PRECISE

Why is PRECISE needed?

Conservation decisions are ultimately made at specific locations: a woodland to restore, a hedgerow to create, or a habitat to protect.

However, ecological research often provides average effects across many studies, sites or regions. While these averages can be informative, they rarely answer the question practitioners actually need to know: what is likely to happen here?

A conservation intervention that works well in one location may have very different outcomes elsewhere because of differences in climate, soils, landscape structure, species communities and management history.

PRECISE aims to bridge this gap by developing methods that make site-specific predictions about the likely effects of environmental interventions.

Our approach

PRECISE combines ideas from ecology, causal inference and machine learning.

Rather than asking only whether an intervention works on average, we ask:

• Where is an intervention likely to be most effective?

• Which environmental conditions influence success or failure?

• How much variation should we expect among sites?

• How can we make reliable predictions from large ecological datasets?

To answer these questions, we will develop and test a range of modern causal prediction approaches that have already transformed fields such as medicine, economics and marketing.

A central principle throughout the project is that prediction should be grounded in causal understanding. We are interested not simply in identifying patterns in data, but in understanding cause-and-effect relationships and using them to make more useful predictions for conservation and restoration.

What will the project do?

PRECISE will use virtual landscapes and ecological simulations to evaluate how different modelling approaches perform under realistic ecological conditions.

Our primary test case is tree planting and biodiversity, a globally important conservation issue where outcomes are known to vary substantially from place to place.

The project will:

• Simulate biodiversity responses to tree planting across realistic landscapes

• Generate virtual datasets that mimic different ecological sampling and monitoring programmes, from structured surveys to citizen science data

• Evaluate how important data characteristics, such as sample size, covariate coverage, treatment imbalance and spatial structure, influence predictive performance

• Compare conventional ecological approaches with emerging causal machine-learning methods across a wide range of ecological scenarios

• Identify when different analytical workflows perform well, when they fail, and why

• Develop practical guidance that helps ecologists select appropriate methods for the data and questions they are working with

• Apply and test these workflows using real-world datasets from the UK and Scandinavia

Ultimately, PRECISE aims to establish a framework for making robust site-level ecological predictions and provide evidence-based guidance on how different causal workflows should be applied to different types of ecological data. By doing so, the project will help lay the foundations for Precision Ecology.

Open science

All methods, workflows and code developed through PRECISE will be openly shared wherever possible.

The project will produce:

• Reproducible R and Python workflows

• Open-access tutorials and training resources

• Best-practice guidance documents

• Real-world case studies

• Publications and blog posts - watch this space!

• Workshops and webinars

We hope these resources will help researchers and practitioners apply precision approaches across a wide range of environmental challenges.

Funding

PRECISE (Precision Ecology for a Changing World) is funded by the Natural Environment Research Council (NERC).

• Duration: 30 months (2.5 years)

• Lead institution: University of Southampton

• Total project value: £926k

Get involved! :)

PRECISE is intended to be a collaborative and outward-facing project. We are keen to connect with researchers, conservation organisations, monitoring networks and practitioners interested in:

• Causal inference in ecology

• Ecological prediction

• Conservation planning

• Biodiversity monitoring

• Evidence synthesis

• Machine learning for environmental science

• Precision Ecology

If you are interested in collaborating, discussing ideas, exploring applications, or learning more about the project, please get in touch!

Contact:
Dr Rebecca Spake
University of Southampton
r.spake[at[soton.ac.uk

Project team

University of Southampton: Rebecca Spake, Songyan Zhu 

UK Centre for Ecology & Hydrology (UKCEH): James Bullock, Emma Gardner, Rob Cooke

Norwegian Institute for Nature Research (NINA): Matt Grainger

Collaborators and project partners

PRECISE benefits from the expertise of several international collaborators:

Elizabeth Tipton, Northwestern University, USA; Tord Snäll, Forestry Research Institute of Sweden (Skogforsk); Shinichi Nakagawa, University of Alberta, Canada