Comments on “Mechanistic insights into tropical circulation and hydroclimate responses to future forest cover change” by Fahrenbach et al.

The manuscript presents an in-depth analysis of the biophysical effects of land-use changes, based on ad hoc simulations conducted with several Earth System Models (ESMs) participating in CMIP6-LUMIP. The results presented in the main text and supplementary material provide valuable insights into model responses to (mostly) increases in forest fraction. These increases result from either more intensive afforestation or reduced (avoided) deforestation under scenario SSP1 compared to SSP3, with the largest changes occurring in tropical Africa. In addition to the modeled changes in water fluxes and other key variables, the study computes several metrics designed to help understand the mechanisms driving changes in the surface water balance (P − ET).

Main comments:

As noted, the study is comprehensive and, based on the model intercomparison, provides clear conclusions regarding robust changes in precipitation (P), evapotranspiration (ET), and consequently P − ET (Conclusion 1), as well as on the independence of land use-induced effects from the background climate (Conclusion 2; note that it seems unusual not to include a figure in the main text to support this conclusion).

However, regarding the mechanisms of change (Conclusion 3) — where this study invests more effort and could be more innovative — the authors, in my view, overcomplicate the analysis, overlook well-known causal chains, and fail to provide a credible explanation. Understanding the biophysical effects of land cover changes is clearly not straightforward. The change in a given variable depends on (1) the direct impact of surface forcing (i.e., changes in land surface properties), which can involve various processes (e.g., changes in radiative or turbulent fluxes that alter the surface energy balance), and (2) the atmospheric responses to (1). The atmospheric response is key, as it feeds back onto surface variables, either amplifying or damping the initial effect (e.g., changes in water recycling), and can export the impact beyond the region initially perturbed. The resulting net effect of, for example, afforestation, depends largely on the region, climate, and the spatial scale of the modified area, among other factors.

Mechanisms of change may be analyzed from the different optics (more typically modification in water or energy budgets either at the surface or the atmosphere). This paper focuses on the atmospheric water balance, with some useful simplification and decomposition of it terms.

Starting the results description, it reads (lines 211-212): "This study seeks to identify the mechanisms driving the ∆(P −E) pattern over the tropics from a dynamics perspective and to reconcile the apparent mismatch between the tropical ∆(P − E) and ∆NEI”. It is not clear what the “apparent mismatch” refers to. Figure 2 shows a clear (and expected) response to tropical afforestation: increased ET and concomitant surface cooling. In turn, this change supplies moisture and latent heat to the atmosphere (Figures S4 and S10). Figure S4 also shows that the increase in NEI is primarily due to latent heating, partially offset by other radiative effects. Why should we expect a different result in this case?

In several parts of Section 3, it is stated that dynamic effects in the lower troposphere dominate or explain the changes in P − ET (e.g., lines 238–239, 248, 293–294, 318–319), leading to Conclusion 3. As noted earlier, the atmospheric response is indeed key, but it does not explain the primary response of the models to tropical afforestation—namely, the increase in ET (leading to the reduction in P − ET). As shown by numerous previous studies—many of which are cited in the introduction—this increase in ET is a direct consequence of changes in surface properties such as increased LAI, canopy conductance, and turbulence. This pattern clearly dominates in this set of simulations.

This response is clearer during the dry season, as observed in central-southern Africa during the austral winter, where the change in P − E corresponds almost entirely to ∆ET (Fig. 3). Naturally, a change rooted at the surface is then transmitted to the atmosphere, which can be analyzed through the water budget. In this case, increased ET leads to more humid air (Fig. S10), changes in atmospheric motion and moisture convergence, as illustrated by the omega approximation (Fig. 3f). However, this does not imply that changes in vertical motion and regional circulation are the primary causes of changes in P − ET, as the authors suggest; rather, these are atmospheric responses to surface forcing. I agree that the mechanisms discussed in the paper are relevant—particularly for explaining changes in P, when present—which may, in turn, modulate ∆ET, but the explanation and conclusions should carefully follow a consistent causal chain and avoid reversing it.

Another interpretation that seem at least partially incorrect, yet presented as “true” throughout the paper, including in the conclusions and both abstracts, is that the reduction in (near-)surface wind is due to the drag effect of increased surface roughness. In contrast to the previous case, here the authors attribute an atmospheric response to afforestation entirely to a change in a surface property (i.e., roughness), without providing convincing evidence. While this should be a contributing factor, other well-known mechanisms could also contribute to—or even primarily drive—this response. One common mechanism involves temperature-induced changes in regional (monsoonal) circulation, which is completely overlooked in this case, despite all relevant indicators being present: a significant surface cooling and a concomitant sea-level pressure increase in central Africa (Fig. S10). Given that the mean pressure gradient and wind are directed toward the interior of the continent (Fig. 5), the resulting pressure increase would be expected to weaken the monsoonal circulation. The change in wind stress is not definitive evidence of the proposed mechanism, as it may instead result from changes in the low-level circulation. Moreover, the authors do not specify how wind stress was calculated.

These main comments affect a core conclusion of the paper, so the recommendation is for major revisions. However, all of the issues relate to the interpretation of results, many of which could be addressed through a re-assessment of the existing analyses.

Some specific comments:

• Lines 37–39: Canopy conductance/resistance is also a key factor.
• Line 66: Runoff is defined locally (or in a grid cell in a model). The integrated runoff over a basin leads to river streamflow.
• Line 100: 1000 what? (Units are missing)
• Definitions in Section 2.2.1 (and throughout the text): Moisture, wind, and vertical velocity are, by definition, zero at the surface. These quantities must therefore be near-surface values. What level do they correspond to — 2 m, 10 m, or the lowest atmospheric model level? This is particularly relevant for the wind-based metrics used in the paper.
• Explicitly state pressure vertical velocity throughout the text, as it is omitted in several sections.
• Line 171: The phrase “three-dimensional” is unnecessary here.
• Lines 195–196: P − E defines surface runoff.
• Line 258: Again, there is no such thing as “surface vertical wind”. The level used for near-surface circulation analysis must be clearly defined.
• Lines 263–266 and Fig. 5: This is very confusing. Why not directly use the pressure levels provided in the model outputs?
• Lines 422–424: I agree that having a large model ensemble allows for more robust conclusions, but model differences are also of great interest. The authors could elaborate more on this in the discussion.
• As noted in the manuscript, a large ensemble of simulations also allows for an increased signal-to-noise ratio. Yet, although scientifically relevant, the signal may not be particularly significant from the perspective of its impact on natural or human systems. In this sense, the authors could further discuss the intensity of the projected changes and their implications for, e.g, water availability, temperature, etc.
• Figure S6: This figure shows absolute values (not changes), correct? If so, the delta symbol should be omitted.

If the symbol Δ refers to different states in forest cover, this study seeks to identify the mechanisms driving the Δ(P −E) pattern over the tropics from a dynamics perspective and to reconcile the apparent mismatch between the tropical Δ(P −E) and ΔNEI, where P is precipitation, E is evaporation and NEI is net energy input into the atmosphere.  The methodology is based on comparing future scenario simulations from seven multi-ensemble models participating in the Land Use Model Intercomparison Project (LUMIP).

The main finding are as follows, in which are most pronounced over Africa

1. Afforestation and avoided deforestation lead to a robust increase in precipitation and evapotranspiration, alongside widespread decreases in net precipitation (precipitation minus evaporation) in the tropics.
2. The tropical hydroclimate response to future forest cover change is largely independent of the background climate under low- and medium-warming scenarios.
3. The changes in net precipitation over Africa are driven by the competing effects of surface drag-induced reduction of lower-tropospheric winds and net energy input-induced strengthening of deep-convective upper-tropospheric circulations.

The text itself is exceptionally well written.  The review part is particularly interesting.   The emphasis is on mechanisms rather than on a simple description of differences between simulated climate. 

I would have only two comments:

• Why is the response stronger over Africa? Is there a signal over the Amazon?
• Why aren’t the PBL processes considered? Where is the PBL top in Fig. 8?