we do not intend to draw conservation attention away from C. undulatus. Our data relate only to the Kia fishing grounds, where human population density is low, the spot‐checking strategy is effective for reliably spearing large numbers of fish, particularly B. muricatum, and fisheries have only recently begun to be commercialized; such conditions are increasingly rare. Instead, we recommend that regional managers assess the state of their fisheries and the dynamics affecting the vulnerability of the fishes to fishing pressure based on local‐scale, fisheries‐independent data, where resources permit.

Efficacy and safety of Canagliflozin in subjects with type 2 diabetes mellitus on background metformin

Question: Canagliflozin (CANA) is a sodium glucose co-transporter 2 inhibitor in development for the treatment of type 2 diabetes mellitus (T2DM). This study evaluated the efficacy and safety of CANA in subjects with T2DM inadequately controlled with metformin (MET). Methodology: In this randomised, double-blind, Phase 3 study, subjects with T2DM on stable MET (N = 1,284) received CANA 100 or 300 mg, sitagliptin (SITA) 100 mg, or placebo (PBO) daily (2:2:2:1) for a 26-week, PBO- and active-controlled period (results reported here) followed by a 26-week, active-controlled period (PBO subjects switched to SITA; results to be reported elsewhere). Primary endpoint was change from baseline in HbA1c at Week 26 for CANA versus PBO. Secondary endpoints included proportion of subjects reaching HbA1c < 7.0%, change in fasting plasma glucose (FPG), 2-h postprandial glucose (PPG), and systolic BP, and percent change in body weight, HDL-C, and triglycerides. Statistical comparisons for SITA versus PBO or CANA at Week 26 were not performed (not pre-specified). Adverse events (AEs) were recorded throughout the study. Results: Mean baseline characteristics were similar across groups (age, 55.4 y; HbA1c, 7.9%; FPG, 9.4 mmol/L; BMI, 31.8 kg/m2). At Week 26, CANA 100 and 300 mg reduced HbA1c relative to PBO (-0.62% and -0.77%; P< 0.001), with a decrease of -0.66% with SITA versus PBO. More subjects reached HbA1c < 7.0% with CANA 100 and 300 mg and SITA than PBO (45.5%, 57.8%, 54.5%, 29.8%; P= 0.0 for both CANA doses vs. PBO). CANA 100 and 300 mg and SITA decreased FPG (-1.7, -2.2, -1.3 mmol/L) and 2-h PPG (-2.1, -2.6, -2.2 mmol/L) versus PBO (P< 0.001 for both CANA doses vs. PBO). CANA 100 and 300 mg reduced body weight relative to PBO (-2.5% and -2.9%; P< 0.001); no change was seen with SITA. CANA 100 and 300 mg and SITA were associated with decreased systolic BP (-5.4, -6.6, -3.3 mmHg) and increased HDL-C (6.6%, 8.4%, 1.3%) relative to PBO (P< 0.001 for both CANA doses vs. PBO); LDL-C was increased versus PBO (7.9%, 12.2%, 5.5%). Overall AE rates were modestly higher with CANA 100 mg (61.1%) than CANA 300 mg, SITA, or PBO (55.6%, 55.2%, 58.5%); serious AE and AE-related discontinuation rates were low across groups. Rates of genital mycotic infections (female: 8.8%, 9.4%, 1.0%, 0%; male: 4.0%, 2.4%, 1.2%, 1.1%) and osmotic diuresis-related AEs (ie, pollakiuria [5.7%, 2.7%, 0.5%, 0.5%], polyuria [0.5%, 0.5%, 0%, 0%]) were higher with CANA 100 and 300 mg than SITA or PBO; these AEs led to few discontinuations. Urinary tract infection rates were higher with CANA 100 and 300 mg and SITA than PBO (5.4%, 3.5%, 3.6%, 2.2%); most events were mild to moderate in severity. Documented hypoglycaemia rates were higher with CANA 100 and 300 mg than SITA or PBO (4.3%, 4.6%, 1.4%, 1.6%). Conclusions: CANA 100 and 300 mg significantly improved glycaemic control and reduced body weight compared with PBO at Week 26 and were generally well tolerated in subjects with T2DM on background MET.

A review of current statistical methodologies for in-storage sampling and surveillance in the grains industry.

Effective, statistically robust sampling and surveillance strategies form an integral component of large agricultural industries such as the grains industry. Intensive in-storage sampling is essential for pest detection, integrated pest management (IPM), to determine grain quality and to satisfy importing nations biosecurity concerns, while surveillance over broad geographic regions ensures that biosecurity risks can be excluded, monitored, eradicated or contained within an area. In the grains industry, a number of qualitative and quantitative methodologies for surveillance and in-storage sampling have been considered. Primarily, research has focussed on developing statistical methodologies for in-storage sampling strategies concentrating on detection of pest insects within a grain bulk; however, the need for effective and statistically defensible surveillance strategies has also been recognised. Interestingly, although surveillance and in-storage sampling have typically been considered independently, many techniques and concepts are common between the two fields of research. This review aims to consider the development of statistically based in-storage sampling and surveillance strategies and to identify methods that may be useful for both surveillance and in-storage sampling. We discuss the utility of new quantitative and qualitative approaches, such as Bayesian statistics, fault trees and more traditional probabilistic methods and show how these methods may be used in both surveillance and in-storage sampling systems.