Modeling integrated soil fertility management for maize production in Kenya using a Bayesian calibration of the DayCent model

Laub Moritz, Necpalova Magdalena, Van de Broek Marijn, Corbeels Marc, Ndungu Samuel Mathu, Mucheru-Muna Monicah Wanjiku, Mugendi Daniel, Yegon Rebecca, Waswa Wycliffe, Vanlauwe Bernard, Six Johan. 2024. Modeling integrated soil fertility management for maize production in Kenya using a Bayesian calibration of the DayCent model. Biogeosciences, 21 (16) : 3691-3716.

https://doi.org/10.5194/bg-21-3691-2024

Article de revue ; Article de recherche ; Article de revue à facteur d'impact Revue en libre accès total

Prévisualisation

Version publiée - Anglais
Sous licence

.
612905.pdf
Télécharger (6MB) | Prévisualisation

Url - jeu de données - Entrepôt autre : https://doi.org/10.25502/wdh5-6c13/d / Url - jeu de données - Entrepôt autre : https://doi.org/10.25502/be9y-xh75/d

Résumé : Sustainable intensification schemes such as integrated soil fertility management (ISFM) are a proposed strategy to close yield gaps, increase soil fertility, and achieve food security in sub-Saharan Africa. Biogeochemical models such as DayCent can assess their potential at larger scales, but these models need to be calibrated to new environments and rigorously tested for accuracy. Here, we present a Bayesian calibration of DayCent, using data from four long-term field experiments in Kenya in a leave-one-site-out cross-validation approach. The experimental treatments consisted of the addition of low- to high-quality organic resources, with and without mineral nitrogen fertilizer. We assessed the potential of DayCent to accurately simulate the key elements of sustainable intensification, including (1) yield, (2) the changes in soil organic carbon (SOC), and (3) the greenhouse gas (GHG) balance of CO2 and N2O combined. Compared to the initial parameters, the cross-validation showed improved DayCent simulations of maize grain yield (with the Nash–Sutcliffe model efficiency (EF) increasing from 0.36 to 0.50) and of SOC stock changes (with EF increasing from 0.36 to 0.55). The simulations of maize yield and those of SOC stock changes also improved by site (with site-specific EF ranging between 0.15 and 0.38 for maize yield and between −0.9 and 0.58 for SOC stock changes). The four cross-validation-derived posterior parameter distributions (leaving out one site each) were similar in all but one parameter. Together with the model performance for the different sites in cross-validation, this indicated the robustness of the DayCent model parameterization and its reliability for the conditions in Kenya. While DayCent poorly reproduced daily N2O emissions (with EF ranging between −0.44 and −0.03 by site), cumulative seasonal N2O emissions were simulated more accurately (EF ranging between 0.06 and 0.69 by site). The simulated yield-scaled GHG balance was highest in control treatments without N addition (between 0.8 and 1.8 kg CO2 equivalent per kg grain yield across sites) and was about 30 % to 40 % lower in the treatment that combined the application of mineral N and of manure at a rate of 1.2 t C ha−1 yr−1. In conclusion, our results indicate that DayCent is well suited for estimating the impact of ISFM on maize yield and SOC changes. They also indicate that the trade-off between maize yield and GHG balance is stronger in low-fertility sites and that preventing SOC losses, while difficult to achieve through the addition of external organic resources, is a priority for the sustainable intensification of maize production in Kenya.

Mots-clés Agrovoc : modèle de simulation, changement climatique, gaz à effet de serre, matière organique du sol, fumier, modèle mathématique, théorie Bayésienne, rendement des cultures, gestion intégrée de la fertilité des sols, ressource minérale, carbone