PREDICT

Summary

The PREDICT project is advancing understanding of climate tipping points in the biosphere using Earth observation data and new methods to detect early warning signals. The project harnesses satellite-derived data records of a suite of Essential Climate Variables generated by the ESA Climate Change Initiative to monitor resilience changes in key tipping elements including Amazon rainforest dieback, dryland vegetation, and permafrost thaw. Statistical methods, process-based modelling, and cross-ECV data approaches are employed to create a prototype biosphere resilience sensing system. This will provide critical insights into when and where potentially irreversible changes may occur and inform more targeted climate adaptation and mitigation strategies. The Global Systems Institute at the University of Exeter is the project lead in collaboration with the University of Leicester and UK Centre for Ecology & Hydrology.

Background

Tipping points in the Earth's climate system represent critical thresholds where relatively small perturbations can trigger self-propelling, abrupt, and potentially irreversible changes in large parts of the Earth system. These tipping elements, once pushed beyond their critical thresholds, can lead to disproportionate and long-lasting impacts on ecosystems and human societies compared to more gradual climate change.

Evidence from past abrupt climate changes and future climate model projections suggests that several elements of the Earth's climate system, particularly in the biosphere, can exhibit tipping point behaviour. Ongoing climate change has already advanced to a point where we are at risk of triggering damaging tipping points in parts of the biosphere coupled to the atmosphere, such as the Amazon rainforest or boreal permafrost. Understanding and providing early warning of these tipping points is crucial for informing adaptation and mitigation strategies.

Current approaches to climate tipping point early warning are dominated by manual analysis of one or two generic early warning signals (usually rising autocorrelation and variance), which has clear limitations because they rely on restrictive assumptions that don't necessarily hold for real systems. Additionally, these methods often miss information contained in higher-order statistical moments and spatial patterns that could provide more robust early warning signals.

The unprecedented amount of data from remote sensing, coupled with innovative models and computing, presents the opportunity to develop multiple tools for detecting tipping points in the biosphere. Remote sensing with its global coverage at fine temporal and spatial resolution can uniquely identify and monitor these tipping systems, phenomena, and their interactions across scales.

Earth observation data must meet several criteria to be useful for tipping point applications, including: salient variables correlated with key processes, accurate analysis-ready data, sufficient spatial coverage and resolution, adequate temporal resolution, long enough temporal duration, and low data latency. However, current remote sensing capabilities have some limitations in meeting these criteria, such as data discontinuities, uncertain data quality, challenges in merging sensors, and difficulties in interpreting resilience estimates.

PREDICT seeks to address these challenges by developing an integrated approach that combines multiple lines of evidence from Earth observations, models, statistical techniques, and cross-ECV approaches to significantly advance our understanding of climate tipping points and ecosystem resilience, their underlying processes, interactions, and potential impacts.

Aims and objectives

The PREDICT project aims to develop and advance methodologies for detecting early warning signals of climate tipping points in the biosphere using Earth observation data, particularly the Essential Climate Variable (ECV) datasets from ESA's Climate Change Initiative .

The main objectives include:

5. Deliver novel, high-impact publications driving the scientific frontier on biosphere tipping points. The project aims to produce peer-reviewed articles that significantly advance our understanding of tipping elements and early warning signals.

A central outcome will be the development of a prototype biosphere resilience sensing system that integrates EO data, process understanding, and modelling tools to provide a multi-scale assessment of resilience in key biophysical tipping elements. This system will enable monitoring of critical variables, estimation of stability metrics, and dynamic mapping of tipping risks at spatial scales from local landscapes to global biomes.

The project will also focus on establishing a tipping point sensing system that combines data and models to identify and anticipate potential tipping points across scales, providing a unifying research framework that could inform resilience-based ecosystem management and governance at multiple levels.

Project Plan

The PREDICT project is structured around six work packages (WPs):

• WP1: Project Management, Communication & Outreach - Led by Dr. Jesse Abrams (UoE), this WP will establish a rigorous project management framework, maintain communication with ESA and stakeholders, implement data management policies, and develop a comprehensive communication and outreach strategy. Deliverables include a project management plan, monthly and quarterly reports, and an executive summary.
• WP2: Earth Observation Data - Led by Dr. Robert Parker and Dr. Darren Ghent (UoL), this WP will underpin all EO-related activities, ensuring ESA CCI ECVs are fully utilized. Tasks include literature review, data identification, uncertainty assessment, cross-ECV consistency evaluation, and integration into ESMValTool recipes. This WP will contribute to inventory documentation and roadmap development.
• WP3: Project Definition and Synthesis - Led by Prof. Valerio Lucarini (UoL) and Prof. Timothy Lenton (UoE), this WP will perform a comprehensive scientific study demonstrating the value of satellite data in investigating biosphere tipping points. It will review existing methodologies, catalogue evidence of tipping elements, analyse multisensor data, assess uncertainties, and develop a scientific roadmap.
• WP4: Development of Methodology - Led by Dr. Joseph Clarke and Prof. Peter Cox (UoE), this WP will develop novel methods for early warning signal detection. Tasks include literature review, statistical and deep learning method development, process-based indicator creation, and cross-ECV data combination approaches, followed by feasibility studies and benchmarking.
• WP5: Tipping Elements Analysis - Led by Dr. Joshua Buxton and Dr. Chris Boulton (UoE), this WP will apply the developed methodologies to three key case studies: Amazon Rainforest dieback, Sahelian dryland greening, and permafrost thaw. Tasks include investigating biophysical processes, selecting case study regions, implementing methods, combining multi-source data, and evaluating performance.
• WP6: Uncertainty Characterization - Led by Dr. Chris Huntingford (UKCEH), this WP will define strategies for understanding uncertainties and error propagation. Tasks include quantifying uncertainties due to Earth System Model differences, variations between early warning signal methods, and bounds associated with each method, culminating in overall uncertainty bounds on triggering conditions.

This work programme leverages the team's expertise in EO algorithm development, ecosystem modelling, and tipping point science to deliver a step-change in our ability to understand, monitor, and forecast the resilience of critical biophysical tipping elements.