Brody et al. are interested in selecting latitudes at which to emit aerosol precursor gas into the stratosphere in order to optimize climate change with regard to some scalar metrics such as temperature or precipitation changes. They do so on the basis of available climate model simulations and apply a simple emulator in the optimisation process.

The study in general is well conducted and the manuscript is well written.

As the only major suggestion, it would be very useful to test the results by implementing the optimal choice of injection into a different Earth system model (see also lines 142-143 that somewhat alludes to this idea).

Else I propose some minor edits.

l95 – It would be useful to motivate this surprising use of another, very differently conducted, simulation at this point.

l114-115 – to this point, the reader assumed the assumption of linearity is “validated” (e.g. abstract, line 11). If it turns out very problematic, this needs to be explained instead of assuring the reader about validation.

l121 / Table 1 – acronym “SSI” needs to be explained

l146 / Eq. 1 – a noise/natural variability term is missing

l152 – what quantity related to ITCZ, its annual-mean zonal-mean latitude?

l155 – or higher moments of the quantities, especially extremes would be interesting.

l174 – define “adequately” linear

l179 / Eq. 2 – a noise term is missing

l184 / Table 2 – “values”. Also I gather \alpha is for the SSP2-4.5 column, and all other columns list \mu? This should better be explained in the caption. The signs in the ITCZ location require explanation.

l192 – define T_0 (=T_0,warming)

l202 – How is this possible (in terms of seeding), a negative coefficient (see also line 222)? And also, why such a complicated combination, instead of straightforward combination?

l209 /Table 3 – additional digits required for P_0 and SSI.

Brody and coauthors provide a novel and useful analysis of design space of stratospheric aerosol injections in CESM2-WACCM. Specifically, it is interesting to see the trade-offs between optimizing temperature and precipitation responses extends to the pattern of response when designing SAI interventions, in addition to the well known trade offs in the global mean responses. The manuscript is well written and is largely ready for publication. However, I believe the analysis would be benefit substantially from a more rigorous assessment of the uncertainty due to natural variability. I am concerned that internal variability presents a significant constraint on the ability to optimize climate responses that must be quantified. Particularly when computing the sum of anomalies, the noise in the pattern can increase substantially as the number of signals being combined is increased (i.e., the sum of variances than the mean of variances). One approach may be to compute the Pareto front with all 18 possible combinations of individual ensemble members from each of the existing strategies. This could be used to generate a "cloud" of such Pareto fronts and uncertainty ranges for the dots in Fig. 6 and 8. 

Additional minor comments follow.

- Line 4: "latitudes of injection" -> "latitudes and altitudes of injection"

- Line 5: ", managing up to" -> ". For example, managing up to"

- Line 113:" the Atlantic Meridional Overturning Circulation" -> "rate dependent responses, such as the Atlantic Meridional Overturning Circulation (Hankel 2024)". Hankel, C. 2024 https://www.pnas.org/doi/abs/10.1073/pnas.2411357121

- Line 205-206: How is the standard error computed? Is is computed using variance across an initial condition ensemble?

- Line 370: "at low cooling" -> "at lower cooling"

- Line 403: "greater accuracy" -> "greater precision"

- Fig 6,8,9: I assume since the 1C cooling is the reference point, the overall emission magnitude can change? It would be useful to see total emissions would help see if there are overall SAI efficiency differences between the optimizations