| 04 Dec 2025
Asymmetric Decline in Hydrological Efficiency of China's Natural and Planted Forests
Abstract. The vegetation transpiration fraction (TF) is a key parameter linking terrestrial water and carbon cycles. Against the backdrop of global greening and climate change, the response of TF sensitivity to Leaf Area Index (LAI) changes (θ), and the relative roles of soil moisture (SM) and atmospheric drought (VPD), remain unclear, especially lacking a systematic comparison between China's Natural Forests (NF) and Planted Forests (PF). This study utilized multi-source datasets from 1990–2020 (including forest types, GLEAM, and ERA5-Land) and employed methods including sliding windows, partial correlation, ridge regression, and mediation effect models to systematically analyze the spatiotemporal dynamics of θ in NF and PF, and to quantify the independent contributions and dynamic shifts of SM and VPD. Results show: 1) Forest θ spatially increases from humid to semi-arid regions (NF > PF); temporally, θ shows a widespread significant decline, with PF declining more (mean of −0.262 %∙m−2 · m² · decade⁻¹) than NF, especially in semi-arid/semi-humid transition zones. 2) θ exhibits a "hump-shaped" nonlinear response to the joint SM-VPD gradient, peaking at moderate SM and medium-high VPD. 3) The key hydrological drivers of θ are undergoing a dynamic shift from "atmospheric demand" (VPD) to "soil supply" (SM); the independent control of SM (βSM) has significantly strengthened over time, while that of VPD (βVPD) has weakened. 4) The two forest types show distinct response mechanisms: NF is more sensitive to VPD, while PF is more sensitive to SM stress. Overall, China's forests are shifting towards a more "conservative" water-use strategy, and the enhancing effect of LAI on TF has significantly weakened under strengthening SM constraints and VPD stress. The differentiated high sensitivity – NF to atmospheric drought and PF to soil drought – provides critical insights for forest water resource management and afforestation planning under future climate change scenarios.
The paper by Zhang et al. calculated the transpiration fraction (TF), estimated its sensitivity to leaf area index (denoted as θ), examined how the mean and trend of are influenced by vapor pressure deficit (VPD) and soil moisture (SM), and quantified the contribution of climate change to changes in θ through alterations in SM and VPD. These analyses are performed separately for natural and plantation forests in China, with direct comparisons made between the two ecosystem types.
Although this study seeks to address an important question—specifically, the relative roles of VPD and SM in mediating the coupling between plant physiological processes (transpiration) and structural attributes (leaf area), as well as potential differences in these roles between natural and plantation forests—I find that neither the methodology nor the presented results are sufficiently convincing. Several major concerns remain unaddressed. Unless these are satisfactorily resolved in a revision, I do not consider the manuscript suitable for publication in Hydrology and Earth System Sciences.
Major comments
Comment 1: The text in all figures is excessively small, making it difficult to read. This significantly impairs the clarity and accessibility of the data presentation. I strongly recommend enlarging the font sizes (including axis labels, legends, titles, and annotations) and improving the overall figure resolution to enhance readability!
Comment 2: As demonstrated by Wei et al. (2017), the relationship between the transpiration fraction (TF = T/(T + Es), neglecting interception loss) and LAI is inherently non-linear. Consequently, the θ is expected to diminish with ongoing vegetation greening. Although their study omitted Ei, it robustly illustrates the non-linear response of TF to increasing LAI. This non-linearity arises from biophysical constraints rather than direct influences of VPD or SM.
Specifically, as LAI increases, Es rapidly approaches its minimum due to enhanced canopy shading, resulting in a substantial decline in the absolute value of ∂Es/∂LAI. Interception evaporation becomes largely independent of further LAI increases, as it is primarily governed by rainfall patterns (frequency and intensity). T also becomes progressively constrained by available energy or water supply, leading to a marked reduction in ∂T/∂LAI at higher LAI values.
Under these conditions, ∂T/∂LAI, ∂Es/∂LAI, and ∂Ei/∂LAI all approach zero, causing θ to decrease toward zero. These mechanisms indicate that trends in θ may not be driven by shifts in plant water-use strategies, and the effects of VPD and SM on θ are indirect—operating primarily through modulation of absolute transpiration rates rather than directly altering the sensitivity of TF to LAI.
Comment 3: This study quantifies the influences of VPD and SM on θ from both spatial gradients and temporal dynamics perspectives. However, VPD and SM operate on distinctly different timescales: VPD fluctuates relatively rapidly compared to SM, and vegetation can respond to these changes over short timescales. For instance, stomata may partially close in response to elevated VPD, leading to quick adjustments in transpiration. In contrast, vegetation responses to changes in SM are typically slower and more gradual, involving physiological acclimation and structural adjustments over longer periods. Furthermore, during the warm growing season under dry soil conditions, VPD can be amplified through land–atmosphere feedback, creating a coupled effect that makes it difficult to disentangle the independent contributions of VPD and SM.
These differences in response timescales and potential feedbacks raise questions about the interpretation of the results presented in the study. I believe this issue warrants further careful consideration and discussion to strengthen the conclusions regarding the relative roles of VPD and SM in driving changes in θ.
Comment 4: [Line 79-82] The GLEAM4 and SiTHv2 transpiration are generated using semi-empirical modelling approaches that rescale potential evapotranspiration based on a stress factor, which incorporates vegetation optical depth (VOD) and simulated root-zone SM. The only mechanisms through which physiological responses could theoretically be represented are (i) the use of VOD as a proxy for canopy characteristics and (ii) the assimilation of surface soil moisture retrieved from satellite observations (applied only in GLEAM4).
However, the relatively low signal-to-noise ratio in these datasets may introduce substantial uncertainties into the estimated transpiration values. Consequently, it remains uncertain to what extent θ reflects physiological regulation by plants. Given this ambiguity, the interpretation and wording regarding the physiological implications of θ should be considerably more restrained and carefully qualified.
Comment 5: [Line 139-143] The description of GLEAM4 is inaccurate. Its data assimilation is limited to surface soil moisture. Furthermore, the hybrid machine-learning model is implemented solely within the vegetation transpiration module.
Comment 6: The methodology employed in this study may be influenced by the strong collinearity between VPD and SM, particularly regarding the application of partial correlation and mediation effect models. To ensure that the findings are robust and not merely artefacts of the chosen methods, it is recommended to provide collinearity diagnostics (e.g., VIF or tolerance levels) to quantify the degree of redundancy between these input variables.
Comment 7: The Discussion section is currently characterised by excessive reiteration of the results. I recommend streamlining this section to enhance clarity and conciseness. Additionally, the manuscript would be significantly strengthened by a dedicated discussion on methodological limitations, which is essential for a nuanced interpretation of the findings.
Specific comments
Line 15-16: There is a conceptual mismatch in grouping forest types with reanalysis products like GLEAM and ERA5-Land. Furthermore, the sliding window approach is a temporal analysis framework, whereas partial correlation, ridge regression, and mediation models are statistical inference methods.
Line 75-79: References of these sentences?
Line 123: This dataset, …., was produced by machine learning algorithm …
Line 188: Please clarify the identities of the left and right bars in panel (b). It is currently unclear what variables or groups these bars represent.
Line 214-216: The temporal scales of transpiration and evapotranspiration used for the calculations are not specified. Please clarify whether these analyses were conducted on a daily or monthly basis.
Line 273-279: The section 'Trend Analysis and Testing' should be integrated into Section 2.3.4, as the latter mentions trend analysis but lacks a detailed explanation.
Line 320: The 'hump-shaped' pattern described by the authors is not readily apparent in Figure 4a.
Line 458-459: The phrase 'non-linear trend of increasing' is ambiguous.