Future scenarios assessed by climate science are not predictions of how society will evolve but provide a range of possible futures. It is a societal choice which pathway we try to follow. Climate science and impacts modelling allow us to see the implications of these choices with the aim to guide us to a sustainable 21st century and beyond. Emissions scenarios are often produced using Integrated Assessment Models (IAMs). IAMs are typically optimization models that have mathematical representations of the global economy, energy systems and land-use and agricultural sectors, solving a global system, maximising economic welfare or minimising system costs subject to constraints. Inputs to IAMs are provided by exogenous socioeconomic narratives, such as the Shared Socioeconomic Pathways (SSPs), which specify regional-level population, Gross Domestic Product (GDP), urbanisation, and education pathways to 2,100 (Riahi et al., 2017). Constraints implemented in IAMs may include climate constraints, such as a total allowable cumulative carbon budget acting as a proxy for temperature or year-2100 radiative forcing. An end-of-century radiative forcing constraint is the basis of the well-known Representative Concentration Pathways (RCPs) for the Couple Model Intercomparison Project Phase 5 (CMIP5) (Taylor et al., 2012) and SSP-RCPs for CMIP Phase 6 (CMIP6) (Eyring et al., 2016), examples being RCP4.5 and SSP2-4.5 where the year-2100 radiative forcing is limited to 4.5 W m⁻² (O’Neill et al., 2016). Land use patterns provided by IAMs can also provide boundary conditions for Earth System Models (ESMs) that have a dynamic representation of vegetation cover distribution.

Previously, physical climate models were always run with pre-calculated concentrations of greenhouse gases (GHGs) in future scenarios. However, the development of ESMs, that fully resolve the physical climate system and its coupling with land and ocean carbon cycles, allows for a more realistic forcing with CO₂ emissions (Friedlingstein et al., 2006; Hajima et al., 2024). For CMIP Phase 7 (CMIP7), there is a desire that CO₂ emissions-driven becomes the default configuration (Sanderson et al., 2023). For the IPCC AR6, around a dozen ESMs resolved land and ocean carbon cycles and could be driven by CO₂ emissions (Arora et al., 2020). These models are incredibly useful for determining the Earth system response to increasing CO₂ emissions, but also the Earth System response to zero or negative emissions (Jones et al., 2016; MacDougall et al., 2020). A few models now have methane emissions-driven capability (Folberth et al., 2022), but other greenhouse gases must be provided with pre-calculated concentrations (Smith and Gasser, 2022). Hence, most climate model experiments that investigate future scenarios, as assessed by IPCC, are run with prescribed concentrations for all GHGs, even in those ESMs that have emissions-driven capability (Sanderson et al., 2023). The GHG concentrations for input into ESMs are calculated using simpler climate models (Meinshausen et al., 2011; Smith et al., 2018) from the IAM-simulated GHGs future emissions pathways.

In the early days of scenario-building such as those from the Special Report on Emissions Scenarios (SRES) (Nakicenovic et al., 2000), “business as usual” tended to denote pathways with constant or increasing greenhouse gas emissions in the future leading to several degrees Celsius of warming in 2100. This continued with the development of the RCPs, where RCP8.5, the highest emissions scenario used for ESMs in CMIP5 and AR5, was described as “business as usual” by its developers (Riahi et al., 2011). Although CO₂ emissions globally are not yet declining, a shift away from coal in favour of gas, and rapid scale up of renewable energy globally, has slowed the growth of CO₂ emissions. To track the highest SSP-RCP emissions scenario prepared for CMIP6 and assessed in the IPCC AR6, SSP5-8.5, would require a large-scale reversal back to coal power, which appears to be unlikely. It has been argued that emissions scenarios such as SSP5-8.5 no longer represent “business as usual” (Hausfather and Peters, 2020), as they would now require an unprecedented acceleration in fossil fuel production. In addition, the SSP5-85 frames an amount of CO₂ emitted by human activities by the end of the century that exceeds the amount of fossil fuel reserves as assessed by IPCC AR6 (Canadell et al., 2021). Nevertheless, climate outcomes that are representative of these very high emissions scenarios could still occur, should carbon cycle feedback be strong, climate sensitivity high, or some Earth system tipping thresholds crossed (Sarofim et al., 2023).

Low-end scenarios have typically become more ambitious throughout time. Early scenarios such as the SRES sets did not create pathways that were expected to limit global mean surface temperature to below 2°C. Increasingly, scientific and policy discourse was coalescing on 2°C as a warming level that should be avoided to prevent the most catastrophic impacts of climate change (Jaeger and Jaeger, 2011), and therefore a renewed focus on plausible low-end scenarios resulted in IAM groups investigating these trajectories. In time for the IPCC Fifth Assessment Report (AR5), several such scenarios were developed, which for the first time seriously considered the necessary net negative CO₂ emissions in the second half of the 21st century. This includes RCP2.6, a scenario for Earth System models that limited warming to 2°C with net negative CO₂ emissions around 2070 (van Vuuren et al., 2011). Risk assessment of climate impacts still requires knowledge of plausible low-likelihood high-impact outcomes, and this includes higher than expected climate forcing from human activity (Wood et al., 2023).

The 26^th Conference of Parties (COP26), held in Paris in 2015, advocated for a more stringent warming limit, with “holding the increase in the global average temperature to well below 2°C above pre-industrial levels and pursuing efforts to limit the temperature increase to 1.5°C above pre-industrial levels.” New scenarios were required that implemented even deeper mitigation, requiring CO₂ emissions reductions of 50% by 2030 relative to 2010, net zero emissions by 2050, and strong action on short-lived climate forcers such as methane and black carbon. In some circumstances, the definition of 1.5°C-consistent allows for a small, temporary overshoot of this warming level, acknowledging the geophysical and socioeconomic difficulties in avoiding peak warming exceeding 1.5°C (Dvorak et al., 2022). The latest scenario designed for CMIP6, SSP1-1.9, is one such representative of this set (O’Neill et al., 2016). Such scenarios typically require large amounts (and rapid deployment) of carbon dioxide removal (CDR) technologies such as direct air capture and bioenergy with carbon capture and storage, but their plausibility at the scales required has been questioned (Anderson and Peters, 2016; Fuss et al., 2018; Smith et al., 2016). IAMs and ESMs are increasingly including specific CDR processes in order to assess their implications, side-effects and co-benefits (Keller et al., 2018; Merfort et al., 2023). Earth system modelling can help establish both the demand for CDR and also the feasibility of supply (Jones, 2022).

Scenarios such as the SSP-RCPs describe storylines and mitigation pathways of how society may develop over the course of the 21st century. It is not possible for any scenario to describe the exact emissions and land use pathway that the world will take in the future. There are many unknown unknowns in both the human system and the climate system, and on how society will respond to climate change. Furthermore, the IAMs that are used to construct the scenarios do not include climate change impacts on the rest of the human-Earth system. This is by design in the SSPs but may lead to pathways that while feasible in an IAM may be implausible in the real world. For example, high-end scenarios such as SSP5-8.5 project a warming in excess of 4°C in 2100 relative to pre-industrial (Tebaldi et al., 2021), but continues to grow the economy at 3% per year (Riahi et al., 2017), thus neglecting any negative impacts of climate change on human systems (e.g., health, economy, agriculture, etc.) (Woodard et al., 2019). For low-end scenarios, the feasibility concerns revolve around the plausibility of the rate of emissions reduction that could be achieved in the coming decades and on the rate of deployment and scale of negative emissions technologies (Fuss et al., 2018; Heck et al., 2018; Séférian et al., 2018).

Climate change is primarily driven by anthropogenic CO₂ emissions. An advance for the IPCC AR5 was the realisation that global warming is largely scenario independent and approximately linearly related to the total cumulative emissions of CO₂ (Allen et al., 2009; Gregory et al., 2009; Matthews et al., 2009). Other non-CO₂ emissions affect the warming, and can contribute to both rapid mitigation efforts and Earth system feedbacks, but as a good approximation the degree of warming we may expect in the future is principally proportional to the total amount of CO₂ we emit. This linearity is perhaps surprising and reasons for it are still not fully understood as they include inter-related processes, such as ocean heat and carbon uptake, but also unrelated processes, such as changes in radiative forcing from CO₂ vs. changes in atmospheric CO₂ airborne fraction (Goodwin et al., 2014; MacDougall and Friedlingstein, 2015).

When analysing high-end emissions scenarios, one can approximate the degree of warming expected from the knowledge of the cumulative carbon emissions. The relationship between them is characterised by a metric known as TCRE—the Transient Climate Response to cumulative CO₂ Emissions. IPCC AR6 assessed the most likely value of TCRE as 0.45°C per 1,000 GtCO₂, with a likely range of 0.27–0.63°C. Given current CO₂ emissions of around 40 GtCO₂ per year, the CO₂-induced current warming is expected to be around 0.18°C per decade (neglecting natural climate variability), slightly below the observed rate of warming of about 0.26°C per decade given the warming contribution of non-CO₂ greenhouse gases (Forster et al., 2024). By 2100, the high emissions scenarios SSP3-7.0 and SSP5-8.5 have cumulative CO₂ emissions, from the year 1850, of 8,120 GtCO₂ and 10,590 GtCO₂, leading to a warming of 3.7 and 4.8°C, assuming the TCRE central estimate, similar to the IPCC assessed warming (3.6 and 4.4°C respectively). Table 1 shows the approximate expected warming from CO₂ emissions and TCRE along with the more comprehensively assessed estimate from IPCC. The level of agreement is better than the uncertainty in TCRE. This implies that TCRE can be used as a simple and rapid technique to estimate the level of global warming for any scenario by converting from cumulative emissions to a temperature change.

IPCC AR6 showed that the fraction of emissions absorbed by land and ocean reservoirs decreases for higher levels of emissions. Both land and ocean continue to take up carbon, but the efficiency of the sinks (defined as ratio of emissions taken up) declines. Land and ocean absorb about 70% of anthropogenic CO₂ emissions under the low-end SSP1-1.9 scenario, but less than 40% under the high-end SSP5-8.5 scenario. This decline in carbon sink efficiency is partly due to feedbacks from the climate which inhibits the sink strength, and partly due to the change in rate of anthropogenic emissions (Canadell et al., 2021).

When anthropogenic CO₂ dissolves into the ocean, it causes seawater to become more acidic, which then affects organisms especially those that form calcium carbonate shells. The phenomenon is termed ocean acidification, which is progressing steadily on a global scale. Since the late 1980s, the pH of the open ocean surface water has been decreasing at a pace of 0.0166 ± 0.0010 per decade (Ma et al., 2023). Anthropogenic acidification is also found in the deep layers of the ocean, and significant acidification has been detected even at a depth of 1,000 m (Lauvset et al., 2020). The CMIP6 models project that a global average pH decline of 0.44 ± 0.005 from the late 19th century (1870–1899) to the end of the 21st century (2080–2099) under SSP5-8.5, contrasted to 0.16 ± 0.002 for the SSP1-2.6 (Kwiatkowski et al., 2020). If such a large pH decrease is realized in the future, aragonite unsaturation could occur, especially in the polar regions (Figuerola et al., 2021; Yamamoto et al., 2012), with a potential direct impact on aragonite shell-forming organisms. Acidification is also projected to progress in the middle and deep ocean layers, although there are greater uncertainties than in the surface layer due to the contribution of transport processes in ocean circulation (Watanabe and Kawamiya, 2017).

As with the open ocean, coastal areas are also being acidified due to anthropogenic CO₂ invasion. In addition, the background biological activity in coastal areas is vigorous, so the amplitude of diurnal and seasonal fluctuations is large. The result is that the pH can be temporarily very low and is more likely to fall below a critical value than in the open ocean (Cornwall et al., 2013). Moreover, relatively small-scale phenomena such as coastal upwelling can cause large pH fluctuations, potentially with a serious impact on fisheries (Feely et al., 2008).

In addition to ocean acidification, consequences of anthropogenic CO₂ emissions on the ocean include frequent marine heat waves and oxygen depletion. Although each event affects the marine ecosystem individually, it has been pointed out that strong acidification and oxygen depletion may occur simultaneously in areas where marine heat waves occur (Bindoff et al., 2019; Gruber et al., 2021). Understanding these combined effects, along with a detailed understanding of acidification in mid-deep layers and coastal areas mentioned above, is critically needed.

Although our understanding is that global warming levels increase linearly with emissions, there are likely to be abrupt changes or tipping points in the system, which can happen at local-to-regional scales, not necessarily driving a large acceleration in the rate of climate change, but potentially leading to substantial regional to global impacts (Lenton et al., 2008). In the past transitions occurred at slower rates (see Figure 1) and were driven by biogeochemical feedbacks, paced by evolution and tectonics. The latest assessments suggest that multiple tipping points are at non-negligible levels of risk for global warming as low as 1.5°C (Armstrong McKay et al., 2022). The current rate of radiative forcing increase is about 25 times faster than over the last deglaciation. This implies that slow elements of the Earth system, such as ice sheets, are impacted by high rates that may—or may not—be buffered in the case of an overshoot scenario (Ritchie et al., 2021). At progressively higher warming levels, the likelihood of large-scale singular events to occur increases. Outcomes such as Arctic or Antarctic ice sheets instabilities, an Atlantic Meridional Overturning Circulation (AMOC) collapse, or a dieback of the Amazon Forest are plausible. Still, there is little quantitative understanding of the climate thresholds which may trigger these events. Tipping points may be detected by various techniques, including critical slowing down, with some studies reporting these signals as being detected (Boulton et al., 2022) while other studies reporting no evidence yet (Tao et al., 2023). In the case of the Amazon forest, regional drying and warming, in conjunction with human-land-use/deforestation may push Amazon forest beyond the limits of resilience and lead to a widespread loss of tropical forest (Flores et al., 2024; Lapola et al., 2023). However, how such a threshold may change under elevated CO₂, and associated improved water-use efficiency, remains a critical research gap (Good et al., 2011). As mentioned before, observations are already showing an increase in tree mortality and a decline in the forest carbon sink, which suggests the risk is already present. In the future, projections assessed by IPCC suggest drying in the Amazon forest region—especially during June–July–August, which increases with global temperature and higher scenarios. In such scenarios, tropical forest burned area also increases markedly with up to 30% loss of carbon at 4°C under the highest emissions scenario (Burton et al., 2022).

Historically, the family of low-end scenarios emerged in climate assessment at the time of IPCC AR5, where the methodological template of the RCPs accounted for mitigation options in the portfolio of climate solutions (van Vuuren et al., 2011). As such, the introduction of this family of scenarios represented a step change in the provision of future climate projections, contrasting with their predecessors, the SRES scenarios (Nakicenovic et al., 2000).

Contrasting with high-end scenarios, low-end scenarios are defined through several overarching properties that have been assessed in IPCC reports since AR5. The first key property of low-end scenarios is the prominent role of key geophysical constraints, such as the remaining carbon budget. This geophysical quantity is estimated from the total amount of CO₂ we can still emit to limit global warming to a certain level while having removed past emissions. The total amount is derived from TCRE. Knowing the current warming level allows to estimate the remaining carbon budget, also accounting for warming from non-CO₂ forcing (Matthews et al., 2020; Rogelj et al., 2016), and the Zero-Emission Commitment (ZEC), i.e., the long-term warming or cooling after CO₂ emissions cease (Jones et al., 2019; MacDougall et al., 2020).

For instance, the IPCC AR6 estimated that the remaining global carbon budget for a 50% chance of staying within 1.5°C was 500 GtCO₂ from January 2020 onwards. More recent estimates, accounting for the CO₂ emissions since 2020 and updated estimates of the future warming from non-CO₂ agents, are as low as 200 GtCO₂ from January 2024 onwards (Forster et al., 2024). Importantly, this still applies to overshot and CO₂ removal, so even if we exceed a target warming level, the concept of a carbon budget helps us plan the scale of action required to return to it.

The fact that low-end scenarios are constrained to a given carbon budget, implies that at a point in time, human-induced CO₂ emission will pass from positive to negative. This point in time, when emissions reach “net zero,” represents the second overarching geophysical properties of low-end scenarios. Indeed, the recent IPCC AR6 WGIII report shows that limiting human-caused global warming to a specific level requires limiting cumulative CO₂ emissions, i.e., reaching net zero or net negative CO₂ emissions, along with strong reductions in other GHG emissions. These low-end scenarios also clarify the difference between net zero CO₂ and net zero GHGs, with net zero GHGs (or climate neutrality) requiring all remaining CO₂ and non-CO₂ GHG emissions to be counterbalanced by durably, ideally permanently, stored CO₂ removals.

In the latest IPCC AR6 WGIII scenario database, about 70% of the low-end scenarios halting warming to 2°C deploy net negative CO₂ emissions and do not reach net zero GHG emissions by the end of the century.

Beyond “net zero” many scenarios also extend to having globally net-negative emissions of CO₂—a case which requires total amounts of CDR to exceed any residual positive emissions from human activity. Net negative emissions are required to enable reductions in global temperature and so-called “overshoot” scenarios. Many scenarios achieving 1.5°C by the end of this century entail periods of overshoot and subsequent net-negative emissions. Different CDR techniques have different consequences, including on land or marine ecosystems, biodiversity, albedo (and regional temperature). Hence we need to consider the concept of “climate neutrality” as well as simply carbon neutrality (Zickfeld et al., 2023).

Low-end scenarios also helped to challenge an outdated concept where halting global warming only requires stabilised greenhouse gases concentration, not zero emissions. In a geophysical sense, stabilising greenhouse gases concentration would stabilise the radiative forcing from greenhouse gases. This would lead to continuous global warming for centuries because of the slow accumulation of heat in the ocean.

In order to stabilise the climate (i.e., no significant further warming), recent research shows that CO₂ emissions have to cease (MacDougall et al., 2020; Solomon et al., 2009). After CO₂ emissions stop, the ocean and land will continue to absorb the excess CO₂ from the atmosphere, leading to a slow decline in atmospheric CO₂ concentration and, hence, in radiative forcing, which balances the delayed heat uptake by the ocean. The ZEC has been assessed with ESMs in the ZECMIP experiments where ESMs were driven by CO₂ emissions, emitting 1000GtC total before CO₂ emissions were set to zero. 50 years after emissions cease, the simulated ZEC range from −0.36°C to 0.29°C, with a multi-model ensemble mean of −0.07°C, i.e., essentially zero. However, long-term changes in some slow components of the Earth System are expected to continue even when emissions cease, and to be irreversible on human time scales, this is the case for permafrost thaw and carbon release, ice sheets mass loss, and sea level rise.

When CO₂ emissions become negative, via CDR, the land and ocean carbon sinks are expected to further reduce and eventually, potentially turning into CO₂ sources, in particular for the land sinks, driven by the decline in photosynthesis due to decreasing CO₂ concentrations. The ocean on the other hand, is expected to keep absorbing CO₂ for longer (Jones et al., 2016). However, reversing emissions will likely mean reversing global levels of warming. This is quantified by the same TCRE relationship, so every removal of 1,000 GtCO₂ will lead to about 0.45°C of cooling. IPCC assesses a level of asymmetry in the CO₂ response to negative emissions, but due to the logarithmic dependence of the radiative effect, this means there is likely very little asymmetry in the temperature response. Positive or negative values of the ZEC lead to positive or negative hysteresis behaviour of the global system under negative emissions of CO₂ (Koven et al., 2022).

Although the dominant part of the negative CO₂ emissions is assumed to come from the land sector, novel negative CO₂ emissions technologies that encompass marine-based CO₂ removal are increasingly considered (e.g., Kwiatkowski et al., 2023). Various ocean-based CDR measures have been proposed to mitigate climate change, however, some of the proposed schemes may either promote (nutrient addition) or counteract (alkalinisation) ocean acidification, so it will also be important to also consider CDR from the perspective of ocean acidification.

Air bubbles enclosed in ice cores provide direct evidence of past CO₂ concentrations (Bereiter et al., 2015; Lüthi et al., 2008). The highest CO₂ concentration recorded in polar ice over the past 800,000 years before the industrial era was 300 ppm (Nehrbass-Ahles et al., 2020). Although these values are well below those reached this century, there are at least two periods, the Last Interglacial (LIG, about 125 kyr ago) and Marine Isotope Stage 11 (MIS11, about 400 kyr ago), when sea level likely exceeded current levels (Dutton et al., 2015). Paleo reconstructions of regional sea-level rise rely on evidence for physical, geochemical, and biological features such as coral reefs, shorelines, saltmarshes, or floodmarks that can be dated to the time period of interest. Global sea-level change is derived by combining present and paleo-elevation estimates for regional sea-level with models of glacial isostatic adjustment (Kemp et al., 2015). The first and best studied of these is the LIG when sea level was higher than today, although the exact sea level maximum remains under discussion (Dumitru et al., 2023; Dutton et al., 2015; Dyer et al., 2021). It appears likely that both the Greenland and West Antarctic Ice Sheets were smaller than they are today. The polar warmth experienced in the LIG was comparable to that of the near future. Recent studies suggest that the Greenland Ice Sheet may have lost the ice equivalent to about 3 m of sea level (Sommers et al., 2021), while the West Antarctic Ice Sheet could have lost 4 m of sea level equivalent (Golledge et al., 2021). Insight on sea level can be gained going further back in time to the Pliocene (5 to 3 million years ago), when global mean temperatures was around 4°C higher than today and CO₂ concentrations above 400 ppm (Burke et al., 2018). Despite very few sites having preserved direct sea-level proxies, data from coastal caves in Mallorca suggest that sea-level highstand was around ~17 m higher than present in the mid-Pliocene (~3 Ma) and ~ 25 m higher in the Early Pliocene Climatic Optimum (~5 Ma). These estimates agree with those obtained from data and models, that is 15 m on the United States East Coast for the mid-Pliocene (Moucha and Ruetenik, 2017), and 17.5 m from the Patagonian shorelines for the early Pliocene (Hollyday et al., 2023).

The safe limit for peatland/wetland carbon is first and foremost reliant on having waterlogged soils, but no paleo evidence suggests an upper temperature limit. This is corroborated by modern-day peatlands being present in tropical and temperate areas, with mean annual temperatures above 25°C, wherever moisture is maintained all (or most) year-round. Examples of past time periods with high temperatures leading to unprecedented levels of biomass production exist, i.e., the Carboniferous/Permian periods when the wet early forests (made of spore bearing arborescent plants growing on wet soils) colonised coastal areas of land and stored away all the biomass that we now know as coal seams; this process of carbon burial led to rapid CO₂ drawdown (Feulner, 2017). Some of that rapid accumulation of peat deposits during this warm period may have also been related to fungi not having yet evolved to decompose lignin (Dighton, 2007; Hibbett et al., 2016). Paleo evidence of limits for the hydrological cycle, especially in warm conditions, do exist, and they suggest seasonality of rainfall is important, as any period where peat deposits are exposed to oxic conditions will lead to rapid decomposition (Garcin et al., 2022). This all means that high temperatures may not destabilise peatlands directly, especially if plants adapted to warmer conditions are able to colonise them, but hydrological cycle changes leading to more arid or more seasonal hydroclimates would.

Despite considerable warmth experienced in the Arctic during the LIG (Otto-Bliesner et al., 2021), methane only reached concentrations comparable to pre-industrial level (Köhler et al., 2017; Loulergue et al., 2008). In addition, isotopic measurements of ¹⁴CH₄ in ice suggests that old carbon reservoirs were not an important component of the methane increase after the last glacial period (Dyonisius et al., 2020). This provides some reassurance that a rapid release of methane from permafrost or marine hydrates has not occurred during warm periods when substantial amounts of permafrost undoubtedly melted.