Is soil analysis data of real value for smallholder farmers?

Fertilizer recommendations in developed countries are widely based on soil testing, targeted on maintenance of soil P and K fertility and correction of pH. In sub-Saharan Africa, smallholder fields are far more variable in terms of soil fertility, reflecting large differences in historical land-use, and in general have poor fertility. Soil analysis is expensive, and comes with large amounts of uncertainty. At best, it reflects what farmers already know: manured fields are much better than those not manured. For soil scientist and agronomists without such knowledge, the value of soil sample analysis is beyond question. The objective here is to evaluate how useful soil sample analysis is in relation to what farmers need to know. To this end, we reflected on observed variation in yield and on responses to fertilizer applications in on-farm experiments in relation to soil analysis. Here, we evaluated maize response to fertilizers in experiments conducted in Western Kenya and Tanzania. The Kenyan dataset included 46 on-farm experiments in the long-rain season of 2018, combined with a dataset including 23 on-farm NOT trials running from 2013 to 2018. For Tanzania, a dataset from TAMASA nutrient omission trials was included. We found that yield responses vary enormously, and were only partly explained by variables that were determined by soil analysis. The majority of variation was found on farm, i.e. at very short distances. At best, soil analyses informed about fields were a higher probability of a response is expected. In particular, knowledge about soil K content was important, yet strongly related to historical manure use. We conclude that offering a limited number of fertilizer composites with different ratios of N, P and K in combination with a N top-dress fertilizer, adjusted to soil map units is more than sufficient for proper fertilizer recommendations. Local adjustments need to be made in combination with farmer assessments, without the need for expensive soil testing, combined with limited plant sample analysis for monitoring purposes.