Why is the spatial variability of millet yield high at farm level in the Sahel? Implications for research and development

Abstract

Millet yields are highly variable even within the same farm in the Sahel. A conceptual model of the phenomenon was designed and quantitative analysis of key relationships was conducted based on the available published studies. We found that the high spatial variability of millet yield is due to two main edaphic factors: soil fertility properties and water availability. It is still unknown whether the spatial variability of the two main factors is inherent to Sahelian soils. However, some biotic and abiotic factors induce and even maintain the variability of the main factors. The biotic factors include the presence of trees, termite’s activity, fertility management practices, pests, and diseases. Abiotic factors include rainfall (amount, intensity, and distribution) and wind erosion. The major soil chemical properties which strongly influence spatial yield variability include organic carbon, pH, aluminum saturation, and available phosphorus. Millet yield was positively and strongly correlated with P (r²=0.80, p < 0.001), moderately with pH (r²=0.30, p < 0.001), but strongly and negatively with the logarithm of aluminum saturation (r²=0.54, p < 0.001).

Water availability is affected mainly by the soil physical properties, which strongly impact millet yield variability through infiltration rate and water retention capacity. The underlying properties are soil structure, texture, and soil surface conditions. The inadequate consideration of soil heterogeneity could also explain the low level of farmer adoption of the recommendations regarding soil fertility improvement based on experimental research or decision support. This situation calls for improved precision agriculture technologies that are compatible with Sahelian farmers’ socio-economic conditions.

Keywords:

• High spatial variability
• modeling
• pearl millet
• Sahel
• soil physical and chemical properties
• yield variability

Acknowledgments

The authors gratefully acknowledge many valuable comments from anonymous reviewers as well as the kind editing and proofreading of the entire manuscript by Colin M. Minielly.

Additional information

Funding

We thank the CerLiveTrees (CLT) project under the West and Central African Council for Agricultural Research and Development (CORAF/WECARD) for funding this work.