the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Feb 2026
Selection of soil biochemical indicators according to seasonal variation and vegetation cover for long-term soil monitoring in a mountain valley of the Alps
Abstract. The complexity of soil organic matter and the multifunctional role of its components on soil processes make the characterization of soil ecological status challenging. Due to its ready responsiveness to environmental changes, the soil microbial community has gained increasing attention for its relationship to the dynamics of C pools and soil chemical and physical processes. Its activity can be monitored by the enzymatic profile, which enables the detection of early changes in soil status, supported by direct or indirect measurement – e.g., by double-stranded DNA (dsDNA) – of microbial biomass and parameters, such as dissolved fractions of C and N, which are linked to soil activity as rapidly available energy sources. This study analyzed the seasonal response of these indicators in a subalpine ecosystem, using sampling date and vegetation cover as predictors capable of capturing long-term and short-term changes in the ecosystem, respectively. Most of the bioindicators showed higher values in the warmest and least rainy summer season. In the cold season, two distinct trends were evident: the values of dsDNA and enzyme activities decreased to their minimum in early winter and rose to their maximum in late winter, while those of soil organic matter (SOM), dissolved C, and N continued to decline until the end of winter. The study also found that the dynamics of SOM in the woodland and meadow ecosystems differed, with the former achieving the highest SOM content during the summer period of greatest plant and faunal activity. Overall, this study suggests that the use of bioindicators and high-throughput techniques can contribute to improving soil quality assessment and monitoring. Additionally, they can be used to characterize humus forms and motivate the preservation of Alpine meadows and surrounding wooded habitats for their non-wood products.
- Preprint
(644 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-250', Anonymous Referee #1, 06 Mar 2026
-
AC1: 'Reply on RC1', Gilberto Bragato, 02 Apr 2026
COMMENTS AND SUGGESTIONS IN COMMON TO BOTH REFEREES
In general, whilst adding a figure to show the area investigated and the spatial distribution of the sampling locations (Fig. 1) and expanding the final table, we have reduced the length of the manuscript by shortening the Introduction section and combining Tables 1 and 3.
SAMPLING DESIGN AND LOCATIONS
Number of replicates
Lines 85-95 have been amended to clarify the sample selection method used.
The total number of samples used in the data analysis, conventionally denoted by n, ranged from 75 to 79, depending on the variable considered.
In Tables 1 and 3, we added a further column reporting the number of samples by date and vegetation cover. A reference to the number of samples reported in Table 1 has been included in the captions of Figs. 2–4.
Study area and sampling locations
We added an additional figure (Fig. 1) and reported the surface area (in hectares) of the two vegetation covers in the revised manuscript.
We did not report the distance between samples because the selection was made according to the design-based approach, which assumes randomly distributed sampling locations inside the sampling areas.
Similarly, the number of samples per sampling location is not reported because we did not use compositing based on our experience in soil sampling and the statements on compositing made by de Gruijter et al. (2006; pages 52–55). On page 55, they point out that compositing introduces “error by imperfect mixing of the composites and error by subsampling the mixed composite”. The second source of error becomes significant in current biochemical soil analyses that use small aliquots (at most a few grams) taken from samples that, when composite, weigh several hundred grams.
We did not re-sample the sampling locations over time because we were interested exclusively in the sensitivity of response of biochemical variables. The results presented in the manuscript served as the starting point for testing some of the panel-based monitoring designs proposed by Brus and de Gruijter (2011) in the years that followed.
DATA ANALYSIS
PCA
PCA aims to reduce the dimensionality of the data matrix, but it also enables the exploration of relationships between variables correlated with the same principal component and helps to guide further analysis to be carried out using linear models. We specifically used it to provide a rough preview of the results and to justify the weighting of enzymatic activities relative to dsDNA.
As regards the standardisation of variables, most R packages that perform PCA start with the correlation matrix, which centres and standardises the variables by default. We used ‘FactoMineR’ package (Lê et al., 2008), which adopts a singular value decomposition algorithm to decompose the correlation matrix.
Linear models and ANOVA
ANOVA is a special case of linear modelling. We opted for the less specific but more flexible approach of linear models because, in a single analysis session, they allow to test the assumptions of homoscedasticity and normality, to suggest any necessary linear transformations, to select the model that best fits the data, and to compare treatment means using the t-test.
We tested the assumptions of normality and homoscedasticity by examining the residuals using the standard graphical methods of linear model analysis (functions ‘lm’ and ‘plot’ from the ‘stats’ package in R).
We therefore compared models with and without interaction terms using the ‘gls’ function from R’s ‘nlme’ package. For all variables, the optimal model was without an interaction term. We have specified this information in the revised text.
However, the reviewers’ comments helped us to reconsider Fig. 4, replacing EST with BG, whose trend was equivalent, but which is much more widely used in characterising the soil C cycle. This amendment is set out in lines 201–207 of the revised text
Regarding significant markers, we opted for p < 0.01 as a robust probability threshold for the significance of differences and preferred to highlight them in bold as this seemed more visible to the reader. A brief explanation has been introduced in the text.
In the end, we have updated the reference to the WRB classification to reflect the 2022 version.
References
Brus, D.J., de Gruijter, J.J: Design-based Generalized Least Squares estimation of status and trend of soil properties from monitoring data, Geoderma, 164, 172–180, https://doi.org/10.1016/j.geoderma.2011.06.001, 2011.
de Gruijter J.J., Brus D.J., Bierkens M.F.P., Knotter M.: Sampling for natural resource monitoring, Springer, Berlin, https://doi.org/10.1007/3-540-33161-1, 2006.
Added in the revised text
Del Duca S., Aponte C., Trasar-Cepeda C., Vitali F., Esposito A., Pastorelli R., Bragato G., Fornasier R., Sagova-Mareckova M., Suhadolc M., Mocali S.: Effects of sample storage conditions on agricultural soil bacterial diversity and functionality, Appl. Soil Ecology, 212, 106218, https://doi.org/10.1016/j.apsoil.2025.106218, 2025.
IUSS Working Group WRB: World Reference Base for Soil Resources: International soil classification system for naming soils and creating legends for soil maps, 4th edition, International Union of Soil Sciences (IUSS), Vienna, Austria, 2022.
Lê, S., Josse, J. & Husson, F.: FactoMineR. An R package for multivariate analysis, J. Stat. Softw., 25, 1-18, http://doi.org/10.18637/jss.v025.i01, 2008.
Wang F., Che R., Deng Y., Wu Y., Tang L., Xu Z., Wang W., Liu H., Cui X.: Air-drying and long-time preservation of soil do not significantly impact microbial community composition and structure, Soil Biol. Biochem., 157, 108238, https://doi.org/10.1016/j.soilbio.2021.108238. 2021.
Specific response to Reviewer # 1
Cite the latest research to support the use of air-dried soil for dsDNA/EEA analysis
Thank you very much for the suggestion. We have read and included the article you suggested (Wang et al., 2021), as well as a more recent one (Del Duca et al., 2025) of which we are co-authors (lines 100-102). Prior to 2020, our laboratory conducted several comparisons between moist and dry samples analysed within a few weeks of drying and several years later. In the case of EEAs, for instance, the analyses carried out on the samples subsequently used for the 2016 article (Bragato et al., 2016) showed no significant difference in values between moist and dry samples when the storage time was less than one month.
Abbreviations
As suggested, we added all abbreviations at first mention both in the text and in Figures.
Critical parameters of enzyme assays
The method we used for the determination of enzyme activities does not require pre-incubation, whilst the extraction was carried out at 25 °C with the buffer solution described in the text.
Quality control details have been added in the revised text.
Delete claims
The sentence with the claim "can be used to characterize humus forms" was deleted from the Abstract.
The claim "support long-term monitoring" was deleted and the suggested sentence "provide baseline data for future long-term monitoring" was incorporated in the last sentence of the Abstract.
Specify indicator selection criteria
Following the reviewer’s suggestions, we have specified 0.6 as the quantitative thresholds for selecting non-collinear indicators for principal component analysis (PCA) and the most representative models (R²adj ≥ 0.25), and we have briefly explained why we chose p<0.01 in the linear modelling (robustness of the model parameters).
Exclusion of BG and ALP
Thanks to your suggestions, we have reconsidered the use of BG and have used it in place of EST in Fig. 4, as BG is more frequently used as an indicator of the biogeochemical C cycle in soil. Furthermore, the fact that BG shows higher values in meadows is a finding we had already observed in other studies, but which we had not had the opportunity to publish, and which we felt it was important to mention in the revised text. The trend for ALP, on the other hand, is like that of the other enzymes, but the greater variability observed, particularly in December 2012, resulted in a model in which the sampling dates were of little explanatory value. All these points are highlighted in the revised text.
Brief note on study limitations
The note was added in the revised text.
Overgeneralization
We have revised the text in the hope of removing any over-generalisation.
Critical formatting errors
All reported errors have been corrected, except for the term “bead-beated”, as it is spelled correctly.
Citation: https://doi.org/10.5194/egusphere-2026-250-AC1
-
AC1: 'Reply on RC1', Gilberto Bragato, 02 Apr 2026
-
RC2: 'Comment on egusphere-2026-250', Anonymous Referee #2, 06 Mar 2026
The study of Gronchi et al. investigates seasonal variation in several indicators (enzyme activities, DNA, soil C and N) in a subalpine meadow and hazelnut grove. The focus is on the impact of sampling time and vegetation cover on the indicators. The selected indicators are considered to be potentially relevant for soil monitoring, particularly for changes in soil organic matter. The investigation of such indicators over time is valuable, but several major points need thorough improvement:
- While the study of the indicators is motivated by their potential value in soil monitoring, the soil monitoring context is not sufficiently discussed. In recent years, there have been several steps taken in Europe to advance soil monitoring (e.g., within LUCAS program), and different cost-effective indicators have been proposed for soil health assessment. The suggested indicators need to be discussed in the context of the existing efforts, what could they bring more? I am concerned that none of these monitoring developments are acknowledged and that generally no studies published after 2020 are referenced in the manuscript.
- Information on the sampling design remains unclear. To judge heterogeneity and temporal development, we need to know: sampling area (m2) per vegetation cover, number of sampling locations and distance between them per time point and vegetation cover, number of samples per sampling location. A map of the sampling locations might also be helpful. Also, were the sampling locations not the same for the different time points?
This info should be added in the method section (e.g., L.99), and the number of replicates should be clarified in the result section. For example, in Table 1, the n is given as 75 and 79, but were these actually the number of samples per sampling time x vegetation cover combination for each variable (seems like a high number)? Or is the overall number of samples summed up over sampling times and vegetation cover? For the figures, the info is missing. - For the data analysis with PCA and linear models, the reasoning and reporting need clarification. First, the value of the PCA remains unclear to me. In the discussion, the focus lies on correlations, but could they not be more easily determined without PCA? Moreover, the PCA scores from the sampling locations and the mentioned clustering of vegetation and sampling times are not shown and the data treatment (scaling) is not described. Then, the linear models are mainly used to infer differences between sampling times and vegetation cover; therefore, models using these two factors and then post hoc tests to compare the levels would be more meaningful? The output could be combined with Table 1 and 3, reducing data repetition. Finally, if the goal is to determine which indicators are influencing SOM, it might be useful to model SOM based on the indicators?
Selected minor issues:
L.9: First three sentences of the abstract are a bit unclear, e.g., what is meant by the soil ecological status and soil status?
L.15: Indicators, better use bioindicators as in the following for consistency
L.58: 0.29 or 0.33 to match the air-dry or field-moist
L.64: What are the criteria relevant for your study proposed by the Brus and de Gruijter (2011)? At the moment, the connection remains unclear.
L.110: What is the maximum speed and the g force applied?
L.119: “Summary statistics” instead of “statistics”?
L.129, L.143, L. 166: These descriptive sentences are distracting from the shared information. Please describe the main patterns and mention where the data is presented (e.g., Table x) or if differences were significant (p-values).
Figure 2 and Figure 2: Units are not correctly displayed.
L. 211: The concluding sentence needs an explanation of the mechanisms suspected behind these trends.
Citation: https://doi.org/10.5194/egusphere-2026-250-RC2 -
AC2: 'Reply on RC2', Gilberto Bragato, 02 Apr 2026
COMMENTS AND SUGGESTIONS IN COMMON TO BOTH REFEREES
In general, whilst adding a figure to show the area investigated and the spatial distribution of the sampling locations (Fig. 1) and expanding the final table, we have reduced the length of the manuscript by shortening the Introduction section and combining Tables 1 and 3.
SAMPLING DESIGN AND LOCATIONS
Number of replicates
Lines 85-95 have been amended to clarify the sample selection method used.
The total number of samples used in the data analysis, conventionally denoted by n, ranged from 75 to 79, depending on the variable considered.
In Tables 1 and 3, we added a further column reporting the number of samples by date and vegetation cover. A reference to the number of samples reported in Table 1 has been included in the captions of Figs. 2–4.
Study area and sampling locations
We added an additional figure (Fig. 1) and reported the surface area (in hectares) of the two vegetation covers in the revised manuscript.
We did not report the distance between samples because the selection was made according to the design-based approach, which assumes randomly distributed sampling locations inside the sampling areas.
Similarly, the number of samples per sampling location is not reported because we did not use compositing based on our experience in soil sampling and the statements on compositing made by de Gruijter et al. (2006; pages 52–55). On page 55, they point out that compositing introduces “error by imperfect mixing of the composites and error by subsampling the mixed composite”. The second source of error becomes significant in current biochemical soil analyses that use small aliquots (at most a few grams) taken from samples that, when composite, weigh several hundred grams.
We did not re-sample the sampling locations over time because we were interested exclusively in the sensitivity of response of biochemical variables. The results presented in the manuscript served as the starting point for testing some of the panel-based monitoring designs proposed by Brus and de Gruijter (2011) in the years that followed.
DATA ANALYSIS
PCA
PCA aims to reduce the dimensionality of the data matrix, but it also enables the exploration of relationships between variables correlated with the same principal component and helps to guide further analysis to be carried out using linear models. We specifically used it to provide a rough preview of the results and to justify the weighting of enzymatic activities relative to dsDNA.
As regards the standardisation of variables, most R packages that perform PCA start with the correlation matrix, which centres and standardises the variables by default. We used ‘FactoMineR’ package (Lê et al., 2008), which adopts a singular value decomposition algorithm to decompose the correlation matrix.
Linear models and ANOVA
ANOVA is a special case of linear modelling. We opted for the less specific but more flexible approach of linear models because, in a single analysis session, they allow to test the assumptions of homoscedasticity and normality, to suggest any necessary linear transformations, to select the model that best fits the data, and to compare treatment means using the t-test.
We tested the assumptions of normality and homoscedasticity by examining the residuals using the standard graphical methods of linear model analysis (functions ‘lm’ and ‘plot’ from the ‘stats’ package in R).
We therefore compared models with and without interaction terms using the ‘gls’ function from R’s ‘nlme’ package. For all variables, the optimal model was without an interaction term. We have specified this information in the revised text.
However, the reviewers’ comments helped us to reconsider Fig. 4, replacing EST with BG, whose trend was equivalent, but which is much more widely used in characterising the soil C cycle. This amendment is set out in lines 201–207 of the revised text
Regarding significant markers, we opted for p < 0.01 as a robust probability threshold for the significance of differences and preferred to highlight them in bold as this seemed more visible to the reader. A brief explanation has been introduced in the text.
In the end, we have updated the reference to the WRB classification to reflect the 2022 version.
References
Brus, D.J., de Gruijter, J.J: Design-based Generalized Least Squares estimation of status and trend of soil properties from monitoring data, Geoderma, 164, 172–180, https://doi.org/10.1016/j.geoderma.2011.06.001, 2011.
de Gruijter J.J., Brus D.J., Bierkens M.F.P., Knotter M.: Sampling for natural resource monitoring, Springer, Berlin, https://doi.org/10.1007/3-540-33161-1, 2006.
Added in the revised text
Del Duca S., Aponte C., Trasar-Cepeda C., Vitali F., Esposito A., Pastorelli R., Bragato G., Fornasier R., Sagova-Mareckova M., Suhadolc M., Mocali S.: Effects of sample storage conditions on agricultural soil bacterial diversity and functionality, Appl. Soil Ecology, 212, 106218, https://doi.org/10.1016/j.apsoil.2025.106218, 2025.
IUSS Working Group WRB: World Reference Base for Soil Resources: International soil classification system for naming soils and creating legends for soil maps, 4th edition, International Union of Soil Sciences (IUSS), Vienna, Austria, 2022.
Lê, S., Josse, J., Husson, F.: FactoMineR. An R package for multivariate analysis, J. Stat. Softw., 25, 1-18, http://doi.org/10.18637/jss.v025.i01, 2008.
Wang F., Che R., Deng Y., Wu Y., Tang L., Xu Z., Wang W., Liu H., Cui X.: Air-drying and long-time preservation of soil do not significantly impact microbial community composition and structure, Soil Biol. Biochem., 157, 108238, https://doi.org/10.1016/j.soilbio.2021.108238. 2021.
Specific response to Reviewer # 2
Soil monitoring context
In our study, we focused exclusively on the selection of indicators and did not address the issue of spatio-temporal monitoring designs, as this is a complex matter heavily influenced by the size of the areas to be monitored.
Few articles on spatio-temporal soil monitoring have appeared since 2020. The most comprehensive is that by Froger et al. (2025), which compares LUCAS data with that of several European monitoring networks at national scale. Without going into the merits of the LUCAS programme (which, to my knowledge, has never been discussed in depth within the soil science community), the aim is to assess trends at national or European scale, with observation densities exceeding hundreds of square kilometres. At these scales, biochemical indicators other than soil organic matter/carbon are scarcely responsive, whereas they can become so at the catchment/districts level, which was the focus of our study.
Combination of Tabs. 1 and 3
Tabs. 1 and 3 have been combined in the revised text. In this way, we have reduced both the length of the manuscript and the number of repetitions.
Selected minor issues
L.9: The term ‘soil ecological status’ is incorrect. In the revision, we have replaced it with ‘soil status’ (L. 25).
L.15: We have replaced “indicators” with “bioindicators” as suggested.
L.58: “0.29 or 0.33 to match the air-dry or field moist”. We changed the sentence as follows “0.29–0.33 hours per sample, whether the initial sample was air-dried or field-moist, respectively”.
L.64: As mentioned above, the study was preparatory. Once selected, we used these variables to test some of the space-time panel designs proposed by Brus and de Gruijter (2011).
L.110: Centrifugation was carried out at 20.000 g for 5 min for both dsDNA and EEAs. We have included the information in the revised text.
L.119: “Summary statistics” is the right term, we introduced it in place of “statistics” and “descriptive statistics”).
L.129, L.143, L. 166: the text was modified in accordance with suggestions of the Referee.
Figs. 2 and 3: Units were modified as suggested.
L. 211: We have removed the final sentence from the Conclusions. The discussion regarding the use of dsDNA and EEAs in the characterisation of humus forms would require specific further investigations, which are effectively beyond the scope of this study.
References
Froger, C., Tondini, E., Arrouays, D., Oorts, K. et al. Comparing LUCAS Soil and national systems: Towards a harmonized European Soil monitoring network. Geoderma, 449, 117027, 2025.
Citation: https://doi.org/10.5194/egusphere-2026-250-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 192 | 52 | 29 | 273 | 54 | 70 |
- HTML: 192
- PDF: 52
- XML: 29
- Total: 273
- BibTeX: 54
- EndNote: 70
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear Editors and Authors,
This manuscript investigates the seasonal dynamics of soil biochemical and microbial indicators under two vegetation covers (meadow vs. hazel groves) in an Alpine subalpine ecosystem, with the aim of screening cost-effective, sensitive indicators for long-term soil monitoring. The study provides valuable baseline seasonal dynamic data for the target region. However, these issues must be fully addressed through major revisions before acceptance.
e.g. Line 15-16: sampling date = short-term seasonal variation, vegetation cover = long-term anthropogenic management? If true, it is better to rephrase it: using sampling date (capturing short-term seasonal climatic variation) and vegetation cover (capturing long-term anthropogenic management effects) as key predictors to analyze the seasonal response of these indicators in a subalpine ecosystem.
resolve Figure-text inconsistencies (Figures 2 and 3 captions vs. text descriptions), Figure 2: Text describes dsDNA and ACP seasonal trends, but the caption states dsDNA and ARYS Figure 3: Text describes ACP and EST (standardized to dsDNA), but the caption states ARYS and EST
Renumber the duplicate "2.2 Data analysis" to "2.3 Data analysis".
Line 104: The solution ratio “1, 4 w/v” is a punctuation error (revise: 1:4 w/v, the standard expression of solid-liquid ratio).
Line 109: should be bead-beaten?
Line 135: The “former” is ambiguous, directly use TDN instead Line 140：“March 2012” is inconsistent with the sampling date (March 2013) in the methods section, it should be March 2013.
Horvath (2007) in the text is misspelled—the reference list is Horwath (2007)