The British and Irish Region of the International Biometric Society will be hosting a meeting “Advanced Topics in spatial sampling”.
This will be a two-part meeting. The morning session will consist of a 3 hour workshop run by Professor Murray Lark (University of Nottingham) for approximately 30 participants and will include aspects of both design-based and model-based approaches to spatial sampling schemes. The workshop will include both lectures and practicals aiming to cover topics such as spatial coverage, nested designs and spatially balanced designs.
The afternoon session will be a series of 3 webinars by Professor Janine Illian (University of Glasgow), Dr Peter Henrys (Centre for Ecology and Hydrology) and Dr Eleni Matechou (University of Kent) and will discuss some advanced topics in spatial sampling such as monitoring networks, citizen science data and preferential sampling.
Abstracts will be provided in due course and posted below.
Venue: Zoom – joining details will be circulated prior to the event
Morning workshop: 09:30-12:30 (GMT)
Afternoon webinar: 13:30-15:45 (GMT)
Note both events are free of charge for BIR members, for non-members there is a £30 registration fee to attend the morning workshop. Further details on BIR membership can be found at https://members.biometricsociety.org/bir/membership/fees-renewals and is free for students and members of the wider IBS community.
Organiser: Kirsty Hassall, firstname.lastname@example.org
Spatial sampling: some principles and practicalities.
The objective of this workshop is to examine approaches to sampling variables in space. We shall consider both probability sampling for the estimation of population parameters and purposive sampling to support tasks such as spatial mapping. We shall look at practical tools to undertake conceptually simple but practically challenging tasks such as stratified random sampling of irregular regions. We shall examine methods to achieve spatial coverage sampling, including cases where the sample must incorporate fixed locations. We shall examine how sampling density can be selected for geostatistical mapping and shall examine methods including balanced sampling to make use of exhaustive covariates such as remote sensor data in sampling design.
The workshop will be practical and interactive using the R platform, and packages including BalancedSampling and spcosa. Participants should have basic R skills. The spcosa package requires rjava, so participants are advised to install it and check it in advance of the workshop to ensure that their R and Java specifications are compatible. Scripts and data sets will be provided in advance of the workshop.
Spatial modelling – a focus on sampling and observation processes
All statistical modelling of complex data structures involves an abstraction to the essential properties of interest into quantifiable units and associated random variables. In addition, it also often goes along with simplifying assumptions as part of the abstraction process, typically for practical reasons. As a result, methodology can tend to be far removed from reality and hence be of little practical relevance. In the context of spatial modelling, the usual abstraction reduces the available information measurements at locations in space, whose spatial behaviour is then analysed or modelled. Classical simplifications often concern assumptions of homogeneity, isotropy and simple observation processes with known detection probabilities, often for computational reasons. Recent computational improvements however, allows us to relax some of these assumptions and include sampling and
observation processes explicitly in the model. This talk illustrates how specific sampling and observation processes can be explicitly included as part of a model, illustrating the principles behind the software inlabru with a number of ecological examples.
Monitoring the status, trends and impacts on vegetation at national scales: current practices and designs for the future
Environmental monitoring schemes are rarely designed with a single scientific question to answer - often there are multiple purposes for which such data are required. For example. understanding the status (or spatial distribution) and trends of particular metrics and the potential effects of specific covariates on these. When designing a scheme, these multiple issues inevitably mean there is no simple, single optimal design. In this talk I will describe the design of a national monitoring scheme across Wales and how we used a mixed design to try and overcome these issues. I will then describe the potential benefits of integrating data from multiple schemes and how these may be combined within a consistent modelling framework. Finally, the potential to use this integrated framework within an adaptive sampling context will be presented.
How to walk the BeeWalk: modelling bumblebee citizen science data
Eleni Matechou, Fabian Ketwaroo, Richard Comont
The BeeWalk is the only citizen science survey specifically targeting bumblebees in the UK. Volunteers walk along transects counting the number of bumblebees they detect and identifying their species and caste, where possible. In this talk, I will present our recently developed Shiny app for modelling BeeWalk data and inferring several key demographic parameters, such as caste-specific phenology and average nest productivity, in the wild. I will also discuss challenges and limitations of both the data and the model and ideas for future research.