the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Oct 2025
Assessing Regional Climate Model Sensitivity to Vegetation Dynamics Informed by Remote Sensing
Abstract. Climate change significantly impacts vegetation ecosystems, and their modification may create feedback loops exacerbating regional effects of global warming. Accurately simulating vegetation dynamics and their interactions with the atmosphere is crucial for understanding and mitigating these impacts. Regional earth system models offer the possibility to study the retroaction between the atmosphere and the vegetation at regional to continental scale by incorporating vegetation dynamics in climate models. In this study, we quantify the sensitivity of the Modèle Atmosphérique Régional (MAR) to vegetation representation at daily to annual scales over a temperate region of Europe, by means of both synthetic experiments and realistic studies.
Our sensitivity study on the Leaf Area Index (LAI) dynamics reveals non-linear responses on meteorological variables, with asymmetric effects relative to the direction of the change. For example, a 92 % reduction in LAI led to an 83.4 % decrease in evapotranspiration and an 88.9 % drop in evaporation. Conversely, a 178.4 % increase in LAI resulted in smaller, yet significant, increases of 29.8 %, 27.4 % respectively. At the seasonal-time scale, evapotranspiration and albedo have the strongest shifts in summer and winter, while relative humidity and rainfall responded more prominently in spring.
Furthermore, we assessed the model performance in simulating daily evapotranspiration and daily maximum temperature during extreme events by comparing simulations incorporating 8-day MODIS LAI data with those based on climatological LAI. Although the improvements were more subtle than those resulting from a change in LAI source, the 8-day observation-based LAI enhanced the model’s capacity to capture shorter events compared to the static LAI. This refinement also helped understanding how various vegetation types respond to extreme events.
The findings highlight the need to integrate dynamic vegetation into regional climate models to enhance their representation of biosphere- atmosphere interactions and provide more accurate tools to assess the impacts of climate change on natural ecosystems.
- Preprint
(13627 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-3907', Pavol Nejedlik, 23 Dec 2025
-
AC1: 'Reply on RC1', Thomas Dethinne, 28 Jan 2026
The authors would like to thank Dr. Nejedlik Pavol for his time and effort in evaluating our manuscript and for providing constructive comments and suggestions. Please find our detailed responses to the reviewer’s comments in the attached file. For clarity, the reviewer’s statements are shown in black, and the authors’ responses are shown in blue.
-
AC1: 'Reply on RC1', Thomas Dethinne, 28 Jan 2026
-
RC2: 'Comment on egusphere-2025-3907', Anonymous Referee #2, 25 Dec 2025
This preprint presents a multi-experiment sensitivity study of the MAR regional climate model (SISVAT surface scheme) to vegetation forcing by Leaf Area Index (LAI), replacing the default MERRA2-based LAI climatology with MODIS LAI (climatology and 8-day) and adding synthetic LAI perturbations, with the main finding of strong, nonlinear/asymmetric impacts on ET/evaporation/soil moisture and weaker effects on temperature and especially precipitation; however, several core methodological and reporting issues (notably unit/definition inconsistencies, confounding changes between baseline experiments, incomplete specification of the LAI perturbation procedure, and significance testing that likely ignores autocorrelation/spatial dependence) currently prevent unambiguous attribution of the simulated differences to vegetation dynamics and weaken the robustness of the conclusions.
Major comments:
- The manuscript contains critical internal inconsistencies in units/definitions that must be resolved before results can be interpreted: Table 3 reports summer ET/rainfall values (e.g., 120–156 “mm” ET; 230–232 “mm” rainfall) while text refers to “mean daily”/“cumulative mean daily” in ways that imply contradictory units (mm/day vs mm/season); soil moisture is reported as ~65–68% in Table 3 but described as 0.72%→0.56% in text (likely fractions mis-labeled as percent); “evaporation” vs “evapotranspiration” is not defined (soil evaporation only vs soil+interception, etc.), which is essential for interpreting large asymmetries.
- The key comparison MARref vs MARMODISclim is confounded and does not isolate spatial/temporal resolution effects: (i) MERRA2 LAI climatology is for 1961–1990 while MODIS climatology is 2012–2022 (different climate/land-management era), and (ii) MAR applies vegetation-type correction coefficients to MERRA2 LAI but not to MODIS LAI (except partially via MARsector). The reported “improvements” and sensitivity could be driven by rescaling/calibration rather than information content/resolution. Add bridging experiments (e.g., MERRA2 without coefficients; MODIS with comparable scaling; bias/quantile matching at coarse scale) or quantify the portion of LAI differences attributable to scaling vs source.
- The LAI “Gaussian noise” design is insufficiently specified and is closer to a large systematic bias injection than random noise: setting the noise mean to ±µ or ±2µ of daily mean LAI creates strong deterministic shifts, and missing details about (a) clipping/capping of LAI (0 bound and realistic maxima), (b) frequency/location of negative values and how handled, and (c) spatial/temporal correlation structure (pixel-wise white noise vs coherent field vs correlated random field) make the resulting nonlinearity/asymmetry difficult to trust. Report the full perturbation formulation, bounds, clipping rates, and resulting LAI distributions by land-cover class and season.
- The strong asymmetry (large response to LAI decrease vs smaller response to LAI increase) may be partly an artifact of hard bounds/clipping and diminishing returns; it must be demonstrated as physical rather than numerical by showing (i) whether LAI hits 0 frequently in negative experiments, (ii) whether positive experiments saturate due to parameter caps (e.g., canopy resistance, albedo formulation), and (iii) sensitivity under multiplicative bounded perturbations (LAI×(1+ε)) rather than additive bias that can force zeros.
- Statistical significance claims are likely overstated: daily meteorological/flux time series are autocorrelated, precipitation is non-normal/zero-inflated, and spatial maps involve multiple comparisons; simply applying p<0.05 on daily samples (and across pixels) without effective sample size correction, block bootstrapping, or field-significance/FDR control is not defensible. Recompute significance with autocorrelation-aware methods and address spatial multiple testing.
- Interpretation of weak precipitation sensitivity must account for model configuration constraints: the domain is small and strongly forced by ERA5 (6-hour boundaries + nudging aloft), so synoptic control may dominate and suppress land-surface feedbacks on rainfall. If concluding limited rainfall sensitivity, provide process diagnostics (convective vs stratiform partition if available; low-level moisture convergence; PBLH/LCL/CAPE changes; moisture budget) to show whether ET changes could plausibly influence convection under this setup.
- The reported albedo response direction (“higher LAI increases albedo”) is not generally expected and depends on canopy optics, background soil, snow masking, and land-cover; without mechanistic evidence this will be questioned. Provide SISVAT albedo parameterization details (how LAI enters), and stratify albedo responses by land-cover and snow/no-snow conditions to justify the sign and seasonality.
- The statement that LAI underestimation contributes to a multi-year “spurious drying trend” is not supported without a closed water balance diagnosis. Provide P–ET–runoff–drainage–Δstorage terms (and layer-resolved soil moisture) to identify which term drives drift and how it changes across LAI experiments; otherwise the drift could arise from precipitation bias, runoff/drainage parameterization, soil texture/rooting depth, or stomatal stress representation.
- The added value of 8-day LAI during extremes is asserted as “subtle” and sometimes non-significant; strengthen this with event-based verification (phase/timing error, peak bias, tail metrics like TXx/TX95p, ET percentiles), and ensure that smoothing choices in the MODIS climatology (32-day moving average) are not inadvertently damping drought signals you aim to test.
- Reproducibility is insufficient for a sensitivity paper: key implementation choices (aggregation from 500 m to 5 km; sector masking rules; handling of missing MODIS pixels; temporal interpolation step usage in SISVAT; any LAI caps) must be described so the experiments can be repeated and the sensitivity attributed to defined perturbations rather than undocumented preprocessing.
Minor comments:
- Replace ambiguous phrasing (“cumulative mean daily”) with explicit definitions (e.g., “seasonal total (mm)”, “mean daily (mm/day)”, “domain-mean daily”). Ensure every figure/table uses consistent units and labels.
- Clarify the soil moisture variable (volumetric water content vs saturation fraction), which soil layers are included (top layer vs root zone vs integrated 0–7 m), and whether values represent sector-weighted or grid-cell means.
- Provide explicit sign conventions for surface fluxes (sensible heat in particular), since “positive toward the surface” differs from common conventions and affects interpretation.
- Specify whether “evaporation” includes canopy interception evaporation; if possible, report ET partition (transpiration, soil evaporation, interception) because LAI primarily changes transpiration and interception pathways.
- The SPEI description mixes “SPEI3” and “90-day”; use standard notation (e.g., SPEI-3) and explain the averaging/selection procedure for events more precisely (thresholds, windowing, persistence).
- The MODIS climatology period (2012–2022) overlaps drought years and differs from evaluation years; note the potential bias explicitly and, if feasible, add sensitivity to climatology period choice.
- Quantify the impact of MODIS quality filtering and gap handling on spatial sampling (cloud-driven biases), especially in winter/spring.
- Figure interpretations would benefit from stratification by land-cover class (forest/crops/grass) beyond national averages; many responses (albedo, ET) are class-dependent.
- Where you mention “model instability” (MAR−2µ), provide diagnostics (runoff, drainage, soil moisture bounds, energy closure) rather than a qualitative statement.
- Clean up typos/formatting (missing spaces, “n response”, duplicated punctuation) and ensure references are consistent (e.g., IPCC citation formatting).
Citation: https://doi.org/10.5194/egusphere-2025-3907-RC2 -
AC2: 'Reply on RC2', Thomas Dethinne, 28 Jan 2026
The authors would like to thank the anonymous reviewer for his/her time and effort in evaluating our manuscript and for providing constructive comments and suggestions. Please find our detailed responses to the reviewer’s comments in the attached file. For clarity, the reviewer’s statements are shown in black, and the authors’ responses are shown in blue.
Data sets
MAR Belgium sensitivity test to LAI input (2015-2021) Thomas Dethinne https://doi.org/10.5281/zenodo.15490382
MAR Belgium sensitivity test to LAI input (2022-2024) Thomas Dethinne https://doi.org/10.5281/zenodo.16761004
Model code and software
MAR code MAR team https://gitlab.uliege.be/tdethinne/mar-modis
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 307 | 169 | 28 | 504 | 20 | 17 |
- HTML: 307
- PDF: 169
- XML: 28
- Total: 504
- BibTeX: 20
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments
Although the article does not introduce new scientific methods, the methods used are fully sufficient for a detailed analysis of the specific phenomenon under investigation.
The language and the interpretation of the results are clear, and it shows very well the influence of LAI on atmospheric parameters produced by MAR. In addition to the objectives of the article, it also clearly shows the role of forests in the natural environment.
Nevertheless, here are some points for possible amendments and explanation.
-the title does not fully reflect the content of the manuscript as the vegetation dynamic is represented exclusively only by one parameter - LAI. Therefore, LAI should be reflected in the title. Further to that, the role of LAI in vegetation dynamics should be briefly discussed in text.
-the finding that increased LAI leads to increased evaporation is somehow surprising. Does it have any physical explanation, or could it rather be attributed to MAR procedure?
-it should be clarified what data entered the comparison in Sec. 3.1. The terms „observational-modelled data“ (row 255) and „observed data“ (row 259) should be defined/described.
-the sensitivity test in section 3.4 should be supported by some numeric or graphic interpretation.
-the statement „...that changes in LAI influence the distribution of rainfall events (Fig. 8(a)) sounds like LAI is influencing the raifall events didtribution.The above fact is rather attributed to the seasonal distribution of both LAI and rainfall events. Explain please, if not.
Typos
There are very few typos in the text. Nevertheless, one more check for typos, like in rows 31 and 465 is recommended.
-row 31 . n response, ... (In response,...)
-row 465 ...a led to... (skip a), ...LAI compared to s static one,... (change to s to a)
Formal notices
-the areas marked with a thin dashed line in Fig. 1 should be defined in the legend to Fig. 1.
-it should be stressed that 10 years of phenological data is quite a short time to compute useful climatology (row 193). The word „averages“ instead of „climatology“ should be rather used.
-it is difficult to distinguish some of the lines in Fig, 4a.
-terms MARm2m and MAR2m (row 382) should be clearly defined.
-rows 468 and 469. ... maximum daily air temperature (Fig. 11a) and daily average evapotranspiration (Fig. 11b). Figs 11a and 11b show the ET and Tmax in reverse order from that listed in the text.
-Fig. 11. The description in the legend and in the figure itself should be identical. MARMERRAclim is shown in the figure while MARref in the legend.
-row 353. There is no Appendix A1. There is Appendix A, Table1.