Climate Change’s Influence on Dairy Farming in Punjab, Pakistan: Effects on Milk Production, Farmers’ Views, and Future Adaptation Strategies

Dear Reviewer, thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the revisions/corrections highlighted/in track changes in the resubmitted files.

Comments 1: [Unexplained Variation in Milk Production Across Farms.]

Response 1: We acknowledge the observed variability in milk production across farms. However, this variation primarily stems from differences in herd size rather than per-animal productivity or climate adaptation strategies. Most farmers with higher yields owned a larger number of milking animals, not necessarily more productive ones. We have clarified this point in the revised descriptive statistics section to avoid misinterpretation in section 4.1 and lines from 370 to 372.

Comments 2& 3: [Did the authors check if some farms already started adaptation strategies? And did they try to analyze this statistically? This kind of information would help explain the results better and show real-life examples for other farms.]

Response 2&3: Thank you for pointing this out, as most of the respondents in our sample were smallholder farmers with limited access to technological and infrastructural resources. During the interviews, very few reported employing formal adaptation strategies such as cooling systems or feed optimization technologies. This lack of widespread adaptation among the sampled farmers is one of the key motivations for conducting this research. Nevertheless, we have now clarified this point in the perception analysis section to strengthen the rationale for policy-focused adaptation interventions. We added a brief note in section 4.3 and lines 428 to 430

Comments 4: [Discussion Section Needs More Depth.]

Response 4: We appreciate the reviewer’s insightful feedback. In response, we have strengthened the discussion section by incorporating comparative insights from recent international studies that report similar heat stress impacts on dairy production in regions like the Mediterranean, Korea, and Tunisia. These additions provide a broader context for interpreting our findings. Furthermore, we added a paragraph elaborating on the policy relevance of our projections, emphasizing the need for region-specific interventions such as climate-resilient infrastructure, cooling technologies, and targeted subsidies for smallholder farmers. These revisions enhance the policy implications and global alignment of our study in section 5, lines 686 to 688 and 716 to 720

Comments 5: [The discussion lacks sufficient policy insight. Are the recommendations realistic and feasible for implementation in Pakistan?]

Response 5:  We thank the reviewer for this insightful comment. To address the concern, we revised the discussion section by explicitly evaluating the feasibility and realism of the proposed adaptation measures. We anchored our recommendations in the institutional and socio-economic context of Pakistan, referencing national policy frameworks in discussion section and also adoption policies are explained in section 5.1 Implications for Dairy Farming & Adaptation Strategies. Hope so we could satisfy the reviewers.

 These additions ensure the proposed actions are context-sensitive, actionable, and aligned with ongoing efforts to enhance climate resilience in smallholder dairy systems

Comments 6: [Did the authors check whether some farms already adopted climate adaptation strategies? Differences in production might reflect management practices, not just climate exposure."]

Response 6:  We appreciate the reviewer’s valuable observation. Our fieldwork already revealed that most participating farmers were smallholders with limited financial capacity and access to climate adaptation technologies (e.g., cooling infrastructure or heat-tolerant breeds). This lack of widespread adaptation practices is one of the key motivations behind our study. While the primary focus was not to compare adaptation levels across farms statistically, our structured perception survey did gather insights into existing practices. These responses were qualitatively integrated into the analysis and are reflected in Sections 4.3 and 5.1. To further clarify this point, we have added a statement highlighting the generally low prevalence of formal adaptation strategies among smallholder farmers and the resulting need for policy-driven interventions

Comments 7: [Even though there are many results shown, the discussion part could be more detailed and better connected to international research. Right now, comparisons with other studies are mentioned only briefly. Also, there is no deeper discussion about what the results mean for climate policy in Pakistan – are the recommendations realistic? Are they possible to implement?”.]

Response 7: We thank the reviewer for highlighting the need to deepen the discussion and strengthen connections with international literature and policy relevance. In response, we expanded Section 5 (Discussion) to include comparative insights from studies in Tunisia, Korea, Italy, and Ethiopia, highlighting parallel patterns of climate-induced yield reduction and adaptation strategies. We also revised Section 5.1 to critically evaluate the feasibility of our proposed adaptation measures in the Pakistani context. Specific attention was given to institutional barriers, financial access, and implementation capacity. These additions strengthen the link between empirical findings, global literature, and actionable policy recommendations.

Dear Reviewer, thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the revisions/corrections highlighted/in track changes in the resubmitted files.

Comments 1: [The research gap is not clearly defined. The introduction focuses only on the Punjab region, while the Abstract claims broader contributions (e.g., quantifying impacts and guiding policy), which are not well reflected in the gap statement.]

Response 1: We appreciate the reviewer’s insightful suggestion. In response, we have revised the concluding paragraph of the Introduction to more clearly articulate the research gap and the innovative integration of perception-based analysis with econometric forecasting. Specifically, we now emphasize how the study addresses the underexplored alignment between smallholder perceptions and empirical production trends in Punjab. This mixed-methods design enhances explanatory depth and forecasting capacity, as now reflected in the updated Introduction. “[updated introduction portion in the manuscript lines 76,77 ]”

Comments 2 &3: [suggestions from reviewers is methodological framework, particularly the mixed-method design combining perception surveys and regression models (FE, RE, interaction, GMM), is detailed in the Introduction. This section could be shortened and moved to the Methodology section to enhance structural clarity.]

Response 2&3: Thank you for the suggestion. We agree that methodological details belong primarily to the methodology section and have ensured full elaboration. However, we respectfully retained a concise summary in the Introduction to establish coherence between the study’s research gap, objectives, and analytical design. This is aligned with standard academic practice, especially for interdisciplinary research, to help readers understand how the mixed methods approach directly addresses the research problem. Additionally, as some reviewers recommended condensing the Methodology section due to length, this brief overview in the Introduction ensures early clarity without repetition. Nonetheless, we have carefully reviewed the Introduction to ensure it remains succinct and does not overly elaborate on procedures.

Dear Reviewer, thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted file.

Comments 1& 2 combined: [ [1] The justification for selecting 200 out of 450 farmers for quantitative analysis needs elaboration. Were these farmers randomly chosen, or were they selected based on data completeness?

[2] Clarify how milk production logs were maintained (self-reported vs. monitored) and whether any validation steps were taken to ensure accuracy.]

Response 1&2: We thank the reviewer for these insightful comments. From the total sample of 450 dairy farmers, 200 were selected for quantitative analysis based on the completeness and reliability of their milk production records spanning 2017–2024. These farmers maintained detailed daily production logs as part of regular farm routines. To ensure data accuracy, their logs were cross-validated against records maintained by local middlemen, who purchase the milk and keep accurate accounts for payment purposes. This triangulated verification process ensured a high-quality dataset for panel regression and forecasting analyses. These clarifications have now been briefly and concisely incorporated into Section 2.4 of the revised manuscript to maintain overall brevity in the methods section, as advised. “[updated text in the manuscript from line 180 to onward 188]”

Comments 3: [Briefly explain why GMM was preferred over other time-series methods (e.g., ARIMA) for forecasting.]

Response 3: We appreciate the reviewer’s request for clarification. The dynamic panel GMM model was chosen over univariate time-series approaches such as ARIMA due to the panel structure of the data, which includes both farm-level and time-series dimensions. GMM is particularly well-suited for modeling lagged effects, unobserved heterogeneity, and potential endogeneity, making it more appropriate for this study than traditional time-series techniques. A concise justification has been added to Section 3.5 of the revised manuscript.

“[updated text in the manuscript accordingly from line 304 to 310]”

Comments 4: [The regional analysis (Figure 3) lacks geographic labels (e.g., district names). Specify which regions (1–4) correspond to which districts.]

Response 4: We thank the reviewer for highlighting this oversight. In the revised manuscript, we have added a clear explanation of the regional grouping used in Figure 3. Specifically, Region 1 = Faisalabad, Region 2 = Jhang, Region 3 = Toba Tek Singh, and Region 4 = Chiniot. This clarification has been added to both the main text (Section 4.4) and the figure caption to enhance readability. “[add in the manuscript in line 428, 429 and 437 and 438]”

Comments 5: [The positive coefficient for humidity² (Model 3) suggests a plateau/reversal effect. Discuss potential physiological or management reasons for this nonlinearity.]

Response 5: The authors acknowledge this valuable observation. The positive coefficient for the squared humidity term in Model 3 suggests a nonlinear response, whereby initial increases in humidity negatively affect milk production, but the marginal impact diminishes at higher levels. This effect is likely attributable to a combination of thermoregulatory adaptation mechanisms in dairy cattle and on-farm mitigation strategies, such as misting systems, shaded housing, and adjusted feeding schedules. To address this, a scientifically grounded explanation has been added to Section 4.5 of the revised manuscript to interpret this curvature more rigorously. [page number 491 to 493]

Comments 6: [Expand on how perceptions were validated against empirical data. For example, did farmers in hotter regions report greater concerns, aligning with steeper declines in yield?]

Response 6: The author appreciates this insightful suggestion. To address this, a linking explanation has been added at the end of Section 4.3. The perception survey data were validated by the regional regression analysis in Section 4.4, where it was observed that Region 1 (Faisalabad) which exhibited the strongest negative climate effects on milk production was also where farmers reported the highest perceived impact of temperature and humidity. This coherence supports the internal validity of the farmer-reported data and its relevance to regional climate vulnerability. [As revised manuscript and these changes can be found on – page number, 11 and 12, and line 425 to 429 and 460 to 462 respectively.]

Comments 7: [The climate scenarios could be contextualized within IPCC RCP pathways for broader relevance.]

Response 7: We sincerely appreciate this valuable suggestion. In response, the manuscript has been updated to clarify that the applied climate scenarios (+2°C and +10% relative humidity) are aligned with IPCC AR6 projections for the South Asian region. Specifically, they are consistent with Shared Socioeconomic Pathways (SSPs) such as SSP2-4.5 (stabilization scenario) and SSP3-7.0 (regional rivalry). While not directly modeled as formal RCP or SSP trajectories, these scenario deltas serve as realistic, mid-range stress simulations to assess potential climate-induced risks in the dairy sector. A brief explanation has been added to Section 4.6 to strengthen policy relevance and improve interpretability. Updated manuscript with these changes it can be found – page number 16, and lines 569 to 575.]

Comments 8: [Simplify terms like "thermoneutral zone” or "heteroscedasticity" for interdisciplinary readers.]

Response 8:  We appreciate the reviewer for highlighting the importance of improving accessibility for interdisciplinary readers. In response, we have revised technical terms such as “thermoneutral zone” and “heteroskedasticity” by including brief clarifications upon their first mention in the manuscript. These adjustments enhance readability without compromising scientific accuracy and ensure the manuscript can be better understood by readers from diverse backgrounds, including those in agricultural economics, climate policy, and environmental sciences. “[revised text in the manuscript line 627 to 630]”

Comments 9: [Some sections (e.g., methodology) are overly detailed; condense where possible (e.g., Hausman test results).]

Response 9:  We agreed with the reviewer for pointing out the need to streamline methodological sections for improved readability. Accordingly, we have condensed the discussion of the Hausman test in Section 3.1 to retain only the essential statistical result and interpretation. Non-critical procedural details have been removed, ensuring that the section remains focused, concise, and accessible without losing analytical rigor. Similar minor refinements have been made in other methodological subsections to enhance clarity and flow [condensed the part on – page number6, and from lines 253 to 255.]

Comments 10: [In the Abstract, mention the sample size (450 farmers) and time frame (2017–2024) upfront for context.]

Response 10: We appreciate the reviewer’s thoughtful observation. While we recognize the importance of stating the sample size and study duration clearly, we have retained the current structure of the Abstract to maintain a smooth narrative progression from problem context to methodological approach and key findings. Introducing numerical specifics in the very first sentence was avoided to preserve a more general and engaging lead-in. Nevertheless, both the sample size (450 farmers) and time span (2017–2024) are explicitly included early in the abstract and detailed in the Methods section, ensuring transparency and completeness of study design.

Comments 11: [Update citations to reflect the latest IPCC reports (e.g., AR6 instead of AR5.]

Response 11:  We thank the reviewer for pointing out the need to ensure the most recent climate science is reflected in our manuscript. In response, we have updated all relevant references to the Intergovernmental Panel on Climate Change (IPCC) to cite the Sixth Assessment Report (AR6) instead of earlier versions (e.g., AR5). The revised citations now reference the latest findings on projected temperature increases and humidity trends, in alignment with current climate scenario modeling standards. These updates enhance the scientific currency and credibility of the manuscript. [Revised relevant part in manuscript – page number 1, and lines 43 to 46.]

Comments 12: [Check consistency in reference formatting.]

Response 12: We thank the reviewer for highlighting the consistency in reference formatting. We have carefully reviewed and revised the reference list to ensure uniform formatting throughout the manuscript. Journal names have been standardized based on the style guidelines of the Journal of Agriculture (MDPI), with consistent use of either full titles or standard abbreviations as required. This revision ensures professional presentation and adherence to the journal’s citation style.

Comments 13: [In line 165, what is the purpose for Figure 2.1? Lack of linkage with any section]

Response 13: Thank you for pointing this out. We acknowledge that the placement and purpose of Figure 2.1 were not clearly linked with the corresponding text. To enhance clarity and maintain logical flow, we have revised the manuscript to explicitly reference and explain Figure number as 1and also within the context of sampling design. This ensures a clear connection between the figure and the description of the sample distribution and validation process.

Comments 14: [In line 219, here is a Figure 1 while Figure 2.1 already appeared in line 165. Check the figures and their formatting.]

Response 13: Thank you for highlighting this inconsistency. We have reviewed and corrected the figure numbering to ensure a sequential and consistent format throughout the manuscript. Figure 1 now correctly appears before Figure 2, and all subsequent figures have been renumbered accordingly to reflect their proper order of appearance in the text.

Comments 1: [The stratification and random sampling might introduce bias if not adequately balanced across different farm sizes and geographic diversities within the selected districts.”]

Response to comment 1: Thank you for pointing this out. We agree with this comment. Therefore, we appreciate the reviewer’s concern regarding potential sampling bias. Section 2.2 (lines 109 to 130) now reports a χ² test of representativeness (χ² = 0.59, p = 0.90) and includes Figure 2a, which shows that the sample proportions across Faisalabad, Jhang, Chiniot, and Toba Tek Singh align closely with official district farm‑household counts. This demonstrates that the sampling frame is balanced and unbiased.” Appendix Table A1 provides the full χ² calculation (χ² = 0.59, p = 0.90).

Comments 2: [The article uses ARIMA and panel regression models without extensive discussion of the validity of the model assumptions]

Response to comment  2: Dear reviewer, for the panel regressions, a Hausman test (χ² = 0.011, p = 0.994) confirms that Fixed‑Effects are appropriate; Breusch‑Pagan (χ² = 1.24, p = 0.27) in table 3 and Wooldridge (F = 1.61, p = 0.20) indicate homoscedastic, serially independent residuals, and VIF = 1.001 rules out multicollinearity. The same temperature and humidity elasticities remain highly significant when we re‑estimate log‑linear, 2‑month‑lag, and 3‑month‑lag specifications, demonstrating specification robustness in table 5. For the ARIMA model, Table 4 now includes the stationarity tests: Augmented Dickey–Fuller τ = –4.754 (lag = 4, p = 0.010) and KPSS = 0.31 (< 0.46 critical), both confirming that the differenced series is stationary. Residual diagnostics show no autocorrelation (Ljung‑Box Q = 25.029, p = 0.069) and  RMSE and MAPE Together, these results demonstrate that the panel and ARIMA models satisfy all key assumptions and that the reported climate‑yield relationships are robust to alternative specifications.]”

Comments 3: [Why is there a need to model milk production concerning climate variables, and later to forecast it using the ARIMA approach (you have only eight years, right??). What is the gain here in terms of results? Or novelty? Or contribution?]

Response to comment  3: Dear Reviewer, our study adopts a sequential modeling strategy to serve dual objectives: explanation and projection. Panel regression models quantify the immediate effects of temperature and humidity on milk production, while the seasonal ARIMA model estimated on 96 monthly observations (2017–2024) forecasts future trends under IPCC scenarios (+2°C, +10% RH). Despite the eight-year span, the data meet Box-Jenkins criteria for ARIMA estimation. This integrated design links micro-level climate sensitivity with macro-level yield forecasting, offering both scientific insight and policy relevance. It has also been explained in the introduction. Please refer to the line in the introduction (69,74)

Comments 4: [“You present several sub-models before introducing the generalised regression model. Is there a need to have all the models, or is there a difference in the results? Wouldn’t it be better to choose a unique model to model milk production?”]

Response to comment  4 Thank you for this insightful suggestion. We understand the concern regarding potential model redundancy. However, the inclusion of all four models—Fixed Effects (Model 1), Random Effects (Model 2), Quadratic (Model 3), and Interaction (Model 4)—was a deliberate decision based on both empirical robustness and theoretical reasoning. Each model serves a distinct analytical purpose also explained in section 3 empirical analysis( line 185)

1. Model 1 (Fixed Effects) estimates the average linear effect of temperature and humidity while controlling for unobserved farm-level heterogeneity. The Hausman Test (χ² = 0.011, p = 0.994) confirms the appropriateness of the Fixed Effects model over Random Effects.
2. Model 2 (Random Effects) is reported for completeness and comparison. Its coefficient estimates align closely with Model 1, further validating the consistency of the core results.
3. Model 3 (Quadratic) detects non-linear physiological threshold effects. The negative coefficient of temperature² (−0.199, p < 0.001) identifies a yield-declining threshold beyond 27°C. Similarly, a positive humidity² term (0.012, p < 0.001) reveals a U-shaped humidity effect.
4. Model 4 (Interaction) captures the combined effects of temperature and humidity. The significant interaction term (−0.068, p < 0.001) shows that humidity amplifies the heat stress effect on milk production.

The explanatory power of the models increases sequentially:

• Model 1 (Linear FE): R² = 0.238
• Model 3 (Quadratic): R² = 0.3309
• Model 4 (Interaction): R² = 0.3553

These gains in fit and interpretability justify the inclusion of each model. elaborate it in regression analysis section 4.5. To address the reviewer’s concern, we have now clearly emphasized in methodology section then regression result section about Models selection from panel to quadratic and interaction model in the revised manuscript, while the preceding models are retained as essential diagnostic steps and robustness checks. This modeling progression follows established best practices in climate-agriculture literature. Additionally, we have added a model comparison table to improve transparency and facilitate interpretation.

Comments 5: ["Some of the robustness checks are intuitive since they consist of post-estimation tests, so consider it to be more synthetic in this part."]

Response to comment  5: Dear reviewer, we agree that the robustness checks should be concise and purposeful. In response, we streamlined the robustness section to focus only on the most essential diagnostics that validate the reliability and consistency of the regression models. These include the log-transformed model to address potential nonlinearity, lagged models to test delayed climate effects, and standard diagnostic tests such as Variance Inflation Factor (VIF), Breusch-Pagan, and Durbin-Watson tests to confirm the absence of multicollinearity, heteroskedasticity, and autocorrelation.

Rather than elaborating each test in depth, we now present the results in a summary table (Table 5) and include brief interpretation in the text. This revised presentation ensures the robustness checks serve their intended function without overcomplicating the narrative, maintaining both methodological rigor and clarity for the reader.

 Comments 6: [More importantly, you chose as climate variables the ones on average. Average temperature and precipitation, correct? Did you consider including heat waves? An average temperature doesn’t capture the reaction of the milk production quite directly. Don’t you think so?”]

Response to comment  6: Dear reviewer, we agree that extreme climatic events such as heatwaves can provide a more direct and granular understanding of dairy stress responses. However, in the present study, we focused on average monthly temperature and relative humidity for two key reasons:

First, our empirical models were based on eight years of consistent farm-level milk production data aligned with monthly meteorological records. The use of averages allowed for harmonization across both datasets and supported model stability, particularly for panel regression and ARIMA forecasting. Second, heatwaves and other extremes are episodic and require high-frequency (e.g., daily) time series data, which were not uniformly available across all surveyed districts for the full 8-year span.

Nevertheless, we fully acknowledge the value of incorporating climate extremes in future analyses. In the revised manuscript, we have clearly stated this as a methodological limitation and proposed it as a priority for future research. We believe this clarification will enhance the transparency of our methodological choices while aligning with best practices in climate-agriculture studies.

Comments 7: [There is a need to re-design the methodology used. Many models and many techniques elaborate quite limited panel data (at least in the time dimension), create confusion, and do not produce coherent and in-line results. If you want to forecast milk production after a panel data analysis (considering climate variables), try a technique for forecasting in a panel data set. The approach you chose gives the idea of a set of techniques used randomly and does not reflect a well-structured and motivated methodology]

Response to comment  7: We thank the reviewer for raising this valuable concern. In response, we have strengthened the justification and clarified the structure of our methodological approach to ensure coherence, transparency, and alignment with the research objectives.

Our multi-model framework is purposefully structured to address distinct but interconnected empirical questions:

1.       Panel Regression Models (Models 1–4) were used to identify the causal effects of temperature and humidity on milk production, allowing us to:

• Isolate farm-level heterogeneity (Fixed Effects)
• Capture non-linear threshold responses (Quadratic Model)
• Assess interactive climatic stressors (Interaction Model)
• Test spatial variability (Regional Models)

Each model progressively deepens the analysis rather than overlapping. For example, the interaction model (Model 4) revealed that high humidity exacerbates heat-related losses, a finding not evident from linear models alone. The higher R² in the interaction model further supports its empirical strength.

2.       ARIMA Forecasting was introduced after establishing the climatic relationship in the panel models. Rather than being a stand-alone method, ARIMA complements the regression analysis by offering predictive insights under future climate scenarios (e.g., +2°C and +10% RH) using past milk production trends.

To avoid the impression of random model selection, we have revised the methodology section to clearly specify the unified version integrating linear, quadratic, and interaction effects.

We believe these revisions have significantly improved the structure and alignment of our methodology, ensuring that each model contributes uniquely to a coherent analytical narrative.

Comments 8: [“Some more elaborated policy implications are needed. The ones you provided are quite general, and not directly related to your findings.”]

Response to comment  8: Dear reviewer, we have substantially revised Section 5.1: Implications for Dairy Farming and Adaptation Strategies to ensure that the policy recommendations are directly grounded in the study’s empirical findings, particularly the regression, ARIMA forecast, and climate scenario models. 

The updated section now includes:

• Region-specific adaptation strategies, based on insights from the regional regression models and scenario simulations, particularly identifying high-risk tehsils with compounded temperature-humidity exposure.
• Technological interventions such as heat-tolerant breed development, precision livestock farming, and climate-controlled housing, contextualized within feasibility constraints of smallholder dairy systems in Punjab.
• Policy instruments like climate-indexed livestock insurance, conditional cash transfers, and subsidy targeting frameworks, aligned with the study’s simulation of localized milk yield declines.
• Comparative policy learning, referencing adaptation success stories from Tunisia and northern India, to ensure relevance and replicability in Punjab’s agro-climatic conditions.

Comments 1: [“According to the duration of adaptation formation, there are three strategic directions of adaptive processes. It is not entirely clear from this study which direction the authors have chosen. The keywords include adaptation strategies. The study does not describe them and their application to the region.”]

Response 1: [Dear reviewer, we have clearly stated the strategic direction of adaptation followed in this study. Specifically, the study adopts a combined incremental and anticipatory adaptation approach. Incremental adaptation focuses on immediate and low-cost measures such as improved housing, ventilation, modified feeding regimes, and awareness campaigns. In parallel, anticipatory adaptation is reflected in our climate scenario modeling and simulation of future milk yield losses, which guide long-term planning for climate-resilient breeds and early-warning systems.

These strategies have been aligned with Punjab’s regional climate risks and socio-economic conditions. We have also included comparative references to successful adaptation initiatives in regions with similar agro-climatic conditions, such as Tunisia and northern India. The revised version of Section 5.1: Implications for Dairy Farming & Adaptation Strategies now explicitly outlines the adaptation pathways chosen and their applicability to vulnerable farming communities in Pakistan] Thank you for pointing this out. I/We agree with this comment. Therefore, I/we have explain. [ Section 5.1: Implications for Dairy Farming & Adaptation Strategies page 18 line 640 to 658.]

Comments 2: [The article does not compare this region with the experience of similar regions in the world regarding climate and dairy farming, as well as the solution to this problem.]

Response 2: Dear reviewer, , we have revised Section 5.1 to include comparative insights from other climate-vulnerable dairy regions. Specifically, we now reference successful adaptation strategies adopted in Tunisia and northern India—regions that, like Punjab, experience prolonged periods of heat stress and high humidity. – page number, 18.]

Comments 3: [“The authors also need to provide a clear definition of heat stress. It can be considered in different aspects.”.]

Response 3:  Dear reviewer, we have added a clear and concise definition of heat stress in the Introduction. (Page 1 line 43)

Furthermore, in the Discussion and Section 5.1, we elaborate on how heat stress is measured in terms of climatic indicators (e.g., temperature-humidity index) and perceived by farmers in the context of their local experiences.

Comments 4: [“To assess the farmers’ understanding of climate change and associated risks, structured interviews were conducted… What kind of questions were in the survey and can they be adapted for all farmers? Explain and justify.”]

Response 4: Dear reviewer, we have added a detailed explanation of the structured questionnaire in the Methodology section (Section 2.2 or 2.3). The survey included both closed and Likert-scale questions focusing on farmer awareness of climatic trends, observed changes in temperature and humidity, and perceived impacts on milk production. Questions also addressed seasonal variation in milk yield, extreme weather events (e.g., heatwaves), coping practices (e.g., shade, watering, feed adjustment), and perceived adaptation needs.

The survey was designed with simplicity, clarity, and cultural sensitivity to ensure it could be administered effectively across varying literacy levels. Local enumerators were trained to assist participants where necessary, making the tool adaptable for both literate and less-educated farmers. The standardized nature of the questionnaire also ensures comparability across all respondents, which supports generalizability of results across Punjab’s diverse farming population. We have revised the manuscript accordingly and added justification for the survey's broad applicability

Comments 5: [“For the quantitative assessment, daily milk production records were collected from 200 farmers over an eight-year period, from 2017 to 2024 (line 156–157), and before that the article mentioned a period of ten years. What period was used for the study?”.]

Response 5: Dear reviewer, the actual period used for quantitative milk production data analysis was eight years (2017–2024), based on availability of consistent and complete records across all selected farms. Earlier references to a ten-year period referred more generally to the recall period used in farmer perception surveys, where respondents were asked to reflect on climate-related changes and experiences over the past decade. To eliminate confusion, we have now clarified the distinction between these two timeframes in the revised

Comments 6: [A fixed effects model (FEM) is preferred to a random effects model (REM) (line 181). Justification.]

Response 6: Thank you for highlighting the need to justify the selection of the Fixed Effects Model (FEM) over the Random Effects Model (REM). We have now explicitly incorporated the statistical rationale behind this choice into the manuscript, to determine the most appropriate model, a Hausman test was applied (χ² = 0.011, p = 0.994) (page 12 line 411 to 414)

Comments 7: [“ARIMA models are used in some cases where the data show evidence of non-stationarity. When seasonality is evident in the time series, seasonal variance can be applied to remove the seasonal component.”]

Response 7: Dear reviewer, to ensure the validity of the ARIMA model application, we conducted formal diagnostic tests to verify both stationarity and seasonality in the milk production time series.

Specifically, we applied the Augmented Dickey-Fuller (ADF) test, which confirmed that the differenced series is stationary (τ = –4.754, p = 0.01). Additionally, seasonal patterns were clearly visible in the monthly data, consistent with climate-related livestock productivity cycles. Consequently, we employed a Seasonal ARIMA (SARIMA) model with specification ARIMA (1,0,0)(1,1,0)[12], which captures both short-term autocorrelation and seasonal dependencies.

We have updated the methodology and results sections to include this justification, ensuring clarity and model robustness. (Page14 line 484 to 486)

Comments 8: [“The conclusions need to be clarified and expanded.”]

Response 8: We thank the reviewer for this important suggestion. In response, the Conclusion section has been revised to explicitly summarize the core empirical findings, connect them to the adaptation implications, and propose concrete recommendations for future research. The updated version now reflects the significance of our ARIMA forecasts, multi-model regression analysis, and farmer perception alignment providing a cohesive end to the study and emphasizing policy relevance.

Comments 9: [“The references in the article need to be corrected.”]

Response 9:  Thank you for pointing this out. We have carefully reviewed and corrected all references to ensure consistency, completeness, and alignment with the journal's citation style. This includes updating citation years, standardizing author formats, and ensuring that all in-text citations correspond accurately to entries in the reference list. Additional peer-reviewed and recent literature has also been integrated into the Introduction and Discussion sections to further strengthen the academic foundation of the study.  

Comments 1: [“Although the literature used is appropriate and sufficient, a typical literature review is missing at the beginning. And the introduction is modest in this respect (15 items). Perhaps it would be worth supplementing the article with a literature review (using the one from the discussion) or expanding the introduction with more literature?”]

Response 1: We sincerely thank the reviewer for highlighting the need to expand the literature foundation of the introduction. In response, we have revised the introduction to integrate a more comprehensive review of recent and relevant studies. [page 1 line 33 to42.]

Comments 2: [“Line 10 – In the summary, it should be noted that the study concerned Pakistan.”]

Response 2: We thank the reviewer for this valuable suggestion. We have revised the abstract to explicitly mention that the study is based in Faisalabad region of Punjab Pakistan. (line 10)

Comments 3: [Line 114 – It is necessary to explain what a ‘Tehsil’ is.”]

Response 3: Thank you for pointing this out. We have clarified the term “Tehsil” in the revised manuscript. A sentence has been added to Section 2.2 (Sampling Strategy and Study Area  (line 134,135)

Comments 4: [Line 116 – It is necessary to explain what the ‘Union Council and Farmer Selection’ is and how it works.”]

Response 4: Thank you for the helpful observation. We have elaborated on the Union Council and farmer selection process in Section 2.2 of the revised manuscript.  (Line 136,137)

Comments 5: [“Point 2.2 – It is necessary to explain exactly where the list of milk producers from which the random research sample was based was taken from. Unfortunately, this point is not explained in detail.”.]

Response 5: Thank you for this important comment. We have now clarified the source of the sampling frame in Section 2.2. Specifically, the list of milk producers was obtained from the district livestock departments, tehsil level and Union Council level livestock extension offices in each selected tehsil.

Comments 6: [“Point 4.2 – Is it possible to determine how many people in a family are involved in milk production on the farm on average?”]

Response 6: Thank you for this thoughtful observation. In response, we have added clarification in Section 4.2 of the manuscript. Based on our socio-economic survey data, we found that on average, 3 to 5 family members are actively involved in various aspects of milk production on each farm. These activities include feeding, milking, fodder collection, animal care, and local distribution. This reflects the labor-intensive nature of smallholder dairy farming in Punjab, where family labor remains a crucial resource due to limited mechanization and hired labor.

Comments 7: [“The source, i.e. own study, should be provided below tables and graphs.”]

Response 7: Thank you for this helpful suggestion. We have now added the source notation “Source: Authors’ own field survey (2024)” all relevant tables and figures that are based on primary data collected from dairy farmers. 

Comments 1: [There is a misalignment between the stated research aim of exploring smallholder perceptions and the rest of the paper, which focuses on climate variables and milk production. Only subsection 4.3 discusses farmer perceptions, creating a disconnection between the article’s title, objectives, and methodology.”.]

Response 1: We appreciate the reviewer’s observation. To address the concern, we have made the following refinements to Subsections 2.2, 2.3, and 2.4 to ensure clear alignment between the study’s stated objective (exploring smallholder perceptions and climate forecasting) and the methodological narrative:

• In Subsection 2.2, we have clarified that the stratified sampling design was not only intended for geographic representation but also for capturing a diversity of farm sizes, which includes a focus on smallholder perceptions.
• In Subsection 2.3, we have emphasized that the structured perception survey was a core component of the study, forming the basis of our perception analysis (Subsection 4.3). We have also clarified that this section directly supports our objective of aligning perceptions with empirical climate impacts.
• In Subsection 2.4, we now clarify that the same farmers who provided perception data also maintained detailed milk production records, thus forming an integrated dataset used in both regression analysis and ARIMA forecasting. This dual-data approach justifies the methodological linkage between perceptions and forecasting, aligning with the paper’s mixed-method design.

Additionally, the distinction between the perception analysis and the climate modeling components has been explicitly stated in both the methods and the discussion sections. This ensures that both dimensions of the study—subjective (perception) and objective (climate effects)—are methodologically and thematically aligned.

Comments 2: ["The ARIMA modelling is about forecasting of milk production under specific climate scenarios – it is not about the perceptions of the producers."]

Response 2: Thank you for this observation. We agree that the ARIMA modeling is distinct from the perception analysis and serves a different purpose. In the revised manuscript, we have clearly delineated these components in both the methodology and results sections. Specifically, the perception analysis (Section 2.3 and 4.3) captures qualitative farmer insights regarding climate-induced risks, while the ARIMA model (Section 4.6 and 6.1) uses historical milk yield data to project future trends under climate stress scenarios. These components complement each other, with the ARIMA model offering forward-looking projections and the perception data providing immediate, on-ground experiential context. This dual approach ensures a more holistic assessment of climate risk in dairy farming and aligns with the mixed-methods framework of the study.

Comments 3: ["Furthermore, IPCC has two main Representative Path Concentration (RPC) scenarios: the pessimistic (RPC 8.5) and the optimistic (4.5) - it is not clear in the research.".]

Response 3: We appreciate the reviewer’s note regarding the IPCC’s Representative Concentration Pathways (RCPs). While the study does not adopt formal RCP 8.5 or RCP 4.5 pathways in full due to regional constraints in downscaled data, the simulated scenarios of +2°C temperature and +10% relative humidity reflect moderate, regionally plausible stress increments aligned with IPCC’s mid-century projections for South Asia. These increments were selected to stress-test milk productivity under realistic future conditions. A clarification regarding this alignment with IPCC frameworks has now been added to the methodology and discussion sections for transparency.

Comments 4: [The results about perceptions appear only in the subsection 4.3..]

Response 4: Thank you for this helpful observation. We agree that farmers' climate risk perceptions are central to the study’s theme and should be emphasized more clearly throughout the manuscript. In response, we have made two key improvements:

1. Integration in Results and Discussion: Farmer perception results have been better highlighted in Section 4.3, with additional explanation on how these perceptions complement the regression and forecasting findings.
2. Linkage in Introduction and Conclusion: The revised Introduction now emphasizes the mixed-methods design to bridge perception with empirical trends, and the Conclusion reiterates how understanding farmer perceptions is essential for designing adaptive dairy policies.