Publisher’s note: the content of this comment was removed on 13 February 2026 since the comment was posted by mistake.

Publisher’s note: the content of this comment was removed on 13 February 2026 since the comment was posted by mistake.

We sincerely thank the referee for the critical and insightful comments on our manuscript. We particularly appreciate the referee’s emphasis on the inherent nonlinearity and saturation in the TF–LAI relationship and the potential bias introduced by imposing a linear sensitivity. This feedback has motivated us to re-evaluate and strengthen both our methodological framework and the interpretation of θ in the revised manuscript.

Major Comment 1:

Response: We will comprehensively improve figure readability by increasing font sizes (axes, legends, titles, annotations), exporting higher-resolution figures, and optimizing multi-panel layouts and spacing to ensure clarity in the journal format.

Major Comment 2:

Response: We agree this is a key point. To avoid bias from forcing a nonlinear TF–LAI relation into a linear form, and to better reflect biophysical constraints, we will revise both the methodology and interpretation:

1. Method: we will extend θ estimation from a simple linear sensitivity to a bounded nonlinear framework, applying a logit transform to keep TF within (0,1) and including a quadratic term to capture curvature/saturation. We will retain the linear approach as a benchmark and provide robustness checks comparing both approaches.
2. Interpretation: we will explicitly define θ as an emergent marginal sensitivity/elasticity of TF to LAI, whose long-term changes can be driven by component saturation and energy/water constraints, and thus should not be over-interpreted as a direct shift in plant water-use strategy. We will clarify that VPD/SM effects on θ are largely indirect via constraints on absolute transpiration and partitioning.

Major Comment 3:

Response: We agree and will strengthen the manuscript in two ways:

1. Method clarification: we will state explicitly that analyses are based mainly on growing-season aggregated annual series and sliding-window low-frequency variability to reduce synoptic noise, and that partial correlation/regression quantify statistical independent contributions.
2. Discussion: we will add a dedicated discussion that SM–VPD feedbacks under drought can limit strict causal separability; thus we will interpret results as converging statistical evidence across methods and explicitly state the boundary in the limitations section.

Major Comment 4:

Response：

We agree and will temper the physiological interpretation. In the revision, we will avoid equating θ with a directly observed physiological parameter and instead present it as an emergent ecohydrological sensitivity metric within the gridded product framework. We will add a dedicated limitations paragraph on dataset uncertainty and interpretational boundaries and revise wording to be more cautious (e.g., “may suggest/indicate” rather than “demonstrate”).

Major Comment 5:

Response: Thank you. We will correct the GLEAM4 description following official documentation: assimilation is primarily applied to surface soil moisture, and the hybrid ML component is implemented mainly within the vegetation transpiration module rather than across the entire model.

Major Comment 6:

Response: We agree and will add explicit collinearity diagnostics. In the revision, we will:

1. compute VIF (or tolerance) to quantify redundancy between SM and VPD and report distributions by climate zone and NF/PF in the Supplement;
2. clarify that ridge regression is adopted specifically to stabilize coefficient estimation under collinearity, and use relative contribution (RC) to compare the roles of VPD and SM;
3. emphasize that conclusions are supported by convergence across multiple methods rather than a single model.

Major Comment 7:

Response: We will streamline repetitive result summaries and refocus the Discussion on mechanisms, comparison with prior studies, and possible sources of discrepancies. We will also add a dedicated “Methodological limitations” subsection addressing dataset uncertainty, SM–VPD coupling/feedback, scale mismatch, and interpretational boundaries of θ.

Specific Comments Line 15–16:

Response: We will revise the text to clearly separate datasets (forest-type maps and gridded products) from analytical frameworks (sliding window for temporal evolution) and inference methods (partial correlation, ridge regression, and mediation analysis).

Specific Comments Line 75–79:

Response: We will add appropriate references to support these statements and ensure that each key claim is properly cited.

Specific Comments Line 123:

Response: We apologize for the grammatical error. We will correct this sentence in the revised manuscript to read: "This dataset ... was produced by a machine learning algorithm ...".

Specific Comments Line 188:

Response: We will update the legend of Figure 2b to explicitly state that the left bars represent Natural Forests and the right bars represent Planted Forests.

Specific Comments Line 214–216:

Response: We will clarify that TF and θ are computed from growing-season aggregated annual series (growing-season accumulated values), and that an 11-year sliding window is applied on the annual series to generate a θ time series for trend analysis.

Specific Comments Line 273–279:

Response: We will integrate the “Trend Analysis and Testing” content into Section 2.3.4 so that trend estimation/testing and attribution steps are described coherently in one place.

Specific Comments Line 320:

Response: We appreciate this observation. We will revise the description in the revised manuscript to more accurately characterize the actual distribution pattern shown in Figure 4a. Specifically, we will modify the text to strictly align with the observed data trends.

Specific Comments Line 458-459:

Response: We agree that the phrase is ambiguous. We will replace “non-linear trend of increasing” with a precise description consistent with the results, e.g., “a significant increasing trend with pronounced interannual variability” (or other wording strictly aligned with the observed pattern).

We are grateful to the referee for the positive evaluation of our work and the recognition of its significance in understanding forest-water relations. The constructive suggestions, particularly regarding the validation of transpiration sensitivity and the separation of climatic drivers, have been very helpful in improving the quality of this study.

Comment 1:

Response: We agree that in situ validation is important. In the revision, we will conduct a feasible “minimum validation” using available eddy-covariance and/or sap-flow sites by constructing TF and θ consistently with our definitions and evaluating whether the sign/magnitude of θ variability aligns with our gridded results and climate drivers. If site coverage is limited, we will explicitly state the sampling limitations, treat the validation as qualitative/semi-quantitative consistency checks, and outline future expansions for systematic validation.

Comment 2:

Response: We will provide an operationally clear separation strategy and add diagnostics/robustness evidence:

1. Spatial gradients: use 2D binning with a control-variable logic (compare θ across VPD groups within similar SM bins, and vice versa) to estimate net effects;
2. Temporal dynamics: use partial correlation (net association controlling the other variable) alongside ridge regression (stabilized coefficients and contribution decomposition under collinearity);
3. Diagnostics: add VIF/tolerance to quantify collinearity and demonstrate that results are not methodological artefacts. We will rewrite Sections 2.3.3–2.3.4 as clear step-by-step procedures and include equations/flowcharts if needed (possibly in the Supplement).

Comment 3:

Response: We will examine the potential role of rising CO₂ by adding a time-varying CO₂ covariate and conducting sensitivity analyses, and we will explicitly discuss the interpretational limits (correlation vs causation).

We will explore available gridded forest-age products and, where feasible, perform stratified analyses; otherwise, we will clearly state this as a limitation and outline it as future work.

Comment 4:

Response：We will explicitly clarify in the revised manuscript that the Root-Zone Soil Moisture (SMrz) from GLEAM 4.2 is a vegetation-specific variable. According to the GLEAM formulation (Martens et al., 2017), SMrz represents the integrated soil moisture over the effective rooting depth, which is 0–100 cm for low vegetation and 0–250 cm for tall vegetation (forests).

Since our study focuses on forest ecosystems, the GLEAM SMrz data effectively characterizes the water status of the entire soil profile (0–250 cm) accessible to tree roots. This captures the hydraulic conditions more comprehensively than surface-only metrics. We have added this definition and the relevant citation to Section 2.1 to ensure methodological precision.

Comment 5:

Response: We will clarify that RH is already explicitly incorporated through the VPD calculation as an integrated metric of atmospheric demand. Including RH as an additional independent driver alongside VPD would introduce redundancy and stronger collinearity, undermining attribution. We will add this rationale in the revision and, if helpful, provide a sensitivity analysis in the Supplement by examining RH–θ relationships in a model setup that avoids simultaneously using RH and VPD.

Comment 6:

Response: We will harmonize the description: the AI framework yields four zones (humid, semi-humid, semi-arid, arid), but the arid zone is excluded from core analyses due to insufficient forest samples for robust statistics. We will state this consistently in the Methods, Results, and figure captions (definition vs included zones).

Comment 7:

Response: We will incorporate the NF and PF fractions for each climate zone into Fig. 2a

Comment 8:

Response: Thank you. We will correct the typo and carefully check notation throughout the manuscript to ensure consistency between all β symbols and their definitions (e.g., standardized regression coefficients).

Comment 9:

Response: We will comprehensively polish the manuscript for English clarity and consistency and will have the revised version proofread by a native English speaker before resubmission.