Peer review

Problem: Although cassava exports large nutrient amount from the soils, there is mixed evidence on its response to fertilizer. Objective: This paper aimed at better understanding variable cassava responses to nutrient application and associated soil causes. Methodology: A network of multi-location two-year on-farm trials, evaluating in contrasting DRC agroecological zones (Tshopo and Kongo Central (KC)) and in Burundi, cassava response to macronutrients (DRC and Burundi) and micronutrients (Burundi), was conducted. Two stochastic frontier models were fitted per DRC agroecological zone and per experiment in Burundi (macronutrient or micronutrient), with observation number ranging between 290 and 490 per scenario of model fitting. The best fitted model was used to estimate soil parameters’ contribution to yield response and to calculate the soil-efficient yield gap (YSEG). For nutrient combination “k” and in farm “i”, YSEG was calculated as ratio of yield obtained in farm “i” with “k” to the highest yield (derived from the best fitted model) attained from “k” in farms with similar soil fertility as farm “i”. Results: In Tshopo, only NK-combination yielded more than no-fertilizer, while all macronutrient combinations, except PK, did in KC and Burundi. Yield response to micronutrient was remarkably poor. YSEG averaged 70 %, 63 % and 54 % in Tshopo, Burundi and KC, respectively, indicating that, if nutrient use had been efficient, cassava yield should have been 30–50 % higher without increasing nutrient input. Soil fertility significantly determined yield in all zones, but soil parameters effects and the way in which their effect occurred were agroecology specific. In Tshopo and for micronutrients in Burundi (i.e., low yield response), there were complex interactions of soil parameters with the applied nutrients. For instance, soil N, exchangeable K and Mg significantly interacted with applied N in Tshopo, indicating that the farms might have simultaneously optimal levels of these three soil parameters to respond to the applied N. For most of the applied macronutrients in KC and Burundi, only one soil parameter had significant interaction, indicating that farms with optimal level of that soil parameter responded to the applied nutrient irrespective of the other soil parameters’ levels. In Burundi, YSEG was lower with no-fertilizer than with fertilizer application, indicating a lesser effectiveness of the native soil nutrients. Conclusion: We conclude that soil fertility management (SFM) should be at the latest stage of prioritization in Tshopo. In Burundi and KC, SFM should be key priority, with focus on improving uptake of native soil nutrients in Burundi.

Text in English