Indicators of Global Climate Change 2023: annual update of key indicators of the state of the climate system and human influence

Human-induced warming, also known as anthropogenic warming, refers to the component of observed global surface temperature increase attributable to both the direct and indirect effects of human activities, which are typically grouped as follows: well-mixed greenhouse gases (consisting of CO2, CH4, N2O and F-gases) and other human forcings (consisting of aerosol–radiation interaction, aerosol–cloud interaction, black carbon on snow, contrails, ozone, stratospheric H2O and land use) (Eyring et al., 2021). The remaining contributors to total warming are natural: consisting of both natural forcings (such as solar and volcanic activity) and internal variability of the climate system (such as variability related to El Niño/La Nina events).

An assessment of human-induced warming was provided in two reports within the IPCC's sixth assessment cycle: first in SR1.5 in 2018 (Chap. 1 Sect. 1.2.1.3 and Fig. 1.2 (Allen et al., 2018), summarised in the Summary for Policymakers (SPM) Sect. A.1 and Fig. SPM.1 (IPCC, 2018)) and second in AR6 in 2021 (WGI Chap. 3 Sect. 3.3.1.1.2 and Fig. 3.8 (Eyring et al., 2021), summarised in the WGI Summary for Policymakers (SPM) Sect. A.1.3 and Fig. SPM.2 (IPCC, 2021b) and quoted again without any updates in SYR Sect. 2.1.1 and Fig. 2.1 (IPCC, 2023a) and SYR Summary for Policymakers (SPM) Sect. A.1.2 (IPCC 2023b)).

Warming period definitions in the IPCC sixth assessment cycle

Temperature increases are defined relative to a baseline; IPCC assessments typically use the 1850–1900 average temperature as a proxy for the climate in pre-industrial times, referred to as the period before 1750 (see AR6 WGI Cross Chapter Box 1.2).

Tracking progress towards the long-term global goal to limit warming, in line with the Paris Agreement, requires the assessment of both what the current level of global surface temperatures are and whether a level of global warming, such as 1.5 °C, is being reached. Definitions for these were not specified in the Paris Agreement, and several ways of tracking levels of global warming are in use (Betts et al. 2023); here we focus on those adopted within the IPCC's AR6 (Fig. 6). When determining whether warming thresholds have been passed, both AR6 and SR1.5 adopted definitions that depend on future warming; in practice, levels of current warming were therefore reported in AR6 and SR1.5 using additional definitions that circumvented the need to wait for observations of the future climate. AR6 defined crossing time for a level of global warming as the midpoint of the first 20-year period during which the average observed warming for that period, in GSAT, exceeds that level of warming (see AR6 WGI Chap. 2 Box 2.3). It then reported current levels of both observed and human-induced warming as their averages over the most recent decade (see AR6 WGI Chap. 3 Sect. 3.3.1.1.2). This still effectively gives thepage2639 warming level with a crossing time 5 years in the past, so it would need to be combined with a projection of temperature change over the next decade to give a 20-year mean with crossing time at the current year (Betts et al., 2023); we do not focus on this here due to the need for further examination of methods and implications. SR1.5 defined the current level of warming as the average human-induced warming, in global mean surface temperature (GMST), of a 30-year period centred on the current year, extrapolating any multidecadal trend into the future if necessary (see SR1.5 Chap. 1 Sect. 1.2.1). If the multidecadal trend is interpreted as being linear, this definition of current warming is equivalent to the end point of the trend line through the most recent 15 years of human-induced warming and therefore depends only on historical warming. This interpretation produces results that are almost all identical to the present-day single-year value of human-induced warming (see Fig. 6 and results in Sects. 7.3 and S7.3), so in practice the attribution assessment in SR1.5 was based on the single-year-attributed warming calculated using the Global Warming Index, not the trend-based definition.

Updated assessment approach of human-induced warming to date

This paper provides an update of the AR6 WGI and SR1.5 human-induced warming assessments, including, for completeness, all three definitions (AR6 decade-average, SR1.5 trend-based and SR1.5 single-year). The 2023 updates in this paper follow the same methods and process as the 2022 updates provided in Forster et al. (2023). Global mean surface temperature is adopted as the definition of global surface temperature (see Sect. S7.1). The three attribution methods used in AR6 are retained: the Global Warming Index (GWI) (building on Haustein et al., 2017), regularised optimal fingerprinting (ROF) (as in Gillett et al., 2021) and kriging for climate change (KCC) (Ribes et al., 2021). Details of each method, their different uses in SR1.5 and AR6, and any methodological changes are provided in Sect. S7.2; method-specific results are also provided in Sect. S7.3. The overall estimate of attributed global warming for each definition (decade-average, trend-based and single-year) is based on a multi-method assessment of the three attribution methods (GWI, KCC, ROF); the best estimate is given as the 0.01 °C precision mean of the 50th percentiles from each method, and the likely range is given as the smallest 0.1 °C precision range that envelops the 5th to 95th percentile ranges of each method. This assessment approach is identical to last year's update (Forster et al., 2023); it is directly traceable to and fully consistent with the assessment approach in AR6, though it has been extended in ways that are explained in Sect. S7.4.

Results

Results are summarised in Table 6 and Figs. 6 and 7. Method-specific contributions to the assessment results, along with time series, are given in Sect. S7.3. Where results reported in GSAT differ from those reported in GMST (see Sect. S7.1), the additional GSAT results are given in Sect. S7.3.

The repeat calculations for attributable warming in 2010–2019 exhibit good correspondence with the results in AR6 WGI for the same period (see also Sect. S7). The repeat calculation for the level of attributable anthropogenic warming in 2017 is about 0.1 °C larger than the estimate provided in SR1.5 for the same period, resulting from changes in methods and observational data (see AR6 WGI Chap. 2 Box 2.3). The updated results for warming contributions in 2023 are higher than in 2017, also due to 6 additional years of increasing anthropogenic forcing. Note, also, that the SR1.5 assessment only used the GWI method, whereas these annual updates apply the full AR6 multi-method assessment (see Sect. S7.4 for details and rationale). A repeat assessment using the SR1.5 trend-based definition (see Sect. 7.1) leads to results that are very similar to the single-year results reported in Table 6b; best estimates across all components for single-year and trend-based definitions are identical to each other for 2023 and identical to or well within the uncertainty range for 2017 (Sect. S7.3 Table S4).

In this 2024 update, we assess the 2014–2023 decade-average human-induced warming at 1.19 [1.0 to 1.4] °C, which is 0.12 °C above the AR6 assessment for 2010–2019. The single-year-average human-induced warming is assessed to be 1.31 [1.1 to 1.7] °C in 2023 relative to 1850–1900. This best estimate for the current level of human-induced warming reaches the 1.3 °C threshold for the first time. The best estimate is below the observed temperature in 2023 (1.43 [1.32 to 1.53] °C; see Sect. 6), but note the overlap of uncertainties. These best estimates for decade-average and single-year human-induced warming are both 0.05 °C above the value estimated in the previous update for the year 2022 (Forster et al., 2023). The increase of 0.05 °C in the single-year anthropogenic warming level since last year's assessment (Forster et al., 2023) is broken down in the following way: (i) around half of the increase is due to a revision of the historical period due to the additional year of observations (i.e. the 2023 analysis of single-year warming for 2022 is 0.026 °C warmer than the 2022 analysis of single-year warming for 2022), and (ii) around half of the increase is due to the additional year itself (i.e. the 2023 analysis of the single-year warming for 2023 is 0.025 °C warmer than the 2023 analysis of the single-year warming for 2022). Therefore, the variability in human-induced warming from adding 1 particularly hot year of observations (which 2023 was; see Sect. 6) corresponded to advancing only 1 additional year at the current warming rate (0.026 °C per year; see Sect. 7.4), which is significantly smaller than other sources of uncertainty in the assessment. At the attribution level, this slightlypage2640 strengthened anthropogenic warming is comprised of a slightly larger greenhouse-gas-attributable warming, partially offset by a slightly stronger aerosol-attributable cooling.

WGI AR6 found that, averaged for the 2010–2019 period, essentially all observed global surface temperature change was human-induced, with solar and volcanic drivers and internal climate variability making a negligible contribution. This conclusion remains the same for the 2014–2023 period. Generally, whatever methodology is used, on a global scale, the best estimate of the human-induced warming is (within small uncertainties) similar to the observed global surface temperature change (Table 6).

Rate of human-induced global warming

Estimates of the human-induced warming rate refer to the rate of increase in the level of attributed anthropogenic warming over time; this is distinct from the rate of increase in the observed global surface temperature (Sect. 6), which is affected by internal variability such as El Niño and natural forcings such as volcanic activity (Jenkins et al., 2023). The rate of anthropogenic warming is driven by the rate of change of anthropogenic ERF, meaning variations in the rate of climate forcing over time correlate with variations in the rate of attributed warming (see Fig. 8).

\hack{\newpage}A very simple estimate of the rate of human-induced warming and effective radiative forcing was made last year by Forster et al. (2023), which indicated that warming rates were unprecedented, surpassing 0.2 °C per decade (although no uncertainty range was given). That rate calculation was based on annual changes in decade-average anthropogenic warming levels from the GWI method (see Sect. S7.2). This year, attributed anthropogenic warming rates are calculated for all attribution methods using linear trends, as used in AR6, with the overall rate estimate updated in a manner that is fully traceable to and consistent with the rate assessment in AR6.

SR1.5 and AR6 definitions of warming rate

In recent IPCC assessments the definition of warming rate follows two approaches, both of which rely to some extent on expert judgement. In SR1.5 several studies were considered, each defining the rate of warming in various ways and over various timescales; the assessment concluded that the rate of increase in anthropogenic warming in 2017 was 0.2 °C per decade with a likely range of 0.1 to 0.3 °C per decade. In AR6 WGI, the rate of anthropogenic warming utilised three methods (GWI, KCC and ROF; see Sect. S7.2), with the rate defined consistently across all three as the linear trend in the preceding decade of attributedpage2641 anthropogenic warming. While the best-estimate trends reported in AR6 were all higher than the SR1.5 assessment, Eyring et al. (2021) concluded that there was insufficient evidence to change the SR1.5 assessed anthropogenic warming trend in the AR6 WGI report, which therefore remained unchanged from SR1.5 at 0.2 °C per decade (with a likely range of 0.1 to 0.3 °C per decade). Both the SR1.5 and AR6 assessments were given to a precision of 0.1 °C per decade only.

Methods

Following AR6's definition, the rate of warming is defined here as the rolling 10-year linear trend in attributed anthropogenic warming, calculated using ordinary-least-squares linear regression. Note that, as with the level of anthropogenic warming, this decadal approach means the rate of warming in a given year is the trend centred on the preceding decade (i.e. it is 5 years out of date). Each of the three attribution methods used to calculate the level of warming ispage2642 used again here to estimate separate anthropogenic warming rates.

Note that only the GWI methodology relies on the updated historical forcing time series presented in Sect. 4, with the other two methods (ROF and KCC) relying on CMIP6 SSP2-4.5 simulations, which are increasingly out of date (see Sect. S7.2). Very recent changes in anthropogenic forcing, for example, desulfurisation of shipping fuels or the impact of COVID-19, may therefore not be captured fully in the decade-average trend. Further, the anthropogenic forcing record used for attributing warming contains small contributions from biomass burning in the natural environment because of difficulty separating this in estimates of anthropogenic aerosol emissions. It is not expected that either of these effects substantially biases the globally averaged rate of warming estimated here.

Results

Estimates from the GWI (based on observed warming and forcing) and KCC (based on CMIP simulations) both report results in terms of GMST and are in close agreement across each time period. Estimates derived with the ROF method (also based on CMIP simulations) are also reported for GMST here and are more strongly influenced by residual internal variability that remains in the anthropogenic warming signal due to the limitations in size of the CMIP ensemble, as reflected in their broader uncertainty ranges. Given that the ROF results are in this sense outlying, the standard approach of taking the median result for the overall multi-method assessment is adopted.

Results for human-induced warming rate are summarised in Table 7 and Fig. 8. For the purpose of providing annual updates, we take the median estimate at a precision of 0.01 °C per decade, resulting in an overall best estimate for 2014–2023 of 0.26 °C per decade. This increased rate relative to the AR6 assessment of 0.2 °C per decade is broken down in the following way: (i) 0.03 °C per decade of the increase is from a change in rounding precision (updating the AR6 assessment for the 2010–2019 warming rate from 0.2 °C per decade to 0.23 °C per decade), (ii) 0.02 °C per decade of the increase is due to methodological and dataset updates (updating the 2010–2019 warming rate from 0.23 °C per decade to 0.25 °C per decade; this includes the effect of adding 4 additional observed years, which affects the attribution for the entire historical period), and (iii) only 0.01 °C per decade of the increase is due to a substantive increase in rate forpage2643 the 2014–2023 period since the 2010–2019 period (updating 0.25 °C per decade for 2010–2019 to 0.26 °C per decade for 2014–2023). The spread of rates across the three attribution methods remains similar to their spread in AR6 and hence do not support a decrease in the uncertainty range in this update. However, to better reflect the closer agreement of the 5% floors and the larger spread in the 95% ceilings of the three methods, and high rate from the ROF method, we update the uncertainty range for the rate of human-induced warming from [0.1–0.3] °C per decade in AR6 to [0.2–0.4] °C per decade, leaving the precision and range unchanged, noting that this is asymmetric around the central estimate. Therefore, the rate of human-induced warming for the 2014–2023 decade is concluded to be 0.26 °C per decade with a range of [0.2–0.4] °C per decade).

Figure 8 and Table 7 include a breakdown of well-mixed GHGs and other human forcings (including aerosols) and natural forcing contributions since pre-industrial times. The rate time series with ensemble uncertainty are depicted by the GWI method, which is based on observed warming and historical forcing. The rate of total attributable warming (the sum of anthropogenic and natural, not plotted) has good correspondence with the reference plotted observed warming rates. The rates for the attributed warming also correlate closely with the forcing rates. Warming rates have remained high due to strong GHG warming from high emissions and declining aerosol cooling (Forster et al., 2023; Quaas et al., 2022; Jenkins et al., 2022).