SPIA Bangladesh Study 2025: Updating the Green Revolution December 2025 Saumya Singla, Tanjim Ul Islam, Fuad Hassan, Isabella Monteiro, James Stevenson, Kyle Emerick Cover Photo: Women farmers growing rice. Credit: b.a.sujan/Map PhotoAgency/IRRI The Standing Panel on Impact Assessment The Standing Panel on Impact Assessment (SPIA) is an external, impartial panel of experts in impact assessment appointed by the CGIAR System Council and accountable to it. SPIA is responsible for providing rigorous, evidence-based, and independent strategic advice to the broader CGIAR System on efficient and effective impact assessment methods and practices, including those measuring impacts beyond contributions to science and economic performance, and innovative ways to improve knowledge and capacity on how research contributes to development outcomes. https://iaes.cgiar.org/spia Permissions SPIA encourages fair use of this material under the terms of Creative Commons Attribution 4.0 (BY-NC-SA 4.0), providing proper citation is made. Recommended Citation Singla, S., Ul Islam, T., Hassan, F., Monteiro, I., Stevenson, J.,Emerick, K. (2025). SPIA Bangladesh Study 2025: Updating the Green Revolution. Rome. Standing Panel on Impact Assessment (SPIA) Full replication code and data are available at: https://github.com/CGIAR-SPIA/SPIA-Bangladesh-Study-2025 Authors Saumya Singla, Tanjim Ul Islam, Fuad Hassan, Isabella Monteiro, James Stevenson, Kyle Emerick Editor Samantha Collins Design and layout Luca Pierotti Countries and Territories Disclaimer Boundaries used in the maps do not imply the expression of any opinion whatsoever on the part of CGIAR concerning the legal status of any country, territory, city, or area, or its authorities, or concerning the delimitation of its frontiers or boundaries. Border lines are approximate and cover some areas for which there may not yet be full agreement. The term country also refers, as appropriate, to territories or areas. https://iaes.cgiar.org/spia https://github.com/CGIAR-SPIA/SPIA-Bangladesh-Study-2025 SPIA Bangladesh Study 2025: Updating the Green Revolution Saumya Singla, Tanjim Ul Islam, Fuad Hassan, Isabella Monteiro, James Stevenson, Kyle Emerick i Contents Abbreviations vii Acknowledgments viii Executive Summary 1 1. Introduction 7 2. Context 8 2.1 Geographic Overview 8 2.2 Agriculture in Bangladesh’s Structural Transformation 9 2.3 Existing Hazards and Vulnerability to Climate Change 11 3. Methods and Data 12 3.1  Identifying CGIAR-Related Innovations 12 3.2 Bangladesh Integrated Household Survey 12 3.2.1 Sampling Villages and Households 13 3.2.2 Attrition 14 3.2.3 Overall Data Collection Approach 14 3.3 Measurement Approaches 15 3.3.1 Varietal Adoption 15 3.4 Data Analysis 20 3.4.1 Estimating Reach 20 3.4.2 Correlates of Adoption 21 4. CGIAR-Related Innovations in Bangladesh 22 4.1 Crop Germplasm Improvement 24 4.1.1 Rice 24 4.1.2 Wheat 25 4.1.3 Maize 26 4.1.4 Potato 26 4.1.5 Sweetpotato 27 4.1.6 Lentil 27 4.1.7 Groundnut 28 4.1.8 Chickpea 28 4.2 Aquaculture and Fisheries Improvement 29 4.3 Natural Resource Management 31 4.4 Mechanization 32 5. Insights From All Waves of the Bangladesh Integrated Household Survey 34 5.1 Crop Agriculture Trends 34 5.2 Agricultural Entries and Exits 35 5.3 Aquaculture Trends 39 ii 6. Adoption Rates of CGIAR-Related Innovations 45 6.1 Crop Germplasm Improvement 45 6.1.1 Boro Rice DNA Fingerprinting 45 6.1.2 Self-Reported Reach Estimates for Rice Across All Plots by Season 53 6.1.3 Wheat 56 6.1.4 Maize 56 6.1.5 Lentil 56 6.1.6 Potato 57 6.1.7 Sweetpotato 57 6.1.8 Groundnut 57 6.1.9 Chickpea 57 6.2 Aquaculture Innovations 58 6.3 Natural Resource Management 61 6.3.1  Irrigation 61 6.3.2 Alternate Wetting and Drying 63 6.3.3 Digital agronomic applications 66 6.4 Farm Mechanization 67 6.4.1 Axial Flow Pumps 68 6.4.2 Two- and Four-Wheeled Power Tillers 72 6.4.3 Mechanical Reapers 73 7. Who Are The Adopters? 74 7.1 Crop Varieties 74 7.2 Aquaculture Innovations 76 7.3 Natural Resource Management 76 7.4 Mechanization 78 8. Where Are The Adopters? 79 8.1 Rice Varieties 79 8.2 Aquaculture Innovations 81 8.3 Natural Resource Management 83 9. Discussion 84 10. Conclusion 87 References 88 Appendices 93 Appendix A. Details of Survey Field Implementation 93 Appendix B. List of Rice Varietal Releases 94 Appendix C. Bioinformatic Analysis of the Genotyped Boro Rice Samples 103 Appendix D. Details of Wheat Varietal Releases 104 Appendix E. Details of Maize Varietal Releases 106 iii List of Figures Figure 1: Number of rural households adopting each CGIAR-related innovation in Bangladesh in 2024, in millions 3 Figure 2: Map of Bangladesh showing eight administrative divisions 8 Figure 3: Natural hazard co-occurrence by upazila 11 Figure 4: Distribution of BIHS upazilas (sub-districts) 21 Figure 5: Distribution of hatcheries and upazila sales data for GIFT tilapia seed 30 Figure 6: Incidence of cultivation of major crops in Bangladesh (2011-2024) 34 Figure 7: Time trend in households participating in agriculture in Bangladesh (2011- 2024) 35 Figure 8: Proportion of agricultural households in 2018-19 (left panel), proportion of net agricultural exits in 2024 (right panel) 36 Figure 9: Foreign remittances as a main earning source by division (%) 37 Figure 10: Percentage of aquaculture households across the Panel Waves 39 Figure 11: Average number of ponds per household (left panel), average size of pond per household (right panel) 40 Figure 12: Distribution of ponds per household in 2024 40 Figure 13: Average number of harvests per household (left panel), average harvest per household (kgs) (right panel) 41 Figure 14: Aquaculture households (%) by division across the panel 42 Figure 15: Household adoption by fish breed (% of fish-cultivating households) 43 Figure 16: Average tilapia and rohu harvest per household (kg/annum) 44 Figure 17: Crop cultivation of agricultural households in the past three seasons (%) 45 Figure 18: Adoption of major Boro varieties for DNA fingerprinting plot in 2023-24 (%) 47 Figure 19: Mixing of Varieties on DNA Fingerprinting Plot 48 Appendix F. Details of Potato Varietal Releases 107 Appendix G. Details of Sweetpotato Varietal Releases 110 Appendix H. Details of Lentil Varietal Releases 111 Appendix I. Details of Groundnut Varietal Releases 112 Appendix J. Details of Chickpea Varietal Releases 113 Appendix K. Unassigned Samples from Rice DNA Fingerprinting and Their Characteristics 114 Appendix L. Farmer Self-Reported Data on Certainty of Responses and Reliability and Quality of Varieties Cultivated 116 Appendix M. Further Information on Irrigation Sources 117 Appendix N. Definition of Variables Used in Analysis of Household Correlates of Adoption 119 Appendix O. Covariates of Adoption for Additional Mechanization Innovations 120 Appendix P: Covariates of Adoption for Additional CGIAR Variety Innovations 122 iv Figure 20: Variety-wise division of ‘local’ and ‘improved’ self-reports with corresponding self-reported variety names 49 Figure 21: DNA fingerprinting results (left side) vs. self-reported data (right side) comparison for major Boro varieties 50 Figure 22: Misclassification of DNA fingerprinting samples. Left panel: Scatterplot of false negatives vs release year; right panel: false negatives by variety 51 Figure 23: Correlates of misclassification of DNA fingerprinted rice 52 Figure 24: Percentage of aquaculture households raising small indigenous species of fish (2015-2024) 59 Figure 25: Percentage of rainfed plots (2015-2024) 61 Figure 26: Primary water source for percentage of agricultural plots in the dry season 62 Figure 27: Primary fuel source for irrigated plots in the dry season (%) 63 Figure 28: Water management practices of plots in the Boro season in 2024 64 Figure 29: Duration and frequency of drying cycles on irrigated Boro plots in 2024 64 Figure 30: Divisional distribution of AWD practices (left panel) and water shortages (right panel) on Boro plots in 2024 65 Figure 31: Duration and frequency of drying cycles on irrigated Boro plots in 2024 66 Figure 32: Agricultural technology adoption trends over time, 2015-2024 67 Figure 33: Rented vs owned agricultural technology in 2024 68 Figure 34: Axial Flow Pump Usage Over Time (2015-2024), National 68 Figure 35: Axial flow pump usage in Barisal and Chittagong vs other divisions in Bangladesh 69 Figure 36: Axial flow pump adoption by division (2024 vs 2018) 70 Figure 37: Axial flow pump usage across Feed the Future and non-Feed the Future zones 71 Figure 38: 2-Wheeled Power Tiller usage over time across Feed the Future and non-Feed the Future zones (2015-2024) 72 Figure 39: Covariates for CGIAR-related Boro varieties (left panel) and CGIAR-related varieties for all non-rice crops (right panel) 75 Figure 40: Covariates of adoption of G3 rohu strain (left panel) and GIFT tilapia (right panel) 76 Figure 41: Covariates of adoption of Alternate Wetting and Drying 77 Figure 42: Covariates of adoption of axial flow pumps 78 Figure 43: District-wise share of CGIAR-related Boro rice varieties 79 Figure 44: District-wise share of BR-28 and BR-29 (left panel) and more recent CGIAR-related releases (post 2005, right panel) 80 Figure 45: Self-reported Boro rice variety in 2018-19 at the plot level (%) 81 Figure 46: Three distinct geographic distributions of aquaculture innovations 82 Figure 47: Proportion of households using AFPs at the division level (as a share of all households that apply irrigation in 2024) 83 Figure 48: Self-reports for 'Not-Assigned' samples (in reference list) 114 Figure 49: Self-Reports for ‘not assigned’ samples (not in reference list) (%) 115 Figure 50: Primary pump type (% of agricultural households irrigating in the Boro season) 117 Figure 51: Histogram of plots using an axial flow pump in 2024 (share of plots on x axis, share of villages on y axis) 118 v Figure 52: Division-wise distribution of households using both axial flow pump and tubewells in 2024 118 Figure 53: Covariates of adoption for 2-wheel and 4 wheel power tillers 120 Figure 54: Covariates of adoption for shallow tubewells 121 Figure 55: Covariates of adoption for 2-wheel and 4 wheel power tillers 122 List of Tables Table 1: BIHS sample villages per division 13 Table 2: Household weight calculation for SPIA-BIHS 2024 14 Table 3: Snapshot of stocktake exercise 23 Table 4: Number of rice varieties released, by germplasm origin, 1970-2022 25 Table 5: Number of wheat varieties released, by germplasm origin, 1990-2019 25 Table 6: Number of maize varieties released, by germplasm origin, 1980-2019 26 Table 7: Number of potato varieties released, by germplasm origin, 1990-2019 27 Table 8: Number of sweetpotato varieties released, by germplasm origin, 1980-2019 27 Table 9: Number of lentil varieties released, by germplasm origin, 1990-2019 28 Table 10: Number of groundnut varieties released, by germplasm origin, 1980-2019 28 Table 11: Number of chickpea varieties released, by germplasm origin, 1980-2019 28 Table 12: Comparison of wealth measures for households remaining in vs exiting agriculture using t-test 38 Table 13: Comparison of wealth measures for households outside agriculture vs entering agriculture using t-test 38 Table 14: Comparisons between households remaining, entering, and exiting aquaculture within 2018 and 2024 43 Table 15: DNA fingerprinting results reach estimates for rice in Boro 2023-24 for the PPS sampling selected plot 47 Table 16: Comparison of reach estimates between DNA fingerprinting and self-reported varieties in Boro 2023-24 for the PPS sampling plot 48 Table 17: Combined DNA fingerprinting plot and self-reported other plots’ reach estimates for Boro rice 53 Table 18: Comparison of major Aman variety adoption between the SPIA-BIHS survey and Kretzschmar et al. (2018) 54 Table 19: Self-reported reach estimates for rice for Aman season across all plots 55 Table 20: Self-reported reach estimates for rice for Aus season across all plots 55 Table 21: Self-reported reach of crop germplasm innovation 58 Table 22: Reach estimates for improved fish strains in the past one year 59 Table 23: Reach estimates for improved fish strains in the past one year 60 vi Table 24: Reach estimates for Alternate Wetting and Drying (1 plot drying cycle of at least 5 days) 66 Table 25: Reach estimates for CGIAR-related mechanization innovations 73 Table 26: Summary table on reach of CGIAR-related innovations in Bangladesh, 2024 86 Table 27: Rice varietal releases 94 Table 28: Wheat varietal releases 104 Table 29: Maize varietal releases 106 Table 30: Potato varietal releases 107 Table 31: Sweetpotato varietal releases 110 Table 32: Lentil varietal releases 111 Table 33: Groundnut varietal releases 112 Table 34: Chickpea varietal releases 113 Table 35: Definition of variables used in analysis of household correlates of adoption 119 vii Abbreviations AFP Axial Flow Pump AWD Alternate Wetting and Drying BARC Bangladesh Agricultural Research Council BARI Bangladesh Agricultural Research Institute BFRI Bangladesh Fisheries Research Institute BIHS Bangladesh Integrated Household Survey BINA Bangladesh Institute of Nuclear Agriculture BRRI Bangladesh Rice Research Institute CIAT International Center for Tropical Agriculture CIMMYT International Maize and Wheat Improvement Center CIP International Potato Center CRP CGIAR Research Program CSISA Cereal Systems Initiative in South Asia DATA Data Analysis and Technical Assistance FAO Food and Agriculture Organization of the United Nations GIFT Genetically Improved Farmed Tilapia ICARDA International Center for Agricultural Research in the Dry Areas ICRISAT International Crops Research Institute for the Semi-Arid Tropics IFPRI International Food Policy Research Institute IRRI International Rice Research Institute IWMI International Water Management Institute ILRI International Livestock Research Institute NARES National Agricultural Research and Extension Systems NARS National Agricultural Research Systems NRM Natural Resource Management PANI Program for Advanced Numerical Irriagtion SIS Small Indigenous Species viii Acknowledgments We are grateful to the many people who have contributed to this report. Dr Humnath Bhandari, IRRI Country Representative for Bangladesh has been unceasingly supportive in hosting SPIA’s team in the country and providing advice on how we should approach this work. We are grateful to Dr Akhter Ahmed, Dr Mehrab Bakhtiar, and Julie Ghostlaw in allowing us to add an additional survey wave to the Bangladesh Integrated Household Survey and thank them for the work they have done to curate and protect this international public good over the past fifteen years. Professor Sattar Mandal did a wonderful job of facilitating our consultation workshop in 2023, which fed directly into the design of the study. We are similarly grateful to all participants in that workshop, but particularly the members of the Steering Committee, namely Professor Karen Macours (former SPIA Chair), Dr Tim Krupnik (CIMMYT Country Representative) and Temina Lalani-Shariff (CGIAR Regional Director, South Asia). We thank the directors and staff of Data Assistance and Technical Assistance (DATA) for their flexibility in adapting the BIHS survey to meet the specific needs and unusual features that SPIA introduced. Finally, we thank all the Bangladeshi citizens who graciously gave their time to participate in responding to the survey questions. 1 SPIA Bangladesh Study 2025: Updating the Green Revolution Executive Summary This report presents the results of a comprehensive study on the adoption and diffusion of CGIAR-related agricultural innovations in Bangladesh. We focus on innovations across diverse CGIAR research avenues: crop germplasm, aquaculture and fisheries, climate change adaptation, digital apps, natural resource management innovations, and mechanization. This is the fifth country report, following Ethiopia (2018-19 data), Ethiopia (2021-22 data), Uganda (2021-22 data) and Vietnam (2022-24 data). To produce this report, SPIA commissioned a new survey wave of the Bangladesh Integrated Household Survey (BIHS), focused primarily on the country’s agriculture. The BIHS is a nationally representative panel dataset with previous survey waves implemented in 2011-12, 2015, and in 2018-19. It was originally designed by the Bangladesh Office of the International Food Policy Research Institute (IFPRI) and implemented by the survey firm Data Analysis and Technical Assistance (DATA). In late 2023, SPIA reached an agreement with the International Food Policy Research Institute (IFPRI) to manage BIHS Wave 4 independently. SPIA led the survey design and analysis, in consultation with IFPRI. This wave followed the same sample of households surveyed in Wave 3. The 1960s and 1970s marked the establishment of the first global agricultural research centers that now form the CGIAR network. Centers like the International Rice Research Institute (IRRI) and the International Maize and Wheat Improvement Center (CIMMYT) quickly made significant contributions to crop development in Bangladesh, particularly in rice and wheat. Fifty years after the Green Revolution, farmers in Bangladesh can now access a wide range of new agricultural technologies developed by both domestic and international sources. IRRI began working in Bangladesh in the early 1970s to boost rice production, which was essential for the nation's food security. Since then, this initial narrow focus on rice has diversified significantly. Today, five CGIAR centers – CIMMYT, IFPRI, the International Potato Research Center (CIP), WorldFish, and IRRI – are embedded within Bangladesh’s agricultural research landscape, working in close coordination with local institutions. In addition, Bangladesh's national agricultural research system has also collaborated with other CGIAR centers, including the Alliance of Bioversity International and CIAT, the International Water Management Institute (IWMI), and the International Livestock Research Institute (ILRI). Prior to data collection, we conducted an extensive stocktaking exercise to map CGIAR activity in Bangladesh. This exercise delineated the scope of CGIAR research initiatives in the country from 1970 to 2022 and identified key areas to guide the prioritization of data collection on CGIAR-related developments. The stocktaking process involved three interrelated phases: desk- based research, interviews with stakeholders, and review/consultation. The resulting stocktake includes 64 CGIAR-related innovations and 57 instances where CGIAR research is reported to have influenced government or development institution policies. A common pathway for such policy influence is the scaling up of government pilot programs following CGIAR-led impact evaluations. In other cases, CGIAR expertise has helped shape policy frameworks at both regional and national levels. From the initial list of 64 CGIAR-related innovations, we identified several that we considered high priority and presented these for discussion during a consultation workshop in Dhaka in 2 SPIA Bangladesh Study 2025: Updating the Green Revolution July 2023. Based on the consultation feedback, the list was refined to 46 innovations that were believed to be widely adopted and could be measured. These became the focus of data collection in the BIHS. While some were already a part of the BIHS, others necessitated the development of specialized measurement protocols. A novel feature of this wave was the collection of paddy leaf samples during the Boro season for DNA fingerprinting. Several potential DNA fingerprinting candidates – such as lentil and wheat – were omitted because relevant data had already been collected in recent CGIAR surveys. These included lentil (Yigezu et al. 2022), wheat (Gade et al. 2021), and Aman rice (Kretzschmar et al 2018). Given that previous BIHS rounds had already collected rich panel data across a broad range of topics, Wave 4 deliberately prioritized agriculture-related modules to generate more detailed information for studying innovation adoption. As a result, a shorter version of the main BIHS questionnaire was administered. Drawing on our data collection, we estimated the potential reach of these innovations among Bangladeshi agricultural households. The total reach across all innovations is estimated to range from 8.05 to 9.37 million. We present a range of reach estimates to reflect varying levels of confidence in attributing adoption to CGIAR involvement. This approach allows us to distinguish innovations with clear, direct CGIAR contributions from those involving broader dissemination efforts. For example, in the case of germplasm, sophisticated DNA fingerprinting techniques enabled us to confidently assess CGIAR’s role. Similarly, for aquaculture, we could identify strains with a direct CGIAR contribution. We have considered these in the lower bound. In contrast, attribution is more challenging for innovations such as agronomic practices or digital tools, where multiple actors are involved and diffusion pathways are less easily traced, and hence they are considered in the upper bound of CGIAR’s reach. In Figure 1, the large reach observed for rice is owed to rice being the staple crop in Bangladesh, cultivated by around 80% of the agricultural households. Notably, the success of rice germplasm dissemination remains a cornerstone of CGIAR’s impact, as validated through the DNA fingerprinting exercise conducted during the Boro season. Other crops, by contrast, have a reach of under a million households, covering about 1-5% of agricultural households in the BIHS. Around 24% of agricultural households are engaged in aquaculture, with a noticeable trend toward consolidation in commercially viable species such as rohu, tilapia, and carp. As for CGIAR’s reach, G3 rohu has already reached over 200,000 households despite its recent release. 3 SPIA Bangladesh Study 2025: Updating the Green Revolution Figure 1: Number of rural households adopting each CGIAR-related innovation in Bangladesh in 2024, in millions Adoption of CGIAR Technologies in Bangladesh 0 1 2 3 4 5 5.5Rice varieties: Boro season Rice varieties: Aman season Alternate Wetting and Drying Axial Flow Pump Rice varieties: Aus season Wheat varieties GIFT Tilapia Two-wheeled mechanical reaper Small Indigenous Species G3-Rohu Maize varieties Potato varieties Peanut varieties Lentil varieties 3.5 1.7 1.6 0.4 0.3 0.2 0.2 0.2 0.2 0.1 0.1 0.1 0.1 6 Million Note: Dark blue bars reflect lower-bound reach numbers, light blue bars reflect upper-bound reach numbers. Insights from the DNA fingerprinting exercise, combined with self-reported data, reveal important patterns in varietal turnover. Newer stress-tolerant and micronutrient-enriched rice varieties have yet to fully replace the older varieties. The weighted year of release shows that varieties are on average 20 years old. This suggests that while CGIAR-developed varieties are widely adopted, older releases – particularly BRRI Dhan 28 and 29 – continue to dominate. Interestingly, those who are indeed cultivating newer varieties often report them as BRRI Dhan 28 and 29. Investigations of seed bag images collected during the survey revealed that these newer varieties are often marketed as older varieties. This may reflect a strategy by seed companies to leverage the stronger brand recognition of legacy varieties, like BRRI Dhan 28, thereby making it easier to sell newer ones. It could also reflect quality issues during the seed multiplication process. While varietal replacement has been slow, some recent varieties have spread fast, including BRRI Dhan 74, 89, and 100. These results show why there have been efforts by numerous stakeholders to replace BRRI Dhan 28 and 29 with newer varieties (Lojo, n.d.)There is considerable heterogeneity in household and plot-level characteristics among adopters of CGIAR-related innovations, with different factors driving adoption across innovation types. For crop germplasm, adoption seems to be concentrated among less wealthy households. In the case of Boro rice, these include households with lower savings, and in the case of non- rice crops like wheat, maize, lentil, groundnut, and potato, those with a smaller farm size and no outside employment. CGIAR-related aquaculture innovations, like G3 rohu and Genetically Improved Farmed Tilapia (GIFT), are broadly accessible to households engaged in aquaculture, 4 SPIA Bangladesh Study 2025: Updating the Green Revolution without strong differentiation. In the case of Alternate Wetting and Drying (AWD1), the physical characteristics of the plots suggest that farmers practice AWD for those that have higher irrigation demands and pumping costs. Similarly, larger plots located further from the main pump or canal are more likely to be irrigated using an axial flow pump (AFP). The geographic distribution of adopters for different innovations is concentrated in different parts of the country, showing that CGIAR reach is spread across the whole country. For instance, the Boro rice adopters are predominantly concentrated in central areas, i.e., Dhaka and Sylhet divisions, largely driven by the continued dominance of older varieties, BRRI Dhan 28 and 29. In contrast, newer varieties released after 2005 have higher levels of dispersion across the country, possibly reflecting a successful shift in targeting efforts towards more diverse and peripheral regions. In aquaculture, the distribution of GIFT tilapia and G3 rohu adopters is relatively geographically balanced. Households reporting GIFT tilapia fingerling purchases are concentrated in regions such as Sylhet, Dhaka, and parts of Chittagong. Despite its recent release, G3 rohu has already gained traction in Rangpur and Sylhet, where 11-15% of all reported fingerling purchases were G3. AWD of Boro plots has proliferated at a higher rate in Barisal and Mymensingh than in other districts. Similarly, AFPs, which were primarily tested and disseminated by CIMMYT in Barisal and Chittagong, also show a higher adoption in these divisions. This report underscores the importance of continued efforts in the country to unpack the impact of CGIAR-related innovations. To build on these initial findings, upcoming efforts, such as the use of DNA fingerprinting in aquaculture, will enable more precise measurement. Looking ahead, SPIA will continue to leverage the panel structure of the BIHS through 2030 to further deepen the understanding of the reach and impact of CGIAR innovations in Bangladesh. 1 It is important to clarify that by AWD, we refer to farmers who reported drying their fields at least once for a period of five days. This differs from the IRRI-defined AWD method, which involves using a 'field water tube' or plastic pipe to monitor water depth. Farmers who use the pipe represent a small subset of those who report drying their plots. A middleman offers maize and wheat grain and seed for sale at a market in Nur Islam, Dinajpur, Bangladesh. Credit: S. Mojumder/Drik/CIMMYT Regeneration of wheat wild relatives under screenhouse conditions in CIMMYT, to have enough healthy and viable seed for distribution when necessary. Credit: Rocio Quiroz/CIMMYT SPIA Bangladesh Study 2025: Updating the Green Revolution 7 1. Introduction Bangladesh has long been a high-priority country for CGIAR. This is reflected in numerous active research programs in the country, involving most CGIAR centers. Having identified Bangladesh as suitable for a SPIA country-level study, we began work on the stocktaking exercise for this report in June 2022, carrying out desk research and interviewing CGIAR researchers and external stakeholders. In July 2023, in Dhaka, we held a consultation workshop with CGIAR research colleagues and representatives from government ministries and the Bangladesh Agricultural Research Council (BARC). We then laid the plans for the survey fieldwork that was carried out between January and June 2024 and reported on here. This report is the fifth in a series following Ethiopia (2018-19 data), Ethiopia (2021-22 data), Uganda (2021-22 data) and Vietnam (2022-24 data). Data collection concluded in early 2024, just before the onset of widespread political unrest in Bangladesh. Beginning in June, student- led protests escalated into a nationwide movement that culminated in a change of government in August 2024. The findings in this report, therefore, reflect conditions immediately prior to this major political and economic disruption. In this report, we aim to estimate the reach of CGIAR-related agricultural innovations in Bangladesh and to understand the geographic distribution and socioeconomic correlates of their adoption by households nationwide. It is organized as follows. Section 2 introduces the Bangladeshi context, focusing on agriculture in the context of structural transformation of the economy, and the ongoing vulnerabilities the country faces from natural hazards, exacerbated by climate change. Section 3 presents the methods used to identify and measure the adoption of CGIAR-related innovations among households in the Bangladesh Integrated Household Survey (BIHS) sample, via a special stand-alone survey round administered by SPIA in 2024. Section 4 provides details of the stocktaking process we undertook to identify agricultural innovations believed to be operating at scale in Bangladesh, and which helped us set data collection priorities. In Section 5, we use earlier waves of the BIHS to examine secular trends in the country’s agricultural development prior to our main fieldwork. Section 6 provides our adoption estimates for CGIAR-related innovations using the BIHS (2024). For some innovations, we also examine changes in adoption from earlier waves of the BIHS, going back to 2012. In Section 7, we ask, “Who are the adopters?”. This is to determine whether innovations reach subpopulations that are of particular interest to CGIAR. Section 8 documents where these adopters are located. In Section 9, we discuss our results and suggest priorities for future data collection efforts, prior to our conclusion, set out in Section 10. SPIA Bangladesh Study 2025: Updating the Green Revolution 8 2. Context 2.1 Geographic Overview Bangladesh is relatively young as a nation state, gaining independence from post-colonial Pakistan in the liberation war of 1971. It is surrounded by neighboring India, apart from a small border with Myanmar’s mountainous and forested regions to the south-east. Many rivers - principally the Padma, the Jamuna, and the Meghna - bring Himalayan glacial snowmelt as well as rain deposited by the monsoon down from the north of Bangladesh and out into the Bay of Bengal. Much of the land in this riverine country is in the Ganges Delta - the largest delta system in the world, and one of the most densely populated regions anywhere on Earth. The nation’s capital, Dhaka, is situated within the Ganges River Floodplain. Its greater metropolitan area is home to over 22 million people. 2 Figure 2: Map of Bangladesh showing eight administrative divisions2 2 Armanaziz, Peter Fitzgerald, Cacahuate, JamesA, CC BY-SA 4.0. SPIA Bangladesh Study 2025: Updating the Green Revolution 9 Bangladesh is divided into eight major administrative divisions, each with its own unique landscape that influences agricultural practices, crop varieties, fish species prevalence and overall productivity. The Dhaka Division has fertile alluvial soils that are formed by the deposition of sediments carried by rivers and streams. These are ideal to support rice paddy production, as well as fruits and vegetables, with the added advantage of proximity to urban markets. Chittagong is a low-lying coastal area categorized as a Coastal Saline Zone. It is home to the country's busiest seaport and the Chittagong Hill Tracts, which is a hilly and forested region. Saline intrusion influences agricultural selections in both coastal regions. Barisal is often called ‘the Land of Rivers’ due to its dense network of waterways, while Khulna has the largest mangrove forest in the world. The northeastern region of Sylhet is renowned for its tea plantations and picturesque landscapes, while the northern regions of Rajshahi and Rangpur are prone to drought conditions. In 2015, Mymensingh was constituted as the newest division, having been segregated from Dhaka. Each division is further administratively subdivided into districts (or zilas). Bangladesh currently consists of 64 districts. There are 400 sub-districts (upazilas) which are the administrative entities within districts. Unions, or union councils, constitute the most basic rural administrative entities that further segment each upazila, with over 4,500 nationwide. 2.2 Agriculture in Bangladesh’s Structural Transformation The 1960s and 1970s saw the founding of the first of the global agricultural research centers that now make up CGIAR. The International Rice Research Institute (IRRI) and the International Maize and Wheat Improvement Center (CIMMYT) quickly made significant advances in crop development in relation to rice and wheat that have relevance for Bangladesh. Fast forward fifty years from the days of the Green Revolution, and Bangladeshi farmers now have access to a wide range of new agricultural technologies from several different domestic and international sources. IRRI started working with Bangladesh in the early 1970s to increase rice production, which was essential for the nation's food security. This initially narrow focus on rice has diversified over the decades since with five CGIAR centers – CIMMYT, the International Food Policy Research Institute (IFPRI), the International Potato Research Center (CIP), WorldFish, and IRRI – having offices in the country, and through close collaboration with regional organizations. Bangladesh's national agricultural research system has also collaborated with other CGIAR centers such as the Alliance of Bioversity International and CIAT3, the International Water Management Institute (IWMI), and the International Livestock Research Institute (ILRI). In the 1970s, Bangladesh faced widespread famine due to its rapidly growing population and limited food grain availability per person. In recent decades, pressures on its food system have eased as the country has seen sustained growth in rice production. Average yields have risen from approximately one tonne per hectare in the early 1970s to three tonnes per hectare by 2013 (Gautam and Faruqee, 2016) while the population increased 2.3-fold over this same 3 Bioversity International and the International Center for Tropical Agriculture (CIAT) merged in 2020 to become the Alliance of Bioversity International and CIAT. SPIA Bangladesh Study 2025: Updating the Green Revolution 10 period, from 67.5 million in 1970 to 154.0 million in 20134. This is lower than the yield increase, thus avoiding the Malthusian trap, where population growth outpaces the growth of food production and resources. While this is no mean achievement, challenges remain. The Food and Agriculture Organization of the UN (FAO) listed Bangladesh as one of 36 member states classified as “protracted food crisis countries” (FAO, 2024). The prevalence of undernourishment in the population in 2021-23 was estimated at 11.9%, only modestly lower than the 13.7% estimated for 2004-06 (FAO, 2024). Progress on food security appears similarly challenging. 30.5% of the population in 2021-23 were considered either moderately or severely food insecure, modestly lower than the 32.2% estimate for 2014-16. However, for context, the South Asian nations of Afghanistan, Nepal, Pakistan, and Sri Lanka all saw substantial increases in these measures over the same period. The Government of Bangladesh updated their policy guiding government action in these areas from the National Food Policy (2006) to the National Nutrition Policy (2015) through to the National Food and Nutrition Security Policy (NFNSP, 2020). The plan of action for the NFNSP recognizes a greater role for the private sector in shaping food security and nutrition outcomes, and aims for nutrition-sensitive food systems. Since independence in 1971, economic growth rates have steadily risen. The estimated rate of economic growth of 4% year-on-year average for 1990 to 2020 puts the country comfortably in the top 10% of performers globally. The country’s “recovery from financial turmoil, increasing trade openness, and more foreign direct investment” (Beyer and Wacker, 2022, p.24) are governance factors associated with its strong performance, led by its extensive ready-made garment industry. In terms of the contribution of agricultural progress to underpinning structural transformation and dynamic economic growth in recent decades, Emran and Shilpi (2018) highlight some descriptive statistics using three rounds of the Household Income and Expenditure Survey (HIES) from 2000, 2005, and 2010. During the period 2000-2010, rice yields grew by 3.8% annually, accompanied by both an increase in agricultural wages and a reduction in the amount of hired labor. Sen et al (2021) examine the phenomenon of agricultural exits, or more accurately, the de-prioritization of agriculture in the economic life of rural Bangladesh. They find that non-farm orientation increased over 2000-2013, and that household members are increasingly engaged in salaried work. The implication is that structural transformation is not entirely driven by mass permanent migration to cities, but rather that rural areas, particularly peri-urban rural areas, are changing rapidly. Aquaculture has boomed in Asia, and Bangladesh in particular, since 2000. This growth is driven by strong domestic demand and underwritten by technological innovations and a major deepening in the capitalization of the value chain at all levels (Hernandez et al, 2018). The sector’s importance is reflected in policy instruments such as the National Aquaculture Development Strategy and Action Plan of Bangladesh (2013 – 2020). In Bangladesh, most aquaculture production is destined for domestic consumption, particularly by urban dwellers who consume aquaculture products at higher rates per capita than those in rural areas (Toufique and Belton, 2014). Commercial aquaculture – operations where produce is sold – now dwarfs subsistence fish farming. Hernandez et al (2018) find a 207% growth in hatcheries over ten years, far outstripping the growth in the number of fish farmers (63% increase) or total output (117% increase). As the 4 https://data.worldbank.org/. https://data.worldbank.org/ SPIA Bangladesh Study 2025: Updating the Green Revolution 11 authors note, this trend strongly suggests a major “shift to purchased seed” (p.461), providing opportunities for technological change through new, improved fish strains. Over the same period, the number of feed mills jumped from around 8 in 2004 to approximately 100 by 2014, and the number of feed dealers approximately doubled. 2.3 Existing Hazards and Vulnerability to Climate Change Agricultural development in Bangladesh faces multiple, compounded climate-related hazards. These include river floods, coastal floods, tropical cyclones, drought, heat stress, landslides, and air pollution. Figure 3 shows the extent to which any given upazila (subdistrict) is exposed varies across the country, but from a global perspective, Bangladesh is high on every list in terms of vulnerability to the projected temperature and sea level rises over the coming decades. Climate change adaptation has thus been an important framing concept for agricultural research in Bangladesh since at least 2011. Figure 3: Natural hazard co-occurrence by upazila Note: The color-coding indicates how many out of seven natural hazards – riverine floods, coastal floods, heat stress, drought, tropical cyclones, landslides, and air pollution – each upazila ranks in the highest decile for exposure to. Source: United Nations Office for Coordination of Humanitarian Affairs (2020), in World Bank Country Climate and Development Report (2022). SPIA Bangladesh Study 2025: Updating the Green Revolution 12 3. Methods and Data 3.1  Identifying CGIAR-Related Innovations Before collecting data, we produced an extensive stocktake of CGIAR activity in Bangladesh. This exercise delineated the scope of CGIAR Research Initiatives in the country from 1970- 2022 and documented pertinent information to aid in prioritizing data collection for CGIAR- related developments. Stocktaking involves three interrelated phases: desk-based research, interviews with stakeholders, and review/consultation. We retrieved information from the CGIAR Results Dashboard for 2017-20215, and from the updated 2022 version, which aims to track real-time progress to 20306. We reviewed reports from the CGIAR Research Programs (CRPs) that ran from 2011 to 2021, Annual Reports from the Challenge Programs, as well as bilateral/center-specific project reports. The goal was to construct the universe of CGIAR-related innovations from the past two decades, as well as record claims of policy influence. In addition, 15 interviews were conducted with impact assessment focal points, country representatives, CGIAR, and National Agricultural Research System (NARS) scientists. Following the completion of an initial draft, we conducted a consultation workshop in Dhaka on 10-11 July 2023, with stakeholders from ministries, National Agricultural Research System (NARS) scientists and CGIAR researchers, and management. The primary objectives were to gather feedback about the stocktake, to identify the priority innovations therein, and to discuss possible data sources for supporting our understanding of technology dissemination and adoption. This resulted in a longlist of 64 innovations, from which we prioritized 46 innovations for data collection in the BIHS. These form the basis of this report. We also collect 61 examples of possible policy influence that we include in the stocktake for future investigation, but that are not suited for interrogation through a household survey. The full stocktake table is posted here, and we include some example entries in Section 4. 3.2 Bangladesh Integrated Household Survey Following the 2023 consultation workshop, SPIA determined that the best course of action for data collection was to commission a new survey wave of the Bangladesh Integrated Household Survey (BIHS). The BIHS is a nationally representative panel dataset. Three previous survey waves were implemented in 2011-12, 2015, and in 2018-19, designed by the Bangladesh Office of the International Food Policy Research Institute (IFPRI) and implemented by a contracted survey company, Data Analysis and Technical Assistance (DATA). SPIA approached IFPRI about the possibility of carrying out an independently managed wave of the BIHS, and an agreement was reached in late 2023. For BIHS Wave 4, SPIA would be responsible for all aspects of the design and analysis, but would consult with IFPRI on design issues. DATA would remain the survey company, allowing for continuity for the surveyed households. 5 https://www.cgiar.org/food-security-impact/results-dashboard-2017-2021/. 6 https://www.cgiar.org/food-security-impact/results-dashboard/. https://osf.io/cunf9 https://www.cgiar.org/food-security-impact/results-dashboard-2017-2021/ https://www.cgiar.org/food-security-impact/results-dashboard/ SPIA Bangladesh Study 2025: Updating the Green Revolution 13 3.2.1 Sampling Villages and Households The survey follows a stratified sample design selected in two stages using the sampling frame developed from the community series of the population census 2001. As there were seven divisions in Bangladesh when the first round took place, each of them formed a separate stratum7. At the first stage, 275 villages (our Primary Sampling Units - PSUs) were selected with probability proportional to the number of households in the village. The distribution of villages per division is listed in Table 1. Table 1: BIHS sample villages per division Division Number of sample villages Barisal 21 Chittagong 48 Dhaka 87 Khulna 27 Rajshahi 29 Rangpur 27 Sylhet 36 Total 275 In the second sampling stage, IFPRI researchers used a linear systematic sampling8 scheme to select 20 households from each cluster with an equal probability. Thus, the total sample size for the first BIHS round in 2011/12 was 5,500 households. In subsequent rounds, the original households may have split (and some have attrited). The BIHS follows both the original and the split households. The measure of village size that was originally used for selecting the PSUs is the total number of households as of the 2001 population census. The current value of this size measure could have changed significantly; therefore, we updated the number of households at the selected PSU level. For the SPIA BIHS 2024 round, we use the population census of 2022. Survey weights should not vary widely within a division/stratum, as a wide variation of weights within a division/stratum could unnecessarily increase the variances of the estimates. Base weight is the inverse of the selection probability of an ultimate sampling unit (household). To ensure similar or uniform weight within a division/stratum, Table 2 illustrates the procedure of computing sampling weights for the national survey. The final weight is calculated by dividing the predicted number of households in 2024 by the number of observations in 2024 for each division. 7 The Mymensingh division was only officially formed as being distinct from the Dhaka division in 2015, so the BIHS surveys are not statistically representative of this division. Households in the territory of the current Mymensingh division are included in the sample but form part of the Dhaka division sample. 8 Linear systematic sampling is a method where samples aren’t repeated at the end, and ‘n’ units are selected to be a part of a sample having ‘N’ population units. Here, a skip pattern is created following a linear path. The selection process follows a predetermined sequence, such as selecting every 5th member, then every 7th member, then every 9th member, and so on. SPIA Bangladesh Study 2025: Updating the Green Revolution 14 Table 2: Household weight calculation for SPIA-BIHS 2024 Division Total households in 2001 Total households in 2022 (per 2022 census) Growth rate Predicted household number in 2024 Number of observations in 2024 Adjusted final weight of SPIA- BIHS round Barisal 1,410,100 1,663,967 0.79% 1,677,136 419 4,002.71 Chittagong 3,326,980 4,922,971 1.88% 5,015,693 991 5,061.24 Dhaka 5,357,120 8,408,048 2.17% 8,590,477 1670 5,144 Khulna 2,468,280 3,383,245 1.51% 3,434,428 551 6,233.08 Rajshahi 2,972,460 4,127,351 1.58% 4,192,371 613 6,839.11 Rangpur 2,690,360 3,530,386 1.30% 3,576,365 555 6,443.9 Sylhet 1,209,260 1,783,477 1.87% 1,816,783 755 2,406.34 Notes: The figures in Columns 2 and 3 are taken from the census data of the Bangladesh Bureau of Statistics. The total number of households in 2001 and 2022 is the village-level population for each division. The growth rate is the yearly growth rate calculated from the figures in Columns 2 and 3. The predicted household number is calculated by multiplying the total number of households in 2022 by the growth rate. The adjusted final weight is calculated by dividing the predicted household number by the number of observations in 2024. 3.2.2 Attrition The 5,503 original households from the 2011/12 Wave are considered for attrition calculations. For split households (those in which members have left since the prior wave), if at least one of the newly created split households consented and completed a survey, the original household is considered not to have been subject to attrition. For the period 2015-2018, there was an anomaly: 92 original households (of which five had split, thus 97 households in total) were not surveyed successfully in 2015 but were surveyed successfully in 2018. We have not considered these 92 households as having been subject to attrition. The attrition for each wave is calculated based on the number of original panel households in the previous round, and the percentages are as follows: BIHS 2015 =2.74, BIHS 2018-19= 4.37, and SPIA-BIHS 2024 = 5.43. 3.2.3 Overall Data Collection Approach BIHS Wave 4 followed the same sample households from Wave 3. A novel feature of Wave 4 was the collection of leaf samples from rice paddy during the Boro season9. To accommodate the additional survey burden of SPIA’s specific interests in agricultural innovations, we administered a shorter version of the main BIHS questionnaire compared to prior rounds. For example, the sub- module pertaining to land preparation, fertilizer and pesticide use, and the sub-module on input costs and returns were only asked for one randomly selected plot for all agricultural households. We added necessary questions for assessing the adoption of the relevant CGIAR innovations from the stocktake as outlined in the following sections. All Boro-cultivating households were visited, and leaf samples were taken for DNA fingerprinting to identify the varieties farmers are cultivating. We employed probability proportional to size (PPS) sampling to randomly select a plot from all Boro plots cultivated by the household, and questions were asked for that plot. This was done to ensure that there was no bias in selecting 9 The Boro season is one of the main rice-growing seasons in Bangladesh and parts of South Asia. It refers to the period when Boro rice is cultivated, typically from January to April/May. SPIA Bangladesh Study 2025: Updating the Green Revolution 15 the plot based on enumerator preference, owing to the distance of different plots under operation by the household. For households cultivating Boro paddy, the sub-modules on DNA fingerprinting, on land preparation, fertilizer and pesticide use, and one on input costs and returns were asked for the same randomly selected plot. If an agricultural household had no rice plots in Boro, then we took the total agricultural land area in the past four seasons and randomly selected a plot based on PPS to be used as the focal plot for our sub-modules on land preparation, fertilizer, pesticide use, and input costs and returns. We then took the plot boundaries using GPS for the selected Boro paddy fields to obtain an accurate estimate of the plot area. SPIA was an active partner during survey implementation. SPIA instituted a process of high- frequency checks and audited data quality using random audio audits of segments of the interviews. Resurveys were carried out when data collection was found to be insufficiently rigorous. Further details can be found in Appendix A. 3.3 Measurement Approaches 3.3.1 Varietal Adoption During the July 2023 stakeholder consultation, we discussed measurement priorities, identifying groups of CGIAR-related innovations with the potential for large-scale adoption. Within the crop germplasm improvement group, rice varieties were identified as the top priority for DNA fingerprinting. In Bangladesh, rice is the principal crop and is grown during three distinct seasons: Boro (January to April/May - also known as the Rabi season); Aus (March- June); and Aman (JuneJuly to October/November). Most Aus and Aman rice is rainfed or uses limited supplemental irrigation, whereas Boro rice requires consistent irrigation due to minimal rainfall during the Rabi period. An earlier study by Kretzschmar et al (2018) applied DNA fingerprinting to varietal identification in the Aman rice season. To complement this study, we decided instead to focus our DNA fingerprinting efforts on the Boro season. This meant that for Aman varietal adoption (and the relatively minor Aus rice season), and varietal adoption in wheat, lentil, maize, potato, and sweetpotato, we would rely on farmer self-reported data. The lack of a nationally representative DNA fingerprinting effort on estimating the adoption of these other non-rice crops reflects their much less widespread cultivation compared to rice. Rice is cultivated by around 80% of agricultural households in our survey, whereas the value for the non-rice crops ranges from 1-5%. There is also evidence to suggest that farmer self-reported data for lentil is quite accurate, given the small number of varieties used by farmers (Yigezu et al., 2022) obviating the need for the more expensive and complicated DNA fingerprinting data collection. Details of the rice varietal releases for all seasons, including the complete set of reference materials used for DNA fingerprinting in the Boro season, are detailed in Appendix B. The Bangladesh Rice Research Institute (BRRI) has released 115 rice varieties in total, while the Bangladesh Institute of Nuclear Agriculture (BINA) has released 27. The relationship between the International Rice Research Institute (IRRI) and BRRI has been strong for several decades (IRRI, n.d.), so we have 115 varieties that are CGIAR-related. Most of these varieties exhibit one or more of three key traits: higher yield, stress tolerance (drought, flood, salt), and higher levels of micronutrients due to biofortification. The prior BIHS survey wave (Wave 3, 2018) has SPIA Bangladesh Study 2025: Updating the Green Revolution 16 self-reported data from farmers on rice varieties up to the BRRI Hybrid 4 and BRRI Dhan 69 series. We expanded the varietal adoption options to include varieties released since 2018, up to BRRI Hybrid Dhan 8 and BRRI Dhan 106. 3.3.1.1 Field Protocol for Rice Leaf Collection To allow for consistent interpretation, we designed the protocol for leaf collection using the approach developed for Vietnam (Kosmowski et al., 2024). SPIA procured all necessary supplies for collecting leaf samples, including rolls of adhesive barcode labels, 50 ml plastic pots, silica gel, leaf punches, alcohol-based wipes, and sample-holding bags. To assemble the kit for enumerators, we developed 3,000 unique barcodes for labeling the plastic containers into which the leaf samples would be deposited. Each 50 ml plastic pot was filled with 21 grams of dried silica gel in packets. The silica gel ensures the correct drying of the punched leaves. SPIA conducted two on-farm pilots of the plant tissue collection modules in January 2024 to help ensure smooth fieldwork operations. Enumerators identified the plot for sample collection based on the random sampling protocol programmed into SurveyCTO – a comprehensive data collection platform used for designing, deploying, and managing surveys. Enumerators were instructed to take leaf samples from healthy plants within the plot, meaning that there were no obvious signs of disease on the leaves. This was to ensure that the tissue had a high DNA concentration. Enumerators were instructed that sampled plants should be away from the edge of the plot, preferably taken by walking three steps into the interior of the field. The next step was to excise a young, growing leaf and fold it three times to create four layers of leaf tissue. The enumerator then placed the front side of the leaf punch and the leaf within the plastic pot for the sample and punched through all four layers, thereby creating four leaf discs and ensuring that all the leaf discs fall inside the pot. After punching, the enumerators sealed the pot tightly and immediately scanned the barcode when prompted by the SurveyCTO CAPI application. The enumerators then used alcohol wipes to thoroughly clean the leaf punch before beginning another sample collection. 20% of the rice plots were selected for technical replication, meaning that the process was repeated for a second plant, resulting in two distinct individual samples for each plot. Once the samples were taken and scanned, they were shipped at an ambient temperature and stored in IRRI’s Bangladesh office. Following the completion of fieldwork, SPIA recruited research assistants to work at the IRRI Bangladesh office to transfer the leaf tissue samples from the barcoded pots to 96-well plates used by genotyping laboratories. The research assistants scanned the barcodes and ensured correct placement of the material in the sample wells using the Coordinate app.10 SPIA researchers then sealed the filled 96-well plates using heat-sealing foil and prepared them for shipping to the genotyping laboratory in October 2024. 3.3.1.2 Reference Library Compilation SPIA consulted with the Genetic Resources and Seed Division of BRRI and the Plant Breeding Division of BINA to construct the rice varietal reference library. Both supplied 5 mg of breeders’ rice seed, with 114 BRRI varieties and 15 BINA cultivars requested. This reference material was 10 https://excellenceinbreeding.org/toolbox/tools/coordinate. https://excellenceinbreeding.org/toolbox/tools/coordinate SPIA Bangladesh Study 2025: Updating the Green Revolution 17 specified to be healthy rice seed devoid of pests and fungi. We received reference samples of 99 rice varieties from BRRI and 11 from BINA and placed each variety in an airtight plastic container. The discrepancy between the 129 varieties requested and the size of our final reference library (110) is explained by the fact that BRRI and BINA could only provide us with the varieties that they had in their stores, the remaining 19 being unavailable. The full list is provided in Appendix B. The reference rice seeds were dried for 24 hours then were meticulously peeled, preserved, and their plastic containers labeled with barcodes . Before grinding, all the viable seeds from BRRI and BINA underwent additional screening for pests (weevils in particular). We employed a small sample grinder to pulverize the rice seeds and meticulously disinfect the apparatus with ethanol between reference samples to prevent cross-contamination. Rice flour from each reference sample was placed into the 96-well plates, securely sealed, and their locations systematically documented using the Coordinate application. Each well in the plates was given 0.66 ml of rice flour. The final, packaged reference library received a phytosanitary certificate from the Plant Protection Division of the Department of Agricultural Extension before it was shipped to AgriPlex Genomics in the United States of America for processing alongside the leaf samples obtained through on-farm sampling in the BIHS. 3.3.1.3 Sample Genotyping Reference and field samples that were shipped to AgriPlex Genomics were received in good condition11 and subjected to DNA extraction. The field samples comprised 2,202 leaf tissue samples, while the submitted references comprised 110 seed samples ground to a fine flour. These were genotyped on the PlexSeqTM platform, using the IRRI Rice Custom Amplicon Version 4 SNP panel (Arbelaez et al., 2019) – a mid-density panel with 1,024 markers (AgriPlex Genomics, Cleveland, OH, USA). 3.3.1.4 Bioinformatic Analysis After genotyping, Agriplex Genomics delivered the genotype data in score format, indicating the type of single-nucleotide base present at each marker position. Bioinformatic analysis entailed comparing the similarity between the samples and references at each marker position and generating a genetic distance matrix using the Identity by State (IBS) method. This enabled the identification of samples that were closely related to the references. A detailed description of the bioinformatic analysis carried out on Boro rice varieties is provided in Appendix C. 3.3.1.5 Crop Varietal Adoption: Non-Rice Crops In partnership with the various auspices of the Bangladesh Agricultural Research Council (BARC), CGIAR has collaborated on crop germplasm leading to varietal releases for wheat, maize, groundnut, lentil, potato, and sweetpotato. Details of those varieties are provided in Appendices D to J. Farmers self-report varietal adoption by variety name, following the results of Yigezu et al (2022). 11 Sample processing was delayed by approximately two months. Initially, the procured plates and seals did not fit together tightly. SPIA then procured heat-sealing foil and shipped it to Bangladesh for the team to use. In the process of heat-sealing, some of the plasticware on the plates melted slightly around the top, confounding the robotics used by AgriPlex and requiring their technician to remove some excess plastic with a razor in a painstaking manual process. The samples were not affected by the process as we had been very conservative with our protocol, ensuring each sample was desiccated individually in their own pots. Thus, DNA recovery rate from this exercise remained high. SPIA Bangladesh Study 2025: Updating the Green Revolution 18 3.3.1.6 Aquaculture Innovations Our stocktake featured 14 aquaculture innovations, including genetically improved strains, polyculture systems that incorporate nutrient-rich small fish, community-managed fisheries, and digital platforms that enhance market linkages along the aquaculture value chain. Of these, we identified eight that have been successfully scaled or have the potential for large-scale adoption to be the focus of data collection. Genetically improved rohu carp and Genetically Improved Farmed Tilapia (GIFT) have a system of hatcheries that multiply these strains to make them widely available to aquaculture farmers. To assess the reach of these genetic advancements, we rely on self-reported farmer data on fish strains harvested from each pond. We also collected data on fingerling sources and hatcheries. Additionally, we conducted interviews with the most prominent fish dealer in each village’s largest market, as well as with hatchery and nursery managers. To estimate the adoption of WorldFish-promoted improved pond management practices, we also gathered detailed data on practices at a pond level. In capture fisheries, WorldFish’s efforts in Bangladesh have primarily focused on conserving, restoring, and managing Hilsa populations, particularly through Phases One and Two of the Enhanced Coastal Fisheries (ECOFISH) project. To assess the reach and sustainability of these initiatives, four years after project completion, our survey includes questions on awareness of fishing bans, regulations governing sanctuary fishing, the presence of fish guards, and compensation mechanisms for livelihood disruptions caused by these restrictions. Small indigenous fish species (SIS), such as Mola, are rich in micronutrients and can be regularly harvested and consumed. WorldFish researches and promotes the use of carp-Mola polyculture systems with an emphasis on household nutritional outcomes. To capture the adoption, the household survey includes a question on whether SIS are cultivated alongside carp, ensuring continuity with previous rounds of the BIHS. To evaluate the integration of digital platforms into the aquaculture sector, our survey includes a dedicated section on application usage, adoption timelines, and perceived business benefits. These insights will help assess how digital technologies enhance market connectivity and whether they contribute to improved efficiency and profitability for fish farmers. 3.3.1.7 Natural Resource Management Innovations CGIAR research projects have addressed a range of natural resource issues in Bangladesh over the past two decades. The stocktake examined 13 Natural Resource Management (NRM) innovations either arising from this body of research or promoted by CGIAR researchers while carrying out research for development interventions. Among these, Alternate Wetting and Drying (AWD) was identified as having been promoted at scale. AWD irrigation technology was introduced in Bangladesh in 2004 as an IRRI initiative, with the first trials conducted in 2005 by the Bangladesh Rice Research Institute (BRRI) in collaboration with CGIAR. In AWD, irrigation water is applied a few days after the ponded water disappears, creating alternating flooded and non-flooded conditions. The non-flooded period can range from 1 to over 10 days, depending on factors such as soil type, weather, and crop growth stage. AWD is of interest to CGIAR primarily from the perspective of its potential to reduce greenhouse gas emissions from rice production, as interruption to continuous flooding limits the generation of methane-producing SPIA Bangladesh Study 2025: Updating the Green Revolution 19 bacteria. Farmers will interpret and implement AWD differently, so it is fair to consider it as a management principle rather than a specific practice (Stevenson et al, 2019). IRRI and partners found a practical way to implement AWD safely by using a perforated PVC pipe to monitor water depth below the surface of the soil. Irrigation water can then be applied when the water level drops to 15 cm below the soil surface. This is a good rule of thumb as a safe limit, as below this, the standing rice crop could suffer yield losses. It is then reflooded to a depth of 5 cm above the soil surface. During the critical flowering stage, the field should remain flooded with periodic topping up, while during the grain-filling and ripening stages, the water level can again be allowed to drop to 15 cm below the surface before re-irrigation. To estimate AWD adoption, we employed a multi-pronged approach. First, we asked farmers whether they were familiar with the practice. Second, we inquired whether they had adopted it. Third, we gathered detailed information on the frequency and duration of drying their rice plots. These estimates help assess whether farmers are practicing AWD even if they do not use the PVC pipe. Additionally, we asked about decision-making processes related to irrigation, including who makes these decisions, when they are made, and whether farmers participate in village- level irrigation committees. Finally, we mapped plot boundaries using GPS to attempt adoption estimation through remote sensing. Beyond AWD, we designed survey questions to assess the adoption of various NRM-related mobile apps developed and disseminated through CGIAR Research Programs. These include Right Haat, Digital Feed Supply, Program for Advanced Numerical Irrigation, Rice Crop Manager, and Macher Gari. To understand broader app usage trends, we also inquired about the use of popular consumer apps in Bangladesh, such as Bkash, Nagad, and Rocket, to determine whether households engage with mobile applications in other areas of their lives. Additionally, we asked about the primary purposes of app usage and whether these technologies have contributed to improving their businesses. CGIAR has also worked in the sphere of index-based insurance through WorldFish. The first activities relating to index-based insurance in Bangladesh pertain to WorldFish’s role in the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS). We asked households about their knowledge of crop and livestock insurance and their participation in index-based flood insurance schemes. 3.3.1.8 Farm Mechanization The Government of Bangladesh has consistently promoted mechanization as a means for fostering agricultural intensification towards achieving agricultural self-sufficiency (Mainuddin & Kirby, 2015). SPIA’s consultation meetings held in Dhaka in 2023 identified several priority innovations regarding mechanization that CGIAR has played a role in promoting, particularly through fostering both the rental market for mechanization services and the supporting services that underpin that offer. Focus innovations for the promotion of specific business models, particularly in the Feed the Future12 zone under the Cereal Systems Initiative in South Asia (CSISA)13 include axial flow pumps (AFPs), power tillers, and mechanical reapers. The survey 12 Feed the Future is a USAID Initiative that aims to reduce global hunger and poverty by improving agriculture, nutrition, and resilience in developing countries. https://www.feedthefuture.org/. 13 https://csisa.org/tag/feed-the-future/. https://www.feedthefuture.org/ https://csisa.org/tag/feed-the-future/ SPIA Bangladesh Study 2025: Updating the Green Revolution 20 examines usage, ownership, participation in rental markets, associated costs, and year of purchase/ sale. Furthermore, it specifically investigates the reasons behind farmers’ decisions to adopt or discontinue a particular technology and whether they integrate these technologies with other equipment. To ensure accurate responses, a visual aid comprised of images of different machines was incorporated into the survey operation. The visual aid helped farmers correctly identify the technologies they used, helping to reduce measurement error. 3.4 Data Analysis 3.4.1 Estimating Reach In Section 6, we present the adoption rates for key CGIAR innovations in both agriculture and aquaculture, based on data from the BIHS sample of rural households. Adoption statistics are reported at the household, village, and division levels. All figures incorporate sampling weights to ensure national-level representativeness. To capture the extent of CGIAR’s direct contribution, we organize innovations into two distinct categories for the purposes of estimating their reach: those that make it into a conservative lower-bound estimate and those that are only in an upper-bound estimate. For each innovation discussed, we draw on information from our stocktake to determine whether its dissemination can be attributed to CGIAR. For example, in the case of rice germplasm, we identified varieties that are CGIAR-related through parentage and used DNA fingerprinting of Boro rice plots to estimate their adoption. This conservative definition of CGIAR's contribution, based on embodied technologies – those in which the research contribution is captured in a product used by farmers – gives us the lower bound of CGIAR’s contribution. For other innovations, such as Alternate Wetting and Drying (AWD) and axial flow pumps (AFPs), we acknowledge that CGIAR had a role in their design and/or dissemination, but is only one of several actors who have promoted these innovations. It is important to clarify that by AWD, we refer to farmers who reported drying their fields at least once for a period lasting five days. This differs from the International Rice Research Institute (IRRI)-defined AWD method, which involves the use of a ‘field water tube’ or plastic pipe to monitor water depth. Given that adoption may not be solely attributable to CGIAR’s efforts, we classify these innovations under the upper-bound category to reflect a broader, shared contribution. As outlined in Section 3.2, BIHS weights were constructed based on the 2022 population census, enabling household-level statistics to align with the rural household population of Bangladesh and thus providing accurate estimates of reach. BIHS’s representativeness is depicted in Figure 4 where we can see the spread of the upazila (sub-districts) across the country. We are depicting this map at the sub-district level because this is the most granular level of location data we can disclose as per the Institutional Review Board (IRB) protocols. Within these 257 sub-districts, BIHS covers 275 villages. DNA fingerprinting data for rice were collected for 98.3% of BIHS households cultivating rice during the 2024 Boro season.14 Together, rural households across all surveyed regions represent 93% of Bangladesh’s rural population, offering a reliable approximation of national adoption rates and the 14 The households without leaf samples were identified as Boro households during the resurveyed sub-sample period described in Appendix A, when the Boro season had already ended. SPIA Bangladesh Study 2025: Updating the Green Revolution 21 number of households using CGIAR germplasm for rice. For CGIAR innovations measured in the 2012, 2015, 2018, and 2024 BIHS surveys, we report nationally representative changes in adoption over the past 12 years. The consistent survey interval and extended timeframe allow us to examine both agricultural trends (Section 5) and changes in adoption rates (Section 6), though only for the subset of innovations that had already been measured in prior survey rounds. While BIHS datasets for 2012, 2015, and 2018 are publicly accessible, BIHS 2024 will be available with the release of this report. For each innovation, the proportion of adopters was calculated over the population deemed most relevant for the innovation’s application. Figure 4: Distribution of BIHS upazilas (sub-districts) Note: Total number of upazilas = 257 3.4.2 Correlates of Adoption A set of variables was selected to describe the plot-level and household-level characteristics of households adopting CGIAR technologies (reported in Section 7). Innovations adopted by fewer than 5% of households were omitted from this analysis. Detailed definitions of each variable are available in Appendix O. Continuous variables were standardized, and an asset index was created based on the first principal component of all household assets that the household owned, as reported in BIHS 2018. To assess the impact of the chosen variables on the adoption rate, a multivariate regression model was constructed with village-level clustering to reflect clustered sampling at the village level. SPIA Bangladesh Study 2025: Updating the Green Revolution 22 4. CGIAR-Related Innovations in Bangladesh Here we describe the CGIAR-related innovations for which adoption data are collected. We first defined the universe of CGIAR research through a desk review of multiple sources and interviews with CGIAR researchers. This stocktake methodology aims to comprehensively account for all CGIAR research activities over the last two decades. We define innovation in CGIAR research as any new technology, practice, decision-support tool, or policy/institutional design that required research engagement for its development and/or dissemination and was previously unknown to its users. CGIAR-related initiatives for developing and/or disseminating these innovations encompass CGIAR Research Programs (CRPs) as well as bilaterally funded projects, whose research outputs contributed to advancing or spreading the innovation. We further categorized innovations across five primary domains of CGIAR research: crop varietal enhancement, aquaculture and fisheries, natural resource management, mechanization, and cross-cutting research. We used discussions with CGIAR scientists and government partners to identify innovations that were thought to be adopted on a large scale. Using the framework summarized in Table 3, for each innovation (Column 1), we began by describing the CGIAR-related initiatives responsible for its development and/or dissemination (Column 2), followed by a description of the innovation (Column 3), whether it has an observable feature that would allow us to measure adoption in a survey setting (Column 4), the scope and geographic context of activities to promote the innovation (Column 5), and prior evidence that would support the claim that the innovation has been adopted at scale (Column 6). Table 3 shows an excerpt of the stocktake for two innovations as examples – Genetically Improved Farmed Tilapia (GIFT) and axial flow pumps (AFPs). Table 3: Snapshot of stocktake exercise Innovation CGIAR-related efforts for development and/or dissemination Description Observable feature Scale and location Evidence of adoption Genetically Improved Farmed Tilapia (GIFT) Breeding efforts by WorldFish with BFRI since 1994. Every year (starting in 2001), a new generation of GIFT is produced by WorldFish. Specific projects to promote or develop include: 1. Scaling Systems and Partnerships for Accelerating the Adoption of Improved tilapia Strains Small-Scale Fish Farmers (SPAITS) Project 2. Quality broodstock and Tilapia Breeding Nucleus (TBN) genetic selection program (2005 – 2015) The current generation is G17 with much higher weight gain. Dissemination years: 1994, 1996, 1997, 2005, 2008, 2009, and 2012 1. Development of communication products and dissemination – posters, factsheets, and training manuals in Bangla; Training of 500 tilapia farmers, training of tilapia hatchery owners, and other actors Household has farmed hatchery-produced GIFT- derived tilapia. Identification using fish seed source (current study) or DNA fingerprinting (for future rounds) 1. Rangpur, Khulna, and Mymensingh Divisions. 1. In 2016, 2,360,000 improved GIFT fry were produced and distributed/ sold from these TBNs among 47 tilapia hatcheries in the Rangpur, Jessore, Narail, Fardidpur, Khulna and Barisal regions Axial Flow Pump (AFP) CSISA-MI: CIMMYT and International Development Enterprises (iDE), BARI and BARC (2013-18) Developed Bangladeshi prototype AFPs and compared hydraulic, energetic, and economic performance of AFPs and conventional centrifugal pumps (CEN). Supported the development of the service provider/rental market for AFPs Household has used an AFP on at least one plot. Visual- aid protocol to be used. Southern Bangladesh (Khulna, Patuakhali, Bagerhat). Approx. 1,017 AFP service providers developed. 47,808 farmers reported to have been reached through the AFP service providers. Acronyms = Bangladesh Agricultural Research Council (BARC), Bangladesh Agricultural Research Institute (BARI), Bangladesh Fisheries Research Institute (BFRI), Cereal Systems Initiative for South Asia - Mechanization and Irrigation (CSISA-MI), International Maize and Wheat Improvement Center (CIMMYT), Tilapia Breeding Nucleus (TBN). SPIA Bangladesh Study 2025: Updating the Green Revolution 24 The stocktake comprises 64 CGIAR-related innovations and 61 cases where research is considered to have influenced the policies of government and/or development institutions. A common route for claimed CGIAR policy influence is via the expansion of government pilot programs following CGIAR-led impact evaluations. In other cases, CGIAR expertise has shaped policy frameworks at regional or national levels. From the 64 CGIAR-related innovations, we identified several that we considered high priority and discussed them in detail at the consultation workshop in Dhaka in July 2023. We then reduced the list to 46 innovations that were thought to be widely adopted and that would be the focus of the data collection activities in the BIHS. Some were already part of the BIHS, others required minimal adjustments, and a few needed specialized measurement protocols. Several potential DNA fingerprinting study opportunities were omitted because relevant data had already been collected in another recent CGIAR survey, namely Yigezu et al. (2022) for lentil, Gade et al. (2021) for wheat, and Kretzschmar et al. (2018) for Aman rice. The balance of 18 innovations were either not considered likely to have been adopted at scale or were too challenging to meaningfully collect data as they lacked a clear CGIAR ‘signature’ that could be observed in a survey. This section outlines the innovations that met the inclusion criteria for each core domain. Additionally, we highlight excluded technologies when discussing future research priorities. 4.1 Crop Germplasm Improvement The largest group of CGIAR-related innovations we identified fall into this category. They include rice (5 innovations), wheat (3), maize (3), potato (4), sweetpotato (5), lentil (3), groundnut (2), and chickpea (4). To describe the CGIAR contribution to breeding varieties, we use innovation to refer to a specific trait or a cluster of traits. Therefore, each innovation may have a single variety associated with it, or may have several, as outlined in the following sub- sections. 4.1.1 Rice Following the initial decades of progress in rice improvement during the Green Revolution era, rice breeding in Bangladesh in the 1990s turned towards disease- and insect-tolerant/resistant varieties using IRRI parent lines, introducing 31 such varieties. There has been considerable work in developing stress-tolerant varieties in this century (particularly for flood, salinity, and drought). Sixteen salt-tolerant varieties have been developed by the Bangladesh Rice Research Institute (BRRI) (12) and the Bangladesh Institute of Nuclear Agriculture (BINA) (4). BRRI has also released eight drought-tolerant varieties. Since 2010, BRRI (4) and BINA (3) have released seven submergence-tolerant varieties with the sub-1 gene. In 2004, HarvestPlus began its efforts to address to develop biofortification of rice in collaboration with BRRI and the International Rice Research Institute (IRRI). In 2013, BRRI succeeded in developing and releasing the first-ever biofortified-zinc rice variety in the world, BRRI Dhan 62, through HarvestPlus support. Since then, ten such zinc-enriched varieties have been released. In addition, short-duration (13) and lodging-tolerant (11) rice varieties have been released by BRRI and BINA. In total, 139 rice varieties have been released (Table 4), designed for one or SPIA Bangladesh Study 2025: Updating the Green Revolution 25 more of Bangladesh’s growing seasons – Boro, Aus, and Aman. The complete list of varieties is provided in Appendix B. Table 4: Number of rice varieties released, by germplasm origin, 1970-2022 Germplasm origin 1970- 1979 1980- 1989 1990- 1999 2000- 2009 2010- 2019 2020- 2023 Total Pure line selection BRRI/BINA 1 1 4 1 16 1 24 IRRI/HarvestPlus 9 13 15 14 49 15 115 Total 10 14 19 15 65 16 139 Acronyms = Bangladesh Rice Research Institute (BRRI), Bangladesh Institute of Nuclear Agriculture (BINA), International Rice Research Institute. 4.1.2 Wheat In the early 2000s, wheat breeding addressed stress (intensifying heat and salinity) and disease problems (leaf blight and leaf rust tolerance; Ug 99-tolerance) in the Bangladeshi wheat crop. In the latter part of the 2010s, zinc-enriched wheat varieties were introduced. Leaf-blight- and leaf-rust-tolerant varieties were introduced through five varieties (Gourob, 1998; Satabdi (also known as BARI Gom 21), 2000; Bijoy, 2005; BARI Gom 28, 2012; and BARI Gom 31, 2017). Salt-tolerant varieties were developed jointly by BARI’s Wheat Research Center (WRC) and the International Maize and Wheat Improvement Center (CIMMYT), leading to the release of BARI Gom 25 in 2010. Heat- and blast-tolerant varieties were introduced by crossing three introduced wheat varieties. Four varietal releases meet this description: Sufi / BARI Gom 22, 2005; Prodip/ BARI Gom 24, 2005; BARI Gom 26, 2010; BARI Gom 30, 2014. Ug99, the deadly wheat stem rust, was a sustained target for breeding through the partnership between CIMMYT and WRC, as well as being a core theme of the activities under the Cereal Systems Initiative for South Asia (CSISA). Four Ug 99-tolerant wheat varieties were released: BARI Gom 27, 2008; Francolin, 2012; BARI Gom 29, 2014; BARI Gom 31, 2017. Finally, following a 2016 outbreak of the mysterious wheat blast disease, BARI Gom 33 was released in 2017. This variety is blast-resistant and zinc-enriched. In total, 33 wheat varieties have been released since 1974 (Table 5). The complete list of varieties is provided in Appendix D. Table 5: Number of wheat varieties released, by germplasm origin, 1990-2019 Germplasm origin 1970- 1979 1980- 1989 1990- 1999 2000- 2009 2010- 2019 Total Pure line selection BARI 0 0 0 0 1 1 CIMMYT 11 5 3 4 9 32 Total 11 5 3 4 10 33 Acronyms = Bangladesh Agricultural Research Institute (BARI), Bangladesh Institute of Nuclear Agriculture (BINA), International Wheat and Maize Improvement Center (CIMMYT). SPIA Bangladesh Study 2025: Updating the Green Revolution 26 4.1.3 Maize The earliest maize work from the 1980s to the early 2000s focused on improved maize hybrids derived from CIMMYT populations, leading to ten varietal releases (Suvra, 1986; Barnali, 1986; BARI Maize 5, 1998; BARI Maize 6, 1998; BARI Maize 7, 2002; BARI Hybrid Maize 2, 2002; BARI Hybrid Maize 6 and 7, 2006; BARI Hybrid Maize 8 and 9, 2007). Aiming to address nutritional outcomes, protein-enriched maize was introduced with the release of BARI Hybrid Maize 3 in 2002 and BARI Hybrid Maize 5 in 2004. Stress-tolerant varieties for heat, drought, and salt have also been released. These include several heat-tolerant varieties developed by the Heat Tolerant Maize for Asia (HTMA) project with BARI and CIMMYT as collaborators, namely BARI Hybrid Maize 14, 2017; BARI Hybrid Maize 15, 2017; and BARI Hybrid Maize 17, 2019. Drought-tolerant varieties were selected from a CIMMYT trial and later single-crossed by BARI researchers, leading to BARI Hybrid Maize 12, 2016, and BARI Hybrid Maize 13, 2016. A salt- tolerant maize hybrid was produced by BARI single crossing a hybrid that had been selected from a CIMMYT trial, namely BARI Hybrid Maize 16, 2016. In total, 25 maize varieties have been released since 1980 (Table 6). The complete list of varieties is provided in Appendix E. Table 6: Number of maize varieties released, by germplasm origin, 1980-2019 Germplasm origin 1980-1989 1990-1999 2000-2009 2010-2019 Total Pure line selection BARI 1 1 2 0 4 CIMMYT 2 2 11 6 21 Total 3 3 13 6 25 Acronyms = Bangladesh Agricultural Research Institute (BARI), International Wheat and Maize Improvement Center (CIMMYT). 4.1.4 Potato The International Potato Center (CIP) and the Tuber Crops Research Centre (TCRC) of BARI have worked towards developing improved potato varieties in Bangladesh since the 1990s. Stress-tolerant varieties include heat- and saline-tolerant potatoes. Heat-tolerant potatoes include BARI ALU-01 (HEERA), released in 1990; BARI ALU-12 (Dheera), released in 1993; and BARI ALU-73, released in 2016. Varieties bred for their saline tolerance are BARI ALU -22 (SAIKAT), released in 2004; BARI ALU-79, released in 2017; and BARI ALU-72, released in 2016. Virus- and disease-resistant potatoes were also developed. BARI ALU-81, released in 2019, was bred to be virus-resistant. The Root and Tuber Crops Research and Development Programme for Food Security in the Asia and Pacific Region (2011-15) developed late blight- resistant potatoes: BARI ALU-46, released in 2013, and BARI ALU-53, released in 2014. Potato varieties grown from true potato seeds were released in 1997 (BARI TPS-1 and 2). Finally, table potatoes: BARI ALU-78, released in 2017; BARI ALU-87, released in 2019; BARI ALU- 88, released in 2019. In total, 93 potato varieties have been released in Bangladesh since the 1990s (Table 7). The complete list of varieties is provided in Appendix F. SPIA Bangladesh Study 2025: Updating the Green Revolution 27 Table 7: Number of potato varieties released, by germplasm origin, 1990-2019 Germplasm origin 1990-1999 2000-2009 2010-2019 Total Pure line selection TCRC BARI 12 14 51 77 CIP 5 1 10 16 Total 17 15 61 93 Acronyms: Bangladesh Agricultural Research Institute (BARI), International Potato Center (CIP), Tuber Crops Research Centre (TCRC). 4.1.5 Sweetpotato Orange-fleshed sweetpotato varieties were first introduced to Bangladesh in the 1980s by TCRC BARI, while the 2000s marked the first decade in which varieties with CIP germplasm (BARI SP-6 and 7 in 2004) began being released in Bangladesh (Table 8). The Scaling Up Sweetpotato Through Agriculture and Nutrition (2013-2019) Program introduced the following, orange- fleshed varieties: BARI SP-12 and 13, released in 2013; and BARI SP-14 and 15, released in 2017. Yellow-fleshed sweetpotatoes include BARI SP-8 and 9, released in 2008, and BARI SP 10 and 11, released in 2013. The complete list of varieties is provided in Appendix G. Table 8: Number of sweetpotato varieties released, by germplasm origin, 1980-2019 Germplasm origin 1980-1989 1990-1999 2000-2009 2010-2019 Total Pure line selection TCRC BARI 4 1 0 0 5 CIP 0 0 4 6 10 Total 4 1 4 6 15 Acronyms: Bangladesh Agricultural Research Institute (BARI), Tuber Crops Research Centre (TCRC), International Potato Center (CIP). 4.1.6 Lentil The International Center for Agricultural Research in the Dry Areas (ICARDA) has contributed to lentil breeding efforts in Bangladesh with germplasm developed under partnerships with both BARI (the BARI Masur series of varieties) and BINA (the BINA Masur series). Disease- resistant lentils include stemphylium-blight-resistant and rust-, foot-, and root-rot-resistant varieties. Lentils with stemphylium-blight resistance were released, and in later years (from 2010), involving the HarvestPlus program, resulting in BARI Masur 4, 1996; BARI Masur 5, 2006; BARI Masur 6, 2006; BARI Masur 7, 2011; and BARI Masur 8, 2015. BARI Masur 4 to 8 are also micronutrient-enriched (with zinc and/or iron). CGIAR-related lentil varieties with rust-, foot-rot-, and root-rot-resistance are BARI Masur 1, 1991; BARI Masur 2, 1993; BINA Masur 4, 2011; BINA Masur 5, 2011; BINA Masur 6, 2011; and BINA Masur 7, 2013. Finally, a drought- tolerant lentil targeting the North-Western drought-prone districts was released as BINA Masur 10, 2016. In total, 19 lentil varieties have been released since 1990 (Table 9). The complete list of varieties is provided in Appendix H. SPIA Bangladesh Study 2025: Updating the Green Revolution 28 Table 9: Number of lentil varieties released, by germplasm origin, 1990-2019 Germplasm origin 1990-1999 2000-2009 2010-2019 Total Pure line selection BARI/BINA 1 3 3 7 ICARDA/HarvestPlus 3 2 7 12 Total 4 5 10 19 Acronyms: Bangladesh Agricultural Research Institute (BARI), Bangladesh Institute of Nuclear Agriculture (BINA), International Center for Agricultural Research in the Dry Areas (ICARDA). 4.1.7 Groundnut The International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) has released four short-duration groundnut varieties: BARI Chinabadam-6 released in 1998; BARI Chinabadam-8 released in 2006; BARI Chinabadam-9 released in 2009, and BARI Chinabadam-10 released in 2016. In 2004, they also released one high-yielding groundnut variety: BARI Chinabadam-7. In total, 14 groundnut varieties have been released since 1980 (Table 10). The complete list of varieties is provided in Appendix I. Table 10: Number of groundnut varieties released, by germplasm origin, 1980-2019 Germplasm origin 1980-1989 1990-1999 2000-2009 2010-2019 Total Pure line selection BARI 2 1 4 2 9 ICRISAT 0 1 3 1 5 Total 2 2 7 3 14 Acronyms: Bangladesh Agricultural Research Institute (BARI), International Crops Research Institute for the Semi- Arid Tropics (ICRISAT). 4.1.8 Chickpea BARI and ICRISAT have released improved chickpea varieties that include those that are disease-tolerant and disease-resistant, high-yielding, and stress-tolerant. Improved varieties include BARI Chola-1, released in 1987, and BARI Chola-3, released in 1993. BARI Chola-2, released in 1993, is wilt-disease-tolerant, while BARI Chola-4, released in 1996, is tolerant to Fusarium wilt. BARI Chola-5, released in 1996, is a high-yielding variety. BARI Chola 7-9 varieties, released in 1998, are resistant to Fusarium wilt disease. BARI Chola-10, released in 2017, is heat tolerant. In total, 10 chickpea varieties have been released since 1980 (Table 11). The complete list of varieties is provided in Appendix J. Table 11: Number of chickpea varieties released, by germplasm origin, 1980-2019 Germplasm origin 1980-1989 1990-1999 2010-2019 Total Pure line selection BARI 0 0 0 0 ICRISAT 1 8 1 10 Total 1 8 1 10 Acronyms: Bangladesh Agricultural Research Institute (BARI), International Crops Research Institute for the Semi- Arid Tropics (ICRISAT). SPIA Bangladesh Study 2025: Updating the Green Revolution 29 4.2 Aquaculture and Fisheries Improvement Fish is Bangladesh's second most important food source after rice. It accounts for approximately 60% of the nation's animal protein intake and provides essential micronutrients. Projections indicate that by 2030, demand could rise to nearly five million tonnes annually.15 Between 1984 and 2024, Bangladesh’s aquaculture industry experienced dramatic growth, with farmed fish production increasing from 124,000 to 2.1 million tonnes annually. Aquaculture now accounts for 55% of the country’s fish supply, up from just 16% four decades ago. This 25-fold expansion in the farmed fish market is thought to have been supported by genetically improved strains that enhance production (Gjedrem & Rye, 2018; Olesen et al., 2015) and strengthen disease resistance (Barría et al., 2021; Houston, 2017). The sector has focused on species that are both economically viable and culturally significant. Among these, rohu (Labeo rohita), silver carp (Hypophthalmichthys molitrix), and tilapia (Oreochromis niloticus) have played major parts in the sector’s rapid growth. These species are adapted to polyculture systems. This maximizes land use in one of the most densely populated countries and mitigates risk as each species has different abiotic and biotic stressors. Their status as dietary staples is also significant in determining their large share of the aquaculture sector. The Aquaculture for Income and Nutrition (AIN) project, launched in 2011 by the United States Agency for International Development (USAID) in collaboration with WorldFish, has been a key initiative in promoting aquaculture development. This program focused on developing improved fish strains, enhancing technology, and building capacity in hatcheries and nurseries to promote broader dissemination and adoption among small- and medium-scale households. Simultaneously, WorldFish has collaborated with the Bangladesh Fisheries Research Institute (BFRI) on tilapia breeding since 1994. Beginning in 2001, WorldFish produced a new generation of Genetically Improved Farmed Tilapia (GIFT) annually. The current generation, G17, demonstrates significantly higher weight gain, underscoring the success of these focused breeding initiatives. Rohu (Labeo rohita) is Bangladesh’s most widely farmed carp, with production reaching 386,000 tonnes in 2019–2020 (DoF, 2020; FAO, 2020). The Carp Genetic Improvement Program (CGIP) has sought to advance productivity, particularly through the Rohu Genetic Improvement Program, which began in 2011 when WorldFish started breeding fast-growing rohu (along with Catla and Silver Carp) through the Agriculture for Income and Nutrition Initiative. In 2020, a multiplier population of highly ranked generation-three (G3) families from the WorldFish Rohu Genetic Improvement Program was released to hatcheries across Bangladesh for further development into broodstock. The G3 rohu is a genetically improved strain with superior growth rates, disease resistance, and productivity. On-farm trials in 2022 at 19 hatcheries in Jashore, Natore, and Rajshahi demonstrated that G3 rohu weighed 37% more than unimproved strains. WorldFish has played a key role in disseminating G3 rohu, initially targeting the Feed the Future zone, and later expanding nationwide through partnerships with government agencies, NGOs, and private sector stakeholders. By 2022, spawns from the G3 rohu line had been widely propagated across hatcheries, where they were developed into broodstock for large-scale production. By 2023, over 30 commercial hatcheries and 24 nurseries across Bangladesh were supplying genetically 15 https://worldfishcenter.org/project/feed-future-bangladesh-aquaculture-activity https://worldfishcenter.org/project/feed-future-bangladesh-aquaculture-activity SPIA Bangladesh Study 2025: Updating the Green Revolution 30 improved rohu seed to farmers. WorldFish has been actively tracking the dissemination process, monitoring 48 nurseries (24 offering non-G3 rohu and 24 offering G3 rohu) across two divisions, Khulna and Rajshahi, to evaluate the program’s impact and reach. Genetically Improved Farmed Tilapia (GIFT) has been central to WorldFish’s work in Bangladesh since 1987, with various projects aimed at enhancing nutrition and reducing ecological impact. Developed through a long-standing collaboration between WorldFish (formerly ICLARM) and BFRI since 1994, GIFT tilapia was selectively bred for faster growth, higher yields, and greater survival rates compared to other strains. New GIFT tilapia generations were introduced in multiple phases (1994, 1996, 1997, 2005, 2008, 2009, and 2012), with the latest, G17, demonstrating superior weight gain. The expansion of tilapia hatcheries accelerated between 2011 and 2015 (Rossignoli et al, 2021), providing the conditions for the widespread adoption of GIFT tilapia. While hatcheries from 2006 to 2015 supplied diverse strains, this diversity declined after 2015, with newer hatcheries (2016- 2020) primarily offering GIFT tilapia (see Figure 5 for a map of GIFT tilapia hatcheries in 2020). Broodstock management plays a critical role in seed quality, and the share of multiplier hatcheries sourcing seed from breeding nuclei or cohort-based breeding systems has steadily increased. The proportion of hatcheries using elite germplasm rose from less than 10% in 2012 to over 35% in 2020 (Rossignoli et al, 2021). Recent data from WorldFish highlight a well-established diffusion network that has driven the dramatic growth of tilapia production in Bangladesh. This structured dissemination process continues to support the industry's expansion and sustainability. Figure 5: Distribution of hatcheries and upazila sales data for GIFT tilapia seed Source: Rossignoli et al, 2021 SPIA Bangladesh Study 2025: Updating the Green Revolution 31 Beyond providing policy recommendations for the government’s Hilsa fishery management efforts, WorldFish has focused on establishing co-management bodies in the six Hilsa sanctuaries within marine protected areas and promoting alternative income-generating activities in these regions. The WorldFish Silver Carp Genetic Improvement Program (WSCGIP) focuses on enhancing growth rates in silver carp through pedigree-based selection. Significant additive genetic variation in growth has been previously documented in Bangladeshi silver carp (Gheyas et al., 2009). The base population (Generation 0) was established in 2017 using broodstock from multiple Bangladeshi hatcheries (Hamilton et al., 2021), and the first selected generation was successfully spawned in 2019. The Carp-Mola polyculture approa