British Irish Region IBS

Advanced Topics in Spatial Sampling

By Kirsty Hassall posted 28 days ago


A series of events on the topic of advanced spatial sampling was held on the 24th March 2021. The day was split into two parts with the morning dedicated to a workshop on Spatial sampling: some principles and practicalities and was run by Professor Murray Lark and the afternoon dedicated to a series of three webinars by Professor Janine Illian, Dr Peter Henrys and Dr Eleni Matechou. Both events were run online using zoom and with great participation. We had a sell-out of 31 participants to the morning workshop and around 58 attendees to the free afternoon webinars.

The morning workshop focussed on three principles, sampling to estimate the mean, sampling to predict (i.e. to map) and sampling to investigate spatial variability. Murray took participants through these different approaches highlighting that very different sampling strategies are needed to address each objective. The course was highly interactive with the ability to follow examples in R and to work through practicals after each section. Participants asked questions throughout and showed great engagement. Attendees came from a broad geographical spread – a clear advantage of the virtual framework.

The first speaker of the afternoon was Janine Illian (University of Glasgow) whose talk was entitled Spatial modelling – a focus on sampling and observation processes. With some fantastic photographic props Janine introduced some of the key aspects that need to be considered in ecological modelling, and by association the sampling needed to obtain data. These include, the spatial scale, landscape features and domains, interactions between species and of course the practicalities of measurement. All of these introduce some constraints on the observation process which was introduced as an operation of the underlying point process to create a thinned point process. Janine went on to describe how the package inlabru allows users to implement such observation processes and use the underlying inla framework for inference.  A number of examples were used to demonstrate this approach including predicting the re-establishment of cranes in the UK and modelling orang-utan conservation in Borneo.

Peter Henrys (Centre of Ecology and Hydrology) gave the second talk entitled Monitoring the status, trends and impacts on vegetation at national scales: current practices and designs for the future. Pete’s talk focussed on the challenges of designing long-term monitoring studies due to the many different objectives from different stakeholders some of which are not even known at the start of the survey. Increasingly, multiple sources of data are available to tackle similar questions, and so the second half of the talk focussed on different methodologies for combining such data and the link with the observation model, a theme continued from the first talk. A critical warning was issued that in some cases, combining data has detrimental impacts on the estimation when correlation is present. The final part of the talk touched on adaptive designs and how these can help the general framework for national monitoring, but of course the issue of multiple or ill-defined objectives remain.

The final speaker of the afternoon was Eleni Matechou (University of Kent) with her talk, How to walk the BeeWalk: modelling bumblebee citizen science data. Eleni’s talk introduced a citizen science project aimed at quantifying the abundance of bumble bee species throughout the UK. Interestingly, volunteers needed to be able to identify bees of different castes in order to construct a phenological model of the bee lifecycle. The resulting data were incredibly sparse limiting the ability to model things spatially due to quite complex observational processes. Despite this, Eleni showed how they have been successful in modelling the temporal dynamics of the bee lifecycle and how this varies year-to-year and in different species. The modelling is openly available on an R shiny app for users to explore. Discussions concentrated on how the spatial aspect could be incorporated generating new ideas for future research.

All speakers gave fantastic talks with good cross-links between the talks. There were questions after all talks which stimulated some good discussions and opportunities for new collaborations. All in all, a successful event. Slides are now available under the BIR resources, or can be found here. Details of the meeting are available under past events.