Sub-Saharan Africa (SSA) faces a looming challenge: meeting a projected 233% increase in maize demand by 2050. This surge is driven by population growth and changing dietary habits. Over the past two decades, SSA has responded to rising demand primarily through expanding maize cultivation—an area now comparable to the US Corn Belt. This approach, however, is unsustainable. Continued land expansion threatens natural ecosystems and pushes agriculture into marginal lands with lower productivity. Moreover, it may not suffice to meet the projected demand, potentially necessitating costly imports—a significant burden given many SSA countries' limited financial resources and recent volatility in commodity prices. The average maize yield in SSA is substantially below its potential, roughly five times lower than what climate and soil conditions would allow. This significant yield gap presents a critical opportunity. By focusing on yield intensification rather than land expansion, SSA can significantly improve food security and reduce reliance on imports. However, despite substantial investments in agricultural research and development (AR&D), yield gains in SSA have lagged far behind other regions like Southeast Asia and South America. This discrepancy highlights a need for a more targeted and effective approach to AR&D. Previous research on maize yield constraints in SSA has faced limitations: small sample sizes, narrow ranges of management practices studied, and a lack of comprehensive analysis across diverse environments. This study addresses these limitations by employing a large-scale dataset, advanced statistical techniques, and high-resolution spatial data to identify and quantify the impact of various agronomic practices on maize yields across diverse agro-ecological zones. The goal is to pinpoint effective, scalable solutions for enhancing maize production in SSA without excessive land conversion.

A substantial body of existing literature highlights the significant yield gap in Sub-Saharan African maize production. Studies have pointed to various factors limiting yields, including nutrient deficiencies, pest infestations, poor cultivar selection, and suboptimal farming practices. However, many previous studies suffered from methodological limitations, such as small sample sizes, limited geographic scope, and a focus on individual practices rather than integrated management systems. Some research has utilized farmer surveys to assess yield constraints, but these are prone to inaccuracies due to potential overestimation of yields, especially in small plots. There’s also been a call to embrace diverse agricultural approaches, including conservation agriculture, agroecology, climate-smart practices, regenerative agriculture, nature-based solutions, and digital agriculture. While these approaches are beneficial for food security, climate change mitigation, and resource conservation, the authors express concern that a broader focus might shift attention away from the urgent need to improve yields in SSA through proven agronomic solutions.

This study analyzes a comprehensive dataset of 14,773 smallholder maize fields across five regions of SSA: Nigeria, Rwanda-Burundi, Zambia, Tanzania, and Uganda-Kenya. The data, collected by One Acre Fund between 2016 and 2022, includes detailed information on maize yields, management practices (cultivar selection, fertilizer application, pest control, planting date, plant density, weeding etc.), and environmental factors. The dataset includes fields from farmers both participating and not participating in One Acre Fund's programs, which enhances the range of management practices observed. Addressing the inherent variability in climate, soil, and management practices, the researchers employed a stratified approach. They grouped fields into 25 distinct climate zones based on growing-degree days and aridity index, using high-resolution spatial data to account for local variations. Within each climate zone, a conditional inference tree analysis was performed, incorporating soil, terrain, and in-season weather variables to control for residual environmental variation. This approach allowed for the identification of significant agronomic practices influencing yields across various environmental contexts. To further validate the findings and enhance understanding of individual practice effects, a machine-learning model (Gradient Boosting Machine) was applied. This model, trained on the entire dataset, helped assess the contribution of each management practice to yield variability and provided insights into their individual response functions, also accounting for the synergistic interactions between practices. Shapley Additive Explanation (SHAP) values were used to quantify these individual practice effects. Finally, the researchers performed a scenario assessment to project the potential impact of widespread adoption of identified yield-improving practices on SSA's maize self-sufficiency by 2050. This involved extrapolating the observed yield gains from the study area to the entire maize-growing region of SSA, considering different scenarios: maintaining current yield increase rates and accelerating yield gains to match the levels observed in high-performing farms in the dataset. These scenarios were compared against projected maize demand by 2050, calculated based on population projections and per capita maize consumption changes.

The analysis revealed several key agronomic practices that significantly impacted maize yields across diverse SSA environments. These included: higher nitrogen (N) and phosphorus (P) fertilizer application rates, proper fertilizer placement (in-hole application vs. surface broadcasting), use of hybrid seeds instead of open-pollinated varieties, earlier sowing dates, higher plant densities, and effective pest control. Synergistic effects were also noted; for instance, the yield benefit of hybrids was maximized when combined with early sowing and high N fertilization. Interestingly, the study did not find a significant impact of hybrid traits (crop cycle duration, year of release, disease tolerance) on yields—suggesting that poor agronomic practices might override the benefits of superior genetics. The conditional inference tree analysis indicated a substantial potential for yield improvement. Farmers adopting a suite of improved practices (hybrid seeds, high N rates, high plant densities, and early sowing) achieved yields 2.4 times higher than those using conventional practices. This corresponds to approximately a 30% yield gap closure, adding 2.5 t/ha in average yield. The machine-learning model confirmed these findings, providing additional insights into the individual contributions of each practice. Extrapolating these findings to the entire SSA maize-growing area, the study suggests that the widespread adoption of improved agronomic practices could double current maize output, increasing production from 80 million tons to 168 million tons. This could substantially reduce the need for land expansion and imports to meet the projected maize demand by 2050. In contrast, if current yield improvement rates persist, SSA would need an additional 28 million hectares of land or 76 million tons of imports to meet demand. This highlights the urgent need for targeted yield intensification efforts.

This study's findings underscore the critical role of improved agronomic management in achieving maize self-sufficiency in SSA. While the positive effects of such practices have been known for decades, recent calls for diverse, often broadly defined agricultural approaches might distract from the urgent need for direct yield improvements. This research directly demonstrates the potential of existing technologies and practices to achieve substantial yield gains. The results, validated through a large-scale dataset and rigorous statistical methods, provide strong evidence for prioritizing investments in and policy support for improved agronomic practices related to soil fertility, cultivar selection, planting date, plant density, and pest control. Scaling up these practices can enhance food security and smallholder farmers’ income. Failure to address the yield gap could have severe negative socio-economic and environmental repercussions.

This study demonstrates the significant potential for doubling maize yields in Sub-Saharan Africa through the adoption of improved agronomic practices. Widespread implementation of these practices—including optimized cultivar selection, enhanced nutrient management, effective pest control, and improved crop establishment techniques—could drastically reduce the need for further cropland expansion and costly maize imports. Future research should focus on understanding the barriers to adoption of these practices, developing targeted interventions to overcome these barriers, and investigating further yield improvements beyond the 30% gap closure achieved here.

The study's reliance on observational data and predictive models, rather than controlled field experiments, limits the ability to definitively establish causal relationships. While the large dataset and rigorous statistical methods mitigate this limitation, the possibility of unmeasured confounding factors remains. Also, although efforts were made to minimize data uncertainties and inaccuracies typical of farmer-reported information, inherent uncertainties in farmer surveys in developing countries still exist. Finally, while the regional extrapolation offers valuable insights, local conditions may require tailoring of recommendations to specific environmental contexts and farming systems.