I appreciate the authors' work to respond to my concerns (especially the cross-validation against published studies in the results and discussion sections) and improve the quality and readability of this manuscript. I have no further general questions but one major request, to simplify the languages used in this manuscript. The following minor revision suggestions (marked in italic format) are based on the line number in the manuscript version with tracking changes and hopefully can help authors to find directions about how to squeeze out unnecessary wordings, but authors shall double check their writings, especially the newly added content in discussion section:
Line 66: “There may also be a flux of fossil fuels directly into the ocean or land surface via for instance fossil fuel extraction and other leaks (Roser and Ritchie, 2022), but these are not generally included in Earth system models.”
can be simplified.
“There may also be a direct flux of fossil fuel extraction and other leaks (Roser and Ritchie, 2022) into the ocean or land surface, but are not included in Earth system models.”
Line 71: “Figure 5.31 of Canadell et al. (2021) shows the cumulative carbon emissions against global mean temperature change for several projections. That figure shows a strong correspondence between emissions and warming which appears to be scenario independent. “
These two sentences can be merged.
“Figure 5.31 of Canadell et al. (2021) shows the cumulative carbon emissions against global mean temperature change for several projections, but the strong correspondence between emissions and warming appears to be scenario independent.”
Line 83: “For instance, atmospheric carbon allocation is 30% in SSP1-1.9 of the carbon remaining in the atmosphere in the year 2100, but in SSP5-8.5, that value is 62%. While the land and ocean carbon uptake are expected to remain approximately equal, the uncertainty is much larger for the land carbon sink than the ocean. In the land, some of the uncertainty is due to the balance of increased land carbon accumulation in the high latitudes and loss of land carbon in the tropics (Canadell et al., 2021). Further uncertainty arises from the challenges of forecasting the water cycle, including droughts that reduce carbon absorption potential of the land surface. On the other hand, the ocean CO2 sink is strongly dependent on the emissions scenario. This absorption of carbon into the ocean reduces the mean global buffering capacity and drives changes in the global ocean’s carbonate chemistry (Jiang et al., 2019; Katavouta and Williams, 2021). These projections are based on data from the Coupled Model Inter-comparison Project (CMIP), and the most recent CMIP round, CMIP6, is described in sec. 1.1.”
can be simplified.
“For instance, projected 2100 atmospheric carbon allocation from CMIP6 is 30% in SSP1-1.9 but rises to 62% in SSP5-8.5. While the land and ocean carbon uptake are expected to remain approximately equal, the uncertainty is much larger for the land carbon sink than the ocean. Uncertainty from the land sink is a tradeoff between the accumulated land carbon in the high latitudes and loss of land carbon in the tropics (Canadell et al., 2021) and the challenges of forecasting the water cycle, including droughts that reduce carbon absorption potential of the land surface. On the other hand, continuous absorption of carbon into the ocean reduces the mean global buffering capacity and drives changes in the global ocean’s carbonate chemistry (Jiang et al., 2019; Katavouta and Williams, 2021), building a strong dependency on the choice of scenarios.”
Line 98: “Earth System models (ESMs) are one of the main tools to study the climatic impact of the combustion of fossil fuels, and they are the only tools capable of projecting the future coupled carbon-climate system. The Sixth Coupled Model Inter-comparison Project (CMIP6) (Eyring et al., 2016) is the most recent in a series of global efforts to standardise, share and study ESM simulations. To participate in CMIP6, models must simulate a suite of standard simulations and meet certain model quality and data standards. These standard simulations (also known as the Deck) include a pre-industrial control, a historical simulation, a gradual 1% CO2 growth experiment and a rapid 4xCO2 experiment. The quality requirements include a drift in the air-sea flux of CO2 of less than 10 PgC per century, and a drift in the global volume mean ocean temperature of less than 0.1 degrees per century (Jones et al., 2011; Eyring et al., 2016; Yool et al., 2020).”
can be shortened.
“Earth System models (ESMs) are the only tools capable of projecting the future coupled carbon-climate system. The Sixth Coupled Model Inter-comparison Project (CMIP6) (Eyring et al., 2016) is the most recent global effort to standardise, share and study ESM simulations. CMIP6 designed standard simulation protocols (also known as the Deck) include a pre-industrial control, a historical simulation, a gradual 1% CO2 growth experiment and a rapid 4xCO2 experiment. For quality assurance, only results with a global drift per century lower than 10 PgC in the air-sea CO2 flux and lower than 0.1 degrees in the volume mean ocean temperature are accepted (Jones et al., 2011; Eyring et al., 2016; Yool et al., 2020).”
Line 108: “In order to make projections of the future anthropogenic climate drivers, multiple scenarios were proposed in the ScenarioMIP project to cover a wide range of potential futures. ScenarioMIP expands upon the CMIP6 core simulations and multiple scenarios are available for modellers to use to generate simulations (O’Neill et al., 2016). We include the scenarios: SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5 (O’Neill et al., 2016; Riahi et al., 2017). Scenario names in CMIP6 are comprised of a general future pathway (SSP1-SSP5) followed by an estimate of the radiative forcing at the year 2100 in units of Wm−2. These scenarios cover a wide range of possible futures, including sustainable development in the SSP1-1.9 and SSP1-2.6 scenarios. The intermediate emissions scenario or “middle of the road” pathway in SSP2-4.5 extrapolates historic and current global development into the future with a medium radiative forcing by the end of the century. The regional ri90 valry scenario, SSP3-7.0, revives nationalism and regional conflicts, pushing global issues into the background and resulting in higher emissions. Then finally, the enhanced fossil fuel development in SSP5-8.5 is a scenario with the highest feasible fossil fuel deployment and atmospheric CO2 concentration (Riahi et al., 2017).”
can be more succinct.
“ScenarioMIP expands upon the CMIP6 core simulations with multiple scenarios of the future anthropogenic climate drivers to cover a wide range of potential futures (O’Neill et al., 2016). Scenario names in CMIP6 are a general future pathway (SSP1-SSP5) followed by an estimate of the radiative forcing at the year 2100 in units of Wm−2. These scenarios (O’Neill et al., 2016; Riahi et al., 2017) include sustainable development scenarios SSP1-1.9 and SSP1-2.6, the intermediate emissions scenario SSP2-4.5 with a medium radiative forcing by the end of the century, the regional rivalry scenario, SSP3-7.0 pushing global issues into the background and the enhanced fossil fuel development in SSP5-8.5 with the highest feasible fossil fuel deployment and atmospheric CO2 concentration (Riahi et al., 2017).”
Line 121: “Given the same rise in atmospheric CO2 concentration, each ESM will warm by a different amount due to the significant structural and parametric differences between models. The Equilibrium Climate Sensitivity (ECS) is a measure of this sensitivity to CO2. The ECS is given in ◦C Celsius and represents the long-term near-surface air temperature rise that is expected to result from a doubling of the atmospheric CO2 concentration once the model has reached equilibrium. In effect, the ECS is an indicator for how much warming occurs in a model with a doubling of CO2. The most recent 5-95% assessed natural ECS range was between 2 ◦C and 5 ◦C, the likely ECS range was 2.5 - 4 ◦C, and the most likely value was 3 ◦C (Arias et al., 2021, TS6). The wide spread ECS values in climate models is one of the causes of uncertainty for the timing of when forecasts reach certain warming levels. The “allowable emissions” that keep global temperature rise within policy targets are equally impacted (United Nations Treaty Collection, 2015). This has been exacerbated in the latest round of CMIP, as the CMIP6 generation of ESMs has a broader range of sensitivities than previous generations. Several CMIP6 models have a stronger response to atmospheric carbon than any CMIP5 model, and many sit above the likely ECS range (Arias et al., 2021, TS6.)”
can be more succinct.
“The Equilibrium Climate Sensitivity (ECS) is a measure defined as the near-surface air temperature rise in ◦C Celsius from a doubling of the atmospheric CO2 concentration once the model has reached equilibrium. The wide spread ECS values in climate models is one of the indicators of uncertainty for the timing of when forecasts reach certain warming levels. Due to the demand to keep global temperature rise within policy targets (United Nations Treaty Collection, 2015), more scenarios based on “allowable emissions” exacerbated a broader range of ECS in CMIP6 models than previous generations. Several CMIP6 models have a stronger response to atmospheric carbon than any CMIP5 model, and many sit above the likely ECS range. The most recent 5-95% assessed natural ECS range was between 2 ◦C and 5 ◦C, the likely ECS range was 2.5 - 4 ◦C, and the most likely value was 3 ◦C (Arias et al., 2021, TS6).”
Line 140: “Climate change policy can often focus on the climate at specific target years, like 2050 or 2100 (United Nations Treaty Collection, 2015; IPCC, 2021a). However, due to the wide range of ECS values in ESMs, this can mean that ensembles at the year 2100 are composed of a set of models with significantly different behaviours. This wide range in the temperatures and warming rates at a given point in time has knock-on effects on feedbacks and may inhibit the realism and representivity of the ensemble 110 multi-model mean (Hausfather et al., 2022; Swaminathan et al., 2022). Instead of specific target years, we can alternatively focus on model behaviour at specific Global Warming Levels (GWL), such as 2 ◦C, 3 ◦C or 4 ◦C of warming relative to the pre-industrial period. By investigating the system’s behaviour at specific warming levels instead of target years, we can account for the impact of climate sensitivity and make policy relevant assessments while still exploiting the full ensemble of CMIP6 models. This allows us to maintain model democracy, even in a so-called “hot model” ensemble. The 2, 3 and 4 ◦C GWLs were chosen because the 2 ◦ C GWL is a key target set in the 2015 Paris Agreement (United Nations Treaty Collection, 2015) and thought to be a threshold for potentially dangerous climate change.”
Need to be polished.
“Climate change policy focuses on the climate at specific target years, like 2050 or 2100 (United Nations Treaty Collection, 2015; IPCC, 2021a). However, due to the wide range of ECS values in ESMs, models with significantly different behaviours projected wide range of warming rates at a given point in time, which has knock-on effects on carbon feedback and reduce realism and representativeness of the multi-model ensemble mean (Hausfather et al., 2022; Swaminathan et al., 2022). Instead of specific target years, we alternatively focus on model behaviour at specific Global Warming Levels (GWL), including 2 ◦C, 3 ◦C or 4 ◦C of warming relative to the pre-industrial period for policy relevant assessments while still exploiting the full ensemble of CMIP6 models. This allows us to maintain model democracy, even in a so-called “hot model” ensemble. The 2 ◦ C GWL is a key target set in the 2015 Paris Agreement (United Nations Treaty Collection, 2015) and thought to be a threshold for potentially dangerous climate change.”
Line 186: “The NBP is an prognostic variable calculated” shall be “a prognostic variable”
Line 297: “The left side shows the percentage allocation, and the right side shows the totals in PgC.” Move figure explanation to the figure caption.
Line 299/300: “More carbon is allocation” shall be “More carbon is allocated”
Line 300: revise to “is allocate”
Line 335: “Figure 2 only shows the multi-model means, not single models. This means that multi-model means that do not reach the GWL are not included in this figure.”
Move this to the corresponding figure caption.
Line 387” “This figure includes a pair of panes for each experiment scenario. For each pair, the top pane is the cumulative carbon in PgC and the bottom pane shows the percentage.”
Move to the figure caption.
Line 413: “This results in the saw-tooth pattern on the right of this figure. However, this saw-tooth pattern does not appear on the left side of the figure, as the ratios of carbon allocation between land, ocean and atmosphere at a given GWL are not dependent on ECS. ”
can be simplified.
“However, the ratios of carbon allocation between land, ocean and atmosphere at a given GWL are not dependent on ECS, since the pattern is relatively smooth comparing to the C emission.”
Line 426: “Firstly, at a given GWL, higher emission scenarios have a higher atmospheric fraction. In effect, the SSP5-8.5 scenarios have a higher atmospheric fraction than SSP1-1.9 and SSP1-2.6 scenarios, even at the same GWL. Similarly, higher emission scenarios have a smaller land fraction, while the ocean fraction is similar across scenarios at the same GWL. Secondly, warmer GWLs have a larger atmospheric fraction than cooler GWLs. Thirdly warmer GWLs have a smaller land fraction than cooler GWLs. Finally, the ocean fraction is relatively consistent between GWLs and scenarios.”
The expression shall be simplified.
“At a given GWL, higher emission scenarios have a higher atmospheric fraction, but a lower land fraction and a relatively consistent ocean fraction. When comparing the allocation fractions from different GWLs, warmer GWLs have larger atmospheric fractions, lower land fractions and consistent ocean fractions than colder GWLs.”
Line 432: ”The data from fig.4 is re-framed in fig.5 as a series of scatter plots. In this figure, each row represents a different scenario, and each column is a different dataset. These datasets are: the GWL threshold year, the total carbon allocated, the carbon allocation for each domain and the fractional carbon allocation to each domain. The y-axis shows the model’s ECS, and each point is a different GWL, where the squares are 2 ◦C GWL, the circles are 3 ◦C GWL, and the triangles are 4 ◦C GWL.” Move to the figure caption.
Line 437: “For each group of data, the line of best fit is shown and the absolute value of the fitting error (Err) of the slope (M) over the slope is shown in the legend, as Err/M. The fitting error, Err, here is the standard error of the estimated gradient under the assumption of residual normality. This value indicates whether the slope crosses the origin within the 95% confidence limit. If the uncertainty on the slope is greater than the slope itself (and Err/M exceeds unity), then we can assume that the fit is not statistically significant. All groups with three models or fewer that reach the GWL were excluded as this is not enough data points to draw meaningful conclusions.
The goal of this figure is to highlight in broad strokes the ways that ECS interacts with carbon allocation in these models. In most of the fits, the data and the ECS are inversely correlated such that lower ECS models have higher values. This appears to be true for GWL year, total carbon and the individual total carbon fields in the atmosphere, ocean and land. The GWL threshold year and the total carbon allocations both have all absolute Err/M values lower than unity and as such are both related to ECS. In both the ocean and the atmosphere’s total carbon, the absolute value of Err/M is always smaller than one. This means that the total carbon in both the ocean and the atmosphere are linked to ECS with 95% confidence. However, this is not the case for the ocean or the atmosphere’s carbon allocation as a percentage and in many cases absolute Err/M is greater than unity. This means that we can not say that the fraction of carbon allocated to the ocean or to the atmosphere is related to the ECS with 95% confidence. Similarly, this absolute Err/M ratio is not consistently below unity for the land ensembles at all GWLs. This implies that the total or percentage land carbon allocation is likely to be not correlated with ECS.”
This part is way too verbose. I suggest to revise:
“For each group of data, the line of best fit is shown and the absolute value of the fitting error (Err, the standard error of the estimated gradient under the assumption of residual normality) over the slope (M) is shown in the legend, as Err/M. This value indicates whether the slope crosses the origin within the 95% confidence limit (< 1) or not (> 1). All groups with three models or fewer that reach the GWL were excluded as not enough data points to draw meaningful conclusions.
GWL year, total carbon and the individual total carbon allocation fractions are inversely correlated to ECSs. The GWL threshold year and the total carbon allocations both have all absolute Err/M values lower than unity and as such are both related to ECS. The total carbon in both the ocean and the atmosphere are linked to ECS, as their Err/M are smaller than 1. However, the correlations between carbon allocation fraction of the ocean or the atmosphere and ECS are not statistically significant. For land, both total carbon sink and allocation fraction are not significantly correlated to ECS at all GWLs.” |
