A global map of Earth system interactions
Abstract. The intricate interplay of the biophysical processes of the Earth system provides the basis for Earth resilience and human well-being. With local anthropogenic pressures increasing in most regions, there has been a growing need for a systemic understanding of this interplay on a sub-global scale. However, due to inconsistency in the temporal and spatial scales in the corresponding studies, a holistic assessment of the environmental impact of local human activities remains challenging. We take a step in the direction of a uniform framework for estimating, exploring, and communicating the spatially resolved global pattern of crucial Earth system interactions. We focus on the processes of change in carbon dioxide concentration, natural vegetation cover, and surface water runoff, representing the major components of the Earth system of climate, land, and the global water cycle, respectively. In a first step, we quantify local interactions based on historical simulations of the dynamical global vegetation model LPJmL. In a second step, we approach the question of a global partition into coherent interaction zones, illustrating the risk of information loss that comes with established spatial aggregations. Following a top-down approach first, we map the global interaction pattern on common natural partitions of the Earth. Cluster validity indices reveal a close alignment between the effect of land use change on climate and biogeographic classification by biome. In contrast, the effects of land use change and climate change on surface water runoff are best captured by the Köppen-Geiger climate zones. Following a bottom-up approach, we use multivariate spatially constrained clustering to derive integrative global partitions and analyze the interaction profile of the resulting clusters. Showing particularly strong combined effects, we identify several patches of tropical rainforest on the Indomalayan islands, as well as large areas of warm grasslands in Australia, as high-impact clusters with respect to secondary effects of human pressures on land and climate. Our study emphasizes the local nature of the interplay of Earth system interactions as well as both the risks and potential that spatial aggregation entails.
I read this paper by Zoller et al. with interest. The title and abstract led me to accept the request to review, but on reflection I am not sure I am really a suitable reviewer. I would advise the Editor to treat my comments and assessments with due caution.
First of all, it's impossible to not agree with the authors that there is considerable value in communicating the spatially resolved global pattern of crucial Earth System interactions. The fundamental question, which might highlight my lack of understanding, is that the authors do not actually do this according to what I understand to be Earth System interactions. So far as I can tell (in part because the Methodology is not really very thorough) is that they use a global reanalysis at ~50 km (from 1901 to 2013) to force LPJmL offline (uncoupled) from the atmosphere. They do not tell me anything about this reanalysis, only referring to a paper by Lade et al. (2021). I accept that if this paper fully explains the reanalysis then it does not have to be fully re-explained. But the paper I am reviewing here says nothing except the spatial resolution. It does not even name the reanalysis I think. Fundamental to reanalyses are data that form the reanalysis and that cannot conceivably be equally thorough across the whole world. Does this matter? How sensitive are the results presented here to uncertainty in the reanalysis? Looking at Figure 3 I would suggest that a lot of the red in Figure 3b, and almost all the patterns in Figure 3c and 3d are located in places where the reanalysis would be least reliable. So, does this matter? If the reanalyses are perturbed by (say) +/- one standard deviation how do the patterns change? I am not saying the results are wrong, but I am saying that I cannot tell if the results are robust and that is a problem for a reviewer.
Second, and understandably, the authors only use LPJmL. This is a highly respected model, but it is only one model. I am not suggesting that the authors need to repeat this with other dynamic vegetation models but how robust are the results to minor changes in LPJmL? There are plenty of examples in the literature that suggest LPJmL is a good model, but fairly there are also plenty that suggest it has its issues like any global model does. So, do the maps and patterns change if elements of the model are modified? Of course you cannot do a full analysis of this, but some effort to determine which of the results are robust to some of the uncertainty in LPJmL feels warranted.
Next, this paper is about “Earth System interactions” but LPJmL is not coupled. Its forced by a reanalysis. In my naïve thinking about Earth System Interactions I think of a coupled system where LPJmL is coupled to (say) ECEarth. I think of something like the GLACE experiements undertaken and published by Koster et al. We know that land models behave very differently once coupled. It is inconceivable to me that this would not be true of LPJmL. So, how robust are the results to being run offline and uncoupled? Might this affect the conclusions? There is little hint of uncertainty in the conclusions.
Its hard to provide mode detailed comments because the manuscript’s line numbering seems a little inconsistent. But:
Page 3 why is the aggregation step omitted and does it matter?
Page 9, line 187 I doubt this explanation is true – I suspect its linked to an increase in evaporation.
Page 10, line 213 – I thought this was regionally specific?
Finally, all the conclusions might be 100% right, but I cannot tell. I cannot determine if they are broadly real, or merely the consequence of using one model with one reanalysis. I am not sure this meets the criteria for a publication in a significant journal.
So, I am a little challenged in reviewing this paper. It feels like a valid piece of work that might interest a lot of people but I honestly cannot determine whether the results are artefacts of the reanalysis, or of LPJmL, or of the other techniques in the paper. It feels very much like "trust me" and I try to be skeptical when reviewing papers. So, at least for me, this paper needs a fundamental re-write to *not* merely present results, but to provide rigour to demonstrate that the results are sound and can be interpreted beyond this specific modelling system.