Twenty-Year Variability in Water Use Efficiency over the Farming–Pastoral Ecotone of Northern China: Driving Force and Resilience to Drought

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript addresses an interesting investigation of the characteristics of water use efficiency (WUE) variability in the farming-pastoral ecotone of Northern China (FPENC), over the past two decades, with the support of cloud computing platforms (GEE platform). WUE is adopted as a metric for ecosystem resilience.
The document is well written, and results are promising. 
Some review, however, are required, as stated below: 
1-    Some misspellings have been detected along the document. Yellow marks have been used to highlight missing words and misspellings, and also sentences requiring some review.
2-    Authors carried out a trend analysis for some variables using classical methods, but they have not discussed autocorrelation. At least for WUE, it would be relevant to assess the influence of autocorrelation in the results, using some trend decomposition methodology. Some details about the method can be found in:
ALMEIDA, T.A.B.; MONTENEGRO, A.A.A.; MACKAY, R.; MONTENEGRO, S.M.G.L.; COELHO, V.H.R.; DE CARVALHO, A.A.; DA SILVA, T.G.F. Hydrogeological trends in an alluvial valley in the Brazilian semiarid: Impacts of observed climate variables change and exploitation on groundwater availability and salinity. JOURNAL OF HYDROLOGY: REGIONAL STUDIES. v.53, p.101784, 2024.
3-    The main objective of the study is to investigate spatio-temporal variation of WUE across FPENC. Several maps have been developed. However, it would be useful to estimate and explore statistical metrics for variability, like standard deviations.

4-    CHIRPS monthly values have been adopted in the analysis. Such estimates are subject to high uncertainties. It would be useful to provide more details regarding CHIRPS accuracy for monthly rainfall in the region.

Comments for author File: Comments.pdf

Author Response

Comments:

1 Some misspellings have been detected along the document. Yellow marks have been used to highlight missing words and misspellings, and also sentences requiring some review.

Response: Thank you. We have rewritten these misspellings, missing words and sentences.

2 Authors carried out a trend analysis for some variables using classical methods, but they have not discussed autocorrelation. At least for WUE, it would be relevant to assess the influence of autocorrelation in the results, using some trend decomposition methodology. Some details about the method can be found in:

ALMEIDA, T.A.B.; MONTENEGRO, A.A.A.; MACKAY, R.; MONTENEGRO, S.M.G.L.; COELHO, V.H.R.; DE CARVALHO, A.A.; DA SILVA, T.G.F. Hydrogeological trends in an alluvial valley in the Brazilian semiarid: Impacts of observed climate variables change and exploitation on groundwater availability and salinity. JOURNAL OF HYDROLOGY: REGIONAL STUDIES. v.53, p.101784, 2024.

Response: Thank you. We have carefully studied the LOESS method and considered it in analysing the trend of monthly WUE in different vegetation types and the monthly mean WUE based on the pixel scale in the FPENC. Please see Fig. 6 (a, b). We also give more explanation in the methodology part and cited this reference by Almeida et al. (2024). Please see Part 2.3.1, L212-216

“The Mann-Kendall test assumes that the data are serially independent. However, the times series of WUE may have autocorrelation, which might cause errors for the trend test, increasing the significant of data [43]. Therefore, the LOESS (STL) method was adopted to eliminate the influence of autocorrelation in seasonal and trend decomposition of WUE in various vegetation types.”

3 The main objective of the study is to investigate spatio-temporal variation of WUE across FPENC. Several maps have been developed. However, it would be useful to estimate and explore statistical metrics for variability, like standard deviations.

Response: we have further explored the metrics of WUE and various biophysical factors. For example, we added a boxplot of mean values of monthly WUE as Fig. 5. Besides, we also calculated the annual mean values and standard deviatons of the biophysical factors as Fig. 8.

4 CHIRPS monthly values have been adopted in the analysis. Such estimates are subject to high uncertainties. It would be useful to provide more details regarding CHIRPS accuracy for monthly rainfall in the region.

Response: we have compared the montly CHIRPS rainfall values with the obeserved rainfall data which comes from 33 meteorological stations located in the study area for the year 2022. The resultes were good with R-square is 0.84, and RMSE is 24.24 mm. Please see Fig. S7. We have descrpited the verification in part 4.1. Please see L515-518.

“As for our study, we compared the monthly PPT data from CHIRPS with observed data selected from 33 meteorological stations located in the FPENC for the year 2022, and the results shows good with R2 was 0.84 and RMSE was 24.24 mm (Fig. S7).”

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript titled Twenty-year variability in water use efficiency over the farming-pastoral ecotone of Northern China: driving force and resilience to drought can be accepted after Major Revision

 

 

The section Abstract

 

This section should be expanded with at least one additional sentence. Moreover, the most important findings should be more clearly emphasized to better showcase the study’s contributions.

 

The authors are also advised to include at least one additional keyword in the "Keywords" section.

 

 

Introduction

 

 

Can the authors better explain the strategies of mitigations?

 

How the crops are connected with the ecosystems? Explain better.

 

How was the spatial resolution of the satellite images (MODIS)? It is important to know?

 

 

 

Recommendation

 

Considering the above observations, I strongly recommend that the authors consult and cite a relevant study that discusses the role and analyses of crops with satellite recordings and numerical analyses.  In this paper (research) the water use for crops also analyzed.

 

 

The recommended reference is

 

- Valjarević, A., Morar, C., Brasanac-Bosanac, L., Cirkovic-Mitrovic, T., Djekic, T., Mihajlović, M., ... & Kaplan, G. (2025). Sustainable land use in Moldova: GIS & remote sensing of forests and crops. Land Use Policy, 152, 107515. https://doi.org/10.1016/j.landusepol.2025.107515.

 

Materials and Methods

 

Fig.1 Well done map

 

 

 

The authors need to add more explanations about climate of the analyzed area.

 

How the author did this Ecosystem resilience equation, tell me more about it formula?

 

Results

 

This section can be divided on numerical and statistical

 

 

 

Strengths of the Manuscript

 

Long-term Analysis (2003–2022):

The study provides a comprehensive two-decade temporal assessment of water use efficiency (WUE) using high-resolution MODIS data.

 

Integration of Remote Sensing and Biophysical Data:

The combination of GPP and ET from MODIS with climate and vegetation indices (e.g., VPD, LAI, EVI) enables robust spatial-temporal insights.

 

Resilience Assessment Framework:

The paper introduces a ratio-based ecosystem resilience index (Rd = WUEd/WUEm), offering a quantitative approach to drought impact analysis.

 

Vegetation-Type Specific Analysis:

It disaggregates results by 10 vegetation types, revealing unique WUE patterns and resilience levels for forests, grasslands, croplands, etc.

 

Use of Google Earth Engine (GEE):

Application of GEE allows efficient handling of large spatio-temporal datasets, showcasing reproducibility and scalability.

 

Climatic Control Insights Over Two Periods (2003–2012 vs. 2013–2022):

Differentiating dominant WUE drivers over early and later decades provides nuanced understanding of climate-WUE dynamics.

 

Application of Random Forest Algorithm:

The RF model helped identify the relative importance of seven environmental variables, improving variable interpretation beyond linear methods.

 

Policy-Relevant Insights on Ecological Fragility:

The findings underline the dominance of non-resilient areas (93.7%), especially in grassland zones, guiding regional land and water policy responses.

 

 

Limitations of the Manuscript That Can Be Resolved

 

Low Ecosystem Coverage Validation:

Validation of MODIS WUE was limited to a small number of EC sites (only 7), reducing confidence across diverse landscapes.

 

No Consideration of CO₂ Effects on WUE:

Although CO₂ fertilization is known to influence WUE, it was omitted, limiting completeness in the drivers of ecosystem change.

 

Uncertainty in MODIS Products:

Despite general acceptance, MODIS GPP and ET products still carry modeling assumptions and uncertainties, especially in semi-arid zones.

 

Exclusion of Other Drought Metrics:

Drought assessment relied solely on annual precipitation minima, ignoring indices like SPEI or soil moisture anomalies that capture multi-dimensional droughts.

 

Resolution and Aggregation Trade-offs:

Spatial resolution of 500m and use of 8-day composites may obscure fine-scale variations in WUE, especially for heterogeneous land covers.

 

Simplified Resilience Index:

The binary resilience definition (Rd ≥1) lacks nuance for evaluating systems with fluctuating but stable WUE under stress.

 

Underrepresentation of Human Activity Impact:

Land use change and anthropogenic water use (e.g., irrigation) were not explicitly modeled, although highly relevant for cropland regions.

 

No Consideration of Drought Lag Effects:

The potential delayed effects of drought on WUE are acknowledged but not quantified, leaving a gap in ecosystem response modeling.

 

 

The Conclusion section should be further elaborated to enhance the overall clarity and impact of the study's findings

Overall Recommendation

The paper can be accepted after a Major revision to address the specified points.

Wishing the author, the best of luck.

Reviewer #4

Author Response

Summary：Thank you very much for your approval of this manuscript. We are truely grateful for your valuable comments, which have played an important role in improving the quality of our manuscript. In the following, we provide response to each comment point to point. As for the limitations the reviewer recommended, we have tried our best to revise during this revision.

 

Comments

1. The section Abstract

This section should be expanded with at least one additional sentence. Moreover, the most important findings should be more clearly emphasized to better showcase the study’s contributions.

Response: We have added sentences related to the ecological restoration. Please see L20-22.

“The ecological restoration projects in China have mitigated land degradation and maintain the sustainability of dryland. However, the process of greening in drylands has the potential to impact water availability.”

Besides, we have deleted some sentences to make the abstract more concise. The deleted sentences: However, few studies have examined the spatio-temporal patterns of WUE and its driving force over the past twenty years during the vegetation restoration in the FPENC. Besides, the ecosystem resilience of various vegetation types predisposed to drought has not been fully explored in this region.

We have also revised some findings to make it more clearly, such as “59.2% of FPENC showed non-resilient, as grassland occupy the maojoarity of the area, locating in Mu Us Sandy land and Horqin Sand Land.”

The authors are also advised to include at least one additional keyword in the "Keywords" section.

Response: We have addes one keyword, it is “drought”

2. Introduction

Can the authors better explain the strategies of mitigations?

Response: We have added some sentences to better explain the strateries of mitigations. Please see L54-61.

“To address this global crisis, a multiple-faceted measures are required, combing tech-nological innovations, policy interventions, behavioral changes, and nature-based so-lutions. Terrestrial ecosystems adapt by altering their structural composition and mod-ifying their sensitivity to climatic variations. Reforestation and afforestation can enhance the sequestration of CO2 and regulate water cycles. Besides, integrating trees with crops/livestock can enhance carbon storage and reduce soil erosion. Furthermore, precision farming, reduced tillage, and optimized fertilizer can be used to cut methane (CH4) and nitrous oxide (N2O) emissions.”

 

3. How the crops are connected with the ecosystems? Explain better.

Response: We have added some sentences to better explain the relationship between crops and the ecosystems.

Please see L50-54.

“Cropland is one of the most important terrestrial ecosystems, and it is both victims of climate change and part of the solution. Climate change can exerbate the soil evaporation, change the growing duration, and impact the yielding. Besides, the extreme weather events such as droughts, floods and heatwaves lead to crop failures and exacerbate soil erosion.”

4. How was the spatial resolution of the satellite images (MODIS)? It is important to know?

Response: The spatial resolution of MODIS GPP and ET is 500m, Please see L116. Besides, the spatial resolution of MODIS Land Cover Type MCD12Q1 and MODIS LAI (MOD15A2H) are 500m. For the details of the data information used in our study, please refer to the Table 1.

 

5. Recommendation

Considering the above observations, I strongly recommend that the authors consult and cite a relevant study that discusses the role and analyses of crops with satellite recordings and numerical analyses.  In this paper (research) the water use for crops also analyzed.

Response: Thank you very much. We have consulted this article and cited it when discussing the influence of climate change and the cloud computing platforms like Google Earth Engine, Please see L51-54, L98-100.

Climate change can exerbate the soil evaporation, change the growing duration, and impact the yielding. Besides, the extreme weather events such as droughts, floods and heatwaves lead to crop failures and exacerbate soil erosion [1].

Additionally, cloud computing platforms like Google Earth Engine (GEE) facilitate ac-cess to long-term datasets, enabling studies on WUE variability at large [1,29].

The recommended reference is

- Valjarević, A., Morar, C., Brasanac-Bosanac, L., Cirkovic-Mitrovic, T., Djekic, T., Mihajlović, M., ... & Kaplan, G. (2025). Sustainable land use in Moldova: GIS & remote sensing of forests and crops. Land Use Policy, 152, 107515. https://doi.org/10.1016/j.landusepol.2025.107515.

 

 

Materials and Methods

6. 1 Well done map

Response: Thank you

7. The authors need to add more explanations about climate of the analyzed area.

Response: we have add more explanations about the climate of the FPENC area. Please see L142-148.

“Over 70% of annual rainfall occurs between June and September (summer months often experience heave rainstorms, while spring droughts are frequent). ≥10 ℃ accumulated temperatures is 2000-3200 ℃per day, sufficient for one crop per year (e.g., corn, pota-toes). Winter extreme lows temperature of -20 to 30 ℃(central Inner Mongolia can drop below -30℃). Frost-free period was 120-180 days (longer in the east, shorter in the west). The aridity index (annual potential evaporation/precipitation) is 1.5-3.0. Annual evapotranspiration is 1800-2500 m, which is far exceeding precipiation, exacerbating drought risks.”

8. How the author did this Ecosystem resilience equation, tell me more about it formula?

Response: When hydro-climatic conditions change abruptly (such as from a dry to humid year or from humid to a dry year), the ability of the ecosytem to maitain the same structure and functions is called the ecosystem resilience. In our study, ecosystem resilience was defined as the ratio of WUE during drought years (WUEd) to the average annual WUE. We use the averaged annual WUE in the years with SPEI<-1.5 per pixel as the WUEd to analyze the most obvious influence of drought on the ecosystem resilience. We classified the ecosystem resilience based on the standard (Jia et al., 2023): resiliend as Rd ≥1; slightly non-resilient: 0.9< Rd <1; moderately non-resilient:0.8≤ Rd ≤0.9; and severely non-resilient: Rd <0.8.

For more information, please see part 2.3.3.

 

Jia, B.H., Luo, X., Wang, L.H. and Lai, X., 2023. Changes in Water Use Efficiency Caused by Climate Change, CO2 Fertilization, and Land Use Changes on the Tibetan Plateau. Advances in Atmospheric Sciences, 40(1): 144-154.

 

 

Results

9. This section can be divided on numerical and statistical

Response: We have carefully reviewed our manuscript, and we divided each section of the Results Part into numerical and statistical parts.

Details are as follows:

3.1 Trends in WUE

We moved the statistal contents as a single paragraph at the end of the first paragraph. Please see L283-287

“To evaluate the reliability of the numerical results, statistical tests were performed. The annual averaged WUE of the whole FPENC has exhibited a notable decreasing trend over the twenty years (R2=0.29, p<0.01), Specifically it demonstrated a significant de-cline during the first ten years (R2=0.39, p<0.05), while in the last ten years, it showed a slight upward trend (2013-2022) (R2=0.16, p=0.14) (Fig. S2, S3).”

We moved the WUE trend of ten vegetation types at L298-304.

“The annual WUE all showed a sharp decrease in 2010, and it showed an increase trend afterwards (Fig. 6a, b). In terms of ten vegetation types, the annual WUE of evergreen needle forest exhibited the highest value (1.87±0.12 gC H2O-1) and with the WUE of mixed forest rank second afterwards (1.61±0.1 gC H2O-1), and the open shrubland showed the lowest WUE (1.08±0.14gC H2O-1). Especially, the grassland, which occupied the majority area in the FPENC, had a relative low value of WUE (1.16±0.11gC H2O-1) (Fig. 6a, Table 2).”

 

3.2 Trends in biophysical variables
We moved the statistal results (mainly in Table 2) as a singel paragraph below the first paragraph. Please see L344-350.

“The results in Table 2 also showed the same tendecy with Fig.10, especially the EVI, LAI, VPD and VSWC in different vegetation types. The lowest values of EVI and LAI were occurred in open shrubland, followed by grassland and the cropland. The PPT in closed shrubland and mixed forest, with annual means being 556.96 mm and 529.65 mm, respectively. PPT was lowest in open shrubland with average annual total of 407.37 mm. The distribution trends of VSWS and precipitation are consistent across different vegetation types (Table. 2).”

 

 

Limitations of the Manuscript That Can Be Resolved

1. Low Ecosystem Coverage Validation:

Validation of MODIS WUE was limited to a small number of EC sites (only 7), reducing confidence across diverse landscapes.

Response: We have added another four EC sites, including mixed shrubland, two grasslands, and one meadow. The R2 is 0.83, and the RMSE is 0.27gC kg-1H2O-1, and the bias is 0.20gC kg-1H2O-1. These results can further verify the precision of the MODIS WUE. Please see part 4.1. Please see L499-502

“Besides, we have added another four EC sites located in the FPENC (Liu et al., 2019; Dong et al., 2021; Liu et al., 2025). The EC measured the WUE was higher than the MODIS based the WUE with the root mean square error (RMSE) was 0.27 gC kg-1H2O-1 the R-square (R2) was 0.83, and the relative error (bias) was 0.20 gC kg-1H2O-1 (Fig. S6).”

 

2. No Consideration of CO₂ Effects on WUE:

Although CO₂ fertilization is known to influence WUE, it was omitted, limiting completeness in the drivers of ecosystem change.

 Response: Yes, we haven’t considered the CO2 effects on WUE, but we have discussed the effects of CO2, please see L636-642. We would consider the influence of CO2 in the future research work.

“Additionally, our study did not analyze the influencing of CO2 on WUE and Rd. Pre-vious study has indicated that CO2 fertilization can enhance ecosystem resilience, par-ticularly on the Tibetan plateau. Study also showed the CO2 can positively impact WUE over the long term.  Although our study observed a decreasing trend in WUE, this may suggest CO2 had minimal influence, However, a more detail analysis about the impact of CO2 on WUE is warranted in the future investigations. Besides, human activity impact can also be explored in the near future.” 

3. Uncertainty in MODIS Products:

Despite general acceptance, MODIS GPP and ET products still carry modeling assumptions and uncertainties, especially in semi-arid zones.

Response: Previous study has proved that the MODIS GPP would underestimate compared with the observation value. We have compared the WUE based on the MODIS GPP and ET with the WUE calculated from EC data. And the R2 was 0.83, with the RMSE was 0.27 gC kg-1H2O-1 and the relative error (bias) was 0.20 gC kg-1H2O-1. These results validate the precision of MODIS GPP and ET in calculating the WUE. Though the results showed the MODIS WUE was lower than the EC-observed values, the verification results showed that our research can capture the long-term trend and spatio-temporal variation of WUE, thus it would be valuable to guidance the vegetation afforestation in FPENC.  Please see L493-507.

“However, the method has been used previously, and proved to be highly accuracy. The MODIS GPP have proved to be underestimation in arid areas, thus may lead underestimation of WUE, therefore, we tested the MODIS based WUE with other data source for further verification. We compared the annual average WUE values from the MODIS data and that EC dataset, and the sites selection was referred to the Bai et al. (2020), with one shrubland and six grasslands selected (Bai et al., 2020). Besides, we have added another four EC sites located in the FPENC. The EC measured the WUE was higher than the MODIS based the WUE with the root mean square error (RMSE) was 0.27 gC kg-1H2O-1 the R-square (R2) was 0.836, and the relative error (bias) was 0.20 gC kg-1H2O-1 (Fig. S6). Though the results showed the MODIS WUE was lower than the EC-observed values, the verification results showed that our research can capture the long-term trend and spatio-temporal variation value of WUE, thus it would be valuable to guide the vegetation afforestation in FPENC and therefore, it is acceptable. With the fast development of remote sensing product, higher spa-tio-temporal resolution of WUE may be used to do further analysis in FPENC.”

 

4. Exclusion of Other Drought Metrics:

Drought assessment relied solely on annual precipitation minima, ignoring indices like SPEI or soil moisture anomalies that capture multi-dimensional droughts.

 Response: We have considered the other drought metric-SPEI, as this metric can monitor the assessment of dry and wet conditions under climate change, with significant advantages in areas with sinificant temperature changes, making it suitable for climate change research.

Please see Part 2.3.3.

“Ecosystem resilience refers to the capacity of an ecosystem to preserve its structure and functions despite abrupt changes in hydro-climate conditions, such as transitions from a dry to wet years or vice versa. In this study, we define ecosystem resilience (Rd) as the ratio of avereged annual WUE during drought years (WUEd, determined by identifying the year with the heave drought (SPEI<-1.5) per pixel) to the average annual WUE calculated from 2003 to 2022 (WUEm). The spatioal resolution of SPEI with 0.5 °was converted to 500m based on the bilinear interpolation. An ecosystem is deemed resilient if it can maintain or enhance WUE, thereby supporting productivity in water-limited conditions during droughts; specifically, this occurs when WUEd is at least equal to WUEm, yielding an Rd value of 1 or higher. Thus, a greater Rd value (equal to or exceeding 1) signifies a resilient ecosystem, 0.9<Rd<1 indicates that the ecosystem was slightly non-resilient, 0.8≤ Rd ≤0.9 indicates moderately non-resilient, however, when the Rd <0.8, it indicates severely non-resilient.”

5. Resolution and Aggregation Trade-offs:

Spatial resolution of 500m and use of 8-day composites may obscure fine-scale variations in WUE, especially for heterogeneous land covers.

Response: We agree. Although the spatial-temporal resolution of the WUE based on MODIS GPP and ET may lead some uncertainty, however, MODIS GPP and ET products are widely recognized as reliable global datasets for estimating terrestrial carbon uptake and water flux. With the development of remote sensing technologies, high resolution products of WUE may be used in the FPENC in the future.

6. Simplified Resilience Index:

The binary resilience definition (Rd ≥1) lacks nuance for evaluating systems with fluctuating but stable WUE under stress.

Response: Currently, the internationally widely recognized concept of ecosystem resilience is defined as: the capacity of an ecosystem to return to a stable state after disturbance, including the ability to maintain its key characteristics, such as species composition, structure, ecosystem functions, and process rates. Depending on the interpretation of “stable state” existing definitions of resilience can be categorized into two perspectives: engineering resilience and ecological resilience. Engineering resilience is based on the assumption of a single stable state, positing that the system has only one “optimal” equilibrium steady state. When the system deviates into other non-stable states, measures should be taken to restore it to its balanced and stable condition. Ecological resilience is based on the assumption of multiple stable states. Rather than focusing on the time or ability to return to a single stable state, it emphasizes transitions between different stable states. Since the 1970s, while scholars in related fields have been exploring the concept of ecosystem resilience, some researchers have also begun to focus on the attributes and characteristics of resilience. Carpenter et al. argued that assessing ecosystem resilience must explicitly specify system configurations and perturbations of interest.

Due to the multi-stability mechanisms of ecosystems, any external disturbance may trigger an abrupt regime shift, pushing the system into an undesirable state from a management perspective. Quantitative assessment is therefore essential to identify the key drivers of resilience and provide a scientific basis for ecosystem management. Currently, scholars have primarily employed the following methods to study ecosystem resilience: threshold approaches, proxy indicator methods, and experimental approaches.

The influence factors of ecosystem resilience contains biological diversity, ecological memory, habitat conditions, climate, human activity and productivity.

Traditional comprehensive evaluation methods can integrate raw data from multiple assessment indicators through statistical approaches to construct a composite index, thereby providing a holistic and systematic assessment of ecosystem resilience.

Therefore, the contents above will guide our future work related with ecological resilience, maybe we will use the traditional comprehensive evaluation methods to assess the ecological resilience in FPENC.

 

7. Underrepresentation of Human Activity Impact:

Land use change and anthropogenic water use (e.g., irrigation) were not explicitly modeled, although highly relevant for cropland regions.

Response: Thanks for the suggestion. We didn’t consider the human activity impact on the WUE in the FPENC, maybe we will consider the land use change and anthropogenic water use in the future research.

 

8. No Consideration of Drought Lag Effects:

The potential delayed effects of drought on WUE are acknowledged but not quantified, leaving a gap in ecosystem response modeling.

Response: We didn’t analyze the dought lag effects, but we have discussed in the discussion part. We may consider the drought lag effects in our future work related to drought.

 

The Conclusion section should be further elaborated to enhance the overall clarity and impact of the study's findings.

Response: We have revised the Conclusion section. Please see Conclusion section.

Overall Recommendation

The paper can be accepted after a Major revision to address the specified points.

Response: We sincerely appreciate your valuable suggestions.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The article, "Twenty-year variability in water use efficiency over the farming-pastoral ecotone of Northern China: driving force and resilience to drought," presents a methodological development for determining spatiotemporal Water Use Efficiency (WUE) involving the macro variables that are determinants of this process under a climate change scenario.
The authors have developed this methodological approach, which can be replicated in other climatic areas, considering valid information sources and appropriate statistical methodology to determine the weight of the variables considered, with the support of Random Forest and regression analysis.

Author Response

Response to Reviewer #3

The article, "Twenty-year variability in water use efficiency over the farming-pastoral ecotone of Northern China: driving force and resilience to drought," presents a methodological development for determining spatiotemporal Water Use Efficiency (WUE) involving the macro variables that are determinants of this process under a climate change scenario.
The authors have developed this methodological approach, which can be replicated in other climatic areas, considering valid information sources and appropriate statistical methodology to determine the weight of the variables considered, with the support of Random Forest and regression analysis.

Response: Thank you very much for your positive comments. We have revised the citation format in the manuscript. Besides, We have added and revised some discussions, including the limitations of the study and potential future breakthroughs.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript titled Twenty-year variability in water use efficiency over the farm- ing-pastoral ecotone of Northern China: driving force and resilience to drought can be accepted in form that are now presented

 

1. Strong scientific methodology and long-term spatial analysis

The paper integrates multi-source remote sensing data (MODIS GPP and ET, ERA5, CHIRPS, TerraClimate) over a 20-year period (2003–2022), combined with robust statistical techniques like Theil-Sen’s slope, Mann-Kendall test, random forest analysis, and multiple regression. This long-term, multi-scale analysis offers a valuable and technically sound assessment of water use efficiency (WUE) trends, drivers, and ecosystem resilience across a fragile ecological zone.

 

2. Novel insights into ecosystem resilience under climate stress

The authors provide a novel quantification of ecosystem resilience to drought using a ratio-based indicator (Rd = WUEd/WUEm). The identification of resilience levels across 10 vegetation types and the spatial patterns (e.g., grasslands being mostly non-resilient) contribute meaningful ecological insights for land management and adaptation planning under climate change.

 

3. Relevance to global and regional sustainability and climate adaptation goals

The study addresses critical issues at the intersection of climate change, ecological vulnerability, and sustainable land management especially in a transitional farming-pastoral ecotone. The findings are valuable not only for China but also as a transferable framework for assessing ecosystem responses in other semi-arid or drought-prone regions globally.

The paper now is total suitable for acceptance

The main reason for that are

The section Abstract now is much better written and now is acceptable

 

The Discussion section is acceptable and very concise

I have accepted the manuscript in full

Sincerely,

Reviewer#1