This study by Dunkl et al. used six earth system models’ output from the Decadal Climate Prediction Project to evaluate the inter-annual variability (IAV) of terrestrial gross primary production (GPP) and models’ predictability of GPP IAV. Overall, the manuscript is well organized and written. The spatial patterns of GPP IAV and their predictability from environmental drivers are shown for each model. Then the differences in GPP IAV and predictability between models are clearly presented. I recommend its publication after addressing minor issues.

After reading the text, I think implications of the findings should be addressed or discussed in-depth. Now in section 3, results are described but little linkage with findings/questions summarized in the introduction section. It is better to add more discussion and/or implications.

I am wondering whether regional results can be presented, such as Anders Ahlstorm et al., Science 2015. This would help our understanding on GPP IAV from models.

Dunkl et al analyze 6 Earth System Models and asses the predictability of interannual variability in gross primary production on land as a function of temperature, soil moisture, and radiation. 

Overall the writing in the paper is clear, and the authors describe an appropriate quantity of analysis and results. But there was very little in the way of contextualizing the results from their analysis. The authors do tie their findings frequently to the literature (see further comment #2 below), but as a reader I was left wanting to know more about why this mattered and how to go forward. These revisions can be made by the authors, although I think they are more substantial than minor revisions. 

Major comments:

1. It seems like there is a section missing with further discussion of implications. There is a description of the results, but then no discuss that puts these results in context. As a reader I wondered so what? What are the implications? Right now there are only a few sentences in the last paragraph of the conclusions. 

2. In general the results frequently appeals to other papers about why different models do or don’t do something without much inclusion of those explanations here. As a reader I needed more help knowing what those previous papers had found to put these results into context. Specific examples:

line 231-232: "due to a misrepresentation of photosynthesis (O’Sullivan et al., 2020)"

This needs more explanation. Misrepresented how? I skimmed the O’Sullivan paper and I didn’t find a specific “misrepresentation of photosynthesis” described, just that it matches poorly with observational products (but not *why*). This statement implies that we know why, and I’m guessing that we do not.

line 246 "complex phenological scheme" 

What do you mean specifically by complex? What is different about it compared to the other models? How do you know that the performance is better because of the complexity? That evidence isn't shown here.

line 307 "poor representation of soil moisture"

The paper cited (Qiao et al. 2022) uses reanalysis as their "truth" for soil moisture. Reanalysis is just another modeled product so this is a somewhat misleading statement. Better would be "poor match to other modeled products of soil moisture". Further I don't see any obvious indication in that paper that CANESM5 is worse than other models. Soil moisture is notoriously poorly constrained by lack of observations and widely varying between models. 

3. There is a lot of describing baseline climatic regions (i.e. “semi-arid” or similar)  that 1) assume that readers will know what regions the authors are referring to, and 2) fail to take into account that there are background climate biases that could shift those locations across models. I suggest that the authors come up with a way to display results that allows for consistency in background climate across models (mean annual T vs mean annual P space being one option). Or that the authors show analysis that is specific to one “climate” region to demonstrate their point. Just staring at maps it was hard to translate back and forth from the statements in the text to the figures.

4. I found the explanation of the predictable component and predictable fraction challenging. The language wasn't hard to read, but it is a way of quantifying something that I am not familiar with and it was hard to wrap my head around what it should tell me and why. I don't have specific suggestions here but encourage the authors, who know the method well, to consider if they can make it more intuitive to a reader encountering this way of quantifying predictability for the first time. Why is this the type of predictability the authors want to know? What is the interpretation of it?

Minor comments:

 line 155: "For the lead years five to ten, the effects of initialization are assumed to be negligible." I assume you mean atmospheric initialization? Carbon cycle initialization would sure matter still on those time scales! Leaves probably take 20 years to equilibrate in some of these models.

 

 line 239: "GSS and GSE" There are already a lot of acronyms in this paper. I suggest just writing these out.

 

 Figure 4 and discussion of Fig. 4. 

 Soil moisture to what depth? Please specify. Total column? Integrated to a similar depth? Surface? Ideally you would want root-zone weighted soil moisture. It isn’t mentioned how the soil moisture shown here compares to root zone.