Research and analysis

Early warning signals for monitoring supplies

Published 19 August 2026

This is not a statement of government policy.

Rapid projects support government departments to understand the scientific evidence underpinning a policy issue or area by convening academic, industry and government experts at a single roundtable. These summary meeting notes seek to provide accessible science advice for policymakers. They represent the combined views of roundtable participants at the time of the discussion and are not statements of government policy.

This publication considers the following question, taken from a meeting note of a roundtable chaired by Anjali Goswami (Department for Environment, Food and Rural Affairs (Defra) Chief Scientific Adviser) facilitated by the Government Office for Science on 2 March 2026.

How should the United Kingdom (UK) use early warning signals to monitor threats to its supplies of food, water and timber?

Summary of roundtable findings

Single indicators of risk to ecosystems often miss cascading disruptions across food, water and timber supplies.

Complex ecosystems, such as boreal forests, are especially difficult to monitor using existing indicators.

Monitoring sudden changes to ecosystems poses a different set of challenges to monitoring long‑term ecosystem degradation. Future approaches should, however, seek to combine tracking of both kinds of risk to provide a more holistic judgement of an ecosystem’s stability.

Indicators add most value when they have clear owners and pre‑agreed thresholds for response, so that subsequent actions can be proportionate and timely.

The UK has rich monitoring of its own ecosystems, but siloed datasets with limited access undermine integrated, national early warning practices. Other countries have shown that operationally relevant monitoring is feasible.

There are promising monitoring methods that use continuously updated modelling: these blend satellite, hydrology and field observations to deliver rolling short and medium‑term forecasts of ecosystem stress, akin to weather forecasting.

Reliability of available indicators for signalling disruption

1. There is no single, dependable approach to monitoring changes across food, water and timber ecosystems combined. 

2. Traditional early-warning systems (such as those tracking changes in ecological populations) cannot anticipate cascading disruptions – past behaviour and trends in ecosystems are not a good predictor for the future.

3. There has been promising use of using natural-capital indicators (for example changes to soil quality, species, vegetation; Lenton and others, 2022) to judge whether ecosystems can still deliver services we rely on. This approach links ecosystem monitoring directly to service delivery risk and highlights where degradation is likely to affect policy goals.

a. In the forest estate, a composite Forest Biodiversity Index (Forest Research, no date given) is already used as a biodiversity condition metric, illustrating how natural‑capital indicators can be assembled from available datasets.

4. Complex ecosystems, such as those prioritised in Defra’s current assessment, are especially difficult to monitor using existing indicators.

a. Available monitoring tools have been designed for detecting changes to more simple forest ecosystems, for example, where satellite and field indicators align well, and anomalies are easier to interpret (for example, NOAA Climate Resilience Toolkit, no date given).

b. By contrast, boreal forests are complex ecosystems of different types of trees interspersed with peatlands/lakes. They span multiple owners, with long periods of darkness and frequent cloud/snow. Both ground data and satellite sensing are inconsistent in such an ecosystem, leading to missed changes or false alarms.

Tipping points versus gradual ecosystem changes

5. Different indicators are required to detect abrupt tipping points versus evidence of long-term degradation (Global Tipping Points, 2025a). These 2 challenges also require different monitoring strategies and policy responses.

6. Abrupt tipping points are sudden shifts where an ecosystem flips to a new state and does not, or cannot, easily recover (Global Tipping Points, 2025b), such as rapid forest die-off (Rotbarth and others, 2024) or outbreak-driven food supply chain disruption.

a. Indicators for this look for loss of resilience (for example, slower recovery after minor drought, De Keersmaecker and others, 2015) which are typically easier to study. There is a higher chance of irreversible outcomes under these circumstances, which lowers the threshold for evidence needed to escalate actions from early warning signals (EWS).

b. The Atlantic Meridional Overturning Circulation (AMOC1) is a good example of a well-understood EWS; warning signs can be risk assessed effectively to inform decision making (Van Westen and others, 2024; Richie and others, 2019; Carnicer and others, 2019).

7. Indicators for gradual degradation are typically slow, cumulative declines in the condition and functioning of ecosystems, such as biodiversity loss, declining amounts of ground water in rivers, or forest health decline.

a. These changes do not require a formal tipping point to have major impacts on supply, but are harder to measure and can have less influence over decision makers.

b. Standard forestry production models handle gradual trends but are poor at capturing sudden disruptions (such as due to drought (Meyer and others, 2025), storms, pest outbreaks (Mohr and others, 2025)), which have the potential to erase decades worth of supply.

c. Standardised precipitation/streamflow indices are also effective at providing situational awareness for drought and water availability (UKCEH, 2025; UKCEH drought inventory, no date given).

d. Some ecosystem monitoring systems in the UK are on a multi-annual repeat cycle, such as the rolling 5-year National Forest Inventory. Such repeat cycles may not be enough to detect change early.

8. Future monitoring approaches should combine tracking of tipping point risks (sudden, often non-linear, and potentially irreversible; Seddon and Milkoreit, 2025) as well as tracking of degradation (smoother, more gradual changes) to provide a more holistic judgment of an ecosystem’s stability.

Data integration and systemic barriers

9. In the UK, data and models largely remain siloed across climate, biodiversity, hydrology, agriculture and forestry, which prevents consideration of routine, integrated early‑warning indicators at national scale.

a. Many universities, businesses, non-governmental organisations (NGOs) and citizen science groups already monitor environmental data. An effective early warning system should draw on and compile these sources, and not rely solely on government datasets.

10. Government‑collected environmental datasets are often not open access. They can lack consistency and be fragmented across different projects and ecosystems. Data usability is a major barrier, with many datasets difficult to access or interpret without specialist training. This reduces their value for researchers and policymakers and hinders evidence‑based decisions.

a. For example, satellite remote sensing can already detect droughts, heatwaves, floods and peatland water loss, but access to government data (for example, National Forest Inventory) is constrained, which hinders calibration and measurement of uncertainty.

11. International examples show that integrated, rapid monitoring is feasible, but the UK lacks equivalent systems and infrastructure to manage growing data volume, data diversity (the different types of data required to fully support modern ecosystem monitoring) and complexity (including novel ecosystem and microbial data).

a. Improving domestic recycling and processing capability would reduce reliance on imported raw materials and overseas processing.

Improving indicator quality

12. Moving from static analysis to dynamic modelling can help to deliver more rolling hindcasts/forecasts and allows better measurement of uncertainty (Chan and others, 2026), similar to weather forecasting. Iterative ensemble prediction (that is running a model multiple times to get a range of possible scenarios) and data assimilation (regularly pulling in new observational data) offer major advances, enabling improved assessment of biodiversity and ecosystem risks.

13. Deep Artificial Intelligence (AI) learning and advanced analytics can integrate multiple data streams (satellite, climate, hydrology, field data) to generate near real‑time indicators of ecosystem stress. These approaches are in use internationally, but are not yet routine or fully operational:

a. Brazil conducts deforestation monitoring, where satellite data are processed rapidly to detect forest loss and trigger enforcement action (Planet Labs PBC, 2025).

b. Czechia conducts bark‑beetle detection, using high‑frequency remote sensing to guide forest management decisions (Planet Labs PBC, 2019).

How indicators can be used to guide government decisions

15. Indicators should be designed to trigger decisions rather than just describe risk. There are ways, for example, of using different levels of indicator outputs to set thresholds for action (for example watch/alert/action), with each level having clear ownership and pre-agreed responses (for example escalate inspections, activate drought measures).

16. Action should be taken whilst risks are still emerging, not only once the evidence is beyond doubt. While some technical indicators still need development, especially for abrupt ecological change, policy officials already have sufficient warning to act on the drivers of risk, particularly in food and timber supply chains and in preparing for water-related risks (such as drought and scarcity).

a. When the consequences of inaction could be irreversible, scientific evidence needs to be framed in terms that clearly show what is at stake for food supplies, water availability, economic stability and national security so that decision makers can act in time.

Meeting participants

The following participants attended the meeting:

• Anjali Goswami (Chair; Defra CSA)
• Alex Pigot (UCL)
• Chris Clements (University of Bristol)
• Emily Lines (University of Cambridge)
• James Morison (Forest Research)
• John Dearing (University of Southampton)
• Katie Facer-Childs (UKCEH)
• Michael Rice (Client Earth)
• Pete Langdon (University of Southampton)
• Ruth Waters (Natural England)
• Tim Lenton (Exeter University)
• Yannick Wurm (ARIA)

References

Carnicer J, Domingo-Marimon C, Ninyerola M, Camarero JJ, Bastos A, López-Parages J, Blanquer L, Rodríguez-Fonseca B, Lenton TM, Dakos V, Ribas M, Gutiérrez E, Peñuelas J and Pons X (2019) ‘Regime shifts of Mediterranean forest carbon uptake and reduced resilience driven by multidecadal ocean surface temperatures’ 

Chan W, FacerChilds KA, Tanguy M, Magee E, Bulut B, Stringer N, Knight J and Hannaford J (2026) ‘UK Hydrological Outlook using Historic Weather Analogues’ 

Forest Research (no date given) ‘The Forest Biodiversity Index (FOBI) tool’ 

Global Tipping Points (2025a) ‘Global Tipping Points – understanding risks & their potential impact’ 

Global Tipping Points (2025b) ‘Tipping Element Monitoring & Response Facilities (TEMRF): Concept Note’ 

Lenton TM, Buxton JE, Armstrong McKay DI, Abrams JF, Boulton CA, Lees K, Powell TWR, Boers N, Cunliffe AM and Dakos V (2022) ‘A resilience sensing system for the biosphere’ 

Meyer BF, Darela-Filho JP, Gregor K, Buras A, Gu Q-L, Krause A, Liu D, Papastefanou P, Asuk S, Grams TEE, Zang CS and Rammig A (2025) ‘Simulating the drought response of European tree species with the dynamic vegetation model LPJ-GUESS (v4.1, 97c552c5)’ 

Mohr JS, Bastit F, Grünig M, Knoke T, Rammer W, Senf C, Thom D and Seidl R (2025) ‘Rising cost of disturbances for forestry in Europe under climate change’ 

NOAA Climate Resilience Toolkit (no date given) ‘ForWarn Forest Change Assessment Viewer. U.S. Climate Resilience Toolkit’ 

Planet Labs PBC (2019) ‘Bark Beetles Are Decimating Forests: Satellite Data Can Help’ 

Planet Labs PBC (2025) ‘How Colombia and Brazil Tackle Crime in Vast Protected Areas’ 

Rotbarth R, van Nes EH, Scheffer M and Holmgren M (2024) ‘Boreal forests are heading for an open state’ 

Seddon J and Milkoreit M (2025) ‘Regional Leadership on Global Tipping Points. Project Syndicate, 19 November’ 

Ritchie PDL, Smith GS, Davis KJ, Fezz C, Halleck-Vega S, Harper AB, Boulton CA, Binner AR, Day BH, Gallego-Sala AV, Mecking JV, Sitch S, Lenton TM and Bateman IJ (2019) ‘Shifts in national land use and food production in Great Britain after a climate tipping point’ 

UK Centre for Ecology and Hydrology (no date given) ‘Drought Inventory. UKCEH project page’ 

van Westen RM, Kliphuis M and Dijkstra HA (2024) ‘Physics-based early-warning signal shows that AMOC is on tipping course’ 

UKCEH (2025) ‘Hydrological hindcasts enable scientists to prepare worst-case drought scenarios’