Uncategorized

Royal Statistical Society Barnett Award for Byron Morgan

Emeritus Professor of Statistics, Byron Morgan, has been awarded the Barnett Award by the Royal Statistical Society.

Being the leading authority on age-structured modelling of capture-recapture and ring-recovery data, his joint paper was the first to model how survival probabilities were influenced by weather covariates. Another influential paper on integrating mark-recapture-recovery and census data was foundational to the internationally-embraced sub-field of Integrated Population Modelling within statistical ecology. Most recently, he has been at the forefront of developing computationally-efficient methods for co-analysis of the UK Butterfly Monitoring Scheme with citizen science data sources, to give insights to biodiversity in urban versus rural settings. Byron Morgan was also one of the co-founders and first director of the National Centre for Statistical Ecology, a virtual Centre that links up statistical ecologists in the UK, and internationally.

Professor Deborah Ashby, President of the Royal Statistical Society, said: “Professor Morgan has had a great influence on the world of statistics and statistical ecology. His innovative work on computationally efficient methods for co-analysis of the UK Butterfly Monitoring Scheme has led to great insights into biodiversity and he had been a significant figure in creating better networks of statistical ecologists.”

Byron will be presented with the award at the Royal Statistical Society Annual Conference in September 2020 where he will also give a keynote presentation.

Standard
Uncategorized

New publication: Statistical Development of Animal Density Estimation Using Random Encounter Modelling

New publication in JABES: Statistical Development of Animal Density Estimation Using Random Encounter Modelling by Natoya Jourdain, Diana Cole, Martin Ridout and Marcus Rowcliffe.

Abstract:

Camera trapping is widely used in ecological studies to estimate animal density, although these studies are largely restricted to animals that can be identified to the individual level. The random encounter model, developed by Rowcliffe et al. (J Anal Ecol 45(4):1228–1236, 2008), estimates animal density from camera-trap data without the need to identify animals. Although the REM can provide reliable density estimates, it lacks the potential to account for the multiple sources of variance in the modelling process. The density estimator in REM is a ratio, and since the variance of a ratio estimator is intractable, we examine and compare the finite sample performance of many approaches for obtaining confidence intervals via simulation studies. We also propose an integrated random encounter model as a parametric alternative, which is flexible and can incorporate covariates and random effects. A data example from Whipsnade Wild Animal Park, Bedfordshire, south England, is used to demonstrate the application of these methods.

 

 

Standard
Conferences/Meetings/Workshops

“eDNA: Challenges and Opportunities”; RSS meeting on the 7th of May 2020

Environmental DNA (eDNA) is an increasingly popular survey tool for monitoring species distribution. eDNA surveys have been used with a wide variety of species in different landscapes and there is growing evidence that they suffer from lower observation error than existing methods relying on direct observation of the target species.

From detecting single species using quantitative polymerase chain reaction  (qPCR), to studying whole communities using metabarcoding, eDNA is showing great promise in helping us understand species distributions and community compositions.

However, we are yet to fully understand the properties of eDNA, and hence are only beginning to appreciate the opportunities that eDNA surveys bring or the challenges that we need to overcome in the field, in the lab or in implementing eDNA surveys into policy.

This meeting brings together researchers who are leading in the development of new statistical methods for analysing eDNA data, in evaluating the use of eDNA surveys with different species and landscapes, or in embedding eDNA techniques into national or international policy.

Speakers and talks

  • 10.15-11  Kerry Walsh, Environment Agency: “Challenges and opportunities: A regulator’s perspective.”
  • 11-11.30 break and refreshments
  • 11.30-12.15  Naomi Ewald, FreshWater Habitats Trust: “Analysis of eDNA data to inform conservation priorities: case studies of long term species monitoring and short term before-after surveys.”
  • 12.15-13 Francesco Ficetola, University of Milan:  “Environmental DNA to track long-term changes of mountain ecosystem.”
  • 13-14 lunch
  • 14-14.45 Jim Griffin, University College London: “Modelling environmental DNA data; Bayesian variable selection accounting for false positive and false negative errors.”
  • 14.45-15.30
  • Doug Yu, University of East Anglia: “Managing wildlife with eDNA data: salmon, leeches, insects, and forests.”
  • 15.30-16.00 Discussion

The meeting, organised by the Environmental Statistics Section and the Emerging Applications Section of the Royal Statistical Society (RSS) will take place on the 7th of May 2020 at the RSS headquarters (12 Errol St, London EC1Y 8LX).

Follow this link to register for the event.

If you have any questions email e.matechou@kent.ac.uk.

Standard
grants

NERC-funded project on “Integrating new statistical frameworks into eDNA survey and analysis at the landscape scale”

The NERC-funded project, led by Dr Eleni Matechou, with Dr Alex Bush, University of Lancaster, Professor Jim Griffin, Statistical Science, UCL, Professor Richard Griffiths, Durrell Institute of Conservation and Ecology, University of Kent, and Professor Doug Yu, UEA as co-Investigators, is part of NERC’s Strategic Priorities Fund on Landscape Decisions: Towards a new framework for using land assets programme.

The RA on the project will be in place until 31st January 2022 and will work on meeting the four model development and two implementation and knowledge exchange objectives of the project.

Model Development (MD) objectives:
MD1 Account for sampling and analysis errors and taxonomic uncertainty in DNA-based data
Single-species (barcoding) and multi-species (metabarcoding) DNA data sets have more complicated properties than data
collected using traditional techniques using visual, aural or physical observations. For example, eDNA surveys are prone to
false-positive and false-negative errors at both the sampling and analysis stages as well as to taxonomic mismatches. We
will develop new statistical models that account for the probabilistic nature of eDNA data, reliably quantifying all levels of
uncertainty in eDNA surveys.

MD2 Identify important landscape predictors for site-specific species composition and metacommunity structure
The probability of species presences, species interactions and metacommunity structures is affected by landscape
characteristics. Accounting for these effects is necessary to obtain an understanding of the system and how to maintain it
or improve it. Failing to account for important effects can lead to inaccurate inference and biased results. However, the
strength and direction of these effects are typically unknown and we will develop a novel Bayesian modelling framework
and sophisticated algorithms for inferring landscape effects efficiently at the species and metacommunity levels using
eDNA data.

MD3 Perform simulations to identify the optimal study design for DNA-based surveys under different scenarios
The power to correctly infer the metacommunity structure and the landscape effects that shape it using eDNA data
depends on the number of sites sampled, the number of samples collected from each site, the number of PCR replicates
performed for each sample, as well as on landscape characteristics. We will perform extensive simulations to identify the
optimal study design under different scenarios and levels of error to provide practitioners with informed guidelines on how
to design eDNA surveys.

MD4 Identify the effects of errors in survey data or suboptimal study design on the selection of conservation priorities (e.g.
high-value habitat or restoration priorities)
Several conservation planning software exist that aim to infer the most efficient allocation of resources. Our Bayesian
model defined by MD1-2 as well as the power analysis resulting from MD3 provide a valuable predictive tool for species
presences and metacommunity structures, with corresponding measures of uncertainty, across the landscape. Using the
predictions generated by the fitted model in our case study system of UK ponds, we will obtain and compare different
solutions provided by planning software that are frequently employed in decision-making processes and assess how and if
decision-making is informed when uncertainty in inference is explicitly accounted for.

Implementation and Knowledge Exchange (IKE) objectives:

IKE1 Develop R-Shiny apps to implement MD1-4, making them accessible to users
The new statistical framework for eDNA data (MD1-4) will be implemented into R-Shiny apps and be accompanied by
examples of data analyses and corresponding interpretation to enable practitioners with limited understanding of statistics
and no prior knowledge of programming to employ our methods when analysing eDNA data.

IKE2 Disseminate project outputs and software to users via training workshops and project partners
We will organise two training workshops to take place in the last quarter of the project and after MD1-4 and IKE1 are
complete to disseminate the new methods to research users. These will complement ongoing – and fully subscribed –
training workshops that the research team have been running for practitioners in the general area of eDNA and statistical
modelling.

A more detailed description of the project can be found here

Standard
Uncategorized

New paper: Assessing heterogeneity in transition propensity in multistate capture–recapture data

Rachel has published a new paper in the Journal of the Royal Statistical Society with former SE@K PhD student Anita Jeyam and co-supervisor Roger Pradel from CEFE, CNRS Montpellier.

The paper is available open access so you are able to download the pdf of the full paper.  The work proposes a new test which allows you to test for heterogeneity in transition probabilities from capture-recapture data before you fit any models to the data.

Assessing heterogeneity in transition propensity in multistate capture–recapture data

Multistate capture–recapture models are a useful tool to help to understand the dynamics of movement within discrete capture–recapture data. The standard multistate capture–recapture model, however, relies on assumptions of homogeneity within the population with respect to survival, capture and transition probabilities. There are many ways in which this model can be generalized so some guidance on what is really needed is highly desirable. Within the paper we derive a new test that can detect heterogeneity in transition propensity and show its good power by using simulation and application to a Canada goose data set. We also demonstrate that existing tests which have traditionally been used to diagnose memory are in fact sensitive to other forms of transition heterogeneity and we propose modified tests which can distinguish between memory and other forms of transition heterogeneity.

 

 

Standard
Uncategorized

New paper: Estimating abundance from multiple sampling capture-recapture data via a multi-state multi-period stopover model

Rachel McCrea, Richard Griffiths (DICE) and collaborators at the Universities of St Andrews and Edinburgh have had a paper published in Annals of Applied Statistics.

 

The article presents an exciting piece of work using Hidden Markov model structure for the fitting of multi-season, multi-state stopover capture-recapture models.  The framework is general, with many existing capture-recapture models being a special case.

The paper can be accessed here.

Abstract

Capture-recapture studies often involve collecting data on numerous capture occasions over a relatively short period of time. For many study species this process is repeated, for example, annually, resulting in capture information spanning multiple sampling periods. To account for the different temporal scales, the robust design class of models have traditionally been applied providing a framework in which to analyse all of the available capture data in a single likelihood expression. However, these models typically require strong constraints, either the assumption of closure within a sampling period (the closed robust design) or conditioning on the number of individuals captured within a sampling period (the open robust design). For real datasets these assumptions may not be appropriate. We develop a general modelling structure that requires neither assumption by explicitly modelling the movement of individuals into the population both within and between the sampling periods, which in turn permits the estimation of abundance within a single consistent framework. The flexibility of the novel model structure is further demonstrated by including the computationally challenging case of multi-state data where there is individual time-varying discrete covariate information. We derive an efficient likelihood expression for the new multi-state multi-period stopover model using the hidden Markov model framework. We demonstrate the significant improvement in parameter estimation using our new modelling approach in terms of both the multi-period and multi-state components through both a simulation study and a real dataset relating to the protected species of great crested newts, Triturus cristatus.

 

Standard