Evaluation and Geospatial Analysis of Variability in Maize Yield Response to Fertilizer (NPK) Using Modeling in Ghana
Bindraban, Prem S.
Isaac N. Kissiedu
K. El Mejahed
Maize is the main cereal crop in Ghana. However, yields are very low (around 1-1.5 mona), and despite the increase in fertilizer application per hectare (21-22 kg/ha), there are large differences in yields in on-farm and on-station trials. Maize production is hampered by several biotic and abiotic factors that negatively impact its yield response to fertilizer application. Therefore, we sought to understand why, despite fertilizer application, maize yields do not increase consistently over space and time and what major factors explain this variability. To answer this question, we chose a yield-modeling approach based on yield data from on-farm and on-station trials. Quantitative Evaluation of Fertility of Tropical Soils (QUEFTS) and Multiple Linear Regression-Akaike Information Criterion (MLR-AIC) models were used to evaluate observed yield variability, while random forest for spatial predictions framework modeling was used for geospatial analysis and mapping of yield predicted. The QUEFTS model cannot significantly explain yield variability at the station and farm level (R-12% and R-24.6%, respectively). MLR showed that soil physical properties explained more of the yield variability (R-24%) at the station level than environmental parameters (R²=8%), with chemical soil properties explaining the highest fraction (R-41%). At the farm level, environmental covariates (R2-26%) explained more variability in yield response than physical (R-21%) and soil chemical (R-16%) variables. Detailed regression analysis revealed that high temperature and high rainfall combined with shallow rooting depth (<50 cm) were determinants that reduced the effectiveness of fertilizer application. Understanding the yield variability observed in Ghana for better fertilizer recommendations must be done comprehensively because yield variability is the result of the interaction and combination of several covariates. Other covariates, such as management, pest and diseases, and solar radiation, must be considered in further modeling analysis.
NPK fertilizers, Maize
Kouame, K.K.A., P.S. Bindraban, IN. Kissiedu, K. El Mejahed, and W.K. Atakora. 2021. Evaluation and Geospatial Analysis of Variability in Maize Yield Response to Fertilizer (NPK) Using Modeling in Ghana. IFDC FERARI Research Report No. 7