Urgency under uncertainty: Acting on negative and positive tipping points

Abstract

Tipping points in the Earth system have been an important topic in climate research for the last couple of decades. Numerous systems are thought to have non-linear dynamics, such that a slow linear external forcing on them can cause large, irreversible changes that have serious consequences for the climate. Recently, the concept of ‘positive’ societal tipping points has been considered to combat the global climate crisis. Social systems that exhibit these behaviours can promote a more environmentally conscious society with considered interventions. This commentary responds to recent analysis of how negative climate tipping points and positive tipping points are framed in public debate by asking what these framings mean for prediction and decision-making. I argue that highlighting movement toward positive tipping points can spur action because governments often have greater domestic leverage and the faster timescales make progress visible on policy horizons.

Regarding both negative and positive tipping points, Hansson et al. (2025) show public discourse increasingly pairs climatic ‘points of no return’ with promising ‘positive tipping’ in society, producing a race narrative that mixes urgency, hope and confusion over what exactly is tipping. Building on this, we explore how the urgency can differ between these two classes of tipping, based on the uncertainty surrounding them when making predictions about their approach to tipping.

A number of potential tipping points have been identified in the Earth system (Armstrong McKay et al., 2022), the crossing of which would have local, regional and global impacts regarding climate change (Lenton et al., 2023). For example, the collapse of the Atlantic Meridional Overturning Circulation (AMOC) is likely to cause much cooler temperatures over Europe – over 5°C in the UK (van Westen et al., 2024) – as well as altering rainfall patterns which would affect crop production across the region due to increased drying (Ritchie et al., 2020). Globally, it would affect monsoon systems (Ben-Yami et al., 2024) as well as reducing the area suitable for growing certain staple crops by over 50% (OECD, 2021).

The uncertainty around the collapse of the AMOC can potentially cause barriers for it to be integrated into policymaking. In its latest assessment report, the Intergovernmental Panel on Climate Change (IPCC) said with medium confidence that it will not occur by 2100 (Fox-Kemper et al., 2021), while previous assessment reports have said it is very unlikely by 2100 and ‘likely as not’ by 2300 (Collins et al., 2013; Collins et al., 2019). Despite these low probabilities of its collapse occurring, the impacts described above should be hard to ignore, regardless of when they happen. This mirrors what Hansson et al. (2025) find in the media, that tipping points language is used both to warn of the climate crisis and to inspire belief in rapid social change, yet the uncertainty around timing can fuel either fatalism or complacency. This uncertainty is further compounded by the fact that policymakers are likely to implement strategies based on shorter timescales given their time in office (Dryzek and Pickering, 2018), struggling to fit these low probability events into their plans, separate from dealing with the global climate crisis.

A useful tool in assessing the health of a system that is thought to exhibit a tipping point is that of resilience monitoring, providing early warning signals (EWS) of an impending tip (Dakos et al., 2024). Based on the idea of critical slowing down (Wissel, 1984), the system will respond more sluggishly to perturbations the closer it is to tipping. For example, the Amazon rainforest, the dieback of which is a potential tipping point in the climate system, would take longer to recover from the effects of a drought closer to tipping than further away (Boulton et al., 2022). We refer to the ability for a system to recover after a perturbation as its resilience, with a loss of resilience suggesting a movement towards tipping. Most EWS are based on analysing time series of the system in question, looking for changes in statistical indicators on a moving window (such as lag-1 autocorrelation, also known as ‘AR(1)’) (Held and Kleinen, 2004). The tendencies of these indicators are then measured (using Kendall's τ correlation coefficient, for example) to determine if it is generally increasing or decreasing over time (Dakos et al., 2008). Using real world observations, four such climate systems already exhibit EWS and thus are losing resilience, including the aforementioned AMOC and the Amazon rainforest (Boers et al., 2025). Generally, these methods can only tell us that a system is moving towards the tipping point, with little information on when the tipping will occur.

Conveying this message is difficult; with climate change causing stress to these systems, we would expect them to exhibit signals of approaching tipping, but knowing how much time is left before tipping is critical information. There are many factors to consider when tipping might occur, most tie the likelihood of tipping in each system to global mean temperature, providing ‘burning embers’ plots of the probability of each system tipping at various temperature thresholds (Armstrong McKay et al., 2022). This gives the impression that tipping points are just part of the climate change narrative, and that they can be avoided by preventing temperature increases, although even being able to highlight that non-linearities such as these exist in the climate system could still raise awareness of them for policymakers (Laybourn, 2025). In line with other climate research, the idea of the overshoot has now been applied to tipping points (Ritchie et al., 2021; Ritchie et al., 2025), assessing how long a system could survive above a temperature threshold before returning without tipping occurring. While unfortunately this type of research is starting to become a necessity, it is playing a dangerous game skirting closer to the impacts associated with tipping in these systems.

Most climate systems are much more complex than having their health tied to temperature alone. The Amazon rainforest again provides us with an example of a system that is driven by direct actions such as deforestation, as well as changes in climate. The Amazon regulates its own water cycle through evapotranspiration which requires the presence of trees – around 28% of the precipitation in the forest is thought to have come from this process (Zemp et al., 2014), with deforestation causing water recycling feedbacks that would cause further forest loss (Zemp et al., 2017a; Zemp et al., 2017b). As such, curbing deforestation allows the forest to regulate water more efficiently and moves the tipping point further away, a relationship that a temperature threshold alone cannot capture (Wunderling et al., 2025). This relationship can be seen when we look for EWS spatially across the basin. Using Vegetation Optical Depth, a remotely sensed data product that approximates the water content of vegetation, we find across the basin indications of a loss of resilience (Figure 1). However, the spatial heterogeneity of these results show that areas closer to human land use – urban areas, farmlands and roads – and drier areas are losing resilience faster (Boulton et al., 2022), suggesting that temperature alone is not driving these changes in resilience. For most systems, it is not possible to determine specific areas that are losing resilience more than others.

Figure 1.

Early warning signals of Amazon rainforest dieback using vegetation optical depth (VOD) data. (a) Map and (b) histogram of signals observed at grid cell resolution, where Kendall τ values measure the tendency of critical slowing down indicator AR(1), with positive (red) values suggesting a loss of resilience. Full methods are described by Boulton et al. (2022), from which this figure is adapted.

Recently, researchers have attempted to answer the question of how far away the tipping point for the AMOC is, providing a probability function of when tipping is likely to occur (Ditlevsen and Ditlevsen, 2023). This received mainstream media attention when it was highlighted the most likely time of tipping was in 2057, and as early as 2037 using the 95% confidence range. While this was soon contested, due to the uncertainties being too large to allow the method to accurately predict the time to tipping (Ben-Yami et al., 2023), the impact of having some evidence that a climate tipping point may be crossed this century without being linked to the usual temperature threshold has done little to change policy thus far.

Recently, a lot of effort has been placed into bringing the concept of tipping points to a large audience. The Global Tipping Points Report (Lenton et al., 2023; Lenton et al., 2025a) aims to compliment the IPCC assessment reports which up to now have not had a strong focus on tipping points as an overarching theme. Furthermore, the focus here is just as much on ‘positive tipping points’, the concept that many social systems can also undergo radical change when forced towards it, much such as physical, negative tipping points. Systems such as the private transport market are thought to be able to tip from an internal combustion engine dominated state to an electric vehicle (EV) dominated state (Sharpe and Lenton, 2021; Geels and Ayoub, 2023). Other such examples include the transition towards meat-free diets (Lenton et al., 2022), and the uptake of solar photovoltaic by the public (Lenton et al., 2025b; Nijsse et al., 2023), although careful consideration must be given to which systems may exhibit a social or positive tipping point (Milkoreit, 2023).

With the belief that social systems can be modelled in a similar way to the physical climate systems, there is also some evidence of EWS being possible for these positive systems, deemed ‘early opportunity signals’ (EOS). Using these, we can predict the movement towards a sort of tipping that would combat global climate change. With Hansson et al. (2025) highlighting the potential dilution of the urgency of tipping points in the media, it is important to keep the distinction between EWS and EOS. These signals have been observed in the transition to an EV dominated marketplace in the UK (Boulton et al., 2025), using the daily number of advert views of EVs compared to non-EVs on a second-hand car selling website. The spike in these views corresponds to political interventions, such as the announcement of a ban on new petrol vehicles, as well as when petrol shortages are reported in the media (Figure 2). As time goes on, these spikes take longer to return to normal, akin to the Amazon rainforest taking longer to recover from droughts the closer it is to tipping, except this time the undesirable, incumbent state of predominately internal combustion engine vehicles is losing resilience.

Figure 2.

Early opportunity signals in daily advert views of electric vehicles (EVs) on a UK second-hand car selling website. (a) Time series of daily view share of EV adverts (percentage of advert views that day that were for EVs; black line) with detected spikes in this being attributed to external events 1–5: 1) UK Government announces ban on sale of new petrol vehicles by 2035, 2) Ban brought forward to 2030, 3) HGV driver shortage causes petrol availability uncertainty, 4) and 5) Spikes in UK fuel prices. Grey regions show the amount of time the view share takes to return to 75% of pre-spike levels. (b) The number of days each spike takes to return to these levels, showing an increase in interest over time. See Boulton et al. (2025) for full method.

That EOS are potentially able to show that a system is approaching a positive tipping point highlights important distinctions compared to EWS showing the movement of a physical system approaching tipping. Firstly, social systems tend to behave on faster timescales, meaning the direct impact of changes to policy for example, can be observed and changes in an indicator can be easily linked to interventions. Conversely, physical systems behave on much slower timescales and individual events can be hard to spot in their EWS signals. Secondly, the interventions of these social systems are likely to be at governmental level rather than requiring global cooperation, providing the chance for countries to get ‘ahead of the curve’, and help towards commitments they may have made towards global efforts to reduce emissions. These timescales also fit within the timescales that governments can make meaningful action within, compared to longer timescale targets (Dryzek and Pickering, 2018).

The difficultly in conveying EWS of tipping points in the climate system stems from the fact that they are intrinsically linked to global climate change. Although they can highlight specific areas where intervention could help in the case of the Amazon rainforest (Boulton et al., 2022), in most cases, EWS just highlight a movement towards tipping that is associated without specific knowledge of when they could happen besides linking them to temperature increases. Monitoring positive tipping points, however, allows governments to quickly see the impacts they can make to a system (Boulton et al., 2025), whilst at the same time making a meaningful effort to meet their own climate or emissions targets. In their conclusion, Hansson et al. (2025) state: ‘When tipping points are used to describe changes in everyday phenomena, such as price trends, attitude changes and new modes of transport, there is a risk that the concept of climate tipping points will be diluted and trivialized’ and I agree with this point. However, this commentary has aimed to highlight important distinctions between the two, particularly how the uncertainty between them differs, and the importance of urgency in acting on positive tipping point predictions.