Editing IPCC:AR6/SR15/Chapter-1 (section)

== 1.2 Understanding 1.5°C: Reference Levels, Probability, Transience, Overshoot, and Stabilization ==

<span id="working-definitions-of-1.5c-and-2c-warming-relative-to-pre-industrial-levels"></span>
=== 1.2.1 Working Definitions of 1.5°C and 2°C Warming Relative to Pre-Industrial Levels ===

<div id="section-1-2-1-block-1"></div>

What is meant by ‘the increase in global average temperature… above pre-industrial levels’ referred to in the Paris Agreement depends on the choice of pre-industrial reference period, whether 1.5°C refers to total warming or the human-induced component of that warming, and which variables and geographical coverage are used to define global average temperature change. The cumulative impact of these definitional ambiguities (e.g., Hawkins et al., 2017; Pfleiderer et al., 2018) <sup>[[#fn:r51|51]]</sup> is comparable to natural multi-decadal temperature variability on continental scales (Deser et al., 2012) <sup>[[#fn:r52|52]]</sup> and primarily affects the historical period, particularly that prior to the early 20th century when data is sparse and of less certain quality. Most practical mitigation and adaptation decisions do not depend on quantifying historical warming to this level of precision, but a consistent working definition is necessary to ensure consistency across chapters and figures. We adopt definitions that are as consistent as possible with key findings of AR5 with respect to historical warming.

This report defines ‘warming’, unless otherwise qualified, as an increase in multi-decade global mean surface temperature (GMST) above pre-industrial levels. Specifically, warming at a given point in time is defined as the global average of combined land surface air and sea surface temperatures for a 30-year period centred on that time, expressed relative to the reference period 1850–1900 (adopted for consistency with Box SPM.1 Figure 1 of IPCC (2014a) <sup>[[#fn:r53|53]]</sup> ‘as an approximation of pre-industrial levels’, excluding the impact of natural climate fluctuations within that 30-year period and assuming any secular trend continues throughout that period, extrapolating into the future if necessary. There are multiple ways of accounting for natural fluctuations and trends (e.g., Foster and Rahmstorf, 2011; Haustein et al., 2017; Medhaug et al., 2017; Folland et al., 2018; Visser et al., 2018) <sup>[[#fn:r54|54]]</sup> , but all give similar results. A major volcanic eruption might temporarily reduce observed global temperatures, but would not reduce warming as defined here (Bethke et al., 2017) <sup>[[#fn:r55|55]]</sup> . Likewise, given that the level of warming is currently increasing at 0.3°C–0.7°C per 30 years ( ''likely'' range quoted in Kirtman et al., 2013 <sup>[[#fn:r56|56]]</sup> and supported by Folland et al., 2018) <sup>[[#fn:r57|57]]</sup> , the level of warming in 2017 was 0.15°C–0.35°C higher than average warming over the 30-year period 1988–2017.

In summary, this report adopts a working definition of ‘1.5°C relative to pre-industrial levels’ that corresponds to global average combined land surface air and sea surface temperatures either 1.5°C warmer than the average of the 51-year period 1850–1900, 0.87°C warmer than the 20-year period 1986–2005, or 0.63°C warmer than the decade 2006–2015. These offsets are based on all available published global datasets, combined and updated, which show that 1986–2005 was 0.63°C warmer than 1850–1900 (with a 5–95% range of 0.57°C–0.69°C based on observational uncertainties alone), and 2006–2015 was 0.87°C warmer than 1850–1900 (with a ''likely'' range of 0.75°C–0.99°C, also accounting for the possible impact of natural fluctuations). Where possible, estimates of impacts and mitigation pathways are evaluated relative to these more recent periods. Note that the 5–95% intervals often quoted in square brackets in AR5 correspond to ''very likely'' ranges, while ''likely'' ranges correspond to 17–83%, or the central two-thirds, of the distribution of uncertainty.

<div id="section-1-2-1-1"></div>

<span id="definition-of-global-average-temperature"></span>
==== 1.2.1.1 Definition of global average temperature ====

<div id="section-1-2-1-1-block-1"></div>

The IPCC has traditionally defined changes in observed GMST as a weighted average of near-surface air temperature (SAT) changes over land and sea surface temperature (SST) changes over the oceans (Morice et al., 2012; Hartmann et al., 2013) <sup>[[#fn:r58|58]]</sup> , while modelling studies have typically used a simple global average SAT. For ambitious mitigation goals, and under conditions of rapid warming or declining sea ice (Berger et al., 2017) <sup>[[#fn:r59|59]]</sup> , the difference can be significant. Cowtan et al. (2015) <sup>[[#fn:r60|60]]</sup> and Richardson et al. (2016) <sup>[[#fn:r61|61]]</sup> show that the use of blended SAT/SST data and incomplete coverage together can give approximately 0.2°C less warming from the 19th century to the present relative to the use of complete global-average SAT (Stocker et al., 2013 <sup>[[#fn:r62|62]]</sup> , Figure TFE8.1 and Figure 1.2). However, Richardson et al. (2018) <sup>[[#fn:r63|63]]</sup>  show that this is primarily an issue for the interpretation of the historical record to date, with less absolute impact on projections of future changes, or estimated emissions budgets, under ambitious mitigation scenarios.

The three GMST reconstructions used in AR5 differ in their treatment of missing data. GISTEMP (Hansen et al., 2010) <sup>[[#fn:r64|64]]</sup> uses interpolation to infer trends in poorly observed regions like the Arctic (although even this product is spatially incomplete in the early record), while NOAAGlobalTemp (Vose et al., 2012) <sup>[[#fn:r65|65]]</sup> and HadCRUT (Morice et al., 2012) <sup>[[#fn:r66|66]]</sup> are progressively closer to a simple average of available observations. Since the AR5, considerable effort has been devoted to more sophisticated statistical modelling to account for the impact of incomplete observation coverage (Rohde et al., 2013; Cowtan and Way, 2014; Jones, 2016) <sup>[[#fn:r67|67]]</sup> . The main impact of statistical infilling is to increase estimated warming to date by about 0.1°C (Richardson et al., 2018 <sup>[[#fn:r68|68]]</sup> and Table 1.1).

We adopt a working definition of warming over the historical period based on an average of the four available global datasets that are supported by peer-reviewed publications: the three datasets used in the AR5, updated (Karl et al., 2015) <sup>[[#fn:r69|69]]</sup> , together with the Cowtan-Way infilled dataset (Cowtan and Way, 2014) <sup>[[#fn:r70|70]]</sup> . A further two datasets, Berkeley Earth (Rohde et al., 2013) <sup>[[#fn:r71|71]]</sup> and that of the Japan Meteorological Agency (JMA), are provided in Table 1.1. This working definition provides an updated estimate of 0.86°C for the warming over the period 1880–2012 based on a linear trend. This quantity was quoted as 0.85°C in the AR5. Hence the inclusion of the Cowtan-Way dataset does not introduce any inconsistency with the AR5, whereas redefining GMST to represent global SAT could increase this figure by up to 20% (Table 1.1, blue lines in Figure 1.2 and Richardson et al., 2016) <sup>[[#fn:r72|72]]</sup> .

<div id="section-1-2-1-1-block-2"></div>

<span id="figure-1.2"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 1.2'''

<span id="evolution-of-global-mean-surface-temperature-gmst-over-the-period-of-instrumental-observations."></span>
<!-- IMG CAPTION -->
'''Evolution of global mean surface temperature (GMST) over the period of instrumental observations.'''
<!-- IMG FILE -->
[[File:c7a573f15451c4f486ebc4cc479db4c0 figure-1.2-1024x626.png]]

Grey shaded line shows monthly mean GMST in the HadCRUT4, NOAAGlobalTemp, GISTEMP and Cowtan-Way datasets, expressed as departures from 1850–1900, with varying grey line thickness indicating inter-dataset range. All observational datasets shown represent GMST as a weighted average of near surface air temperature over land and sea surface temperature over oceans. Human-induced (yellow) and total (human- and naturally-forced, orange) contributions to these GMST changes are shown calculated following Otto et al. (2015) <sup>[[#fn:r73|73]]</sup> and Haustein et al. (2017) <sup>[[#fn:r74|74]]</sup> . Fractional uncertainty in the level of human-induced warming in 2017 is set equal to ±20% based on multiple lines of evidence. Thin blue lines show the modelled global mean surface air temperature (dashed) and blended surface air and sea surface temperature accounting for observational coverage (solid) from the CMIP5 historical ensemble average extended with RCP8.5 forcing (Cowtan et al., 2015; Richardson et al., 2018) <sup>[[#fn:r75|75]]</sup> . The pink shading indicates a range for temperature fluctuations over the Holocene (Marcott et al., 2013) <sup>[[#fn:r76|76]]</sup> . Light green plume shows the AR5 prediction for average GMST over 2016–2035 (Kirtman et al., 2013) <sup>[[#fn:r77|77]]</sup> . See Supplementary Material 1.SM for further details.

<!-- END IMG -->
<div id="section-1-2-1-2"></div>

<span id="choice-of-reference-period"></span>
==== 1.2.1.2 Choice of reference period ====

<div id="section-1-2-1-2-block-1"></div>

Any choice of reference period used to approximate ‘pre-industrial’ conditions is a compromise between data coverage and representativeness of typical pre-industrial solar and volcanic forcing conditions. This report adopts the 51-year reference period, 1850–1900 inclusive, assessed as an approximation of pre-industrial levels in AR5 (Box TS.5, Figure 1 of Field et al., 2014) <sup>[[#fn:r78|78]]</sup> . The years 1880–1900 are subject to strong but uncertain volcanic forcing, but in the HadCRUT4 dataset, average temperatures over 1850–1879, prior to the largest eruptions, are less than 0.01°C from the average for 1850–1900. Temperatures rose by 0.0°C–0.2°C from 1720–1800 to 1850–1900 (Hawkins et al., 2017) <sup>[[#fn:r79|79]]</sup> , but the anthropogenic contribution to this warming is uncertain (Abram et al., 2016; Schurer et al., 2017) <sup>[[#fn:r80|80]]</sup> . The 18th century represents a relatively cool period in the context of temperatures since the mid-Holocene (Marcott et al., 2013; Lüning and Vahrenholt, 2017; Marsicek et al., 2018) <sup>[[#fn:r81|81]]</sup> , which is indicated by the pink shaded region in Figure 1.2.

Projections of responses to emission scenarios, and associated impacts, may use a more recent reference period, offset by historical observations, to avoid conflating uncertainty in past and future changes (e.g., Hawkins et al., 2017; Millar et al., 2017b; Simmons et al., 2017) <sup>[[#fn:r82|82]]</sup> . Two recent reference periods are used in this report: 1986–2005 and 2006–2015. In the latter case, when using a single decade to represent a 30-year average centred on that decade, it is important to consider the potential impact of internal climate variability. The years 2008–2013 were characterised by persistent cool conditions in the Eastern Pacific (Kosaka and Xie, 2013; Medhaug et al., 2017) <sup>[[#fn:r83|83]]</sup> , related to both the El Niño-Southern Oscillation (ENSO) and, potentially, multi-decadal Pacific variability (e.g., England et al., 2014) <sup>[[#fn:r84|84]]</sup> , but these were partially compensated for by El Niño conditions in 2006 and 2015. Likewise, volcanic activity depressed temperatures in 1986–2005, partly offset by the very strong El Niño event in 1998. Figure 1.2 indicates that natural variability (internally generated and externally driven) had little net impact on average temperatures over 2006–2015, in that the average temperature of the decade is similar to the estimated externally driven warming. When solar, volcanic and ENSO-related variability is taken into account following the procedure of Foster and Rahmstorf (2011) <sup>[[#fn:r85|85]]</sup> , there is no indication of average temperatures in either 1986–2005 or 2006–2015 being substantially biased by short-term variability (see Supplementary Material 1.SM.2). The temperature difference between these two reference periods (0.21°C–0.27°C over 15 years across available datasets) is also consistent with the AR5 assessment of the current warming rate of 0.3°C–0.7°C over 30 years (Kirtman et al., 2013) <sup>[[#fn:r86|86]]</sup> .

On the definition of warming used here, warming to the decade 2006–2015 comprises an estimate of the 30-year average centred on this decade, or 1996–2025, assuming the current trend continues and that any volcanic eruptions that might occur over the final seven years are corrected for. Given this element of extrapolation, we use the AR5 near-term projection to provide a conservative uncertainty range. Combining the uncertainty in observed warming to 1986–2005 (±0.06°C) with the ''likely'' range in the current warming trend as assessed by AR5 (±0.2°C/30 years), assuming these are uncorrelated, and using observed warming relative to 1850–1900 to provide the central estimate (no evidence of bias from short-term variability), gives an assessed warming to the decade 2006–2015 of 0.87°C with a ±0.12°C ''likely''  range. This estimate has the advantage of traceability to the AR5, but more formal methods of quantifying externally driven warming (e.g., Bindoff et al., 2013; Jones et al., 2016; Haustein et al., 2017; Ribes et al., 2017) <sup>[[#fn:r87|87]]</sup> , which typically give smaller ranges of uncertainty, may be adopted in the future.

<div id="section-1-2-1-2-block-2"></div>

<span id="table-1.1"></span>
<!-- START TABLE -->
'''Table 1.1'''

<span id="observed-increase-in-global-average-surface-temperature-in-various-datasets.-numbers-in-square-brackets-correspond-to-595-uncertainty-ranges-from-individual-datasets-encompassing-known-sources-of-observational-uncertainty-only."></span>
<!-- TABLE CAPTION -->
'''Observed increase in global average surface temperature in various datasets. Numbers in square brackets correspond to 5–95% uncertainty ranges from individual datasets, encompassing known sources of observational uncertainty only.'''
<!-- TABLE -->
{| class="wikitable"
|-
| '''Diagnostic / dataset'''
| '''1850–1900 to (1)<br />
''' '''2006–2015'''
| '''1850–1900 to (2)<br />
''' '''1986–2005'''
| '''1986–2005 to (3)<br />
''' '''2006–2015'''
| '''1850–1900 to (4)<br />
''' '''1981–2010'''
| '''1850–1900 to (5)<br />
''' '''1998–2017'''
| '''Trend (6)<br />
''' '''1880–2012'''
| '''Trend (6)<br />
''' '''1880–2015'''
|-
| '''HadCRUT4.6'''
| 0.84
[0.79–0.89]

| 0.60
[0.57–0.66]

| 0.22
[0.21–0.23]

| 0.62
[0.58–0.67]

| 0.83
[0.78–0.88]

| 0.83
[0.77–0.90]

| 0.88
[0.83–0.95]

|-
| '''NOAAGlobalTemp (7)'''
| 0.86
| 0.62
| 0.22
| 0.63
| 0.85
| 0.91
|-
| '''GISTEMP (7)'''
| 0.89
| 0.65
| 0.23
| 0.66
| 0.88
| 0.89
| 0.94
|-
| '''Cowtan-Way'''
| 0.91
[0.85–0.99]

| 0.65
[0.60–0.72]

| 0.26
[0.25–0.27]

| 0.65
[0.60–0.72]

| 0.88
[0.82–0.96]

| 0.88
[0.79–0.98]

| 0.93
[0.85–1.03]

|-
| '''Average (8)'''
| '''0.87'''
| 0.63
| 0.23
| 0.64
| 0.86
| 0.92
|-
| '''Berkeley (9)'''
| 0.98
| 0.73
| 0.25
| 0.73
| 0.97
| 1.02
|-
| '''JMA (9)'''
| 0.82
| 0.59
| 0.17
| 0.60
| 0.81
| 0.82
| 0.87
|-
| '''ERA-Interim'''
| N/A
| 0.26
| N/A
|-
| '''JRA-55'''
| N/A
| 0.23
| N/A
|-
| '''CMIP5 global SAT (10)'''
| 0.99
[0.65–1.37]

| 0.62
[0.38–0.94]

| 0.38
[0.24–0.62]

| 0.62
[0.34–0.93]

| 0.89
[0.62–1.29]

| 0.81
[0.58–1.31]

| 0.86
[0.63–1.39]

|-
| '''CMIP5 SAT/SST blend—masked'''
| 0.86
[0.54–1.18]

| 0.50
[0.31–0.79]

| 0.34
[0.19–0.54]

| 0.48
[0.26–0.79]

| 0.75
[0.52–1.11]

| 0.68
[0.45–1.08]

| 0.74
[0.51–1.14]

|}
<!-- END TABLE -->

Notes:

# Most recent reference period used in this report.
# Most recent reference period used in AR5.
# Difference between recent reference periods.
# Current WMO standard reference periods.
# Most recent 20-year period.
# Linear trends estimated by a straight-line fit, expressed in degrees yr <sup>−1</sup> multiplied by 133 or 135 years respectively, with uncertainty ranges incorporating observational uncertainty only.
# To estimate changes in the NOAAGlobalTemp and GISTEMP datasets relative to the 1850–1900 reference period, warming is computed relative to 1850–1900 using the HadCRUT4.6 dataset and scaled by the ratio of the linear trend 1880–2015 in the NOAAGlobalTemp or GISTEMP dataset with the corresponding linear trend computed from HadCRUT4.
# Average of diagnostics derived – see (7) – from four peer-reviewed global datasets, HadCRUT4.6, NOAA, GISTEMP &amp; Cowtan-Way. Note that differences between averages may not coincide with average differences because of rounding.
# No peer-reviewed publication available for these global combined land–sea datasets.
# CMIP5 changes estimated relative to 1861–80 plus 0.02°C for the offset in HadCRUT4.6 from 1850–1900. CMIP5 values are the mean of the RCP8.5 ensemble, with 5–95% ensemble range. They are included to illustrate the difference between a complete global surface air temperature record (SAT) and a blended surface air and sea surface temperature (SST) record accounting for incomplete coverage (masked), following Richardson et al. (2016) <sup>[[#fn:r88|88]]</sup> . Note that 1986–2005 temperatures in CMIP5 appear to have been depressed more than observed temperatures by the eruption of Mount Pinatubo.

<div id="section-1-2-1-3"></div>

<span id="total-versus-human-induced-warming-and-warming-rates"></span>
==== 1.2.1.3 Total versus human-induced warming and warming rates ====

<div id="section-1-2-1-3-block-1"></div>

Total warming refers to the actual temperature change, irrespective of cause, while human-induced warming refers to the component of that warming that is attributable to human activities. Mitigation studies focus on human-induced warming (that is not subject to internal climate variability), while studies of climate change impacts typically refer to total warming (often with the impact of internal variability minimised through the use of multi-decade averages).

In the absence of strong natural forcing due to changes in solar or volcanic activity, the difference between total and human-induced warming is small: assessing empirical studies quantifying solar and volcanic contributions to GMST from 1890 to 2010, AR5 (Figure 10.6 of Bindoff et al., 2013) <sup>[[#fn:r89|89]]</sup> found their net impact on warming over the full period to be less than plus or minus 0.1°C. Figure 1.2 shows that the level of human-induced warming has been indistinguishable from total observed warming since 2000, including over the decade 2006–2015. Bindoff et al. (2013) <sup>[[#fn:r90|90]]</sup> assessed the magnitude of human-induced warming over the period 1951–2010 to be 0.7°C ( ''likely'' between 0.6°C and 0.8°C), which is slightly greater than the 0.65°C observed warming over this period (Figures 10.4 and 10.5) with a ''likely'' range of ±14%. The key surface temperature attribution studies underlying this finding (Gillett et al., 2013; Jones et al., 2013; Ribes and Terray, 2013) <sup>[[#fn:r91|91]]</sup> used temperatures since the 19th century to constrain human-induced warming, and so their results are equally applicable to the attribution of causes of warming over longer periods. Jones et al. (2016) <sup>[[#fn:r92|92]]</sup> show (Figure 10) human-induced warming trends over the period 1905–2005 to be indistinguishable from the corresponding total observed warming trend accounting for natural variability using spatio-temporal detection patterns from 12 out of 15 CMIP5 models and from the multi-model average. Figures from Ribes and Terray (2013) <sup>[[#fn:r93|93]]</sup> , show the anthropogenic contribution to the observed linear warming trend 1880–2012 in the HadCRUT4 dataset (0.83°C in Table 1.1) to be 0.86°C using a multi-model average global diagnostic, with a 5–95% confidence interval of 0.72°C–1.00°C (see figure 1.SM.6). In all cases, since 2000 the estimated combined contribution of solar and volcanic activity to warming relative to 1850–1900 is found to be less than ±0.1°C (Gillett et al., 2013) <sup>[[#fn:r94|94]]</sup> , while anthropogenic warming is indistinguishable from, and if anything slightly greater than, the total observed warming, with 5–95% confidence intervals typically around ±20%.

Haustein et al. (2017) <sup>[[#fn:r95|95]]</sup> give a 5–95% confidence interval for human-induced warming in 2017 of 0.87°C–1.22°C, with a best estimate of 1.02°C, based on the HadCRUT4 dataset accounting for observational and forcing uncertainty and internal variability. Applying their method to the average of the four datasets shown in Figure 1.2 gives an average level of human-induced warming in 2017 of 1.04°C. They also estimate a human-induced warming trend over the past 20 years of 0.17°C (0.13°C–0.33°C) per decade, consistent with estimates of the total observed trend of Foster and Rahmstorf (2011) <sup>[[#fn:r96|96]]</sup> (0.17° ± 0.03°C per decade, uncertainty in linear trend only), Folland et al. (2018) <sup>[[#fn:r97|97]]</sup>

and Kirtman et al. (2013) <sup>[[#fn:r98|98]]</sup> (0.3°C–0.7°C over 30 years, or 0.1°C–0.23°C per decade, ''likely'' range), and a best-estimate warming rate over the past five years of 0.215°C/decade (Leach et al., 2018) <sup>[[#fn:r99|99]]</sup> . Drawing on these multiple lines of evidence, human-induced warming is assessed to have reached 1.0°C in 2017, having increased by 0.13°C from the mid-point of 2006–2015, with a ''likely'' range of 0.8°C to 1.2°C (reduced from 5–95% to account for additional forcing and model uncertainty), increasing at 0.2°C per decade (with a ''likely'' range of 0.1°C to 0.3°C per decade: estimates of human-induced warming given to 0.1°C precision only).

Since warming is here defined in terms of a 30-year average, corrected for short-term natural fluctuations, when warming is considered to be at 1.5°C, global temperatures would fluctuate equally on either side of 1.5°C in the absence of a large cooling volcanic eruption (Bethke et al., 2017) <sup>[[#fn:r100|100]]</sup> . Figure 1.2 indicates there is a substantial chance of GMST in a single month fluctuating over 1.5°C between now and 2020 (or, by 2030, for a longer period: Henley and King, 2017) <sup>[[#fn:r101|101]]</sup> , but this would not constitute temperatures ‘reaching 1.5°C’ on our working definition. Rogelj et al. (2017) <sup>[[#fn:r102|102]]</sup> show limiting the probability of annual GMST exceeding 1.5°C to less than one-year-in-20 would require limiting warming, on the definition used here, to 1.31°C or lower.

<span id="global-versus-regional-and-seasonal-warming"></span>
=== 1.2.2 Global versus Regional and Seasonal Warming ===

<div id="section-1-2-2-block-1"></div>

Warming is not observed or expected to be spatially or seasonally uniform (Collins et al., 2013) <sup>[[#fn:r103|103]]</sup> . A 1.5°C increase in GMST will be associated with warming substantially greater than 1.5°C in many land regions, and less than 1.5°C in most ocean regions. This is illustrated by Figure 1.3, which shows an estimate of the observed change in annual and seasonal average temperatures between the 1850–1900 pre-industrial reference period and the decade 2006–2015 in the Cowtan-Way dataset. These regional changes are associated with an observed GMST increase of 0.91°C in the dataset shown here, or 0.87°C in the four-dataset average (Table 1.1). This observed pattern reflects an on-going transient warming: features such as enhanced warming over land may be less pronounced, but still present, in equilibrium (Collins et al., 2013) <sup>[[#fn:r104|104]]</sup> . This figure illustrates the magnitude of spatial and seasonal differences, with many locations, particularly in Northern Hemisphere mid-latitude winter (December–February), already experiencing regional warming more than double the global average. Individual seasons may be substantially warmer, or cooler, than these expected changes in the long-term average.

<div id="section-1-2-2-block-2"></div>

<span id="figure-1.3"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 1.3'''

<span id="spatial-and-seasonal-pattern-of-present-day-warming."></span>
<!-- IMG CAPTION -->
'''Spatial and seasonal pattern of present-day warming.'''
<!-- IMG FILE -->
[[File:0d0ae08f34a1c5aefaca52ef4759d334 Figure-1.3-1024x854.png]]

Regional warming for the 2006–2015 decade relative to 1850–1900 for the annual mean (top), the average of December, January, and February (bottom left) and for June, July, and August (bottom right). Warming is evaluated by regressing regional changes in the Cowtan and Way (2014) <sup>[[#fn:r105|105]]</sup> dataset onto the total (combined human and natural) externally forced warming (yellow line in Figure 1.2). See Supplementary Material 1.SM for further details and versions using alternative datasets. The definition of regions (green boxes and labels in top panel) is adopted from the AR5 (Christensen et al., 2013) <sup>[[#fn:r106|106]]</sup> .

<!-- END IMG -->
<span id="definition-of-1.5c-pathways-probability-transience-stabilization-and-overshoot"></span>
=== 1.2.3 Definition of 1.5°C Pathways: Probability, Transience, Stabilization and Overshoot ===

<div id="section-1-2-3-block-1"></div>

Pathways considered in this report, consistent with available literature on 1.5°C, primarily focus on the time scale up to 2100, recognising that the evolution of GMST after 2100 is also important. Two broad categories of 1.5°C pathways can be used to characterise mitigation options and impacts: pathways in which warming (defined as 30-year averaged GMST relative to pre-industrial levels, see Section 1.2.1) remains below 1.5°C throughout the 21st century, and pathways in which warming temporarily exceeds (‘overshoots’) 1.5°C and returns to 1.5°C either before or soon after 2100. Pathways in which warming exceeds 1.5°C before 2100, but might return to that level in some future century, are not considered 1.5°C pathways.

Because of uncertainty in the climate response, a ‘prospective’ mitigation pathway (see Cross-Chapter Box 1 in this chapter), in which emissions are prescribed, can only provide a level of probability of warming remaining below a temperature threshold. This probability cannot be quantified precisely since estimates depend on the method used (Rogelj et al., 2016b; Millar et al., 2017b; Goodwin et al., 2018; Tokarska and Gillett, 2018) <sup>[[#fn:r107|107]]</sup> . This report defines a ‘1.5°C pathway’ as a pathway of emissions and associated possible temperature responses in which the majority of approaches using presently available information assign a probability of approximately one-in-two to two-in-three to warming remaining below 1.5°C or, in the case of an overshoot pathway, to warming returning to 1.5°C by around 2100 or earlier. Recognizing the very different potential impacts and risks associated with high-overshoot pathways, this report singles out 1.5°C pathways with no or limited (&lt;0.1°C) overshoot in many instances and pursues efforts to ensure that when the term ‘1.5°C pathway’ is used, the associated overshoot is made explicit where relevant. In Chapter 2, the classification of pathways is based on one modelling approach to avoid ambiguity, but probabilities of exceeding 1.5°C are checked against other approaches to verify that they lie within this approximate range. All these absolute probabilities are imprecise, depend on the information used to constrain them, and hence are expected to evolve in the future. Imprecise probabilities can nevertheless be useful for decision-making, provided the imprecision is acknowledged (Hall et al., 2007; Kriegler et al., 2009; Simpson et al., 2016) <sup>[[#fn:r108|108]]</sup> . Relative and rank probabilities can be assessed much more consistently: approaches may differ on the absolute probability assigned to individual outcomes, but typically agree on which outcomes are more probable.

Importantly, 1.5°C pathways allow a substantial (up to one-in-two) chance of warming still exceeding 1.5°C. An ‘adaptive’ mitigation pathway in which emissions are continuously adjusted to achieve a specific temperature outcome (e.g., Millar et al., 2017b) <sup>[[#fn:r109|109]]</sup> reduces uncertainty in the temperature outcome while increasing uncertainty in the emissions required to achieve it. It has been argued (Otto et al., 2015; Xu and Ramanathan, 2017) <sup>[[#fn:r110|110]]</sup> that achieving very ambitious temperature goals will require such an adaptive approach to mitigation, but very few studies have been performed taking this approach (e.g., Jarvis et al., 2012) <sup>[[#fn:r111|111]]</sup> .

Figure 1.4 illustrates categories of (a) 1.5°C pathways and associated (b) annual and (c) cumulative emissions of CO <sub>2</sub> . It also shows (d) an example of a ‘time-integrated impact’ that continues to increase even after GMST has stabilised, such as sea level rise. This schematic assumes for the purposes of illustration that the fractional contribution of non-CO <sub>2</sub> climate forcers to total anthropogenic forcing (which is currently increasing, Myhre et al., 2017) <sup>[[#fn:r112|112]]</sup> is approximately constant from now on. Consequently, total human-induced warming is proportional to cumulative CO <sub>2</sub> emissions (solid line in c), and GMST stabilises when emissions reach zero. This is only the case in the most ambitious scenarios for non-CO <sub>2</sub> mitigation (Leach et al., 2018) <sup>[[#fn:r113|113]]</sup> . A simple way of accounting for varying non-CO <sub>2</sub> forcing in Figure 1.4 would be to note that every 1 W m <sup>−2</sup> increase in non-CO <sub>2</sub> forcing between now and the decade or two immediately prior to the time of peak warming reduces cumulative CO <sub>2</sub> emissions consistent with the same peak warming by approximately 1100 GtCO <sub>2</sub> , with a range of 900-1500 GtCO <sub>2</sub>  (using values from AR5: Myhre et al., 2013; Allen et al., 2018; Jenkins et al., 2018 <sup>[[#fn:r114|114]]</sup> ; Cross-Chapter Box 2 in this chapter).

<div id="section-1-2-3-1"></div>

<span id="pathways-remaining-below-1.5c"></span>
==== 1.2.3.1 Pathways remaining below 1.5°C ====

<div id="section-1-2-3-1-block-1"></div>

In this category of 1.5°C pathways, human-induced warming either rises monotonically to stabilise at 1.5°C (Figure 1.4, brown lines) or peaks at or below 1.5°C and then declines (yellow lines). Figure 1.4b demonstrates that pathways remaining below 1.5°C require net annual CO <sub>2</sub> emissions to peak and decline to near zero or below, depending on the long-term adjustment of the carbon cycle and non-CO <sub>2</sub> emissions (Bowerman et al., 2013; Wigley, 2018) <sup>[[#fn:r115|115]]</sup> . Reducing emissions to zero corresponds to stabilizing cumulative CO <sub>2</sub> emissions (Figure 1.4c, solid lines) and falling concentrations of CO <sub>2</sub> in the atmosphere (panel c dashed lines) (Matthews and Caldeira, 2008; Solomon et al., 2009) <sup>[[#fn:r116|116]]</sup> , which is required to stabilize GMST if non-CO <sub>2</sub> climate forcings are constant and positive. Stabilizing atmospheric greenhouse gas concentrations would result in continued warming (see Section 1.2.4).

If emission reductions do not begin until temperatures are close to the proposed limit, pathways remaining below 1.5°C necessarily involve much faster rates of net CO <sub>2</sub> emission reductions (Figure 1.4, green lines), combined with rapid reductions in non-CO <sub>2</sub> forcing and these pathways also reach 1.5°C earlier. Note that the emissions associated with these schematic temperature pathways may not correspond to feasible emission scenarios, but they do illustrate the fact that the timing of net zero emissions does not in itself determine peak warming: what matters is total cumulative emissions up to that time. Hence every year’s delay before initiating emission reductions decreases by approximately two years the remaining time available to reach zero emissions on a pathway still remaining below 1.5°C (Allen and Stocker, 2013; Leach et al., 2018) <sup>[[#fn:r117|117]]</sup> .

<div id="section-1-2-3-2"></div>

<span id="pathways-temporarily-exceeding-1.5c"></span>
==== 1.2.3.2 Pathways temporarily exceeding 1.5°C ====

<div id="section-1-2-3-2-block-1"></div>

With the pathways in this category, also referred to as overshoot pathways, GMST rises above 1.5°C relative to pre-industrial before peaking and returning to 1.5°C around or before 2100 (Figure 1.4, blue lines), subsequently either stabilising or continuing to fall. This allows initially slower or delayed emission reductions, but lowering GMST requires net negative global CO <sub>2</sub> emissions (net anthropogenic removal of CO <sub>2</sub> ; Figure 1.4b). Cooling, or reduced warming, through sustained reductions of net non-CO <sub>2</sub> climate forcing (Cross-Chapter Box 2 in this chapter) is also required, but their role is limited because emissions of most non-CO <sub>2</sub> forcers cannot be reduced to below zero. Hence the feasibility and availability of large-scale CO <sub>2</sub> removal limits the possible rate and magnitude of temperature decline. In this report, overshoot pathways are referred to as 1.5°C pathways, but qualified by the amount of the temperature overshoot, which can have a substantial impact on irreversible climate change impacts (Mathesius et al., 2015; Tokarska and Zickfeld, 2015) <sup>[[#fn:r118|118]]</sup> .

<div id="section-1-2-3-3"></div>

<span id="impacts-at-1.5c-warming-associated-with-different-pathways-transience-versus-stabilisation"></span>
==== 1.2.3.3 Impacts at 1.5°C warming associated with different pathways: transience versus stabilisation ====

<div id="section-1-2-3-3-block-1"></div>

Figure 1.4 also illustrates time scales associated with different impacts. While many impacts scale with the change in GMST itself, some (such as those associated with ocean acidification) scale with the change in atmospheric CO <sub>2</sub> concentration, indicated by the fraction of cumulative CO <sub>2</sub> emissions remaining in the atmosphere (dotted lines in Figure 1.4c). Others may depend on the rate of change of GMST, while ‘time-integrated impacts’, such as sea level rise, shown in Figure 1.4d continue to increase even after GMST has stabilised.

Hence impacts that occur when GMST reaches 1.5°C could be very different depending on the pathway to 1.5°C. CO <sub>2</sub> concentrations will be higher as GMST rises past 1.5°C (transient warming) than when GMST has stabilized at 1.5°C, while sea level and, potentially, global mean precipitation (Pendergrass et al., 2015) <sup>[[#fn:r119|119]]</sup> would both be lower (see Figure 1.4). These differences could lead to very different impacts on agriculture, on some forms of extreme weather (e.g., Baker et al., 2018) <sup>[[#fn:r120|120]]</sup> , and on marine and terrestrial ecosystems (e.g., Mitchell et al., 2017 <sup>[[#fn:r121|121]]</sup> and Boxes 3.1 and 3.2). Sea level would be higher still if GMST returns to 1.5°C after an overshoot (Figure 1.4 d), with potentially significantly different impacts in vulnerable regions. Temperature overshoot could also cause irreversible impacts (see Chapter 3).

<div id="section-1-2-3-3-block-2"></div>

<span id="figure-1.4"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 1.4'''

<span id="different-1.5c-pathways-schematic-1-illustration-of-the-relationship-between-a-global-mean-surface-temperature-gmst-change-b-annual-rates-of-co-2-emissions-assuming-constant-fractional-contribution-of-non-co-2-forcing-to-total-human-induced-warming-c-total-cumulative-co-2-emissions-solid-lines-and-the-fraction-thereof-remaining-in-the-atmosphere-dashed-lines-these-also-indicates-changes-in-atmospheric-co-2-concentrations-and-d-a-time-integrated-impact-such-as-sea-level-rise-that-continues-to-increase-even-after-gmst-has-stabilized."></span>
<!-- IMG CAPTION -->
'''Different 1.5°C pathways Schematic <sup>[[#fn:1|1]]</sup> illustration of the relationship between (a) global mean surface temperature (GMST) change; (b) annual rates of CO <sub>2</sub> emissions, assuming constant fractional contribution of non-CO <sub>2</sub> forcing to total human-induced warming; (c) total cumulative CO <sub>2</sub> emissions (solid lines) and the fraction thereof remaining in the atmosphere (dashed lines; these also indicates changes in atmospheric CO <sub>2</sub> concentrations); and (d) a time-integrated impact, such as sea level rise, that continues to increase even after GMST has stabilized.'''
<!-- IMG FILE -->
[[File:821be06d1277f0d233698c109dc6082d figure-1.4-1024x717.png]]

Different 1.5°C pathways Schematic <sup>[[#fn:1|1]]</sup> illustration of the relationship between (a) global mean surface temperature (GMST) change; (b) annual rates of CO <sub>2</sub> emissions, assuming constant fractional contribution of non-CO <sub>2</sub> forcing to total human-induced warming; (c) total cumulative CO <sub>2</sub> emissions (solid lines) and the fraction thereof remaining in the atmosphere (dashed lines; these also indicates changes in atmospheric CO <sub>2</sub> concentrations); and (d) a time-integrated impact, such as sea level rise, that continues to increase even after GMST has stabilized. Colours indicate different 1.5°C pathways. Brown: GMST remaining below and stabilizing at 1.5°C in 2100; Green: a delayed start but faster emission reductions pathway with GMST remaining below and reaching 1.5°C earlier; Blue: a pathway temporarily exceeding 1.5°C, with temperatures reduced to 1.5°C by net negative CO <sub>2</sub> emissions after temperatures peak; and Yellow: a pathway peaking at 1.5°C and subsequently declining. Temperatures are anchored to 1°C above pre-industrial in 2017; emissions–temperature relationships are computed using a simple climate model (Myhre et al., 2013; Millar et al., 2017a; Jenkins et al., 2018) <sup>[[#fn:r122|122]]</sup> with a lower value of the Transient Climate Response (TCR) than used in the quantitative pathway assessments in Chapter 2 to illustrate qualitative differences between pathways: this figure is not intended to provide quantitative information. The time-integrated impact is illustrated by the semi-empirical sea level rise model of Kopp et al. (2016) <sup>[[#fn:r123|123]]</sup> .

<!-- END IMG -->
<div id="section-1-2-3-3-block-3" class="box"></div>

<span id="cross-chapter-box-1-scenarios-and-pathways"></span>