Steven Gregory Candy has a total of 3 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A Trial of a Virtual Fence to Mitigate Roadkill on an Unsealed Rural Road in Tasmania, Australia

    by Steven G. Candy, James A. Bunker, Bruce Englefield

    A commercial roadkill Virtual Fence (VF) mitigation device (iPTE Traffic Solutions) was used in a field trial to test its effectiveness, for which previously published results have been inconsistent, along a 4.9 km segment of road on Bruny Island, Tasmania. A total of 585 days of monitoring roadkill by species was conducted, with six sections that were alternatively switched on or off according to the Crossover and Multiple Before–After–Control–Impact (MBACI) experimental designs that divided monitoring into “off–on” then “on–off” periods. Aggregate counts, for each period by section combination, from daily counts of Tasmanian pademelons (Thylogale billardierii) were modelled, with a total count of 222. The statistical analysis used the MBACI design to estimate the VF effect using a log-odds ratio parameter (LORP) while accounting for local spatio-temporal effects. Both versions of the analysis, either averaged over the three spatial replicates (paired sections) or two temporal replicates (blocks), showed no statistically significant effect of the VF, judged as an LORP estimate not sufficiently below zero. Corresponding percentage reduction estimates of 9% and 16% were derived from the LORP. The corresponding statistical power required to detect a nominal significant reduction of 50% in rate was 0.5 and 0.6, respectively. This study confirms the results from a similar previous field trial in southern Tasmania that this VF is likely to lead to, if anything, only a minor reduction in roadkill.


  • Open Access

    Models of Inter-Moult Period for Antarctic Krill - Lack of Progress and Promulgation of Unreliable Models Calibrated Using Indirect Observations

    by Steven G. Candy

    Aims: Two methods of calibrating regression models of inter-moult period (IMP) as a function of temperature exposure (T) for crustaceans, in particular, Antarctic krill (Euphausia superba) are reviewed in terms of both theoretical and empirical properties in order to make recommendations on the application of the methods and/or the use of the resultant fitted models. Methodology: The method and fitted model that used a meta-analysis of published results from laboratory-reared krill using means of directly observed IMP for a range of controlled, constant temperature regimes has valid theoretical and empirical support. The alternative used moult frequencies obtained as a “byproduct” of 5-d Instantaneous Growth Rate (IGR) experiments carried out at sea for which individual IMPs were not directly observed. Instead mean IMP given T and animal total length (L) was predicted using the moult frequencies disaggregated to binary data of moulted versus not-moulted as dependent variable in the calibration of a logistic regression on T and L. The shape of the daily development rate, R, the inverse of IMP, versus T response curve fitted using direct observations is a classical monotonically increasing curve whereas for combinations of sex/maturity classes the curves fitted using indirect observations are parabola-like with sexually dimorphic concavities of either up or down. Four sources of bias in predictions of mean IMP using the indirect observations and estimation method are described. One source due to an unrepresentative sampling frame can lead to large positive bias in estimated mean IMP based on theory which has been absent until now and that applies a discrete uniform distribution for next moult date corresponding to ideal asynchrony. This bias and that due to IGR experimental measurement error in T cannot be remedied. Conclusion: The indirect method and the corresponding fitted models are unreliable and should not be used.

  • Open Access

    Another Trial of a Solar-Powered, Cooperative Sensor/Actuator, Opto-Acoustical, Virtual Road-Fence to Mitigate Roadkill in Tasmania, Australia

    by Steven Gregory Candy, James A Bunker, Bruce Englefield

    When wildlife and motor vehicles collide, the result for the animals is often death (roadkill). A commercial roadkill Virtual Fence (VF) mitigation device (iPTE Traffic Solutions) was used in a field trial to test its effectiveness along a 4.9-km segment of road on Bruny Island, Tasmania. A total of 585 days of monitoring of roadkill by species was conducted, with six sections alternatively switched on or off, according to a variation of a Crossover and Multiple Before-After-Control-Impact (MBACI) experimental designs that divided monitoring into “off-on” then “on-off” blocks of periods with each “on” period followed by a 16 d “wash-out” period. For the six sections over the four periods that exclude wash-outs, the 24 aggregated values of daily counts of roadkill of Tasmanian pademelons (Thylogale billardierii), that gave a total count of 222, were modelled. Other species were killed in insufficient numbers for analysis to be informative. The statistical analysis exploited the MBACI design to estimate the VF effect using a log-odds ratio parameter (LORP) while accounting for local spatio-temporal effects. Both versions of the analysis, either averaging over the three spatial replicates (paired sections) or two temporal replicates (blocks) gave no statistically significant effect of the VF judged as a LORP estimate not sufficiently below zero. Corresponding percentage reduction estimates of 9% and 16% were derived from the LORP estimates and given parameter uncertainty estimates the power to detect as significant a LORP corresponding to a 50% reduction in rate were 0.5 and 0.6, respectively. This study confirms results from a similar field trial carried out in 2018 also in southern Tasmania and adds to the evidence that this VF is likely to give if anything only a minor reduction in roadkill for a commonly killed species for which researchers have the best chance of detecting any practically useful reduction.

  • Analysis of Roadkill Rates from a Trial of a Virtual Fence for Reducing Wombat Road Mortalities. Are Severely Under-Powered Studies Worth the Effort?

    by Steven G. Candy, Bruce Englefield

    Aims: Evaluation of the effectiveness of methods of roadkill mitigation requires field experiments and thus experimental designs. We evaluated such a published experiment that used a virtual road-fence (VF) system to attempt to reduce roadkill of bare-nosed wombats (Vombatus ursinus) and draw general conclusions on the ability of the implemented design, given overall mean rates of roadkill, to detect a substantial reduction of 50%. Study Design: The study site consisted of contiguous fenced and unfenced road sections that were each monitored for road kill for 995 and 322 days pre- and post-installation, respectively. The study design was an unreplicated BACI (Before-After-Control-Impact) design. Methodology: No statistical analysis of the roadkill data were carried out in the original study. We used a Poisson/log-link Generalized Linear Model fit to their BACI data as a single 2x2 table of counts and define a single interaction parameter as a function of control-adjusted rate reduction due to the VF. This parameter can also be defined as the logarithm of the ratio of odds of a random kill in the fenced section occurring in the post-installation period to the corresponding odds for the unfenced section. A null hypothesis of no effect of the VF corresponds to a log odds-ratio of zero. We use simulation to show that estimates of this parameter have close to a Gaussian distribution and from this derive an estimate of the statistical power of the design to detect a hypothetical effect of a 50% reduction in rate. We also used simulation to estimate the corresponding power to detect this reduction if the BACI had been physically replicated three times. Results: The estimated log odds-ratio and the corresponding percent reduction inferred for the VF were -0.02 (SE=0.53) and 1.8%, respectively, and the statistical power of the design to detect a hypothetical effect of a 50% reduction in rate was only .35 but gave an improved power of .7 under the hypothetical replication. Conclusion: Our results showed that there was very low confidence in this design combined with similar overall roadkill rates being able to detect a 50% reduction. This increased to medium-to-high confidence if the experiment had been replicated in three locations. The very small point estimate of reduction could have been so low compared to a true but modest effect purely by chance, thus our conclusion is that there is very low confidence in any estimate of the VF’s effect obtained using this study’s design in the context of common roadkill rates.


  • Long-term Trend in Mean Density of Antarctic Krill (Euphausia superba) Uncertain

    by Steven G. Candy

    Two recent attempts to model the long-term trend in mean density of Antarctic krill in the southwestern sector of the Atlantic using the KRILLBASE dataset using different statistical methods as well as inclusion versus exclusion of data from “non-scientific” nets have resulted in disparate conclusions. The approach that used a linear mixed model (LMM) fitted to the log of mean density, after standardisation was applied to individual net hauls and with means calculated for 12 spatial strata by years between 1976 and 2016, gave a highly statistically significant linear “regional” decline north of 60oS and, to a lesser degree, south of this latitude. The alternative approach that used a ”hurdle” model fitted to the individual net haul data, excluded regional stratification, and excluded non-scientific nets failed to detect an overall significant decline. The method of modelling log transformed means was reappraised and corrected by applying a meta-analytic LMM approach. Additionally, nonlinear smooths in year by region and a smooth in mean “climatological temperature” were included in the LMM. This model showed on average a mostly consistent decline north of 60oS, however, neither trend was significantly different from a no-trend prediction with the trend north of 60oS highly uncertain. Uncertainty of predictions resulted in only weak power to detect a substantial decline of the order of 70% between 1985 and 2005. These model-based inferences neither strongly support nor reject a general hypothesis that there has been a dramatic decline in density of Antarctic krill in the Southwest Atlantic over this period.