Heterogeneity of crop productivity and resource use efficiency within smallholder Kenyan farms: Soil fertility gradients or management intensity gradients?

The decrease in crop yields at increasing distances from the homesteads within smallholder farms of Sub-Saharan Africa (SSA) is normally ascribed to the existence of within-farm soil fertility gradients. Field observations also suggest that a large part of such variability is concomitantly caused by poor agronomy. To understand the interaction between soil fertility (S factors) and management decisions (M factors) affecting crop variability, we combined field research conducted in western Kenya (Vihiga, Kakamega and Teso districts; rainfall: 1600, 1800 and 1200 mm, respectively) with explorations using the simple dynamic crop/soil model for dynamic simulation of nutrient balances, previously tested for the region. Field measurements indicated within-farm differences in average maize grain yields of 48% (2.7 vs. 1.4 t ha¿1) in Vihiga and of 60% (1.5 vs. 0.6 t ha¿1) in Teso, between fields that were close and far from the homestead, respectively. Extreme values ranged widely, e.g. between 4.9 and 0.3 t ha¿1 for all the farms surveyed in Vihiga, where the average farm size was 0.6 ha. Maize grain yields tended to increase with increasing contents of soil C, total N, extractable P and exchangeable bases. However, the negative relationship between S factors and distance from the homestead was not as strong as expected, and yield variability was better explained by multiple regression models considering M factors such as planting date, plant density, resource use and weed infestation (40¿60% across sites). Then, we analysed the variation in resource (cash, labour, N) use efficiency within farms of different resource endowments with the aid of the simulation model. N balances at plot scale varied from ca. +20 to ¿18 kg ha¿1, from ¿9 to ¿20 kg ha¿1 and from ¿16 to ¿18 kg ha¿1 for the different fields of the high, medium and low resource endowment case-study farms, respectively. Labour productivities ranged between ca. 10 and 38 kg grain man-day¿1 across field and farm types. The results indicate the need of considering within farm heterogeneity when designing soil fertility management interventions. Resource use efficiency was strongly affected by soil quality. As farmers invest more effort and resources in the more productive and less risky fields, the interaction between S and M factors leads to farmer-driven resource use efficiency gradients within smallholder farms.