Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Open AccessArticle

Peer-Review Record

Soil Quality Assessment in Reclaimed Coastal Paddy Fields: A Case Study from Eastern China

Soil Syst. 2026, 10(9), 99; https://doi.org/10.3390/soilsystems10090099

by Caixia Liu^1,2, Chenfei Liang², Hui Zhang¹, Jianyu Yu³, Linhui Liao¹, Jingjing Chen¹, Yulong Wang¹, Qingying Gao^1,*

and Liang Wang^1,*

Reviewer 1:

Evgeny V. Abakumov

Reviewer 2:

Claudete Gindri

Reviewer 3:

Jifu Li

Reviewer 4: Anonymous

Soil Syst. 2026, 10(9), 99; https://doi.org/10.3390/soilsystems10090099

Submission received: 11 July 2026 / Revised: 8 August 2026 / Accepted: 19 August 2026 / Published: 24 August 2026

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The attitude of the authors of the article to the concept of "soil quality" is not entirely clear - it is necessary to make it more accurate. Soil quality should be attributed to another agent of the ecosystem - a plant, a person, a community of organisms. Soil quality cannot be on its own. There is also the concept of forest-growing properties and soil parameters, the class of soil bonity, and so on. How does the concept introduced by the authors relate to previously known terms and approaches?

Section "Introduction" - you need to clearly formulate the working hypothesis and research methodology, limitations of the research and describe what new will give your research proper for soil science as a science

A little more needs to be said about the classification of coastal areas in the world in general and in China in particular. It is also necessary to more clearly justify the need to study the coastal territories in terms of soil quality and soil ecology. It is not at all clear what the origin of these coastal territories is, what their landscape and geomorphological history is

Line 110 – “The soil type is paddy soil formed from coastal saline soil following a long-term amelioration and cultivation spanning 5 to 8 years.” – provide WRB taxonomy names of the soil, and give the soil profile pictures as well. What where the soil horizons types? This is a pedological journal – we have to operate by soil name with precise qualifiers and soil horizon names, but not only soil layers deeps.

Line 108 – you provided only precipitation rate, but the evaporation rate is also important to understanding and interpretation of soil hydrological regime.

What about annual mean temperature and depth of ground waters?

What about soil parent material type, origin and texture?

What was a type of local topography?

Line 148 – you have to substantiate why this method was used – “Available phosphorus (AP) was determined by the Olsen method” – normally the method choice depends from the initial soil pH.

What was the content of fine earth and coarse fraction in soil?

The main question that arises when reading the text of the article - the soil-genetic (from the point of view of soil genesis) interpretation of the data obtained - gets that a certain soil has been studied on an unknown soil-forming rock, located in an indefinite relief, devoid of vegetation and in incompletely defined climatic conditions. It measured and determined some chemical and microbiological parameters. Their values are processed using a fairly popular, but controversial formula for calculating the soil quality index, which is not compared and relatively not analyzed in comparison with those that already exist and on the basis of this, general conclusions are drawn about the need for improved planning of economic activities. Unfortunately, in this form, this article, in my opinion, is not suitable for a classic journal in the field of Soil Science.

How can it happen that the article does not discuss any soil process? Not even talking about the process of humus accumulation, leaching or gley process?

I belong to the Dokuchaev school of genetic fundamental soil science and I do not understand how you can study soils without soil horizons and without soil processes. Soil is a process. Process at vertical soil profile level, process at genetic soil horizon level, process at aggregate level. And in this article, soil is a set of measured chemical and microbiological parameters. In such a design, the article can be considered in any other journal, but not a specialized journal on Soil Science.

Author Response

Comment 1: The attitude of the authors of the article to the concept of "soil quality" is not entirely clear - it is necessary to make it more accurate.

Response 1: We agree with the reviewer that” the concept of ‘soil quality’ is necessary to be more accurate”, We rewritten it more specifically about the concept of “soil quality” in Introduction (Lines 69-71).

Comment 2: Section "Introduction" - you need to clearly formulate the working hypothesis and research methodology, limitations of the research and describe what new will give your research proper for soil science as a science.

Response 2: We sincerely thank the reviewer for this constructive suggestion. We have thoroughly revised the Introduction to explicitly include: (1) a clear statement of our working hypotheses (Lines 124-129), (2) a brief outline of the research methodology (Lines 117-122), (3) an acknowledgment of the study’s limitations, and (4) a specific description of the novelty and significance of our work for soil science. These additions are now presented in the final two paragraphs of the Introduction (Lines 129-133). We believe the revised Introduction now better frames the study and meets the standards required

Comment 3: A little more needs to be said about the classification of coastal areas in the world in general and in China in particular. It is also necessary to more clearly justify the need to study the coastal territories in terms of soil quality and soil ecology. It is not at all clear what the origin of these coastal territories is, what their landscape and geomorphological history is.

Response 3: We sincerely thank the reviewer for these insightful suggestions. We have now added some descriptions in the Introduction to address the three aspects:

(1) Classification: We briefly introduce the global classification of coastal reclaimed soils and the corresponding categories in the Chinese Soil Taxonomy (Lines 43, 48-57).
(2) Origin and landscape history: We clarify that our study area originates from marine sedimentation, with a long history of tidal flat enclosure and polder construction, and describe the current anthropogenic landscape features specifically in Materials and Methods (Lines 141-155).

(3) Justification for soil quality and ecology research: We explain why evaluating soil quality and microbial ecology in these areas is necessary, emphasizing the challenges of salinity, low fertility, and the policy-driven need for biological soil health indicators (Lines 113-116).

Comment 4: Line 110 – “The soil type is paddy soil formed from coastal saline soil following a long-term amelioration and cultivation spanning 5 to 8 years.” – provide WRB taxonomy names of the soil, and give the soil profile pictures as well. What where the soil horizons types? This is a pedological journal – we have to operate by soil name with precise qualifiers and soil horizon names, but not only soil layers deeps.

Response 4: Following your advice, we have made the following revisions:

WRB Classification: We have reclassified the studied soils using the WRB 2022 system. The original coastal sediments correspond to Fluvic Gleyic Solonchaks, and the long-term cultivated paddy soils are now correctly identified as Eutric Hydragric Anthrosols (Protosalic, Fluvic, Gleyic), reflecting their saline parent material, fluvic origin, and current hydragric properties (Lines 148-152).
Soil Profile Description: We have excavated a representative profile in supplementary information (Fig S1) and provided detailed horizon descriptions as “Before reclamation, the soil profile exhibited an A–C horizon sequence with a distinct absence of the B horizon, indicating an early stage of pedogenesis dominated by sedimentary accumulation with limited horizon differentiation and low overall development. Following reclamation, the profile transformed into an A–Ap–B–C sequence with the neo-formation of both Ap and B horizons, reflecting substantially accelerated pedogenesis under anthropogenic intervention and a markedly enhanced degree of soil development.” (Lines 161-156).
Profile Photographs: Representative soil profile pictures with clearly labeled horizons have been added as Fig.S1 in supplementary information.

We believe these additions substantially strengthen the pedological basis of our manuscript and meet the journal's standards. Thank you for helping us improve our work.

Comment 5: Line 108 – you provided only precipitation rate, but the evaporation rate is also important to understanding and interpretation of soil hydrological regime. What about annual mean temperature and depth of ground waters? What about soil parent material type, origin and texture?

What was a type of local topography?

Response 5: We have added detailed descriptions about evaporation rate, annual mean temperature, depth of ground water, soil parent material type, origin, texture and the type of local topography (Line 140-145).

Comment 6: Line 148 – you have to substantiate why this method was used – “Available phosphorus (AP) was determined by the Olsen method” – normally the method choice depends from the initial soil pH. What was the content of fine earth and coarse fraction in soil?

Response 6: We would like to clarify that the pH of all samples was measured prior to analysis, and all values were above 7.5. This confirms that the Olsen method is appropriate, as it is specifically recommended for calcareous soils with pH > 7.5. Accordingly, we used this method for available phosphorus (AP) determination, and have added a sentence explaining this choice in Line 208.

Comment 7: The main question that arises when reading the text of the article - the soil-genetic (from the point of view of soil genesis) interpretation of the data obtained - gets that a certain soil has been studied on an unknown soil-forming rock, located in an indefinite relief, devoid of vegetation and in incompletely defined climatic conditions. It measured and determined some chemical and microbiological parameters. Their values are processed using a fairly popular, but controversial formula for calculating the soil quality index, which is not compared and relatively not analyzed in comparison with those that already exist and on the basis of this, general conclusions are drawn about the need for improved planning of economic activities. Unfortunately, in this form, this article, in my opinion, is not suitable for a classic journal in the field of Soil Science.

How can it happen that the article does not discuss any soil process? Not even talking about the process of humus accumulation, leaching or gley process?

Response 7: We sincerely thank the reviewer for this valuable and constructive advice. We acknowledge that the original manuscript lacked a systematic description of pedogenesis in the study area. In response to this comment, we have rewritten and substantially expanded the relevant sections, providing more detailed information on soil landform, geology, and other environmental settings in supplementary information （Table S3 and Table S4） and Materials and Methods (Lines 141-158).

We fully agree that soil genesis, evolution, and continuity are long-term processes and represent a fundamental area of soil science research. We greatly respect the genetic soil science tradition. However, the scope of our study is more focused: we aim to address a specific issue within this broader context—namely, soil quality assessment in reclaimed areas. We hope that our findings can provide reliable and objective evidence for reclamation evaluation, and may serve as a reference for similar studies in other reclaimed regions.

We acknowledge that SQI calculation and application have certain limitations, and we appreciate the reviewer's critical view on this point. Nevertheless, in our study, the SQI values consistently reflected the expected trend that higher soil quality corresponds to higher SQIs, which supports the validity of our assessment. We have also added a brief discussion on the limitations of the SQI methodology and the need for future comparative studies in the revised manuscript (Lines 534-546, 583-613).

Regarding the journal aims&scopes, we would like to respectfully note that Soil Systems explicitly lists "Impact of soil management on soil processes," "Soil degradation," and "Soil development" among its subject areas. Our study on soil quality assessment in reclaimed areas directly addresses that reclamation drives soil development (from tidal sediments to cultivated soils) and may lead to degradation processes such as salinization. By integrating chemical and microbiological indicators through the SQI approach, our study provides insight into the biological and biogeochemical processes operating in these rapidly evolving soils. Therefore, we believe that our manuscript aligns well with the journal's scope and will be of interest to its readership.

We are open to any further questions or suggestions and will provide honest and thoughtful responses.

Please see the attachment for more details.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,
I appreciate the opportunity to review your manuscript entitled "Soil quality assessment in reclaimed coastal paddy fields based on minimum data set and microbial indicators." Below, I present my constructive evaluation, highlighting areas for improvement and inconsistencies.

Please specify the knowledge gap in the Introduction and Abstract.

Please include a statement of novelty, distinct from practical relevance, in both the Abstract and at the end of the Introduction.

Section 2.1 mentions that the reclaimed areas were developed at different times, but the exact age of each reclaimed area and the time elapsed since conversion to agricultural use are not provided. This is crucial because the age of reclaiming can be a major determinant of salinity, organic matter accumulation, nutrient availability, and microbial development.

Please provide the following information for each site:

Year of reclaiming;
Year of paddy establishment;
Number of years under cultivation;
Previous land use history;

Irrigation water source;
Drainage regime;
Frequency and duration of flooding;
Characteristics of the barrier or coastal infrastructure.

The comparison between YQ, LW, LG, and RA may be confounded by differences in reclamation age, agricultural management, soil texture, initial salinity, and topographic position. Without this information, it is difficult to attribute the observed differences solely to spatial variation among locations.

Line 130:
The number of samples per site varies considerably: YQ (9), LW (18), LG (9), and RA (13). Although it is mentioned that it was based on the consistency of the reclamation history, this imbalance may affect the homogeneity of variances in the ANOVA. Was the homoscedasticity of the data checked before performing the ANOVA? If not, a test such as Welch's test should have been applied.

Lines 93–100 and 403–418
The manuscript concludes that there is spatial heterogeneity in soil quality and links the differences to salinity, metals, and microbial activity. However, the design used appears to be observational and cross-sectional, with a single sampling date, November 2024. Therefore, it is not possible to directly conclude that the coastal reclamation caused the observed changes unless the following are available:

unclaimed reference soils;
a chronosequence with different reclamation ages;
historical data prior to the reclamation; or
a temporal follow-up design.

Please moderate causal expressions such as “driving factors,” “caused by,” “led to,” or equivalents. Instead, you can use expressions such as:

“were associated with”;

“could contribute to”;

“were consistent with”; or

“were identified as variables related to.”

This is especially important in lines 350–354, where the elevated concentrations of Fe, Mn, Cu, and Zn in YQ are attributed to the electrical equipment industry and electroplating operations. This explanation is plausible, but it was not directly demonstrated in the study. No data are presented on:

metal concentrations in sediments;
metals in irrigation water;
wind direction;
distance to industrial facilities;
isotope or source analysis;
contamination history;
geochemical fractions of the metals.

Please present this explanation as an interpretive hypothesis and not as a confirmed conclusion.

“Wang nitrogen et al. 2023” on line 358 appears to be an editing error or a misplaced reference.

Line 242: Correct the spelling inconsistency in the citation of Biau and Scornet.

In equation 7 (lines 234-236), the Sensitivity Index is defined as:

SI = SQImin / SQImax
Since the SQI values are positive and, by definition, SQImax ≥ SQImin, the SI value should theoretically be greater than or equal to 1.0.

However, on line 327, the authors state: "The SI value of the SQIw (0.638) was higher than that of the SQIa (0.625)."
An SI value of 0.638 is mathematically anomalous with the described formula if the data are correct. This suggests an error: either the formula in equation 7 is incorrectly written, or the calculations in the figures and text are wrong. Please correct and clarify this.

The final MDS proposed by the authors contains 14 indicators (pH, MOS, NT, available K, total soluble salts, available N, available Zn, bacterial quantity, bacterial diversity, fungal diversity, BG, CB, NAG, and XYL) out of a total of 27 originally measured (lines 297-300). This represents 52% of the total dataset.

The fundamental purpose of constructing an MDS is to drastically reduce the cost, effort, and time required for soil monitoring by selecting a small and practical number of key indicators (typically between 4 and 7). An MDS with 14 variables, which also includes complex and expensive molecular analyses (Illumina sequencing for fungal and bacterial diversity) and multiple enzyme assays, contradicts the purpose of being a "minimal" set. Please:
Justify the need to retain so many variables.

Review the selection criteria in the PCA. If a Principal Component (PC) contains highly correlated variables (r > 0.60), only the one with the highest factor loading, or sum of correlations, should be retained, instead of maintaining multiple redundant variables.

The authors validate the sensitivity of the SQI using only the mathematical Sensitivity Index (SI) (line 326). Since the study is conducted in paddy fields, which are active agricultural production systems, standard and scientifically robust validation of an SQI requires correlating the obtained indices (SQIa and SQIw) with an output variable from the real system. How do these SQIs relate to rice grain yield, plant biomass, or basal soil respiration? Without validation against real-world productivity, the proposed SQI is merely a mathematical exercise and lacks proven practical utility for farmers or land managers.

Soil samples were collected at a single point in time (November 2024, post-harvest) (line 130).
Microbial communities, fungal diversity, and enzymatic activities are highly dynamic parameters that fluctuate drastically depending on the season, temperature, and, especially, water management in rice paddies (which alternate between flooding/anaerobic and drying phases). Basing a long-term Soil Quality Index (SQI) on biological indicators measured at a single point in time seriously limits its temporal representativeness. Please discuss this limitation and justify why data from a single post-harvest sampling are sufficient to establish a soil quality baseline.

The text refers to "bacterial quantity" and "fungal abundance" as key variables of the Soil Quality Index (MDS) (lines 298 and 323). However, the methodology section only describes in detail the sequencing of amplicons (16S and ITS) to estimate diversity using the Shannon index [176-189]. Illumina MiSeq sequencing yields relative abundance data, not quantitative data. If absolute abundance or "bacterial quantity" was determined by qPCR (which is vaguely implied by the mention of the CFX96 Real-Time System on line 174), please clearly describe the qPCR protocol: the target genes, the primers used, the thermocycling conditions, and the construction of the standard curve.

The authors note that the Yueqing (YQ) area exhibits high concentrations of Fe, Mn, Cu, and Zn due to its history as a center for electrical equipment production and electroplating (Line 352). If these reclaimed areas are currently being used for rice production for human consumption (as mentioned in lines 39–40), the accumulation of heavy metals (especially copper and zinc) poses a serious risk of food safety and phytotoxicity. Further discussion is needed: Do the concentrations of these metals exceed the limits permitted for agricultural soils in China? Could the accumulation of metals in the YQ soil be negatively impacting enzyme activity and be the real cause of its lower Soil Quality Index (SQI), beyond salinity?

Author Response

Comment 1: Please specify the knowledge gap in the Introduction and Abstract. Please include a statement of novelty, distinct from practical relevance, in both the Abstract and at the end of the Introduction.

Response 1: We have specified and included a statement of novelty, distinct from practical relevance the knowledge gap in the Introduction (Lines 127-133) and Abstract (Lines 35-37).

Comment 2:

Please provide the following information for each site:

Year of reclaiming;
Year of paddy establishment;
Number of years under cultivation;
Previous land use history;

Irrigation water source;
Drainage regime;
Frequency and duration of flooding;
Characteristics of the barrier or coastal infrastructure.

Response 2: We have provided detailed information including year of reclaiming, year of paddy establishment, number of years under cultivation, irrigation water source drainage regime and frequency previous land use history, and duration of flooding of each site in supplemental information (Table S3, S4). In this study, we more focus on the total differences of soil quality in study areas. The information reveals notable differences in reclamation age among the sites, ranging from <1 year to approximately 10 years.

In this study, we more focus on the total differences of soil quality in study areas. However, we fully acknowledge that differences in reclamation age, agricultural management, soil texture, initial salinity, and topographic position may confound the comparison among sites and that our cross‑sectional design cannot fully disentangle the individual contributions of these factors.

We have therefore revised our interpretation throughout the manuscript to emphasize that the observed patterns reflect a composite outcome of multiple factors including reclamation history, management practices, and inherent site characteristics rather than attributing the differences solely to spatial variation. We have also explicitly acknowledged the confounding effects of these variables as a limitation in the Discussion section (Lines 583-593). Future studies employing a chronosequence approach or long‑term monitoring would be valuable to more rigorously evaluate the effect of reclamation age on soil quality development in coastal reclaimed areas.

Comment 3: Line 130:
The number of samples per site varies considerably: YQ (9), LW (18), LG (9), and RA (13). Although it is mentioned that it was based on the consistency of the reclamation history, this imbalance may affect the homogeneity of variances in the ANOVA. Was the homoscedasticity of the data checked before performing the ANOVA? If not, a test such as Welch's test should have been applied.

Response 3: We thank the reviewer for raising this important statistical concern. We fully agree that sample size imbalance among the four sites (YQ: 9, LW: 18, LG: 9, RA: 13) may affect the homogeneity of variances in ANOVA.

We applied Welch's ANOVA for indicators with significant Levene test results (p < 0.05). This method does not assume equal variances across groups and corrects the degrees of freedom using the Welch–Satterthwaite approximation, providing more reliable inference when sample sizes are unequal and variances are heterogeneous. In the revised manuscript, we have explicitly stated this procedure in the Materials and Methods section (Lines 306-310) and the Supplementary Information (Table S7 and Table S8). We have confirmed that the observed differences remain robust after this correction.

For the 12 indicators ultimately retained in the MDS (pH, SOM, TN, AK, TWS, AN, available Zn, bacterial quantity, bacterial diversity, fungal diversity, BG, and CB), the results show that most of the retained MDS indicators exhibit significant differences among the four sites (P < 0.05), with the exception of pH, TWS and BG. These findings support the robustness of our MDS selection and subsequent SQI comparisons. We have added a description of the Levene's test and Welch ANOVA procedures, along with the corresponding results, to the Materials and Methods (Lines 315-319).

Table S7. Homogeneity of variances.

Comment 4: Lines 93–100 and 403–418
The manuscript concludes that there is spatial heterogeneity in soil quality and links the differences to salinity, metals, and microbial activity. However, the design used appears to be observational and cross-sectional, with a single sampling date, November 2024. Therefore, it is not possible to directly conclude that the coastal reclamation caused the observed changes unless the following are available:

unclaimed reference soils;
a chronosequence with different reclamation ages;
historical data prior to the reclamation; or
a temporal follow-up design.

Response 4: We agree with the reviewer that our cross-sectional design cannot establish causality between reclamation and soil quality changes. We explicitly acknowledged this as a limitation in the Discussion (Lines 438-447, 494-498,534-546), and clarified that our study aims to provide a baseline assessment and identify correlations, not to infer causation (Lines 129-133). We have also noted that the four reclamation areas differ in age and history, offering a space-for-time perspective, though we acknowledge that this does not replace true temporal monitoring.

Comment 5: Please moderate causal expressions such as “driving factors,” “caused by,” “led to,” or equivalents. Instead, you can use expressions such as:

“were associated with”;

“could contribute to”;

“were consistent with”; or

“were identified as variables related to.”

Response 5: We have therefore moderated all causal language throughout the manuscript (e.g., replacing "driving factors" with "variables associated with," "caused by" with "could contribute to").

Comment 6: This is especially important in lines 350–354, where the elevated concentrations of Fe, Mn, Cu, and Zn in YQ are attributed to the electrical equipment industry and electroplating operations. This explanation is plausible, but it was not directly demonstrated in the study. No data are presented on:

Please present this explanation as an interpretive hypothesis and not as a confirmed conclusion.

“Wang nitrogen et al. 2023” on line 358 appears to be an editing error or a misplaced reference.

Response 6: We thank the reviewer for this constructive comment. We agree that the attribution of elevated Fe, Mn, Cu, and Zn concentrations in YQ to the electrical equipment industry and electroplating operations, while plausible, was not directly demonstrated by our data. We have modified the discussion in the revised manuscript (Lines 433-437) as follows: “The highest concentrations of Fe, Mn, Cu, and Zn were observed in YQ (Table 1). These results may be associated with YQ historical role as a hub for electrical equipment production in China (Liu, et al., 2025). Electroplating in the electrical industry uses Cu, Zn, and Fe, wastewater and exhaust emissions may have accumulated into coastal sediments via sewage irrigation and atmospheric deposition. (Monib et al., 2024). Further target-ed source apportionment studies are needed to verify this hypothesis.”

We believe this revision appropriately reflects the speculative nature of the explanation and addresses the reviewer's concern. All causal language has been moderated throughout the manuscript, and we have ensured that this interpretation is presented as a hypothesis rather than a confirmed conclusion. And the editing error has been corrected into “Wang et al. 2023” (Line 451).

Comment 7: Line 242: Correct the spelling inconsistency in the citation of Biau and Scornet.

In equation 7 (lines 234-236), the Sensitivity Index is defined as:

SI = SQImin / SQImax
Since the SQI values are positive and, by definition, SQImax ≥ SQImin, the SI value should theoretically be greater than or equal to 1.0.

This suggests an error: either the formula in equation 7 is incorrectly written, or the calculations in the figures and text are wrong. Please correct and clarify this.

Response 7: We thank the reviewer for this critical comment. Upon re‑examination, we recalculated the Sensitivity Index (SI) using the formula SI = SQImax / SQImin. The corrected results show that SI-a (0.625/0.308=2.03) is higher than SI-w (0.638/0.327=1.95), indicating that the additive SQI exhibits greater sensitivity than the weighted SQI. This finding contradicts our earlier conclusion that SQI-w is superior to SQI-a based on its higher sensitivity. Given that the SI analysis was not essential to the main conclusions of this study, which are based on SQI comparisons and random forest analysis. We have decided to remove all SI‑related descriptions and discussions from the revised manuscript to avoid confusion and incorrect interpretation. The primary findings and conclusions of this study remain unchanged and are not affected by this removal. We have carefully checked the revised manuscript to ensure that no references to SI remain.

Comment 8: The final MDS proposed by the authors contains 14 indicators (pH, MOS, NT, available K, total soluble salts, available N, available Zn, bacterial quantity, bacterial diversity, fungal diversity, BG, CB, NAG, and XYL) out of a total of 27 originally measured (lines 297-300). This represents 52% of the total dataset.

Response 8: We thank the reviewer for this comment. We have corrected the MDS from 14 to 12 indicators in Lines 380-381 (pH, PHOS, TN, AK, TWS, AN, AZn, bacterial quantity, bacterial diversity, fungal diversity, BG, and CB). We acknowledge that 12 variables exceed the typical 4–7 indicators in some studies. Similar studies in heterogeneous landscapes have reported MDS sizes of 11–15 indicators (e.g., Ding et al., 2024; Tian et al., 2023).

Given the complexity of soil formation in the study area, which is characterized by both the sea-land interaction features prior to reclamation and the profound anthropogenic influences of post-reclamation cultivation practices, it was necessary to retain a diverse set of indicators to comprehensively capture the changes in soil quality. Importantly, we retained biological indicators (diversity and enzyme activities) to address the mechanistic questions central to this study, although we recognize these are costly for routine monitoring. The 12 selected indicators represent the most representative variables across the three principal components extracted, and further reduction would have compromised the coverage of soil quality dimensions. We have acknowledged this limitation and suggested future simplification in the revised Discussion 4.3 (Lines 534-546).

Ding M, Tuo Y, Zheng Y, Luo W, Dai Q, Li J, Shi X, Li J, He X, Xiang P. 2024. Soil Quality Evaluation of Different Forest Types in Liziping Nature Reserve Based on the Minimum Data Set[J]. Chinese Journal of Soil Science, 55(5): 1215 − 1228. DOI: 10.19336/j.cnki.trtb.2023080302

Tian Y, Xu Z, Wang Y, He J, Wang Z. 2023. Soil quality evaluation for different forest plantation of sandy land in Yinchuan Plain, Ningxia. Acta Ecologica Sinica, 2023, 43(4): 1515-1525.DOI: 10.5846/stxb202103180720.

Comment 9: Review the selection criteria in the PCA. If a Principal Component (PC) contains highly correlated variables (r > 0.60), only the one with the highest factor loading, or sum of correlations, should be retained, instead of maintaining multiple redundant variables.

Response 9: We thank you for this valuable comment. We have specified selection criteria of PCA in Discussion (Lines 534-546, 583-613).

Comment 10: Soil samples were collected at a single point in time (November 2024, post-harvest) (line 130).
Microbial communities, fungal diversity, and enzymatic activities are highly dynamic parameters that fluctuate drastically depending on the season, temperature, and, especially, water management in rice paddies (which alternate between flooding/anaerobic and drying phases). Basing a long-term Soil Quality Index (SQI) on biological indicators measured at a single point in time seriously limits its temporal representativeness. Please discuss this limitation and justify why data from a single post-harvest sampling are sufficient to establish a soil quality baseline.

Response 10: We thank the reviewer for this important methodological comment. We agree that microbial communities, enzyme activities, and fungal diversity are highly dynamic parameters that fluctuate with season, temperature, and water management in rice paddies.

We acknowledge this as a limitation of our study and have added a dedicated discussion in the revised manuscript (Lines 583-613). In this paragraph, we justify the use of post-harvest sampling for establishing a soil quality baseline as follows:

(1) Post-harvest represents a stable phase in the rice cropping cycle. By the time of sampling (November), short-term disturbances from fertilization and flooding-drying alternations have subsided, and the soil has equilibrated after the growing season.

(2) Consistent sampling timing across all four sites minimizes confounding due to seasonal variation and enables valid spatial comparisons, which is the primary objective of this study.

(3) The biological indicators included in our MDS showed clear and interpretable relationships with soil physicochemical properties and were identified as important predictors of SQI in random forest analysis. This suggests that these indicators captured meaningful site-specific variation rather than random temporal fluctuations.

Nevertheless, we recognize that a single sampling event cannot fully capture the temporal dynamics of soil quality, and we have explicitly stated that the SQI developed here represents a snapshot assessment rather than a long-term average. We have also suggested that future multi-season or multi-year monitoring studies are needed to validate the temporal stability of the SQI and to understand the seasonal sensitivity of biological indicators in reclaimed coastal paddy systems.

Comment 11: The text refers to "bacterial quantity" and "fungal abundance" as key variables of the Soil Quality Index (MDS) (lines 298 and 323). However, the methodology section only describes in detail the sequencing of amplicons (16S and ITS) to estimate diversity using the Shannon index [176-189]. Illumina MiSeq sequencing yields relative abundance data, not quantitative data. If absolute abundance or "bacterial quantity" was determined by qPCR (which is vaguely implied by the mention of the CFX96 Real-Time System on line 174), please clearly describe the qPCR protocol: the target genes, the primers used, the thermocycling conditions, and the construction of the standard curve.

Response 11: We thank the reviewer for this important methodological comment. We have added more clearly descriptions about qPCR protocol in Material and Method (Lines 235-246) as” The amplified products of quantitative qPCR were recovered by agarose gel electrophoresis, ligated into the pEASY-T3 vector, and then transformed into Escherichia coli DH5α competent cells. Positive clones were screened on ampicillin-containing plates using blue-white colony screening, and selected positive clones were sequenced. The confirmed positive clones were cultivated for plasmid DNA extraction, and the concentration and purity (OD260/OD280) were determined using a microspectrophotometer. The recombinant plasmids were serially diluted to concentrations ranging from 10⁻² to 10⁻⁷ ng μL^-1 and used as standards for qPCR targeting the ITS1 regions and 16S rRNA genes. A single peak in the melting curve confirmed that the qPCR conditions met the required standards and that the primer specificity was satisfactory. The qPCR amplification efficiencies for the 16S rRNA genes and the ITS1 regions were determined to be 96.6% and 98.9%, respectively, with a standard curve correlation coefficient greater than 0.995.”

Comment 12: The authors note that the Yueqing (YQ) area exhibits high concentrations of Fe, Mn, Cu, and Zn due to its history as a center for electrical equipment production and electroplating (Line 352). If these reclaimed areas are currently being used for rice production for human consumption (as mentioned in lines 39–40), the accumulation of heavy metals (especially copper and zinc) poses a serious risk of food safety and phytotoxicity. Further discussion is needed: Do the concentrations of these metals exceed the limits permitted for agricultural soils in China? Could the accumulation of metals in the YQ soil be negatively impacting enzyme activity and be the real cause of its lower Soil Quality Index (SQI), beyond salinity?

Response 12: We sincerely thank for this valuable comment. In response, we have supplemented the heavy metal concentrations in the YQ sediments in Supplementary Table S5.

To address this comment, we have supplemented the heavy metal concentrations in the YQ sediments in Supplementary Table S5. Based on the Chinese "Soil Environmental Quality – Risk Control Standards for Soil Pollution on Agricultural Land" (GB 15618-2018), the risk screening values for paddy soils with pH > 7.5 are as follows: As (20 mg kg⁻¹), Hg (1.0 mg kg⁻¹), Pb (240 mg kg⁻¹), Cd (0.8 mg kg⁻¹), Cr (350 mg kg⁻¹), Cu (200 mg kg⁻¹), Zn (300 mg kg⁻¹), and Ni (190 mg kg⁻¹). As shown in Table S5, the measured concentrations of all eight heavy metals in the YQ sediments are well below these risk screening values (e.g., Cu: 39.6±15.5 vs. 200 mg kg⁻¹; Zn: 124±14.5 vs. 300 mg kg⁻¹), indicating that the current heavy metal levels do not exceed the permitted limits for agricultural soils in China under the current regulatory framework.

However, we acknowledge that the coexistence of heavy metals and salinity at the YQ site, combined with their potential synergistic effects, makes it difficult to unequivocally attribute the lower SQI solely to salinity. As the reviewer rightly pointed out, it remains plausible that metal accumulation, together with salinity, may contribute to enzyme inhibition and reduced soil quality. The specific mechanisms through which heavy metal concentrations and salinity jointly affect enzyme activities, whether through direct toxicity, indirect effects via metal mobilization, or other pathways, remain unclear and warrant further investigation. In future studies, we plan to conduct controlled experiments to disentangle the individual and combined effects of salinity and heavy metals on soil enzyme activities.

We have revised the Discussion section to acknowledge this complexity and to highlight the need for future research on the interactive mechanisms between salinity and heavy metals in coastal reclaimed soils (Lines 438-447).

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

View letter

2 Although Figure 2 is titled "Pearson Correlation Coefficients Between Soil Indicators", the actual figure presented is a schematic overview of the study area.
4 Figure 4 is a composite panel containing three sub-figures, but the specific content of each sub-figure is not clearly specified, which hinders readability. It is recommended to assign appropriate titles to the whole figure and label each sub-figure as (a), (b), and (c) respectively.
Lines 13-31 The description of results in the abstract only provides qualitative comparisons between groups without supporting quantitative data, and simply indicates which variable is higher than another or which comparison shows significance.
Lines 32-33 The number of keywords in the abstract is insufficient to effectively highlight the methodological characteristics and core innovations of this study. It is recommended to add 1 to 2 additional keywords.
Lines 99-101: Hypothesis 2 lacks novelty. Given that a large number of microbial indicators have been measured in this study, the proposition that "microbial indicators are important" is essentially a pre-determined outcome.
Lines 116-117: Fertilization practices in the LG region are described in this section, and described again in Lines 120-122. Please verify the relevant content and consolidate all relevant data into a unified table to improve data readability.
Lines 130-122: The sample sizes of the four study regions vary considerably (9, 18, 9, 13). Please provide a reasonable and detailed explanation for this sampling strategy.
Lines 164-175 Key methodological details for the quantitative analysis are missing. Only primer sequences and reaction systems are reported, while critical parameters including standard curve construction, amplification efficiency and correlation coefficient are not provided. This makes it impossible to verify the reliability of the quantitative results of microbial abundance. It is recommended to supplement complete methodological information for qPCR.
Lines 176-178 The use of only the Shannon index is insufficient to reflect differences in community composition. It is recommended to supplement indicators such as Chao1 or ACE. If other diversity indices have been calculated but not presented, please explain why only the Shannon index is reported.
Lines 198-201 The authors adopted "the top 10% of loading values" as the screening criterion, but did not specify the source of this criterion. Please justify the selection of the 10% threshold in this study to improve methodological standardization.
Lines 211-213 The "Technical Specification for Cultivated Land Fertility Survey and Standard Farmland Classification in Zhejiang Province" is mentioned in the text. It is recommended to add this specification to the supplementary materials.
Lines 222-224 The SQI classification threshold is directly adopted from Marzaioli et al., but the applicability of this threshold to the soils investigated in this study has not been demonstrated.
Lines 264-270 The microbiological results only focus on α-diversity. Incorporating analysis of β-diversity (via PCoA/NMDS/PERMANOVA) will substantially enhance the value of the manuscript.
Lines 298-300 The final Minimum Dataset (MDS) established by the authors includes 14 indicators, while the original indicator set contained 24 indicators. The MDS retains more than half of the original variables, which contradicts to a certain extent the connotation of the "Minimum Dataset" concept. It is recommended to discuss why the final MDS is still relatively large and whether further indicator reduction is possible.
Lines 319-326 First, the explained variance of the random forest model only reflects the ranking of variable importance, but the root mean square error of the model is not reported in the text, which makes it impossible to evaluate the absolute goodness of fit of the model. Second, although the authors used the rfPermute package to obtain P-values via permutation tests for nearly 20 predictors, whether false discovery rate (FDR) correction was performed for multiple testing is not mentioned. Given that a large number of variables are marked as "significant" (P<0.05), false positives resulting from multiple testing are highly likely.
Lines 326-327 The difference in the Sensitivity Index between the two weighting methods is extremely small (0.013), which is insufficient to support the conclusion that SQIw is significantly superior to SQIa. It is recommended to describe this result as "Slightly higher" rather than "higher" and avoid over-interpretation.
Lines 339-355 The interpretation of elevated heavy metal concentrations in YQ linked to the electrical manufacturing industry is speculative, as no source apportionment investigation was conducted. Therefore, overly definitive conclusions on this association should be avoided.
Lines 357-383 The discussion on the dominance of fungi under salinization relies heavily on existing literature rather than evidence obtained from this study. It is recommended to supplement relevant analysis based on soil microbial indicators from this study.

Comments for author File: Comments.pdf

Author Response

Comment 1: Although Figure 2 is titled "Pearson Correlation Coefficients Between Soil Indicators", the actual figure presented is a schematic overview of the study area.

Response 1: We appreciate you pointing out the error. We have corrected the image content (Fig.2).

Comment 2: Figure 4 is a composite panel containing three sub-figures, but the specific content of each sub-figure is not clearly specified, which hinders readability. It is recommended to assign appropriate titles to the whole figure and label each sub-figure as (a), (b), and (c) respectively.

Response 2: We thank the reviewer for this suggestion. We have revised Figure 4 to improve its clarity and readability. The figure now includes an overall title, and the three sub-figures have been clearly labeled as (a), (b), and (c), with each sub‑figure appropriately titled to specify its content. These modifications have been made in the revised manuscript.

Comment 3: Lines 13-31 The description of results in the abstract only provides qualitative comparisons between groups without supporting quantitative data, and simply indicates which variable is higher than another or which comparison shows significance.

Response 3: We thank the reviewer for this valuable suggestion. We agree that the abstract should present quantitative data to support the qualitative comparisons among groups. Accordingly, we have revised the abstract to include specific numerical values and statistical results for the key comparisons.

Specifically, we have added:

(1) Quantitative comparisons of soil properties: For example, YQ had higher contents of soil organic carbon (SOC: 22.82 g kg-1), total nitrogen (TN: 0.14 g kg-1), total water-soluble salts (TWS: 3.09 g kg^-1), cation exchange capacity (CEC: 21.77 cmol(+) kg^-1), and available Fe (44.23 mg kg^-1), Mn (36.63 mg kg^-1), Cu (30.31 mg kg^-1), and Zn (8.79 mg kg^-1) than the other sites. RA exhibited significantly higher activities of β-glucosidase (BG :32.79 nmol g^-1 h^-1), xylanase (XYL: 5.88 nmol g^-1 h^-1), N-acetyl-β-D-glucosaminidase (NAG: 18.00 nmol g^-1 h^-1), leucine aminopeptidase (LAP: 27.01 nmol g^-1 h^-1), and acid phosphatase (PHOS: 54.50 nmol g^-1 h^-1) than the other sites (P < 0.05). LW had the greatest bacterial and fungal abundances, whereas LG displayed the highest fungal diversity.

(2) Quantitative SQI comparisons: RA showed the highest SQI (SQIw:0.49; SQIa:0.48), followed by LW (SQIw0.47; SQIa:0.47), LG (SQIw:0.43; SQIa:0.41), and YQ (SQIw:0.38; SQIa:0.41).

(3) Statistical significance with specific p-values: We have reported specific p‑values from the Welch ANOVA and random forest analyses to support the stated differences (e.g., TWS: p = 0.016).

We believe these revisions make the abstract more informative and quantitatively grounded. The revised abstract has been updated in the manuscript (Lines 19-31).

Comment 4: Lines 32-33 The number of keywords in the abstract is insufficient to effectively highlight the methodological characteristics and core innovations of this study. It is recommended to add 1 to 2 additional keywords.

Response 4: We thank the reviewer for this suggestion. We agree that additional keywords would better highlight the methodological characteristics and core innovations of this study. Accordingly, we have added "minimum data set" and "random forest analysis" to the keyword list. These additions reflect the two key methodological innovations of our work: (1) the construction of a minimum data set for soil quality evaluation, and (2) the application of random forest to identify the most important predictors of soil quality. We believe these revisions will improve the discoverability and representativeness of our manuscript.

The revised keywords are “Keywords: reclaimed coastal land; paddy field; soil enzyme activity; soil quality; minimum data set; random forest analysis” (Lines 38-39).

Comment 5: Lines 99-101: Hypothesis 2 lacks novelty. Given that a large number of microbial indicators have been measured in this study, the proposition that "microbial indicators are important" is essentially a pre-determined outcome.

Response 5: We thank you for your valuable comment. We have detailed hypothesis 2 as “soil microbial indicators, particularly enzyme activities, are strongest indicators of soil quality in reclaimed coastal paddy fields.in Introduction (Lines 127-129).

Comment 6: Lines 116-117: Fertilization practices in the LG region are described in this section, and described again in Lines 120-122. Please verify the relevant content and consolidate all relevant data into a unified table to improve data readability.

Response 6: We thank the reviewer for pointing out this repetition. We have carefully reviewed the fertilization descriptions in the manuscript and found that the LG fertilization regime was indeed described twice (Line 162 and 167) with inconsistent data between the two descriptions. We have now verified the correct fertilization practices for LG based on the field management records: compound fertilizer at 350 kg ha⁻¹, organic fertilizer at 1500 kg ha⁻¹, and urea topdressing at 200 kg ha⁻¹. The second description (375/1500/225) was an error and has been corrected as LW.

To improve data readability and avoid further redundancy, we have consolidated all fertilization information for the four sites into a unified table (Supplementary Table S9). In the main text, we now briefly refer to this table instead of repeating the full fertilization details. This revision ensures consistency and enhances the clarity of the data presentation.

Comment 7: Lines 130-122: The sample sizes of the four study regions vary considerably (9, 18, 9, 13). Please provide a reasonable and detailed explanation for this sampling strategy.

Response 7: We thank the reviewer for raising this important statistical concern. We fully agree that sample size imbalance among the four sites (YQ: 9, LW: 18, LG: 9, RA: 13) may affect the homogeneity of variances in ANOVA.

We applied Welch's ANOVA for indicators with significant Levene test results (p < 0.05). This method does not assume equal variances across groups and corrects the degrees of freedom using the Welch–Satterthwaite approximation, providing more reliable inference when sample sizes are unequal and variances are heterogeneous. In the revised manuscript, we have explicitly stated this procedure in the Materials and Methods section (Lines 315-319) and the Supplementary Information (Table S7 and Table S8). We have confirmed that the observed differences remain robust after this correction.

For the 12 indicators ultimately retained in the MDS (pH, SOM, TN, AK, TWS, AN, available Zn, bacterial quantity, bacterial diversity, fungal diversity, BG, and CB), the results show that most of the retained MDS indicators exhibit significant differences among the four sites (p < 0.05), with the exception of pH, TWS and BG. These findings support the robustness of our MDS selection and subsequent SQI comparisons. We have added a description of the Levene's test and Welch ANOVA procedures, along with the corresponding results, to the Materials and Methods (Lines 315-319).

Comment 8: Lines 164-175 Key methodological details for the quantitative analysis are missing. Only primer sequences and reaction systems are reported, while critical parameters including standard curve construction, amplification efficiency and correlation coefficient are not provided. This makes it impossible to verify the reliability of the quantitative results of microbial abundance. It is recommended to supplement complete methodological information for qPCR.

Response 8: We thank the reviewer for this important methodological comment. We have added more clearly descriptions about qPCR protocol in Material and Method (Lines 235-246) as” The amplified products of quantitative qPCR were recovered by agarose gel electrophoresis, ligated into the pEASY-T3 vector, and then transformed into Escherichia coli DH5α competent cells. Positive clones were screened on ampicillin-containing plates using blue-white colony screening, and selected positive clones were sequenced. The confirmed positive clones were cultivated for plasmid DNA extraction, and the concentration and purity (OD260/OD280) were determined using a microspectrophotometer. The recombinant plasmids were serially diluted to concentrations ranging from 10⁻² to 10⁻⁷ ng/μL and used as standards for qPCR targeting the ITS1/ITS2 regions and 16S rRNA genes. A single peak in the melting curve confirmed that the qPCR conditions met the required standards and that the primer specificity was satisfactory. The amplification efficiency for the soil samples was 96.6%, with a standard curve correlation coefficient greater than 0.995.”

Comment 9: Lines 176-178 The use of only the Shannon index is insufficient to reflect differences in community composition. It is recommended to supplement indicators such as Chao1 or ACE. If other diversity indices have been calculated but not presented, please explain why only the Shannon index is reported.

Response 9: We thank the reviewer for this comment. We acknowledge that the Shannon index alone does not fully capture all aspects of microbial community composition, and we appreciate the suggestion to consider additional diversity indices such as Chao1 or ACE.

In this study, we calculated both richness estimators (Chao1, ACE) and diversity indices (Shannon, Simpson) for all samples. During the minimum data set (MDS) construction, we included bacterial diversity and fungal diversity as two of the original 24 indicators. The selection of Shannon index in the final MDS was determined by the PCA-based screening procedure, which retained indicators with high factor loadings within each principal component while removing redundant indicators through correlation analysis. Compared to Chao1 and ACE, the Shannon index showed higher factor loadings and was less correlated with other retained indicators (e.g., enzyme activities, bacterial/fungal quantities), thereby providing more independent information in the MDS.

However, we agree with the reviewer that the Shannon index alone has limitations in fully representing community composition, as it primarily reflects species evenness rather than richness. We have acknowledged this limitation in the revised Discussion (Lines 494-498) and have suggested that future studies should incorporate a broader range of diversity metrics (e.g., Chao1, ACE, Simpson) to more comprehensively assess microbial community structure in relation to soil quality.

Comment 10: Lines 198-201 The authors adopted "the top 10% of loading values" as the screening criterion, but did not specify the source of this criterion. Please justify the selection of the 10% threshold in this study to improve methodological standardization.

Response 10: We thank the reviewer for this methodological comment. We agree that the source and justification of the "top 10%" screening criterion should be clearly stated to improve methodological standardization.

The selection of the top 10% of factor loadings within each principal component follows established practices in minimum data set (MDS) construction for soil quality assessment. This threshold has been adopted in multiple peer‑reviewed studies. For example, in a recent soil health index study, the authors explicitly stated in Section 2.3.1, “In each PC, the highest absolute value of the rotational load and all variables between it and its 10% difference were selected. When multiple variables met the preceding condition in a PC, Pearson’s correlation analysis (P <0.05) was considered to minimize redundancy. " (Gao et al., 2025). Also, “Highly weighted variables were defined as those with absolute values within 10%of the highest factor loading.” as the screening criterion for MDS construction (Li et al., 2019). From a statistical perspective, using the top 10% of absolute loadings is a common heuristic for PCA‑based feature selection, as it retains the most influential variables within each principal component while excluding those with marginal contributions.

We chose this threshold because: (1) it offers a balance between retaining sufficient information from each principal component and maintaining a parsimonious MDS; (2) it is consistent with the majority of published MDS studies, thereby facilitating comparability across different soil quality assessments; and (3) in our dataset, this criterion effectively identified the most representative indicators within each PC while excluding weakly contributing variables.

We have now added a brief justification and the corresponding reference below to the Materials and Methods section (Line 273) to clarify the source of this criterion.

Gao, M., Hu, W., Li, M., Wang, S., & Chu, L. 2025. Network analysis was effective in establishing the soil quality index and differentiated among changes in land-use type. Soil and Tillage Research. 246, 106352.https://doi.org/10.1016/j.still.2024.106352.

Li, P., Shi, K., Wang, Y., Kong, D., Liu, T., Jiao, J., Liu, M., Li, H., & Hu, F. 2019. Soil quality assessment of wheat maize cropping system with different productivities in China: Establishing a minimum data set. Soil and Tillage Research. 190, 31-40. https://doi.org/10.1016/j.still.2019.02.019.

Comment 11: Lines 211-213 The "Technical Specification for Cultivated Land Fertility Survey and Standard Farmland Classification in Zhejiang Province" is mentioned in the text. It is recommended to add this specification to the supplementary materials.

Response 11: We thank the reviewer for this suggestion. To improve the transparency and verifiability of our methodology, we have added a dedicated entry for the DB33/T 895-2013 standard in the Supplementary Materials. This entry provides the full citation of the standard, including its title, issuing authority, publication date, and a reliable source for accessing the full text. We have also briefly clarified the specific aspects of our study that refer to this standard. This revision ensures that readers can readily consult the original document for a more comprehensive understanding of our methodological framework.

Comment 12: Lines 222-224 The SQI classification threshold is directly adopted from Marzaioli et al., but the applicability of this threshold to the soils investigated in this study has not been demonstrated.

Response 12: We thank the reviewer for this critical and constructive comment. We agree that the SQI classification threshold proposed by Marzaioli et al. (2010) (SQI < 0.55, poor quality; 0.55 ≤ SQI ≤ 0.70, moderate quality; SQI > 0.70, good quality) was originally developed based on Mediterranean ecosystems in Southern Italy, and its direct applicability to the coastal reclaimed paddy soils of eastern China has not been empirically validated.

We chose this threshold for the following reasons:

(1) Widely adopted reference: The Marzaioli et al. (2010) classification has been extensively used as a reference standard in soil quality assessment studies across different regions and land use types, providing a common benchmark for cross‑study comparison. In the absence of region‑specific threshold values for coastal reclaimed soils in China, adopting this widely recognized classification allows our results to be compared with those of other studies.

(2) Interpretive framework, not absolute standard: We acknowledge that this threshold serves as an interpretive reference rather than an absolute standard. Our primary objective was to rank the relative soil quality among the four reclamation areas rather than to make absolute judgments about soil quality classes. The application of this threshold facilitated the interpretation of our SQI values in a broader context.

Nevertheless, we fully agree with the reviewer that the direct transferability of this threshold to our study area requires further validation. We have therefore:

Added a clear statement of this limitation in the revised Discussion section (Lines 524-533), explicitly noting that the Marzaioli et al. (2010) threshold was developed for Mediterranean soils and may not be directly applicable to coastal reclaimed paddy soils without local calibration.

Revised our interpretation to emphasize that the SQI classification in this study is relative and referential, and that future studies should establish region‑specific threshold values through long‑term monitoring and correlation with crop productivity or other soil functions.

We hope these revisions adequately address the reviewer's concern regarding the applicability of the SQI classification threshold.

Comment 13: Lines 264-270 The microbiological results only focus on α-diversity. Incorporating analysis of β-diversity (via PCoA/NMDS/PERMANOVA) will substantially enhance the value of the manuscript.

Response 13: We thank the reviewer for this valuable comment. We have incorporated beta diversity into our calculation and presented the β-diversity results using three complementary approaches: PCoA, PCA, and NMD β-diversity measures the dissimilarity among samples, reflecting the similarity of microbial community composition across sites (Table.1). However, existing studies focus on various microbial communities, and the inclusion of β-diversity in MDS construction remains insufficiently explored. Consequently, it is challenging to assign a unified scoring function to β-diversity—whether it should be classified as a "higher-the-better", "lower-the-better", or "parabolic" indicator remains unclear.

To evaluate whether β-diversity influences our results, we tested both "higher-the-better" and "lower-the-better" scoring functions for beta diversity, while the "parabolic" function could not be calculated due to the lack of a literature‑recommended threshold. After recalculation, the results remained largely consistent with our original conclusions: SQIw outperformed SQIa. In future studies, we plan to further investigate the contribution of β-diversity to soil quality and its potential role in MDS development. Please see the attachment.

Comment 14: Lines 298-300 The final Minimum Dataset (MDS) established by the authors includes 14 indicators, while the original indicator set contained 24 indicators. The MDS retains more than half of the original variables, which contradicts to a certain extent the connotation of the "Minimum Dataset" concept. It is recommended to discuss why the final MDS is still relatively large and whether further indicator reduction is possible.

Response 14: We thank the reviewer for this valuable comment. We have corrected the MDS from 14 to 12 indicators (pH, PHOS, TN, AK, TWS, AN, AZn, bacterial quantity, bacterial diversity, fungal diversity, BG, and CB). We acknowledge that 12 variables exceed the typical 4–7 indicators in some studies. Similar studies in heterogeneous landscapes have reported MDS sizes of 11–15 indicators (e.g., Ding et al., 2024; Tian et al., 2023).

However, given the heterogeneity of our four study areas (differing in reclamation age, parent material, and management), retaining multiple categories of indicators was necessary to capture the full dimensionality of soil quality variation. Importantly, we retained biological indicators (diversity and enzyme activities) to address the mechanistic questions central to this study, although we recognize these are costly for routine monitoring. The 12 selected indicators represent the most representative variables across the three principal components extracted, and further reduction would have compromised the coverage of soil quality dimensions. We have acknowledged this limitation and suggested future simplification in the revised Discussion 4.3 (Lines 532-544).

Comment 15: Lines 319-326 First, the explained variance of the random forest model only reflects the ranking of variable importance, but the root mean square error of the model is not reported in the text, which makes it impossible to evaluate the absolute goodness of fit of the model. Second, although the authors used the rfPermute package to obtain P-values via permutation tests for nearly 20 predictors, whether false discovery rate (FDR) correction was performed for multiple testing is not mentioned. Given that a large number of variables are marked as "significant" (P<0.05), false positives resulting from multiple testing are highly likely.

Response 15: We thank the reviewer for this important methodological comment. We agree that both the model goodness‑of‑fit and the treatment of multiple testing in the random forest analysis require clarification. We have addressed these points as follows:

(1) Model goodness‑of‑fit (RMSE)

The random forest model for SQIw explained 63.02% of the variance (R² = 0.6302) with an RMSE of 0.0467 based on out‑of‑bag predictions, while the model for SQIa explained 60.78% of the variance (R² = 0.6078) with an RMSE of 0.0506. The RMSE values accounted for approximately 10–12% of the observed SQI range, indicating acceptable predictive performance for the random forest models.

(2) Multiple‑testing correction (FDR)

The reviewer raises an important concern regarding multiple testing. In the random forest analysis, we used the rfPermute package with 999 permutations to obtain p‑values for variable importance. However, we acknowledge that we did not apply a formal correction (e.g., FDR/Benjamini‑Hochberg) to the resulting p‑values. Given that 19 predictors were tested, the probability of false positives is indeed elevated when multiple variables are reported as significant at the nominal α = 0.05 level.

To address this, we have now applied the Benjamini‑Hochberg false discovery rate (FDR) correction to the importance p‑values. After correction, the variables that remained significant at FDR < 0.05 were [list retained variables, e.g., XYL, TWS, BG, etc.], confirming the robustness of the most important predictors identified in our original analysis.

Comment 16: Lines 326-327 The difference in the Sensitivity Index between the two weighting methods is extremely small (0.013), which is insufficient to support the conclusion that SQIw is significantly superior to SQIa. It is recommended to describe this result as "Slightly higher" rather than "higher" and avoid over-interpretation.

Response 16: We thank the reviewer for this observation. We fully agree that a difference of 0.013 in the Sensitivity Index is too small to support any substantive conclusion about the superiority of SQIw over SQIa.

We have removed all Sensitivity Index (SI)‑related descriptions and discussions from the revised manuscript.

The comparison between SQIw and SQIa based on SI values (including the 0.013 difference mentioned by the reviewer) no longer appears in the revised manuscript. Our main conclusions regarding SQI are now based on: (1) the actual SQI values themselves (SQIw: 0.46–0.50; SQIa: 0.42–0.48), and (2) the random forest analysis, which identified the key predictors of soil quality. These core findings remain robust and unchanged. This decision of SI delete was made because the SI analysis was not essential to our main conclusions.

We appreciate the reviewer's careful attention to this detail and agree that over‑interpretation of such a small difference should be avoided.

Comment 17: Lines 339-355 The interpretation of elevated heavy metal concentrations in YQ linked to the electrical manufacturing industry is speculative, as no source apportionment investigation was conducted. Therefore, overly definitive conclusions on this association should be avoided.

Response 17: We thank the reviewer for this constructive comment. We agree that the attribution of elevated Fe, Mn, Cu, and Zn concentrations in YQ to the electrical equipment industry and electroplating operations, while plausible, was not directly demonstrated by our data. We have therefore revised the relevant text to present this explanation as an interpretive hypothesis rather than a confirmed conclusion, and we have explicitly acknowledged the lack of direct evidence (e.g., metal concentrations in sediments and irrigation water, isotope or source analysis, contamination history, and geochemical fractions) supporting this attribution.

Comment 18: Lines 357-383 The discussion on the dominance of fungi under salinization relies heavily on existing literature rather than evidence obtained from this study. It is recommended to supplement relevant analysis based on soil microbial indicators from this study.

Response 18: We thank the reviewer for this constructive comment. We believe this revision appropriately reflects the speculative nature of the explanation and addresses the reviewer's concern. All causal language has been moderated throughout the manuscript, and we have ensured that this interpretation is presented as a hypothesis rather than a confirmed conclusion.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript investigates soil quality across four reclaimed coastal paddy-field areas in Zhejiang Province by integrating soil physicochemical properties, microbial abundance and diversity, and extracellular enzyme activities. The authors construct additive and weighted soil quality indices using a PCA-derived minimum data set and subsequently apply correlation and random forest analyses to identify variables associated with soil quality. The topic is relevant, and the inclusion of microbial and enzymatic indicators in the assessment of reclaimed coastal soils is potentially valuable. The sampling effort and the broad range of measured variables could provide useful information on the condition of these agricultural soils. However, the current version contains major methodological inconsistencies in the construction of the minimum data set, calculation and validation of the soil quality indices, statistical interpretation, and reporting of the analytical data. Several conclusions are also more causal and general than can be supported by the observational design. Therefore, I recommend major revision at this stage.

General comments

The construction of the minimum data set is not consistent with the PCA results reported in Table S1.

This is the most important issue in the manuscript. The selection of the MDS variables described in Lines 276–300 does not correspond to the PCA loadings reported in Table S1.

According to the stated criterion, variables whose absolute loading is within 10% of the highest absolute loading for each principal component should be retained. However:

For PC2, the highest loading is for CB (0.41), followed by NAG and PHOS (both 0.39). The manuscript instead identifies BG, NAG, and XYL.

For PC3, the variables meeting the criterion appear to be AK, TWS, bacterial diversity, and XYL. The manuscript includes CB instead of XYL.

For PC5, the highest-loading variables are AN, AK, and XYL, whereas the manuscript again reports CB.

For PC7, fungal diversity has the highest loading (0.57). LAP has an absolute loading of only 0.42 and does not appear to meet the stated 10% criterion.

The treatment of correlated variables in PC1 is also unclear. The manuscript states that only the variable with the highest loading should be retained when correlations exceed 0.6, but subsequently retains SOC, TN, and Zn without fully reporting the pairwise correlations used for this decision.

Furthermore, the final MDS reported in Lines 298–300 differs from Table S2, and the weights reported in Lines 306–309 differ from those in the supplementary material. For example, Table S2 includes LAP but not AN, whereas the main text includes AN but not LAP. The main text also reports weights for only 12 variables, although the proposed MDS contains at least 14 variables.

The complete MDS procedure must therefore be repeated from the original data. The authors should provide:

the standardized or transformed data used for PCA;

the complete loading-selection procedure for each PC;

the relevant pairwise correlations used to remove redundant variables;

the final non-duplicated list of MDS indicators;

the communalities and recalculated weights;

recalculated SQI values, statistical comparisons, figures, and conclusions.

Until this analysis is corrected, the central results of the manuscript cannot be considered reliable.

The identification of “drivers” of SQI using random forest and correlation analysis is circular.

The SQI is mathematically calculated from the same soil variables that are subsequently used as predictors in the random forest analysis and correlated with the SQI. Consequently, finding that enzyme activities, nutrients, microbial abundance, or salinity predict the SQI is largely expected by construction.

This analysis cannot demonstrate that these variables “drive” soil quality. It mainly identifies which components of the index contribute most strongly to the index itself. The same limitation applies to the correlations between the SQI and its component indicators.

The authors should substantially revise the interpretation of these analyses. The results may be described as indicator contributions or statistical associations, but not as evidence of ecological drivers or regulatory mechanisms. A valid analysis of drivers would require an independent response variable, such as crop yield, nutrient-use efficiency, aggregate stability, plant health, or another soil function not used to calculate the SQI.

The sampling design does not allow the observed differences to be attributed specifically to reclamation or to individual soil properties.

The study compares four geographically distinct areas, but each area also differs in climate, fertilization rate, reclamation history, salinity, and potentially industrial and agricultural management. Therefore, “site” is confounded with several environmental and management factors.

The 49 composite samples provide within-area replication, but they do not represent independent replicated reclamation treatments. The conclusions should consequently be restricted to the four investigated areas and should not be generalized to reclaimed coastal paddy fields in eastern China without qualification.

The authors should clarify:

the number of independent farms or fields represented in each area;

the spatial distances among sampling plots;

whether multiple plots originated from the same farm or management unit;

the exact reclamation age of each area and sampling plot;

crop rotation, irrigation, drainage, fertilization history, and residue management;

whether spatial autocorrelation was evaluated.

The scoring functions and SQI calculation are insufficiently described and include mathematical inconsistencies.

The manuscript does not provide the values of L and U used in the ascending and descending membership functions. It is also unclear whether these limits were obtained from the observed minimum and maximum values, regional standards, agronomic sufficiency ranges, or published thresholds. The outer limits required for the parabolic functions applied to pH and bulk density are not reported.

Several units and numerical values in Tables 1 and 2 appear incorrect or incomplete.

The reported units require careful verification.

In Table 1, AN, AP, AK, and DTPA-extractable Fe, Mn, Cu, and Zn are reported in g kg⁻¹. Values such as 40 g kg⁻¹ of available phosphorus or 30 g kg⁻¹ of available copper would be highly implausible. These variables are more likely expressed in mg kg⁻¹.

The reported TN values of 0.10–0.14 g kg⁻¹ should also be verified. In combination with SOC concentrations of approximately 16–23 g kg⁻¹, these values would produce extremely high and biologically unusual C:N ratios. A decimal or unit error appears possible.

Table 2 does not provide units for bacterial and fungal abundance. The authors should specify whether the values represent gene copies per gram of dry soil, values multiplied by a scaling factor, or log-transformed qPCR data.

For both tables, the authors should report:

whether values are means ± SD or means ± SE;

the sample size for each area;

the post hoc test used after ANOVA;

any transformation applied before statistical testing.

The qPCR and amplicon-sequencing methods are not sufficiently described for reproducibility.

The qPCR section lacks several essential details, including:

the exact primer sequences and final primer concentrations;

the amplification programme;

the type and preparation of the standard;

standard-curve efficiency and R²;

technical replication;

no-template and extraction controls;

assessment of PCR inhibition;

calculation of gene-copy abundance;

conversion to copies per gram of dry soil;

transformation of the data before statistical analysis.

The exact 515F/806R primer versions should be reported because some versions also amplify archaeal 16S rRNA genes. Similarly, the fungal primers should be identified by their complete names and sequences.

The sequencing workflow is described as being performed using both “QIIME” and “QIIME2”, while OTU clustering was conducted using UPARSE and USEARCH. The authors should provide software versions and a coherent description of the complete pipeline.

They should also report:

raw and retained read numbers per sample;

sequencing depth and sample coverage;

the normalization or rarefaction procedure used before calculating Shannon diversity;

whether any samples were removed;

negative extraction and PCR controls;

whether sequencing was randomized across batches.

Without normalization or coverage information, differences in Shannon diversity may partly reflect sequencing-depth variation. The deposited SRA datasets should also include complete sample metadata linking each sequence to its sampling area and field.

The statistical analyses require more complete reporting and validation.

The study includes approximately 24 measured indicators, rather than the 27 stated in Lines 93–95. With only 49 samples, PCA involving this number of variables may be unstable. The authors should justify sample adequacy and preferably evaluate the stability of the selected components and MDS variables through bootstrap or resampling procedures.

The manuscript should also specify:

whether PCA was conducted on a correlation matrix or covariance matrix;

whether variables were centred and scaled before PCA;

whether skewed variables, particularly qPCR and enzyme data, were transformed;

how normality and homoscedasticity were assessed;

which multiple-comparison procedure followed ANOVA;

whether multiple-testing correction was applied to the correlation matrix.

Minor and specific comments

Lines 35–48: The opening paragraph contains repeated use of “Therefore” and should be streamlined.

Lines 49–50: Revise to “reflects the capacity of soil to sustain productivity…”.

Lines 53–65: Some references do not appear closely aligned with the statements they support. The authors should verify that each citation directly supports the claimed effects of reclamation.

Lines 76–78: Revise the sentence concerning extracellular polysaccharides; a verb or conjunction is missing.

Lines 93–95: Verify the total number of indicators. The manuscript and supplementary table appear to include 24 rather than 27 variables.

Lines 106–110: Improve the description of climate and provide the exact reclamation age for each area rather than the general range of 5–8 years.

Lines 112–123: LG is described twice. The second occurrence probably refers to LW, but this must be verified because the management information is important for interpreting site differences.

Lines 128–129: “Traditional practice” is too vague. Irrigation regime, flooding–drainage management, and major pest-control practices should be briefly described.

Lines 143–154: Correct grammatical errors such as “using the with potassium dichromate oxidation”. Provide appropriate methodlogical references and quality-control information.

Lines 155–161: Standardize enzyme nomenclature. CB should be described as β-D-cellobiosidase rather than β-cellulase, and XYL as β-xylosidase rather than xylanase, unless different substrates were actually used.

Lines 163–194: Correct “QIIME pipeline” versus “QIIME2” and provide software versions and parameters.

Lines 196–220: Correct “elimate”, define all scoring parameters, and clarify that the denominator in Equation 4 refers to the selected MDS indicators rather than the total data set.

Lines 221–236: Revise the grammar and correct the sensitivity-index equation.

Lines 237–246: Correct the citation spelling “Biau & Scornet” and provide the complete statistical workflow.

Tables 1 and 2: Include sample sizes, definitions of error terms, statistical tests, exact units, and transformations.

Lines 263–269: Resolve the inconsistency concerning the highest fungal diversity.

Lines 276–300: Recalculate the entire MDS and correct SOM to SOC.

Lines 306–327: Ensure that all listed weights correspond exactly to the final MDS and sum to one, allowing for rounding.

Lines 338–355: Avoid attributing elevated available metals to industrial emissions without direct evidence.

Line 358: Correct “Wang nitrogen et al. 2023”.

Lines 368–390: Several correlations are interpreted causally. Replace terms such as “revealing” and “demonstrating” with more cautious language such as “suggesting” or “being consistent with”.

Line 369: The reference to Table 2 appears incorrect because Table 2 does not contain correlations.

Lines 450–466: Management recommendations should be shortened and clearly separated from conclusions directly supported by the study.

Author Response

Comment 1:

The construction of the minimum data set is not consistent with the PCA results reported in Table S1.

This is the most important issue in the manuscript. The selection of the MDS variables described in Lines 276–300 does not correspond to the PCA loadings reported in Table S1.

According to the stated criterion, variables whose absolute loading is within 10% of the highest absolute loading for each principal component should be retained. However:

For PC2, the highest loading is for CB (0.41), followed by NAG and PHOS (both 0.39). The manuscript instead identifies BG, NAG, and XYL.

For PC3, the variables meeting the criterion appear to be AK, TWS, bacterial diversity, and XYL. The manuscript includes CB instead of XYL.

For PC5, the highest-loading variables are AN, AK, and XYL, whereas the manuscript again reports CB.

For PC7, fungal diversity has the highest loading (0.57). LAP has an absolute loading of only 0.42 and does not appear to meet the stated 10% criterion.

The complete MDS procedure must therefore be repeated from the original data. The authors should provide:

the standardized or transformed data used for PCA;

the complete loading-selection procedure for each PC;

the relevant pairwise correlations used to remove redundant variables;

the final non-duplicated list of MDS indicators;

the communalities and recalculated weights;

recalculated SQI values, statistical comparisons, figures, and conclusions.

Until this analysis is corrected, the central results of the manuscript cannot be considered reliable.

Response 1: We sincerely apologize for the table alignment error in the original Table S1. Upon re‑examination, we found that the variable names starting from LAP were misaligned (shifted downward by one row), causing the enzyme activity indicators (AG, BG, CB, XYL, NAG, PHOS) to be incorrectly assigned to different principal components. This error led to the inconsistencies identified by the reviewer. We have now corrected Table S1 and re‑executed the entire MDS selection procedure. All results based on the corrected data are consistent with our original conclusions. The revised Table S1 and the complete MDS selection process are now provided in the revised manuscript and supplementary materials for transparency.

Comment 2:

The identification of “drivers” of SQI using random forest and correlation analysis is circular.

Response 2: We thank the reviewer for this insightful and methodologically rigorous comment. We fully agree that using the same variables to both construct the SQI and then predict it introduces a degree of circularity. We acknowledge that the SQI is mathematically derived from these indicators, and therefore, finding that these indicators are "predictors" of SQI is largely expected by construction. Our original use of the term "drivers" was indeed inappropriate and overstated the causal interpretability of our analysis.

We have therefore substantially revised our interpretation in the revised manuscript as follows:

(1) Terminological revision: We have replaced all instances of "drivers," "driving factors," and similar causal language with more accurate terms such as "important indicators," or "variables” throughout the Abstract, Results, and Discussion sections. This revision avoids implying causal relationships and instead focuses on the quantitative contribution of each indicator to the SQI as constructed.

(2) Clarification of analytical objectives: In the revised manuscript, we now explicitly state that the random forest analysis was used to identify which indicators contribute most strongly to the SQI values, rather than to identify ecological drivers or regulatory mechanisms. This is a tool for dimensionality reduction and indicator prioritization, not for causal inference. The random forest results help us understand which measured variables have the greatest influence on the index itself, thereby guiding future monitoring efforts toward the most informative indicators.

(3) Acknowledgment of limitations: We have added a clear statement in the Discussion section (Lines 560-589) acknowledging that our analysis does not demonstrate causal relationships between soil properties and soil quality. We explicitly note that an independent validation of SQI against external responses (e.g., crop yield, nutrient-use efficiency, aggregate stability, or plant health) would be necessary to establish the SQI's functional relevance. We have also acknowledged that the current study lacks such an independent response variable, and we have identified this as a key limitation that should be addressed in future research.

We believe these revisions appropriately address the reviewer's concern and ensure that our claims are aligned with the actual analytical capabilities of our study design.

Comment 3:

The sampling design does not allow the observed differences to be attributed specifically to reclamation or to individual soil properties.

The authors should clarify:

the number of independent farms or fields represented in each area;

the spatial distances among sampling plots;

whether multiple plots originated from the same farm or management unit;

the exact reclamation age of each area and sampling plot;

crop rotation, irrigation, drainage, fertilization history, and residue management;

whether spatial autocorrelation was evaluated.

Response 3: We have provided detailed information including year of reclaiming, year of paddy establishment, number of years under cultivation, irrigation water source drainage regime and frequency previous land use history, and duration of flooding of each site in supplemental information (Table S3, S4). In this study, we more focus on the total differences of soil quality in study areas. The information reveals notable differences in reclamation age among the sites, ranging from <1 year to approximately 10 years.

We have therefore revised our interpretation throughout the manuscript to emphasize that the observed patterns reflect a composite outcome of multiple factors including reclamation history, management practices, and inherent site characteristics rather than attributing the differences solely to spatial variation. We have also explicitly acknowledged the confounding effects of these variables as a limitation in the Discussion section (Lines 581-592). Future studies employing a chronosequence approach or long‑term monitoring would be valuable to more rigorously evaluate the effect of reclamation age on soil quality development in coastal reclaimed areas.

Comment 4:

The scoring functions and SQI calculation are insufficiently described and include mathematical inconsistencies.

Response 4: We thank the reviewer for this comment. The threshold values (L1, U1, U2, L2) for the parabolic membership functions (pH, bulk density, available Fe, Mn, Cu, and Zn) have been provided in Supplementary Table S10, with a brief description added to the revised Materials and Methods section (Lines 285-287). These thresholds were derived from published regional soil quality standards and literature.

Comment 5:

Several units and numerical values in Tables 1 and 2 appear incorrect or incomplete.

The reported units require careful verification.

For both tables, the authors should report:

whether values are means ± SD or means ± SE;

the sample size for each area;

the post hoc test used after ANOVA;

any transformation applied before statistical testing.

Response 5: We thank the reviewer for carefully checking the units and numerical values in Tables 1 and 2. The reviewer is correct: several units were incorrectly reported in the original manuscript, and we have now corrected them. Specifically, the following corrections have been made:

Total nitrogen (TN): The unit has been corrected from g kg⁻¹ to g 100g⁻¹ (i.e., %). The values (0.10–0.14 g 100g⁻¹) are now consistent with typical TN contents in agricultural soils and yield reasonable C:N ratios when compared with SOC (16–23 g kg⁻¹), giving C:N ratios of approximately 11–16, which are biologically plausible for paddy soils.

Available nitrogen (AN), available phosphorus (AP), available potassium (AK), and DTPA‑extractable Fe, Mn, Cu, and Zn: The unit for these variables has been corrected from g kg⁻¹ to mg kg⁻¹. For example, AP values (~20–60 mg kg⁻¹), AK (~100–800 mg kg⁻¹), and available Zn (~1–10 mg kg⁻¹) are now within the typical ranges reported for agricultural soils in China.

We have carefully verified all other units and numerical values in Tables 1 and 2 to ensure their correctness. The corrections do not affect any of the statistical analyses, SQI calculations, or conclusions, as all calculations were based on the original numerical data with their correct units. The revised tables have been updated in the manuscript (Table 1).

For both tables, we have added notes that values are means ± SD, and the sample size for each area (YQ: 9, LW: 18, LG: 9, RA: 13) in Lines 342-343 and 356.

For the post hoc test used after ANOVA, we fully agree that sample size imbalance among the four sites (YQ: 9, LW: 18, LG: 9, RA: 13) may affect the homogeneity of variances in ANOVA.

We applied Welch's ANOVA for indicators with significant Levene test results (p < 0.05). This method does not assume equal variances across groups and corrects the degrees of freedom using the Welch–Satterthwaite approximation, providing more reliable inference when sample sizes are unequal and variances are heterogeneous. In the revised manuscript, we have explicitly stated this procedure in the Materials and Methods section (Lines 315-319) and the Supplementary Information (Table S7 and Table S8). We have confirmed that the observed differences remain robust after this correction.

For the 12 indicators ultimately retained in the MDS (pH, SOM, TN, AK, TWS, AN, available Zn, bacterial quantity, bacterial diversity, fungal diversity, BG, and CB), the results show that most of the retained MDS indicators exhibit significant differences among the four sites (p < 0.05), with the exception of pH, TWS and BG. These findings support the robustness of our MDS selection and subsequent SQI comparisons. We have added a description of the Levene's test and Welch ANOVA procedures, along with the corresponding results, to the Materials and Methods (Lines 306-310).

Comment 6:

The qPCR and amplicon-sequencing methods are not sufficiently described for reproducibility.

The qPCR section lacks several essential details, including:

the exact primer sequences and final primer concentrations;

the amplification programme;

the type and preparation of the standard;

standard-curve efficiency and R²;

technical replication;

no-template and extraction controls;

assessment of PCR inhibition;

calculation of gene-copy abundance;

conversion to copies per gram of dry soil;

transformation of the data before statistical analysis.

They should also report:

raw and retained read numbers per sample;

sequencing depth and sample coverage;

the normalization or rarefaction procedure used before calculating Shannon diversity;

whether any samples were removed;

negative extraction and PCR controls;

whether sequencing was randomized across batches.

Response 6: We thank the reviewer for this important methodological comment. We have added more clearly descriptions about qPCR protocol in Material and Method (Lines 235-246) as” The amplified products of quantitative qPCR were recovered by agarose gel electrophoresis, ligated into the pEASY-T3 vector, and then transformed into Escherichia coli DH5α competent cells. Positive clones were screened on ampicillin-containing plates using blue-white colony screening, and selected positive clones were sequenced. The confirmed positive clones were cultivated for plasmid DNA extraction, and the concentration and purity (OD260/OD280) were determined using a microspectrophotometer. The recombinant plasmids were serially diluted to concentrations ranging from 10⁻² to 10⁻⁷ ng/μL and used as standards for qPCR targeting the ITS1/ITS2 regions and 16S rRNA genes. A single peak in the melting curve confirmed that the qPCR conditions met the required standards and that the primer specificity was satisfactory. The amplification efficiency for the soil samples was 96.6%, with a standard curve correlation coefficient greater than 0.995.”

Comment 7:

The statistical analyses require more complete reporting and validation.

The manuscript should also specify:

whether PCA was conducted on a correlation matrix or covariance matrix;

whether variables were centred and scaled before PCA;

whether skewed variables, particularly qPCR and enzyme data, were transformed;

how normality and homoscedasticity were assessed;

which multiple-comparison procedure followed ANOVA;

whether multiple-testing correction was applied to the correlation matrix.

Response 7: We thank the reviewer for this careful and constructive methodological review. We have addressed each point raised as follows:

(1) Number of measured indicators (Line 118)

We confirm that a total of 24 indicators were measured in this study, and the previous statement of "27 indicators" was a typographical error. We have corrected this in the revised manuscript (Line 118) to read "24 indicators" consistently throughout the text.

(2) Sample size and PCA stability (49 samples, 24 variables)

The reviewer raises a valid concern regarding the sample‑to‑variable ratio in PCA. We acknowledge that a ratio of approximately 2:1 (49:24) is below the often‑cited rule of thumb (e.g., 5–10 samples per variable). However, we would like to justify our approach as follows:

First, our PCA was not intended as an exploratory factor analysis for latent construct discovery, but rather as a data reduction tool to identify the most representative indicators within each principal component for MDS construction, a common application in soil quality assessment studies where the number of measured indicators often exceeds the sample size.

Second, we assessed the adequacy of our data for PCA using the Kaiser–Meyer–Olkin (KMO) test and Bartlett's test of sphericity. The KMO value was 0.6, which falls within the "mediocre" range according to Kaiser (1974). The highly significant Bartlett's test (p =0.000 < 0.001) indicated that the variables were sufficiently correlated to justify the application of PCA.

Third, the seven retained principal components (eigenvalues > 1) accounted for 74.54% of the total variance, indicating that the extracted components captured the majority of information in the dataset despite the moderate sample size.

While we agree that bootstrap or resampling procedures would further strengthen the stability assessment, we note that such methods are rarely applied in MDS‑based soil quality studies.

(3) PCA implementation details

PCA was performed on the correlation matrix, and all variables were centred and standardized (Z‑score normalization) prior to PCA to eliminate the influence of different units and scales. This is now explicitly stated in the revised Materials and Methods section (Lines 306-310) and Discussion section (Lines 534-536).

(4) Transformation of skewed variables

We examined the distributions of all variables prior to PCA. For the qPCR data (bacterial and fungal quantities) and enzyme activity data, we applied log10 transformation to improve normality and reduce skewness, as these variables typically exhibit right‑skewed distributions. The transformed data were used for all subsequent analyses (PCA, correlation analysis, ANOVA/Welch ANOVA). We have added a description of this transformation step in the revised Materials and Methods section (Lines 315-319).

(5) Assessment of homoscedasticity

Levene's test was performed for each indicator. The detailed Levene's test results and the subsequent Welch's ANOVA corrections for heteroscedastic variables are now provided in Supplementary Tables S7–S8, as noted by the reviewer. This procedure is now clearly described in the Materials and Methods (Lines 315-319).

(6) Multiple‑comparison procedure following ANOVA/Welch ANOVA

For indicators satisfying the homogeneity of variance assumption (Levene's test p ≥ 0.05), ANOVA results were retained. For indicators with significant heteroscedasticity (Levene's test p < 0.05), Tamhane's T2 post‑hoc test was used, which does not assume equal variances. The post‑hoc results are reported in the supplementary materials (Table S9). This information has been added to the revised manuscript (Lines 315-319).

(7) Multiple‑testing correction for correlation matrix

The correlation matrix presented in Figure 2 is intended as a descriptive visualization of pairwise relationships among the measured variables, rather than a hypothesis‑testing tool. Therefore, we did not apply a formal multiple‑testing correction (e.g., Bonferroni) to the correlation p‑values. To avoid over‑interpretation, we have focused our discussion only on correlations that were both statistically significant (p < 0.05) and biologically meaningful, and we have emphasized that these are descriptive associations rather than inferential tests. We have clarified this in the revised manuscript (Lines 544-546).

We believe these revisions and clarifications address all of the reviewer's methodological concerns and enhance the transparency and reproducibility of our statistical procedures.

Comment 8: