Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Overall assessment

The study is interesting and a good fit for Forests. The main issue is the experimental layout: the same treatment sequence was repeated in each row, so treatment is confounded with position and causal comparisons among management regimes are limited. The statistical analysis also does not always respect the plot as the experimental unit, while the forecast and convergence analyses go beyond what the available evidence can support. I recommend major revision.

Major comments

1. Treatment is confounded with plot position

Each treatment occupies the same position in every row (G–RG–R–C), and the manuscript also notes an elevation gradient running largely along this treatment axis. Treatment and position are therefore confounded. Comparing each plot with itself does not remove this problem because spatial differences can also influence how vegetation changes through time.

The later conclusion that grazing was the “decisive component” is therefore too strong. The six grazed plots occupy the first two positions in every row and the ungrazed plots the other two, so the grazed–ungrazed comparison does not separate treatment from position. The manuscript should describe the different trajectories observed under the four management regimes without attributing those differences to treatment alone.

2. Treatment-level inference is not kept consistently at the plot level

Section 2.4.2 correctly identifies the 20 × 20 m plot as the experimental unit because treatments were applied to plots rather than to the 1-m² subplots, but the multivariate analysis then uses all 1200 subplots for treatment-level inference. This is pseudoreplication: the 50 subplots within each plot are subsamples, not independent treatment replicates, so using them as the observational units inflates the effective sample size and makes the treatment p-values too small. Restricting permutations within plot rows does not correct that nesting; treatment-level inference has to come from the plots, and the positional confounding remains.

3. The 2028 forecast is not independently validated

The forecast is built from a single observed transition, 2016–2019, and then extended recursively through 2022, 2025 and 2028. The reported cross-validation tests how well the models reproduce observations from that same transition; it does not show that the ecological relationships remain valid over the following nine years. Species identity and baseline frequency are also strong predictors, so the high cross-validated R² may largely reflect persistence of species frequencies rather than long-term predictive ability.

The authors note that uncertainty accumulates across forecast steps, and agreement among the five learners only shows that the algorithms give similar results. I would treat 2028 as an exploratory continuation of the 2016–2019 relationships. PERMANOVA on model-generated future states cannot independently validate those trajectories.

4. The convergence metric needs stronger justification

The conclusion that the grazed plots converged on Potentillo albae-Quercetum about 1.8 times more strongly than shrub removal alone is based on Euclidean distance to a reference centroid in a two-dimensional NMDS with stress = 0.165. Because NMDS preserves rank relationships rather than exact Euclidean geometry, ratios such as “1.8 times” need to be shown to be robust to the reduced ordination space rather than taken directly from those distances.

The manuscript also does not show that the reference relevés and experimental data were sampled in a sufficiently comparable way for distances between them to carry this meaning. Contemporary and historical relevés are merged with the experimental data and used to define the reference centroids, but their sampling basis and any differences in sampling structure are not described.

Specific comments

1. The manuscript sometimes treats Ellenberg-type indicator values as direct measurements of light, soil nutrients or moisture. They indicate species affinities to those conditions, not measurements of them.
2. The “release rather than recruitment” conclusion is appropriate for the three characteristic species present in all plots at baseline, but less so for the wider flora because the 1-m² subplots were relocated between surveys.
3. The claims that thinning “pays for itself” and that thinning plus grazing is the most likely option for application at scale were not tested here; they come from the broader management discussion rather than this experiment.
4. Section 2.5 still contains journal-template instructions that should be removed; “correct the the language” also needs correction.

Author Response

Dear Reviewer,
thank you for your valuable remarks.
We provide the answers in the attachment.
With best regards
corresponding author

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript “Restoring the Herb Layer of a Thermophilous Oak Forest: Effects of Shrub Removal and Grazing (S Poland)” is a sound, expertly written botanical study that examines the effects of management practices on the herb layer of a thermophilous oak forest. The results are presented clearly and in a contemporary style. The discussion takes a highly critical view of the short duration of the study.

 

INTRODUCTION

Lines 40-44: Could you link the sentence containing the phrase ‘several purposes at once’ more directly to the sentence that lists them ("Cattle grazing, mowing and burning of the herb layer, litter raking...")?

Line 45: --> from the forest habitat

Lines 45-47: Sort out the confusion between singular and plural forms, and don’t introduce the word ‘pressure’ so abruptly – it can be quite a mental leap for readers! --> This pattern of land use and the pressure of use that persisted over the centuries – particularly in areas where livestock farming was practised – led to the emergence of distinctive biocultural agroforestry landscapes.

Line 48: --> The forest stands…

Line 60: Please add a space after ‘floor’

Line 67: Don’t just give a reference; also specify which ‘cascading change’ is meant.

Line 77 and many of the following: Latin species names should always be in italics!

Lines 88-89: and although a growing number of publications demonstrate the value of agroforestry for biodiversity and advocate the restoration of traditional cultural landscapes

 

MATERIALS AND METHODS

Lines 132-133: A thinning operation carried out in 2015 was the most recent forestry measure undertaken there.

Line 136-137: It seems a bit odd – surely that would still have been the case even after the third year of observation?

Lines 140-141: --> Historical sources suggest that the forest which once stood here was most likely used for grazing in the past,

Lines 143-144: Why ‘again dominated’ – are records of vegetation composition from earlier decades available? It doesn’t seem to me that this is the case, though...

Figure 1/ Line 153: Why did you always place the replicates for the treatment types in three adjacent plots (you call them ‘rows’)? Was there any reason not to use a randomised allocation? … Ah, tyou’ll be showing it from line 172. Perhaps you could give us a heads-up?

Line 156: A fence height of 2.2 m? Or a buffer zone of 2.2 m between the fence and the actual plots, which measure 20 m × 20 m?

Line 166: --> One baseline survey…

Lines 182-183: --> should cancel out and should not distort the detected changes.

 

RESULTS

Lines 354-355. --> After three years, both the number and the frequency of species were higher in every variant.

Lines 370-372: --> Tree and shrub species dominated the losses across the board, whilst the strongest increases were recorded among thermophilic plants and meadow herbs.

Figure 3/ Line 409: Prima!

Lines 448-454: You have noted something important here: for the time being, this is simply a matter of frequency, not of new occurrences in an established (60-year-old) forest.

Lines 465-466: --> Succession trends for the twelve plots, observed (2016–2019) and projected (2022–2028), using an ensemble forecast based on the historical proximity method. Please also remember to use full stop.

Figure 5/ Line 486: Why should it happen that way? First a sudden, significant increase, then a decline in the ‘half-shaded forests’ vegetation group, even though the observational data show only a decline. I would be very sceptical!

Figure 6/ Line 504: That’s a good illustration!

 

DISCUSSION

This is certainly a detailed, rather critical and well-conducted discussion. At one point, it even touches on socio-economics.

You use the term ‘succession’ in connection with ‘forecast’; the hedging term ‘vegetation dynamics’ – which applies in all cases – is mentioned nearby. Good. Otherwise, you also use the term ‘treatment-specific successional paths’. Given that the developments appear to be directional in your corresponding visualisation (Figure 4), this is certainly justified. I would be delighted if you could share a few brief thoughts on whether succession can occur across all your treatment variants. And – in light of your observations in the control group – how succession research, such as that in the Chicken Creek project (Germany; Zaplata et al. 2013, Am. Nat.) or the FORBIO Biodiversity Experiment (Belgium; Verheyen et al. 2013, Plant Ecol. Evol.), should contribute. Could a network of such platforms provide a baseline that takes account of changes in what is known as ‘background noise’?

Author Response

Dear Reviewer,
thank you for your valuable remarks.
We provide the answers in the attachment.
With best regards
corresponding author

Author Response File: Author Response.pdf