Summary

The paper by Parry et al. titled “Solar Radiation Modification is projected to increase land carbon storage and to protect the Amazon rainforest” investigate the impact of reducing anthropogenically-induced global warming under SSP585 towards SSP245 level through stratospheric aerosol injection (SAI) in an ensemble of CMIP6 ESMs. Overall, the paper is well written and clear. Though the conclusion that SAI (or most SRM) could increase the land carbon storage through reducing the soil respiration is not novel, they cleverly frame the focus (of their NPP analysis) on the Amazon rainforest under the context of Earth’s tipping element. The uniqueness of the study lies in the use of multi-ESMs, confirming the robustness of earlier results, but the selection of these models should be justified. For instance, do they well simulate the present-day observed states and variability? SRM terms are often used interchangeably with SAI, which is incorrect and the discussions, with respect to uncertainties could be improved.

Main comments

The title suggests that the Amazon rainforest would be endangered if SRM is not applied. But nowhere in the paper it is shown that the models simulate a detrimental effect of climate change should SAI is not applied. Do any of the models capable of simulating Amazon forest dieback under the SSP245 or SSP585? It is not clear if dynamical vegetation is implemented in these models and what are the implications of not including it



Sect.2.2 Validation. It is difficult to validate the patterns shown in Fig. 1 when observational-based estimates are not included. Please include estimates e.g., from Piao et al. (2020). The authors indicate that the spatial pattern is comparable, but how about the spatial magnitudes? Actually, I am not convinced that the comparison with volcanic eruption is useful. As stated, the impact of eruption differs based on climate states (here you can also cite Frolicher et al. 2013;  https://doi.org/10.1002/gbc.20028), and given the large ESMs spread, it looks like the solid black line is insignificant (Fig. 2).



Throughout the paper (inc. Fig. captions), the ‘SAI’ experiments are often referred to as ‘SRM’. Use SAI throughout the text, as SRM refers to a broader radiative-based climate engineering. The definition on L29 is also incomplete, e.g. SRM also includes cirrus cloud thinning, which is not primarily aimed at reflecting sunlight. Differences in SRM methods could lead to considerably different impacts (e.g. for the Amazon precipitation patterns: Park et al., 2019; https://doi.org/10.1029/2019GL084210). Instead, I recommend using the IPCC definition of SRM:

“Refers to a range of radiation modification measures not related to greenhouse gas (GHG) mitigation that seek to limit global warming. Most methods involve reducing the amount of incoming solar radiation reaching the surface, but others also act on the longwave radiation budget by reducing optical thickness and cloud lifetime.”



Information on ‘Fire’ representation is included in Table 2, but its impact is not discussed in the paper. Please clarify.

CO2 fertilization is often credited as the ‘primary’ reasons for the enhanced NPP and land carbon storage. For NPP, this is only true if precipitation remains at a sufficient level, right (i.e., if precipitation decline considerably, higher CO2 will not enhance NPP, as happens in IPSL model)? For land carbon storage, this is true if the vegetation biomass mostly contributes to the land carbon increase. But earlier studies have indicated that the higher land C storage is due to cooling-induced less soil respiration.

Given the emphasis on the Amazon region, the authors should provide a more detailed pretext on how well these ESMs simulate the NPP (e.g., seasonal cycle as compared to flux tower estimates), vegetation or soil carbon storage, mean precipitation, temperature, etc. in the contemporary Amazon region. It is my understanding thats ESM underestimate precipitation in this region (Hagos et al., 2021; https://doi.org/10.1175/JCLI-D-20-0211.1), and would be necessary to elaborate what are the implications of this and other biases on your conclusions.



From the paper, the readers get the message that SAI can be used as an emergency protection against loss of Amazon rainforest. While this is implied by their findings of NPP changes, SAI has also been also shown to induce negative impacts on NPP and land carbon storage in many other regions (high latitude NH as shown in this study). This result should be conveyed accordingly. In addition, the first sentence of conclusion can be expanded to highlight the regional disparity in the benefit of SRM with some examples. Given the controversial topic of SRM, it is the authors’ responsibility to not be partial in delivering their messages.



Finally, it should be mentioned that SRM is a temporary mitigation measure, and it doesn’t address the primary drivers of climate change (growing level of atmospheric CO2), despite the higher land C storage. Hence, when terminated, the issue would likely reappear at a considerably faster rate (Jones et al., 2013, https://doi.org/10.1002/jgrd.50762; Muri et al., 2018, https://doi.org/10.1175/JCLI-D-17-0620.s1), with potential of triggering other tipping element of the Earth system.



Minor comments

Title: Stratospheric Aerosol Injection instead of SRM, and perhaps “SAI has the potential to increase ….”

L1-2: Change SRM definition.

L9: define G6suplhur or remove “under G6suplhur”

L51:  reduce the radiative forcing

L53: remove instead

L56: G6Sulpur => G6sulfur

L56: from a higher

L60: cite Lee et al. (2021); https://doi.org/10.5194/esd-12-313-2021

L102: cite Frolicher et al. (2013); https://doi.org/10.1002/gbc.20028

L112: 1850-1899, but in Figs. 3&18 captions: 1850-1900

L230: This statement appears not well supported, i.e., precipitation could be more important at regional scale. I suggest producing figures similar to Fig. 18, but for the different regions mentioned prior to this sentence. Include precipitation, in addition to T and CO2, for each individual model.

L237: add “(Fig. 8c)” after SSP585.

L244: percentage changes are mentioned in several places. Suggest including them in Table S1.

L245: ‘primarily due to the CO2 fertilization effect’, but not the case for IPSL (Figs. 14&15) despite having similar high CO2.

L256: CESM2-WACCM also show relative weak decline in precipitation here, so it seems that CO2 fertilization alone is insufficient (Fig. 13).

L268: reference studies that show regionally-varying NPP sensitivity to warming (e.g., Cramer et al., 2001, https://doi.org/10.1046/j.1365-2486.2001.00383.x; Tjiputra et al., 2010, https://doi.org/10.5194/gmd-3-123-2010).

L282-4: please elaborate or give examples.

L291: describe G6solar



Table 1: why do CNRM and MPI models starts from 2015 instead of 2020?

Fig 1 caption and elsewhere: remove Gunung from (Gunung Agung)

Fig 3: Looks very smooth. Specify if you applied running average.

Figs 7,17 caption: S3 => S2

Fig 18: improve resolution. Also clarify that this is a multi-models mean(?). If yes, how are these calculated since each model simulates different warming and CO2 evolutions? Did you apply smoothing?

Fig 18 caption: remove comma: ‘and CO2’

Please find my referee report attached