Interdependent adoption of climate change adaptation strategies among rice farmers in northwest Bangladesh Authors: Md Shohidul Islam, Bikash Chandra Ghosh, Apurba Adhikary Published: 2026 ABSTRACT AbstractClimate variability poses serious threats to agricultural productivity and food security, particularly for smallholder rice farmers in climate-vulnerable regions such as northwest Bangladesh. This study examines the adoption of multiple climate change adaptation strategies among smallholder rice farmers in this region. Using a multistage sampling technique, data were collected from 498 farmers. A multivariate probit model and a fractional response model were applied to analyze both the determinants and intensity of adoption of key strategies, including adjusting planting dates, integrated pest management, tolerant rice varieties, improved rice varieties, and rice straw mulching. This joint modeling framework captures interdependence among strategies and provides deeper insights than single-strategy approaches. Robustness was checked using a multinomial logistic regression model. The results indicate that socio-economic factors such as farming experience, access to credit, and membership in farmer groups significantly influence adoption behavior. Strong complementarities among adaptation strategies are also observed, suggesting that farmers tend to adopt these practices in bundles rather than individually. The findings contribute to the literature by highlighting the effectiveness of bundled adaptation strategies in enhancing resilience. They offer important policy implications for designing integrated and context-specific interventions to strengthen adaptive capacity and promote resilient agricultural systems in climate-exposed regions.Keywords: Climate change, Adoption strategies, Multivariate probit model, Fractional response model, Northwestern BangladeshSubject terms: Ecology, Ecology, Environmental sciences, Environmental social sciences Introduction The future of global agriculture largely depends on the ability of farming systems to adopt advanced technologies and sustainable practices to address pressing challenges affecting farmers, land, and crop production1. Among these challenges, climate change remains one of the most critical concerns2. Farmers in developing countries face disproportionately greater challenges than those in developed nations, largely due to limited access to resources and adaptive technologies3. As most South Asian countries are low-income and located in lower latitudes, they are particularly vulnerable to the impacts of climate change, posing serious risks to regional food security4. Cereal production in South Asia is projected to decline by up to 30% between 2001 and 2059, while per capita water availability may decrease by as much as 37%3. Building on the regional vulnerability discussed above, Bangladesh, being heavily reliant on agriculture, faces significant threats to crop production and overall agricultural sustainability due to climate change, particularly in the northwest region5,6. This decline in agricultural productivity is driven by shifting rainfall patterns, rising temperatures, prolonged droughts, frequent flooding, and increasing salinity from sea-level rise7–9. In particular, the yields of two essential rice varieties, Aus and Aman, are projected to decline by approximately 1.5–25.8% and 0.4–5.3%, respectively, by 2050 (Kamruzzaman et al.10. Furthermore, the production of modern Boro rice varieties may decrease by 20% by 2050 and up to 50% by 2070, reflecting the high sensitivity of rice to climatic conditions11. To ensure water and food security while promoting economic growth, environmental sustainability, and climate resilience, Bangladesh must adopt integrated and adaptive strategies12. Given the country’s diverse ecosystems and climate regimes13, addressing climate change also requires coordinated efforts, including financial support, technology transfer, and capacity building. At the farm level, smallholder rice farmers are already adopting various adaptation strategies, such as drought-tolerant rice varieties, adjusted planting dates, crop rotation, and conversion of paddy fields to fish production14. These adaptation practices enhance farmers’ capacity to cope with climate variability and reduce the adverse effects of extreme climatic events15. Northwestern Bangladesh has experienced significant environmental degradation in recent years due to climate change, with rainfall variability emerging as a critical factor affecting rice production. Both water scarcity and excess water pose major challenges for rain-fed agriculture16. The region is an important economic hub, contributing approximately 8% to national GDP through its agricultural resources17,18. Given this dependence, it is highly vulnerable to climate-related risks such as prolonged droughts, groundwater depletion, and flooding. Therefore, integrating climate trends and future projections into agricultural planning, infrastructure development, and disaster preparedness is essential18. In this context, adopting multiple adaptation strategies is particularly important. The combination of modern agricultural technologies with traditional practices highlights the role of indigenous knowledge in cereal crop cultivation. Such approaches enhance climate resilience while also conserving valuable landraces19. Hence, promoting multiple adaptation strategies is essential for ensuring sustainable farming under changing climatic conditions. Recent research reveals that smallholder farmers increasingly adopt multiple climate change adaptation strategies, challenging the traditional assumption that only one strategy would be used20. Consistent with this, Chowdhury et al.21, found that farmers in Bangladesh employ diverse adaptation practices to mitigate climate impacts. Widely documented strategies include crop rotation, drought-tolerant varieties, crop diversification, improved irrigation, adjusted planting dates, non-farm income activities, and microinsurance schemes22. Importantly, studies in developing countries show that the simultaneous adoption of multiple strategies can improve crop yields and household income23,24. Furthermore, Tefera et al.25 identify key factors influencing adaptation behavior, including demographic characteristics, farm size, access to climate information, credit, extension services, and farming experience. Larger farm size is expected to positively influence the likelihood of adopting multiple climate adaptation strategies, while access to credit facilitates the adoption of information-intensive practices such as integrated pest management (IPM). Farming experience also enhances the intensity of multiple strategy adoption among smallholder farmers25. Despite these insights, two critical gaps remain in the literature. First, although several studies acknowledge the adoption of multiple adaptation strategies, most analyses treat these decisions either independently or focus solely on binary adoption outcomes, thereby overlooking potential interdependencies and complementarities among strategies. Second, in the context of Bangladesh, existing studies have largely examined either the determinants of adoption or specific adaptation practices in isolation3,26–33, with only limited attempts to analyze multiple strategy adoption34–39. More importantly, prior studies have not jointly examined both the probability and the intensity of adopting multiple adaptation strategies within a unified econometric framework, leaving an incomplete understanding of farmers’ adaptive behavior. Addressing these limitations, the present study makes both methodological and empirical contributions. Methodologically, it integrates a multivariate probit (MVP) model with a fractional response model to jointly estimate (i) the interdependent decision to adopt multiple strategies and (ii) the intensity of their adoption. This combined approach allows for the identification of complementarities and substitution effects among adaptation strategies, which are typically overlooked in single-equation or single-strategy analyses. Empirically, the study provides one of the first comprehensive applications of this joint framework in the context of smallholder rice farmers in northwest Bangladesh, where financial and institutional constraints critically shape adaptation choices. By simultaneously capturing both the likelihood and extent of adoption, the analysis offers a more complete and policy-relevant understanding of how farmers bundle adaptation strategies under climate stress. While this study focuses on northwest Bangladesh, the insights on interdependent and intensive adaptation behavior have broader relevance for similar agro-ecological regions in South Asia and Sub-Saharan Africa, where smallholder farmers face comparable resource constraints and climate risks. The remainder of the paper is structured as follows: the next section details the study area, sampling methods, and data sources. This is followed by a description of the empirical models employed. Subsequently, the results are presented and discussed, with concluding policy recommendations offered in the final section. Research methodology Study area This study focuses on the northwestern region of Bangladesh, as shown in Fig. 1, specifically the Rajshahi Division, which was purposively selected due to its pronounced vulnerability to drought and increasing water scarcity40. The region is characterized by physical water scarcity and extreme water security risks, as highlighted by previous studies40,41. Map of the North-western region. Within this region, the Natore and Pabna districts were selected for primary data collection through a cross-sectional household survey. These districts are particularly relevant to the study’s objectives, as they are known for recurrent droughts, below-average rainfall, and limited river flows during dry periods42. Natore and Pabna also experience some of the lowest annual rainfall in the country and are among the hottest areas in Bangladesh, with summer temperatures reaching up to 43 °C43–45. The selection of both the study area and participants was made carefully to ensure the findings accurately capture the relationship among climate variability trends. Sampling technique, and data collection The study compiled household lists from the selected villages and employed random sampling to select respondents, with household heads serving as interviewees. A multi-stage random sampling technique was adopted, beginning with two purposively selected districts—Pabna and Natore. From these districts, four upazilas and twelve villages were randomly chosen. Primary data were collected from households to gain a comprehensive understanding of farmers’ perceptions of climate variability and their adaptive responses. Following Perry’s46 recommendation for a minimum sample size of 350 in structured quantitative research, and Bartlett et al.’s47 guideline that a sample representing 5% of the population is sufficient for surveys, the study selected 498 farming households, approximately 10% of each village’s population. This sample size was considered adequate to ensure the reliability and validity of the study’s findings. A well-structured questionnaire was developed based on the research questions derived from the study’s objectives. To assess its validity, a pilot survey was conducted prior to the main data collection. Interviews were then carried out by visiting respondents at their homes. The questionnaire was carefully designed to ensure the collection of accurate and relevant data, and in total, 498 households were interviewed. The explanatory variables Several socioeconomic and demographic factors influence farmers’ adaptation decisions to climate change. Age can play a significant role, as older farmers often rely on traditional practices, while younger farmers tend to be more receptive to adopting new technologies48. Household size also matters, larger households may offer more labor resources, making it easier to implement labor-intensive adaptation strategies49. Education enhances awareness and understanding of climate change, which can lead to the adoption of more effective adaptation practices50. Similarly, farming experience contributes to better knowledge of local climate patterns and adaptive responses51. Larger farm sizes may provide the flexibility to experiment with or adopt a wider range of adaptation strategies52. Geographic and institutional factors also influence adaptation. Greater distances from key resources or markets can hinder access to essential farming inputs, thereby affecting productivity and adaptive capacity49. Market accessibility influences income generation, access to agricultural inputs, and the flow of information53. Access to agricultural extension services is vital, as these services provide critical training and information that support the adoption of adaptation measures48. The frequency of contact with extension agents may build trust and increase responsiveness to climate risks54. Moreover, membership in farmer groups or community organizations can enhance access to shared resources, knowledge, and collective adaptive action55. Income diversification is an important factor in reducing dependence on crop production and lowering vulnerability to climate-related risks49. It includes off-farm activities and livestock rearing, which provide complementary income that can support investment in adaptation practices. Access to credit enables farmers to invest in new technologies and adaptive practices48, while early warning systems improve preparedness and risk management56. Preparedness for future likely climate change impacts refers to the willingness of farmers to take action in response to anticipated climate risks, reflecting long-term risk reduction behavior57. Lastly, a farmer’s primary occupation determines income stability, access to information, and overall adaptability58. Dependent variables The dependent variables in this study represent key climate change adaptation strategies (CCAS) adopted by smallholder rice farmers in northwest Bangladesh. The selected strategies, namely adjusting planting dates (APD), integrated pest management (IPM), tolerant rice varieties (TRV), improved rice varieties (IRV), and rice straw mulching (RSM), were identified based on a pilot survey and consultations with local agricultural extension officers. These strategies reflect the most commonly practiced rice-based adaptation measures in the study area. Other strategies were less prevalent and therefore excluded to maintain a focused analysis of crop-level adaptation. APD refers to shifting the sowing or transplanting time of rice crops to avoid climate-related risks such as drought, erratic rainfall, and temperature extremes. In northwest Bangladesh, where rainfall variability and delayed monsoons are common, adjusting planting dates is a widely used strategy to stabilize production. Prior studies have highlighted its effectiveness in reducing climate-induced yield losses59. IPM involves the combined use of biological, cultural, physical, and chemical methods to manage pest populations in an environmentally sustainable manner. Climate variability in the study region often increases pest incidence, making IPM particularly relevant. It reduces dependence on chemical pesticides while maintaining productivity and environmental balance60. TRV refers to the adoption of drought-tolerant or stress-resilient rice varieties that can withstand water scarcity and adverse climatic conditions. In drought-prone areas of northwest Bangladesh, these varieties play a critical role in reducing the risk of crop failure, although they may offer comparatively lower yields. Existing literature supports their importance in enhancing climate resilience among smallholder farmers61. IRV represents modern high-yielding rice varieties that require sufficient irrigation and input use to achieve optimal productivity. While these varieties offer higher yield potential, their performance is sensitive to water availability, making them particularly relevant in the context of climate variability. The inclusion of IRV reflects the trade-off between productivity enhancement and climate risk in the study region62. RSM involves the application of rice straw as mulch to conserve soil moisture, reduce evaporation, suppress weed growth, and improve soil organic matter. In the water-scarce conditions of northwest Bangladesh, this practice enhances water use efficiency and soil health. Previous studies have identified mulching as an effective adaptation to drought stress63. Additionally, the Share of Adaptation Strategies variable captures the proportion or frequency with which different CCAS are adopted within the farming community or sample, providing insights into strategy preference and overall adaptive behavior64. Multivariate probit model This study uses the Multivariate Probit (MVP) model to examine factors influencing the adoption of multiple climate change adaptation strategies among smallholder rice farmers. The MVP is particularly appropriate because adaptation decisions are interdependent; the choice to adopt one strategy may influence the likelihood of adopting others due to shared socio-economic, institutional, and farm-level factors65. Unlike separate univariate probit or logit models, the MVP explicitly accounts for correlations among error terms across multiple adoption equations, preventing biased or inconsistent estimates and capturing complementarities or substitutive relationships among strategies66–68. Alternative approaches, such as count models or multiple univariate regressions, cannot effectively model these interdependencies or identify how strategies are jointly adopted. While endogeneity remains a potential concern due to self-selection or omitted variables, the MVP provides the most suitable framework for analyzing simultaneous adoption of multiple correlated strategies67. The underlying assumption is that the unobserved error terms jointly follow a multivariate normal (MVN) distribution with a zero mean vector and a correlation matrix with unit diagonal elements. This setup enables the model to capture complementarities or substitution effects among adaptation strategies. Specifically, this study considers the probability that the ith farmer adopts the kth strategy (APD, IPM, TRV, IRV, and RSM), modeled as a latent utility Y*ik: Y*ik = βkX’i + εik. where Xi is a vector of explanatory variables capturing farmer characteristics, institutional factors, and climate variables, and εik is the error term. The observed binary outcome is expressed as: This formulation captures the farmer’s decision to adopt any of the seven adaptation strategies based on net utility. Fractional response model To analyze the extent or level of adoption of climate change adaptation strategies, a logit based Fractional Response Model (FRM) is utilized. This model is suitable when the dependent variable is a proportion, bounded between 0 and 169. Unlike linear models, the FRM naturally accommodates boundary values without transformation, captures non-linear effects, and directly estimates the relationship between predictors and the adoption fraction69. In this case, the dependent variable is the share of adaptation strategies adopted out of the total number available. Following Papke and Wooldridge69, the model estimates the conditional mean of the proportion of strategies adopted, given that some strategies are adopted: In this model, CCSHAREi denotes the fraction of climate adaptation strategies adopted, Zi is a vector of household and farm-level characteristics, and β is the parameter vector to be estimated. A nonlinear function, such as logit or probit, is used to ensure predicted values stay within the [0,1] range. Results This chapter presents the findings of the study, highlighting the adoption of climate change adaptation strategies by rice farmers and the factors influencing their decisions. Descriptive statistics Rice farmers in the study area exhibit varying adoption rates across different climate adaptation strategies, reflecting economic considerations, resource constraints, and practical feasibility. According to Tables 1 and 96.99% of farmers applied pest management practices, highlighting that knowledge-intensive, low-cost strategies with immediate protective benefits are highly attractive and widely feasible. In contrast, only 23.09% adopted tolerant rice varieties, likely due to higher input costs, limited seed availability, or uncertainty about yield benefits. Adjusting planting dates was adopted by 54.82% of farmers, reflecting a moderate-cost strategy that requires careful planning and labor allocation but provides flexibility to respond to changing rainfall patterns. Improved rice varieties were planted by 76.31% of farmers, suggesting recognition of their productivity advantages, though adoption may still be limited by access to credit or market constraints. Straw mulching was implemented by 45.78% of farmers, representing a labor-intensive but low-capital strategy that can improve soil moisture retention and resilience to climatic shocks. Finally, 60.64% of farmers engaged in strategy sharing, indicating the importance of social networks and information exchange in facilitating learning and reducing individual risk. These patterns show that farmers strategically combine multiple strategies to balance costs, labor, risk, and expected returns, demonstrating both economic rationality and adaptive capacity in responding to climate variability Table 2. Table 1Definition and summary statistics of all variables used in the analysis.VariablesDescriptionsMin.Max.MeanSd.Source Continuous variables AgeAge of respondents.187548.6712.08Anzum et al.48,.Households sizeNumber of household members.1144.741.59Rahman et al.49,EducationYears of formal education acquired by the household head.0168.277.00Feinstein et al.50,ExperienceYears of farming experience.06031.1113.89Tripathi and Mishra51,Farm SizeDecimals of cultivated farmland.0.4304.023.48Hossain et al.52,Distance to farmDistance Between Farm and household.0100.860.91Rahman et al.49,Distance of MarketDistance to the primary market in kilometers.050.290.62Gebremedhin and Swinton53,Frequency of Extension serviceNumber of extensions visits in a farming season.0151.000.95Davis54, Dummy variables Yes No Access Extension serviceRespondent’s access to extension service (If have then 1; otherwise, 0).78.71%21.29%0.40Anzum et al.48,.Member of Farmer GroupBeing a member of a farm-based group(If involve then 1; otherwise, 0).18%82%0.38Okumu and Muchapondwa55,Off -farm ActivityRespondent’s engagement in any non-farm income-generating activity (If have then 1; otherwise, 0).51%49%0.50Rahman et al.49,Access creditRespondent’s access to credit (If have then 1; otherwise, 0).38.76%61.24%0.48Anzum et al.48,.Having Early Warning SystemRespondent’s awareness of climate change (If have then 1; otherwise, 0).26.71%73.29%0.44Hellmuth et al.56,Prepared to Future likely climate change (CC) impactsWillingness to take action in response to anticipated future climate change impacts (If take then 1; otherwise, 0).17.07%82.93%0.37Below57,Primary Profession1 If the main occupation of the respondent is farming, otherwise 0.75.90%24.10%0.44Bryan et al.58, Definition and summary statistics of all variables used in the analysis. Being a member of a farm-based group (If involve then 1; otherwise, 0). Table 2Descriptive statistics.StrategiesPercentagesAPDYes54.82No45.18IPMYes96.99No3.01TRVYes23.09No76.91IRVYes76.31No23.69RSMYes45.78No54.22Sharing StrategiesYes60.64No39.36Source: Data analysis, 2025. Descriptive statistics. Source: Data analysis, 2025. Key influences on rice farmers’ implementation of multiple strategies to adapt to climate change This section presents the results of the MVP analysis, which examines the factors influencing smallholder rice farmers’ adoption of multiple climate change adaptation strategies, as detailed in Table 3. The Wald chi-square statistic (χ² (75) = 143.31, p < 0.0000) indicates that the explanatory variables are jointly significant. Moreover, the log-likelihood value of − 1192.6773 further supports the robustness and overall fit of the MVP model employed in this study. Table 3Multivariate Probit Analysis of Rice Farmers’ Climate Adaptation Strategies.VariablesAPDIPMTRVIRVRSMAge −0.018** (0.007) 0.011(0.022)0.009(0.008)−0.007(0.008)−0.010(0.007)Households size0.038(0.041) 0.225** (0.092) 0.002(0.042)0.033(0.04300.034(0.039)Education−0.002(0.009)0.011(0.021)0.001(0.009)0.007(0.010)−0.011(0.009)Experience 0.011* (0.007) 0.016(0.016)−0.007(0.007)0.004(0.007)−0.006(0.007)Farm Size 0.001*** (0.017) 0.165** (0.082) 0.031(0.021) 0.035* (0.018) 0.045** (0.018)Distance to farm0.213(0.073)0.421(0.301)−0.020(0.0668)0.046(0.072)−0.050(0.064)Distance of Market−0.179(0.111)0.383(0.357)0.062(0.114)−0.196(0.126)−0.084(0.102)Access Extension service0.148(0.150)0.252(0.323)−0.001(0.158) 0.394** (0.178) −0.016(0.143)Frequency of Extension service−0.002(0.059)0.324(0.253)−0.034(0.063)−0.020(0.060)0.132(0.083)Member of Farmer Group0.096(0.168)0.100(0.455)−0.126(0.177)−0.146(0.181) −0.278* (0.163) Off -farm Activity−0.011(0.124) 0.751** (0.352) 0.203(0.132)−0.075(0.133)−0.105(0.121)Access credit −0.259** (0.124) 0.194(0.311) −00.239* (0.131) 0.270** (0.132) −0.164(0.121)Having Early Warning System 0.794*** (0.151) 0.773(0.584)0.055(0.160) 0.509*** (0.158) −0.098(0.146)Prepared to Future likely climate change (CC) impacts0.230(0.177)1.252(1.784)0.208(0.195)−0.122(0.193)0.016(0.169)Primary Profession−0.061(0.138)−0.380(0.351)0.104(0.145) 0.318** (0.159) 0.135(0.134)cons0.099(0.368)1.560(0.973)0.266(0.393)−1.590(0.420)0.572(0.3620Source: Data analysis, 2025.*, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Wald chi2 (75) = 143.31; Prob > chi2 = 0.0000; Log likelihood = − 1192.6773; Number of obs. = 498. Multivariate Probit Analysis of Rice Farmers’ Climate Adaptation Strategies. −0.018** (0.007) 0.011 (0.022) 0.009 (0.008) −0.007 (0.008) −0.010 (0.007) 0.038 (0.041) 0.225** (0.092) 0.002 (0.042) 0.033 (0.0430 0.034 (0.039) −0.002 (0.009) 0.011 (0.021) 0.001 (0.009) 0.007 (0.010) −0.011 (0.009) 0.011* (0.007) 0.016 (0.016) −0.007 (0.007) 0.004 (0.007) −0.006 (0.007) 0.001*** (0.017) 0.165** (0.082) 0.031 (0.021) 0.035* (0.018) 0.045** (0.018) 0.213 (0.073) 0.421 (0.301) −0.020 (0.0668) 0.046 (0.072) −0.050 (0.064) −0.179 (0.111) 0.383 (0.357) 0.062 (0.114) −0.196 (0.126) −0.084 (0.102) 0.148 (0.150) 0.252 (0.323) −0.001 (0.158) 0.394** (0.178) −0.016 (0.143) −0.002 (0.059) 0.324 (0.253) −0.034 (0.063) −0.020 (0.060) 0.132 (0.083) 0.096 (0.168) 0.100 (0.455) −0.126 (0.177) −0.146 (0.181) −0.278* (0.163) −0.011 (0.124) 0.751** (0.352) 0.203 (0.132) −0.075 (0.133) −0.105 (0.121) −0.259** (0.124) 0.194 (0.311) −00.239* (0.131) 0.270** (0.132) −0.164 (0.121) 0.794*** (0.151) 0.773 (0.584) 0.055 (0.160) 0.509*** (0.158) −0.098 (0.146) 0.230 (0.177) 1.252 (1.784) 0.208 (0.195) −0.122 (0.193) 0.016 (0.169) −0.061 (0.138) −0.380 (0.351) 0.104 (0.145) 0.318** (0.159) 0.135 (0.134) 0.099 (0.368) 1.560 (0.973) 0.266 (0.393) −1.590 (0.420) 0.572 (0.3620 Source: Data analysis, 2025. *, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Wald chi2 (75) = 143.31; Prob > chi2 = 0.0000; Log likelihood = − 1192.6773; Number of obs. = 498. The results indicate a negative association between age and APD adoption, suggesting that older farmers are less likely to adjust planting dates in response to climate variability, consistent with Adeagbo et al.1,, who observed that older farmers tend to maintain established agronomic practices. In contrast, younger farmers are more likely to adopt improved varieties and modify planting schedules, reflecting their greater openness to innovation and willingness to experiment with new techniques. Household size was positively associated with IPM adoption, implying that larger households can allocate sufficient labor for labor-intensive practices, although they may avoid other strategies that require specialized inputs or management. This aligns with Muneer et al.70,, who highlighted that labor availability shapes the prioritization of specific adaptation practices. Education was not statistically significant for any strategy, suggesting that formal schooling may play a limited role compared to practical experience. Farming experience was positively associated with APD adoption, corroborating Ojo and Baiyegunhi71, who emphasized that experienced farmers are more capable of implementing climate-smart practices due to accumulated knowledge and skill. Farm size was positively associated with the adoption of all strategies except TRV, suggesting that larger farms possess greater resources and capacity to implement multiple adaptation measures, consistent with Muhammad Irshad Ahmad et al.72,, who emphasized that resource availability enables broader climate adaptation. In contrast, TRV adoption may be less dependent on farm size due to its focus on stress tolerance rather than yield maximization. Distance to the farm or market was not statistically significant, indicating that geographic access does not strongly constrain adaptation decisions in this context. Access to extension services was positively associated with IRV adoption, reflecting the critical role of institutional support in promoting high-yielding varieties and enhancing climate resilience, as noted by He et al.73,. However, the frequency of extension visits was not significant, possibly due to variation in service quality or farmers’ selective engagement with advisory support. Surprisingly, membership in farmer groups was negatively associated with RSM adoption, which contrasts with Kahsay and Endalew74 and may reflect that, in the study area, cooperative groups prioritize activities such as credit access, input pooling, or collective marketing rather than labor-intensive soil management practices like mulching. Additionally, differences in local agro-ecological conditions, labor availability, or competing household priorities may limit the time and resources households can dedicate to RSM despite group membership. Off-farm activity positively influenced IPM adoption, likely because additional income from non-farm sources enables farmers to invest in labor- and input-intensive practices. Similarly, livestock rearing provides a complementary source of household income, increasing farmers’ financial capacity to adopt resource-intensive practices such as IPM. This suggests that diversified income sources, both on- and off-farm, play a critical role in enabling investment in sustainable agricultural strategies75. Access to credit was positively associated with IRV adoption but negatively associated with APD and TRV, suggesting that while credit enables investment in high-yielding varieties, it may not encourage labor- or risk-intensive adjustments such as changing planting dates or adopting stress-tolerant varieties, particularly in the local context where these strategies require more labor and careful timing76. Similarly, the negative association between farmer-group membership and RSM may reflect competing labor demands within cooperative activities, differing local priorities, or agro-ecological constraints that limit the adoption of labor-intensive mulching practices. Early warning systems significantly promoted APD and IRV adoption, highlighting the role of timely climate information in facilitating proactive adaptation77. Variables capturing preparations for future climate impacts were not statistically significant, indicating that anticipatory adaptation remains limited among farmers in the study area. Correlation analysis of adaptation strategies based on the multivariate probit model The multivariate probit estimates provide evidence on the interdependence among adaptation strategies, with pairwise correlations reported in Table 4. The coefficients are generally modest in magnitude, although several are statistically significant, indicating limited but meaningful interdependence rather than strong complementarities, as the relatively small correlation values suggest that joint adoption occurs but is not strongly reinforced across strategies. Positive associations are observed between IRV and APD, TRV and IPM, and RSM and IPM, suggesting that the adoption of one practice is associated with a higher likelihood of adopting another. These patterns may reflect complementarities in farm management, where IRV adoption requires timely planting decisions to realize yield potential, while TRV adoption may increase reliance on IPM to safeguard crops under stress conditions, and RSM may enhance field conditions that support more effective pest management. In contrast, negative or insignificant correlations may arise from trade-offs in resource allocation, differences in input and labor requirements, and heterogeneity in farmers’ risk preferences and agro-ecological constraints, which limit the simultaneous adoption of certain strategies. Table 4Interrelationships among adaptation strategies.Pair wise adaptation strategiesCo-efficientRho21 (IPM and APD)0.185(0.167)Rho31 (TRV and APD)−0.085(0.077)Rho41 (IRV and APD) 0.146* (0.077) Rho51 (RSM and APD)−0.033(0.070)Rho32 (TRV and IPM) 0.370*** (0.101) Rho42 (IRV and IPM)−0.054(0.114)Rho52 (RSM and IPM) 0.226** (0.106) Rho43 (IRV and TRV)−0.065(0.084)Rho53 (RSM and TRV)−0.036(0.075)Rho54 (RSM and IRV)0.104(0.075)Source: Data analysis, 2025.*, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Likelihood ratio test of rho21 = rho31 = rho41 = rho51 = rho32 = rho42 = rho52 = rho43 = rho53 = rho54 chi2(10) = 30.6442 Prob > chi2 = 0.0007. Interrelationships among adaptation strategies. 0.185 (0.167) −0.085 (0.077) 0.146* (0.077) −0.033 (0.070) 0.370*** (0.101) −0.054 (0.114) 0.226** (0.106) −0.065 (0.084) −0.036 (0.075) 0.104 (0.075) Source: Data analysis, 2025. *, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Likelihood ratio test of rho21 = rho31 = rho41 = rho51 = rho32 = rho42 = rho52 = rho43 = rho53 = rho54 chi2(10) = 30.6442 Prob > chi2 = 0.0007. Factors influencing the extent of adoption of climate change adaptation strategies The results of the Fractional Response Model presented in Table 5 identify the key socio-economic and institutional factors influencing farmers’ engagement in strategy sharing, which facilitates the adoption of climate change adaptation strategies. The model demonstrates good statistical fit, as indicated by the Wald chi-square statistic (Wald chi² (15) = 46.50, p < 0.001), a log-likelihood value of − 310.56718, and a pseudo R-squared of 0.0697, confirming the robustness of the fractional response specification. Table 5Determinants of Climate Adaptation Adoption in Rice Farmers: Results from Fractional Response Model.Sharing StrategiesCoefficientsMarginal EffectsAge0.992(0.123)−0.002(0.002)Households size1.001(0.065)0.001(0.014)Education1.023(0.015)0.005(0.003)Experience 1.037*** (0.012) 0.008*** (0.002) Farm Size1.043(0.029)0.009(0.006)Distance to farm0.980(0.108)−0.004(0.024)Distance of Market 1.342* (0.230) 0.063* (0.037) Access Extension service0.771(0.183)−0.057(0.051)Frequency of Extension service1.134(0.116)0.027(0.022)Member of Farmer Group 2.092*** (0.572) 0.160*** (0.058) Off -farm Activity0.799(0.160)−0.048(0.043)Access credit1.197(0.243)0.039(0.044)Having Early Warning System 0.493** (0.127) −0.154*** (0.054) Prepared to Future likely climate change (CC) impacts0.946(0.279)−0.012(0.064)Primary Profession0.882(0.200)−0.027(0.049)cons 0.253** (0.154) Source: Data analysis, 2025.*, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Wald chi2 (15) = 46.50; Prob > chi2 = 0.0000; Log likelihood = − 310.56718; Pseudo R2 = 0.0697; Number of obs. = 498. Determinants of Climate Adaptation Adoption in Rice Farmers: Results from Fractional Response Model. 0.992 (0.123) −0.002 (0.002) 1.001 (0.065) 0.001 (0.014) 1.023 (0.015) 0.005 (0.003) 1.037*** (0.012) 0.008*** (0.002) 1.043 (0.029) 0.009 (0.006) 0.980 (0.108) −0.004 (0.024) 1.342* (0.230) 0.063* (0.037) 0.771 (0.183) −0.057 (0.051) 1.134 (0.116) 0.027 (0.022) 2.092*** (0.572) 0.160*** (0.058) 0.799 (0.160) −0.048 (0.043) 1.197 (0.243) 0.039 (0.044) 0.493** (0.127) −0.154*** (0.054) 0.946 (0.279) −0.012 (0.064) 0.882 (0.200) −0.027 (0.049) 0.253** (0.154) Source: Data analysis, 2025. *, **, and *** = statistical significance at the 10%, 5% and 1%, respectively. Figures in parentheses are the standard error. Wald chi2 (15) = 46.50; Prob > chi2 = 0.0000; Log likelihood = − 310.56718; Pseudo R2 = 0.0697; Number of obs. = 498. Farming experience emerged as a significant positive determinant of strategy sharing, with more experienced farmers more likely to engage in collaborative knowledge exchange due to their accumulated practical wisdom and extensive social connections3,59. Membership in farmer groups was also strongly positively associated with strategy sharing, highlighting the critical role of social networks in providing platforms for resource pooling, peer learning, and knowledge transfer78. Distance to the market showed a positive association, suggesting that farmers in more remote areas, facing limited access to formal markets and information sources, rely more on local networks for climate adaptation knowledge79. In contrast, access to early warning systems exhibited a significant negative relationship with strategy sharing, indicating that timely climate information may increase farmers’ self-reliance and reduce dependence on peer knowledge exchange80. Other variables, including education, farm size, and frequency of extension contacts, showed positive but statistically insignificant relationships, possibly because their effects are already captured by stronger determinants such as farming experience and social networks. Age, household size, off-farm activity, access to credit, and primary profession had no significant effect, suggesting that these factors play a limited role in shaping collaborative knowledge exchange in this context78,81. These findings indicate that social capital, experiential knowledge, and institutional support mechanisms are the primary drivers of collective adaptation behaviors among rice farmers. Robustness check by multinominal logistic model To validate the findings from the MVP model, a robustness check was conducted using multinomial logistic regression on five different dependent variables. Multinomial logistic regression is used to predict category membership of a dependent variable based on multiple independent variables. It is widely favored because it does not require assumptions of normality, linearity, or homoscedasticity. Although discriminant function analysis may offer greater statistical power, it is less commonly used due to its stricter assumptions. However, multinomial logistic regression requires careful consideration of sample size and potential outliers82. The robustness checks—including Log Likelihood, LR chi-square, and Pseudo R-squared values—showed varying levels of significance and model fit across strategies. These supplementary analyses reinforce the reliability of the multivariate probit model and increase confidence in the factors influencing the adoption of climate change adaptation strategies. Moreover, this alternative model specifications produced qualitatively similar associations, indicating that potential self-selection does not substantially affect the observed adoption patterns Table 6. Table 6Results of multinominal logistic regression.VariablesAPDIPMTRVIRVRSMAge −0.033** (0.013) 0.035(0.038)−0.012(0.014)−0.010(0.014)0.015(0.012)Households size0.063(0.066) 0.503*** (0.178) −0.013(0.074)0.065(0.075)−0.058(0.063)Education−0.002(0.015)0.018(0.042)0.001(0.016)0.014(0.017)0.021(0.014)Experience 0.021** (0.012) −0.033(0.031)0.008(0.012)0.004(0.013)0.011(0.011)Farm Size0.002(0.026) −0.334* (0.185) −0.051(0.038) 0.062** (0.030) −0.072** (0.029) Distance to farm 0.349*** (0.131) −1.340* (0.691) 0.039(0.119)0.079(0.121)0.090(0.106)Distance of Market−0.288(0.180)−0.898(0.740)0.082(0.262)−0.309(0.218)0.165(0.165)Access Extension service0.242(0.244)−0.688(0.639)−0.105(0.103) 0.694** (0.320) 0.001(0.232)Frequency of Extension service−0.003(0.100)−0.701(0.522)0.071(0.103)−0.017(0.106)−0.175(0.127)Member of Farmer Group0.175(0.277)−0.396(1.039)0.224(0.304)−0.270(0.314)0.426(0.261)Off -farm Activity−0.008(0.202) −1.681** (0.720) −0.273(0.228)−0.111(0.230)0.208(0.195)Access credit −0.418** (0.206) −0.240(0.609) 0.402* (0.223) 0.460** (0.226) 0.276(0.195)Having Early Warning System 1.274*** (0.247) −1.283(1.140)−0.065(0.277) 0.828*** (0.268) 0.193(0.236)Prepared to Future likely climate change (CC) impacts0.405(0.292)−16.356(2227.034)−0.372(0.342)−0.253(0.330)−0.066(0.276)Primary Profession−0.088(0.225)0.858(0.761)−0.149(0.250) 0.532* (0.277) −0.227(0.218)cons−0.153(0.614) −3.270* (1.928) −0.509(0.675) −2.733*** (0.729) −0.965(0.592)Log Likelihood−309.03284−51.09736−264.00298−254.24074−329.66216LR chi267.6832.4310.2236.8627.50Prob chi20.00000.00560.80550.00130.0249Pseudo R20.09870.24090.01900.06760.0400Number of obs498498498498498Note: *indicates significant level at 10%, ** at 5% and *** is for 1%. Results of multinominal logistic regression. −0.033** (0.013) 0.035 (0.038) −0.012 (0.014) −0.010 (0.014) 0.015 (0.012) 0.063 (0.066) 0.503*** (0.178) −0.013 (0.074) 0.065 (0.075) −0.058 (0.063) −0.002 (0.015) 0.018 (0.042) 0.001 (0.016) 0.014 (0.017) 0.021 (0.014) 0.021** (0.012) −0.033 (0.031) 0.008 (0.012) 0.004 (0.013) 0.011 (0.011) 0.002 (0.026) −0.334* (0.185) −0.051 (0.038) 0.062** (0.030) −0.072** (0.029) 0.349*** (0.131) −1.340* (0.691) 0.039 (0.119) 0.079 (0.121) 0.090 (0.106) −0.288 (0.180) −0.898 (0.740) 0.082 (0.262) −0.309 (0.218) 0.165 (0.165) 0.242 (0.244) −0.688 (0.639) −0.105 (0.103) 0.694** (0.320) 0.001 (0.232) −0.003 (0.100) −0.701 (0.522) 0.071 (0.103) −0.017 (0.106) −0.175 (0.127) 0.175 (0.277) −0.396 (1.039) 0.224 (0.304) −0.270 (0.314) 0.426 (0.261) −0.008 (0.202) −1.681** (0.720) −0.273 (0.228) −0.111 (0.230) 0.208 (0.195) −0.418** (0.206) −0.240 (0.609) 0.402* (0.223) 0.460** (0.226) 0.276 (0.195) 1.274*** (0.247) −1.283 (1.140) −0.065 (0.277) 0.828*** (0.268) 0.193 (0.236) 0.405 (0.292) −16.356 (2227.034) −0.372 (0.342) −0.253 (0.330) −0.066 (0.276) −0.088 (0.225) 0.858 (0.761) −0.149 (0.250) 0.532* (0.277) −0.227 (0.218) −0.153 (0.614) −3.270* (1.928) −0.509 (0.675) −2.733*** (0.729) −0.965 (0.592) Note: *indicates significant level at 10%, ** at 5% and *** is for 1%. The results from the multinomial logistic regression model reveal several significant findings regarding the adoption of climate change adaptation strategies. Age was negatively associated with the adoption of APD, indicating that older farmers are less likely to adjust their planting dates. Household size was positively linked to the adoption of IPM, suggesting that farmers with larger households are more likely to engage in integrated pest management. Experience showed a positive relationship with APD, indicating that more experienced farmers are more likely to adopt planting date adjustments. Farm size had a negative relationship with IPM and RSM but was positively associated with the adoption of IRV, suggesting that larger farms are more inclined to adopt improved rice varieties. Distance to the farm was positively associated with APD but negatively related to IPM, implying that farmers with farms farther from home are more likely to adjust planting dates but less likely to adopt integrated pest management. Additional factors such as access to extension services, off-farm activities, and access to credit also played important roles. Access to extension services was positively related to the adoption of IRV, indicating that farmers with extension support are more likely to adopt improved rice varieties. Off-farm activities had a negative relationship with IPM, suggesting that farmers engaged in off-farm work are less likely to adopt integrated pest management practices. Access to credit was negatively associated with APD but positively related to TRV and IRV, implying that credit access enables farmers to adopt improved rice varieties but may discourage the adjustment of planting dates. Having an early warning system was positively related to APD and IRV, suggesting that timely climate information supports the adoption of planting date adjustments and improved rice varieties. Finally, primary profession was positively linked to IRV, highlighting that farmers primarily engaged in agriculture are more likely to adopt improved rice varieties. Conclusion and policy recommendations This study examines how socio-economic characteristics, institutional support, and farm-level factors jointly influence smallholder rice farmers’ adoption and intensity of multiple climate change adaptation strategies in northwest Bangladesh. The multivariate probit results show that farmers’ age, experience, household size, access to credit, and availability of climate information significantly affect adaptation decisions. Membership in farmer groups and access to early warning systems further encourage adoption, while education, farm size, and off-farm activities show mixed or limited effects. The Multinomial Logistic Regression results confirm these findings and highlight complementarities among strategies, as adopting one practice, such as adjusting planting dates, increases the likelihood of using other practices, like improved rice varieties, thereby enhancing overall climate resilience. The Fractional Response Model further reveals that experience, distance to market, membership in farmer groups, and having early warning systems increase the intensity of strategy sharing among farmers. This indicates that both social networks and access to information are critical in spreading adaptive practices across communities. These results indicate that smallholders strategically allocate limited resources but face constraints from financial, informational, and institutional barriers. Policy measures should directly target these constraints. Expanding affordable agricultural credit and concessional loans for climate-vulnerable farmers can reduce financial limitations and encourage the adoption of complementary strategies, as shown in the Multivariate Probit results. Improving the coverage, timeliness, and local relevance of climate information services and early warning systems will enable proactive decision-making and facilitate simultaneous adoption of multiple strategies, supported by all models. Strengthening farmer groups and promoting social networks will enhance strategy sharing and the diffusion of best practices, as evidenced by the Fractional Response Model findings. Investing in farmer training programs that emphasize the benefits of bundled practices can further strengthen knowledge and confidence in complex adaptation approaches. Integrating these measures into national frameworks, such as Bangladesh’s National Adaptation Plan (NAP), can promote sustainable agricultural practices, enhance resilience to climate shocks, support rural livelihoods, and contribute to SDGs 1 (No Poverty), 2 (Zero Hunger), and 13 (Climate Action). This study has several limitations. First, some key explanatory variables may be endogenous due to self-selection, potentially biasing estimated associations; although methods such as 2SLS Probit-IV or Heckman selection models could address this, they were not applied here due to data and modeling constraints, and future research should consider these approaches. Second, the analysis focuses on rice farmers in a drought-prone region of northwest Bangladesh, limiting generalizability to other agro-ecological contexts where integrated farming practices, such as rice–aquaculture systems, may be more prevalent; while such systems are recognized as important climate adaptation strategies, their adoption in the study area is constrained by water availability and related challenges. Future research should explore integrated systems and assess the cost-effectiveness of diverse adaptation strategies to provide actionable, policy-relevant insights. Acknowledgements The authors are grateful to the rice farmers of northwestern Bangladesh for their valuable cooperation and continuous support throughout this research. Their willingness to participate and share their experiences made this study possible. Author contributions Bikash Chandra Ghosh: Conceptualization. Md Shohidul Islam, and Bikash Chandra Ghosh: Methodology, Writing-original draft preparation. Apurba Adhikary, and Md Shohidul Islam: Formal analysis, Data curation. Bikash Chandra Ghosh: Writing-Review and Editing. Bikash Chandra Ghosh: Supervision. All authors have read and agreed to the published version of the manuscript. Data availability The data used in this article will be made available upon request to the corresponding author. Declarations Competing interests The authors declare no competing interests. Ethics Approval Ethical approval to conduct this study was obtained from the Research Ethics Committee of the Department of Economics, Pabna University of Science and Technology, Rajapur, Pabna-6600, Bangladesh (Approval No. PUST/ECON/REC/136). All methods were carried out in accordance with the ethical guidelines for social science research at Pabna University of Science and Technology. Consent to Participate Informed verbal consent was obtained from all participants prior to the interviews. Participation was voluntary, and respondents were assured of anonymity and confidentiality. Consent to Publish Participants consented to the publication of anonymized and aggregated data for academic purposes. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data used in this article will be made available upon request to the corresponding author.