Assessing the effects of plant density and nitrogen on millet yield in Southern Niger using the CERES‑millet model

Background The dryland production environments in Niger Republic (Niger) generate variable crop production risks that reduce crop yields and increase regional food insecurity. Optimal combinations of crop varieties and management are needed to maximize crop water-limited yields in these environments. Methods In this study, we calibrated and validated the CERES-Millet model using data from field experiments. Seasonal analysis (1984–2020) was carried out in 18 selected sites across the three agroecological zones (AEZs) to test the effects of plant density and N fertilization on grain yield. The treatment included five different plant densities (1.5, 3, 6, 9, and 12 plants m−2) and four N rates (0, 20, 40, and 60 kg N ha−1). Three millet varieties (CHAKTI, HKP, and SOSAT-C88) were compared. Millet production risk was assessed at each AEZ using cumulative probability distribution graphs. The acceptable grain yield required to compensate for the minimum production cost of millet in Niger was set to 975 kg ha−1 (75th percentile of the simulated data). Results The CERES-Millet model reasonably reproduced number of days to flowering (d-index > 0.50; RMSE < 2 days), number of days to maturity (d-index > 0.50; RMSE < 2 days), and grain yield d-index > 0.78; RMSE < 100 kg ha−1) for all the three varieties. The results showed that there was significant response to N (40 – 110% yield increase following N application) and plant density (30–80% yield increase by increasing density above 1.5 plants m−2) in all the AEZs depending on variety. The SOSAT was the most responsive variety to N application and plant density in all AEZs. Under low N application (0–20 kg ha−1) and low (1.5—3 plants m−2) to moderate plant density (6 plants m−2), CHAKTI and HKP had the highest production risks. Increasing N application above 20 kg ha−1 mitigate these risks where grain yield was above the 975 kg ha−1 threshold representing the minimum production cost for millet in more than 50% of the years under all plant densities except in Sahel where this threshold was only achieved in < 20% of the years. In all AEZs, increasing plant density above 6 plants m−2 increases this risk under low to moderate N application, but the downside risk was mitigated when N was applied at high rates. Conclusion This study demonstrated N application rate and plant density recommendations must be tailored to specific variety and AEZs to maximize grain yield and reduce volatility in Niger.