Improving Maize Productivity in Northern Benin Through Localized Placement of Amendments and Fertilizers

Université catholique de Louvain (UCL) Faculté des bioingénieurs Earth and Life Institute/ Environmental Sciences

Université d’Abomey-Calavi (UAC) Faculté des Sciences Agronomiques Ecole Doctorale des Sciences Agronomiques

Improving maize productivity in northern Benin through localized placement of amendments and fertilizers

GBÈNOUKPO PIERRE TOVIHOUDJI

June 2018

Thèse en cotutelle présentée en vue de l’obtention du grade de : Docteur en Sciences Agronomiques et Ingénierie Biologique de l’UCL et Docteur en Sciences Agronomiques de l’UAC

Jury members: President : Prof. Jacques Mahillon (UCL, Belgique) Co-President : Prof. Philippe A. Lalèyè (UAC, Bénin) Supervisor : Prof. Charles L. Bielders (UCL, Belgique) Co-Supervisor : Prof. P.B. Irénikatché Akponikpè (UP, Bénin) Readers : Prof. Pierre Bertin (UCL, Belgique) : Prof. Bruno Delvaux (UCL, Belgique) : Prof. Euloge K. Agbossou (UAC, Bénin) : Prof. Bernard Tychon (ULg, Belgique)

CREDITS The four years of this PhD were funded by the West Africa Agricultural Productivity Program (WAAPP-Benin), and the “Climate Change Agriculture and Food Security (CCAFS)” program of the CGIAR. I also benefitted from a doctoral fellowship of the Université catholique de Louvain (UCL). I’m very grateful to all these organizations for their confidence and financial support.

To my wife and children

“The land was lent to you by your children. It was not given to you by your parents; Treat it as it should”.

Old Indian proverb

Acknowledgements

I am so lucky that I had the opportunity of studying at the Université catholique de Louvain (UCL). I have really learned a lot. Now, my stay here almost ends after four years of hard but insightful and rewarding work. I want to express my gratitude to the people who have helped and accompanied me during the journey of my PhD. First and foremost, I owe a debt of gratitude to my promoter, Prof. Charles Bielders whom I affectionately call my “scientific grandfather” because he was the promotor of my co-promotor Pierre, for stimulating my interest in science and encouraging me to pursue rigorous scientific investigations. He gave me countless great suggestions regarding how to learn/write English, how to write good journal papers and elaborate scientific stories. Charles gave me the great opportunity to be a UCL Scholar and has played a huge role in my development since my research proposal setting and during my time at UCL. Besides scientific and academic support, he also provided financial support whenever required during this thesis. Thank you so much, “Big boss”. I thank my co-promoter from University of Parakou (Benin), Prof. Pierre B.I. Akponikpè (affectionately called my “scientific father”) who has supervised me since I was a master student. Akponikpè always kept training me and did all his best to help me become a qualified scientific researcher. Without his support, I could not have obtained the research grant from the West Africa Agricultural Productivity Program (WAAPP-Benin) and the fellowship from the Université catholique de Louvain (UCL). I’m grateful to these organizations for their confidence and financial support. I’m also grateful to the “Climate Change Agriculture and Food Security (CCAFS)” program of the CGIAR for funding the initial trials of this work through a grant offered to Prof. Akponikpè under the “Research Theme # 3” on the modeling of the impact of on-farm fertility management practices on soil carbon dynamics in northern Benin. Thanks a lot, Professor Akponikpè. I thank Prof. Euloge Agbossou (UAC, Bénin), for his availability, critical comments on my manuscripts and intellectual advice that enabled me to cope with other life challenges. I do extend my thanks to my jury members for the advices they gave me throughout the achievement of the study. I appreciated their valuable comments and suggestions including those of the anonymous reviewers aiming at always pushing further my expectations and improving my manuscripts.

vii

Acknowledgements

Thanks to Drs Dagbenonbakin Gustave and Agbangba Emile from INRAB, Drs Fatondji Dougbédji and Ibrahim Ali from ICRISAT/Niger, and Dr Rosaine Yegbemey (Université de Parakou) who supported me during the many challenges of field work and for their enthusiasm and critical review of my research proposal and manuscripts. I am also deeply thankful to CRA/Nord through its current Director Dr Yaoitcha Alain and all the Staff and technicians for the facilities placed at my disposal during the implementation of my field trials. I thank all the members of the Faculté d’Agronomie, Université de Parakou, Benin through Profs. H. Edja, V. Zinsou, A.J. Djenontin, M.N. Baco, I.M. Moumouni, and Dr. C.N. Sossa. I do not forget my friend Kévin Affoukou for his critical reading of my research manuscripts and my fellow Bachelor and MSc students who helped in many and various aspects of research within the framework of this PhD project: Aziz, Fabius, Kader, Joel, Dine, Christian, Maité and O’neil. I would also like to thank all the academics, my friends and colleagues from the GERU group and the Earth and Life Institute of the UCL. During these four years of thesis, I shared really nice moments with them at work but not only. Special thanks to Aimé, Raed, Mokrane, Adèle, and Jean-Claude. I am also grateful to Mme Carine De Meyer and all the technicians. I do not forget the GERU football team (Sebastien, Félicien, Jean-Baptiste, Tanguy, Brieuc and others). The exceptional goalkeeper that I was ( ) will certainly miss you but we will find a good substitute. My PhD. colleagues of the WAAPP-Benin and the Université de Parakou (Complexe de l’Innovation) specially Dr. A. Attingli, Dr. L. Zinsou, Dr. J. Egah, A. Adjogboto, L. Akponikpè, D. Likpété, G. Allakonon, F. Ouidoh, M. Diallo, I. Bello, S. Adéchian are acknowledged for their cooperation, the fruitful discussions and the good time that we shared during the last four years. I deeply acknowledge the moral support, patience and encouragement offered by the family Défourny during my stay at UCL through Maité and her dad Pierre. As Master student from UCL within the framework of this PhD project, Maité accepted to go, stay in Parakou and Ina (northern Benin) to collect data. “Na siara, Maité”. I don’t forget Maité’s sisters (Noémie and Elisabeth) and her boyfriend Loïc. I am also thankful to all my countrymates whom I met in Belgium particularly in LLN for their nice conversations, continuous support and encouragements during my stay. They include but are not limited to P.I. Hounzandji, K. Gnimassou, F. Assogba, Y. Djivoh, N. Kpadonou, H. Prodjinoto, S. Gnonlonfoun, A. Houndji, B. Sinhouenon, J. Lawson, E. Gbedonou and R. Saré. I am really thankful to Prof. Christophe Gandonou for his encouragement and support during his stay in Louvain-la-Neuve.

viii

Acknowledgements

I would like to express my gratitude to my wife Marie Kindji and my children (Ronelle, Ronique and Rosaine) especially for their love, patience and endless support all along the way. Sorry if you did have to cope with my long absence from home: your sacrifice is yet invaluable, but truly appreciated! My utmost acknowledgement goes to my brothers of blood (S.P. Ahannougbé and R. Dohou and their wives) as well as my in-law parents (families Kindji and Aboudou) and Mme R. Djaboutou and his mum for their prayer and the tireless assistance every time I am away. Last but not least, I wish to express my deepest thanks to my former school teacher and headmaster Mr. Pascal Koukoui for his moral support, encouragement and prayer during these last few years. Thanks a lot, Teacher Koukoui; May God bless you and your family. This list is not exhaustive in view of the number of people whose meeting allowed the long and sometime difficult delivery of this work. I therefore apologize to all those who have contributed to this work and who have not been mentioned here. May all of you find here your share of recognition.

Summary

Throughout much of Sub-Saharan Africa, maize production in smallholder farms is characterized by low productivity due to low fertility soils, the scarce availability and use of external inputs and recurrent droughts exacerbated by climate variability. This situation calls for exploring fertilizer and amendment management practices that are efficient and can concurrently improve soil fertility, yields and economic returns. One such technology is the localized application of small quantities of manure and/or mineral fertilizers in the planting holes either at sowing or shortly after planting. Unlike for sorghum and millet, few studies have evaluated the agronomic and economic performances of this technology on maize. Besides, few studies have quantified the risk associated with this technology due to variability in crop management, soil and climatic factors. The general objective of this study was therefore to assess the agronomic and economic potential of localized application of manure and fertilizers in maize-based cropping system in northern Benin, with a long-term goal of developing recommendations. For this purpose, we combined three different approaches: (1) two on-station experiments to assess the agronomic potential and economic profitability of hill-placed manure and mineral fertilizer (or fertilizer microdosing), (2) farmer-field trials to quantify the variability in yield response, economic profitability and risk associated with some of the most promising treatments under real-world conditions and (3) modeling the response of maize across a range of rainfall conditions to further evaluate the sustainability of these practices in northern Bénin. All experiments were carried out in Ina district (Bembèrèkè, northern Benin) from 2012 to 2015. The on-station experiments showed that localized application of manure and fertilizers significantly increased grain and stover yields by 64-132% and 28-131%, respectively, across years. Combining hill-placed manure and fertilizers further increased grain yields by 31-55% on average. The increases in yields under fertilizer microdosing were accompanied by marked increases in nutrient uptake and negative nutrient balances. Nutrient balances were equally or more negative on microdosing fertilized plots than on the unfertilized controls. This was particularly the case for P and K and suggests that microdosing may enhance nutrient mining and should probably not be used for extended periods. The on- farm demonstrations showed a large variability in maize yield responses to fertilizer microdosing which could be partly explained by some measured soil parameters (clay and /or silt, total carbon, exch-Mg, pH) and weed pressure. Overall, absolute yield response tended to decrease with increasing yields in the control plots. Based on the value-cost ratio (VCR) the economic performance of the recommended fertilizer rate was less than that of the microdosing treatments (alone or combined with manure) despite the higher labor cost associated with the latter treatments. Despite the greater variability compared to the control, the risk of no return on investment was nearly nil for microdosing treatments (alone or combined with manure). The long-term scenario analysis using the specifically-parameterized DSSAT model revealed that the application of 2 g of N-P-K15- -1 15-15 fertilizer + 1 g urea per hill (equivalent to 23.8 kg N ha ) improved both the long- term average and the minimum guaranteed yield without increasing inter-annual variability and the economic risk compared to unfertilized plots. Even though combining microdosing with manure (at least at 1 t ha-1) was economically slightly riskier than microdosing alone, this risk remained low since a VCR of 2 could be achieved in almost 100% of the years. This makes the latter more sustainable and appropriate for smallholder farmers than the current recommendation.

Résumé

En Afrique sub-saharienne, la production de maïs dans les petites exploitations se caractérise par une faible productivité due à la faible fertilité des sols, à la disponibilité et à l'utilisation limitées des intrants et aux sécheresses récurrentes aggravées par la variabilité climatique. Une des technologies efficace et qui puisse simultanément améliorer la fertilité des sols, les rendements et les bénéfices économiques est l'application localisée de petites quantités de fumier et/ou d'engrais minéraux au poquet, soit au moment du semis, soit peu de temps après le semis. Contrairement au sorgho et au mil, peu d'études ont évalué les performances agronomiques et économiques de cette technologie sur le maïs. En outre, peu d'études ont quantifié le risque associé à cette technologie en raison de la variabilité des pratiques de gestion, des facteurs édaphiques et climatiques. L'objectif général de cette étude était donc d'évaluer le potentiel agronomique et économique de de cette technologie dans les systèmes de culture à base de maïs au nord du Bénin, avec pour objectif à long-terme de faire des recommandations. À cette fin, nous avons combiné trois approches différentes: (1) deux essais en station pour évaluer le potentiel agronomique et la rentabilité économique, (2) des essais en milieu paysan pour quantifier la variabilité de la réponse, la rentabilité économique et les risques associés aux traitements les plus prometteurs dans des conditions réelles et (3) la modélisation de la réponse du maïs à la fertilisation microdose dans diverses conditions pluviométriques pour évaluer la durabilité de ces pratiques. Toutes les expériences ont été réalisées dans l’arrondissement d'Ina (Commune de Bembèrèkè) de 2012 à 2015. Les essais en station ont montré que l'application localisée de fumier et d'engrais augmentait significativement les rendements en grain et paille de 64-132% et 28-131%, respectivement. La combinaison du fumier et des engrais a d’avantage augmenté les rendements en grains de 31 à 55% en moyenne, comparé à la fertilisation minérale seule. L’augmentation des rendements s’est accompagnée d'une augmentation marquée du prélèvement des nutriments et de bilans nutritifs négatifs. Sauf pour l'azote dans le traitement impliquant la dose recommandée, les bilans étaient tout aussi similaires ou plus négatifs pour les parcelles fertilisées que pour les témoins non fertilisés. Cela indique que la fertilisation microdose peut aggraver l'épuisement des nutriments du sol et ne pourrait être utilisée pendant de longues périodes sur une même parcelle. Les essais en milieu paysan ont montré une grande variabilité dans les réponses du maïs à la fertilisation microdosée qui pourrait s'expliquer en partie par certains paramètres mesurés du sol (argile et/ou limon, carbone total, Mg-échangeable, pH) et la pression des mauvaises herbes. Dans l'ensemble, la réponse absolue tend à diminuer avec l'augmentation des rendements du témoin. En prenant en compte la main-d'œuvre, l'analyse du ratio bénéfice-coût (VCR) a montré que la performance économique de la dose recommandée est inférieure à celle des traitements microdose malgré les coûts de main-d'œuvre plus élevés associés aux derniers. Malgré la plus grande variabilité par rapport au témoin, le risque économique était quasi nul pour la microdose. L'analyse de scénarios à long-terme utilisant le modèle DSSAT a révélé que l'application de 2 g d'engrais NP-K15-15-15 + 1 g d'urée par poquet (équivalent à 23,8 kg N ha-1) a amélioré à la fois le rendement moyen à long-terme et le minimum garanti, sans pour autant augmenter la variabilité inter- annuelle et le risque économique par rapport aux parcelles non fertilisées. Même si la combinaison de la microdose avec du fumier (au moins 1 t ha -1) était économiquement légèrement plus risquée que la microdose seul, ce risque était faible car un VCR de 2 pouvait être atteint dans presque 100% des années. Cela rend ce dernier plus durable et approprié pour les petits exploitants que la recommandation actuelle.

xii

Contents

Acknowledgements vii Summary xi Résumé xii Contents xiii 1 General introduction 1 1.1 Context ...... 2 1.1.1 Economic importance of agriculture in Sub-Saharan Africa ...... 2 1.2 Problem statement and analysis ...... 3 1.2.1 Maize-based cropping systems in Benin: extensive and low inputs ...... 3 1.2.2 Improved manure management for enhancing yields and soil fertility ...... 5 1.2.3 Potential of fertilizer microdosing in improving maize productivity ...... 7 1.2.4 Potential of combining hill-placed manure and fertilizer ...... 10 1.2.5 On-farm research approach for better targeted recommendations ...... 11 1.2.6 Developing decision support tools regarding the fertilizer microdosing ..... 12 1.3 Research questions and objectives ...... 13 1.3.1 Research questions ...... 13 1.3.2 Research objectives ...... 14 1.4 Brief description of the study area ...... 15 1.4.1 Research sites and socio-economical characteristics ...... 15 1.4.2 Biophysical characteristics ...... 16 1.5 Outline of the thesis ...... 19 2 Combined application of hill-placed manure and mineral fertilizer 21 Abstract ...... 22 2.1 Introduction ...... 23 2.2 Materials and methods ...... 26 2.2.1 Experimental site description ...... 26 2.2.2 Experimental design and treatments ...... 27 2.2.3 Trial installation and management ...... 28 2.2.4 Soil sampling and analysis ...... 29 2.2.5 Economic analysis ...... 29 2.2.6 Statistical analysis ...... 31 2.3 Results ...... 31 2.3.1 Rainfall and temperature distribution during the cropping periods ...... 31 2.3.2 Maize grain yield ...... 32 2.3.3 Maize stover yield ...... 33 2.3.4 Yield response to nutrient application ...... 34 2.3.5 Agronomic efficiency (AE) of N, P and K ...... 37 2.3.6 Post-harvest soil status...... 38 2.3.7 Economic performance indicators ...... 41 2.4 Discussion ...... 45 2.4.1 Soil fertility improvement ...... 45 2.4.2 Maize productivity and resource use efficiency ...... 46

xiii

Contents

2.4.3 Implications for nutrient management by farmers ...... 48 2.5 Conclusions ...... 50 3 Fertilizer microdosing in maize cropping systems in northern Benin 51 Abstract ...... 52 3.1 Introduction ...... 53 3.2 Materials and methods ...... 55 3.2.1 Description of study zone ...... 55 3.2.2 Experimental site ...... 56 3.2.3 Experimental design ...... 56 3.2.4 Measurements and calculations ...... 58 3.2.5 Statistical analysis ...... 64 3.3 Results ...... 65 3.3.1 Soil properties of the experimental field...... 65 3.3.2 Rainfall characteristics and plant available water ...... 65 3.3.3 Maize grain and stover yields ...... 67 3.3.4 Yield components ...... 69 3.3.5 Fertilizer and rainfall use efficiency ...... 69 3.3.5 Fertilizer and rainfall use efficiency ...... 71 3.3.6 Nutrient inputs/uptakes and balances ...... 71 3.4 Discussion ...... 78 3.4.1 Maize response to manure application ...... 78 3.4.2 Maize response to fertilizer application ...... 79 3.4.3 Efficiency of fertilizer microdosing across fertility levels ...... 81 3.4.4 Fertilizer microdosing may exacerbate nutrient mining...... 82 3.4.5 Opportunities for smallholder maize farming systems ...... 84 3.5 Conclusions ...... 85 4 Variability in response and profitability following microdosing application 87 Abstract ...... 88 4.1 Introduction ...... 89 4.2 Materials and methods ...... 91 4.2.1 Study sites and farm characteristics ...... 91 4.2.2 Study design and management ...... 91 4.2.3 Monitoring and measurements ...... 93 4.2.4 Baseline soil and manure analyses ...... 94 4.2.5 Economic and risk analysis ...... 95 4.2.6 Statistical analysis ...... 97 4.3 Results ...... 98 4.3.1 Farmer-field sites characteristics ...... 98 4.3.2 Maize grain yields and response to treatments ...... 99 4.3.3 Economic profitability and risk analysis ...... 103 4.3.4 Effects of other management practices and environmental factors ...... 107 4.4 Discussion ...... 112 4.4.1 Effect of treatments on grain yield and economic profitability ...... 112 4.4.2 Understanding variability in yields and responses ...... 113 4.4.3 Explaining variability in yields and responses...... 114 4.4.4 Opportunities and implications for scaling out ...... 116 5.5 Conclusion ...... 117

xiv

Contents

5 Using DSSAT model to support decision making regarding microdosing 119 Abstract………………………………………………………………………… 120 5.1 Introduction ...... 121 5.2 Materials and methods ...... 123 5.2.1 Experimental data ...... 123 5.2.2 CERES-Maize model ...... 125 5.2.3. Model application: long-term simulation experiment ...... 130 5.3 Results ...... 131 5.3.1 Climatic conditions during the experimental years ...... 131 5.3.2 Model calibration (manure and broadcast fertilizer) ...... 132 5.3.3 Model validation ...... 134 5.3.4 Sensitivity analysis ...... 137 5.3.5 Model testing for fertilizer microdosing ...... 141 5.3.6 Long-term scenario analysis under fertilizer microdosing ...... 143 5.4 Discussion ...... 147 5.4.1 Model response to manure and conventional fertilization ...... 147 5.4.2 DSSAT response to fertilizer microdosing practice ...... 147 5.4.3 Long-term scenario analysis under fertilizer microdosing ...... 148 5.5 Conclusion ...... 150 6 Conclusions and perspectives 151 6.1 Introduction ...... 152 6.2 Main findings ...... 152 6.2.1 Potential of localized application of manure and fertilizers ...... 152 6.2.2 Variability in maize response and profitability on farmers’ fields ...... 154 6.2.3 Long-term scenario analysis under fertilizer microdosing ...... 156 6.3 Main implications and policy recommendations ...... 156 6.4 Perspectives for futur research ...... 159 References 163 Appendices 187 Publications and Conferences 197 About the author 201

List of Figures

1.1 Total harvested area, annual total production and maize grain yield (1961-2016) (FAOSTAT, 2016) ...... 4 1.2 Percentage of acreage (%) for selected crop in the study area (Source: Tovihoudji, unpublished survey data) ...... 16 1.3 A map of Benin showing the agro-ecological zones and the study zone ...... 17 1.4 Means and standard deviations of the monthly rainfalls observed between 1961 and 2016 in the Borgou region, northern Benin (Source: ASECNA Parakou Climate Database)...... 19 1.5 A map of Benin showing the municipality of Bembèrèkè and the village of Ina (the experimental site) ...... 20 2.1 Schematic representation of the FYM and fertilizer application spots with respect to maize plants. Application spots are alternated from one year to the next...... 28 2.2 Rainfall distribution from sowing to harvest in 2012, 2013, 2014 and 2015, with an indication of the major crop development stages. The sowing date corresponds to 0 DAS...... 32 2.3 Maize grain (a) and stover (b) yields following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years. Error bars represent standard error of the differences comparing means within a year. Note: Data from the 3M treatments are missing in 2014 and 2015...... 35 2.4 Maize grain yield response to nitrogen (a), phosphorus (b) and potassium (c) as a result of the combined nutrient input from fertilizer and manure over four years (2012-2015)...... 36 2.5 Agronomic efficiency (AE) of N, P and K (in kg grain kg-1 N, P or K) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012-2015)...... 37 2.6 Changes in soil organic carbon (a), available P (b) and exchangeable K (c) in the 0–20 cm soil layer in the vicinity of the planting hills following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate after two (2013), three (2014) and four (2015) cropping years. Error bars represent standard error of the differences comparing means within a cropping year. Note: Data from the 3M treatments are missing in 2014 and 2015...... 39 2.7 Relationship between maize yields, soil organic carbon (a) and available P (b) and between soil organic carbon and available P (c) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate after two (2013), three (2014) and four (2015) cropping years...... 40 2.8 Gross margins (GM) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012-2015). Error bars represent standard error of the differences between means for each year. Note: Data from the 3M treatments are missing in 2014 and 2015 ...... 42

xvii

List of Figures

2.9 a) Benefit cost-ratio (BCR) and b) value-cost ratio (VCR) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012-2015). Error bars represent standard error of the differences between means for each year. Note: Data from the 3M treatments are missing in 2014 and 2015...... 43 2.10 Sensitivity analysis of the VCR for a ±50% fluctuation of fertilizer cost (a) and maize grain price (b) following the application of 3 (3M) or 6 (6M) t ha-1 of manure and 50 (50F) or 100% (100F) of the recommended mineral fertilization rate. Note: due to the lack of data for the 3M treatment in 2014 and 2015, only the data from the first two years (2012 and 2013) were used for the sensitivity analysis...... 44 3.1 Rainfall distribution from sowing to harvest in 2014 and 2015, with an indication of the main crop development stages...... 66 3.2 Cumulative rainfall in-between soil moisture measurement dates, and plant available water (PAW) from sowing to physiological maturity in 2014 and 2015. Error bars are standard errors of differences between PAW means within each measurement date...... 67 3.3 Maize grain (a) and stover (b) yields as influenced by fertilizer application method across five manuring practices in 2014 and 2015. NM and TM refer to no manure and transported farmyard manure; and Cor-0, Cor-1 and Cor-2 refer to corralling in the same year, or 1 and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR mean fertilizer microdosing option 1 and 2 and the recommended fertilizer rate, respectively. Error bars are standard errors of differences between means within each manure stratum. Within each manure stratum bars with the same letter are not significantly different at p < 0.05...... 68 3.4 Fertilizer use efficiency (FUE) across five manuring practices in 2014 and 2015. NM and TM refer to no manure and transported farmyard manure; and Cor-0, Cor-1 and Cor-2 refer to corralling in the same year, or 1, and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR mean microdosing option 1 and 2 and the recommended fertilizer rate, respectively; Error bars are standard errors of differences between means within each manure stratum. Within each manure stratum bars with the same letter are not significantly different at P < 0.05...... 72 3.5 Annual N, P and K inputs (IN; transported manure in black and mineral fertilizer in gray) and uptakes (OUT; hatched pattern) in the no manure (NM) and transported manure (TM) strata as a function of the mineral fertilization treatment. Values are averages of 2014 and 2015. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars are standard errors of differences between means for uptakes. Note: Nutrient balances = IN – OUT...... 74 3.6 N (a), P (b) and K (c) inputs (IN; manure applied through corralling in black and mineral fertilizer in gray) and uptakes (OUT, dotted or hatched patterns) for a three-year corralling cycle as a function of the mineral fertilization treatment. Each compartment is the average of two years. Cor-0, Cor-1, and Cor-2 refer to corralling in the same year, or 1 and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars

xviii

List of Figures

are standard errors of differences between means for uptakes. Note: Nutrient balances = IN – OUT...... 76 3.7 Evolution of the number of maize grains per cob and the 1000-grain weight as a function of the mineral fertilization treatment. Values are averages over all manure strata and both experimental years. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars = standard deviation...... 81 4.1 Location of Ina district (Municipality of Bembèrèkè) in northern Benin and distribution of rain gauges and demonstration sites...... 93 4.2 Cumulative probability density function of grain yields (kg ha−1) for the different treatments across the two years of the trials. MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1...... 101 4.3 Stability analysis of maize grain yields in the control and fertilized treatments across the two years of the trials (2014–2015). Environmental mean is the average yield of all treatments at a given farmer field site. MD1= microdose option 1, MD2=microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1...... 102 4.4 Absolute response to microdose fertilization as a function of yield in control plots (2014-2015). Since yields in the two microdose rates were not significantly different, the average yield of these two treatments was used with a distinction between microdose (MD) alone and microdose + farmyard manure at 3 t ha-1 (MD+FYM)...... 103 4.5 Cumulative probability distributions of value-cost ratios (VCR) following different treatments and scenarios of input and output prices over the two years of the trials (2014–2015). Vertical dashed lines represent a VCR of 2. MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1. See Table 4.2b for an explanation of the different scenarios...... 106 4.6 Maize grain yields as influenced by different levels of weed pressure (a) and previous crop (b) over the two years of the trials (2014–2015)...... 109 5.1 Daily rainfall, solar radiation, maximum and minimum temperatures in 2014 and 2015. The horizontal red line shows the growing period (from planting to harvest)...... 132 5.2 Comparison between observed and simulated time-series of maize LAI (a), above ground biomass (b) and soil water content in the 0-0.2m (c) and 0.20- 0.40 m layers (d) during model calibration in 2014. NM = no manure; NF = no fertilizer; 100F = 100% of the recommended fertilizer rate; 3M= farmyard manure at 3 t ha-1. Error bars =standard deviation (n = 3); RMSE = root mean- square error; E1 = coefficient of efficiency; d = index of agreement...... 133 5.3 Model validation: comparison between observed and simulated maize LAI (a) and aboveground biomass (b) for all treatments not used during calibration (6 dates of measurement per treatment from 20 to 90 days after sowing) as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate (broadcast fertilizer) over two years. Error bars =standard deviation (n = 3); RMSE = root mean-square error; E1 = coefficient of efficiency; d = index of agreement...... 135

xix

List of Figures

5.4 Model validation: comparison between observed and simulated maize grain (a) and aboveground biomass yield (b) at harvest as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate (broadcast fertilizer) over two years. Error bars =standard deviation (n = 3); RMSE = root mean-square error; E1 = coefficient of efficiency; d = index of agreement...... 136 5.5 Sensitivity of grain yield to: (a-d) selected soil chemical and physical variables for four selected treatments and (e) application method and depth of fertilizer N rates under 0 (NM) or 3 (3M) t ha-1 of manure in the 2014 growing season. a-d: SOC (Soil organic carbon), TN (soil total N), and NO3 (Soil mineral NO3), RCN (Runoff curve number), DUL (ﬁeld capacity); e): BR-0=Broadcast, not incorporated; BR-10=Broadcast, incorporated at 0.10 m depth; BD-0=Banded on the surface; BD-10= Banded 0.10 m beneath surface or bottom of the hole...... 139 5.6 Sensitivity of maize aboveground N uptake (a, b), grain yield (c, d) and aboveground biomass (e, f) at harvest to incremental changes in the N stress coefficient from 0 to 90 kg ha-1 of N fertilizer rates using the hole-placement method at 0.10 m depth in 2014 and 2015. Error bars denote standard deviation (n = 3) for the two microdosing rates without manure (NM-MD1 and NM- MD2)...... 140 5.7 Evolution of RMSE values of N uptake (principal axis), yield and aboveground biomass (secondary axis) at harvest to incremental changes in the N stress coefficient for MD1 and MD2 treatments over the two years during the optimization process...... 141 5.8 Comparison between observed (box-and-whisker plots) and simulated (black points) N uptake (a, b) maize grain (c, d), and aboveground biomass yield (e, f) at harvest as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and two microdosing rates (MD1-MD2) in 2014 and 2015 using CTCNP2=0.20. MD1=microdosing option 1; MD2= microdosing option 2. .. 142 5.9 Frequency distributions of maize grain yields as simulated by DSSAT over a 32- year period (1984 to 2015) in response to combined application of manure and fertilizer microdosing. NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure...... 145 5.10 Maize yield variability as simulated by DSSAT over the study period (32 years— 1984 to 2015) in response to combined application of manure and fertilizer microdosing. NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure. INST = inter-annual standard deviation...... 146

List of Tabless

List of Tables

1.1 Characteristics of the bio-climatic zones of Benin...... 15 2.1 Initial soil chemical and physical properties of the experimental field ...... 26 2.2 Input and output prices used in the economic analysis ...... 30 2.3 Summary of the mixed-model repeated measures analysis of variance on all variables over the four cropping years (2012-2015)...... 34 3.1 Quality of the organic amendments...... 57 3.2 Calculated or estimated nutrient flows used for the full balance calculation...... 61 3.3 Soil properties of the experimental fields on unamended plots (initial) as well as after corralling (mean ± SE; n =3)...... 65 3.4 Maize harvest index and yield components as influenced by mineral fertilization across five manuring practices in 2014 and 2015...... 70 3.5 Summary of the results of ANOVA for nutrient uptake (OUT) using log- transformed data over the two years of the trials (2014–2015)...... 73 3.6 Average partial and full balances calculated for the NM, TM and corralled treatments, for different mineral fertilization practices...... 77 4.1 Number of field demonstration sites per year in the five target villages...... 93 4.2 Input and output prices (A), and description of scenarios (B) used in the economic and risk analysis...... 96 4.3 Descriptive statistics of major farmer-field sites characteristics (2014-2015, all data)...... 99 4.4 Effect of treatments on maize grain yield and descriptive statistics across the two years of the trials. The statistical analysis was performed on square root transformed yield data...... 100 4.5. Economic analysis (US$ ha−1) following different treatments over the two years of trial (2014–2015). Gross margin (GM), net return (NR) and Value Cost Ratio (VCR) calculation was based on the average prices for inputs and outputs (Scenario S0; see 4.2B)...... 105 4.6 Proportion of fields (%) with value-cost ratios (VCR) <1 or <2 depending on the treatments and for different scenarios of input and output prices over the two years of the trials (2014–2015). See 4.2B for an explanation of the different scenarios...... 107 4.7 Results of the linear mixed model analysis of individual variables to explain maize grain yields (square root transformed) over the two years of farmer field trials (2014–2015). Non-significant variables (p-values > 0.1) are not shown. There were no significant interactions with microdose and or manure treatments...... 108 4.8 Results of the multivariate linear mixed model analysis to explain the variability in maize grain yields (square root transformed) over the two years of farmer field trial (2014–2015)...... 110 4.9 Linear regression model using absolute yield response (kg ha-1) in microdose plots (mean of MD1 and MD2) as dependent variable over the two years of trial (2014–2015)...... 111 5.1 Default and adjusted genetic coefficients of maize cv. DMR-ESR-W used in CERES-Maize...... 128 5.2 Soil physical and chemical characteristics at the experimental sites used for calibrating and evaluating the CERES-Maize model...... 129 5.3 Results of model calibration for the 3 selected treatments in 2014...... 133 xxi

List of Tabless

5.4 Statistical indicators showing the relationship between simulated and measured maize grain and biomass yield as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and two microdosing rates (MD1-MD2) over the two years (2014-2015)...... 143 5.5 Summary of simulated impact of yearly rainfall variations (32 years - 1984 to 2015) on maize grain yield (kg ha-1) in response to combined application of manure and fertilizer microdosing...... 144 5.6 Summary of simulated impact of seasonal climate variations (32 years—1984 to 2015) on value cost ratio (VCR) in response to combined application of manure and fertilizer microdosing...... 146

xxii

List of Acronyms and variables

1, 2, or…6M 1, 2 or… 6 t manure ha-1 100F 100% of the recommended fertilizer rate 50F 50% of the recommended fertilizer rate AE-N Agronomic efficiency of N AE-P Agronomic efficiency of P AE-K Agronomic efficiency of K AEZ Agro ecological zone ANOVA ANalysis Of Variance APSIM Agricultural Production Systems Simulator BCR Benefit cost ratio CEC Cation exchange capacity CERES Crop Environment Resource Synthesis Cor-0 Corralling in the same year Cor-1 Corralling one year before the experiment Cor-2 Corralling two years before the experiment CRA- Nord Agricultural Research Station of Northern Benin CV Coefﬁcient of variation DAS Days after sowing DM Dry Matter DMR-ESR-W Downy Mildew Resistant, Early-Streak Resistant -White DOY Day of the year DPP Direction de la programmation et de la prospective DSSAT Decision Support System for Agrotechnology Transfer DST Decision Support Tool EF Modeling efﬁciency Eq. Equation Exch. Exchangeable FAO Food and Agricultural Organization (United Nations) FUE Fertilizer use efficiency FYM Farmyard manure GDD Growing degree days GM Gross margin HI Harvest Index HSD Honest Significant Difference ICRISAT International Crops Research Institute for the Semi-Arid Tropics INSAE Institut National de la Statistique et de l’Analyse Economique ISFM Integrated Soil Fertility Management K Potassium LAI Leaf area index LAImax Maximum leaf area index LMM Linear mixed model MAEP Ministère de l’Agriculture, de l’Elevage et de la Pêche Max Maximum MD1 Microdosing option 1 MD2 Microdosing option 2 Min Minimum

xxiii

List of Acronyms and variables

N Nitrogen NF No fertilizer NH4 Ammonium NM No manure NPK N-P-K (15-15-15) fertilizer NO3 Nitrate NR Net return NUE Nutrient Use Efficiency ONASA Office National pour la Sécurité Alimentaire du Bénin OUT Output P Phosphorus P or Prob. Probability PAW Plant available water pH-H2O Soil pH in demineralized water pH-KCl Soil pH in KCl r Correlation coefficient R² Determination coefficient R2 adj. Adjusted determination coefficient REML Restricted maximum likelihood RMSE Erreur quadratique moyenne/root mean-square error RRMSE relative RMSE RR 100% of the recommended fertilizer rate RUE Rainfall or Radiation Use Efficiency S.E.D Standard error of the (mean) difference SD Standard deviation SLN Specific Leaf Nitrogen SOC or OC Soil Organic Carbon SOM Soil organic matter SONAPRA Société Nationale pour la Promotion Agricole SSA Sub-Saharan Africa TLU Tropical Livestock Unit TM or FYM Transported farmyard manure VCR Value cost ratio

xxiv

Chapter 1

General introduction

“Scientific research is one of the most exciting and rewarding of occupations.”

Frederick Sanger

Chapter 1. General introduction

1.1 Context 1.1.1 Economic importance of agriculture in Sub-Saharan Africa

The agricultural sector is the backbone of the economy of the majority of countries in Sub-Saharan Africa (SSA). It accounts for about 30-50% of the gross domestic product, is the main source of income for more than 60% of the population, and provides over 40% of export earnings (FAOSTAT, http: //faostat3.fao. Org /). Over 90% of the food produced on the African continent is grown by smallholder farmers, mostly under rainfed conditions. However, despite the promising prospects that agriculture offers for sustainable development, Africa’s agricultural productivity is still low compared to other countries in the world, leading to chronic food insecurity. Indeed, the average yield of most cereals is low, ranging from 0.5 to 1.5 t ha-1, against a potential yield of more than 5 t ha-1 (i.e, yield achievable under no biotic, water or nutrient constraints; FAOSTAT, http: //faostat3.fao. Org /). Major factors contributing to such low productivity are: i) low inherent and decreasing land quality (soil type, soil fertility, soil water and slope), ii) poor crop and farm management practices (land preparation, choice of crop varieties, planting, fertilization, irrigation, crop protection from biotic factors such as weeds, diseases and pests, etc…), iii) high variability and change in climate factors (precipitation, temperature, radiation, evapotranspiration, wind speed etc…) and iv) limited farm assets (income, labour, training, size and intensity) and socio- economic conditions (institutional, technical and population). Among these factors, soil fertility is considered as the major biophysical limitation for crop productivity (e.g., Bationo and Buerkert, 2001; Giller et al., 2011; Vanlauwe et al., 2011; Affholder et al., 2013; Okumu et al., 2011; Wairegi et al., 2010; Fermont et al., 2009). This is mainly due to the low inherent fertility of the soils, low and decreasing levels of organic matter in the soil, nutrient mining, the low availability or poor quality of organic amendments and the limited or inappropriate use of mineral fertilizers due to high market fertilizer prices, high risk and limited access to credit (e.g., Mapfumo and Giller, 2001; Bationo and Buerkert, 2001; de Ridder et al., 2004; Henao and Baanante, 2006; Bationo and Waswa, 2011; Tittonell and Giller, 2013; Mueller et al., 2012). The changes in rainfall amount and distribution as well as changes in temperature observed in recent years is expected to aggravate the soil fertility constraints (e.g., IPCC, 2007; Gaiser et al., 2011; Traoré et al., 2013; Folberth et al., 2013; Yegbemey et al., 2014). However, several studies

1.2 Problem statement and analysis in SSA reported that crop development and productivity are generally limited more by land management practices and use of nutrient inputs rather than water (Giller et al., 2006; Twomlow et al., 2011; Mueller et al., 2012; Tittonell and Giller, 2013; Srivastava et al.; 2017).

1.2 Problem statement and analysis 1.2.1 Agriculture and maize-based cropping systems in Benin: extensive and low inputs

Agriculture dominates Benin's economic development policy. This sector contributes about 35% to the country's gross domestic product, employs over 70% of the population and generates 75% of its export earnings (INSAE, http://www.insae-bj.org/statistiques/). Among all cereals, maize (Zea mays L.) is the most important staple crop and source of calories in the diets of the population. It occupies about 82% of the total land area under cereals and accounts for about 84% of cereal production (DPP/MAEP, 2010). Its average consumption is estimated at 85 kg / inhabitant / year and can reach 100 kg / inhabitant / year in the large urban centers of southern Benin, notably Cotonou and Porto-Novo (Nago, 1989). For a long time, maize production and consumption were confined to the southern parts of the country, but now it has extended to the northern regions. Recently maize production has also been growing due to demand from neighboring countries, amongst others Nigeria and Niger. From 1961 to 2016, maize production in Benin gradually increased from 219,593 to 1,354,344 tons (FAOSTAT, 2016) (Figure 1.1). During the same period, the land allocated to maize production increased from 375,650 to 968,030 ha and yields grew from 584 to 1422 kg ha-1 (FAOSTAT, 2016) (Figure 1.1). Farming systems in Benin, like most agricultural systems in sub-Saharan Africa, are predominantly rainfed and dominated by smallholder subsistence producers and pastoralists (Livingston et al., 2011; Smale et al., 2011). As part of the Strategic Plan for Strengthening the Agricultural Sector adopted in May 2017, the Government of Benin has promoted eight food crops (ie maize, rice, cassava, yam, pineapple, cashew nuts, oil palm and vegetables) as an essential component of its approach to diversify agricultural production and further fight hunger and poverty (PSDSA, 2017). In the frame of this plan, maize is a strategic crop for improving the livelihoods of smallholder farmers. Thus, increasing maize productivity would be strategically interesting by increasing export revenues and thereby improving national and domestic food security.

Chapter 1. General introduction

Despite its central role in Benin (for both food security and the rural economy), maize productivity remains low, usually below 1500 kg ha-1 (Figure 1.1) and well below potential yields. Among all possible constraints, soil fertility depletion is the most severe threat to food security, sustainable agricultural production and rural development in the country (e.g., Saïdou et al. 2004; Saïdou, 2006; Srivastava, 2010). It is principally the result of mismanagement of agricultural land and historical dynamics of the political-ecological system and regional land policies (Carsky et al., 2001; Yemadjè et al., 2012). In northern Benin, climate variability and change has further contributed to the low productivity in recent years (Tidjani and Akponikpè, 2012; Yegbemey et al., 2014; Akossou et al., 2016). Given the importance of maize for both food security and the rural economy in Benin, and since opportunities for expansion of cultivated land are often limited to marginal lands, improving its production cannot come from area expansion but productivity gains through appropriate management techniques that can restore and maintain the quality of agricultural land and narrow the yield gap.

1600 1600

Area harvested kg)

6 1400 Production 1400

Grain yield )

1200 1200 1 - 1000 1000

800 800

600 600

ha), Production (x10 Productionha), 3 3

400 400 ha(kgyield Grain

200 200

Area (x10 Area 0 0 1960 1970 1980 1990 2000 2010 2020 Year Figure 1.1. Total harvested area, annual total production and maize grain yield (1961- 2016) in Benin (FAOSTAT, 2016)

Traditionally, the fertility of unproductive fields was regenerated principally through medium- or long-duration fallows, but this practice is no longer feasible due to increasing land pressure and competing land-use demands (Samaké et al., 2005; Pascual and Barbier, 2006). In addition, smallholder farmers used traditional organic amendments such as crop residues or cattle manure transported from the homesteads to the fields or applied through corraling, but nowadays these resources are insufficient to cover farmer’s needs (Vanlauwe and

1.2 Problem statement and analysis

Giller, 2006; Valbuena et al., 2012; Castellanos-Navarrete et al., 2014). In an effort to sustainably increase maize production and to contribute to the reduction of the soil fertility degradation problem, a number of technologies have been developed, tested and made available to the extension services through a techno- economic approach known as “Transfer of technology”. These technologies include mineral fertilizers, cover crops, improved agro-forestry and crop rotation etc…. But, the proposed technologies often do not take sufficiently into account the socio-economic conditions of smallholder farmers. Therefore, adoption of these promising technologies proved limited (Versteeg et al., 1998; Adégbola et al., 2006; Adjei-Nsiah et al. 2007; Nederlof and Dangbegnon, 2007). Besides developping new technologies specific to farmer’s socio-economic conditions, Beninese agriculture still needs more methods that can improve nutrient utilization efficiency and economic profitability.

1.2.2 Towards improved manure management for enhancing crop yields and soil fertility

Among the traditional practices, cattle manure is the most important amendment for maintaining soil chemical and physical fertility in smallholder agriculture in SSA (e.g., Bationo et al., 1998; Schlecht et Buerkert, 2004; Rufino et al., 2007; Zingore et al., 2008; Akponikpè et al., 2008; Rahman et al., 2014). However, any substantial increase in soil organic matter and nutrient content would require high amounts of farmyard manure (up to 25 t of manure ha-1) and continuous application over a long period (Mapfumo and Giller, 2001; Zingore et al., 2011; Rusinamhodzi et al., 2013). Such high amounts of manure are usually difficult to apply by smallholder farmers because of it limited availability and lack of means of transportation. Indeed, in the study area, farmers (80%) who generally have less than five heads of cattle would have access to 2765±1827 kg ha-1 of manure (mean±standard deviation; Tovihoudji, unpublished survey) if all manure could be returned to the fields. Given the low availability of manure, recommendations for its use should be very specific to the site and to the socio-economic conditions of the farmers. Otherwise the adoption of recommendations may fail. Numerous attempts have been made to develop solutions that improve the quantity or quality of manure (McIntire et al., 1993; Williams, 1999; Williams et al., 1995) or that increase the resource use efficiency by testing various application and distribution methods at plot scale (Fatondji, 2002; Gandah et al., 2003). Besides improvements in the recycling of organic resources through, for instance, a better integration of crop-livestock activities (Vanlauwe et al., 2010), it is thus necessary to develop approaches that enhance the efficiency of use of the organic

Chapter 1. General introduction amendments. One example of such approach is the zaï system developed in the Sahelian zone, in which small quantities of manure or compost are concentrated in small planting pits (20-30 cm in diameter and 15-20 cm deep). The technique greatly enhances the use efficiency of organic amendments compared to broadcast application (Fatondji, 2002; Fatondji et al., 2006). In the zaï system, the localized application of organic amendments in the planting hills (or ‘hill- placement’) is combined with water harvesting, both of which contribute to the greater productivity of zaï compared to conventional tillage with a broadcast application of organic amendments. However, this technique requires labor investment to dig the holes, but this has not affected the adoption of this technique by some smallholder farmers cultivating marginal lands since the activity is carried out during the dry season after crop harvesting (Wildemeersch et al., 2015). Nevertheless, other studies have demonstrated that hill-placement of organic amendments in the absence of planting pits also leads to increased efficiency and yields (Otinga et al., 2013; Ibrahim et al., 2015a) possibly as a result of better plant uptake of the limited amount of nutrients (especially N and P) due to root proliferation and the smaller nutrient losses favored by this method (Ibrahim et al., 2014, 2015a). In addition, the concentration of the manure around the planting hills may favor moisture retention which would enhance microbial decomposition and nutrient release. Furthermore, in the more humid tropics like northern Benin, where drought is a lesser constraint, hill-placement of organic amendments without having to dig planting pits offers a serious advantage over the zaï as this considerably reduces the labor requirements (Fatondji et al., 2006). However, whether continuous hill-placement of manure at the low rates that more closely match the reality of farmer’s practices may sustain maize production and enhance soil fertility is unknown. Another traditional practice for soil fertility maintenance in SSA is the overnight corralling of livestock (cattle and/or small ruminants) (de Rouw and Rajot, 2004; Schlecht and Buerkert, 2004; Bielders and Gérard, 2015). Corralling is the preferred practice of smallholder farmers because it does not require carts or human labor to transport manure to the fields. In this practice, animals are confined during the dry season for a number of nights to a small part of the field by tying them to poles or by building enclosures of branches, after which they are moved to a different location in the field. Farmers who have no or few cattle generally establish a stubble-grazing contract with Fulani herders in exchange for crop residues at harvest. In addition, the animals trample the manure with urine during corralling, which ensures a partial mixing in the topsoil. During corralling, high amounts of manure and urine are applied in the same field or spot every 3 to 5 years (Schlecht et al., 2004; Suzuki et al., 2014). The application rates are highly variable, depending on the number of animals, species (sheep, goats,

1.2 Problem statement and analysis cattle), duration of penning and the farmer's ability to ensure a good distribution of manure (Brouwer and Powell, 1998; Schlecht and Buerkert, 2004; Akponikpè et al., 2014). For example, Gandah et al. (2003) and Schlecht and Buerkert (2004) observed that 1.5 to 17.5 t DM ha-1 of manure is left in the fields by cattle during corralling in Niger. The average application rates through corralling are generally about 10 t dry matter ha-1 with maximum values up to 30 t ha-1 (Brouwer and Powell, 1998; Michels and Bielders, 2006). Considering these large quantities, the residual effects may extend over several years (Brouwer and Powell, 1998; Esse et al., 2001; Hoffmann et al., 2001; Gandah et al., 2003; Schlecht et al., 2004; Bielders and Gérard, 2015) depending on the quality of manure produced by livestock which depend also on livestock feeding (Delve et al., 2001; Sangaré et al., 2002; Castellanos-Navarrete et al., 2014). However, on corralled plots, significant nutrient losses (through leaching, volatilization and denitrification) occur (Brouwer and Powell, 1998; Diogo et al., 2013), which reduces its fertilizing value and may pollute the environment. The challenge is to efficiently manage this practice in order to improve on-farm nutrient cycling within the livestock- manure-soil-crop continuum. For example, nutrient losses may be reduced or even avoided by reducing the application rates (Gandah et al., 2003) or targeting preferentially low fertility fields (or parts of fields) where low yields are expected, as commonly practiced by farmers (Buerkert et al., 2000).

1.2.3 Potential of fertilizer microdosing in improving maize productivity

In SSA, farmers complain about the shortage of organic matter for composting and the shortage of fodder for livestock (e.g., Valbuena et al., 2014). Several researchers have calculated the amount of manure needed to replace the nutrients removed from fields by crops. The general conclusion is that the application rates required to achieve sustainable crop production are generally much higher than what is available to smallholder farmers (McIntire and Powell, 1995; Harris, 2002; Mapfumo and Giller, 2001). As a result, the use of mineral fertilizers has been strongly encouraged in many places in SSA for solving the problems of nutrient mining and increasing agricultural production (Tittonell et al., 2008b; Chianu et al., 2012; Vanlauwe et al., 2014). The opportunities for intensification through fertilizers in smallholder maize farming systems are limited due to difficulties of farmers to pay mineral fertilizers due to their high cost, lack of credit and poor transport and marketing infrastructure. Moreover, recommended fertilizer rates for maize (200 kg ha-1 of

NPK15-15-15 and 100 kg ha-1 of urea) are often too costly and involve a high

Chapter 1. General introduction financial risk for the poor smallholder farmers (Abdoulaye and Sanders, 2005). In northern Benin, mineral fertilizer is mainly used for cotton, the main cash crop in the country. Cotton production also plays a major role in maintaining food security because it serves as an entry point through which fertilizers are provided to other food crops (Honfoga, 2012). This relationship between cotton and food production is supported by the subsidy system that allows farmers who agree to grow cotton to have additional fertilizer for food crops. This additional quantity received by farmers as a bonus for their commitment to growing cotton is generally considered insufficient, leading producers to divert fertilizer from cotton production to food crops. This situation calls for exploring fertilizer management practices for improving the efficient use of the limited resources of smallholder farmers. In response to the limitations of conventional fertilizer recommendations and to encourage farmers to use mineral fertilizers, scientists at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) developed a precision-farming technique called ‘fertilizer microdosing’, ‘microfertilization’ or ‘microdose fertilization’ (e.g., Buerkert et al. 2001; Muehlig-Versen et al., 2003). Instead of broadcasting fertilizers evenly across the field with or without subsequent incorporation, microdosing consists in hill-placed (or localized) application of small quantities of mineral fertilizer (1-6 g hill−1 of NPK15–15–15 or

DAP18–46–0, equivalent to 4-35 kg N ha-1 and 3-25 kg P ha-1 depending on the crop) together with the seed at sowing. Besides reducing the quantity of fertilizer to be applied, the microdose technique has proven to be an effective technique to increase fertilizer use efficiency and crop yield (sorghum and millet) and to reduce investment costs and financial risk for smallholder farmers. For example, average millet grain yield and farmer income increases ranging from 4% to 144% and 52 to 134%, respectively, over the unfertilized control, have been reported following microdose fertilization under on-farm trials in Mali, Niger and Burkina Faso (Tabo et al., 2007; Aune et al., 2007; Camara et al., 2013). Consequently, it has been presented as a major step along the agricultural intensification pathway for SSA (Aune and Bationo, 2008; Twomlow et al., 2010; Aune and Coulibaly, 2015). Also, Bielders and Gérard (2015) showed this technique to be a risk- reducing strategy in case of late sowing. Therefore, it could be part of adaptation strategies to food crises caused by climate variability. However, unlike for sorghum and millet, few studies have evaluated the performance of fertilizer microdosing on maize. Existing studies involving maize and fertilizer microdosing were so far performed in southern and eastern Africa under drier conditions than those prevailing in northern Benin (Twomlow et al., 2010; Sime and Aune, 2014; Kisinyo et al., 2015). The results of these studies consistently showed that fertilizer microdosing can increase maize grain yields by 19–50%

1.2 Problem statement and analysis across a broad range of soil, farmer management, and climate conditions. A recent study on fertilizer microdosing in the humid forest zone of Ghana also reported an average increase of maize grain yields by 32 to 99% across cropping systems and soil types (Okebalama et al., 2016). The positive effect of hill-placement of phosphorus and ammonium has been largely attributed to both early crop development and a root-growth stimulating effect (e.g., Jing et al., 2010; Ibrahim et al., 2014, 2015a). According to Jing et al. (2012), this enhancement of maize root growth is associated with improved nutrient and water uptake. Indeed, hill-placement of nutrients exerts characteristic effects on root growth and also on root morphology, comprising root elongation, lateral root branching and proliferation and even branching of root hairs (Zhang and Forde, 1998; Hodge, 2004; Robinson, 2001; Li et al., 2014). The close spacing of root tips and root hairs in the localized nutrient patches also locally increases the surface of exudating roots and therefore the effective rhizosphere concentrations of root exudates involved in nutrient mobilization (Figure B.1, appendix; Jing et al., 2012). Subsequently, an increase in lateral roots within the upper soil layer at an early maize growth stage may stimulate the uptake of native phosphorus (a limiting nutrient in African soils) (Ma et al., 2013; Smit et al., 2013) through enhanced production of phosphatases (Li et al. 2004) and P mobilization in the rhizosphere. In addition, the effect of localized application of N (ammonium and urea) plus P fertilizer in maize has been attributed to its ability to induce rhizosphere acidification at the early stage of crop growth due to NH4-induced H+ release (Jing et al., 2010, 2012). Jing et al., (2012) postulated that the induced rhizosphere acidification could promote solubilization of calcium phosphates and also increase the solubility of other nutrients with limited solubility at high soil pH, such as Fe, Zn, and Mn. Early studies have also demonstrated that localized application of nutrients facilitates photosynthesis through enhancing chlorophyll biosynthesis and leaf area development (Raab and Terry, 1994; Jing et al., 2010). Besides plant growth promotion, the improvement of micronutrient status (Fe, Zn or Mn) may have positive effects on photosynthesis and resistance to stresses and diseases (Jing et al., 2012). Last but not least, localized fertilizer placement may be considered as a weed control strategy as it can particularly favor nutrient acquisition by crops at the expense of weeds, allowing them to be more competitive against them (Blackshaw et al., 2004; Melander et al., 2005; Jamil et al., 2012, 2014; Légère et al., 2013). Although, it has been indicated that fertilizer microdosing increases crop yields and fertilizer use efficiency, the application of this technique requires an additional person at the time of sowing to apply the fertilizer. Labor demand is high due to the fact that this technique involves making holes and closing them (Mashingaidze et al., 2013), which may lead to delayed sowing and low yields. But

Chapter 1. General introduction recent research on this technology in Niger has shown that farmers can delay the application of fertilizer until two to three weeks after sowing without significantly reducing yield and economic profitability (Hayashi et al., 2008). Hence the fertilizer can be applied at a time when labour demand is lower. Another aspect that is not sufficiently investigated about this technology is the fact that the large amount of dry matter that is produced may remove more nutrients from the soil than the amount supplied. In other words, microdosing could lead to the depletion of soil nutrients. Although some scientists consider this claim as an exaggeration under the marginal conditions of smallholder farmers in SSA (Buerkert and Schlecht, 2013; Aune and Coulibaly, 2015), Ibrahim et al. (2016) recently confirmed this statement by reporting negative partial nutrient balances of -37 kg N, -1 kg P and -34 kg K ha-1 under fertilizer microdosing in a low-input millet cropping system in Niger. Therefore, besides the agronomic evaluation of microdosing for maize systems, it is necessary to further evaluate its impact on the nutrient balance in order to develop mitigation strategies and enhance its sustainability.

1.2.4 Potential of combining hill-placed manure and fertilizer

Most existing studies have reported substantial positive effects of the combined application of organic amendments and mineral fertilizers in addressing soil fertility depletion and thereby crop productivity (e.g., Akponikpè et al., 2008; Sanginga and Woomer, 2009; Vanlauwe et al., 2010; Rasool et al., 2010; Kihara et al., 2011; Chivenge et al., 2011; Opala et al., 2010; Bedada et al., 2014). This approach is known as “Integrated Soil Fertility Management” (ISFM). Given the constraints that exist in terms of availability and access to manure and fertilizer, it is necessary to find options that encourages smallholder farmers to adopt the ISFM approach in order to achieve optimum efficiency by improving food production. One of the options available may be the combination of fertilizer microdosing with hill-placement of small and realistic amounts of manure (Ncube et al., 2007, Ibrahim et al., 2015a) or exploiting the residual effects of livesock corraling under microdosing (Manyame, 2006; Bielders and Gerard, 2015). Whereas many studies have investigated the interaction between organic amendments and broadcast fertilization, the combined effects between fertilizer microdosing and organic amendments have seldom been studied (Bielders and Gerard, 2015; Ibrahim et al., 2015a, b). It is well established that nitrogen fertilization in the absence of organic amendment supply can increase soil

1.2 Problem statement and analysis acidification (Liu et al., 2010; Zhou et al., 2014; Adams et al., 2016). Given the low application rates involved in microdosing, this acidification process will be slower than in the case of the higher recommended fertilization rates (Adams et al., 2016), but it should nevertheless be avoided. Also, it has been repeatedly demonstrated that the combined application of organic amendments and fertilizers helps buffer the acidification process (Liu et al., 2010; Bado et al., 2012; Adams et al. 2016). In addition, the nutrient mining potential of fertilizer microdosing may be more intense in sole microdose fertilization compared to combined application with manure (Ibrahim et al., 2016).

1.2.5 On-farm participatory research approach for better targeted fertilizer microdosing recommendations

Most of the existing microdosing studies have been carried out on research stations. It is well known that research station results do not necessarily accurately reflect the reality. Responses to mineral fertilizers are often lower and more variable under typical farmers' conditions than responses from research stations (Tittonell et al., 2008b; Sileshi et al., 2010). Though field studies have consistently established the benefits of microdose fertilization in low input farming systems when considering average yield increases, there is a growing concern that such average responses are insufficient to properly assess the agronomic and economic performances of this technology considering the diversity of smallholder farming environments and practices (e.g., Tittonell et al., 2005, 2010, 2011; Zingore et al., 2011; Giller et al., 2011; Chikowo et al., 2014, Bielders and Gerard, 2015; Falconnier et al., 2015; Vanlauwe et al. 2016). Existing studies that reported the variability of response to microdose for a variety of environmental factors and management practices were so far performed mostly for millet in Niger. The responses ranged between nil to +2000 kg ha-1 compared to average yields in unfertilized farmer’s fields (~400-500 kg ha-1) (Buerkert et al., 2001; Bationo et al., 2005; Tabo et al., 2011; Bielders and Gérard, 2015). In southern Zimbabwe, Twomlow et al. (2010) reported a large variability in maize yield response to microdose fertilization, from negative values (~ -250 kg ha-1) to about +2000 kg grain ha-1 across a broad range of soil, farmer management, and seasonal climatic conditions. It is apparent from the above-mentioned studies that maize yield response to microdose fertilization is expected to depend both on environmental conditions (e.g., rainfall, soil) and crop management practices (e.g., planting density, weeding intensity) and their interactions for a given cultivar within the same agro-ecological zone. Thus, characterizing the extent of the yield response

Chapter 1. General introduction variability and its causes remains an important step to develop meaningful and flexible recommendations that allow farmers to use scarce fertilizer and organic resources effectively (Giller et al., 2011). With such understanding, appropriate recommendations with known levels of risk may be issued to smallholder farmers, which would greatly benefit the credibility of the technology and ultimately help its rapid diffusion.

1.2.6 Developing decision support tools regarding the fertilizer microdosing technology

Due to the complexity and dynamics of agro-ecosystems and in order to better understand and support decision making in agricultural systems, the use of dynamic simulation models has been widely recommended. Such models can be used to suitably complement short-term experiments whose results are highly site and year-specific due to high spatial variability in soil properties and high inter- and intra-annual rainfall variability (Akponikpè et al., 2010; 2011). Several models have been developed to assess maize growth, development and yield, among which APSIM (Agricultural Production Systems Simulator; Keating et al., 2003) and DSSAT (Decision Support System for Agrotechnology Transfer; Jones et al., 2003) are the two most widely used. The suitability of the DSSAT/CERES- Maize model for simulating maize growth and yield under different soil, management and climatic conditions in smallholder farming systems in SSA has been widely tested across several regions, including the sub-humid regions of Benin (e.g. Igue et al., 2013; Saidou et al., 2017; Amouzou et al., 2018). However, unlike for conventional fertilization practices, only a few studies have so far attempted to combine experimental and modeling approaches in order to assist decision-making by smallholder farmers and policy makers regarding the use of fertilizer microdosing. Indeed, modeling fertilizer microdosing using soil-plant- atmosphere models such as DSSAT and APSIM represents a specific challenge since such 1-D models are not well suited to deal with localized fertilizer placement. Previous attempts were all based on the use of APSIM in the context of southern and eastern Africa (Cooper et al., 2008; Twomlow et al., 2008; Turner and Rao, 2013). However, none of these studies actually demonstrated that APSIM was capable of properly reproducing crop response to microdose fertilization since simulation results were not compared to measured data. DSSAT/CERES-Maize like most of crop models uses the “N approach” for simulating N dynamics. This approach is based on an empirical parameterisation of the minimum, critical, and maximum nitrogen concentration in plant tissue as functions of growth stage (e.g., Jones and Kiniry, 1986; Sinclair

1.2 Problem statement and analysis and Amir, 1992; Jamieson et al., 1998; Brisson et al., 1998; Bouman and van Laar, 2006; Lizaso et al., 2011). Without further model adaptations, the low amount of N applied by fertilizer microdosing is likely to lead to N deficiency in the modeled plants. This may result in a reduction in green leaf area index (LAI) and/or dilution of specific leaf nitrogen (SLN, g m-2). A reduction in LAI will lead to a decrease in solar radiation interception by the crop which ultimately will result to a decrease in radiation use efficiency (RUE), photosynthetic capacity and biomass production, while the SLN dilution reduces RUE per unit leaf area (Vos et al., 2005; Lemaire et al., 2008; Li et al., 2009; Ratjen and Kage, 2016). As a result, the capacity of the model to capture consequences on the N balance in the physiologically dynamic manner (based on root/shoot biomass) required under fertilizer microdosing technique may be limited. However, the DSSAT CERES- Maize model is flexible with generic crop, soil and plant N management sub- models that can be optimized to simulate a speciﬁc crop and nutrient management practice (e.g., Lizaso et al., 2011; Liu et al., 2012; Ngwira et al., 2014; Corbeels et al., 2016). When accurately calibrated and validated, the model can be used to perform long-term scenario analyses in order to provide information that can enhance strategic decision-making by smallholder farmers and policy makers for enhancing maize productivity, profitability and food security.

1.3 Research questions and objectives 1.3.1 Research questions

Based on the above-mentioned research needs, this research aims at addressing the general question: ‘To what extent can maize productivity, resource use efficiency and profitability be sustainably increased through localized application of amendments and fertilizers in northern Benin?’ This question can be solved by addressing the following sub-questions: 1) Is localized application of realistic amounts of manure and mineral fertilizer a viable option to improve maize productivity, resource use efficiency and profitability? 2) To what extent can the application of fertilizer microdosing to maize contribute towards improved yields, greater nutrient use efficiency and reduced nutrient mining, in combination or not with various manure management practices? 3) How do maize yields and farmers’ economic benefits vary with differences in environmental and management factors within the same agro-ecological

Chapter 1. General introduction

zone under fertilizer microdosing? And to what extent do these factors explain on-farm yield variability and response to microdosing? 4) Can the DSSAT-CERES-Maize model capture maize response to the combined use of fertilizer microdosing and hill-placed manure? And to what extent does fertilizer microdosing effectively and sustainably increase maize productivity and profitability, and reduce risk under long-term climate variation?

1.3.2 Research objectives

This thesis aims at assessing the agronomic and economic potential of localized application of amendments and fertilizers in maize-based cropping system in northern Benin, with a long-term goal of making recommendations, through field trials and decision support tools, for smallholder farmers in order to improve maize productivity, food security and household income.

Four specific research objectives were pursued:

1) Explore the short and medium-term potential of hill-placed farmyard manure and mineral fertilizer alone or in combination for improving maize productivity, soil fertility and economic profitability (Chapter 2). 2) Assess the agronomic potential, nutrient efficiency and nutrient balance of different systems following the application of fertilizer microdosing and organic manure alone or in combination (Chapter 3); 3) Understand variability in maize yields and yield response to the combined application of fertilizer microdosing and manure, and evaluate the consequences of this variability for farmers’ (economic) benefits through on- farm demonstrations (Chapter 4); 4) Evaluate the maize yield variability induced by inter-annual climatic variations following hill-placement of fertilizer and manure using experimental data and the DSSAT model (Chapter 5).

1.4 Brief description of the study area

1.4 Brief description of the study area 1.4.1 Research sites and socio-economical characteristics

The Republic of Benin is located in West Africa between the latitudes 6°30’ and 12°30’ N and longitudes 1° and 30°4’ E (Figure 1.2). It is bordered by Nigeria (East), Togo (West), Niger (North- East), Burkina-Faso (North-West), and the Atlantic sea (South). It is a tropical country with 3 bio-climatic zones (Sudanian in the north, Sudano-Guinean in the centre and Sub-Guineo-Congolian in the south) (Table 1.1) and 8 Agro-Ecological Zones (AEZ) (Sanchez et al., 2012). It covers a land area of 114.763 km² and the population has been estimated at 10, 008,749 inhabitants in 2013 with an average density of 87 inhabitants per km2 (INSAE, RGPH-4, 2013, http://www.insae-bj.org/statistiques/).

Table 1.1. Characteristics of the bio-climatic zones of Benin. Zones Rainfall regime Mean annual Mean annual rainfall (mm) temperature (°C) Sudanian Uni-modal 680 – 950 26 – 40 Sudano-guinean Uni-modal 900 – 1100 25 – 29 Sub-Guineo-Congolian Bi-modal 1200-1300 25 – 29

The present study was conducted in north-eastern Benin (in the department of Borgou) in agro-ecological zone 3 (AEZ 3; Figure 1.2). Located between 8°45’-12°30’ N and 2°00’-3°15’ E, this area represents 45% of the national territory (about 52,093 km2). The total and active population were evaluated at 1,214,249 and 312, 366 inhabitants in 2013, respectively, increasing at an annual rate of 5 % (INSAE, RGPH-4, 2013). The region is considered as the main production zone of food and cash crops. Households in the study area are mainly engaged in farming, from which a majority derive their livelihoods. More than 50% of households have farm sizes smaller than 5 ha. Agricultural production is principally rainfed with traditional techniques (rudimentary tools, no machinery, no animal draught power, no or rare use of inorganic fertilizer). The main food crops produced are maize (Zea mays L.), sorghum (Sorghum bicolor), millet (Pennisetum glaucum L.), and legumes such as cowpea (Vigna inguiculata L.), groundnuts (Arachis hypogaea L.) and soybean (Glycine max L.). Rice (Oryza sativa), cassava (Manihot esculenta Crantz) and yam (Discorea spp.) are generally less important (Figure 1.2). Cash cropping is restricted to cotton and cashew. Sale of cash crops such as cotton is first among income generating activities, followed by food crops such as maize, legumes, rice, cassava and yam (not necessarily

Chapter 1. General introduction surpluses). The average size of the households is 17 persons with 9 members working full time on the farm (Tovihoudji, unpublished survey data). The land:labor ratio (i.e., the number of adults working on the farm per unit area of land available per family) is low in most households, indicating land limitations. The farm is generally fragmented in different fields/plots, usually situated near the houses, but sometimes far away. The cattle breed present in the region is called the Borgou cattle, a crossing between zebu (Bos indicus L.) and taurine (Bos taurus L.). Borgou cattle can withstand heat, are medium trypano-tolerant, humidity-tolerant and also good draught animals given their small size. About 1,500,000 heads (67 % of the national stock) were present in the Borgou-Alibori region in 2013 (ANOPER Bénin, 2014). In addition, this region counts about 850,000 heads of domestic sheep (Ovis aries L.) and goats (Capra hircus L.) (ANOPER Bénin, 2014). Different livestock production systems such as transhumant herders, sedentary agro- pastoralists and farmers using animal power coexist. All this indicates the major role of agriculture and livestock in the livelihoods of population in this region.

Maize Cotton Rice Cassava Millet Yam Legumes Vegetables Others

Figure 1.2. Percentage of acreage (%) for selected crop in the study area (Source: Tovihoudji, unpublished survey data)

1.4.2 Biophysical characteristics

Climate

Based on the rainfall amount, the study area covers the Sudanian and Sudano- Sahelian climatic zones (Figure 1.3). The rainy season is unimodal (5 to 6 months; May-October) with July-September generally the wettest period of the year (Figure 1.4) and allows for only one cropping season per year. From 1961 to 2016, the average annual rainfall and the number of rainy days were 1104 mm

1.4 Brief description of the study area

(±205) and 78 days (± 21), respectively. The average annual temperature is 27.2°C (± 0.6). The relative humidity fluctuates between 30% and 70%. A marking characteristic of the climate in northern Benin is the temporal variability of annual rainfall, with a general trend to decreasing rainfall and aincreasing mean temperature during the past decades (Gnanglè et al., 2011; Yegbemey, 2014). Given this picture, the region is expected to be more affected by projected climate change than the southern part of the country (MEHU, 2011; Yegbemey, 2014). For example, recent studies of MEHU (2011) revealed that the rainfall amount will stay stable (±0.2% variation) in southern Benin, whereas it will decrease by about 13 to 15% in northern Benin by 2100. Predictions indicate that the yield levels of many crops (e.g. maize, cassava, beans, groundnuts, rice, cotton, and sorghum) will significantly decrease by 3 to 18% compared to their current levels by 2025 in northern Benin (MEHU, 2011). These predictions reported maize as the most sensitive to future climate conditions.

Figure 1.3. A map of Benin showing the agro-ecological zones and the study zone

Chapter 1. General introduction

Soils

The study area is characterized by ferruginous tropical soils (87 %), ferralitic soils (10 %) and vertisols (3 %). All our experiments were conducted at the Ina district (municipality of Bembèrèkè, AEZ 3; Figure 1.5) with ferruginous tropical soils. Thus, we will focus only on this type of soil. The ferruginous tropical soil (French soil classification system) corresponds to Lixisols according to the World Reference Base (Youssouf and Lawani, 2002). The agronomic characteristics of these soils are highly variable and heterogeneous (Youssouf and Lawani, 2002). Soil texture consists of friable gravely sand to gravely sandy-loam and is highly leached in most of the region. Their depth is highly variable (0.40-2 m) with reddish color more pronounced in some sites and increasing ferruginous gravel content with depth. Most important constraints and restrictions that limit the ability of these soils include low inherent fertility, moisture stress and hard-setting (Kayombo and Lal, 1993; Lal, 1995; Youssouf and Lawani, 2002). In fact, soils have a low water-holding capacity and are generally medium to well-drained. The pH is slightly acid to neutral (pH-H2O: 5.0 – 6.7). The chemical composition of these substrates shows an accumulation of ferric oxides and hydrates with low concentrations of aluminum oxides and low cation exchange capacity (2 to 5 cmol+ kg-1), but with high base saturation (60-80%). Organic carbon content is very low (0.1-0.6 %) and soils are deficient in N (0.01-0.04%). Most agricultural lands in the region have a slope varying from 1 to 10 %, sometimes with a slightly undulating topography and granite or gneiss as parent material. Despite their constraints and restrictions, the chemical fertility is much better than the ferralitic soils uncountered in the south of the country that are characterized by extreme weathering. Thus they are relatively good growth media for plants (Youssouf and Lawani, 2002).

1.5 Outline of the thesis

350

300

250

200

150

100 Rainfall Rainfall (mm/month) 50

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months Figure 1.4. Means and standard deviations of the monthly rainfalls observed between 1961 and 2016 in the Borgou region, northern Benin (Source: ASECNA Parakou Climate Database).

1.5 Outline of the thesis

The thesis will combine three different approaches: short and medium-term on- station experiments, on-farm participatory demonstration trials and long-term model simulations. Our general approach is based on 1) assessing the agronomic potential and economic profitability of hill-placed manure and mineral fertilizer (or fertilizer microdosing) through two on-station experiments; 2) quantifying through on-farm participatory demonstration trials the variability in yield response, economic profitability and risk associated with some of the most promising treatments and 3) modeling the response of maize across a range of rainfall conditions to further evaluate the sustainability of these practices in northern Benin. This thesis is made up of 4 main sections devoted to specific issues related to the experiments and modeling outcomes and presented in chapter format. Following the general introduction (Chapter 1), Chapter 2 analyses a 4-year on-station field experiment, where the effect of hill-placed manure and mineral fertilizer was investigated, alone or in combination, on maize performance, soil fertility and economic profitability. Chapter 3 presents results regarding maize productivity and nutrient use efficiency, and nutrient balances under combined application of fertilizer microdosing and traditional organic manure management practices. Chapter 4 investigates the performance of and

Chapter 1. General introduction risk associated with the use of the fertilizer microdosing technique over a range of soil, climate and crop management techniques within the same agro-ecological zone through on-farm demonstrations. In Chapter 5, the DSSAT model was calibrated and validated for different fertility management practices (using the experiment of the Chapter 3) in order to support decision making regarding fertilizer microdosing for maize production in northern Benin. This thesis ends with a general discussion (Chapter 6), in which the main findings and the general conclusions and recommendations were provided. Besides, some implications and suggestions were made for future research.

Figure 1.5. A map of Benin showing the municipality of Bembèrèkè and the village of Ina (the experimental site)

Chapter 2

Combining hill-placed manure and mineral fertilizer enhances maize productivity and profitability in northern Benin*

*This Chapter has been published as: Tovihoudji, P.G., Akponikpè, P.B.I., Adjogboto, A., Djenontin, J.A., Agbossou, E.K., Bielders, C.L. 2017. Combining hill-placed manure and mineral fertilizer enhances maize productivity and profitability in northern Benin. Nutrient Cycling in Agroecosystems, 110(3): 375-393. 21

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

Abstract

Throughout much of sub-Saharan Africa, maize production is characterized by low productivity due to the scarce availability and use of external inputs and recurrent droughts exacerbated by climate variability. Within the integrated soil fertility management (ISFM) framework, there is thus a need for optimizing the application of fertilizers and manure to better use the limited nutrient resources and increase crop yield and farmer income. An on-station experiment was conducted in Northern Benin over a 4-year period to evaluate the effect of hill placement of mineral fertilizer and manure on maize yields and soil chemical properties. The treatments consisted in the combination of three rates of manure (0 (NM), 3 (3M) and 6 (6M) t ha-1) and three levels of fertilizer (0% (NF), 50% (50F) and 100% (100F) of the rate recommended by the National Agricultural Research System (76 kg N + 13.1 kg P + 24.9 K ha-1)). On average across the fertilizer rates, hill-placement of manure significantly improved soil organic carbon content (SOC), available P and exchangeable K after 4 years by up to 124%, 166% and 77%, respectively, compared to the initial values in the vicinity of the planting hills. As a result of the nutrient inputs and improved soil properties, yields increased steadily over time for all manure and fertilizer combinations. Value-cost ratios and benefit-cost ratios were > 2 and generally as good or even better for treatments involving 50F compared to NF or 100F. Although applying half the recommended rate of fertilizer without manure as currently done by many farmers appears to make economic sense, this practice is unlikely to be sustainable in the long-term. Substituting 50F for 3M or complementing 50F with 3M are two possible strategies that are compatible with the precepts of ISFM and provide returns on investment at least as good as the current practice. However, this will require greater manure production, made possible in part by the increased stover yields, and access to means of transportation to deliver the manure to the fields.

Keywords: Manure; Fertilizer; Maize yields; Nutrient use efficiency; Profitability

2.1 Introduction

In Sub-Saharan Africa (SSA), agriculture plays an important function in the economic growth and rural livelihoods. However, the low inherent soil fertility and loss of soil fertility through nutrient and organic matter depletion negatively affect agricultural productivity (Bationo and Waswa, 2011; Tittonell and Giller, 2013). According to Henao and Baanante (2006), “the declining fertility of African soils because of soil nutrient mining is a major cause of decreased crop yields and per capita food production and, in the mid to long-term, a key source of land degradation and environmental damage”. These soil fertility constraints have been aggravated in recent years by climate variability and change, and mismanagement of agricultural land (Traoré et al., 2013; Yegbemey et al., 2014). The combination of these constraints has resulted in low yields for food crops. In the case of maize for instance, which is a major food crop across much of SSA (Smale et al., 2013), grain yields seldom surpass 1 t ha-1 in the majority of smallholder farms, compared with a potential yield of 3 to 5 t ha-1 (FAOSTAT, 2016). Traditionally, unproductive fields were regenerated mainly through medium or long-duration fallows, but this practice is no longer feasible due to the increased population leading to pressure on land and competing land-use demands (Samaké et al., 2005; Pascual and Barbier, 2006). Consequently, organic amendments (such as farmyard manure (FYM) or compost) are, more than ever, essential components of soil fertility maintenance strategies (Zingore et al., 2008; Nezomba et al., 2015). Besides their nutrient supply function, the addition of organic amendments is crucial for sustaining soil organic carbon and biological activity levels as well as soil physical and chemical quality (Mando et al., 2005; Zingore et al., 2008). However, the application rates required to achieve sustainable crop production are generally much higher than what is available to smallholder farmers because of limitations in fodder supply, cattle or labor (Mapfumo and Giller, 2001). Besides improvements in the recycling of organic resources through, for instance, a better integration of crop-livestock activities (Vanlauwe et al., 2010), it is thus necessary to develop approaches that enhance the efficiency of use of the organic amendments. One example of such approach is the zaï system developed in the Sahelian zone, in which small quantities of manure or compost are concentrated in small planting pits. The technique greatly enhances the use efficiency of organic amendments compared to broadcast application (Fatondji et al., 2006). In the zaï system, the hill-placed application of organic amendments is combined with water harvesting, both of which contribute to the greater productivity of zaï compared to conventional tillage with a broadcast application of organic 23

Chapter 2. Combined application of hill-placed manure and mineral fertilizer amendments. However, other studies have demonstrated that hill-placement of organic amendments in the absence of planting pits also leads to increased efficiency and yields (Otinga et al., 2013; Ibrahim et al., 2015a). In the more humid tropics, where droughts are a lesser constraint and hence water harvesting is less crucial to achieving a decent harvest, hill-placement of organic amendments without having to dig planting pits offers a serious advantage over the zaï as this considerably reduces the labor requirements (Fatondji et al., 2006). Nevertheless, it has long been recognized that organic resources cannot ensure by themselves the closure of the nutrient balance since they are merely a form of imperfect nutrient recycling (Valbuena et al., 2014). In addition, the livestock-mediated fertility transfers from grazing land to cropland, which traditionally greatly contributed to sustaining crop production, breaks down when the pasture-cropland ratio drops below a certain threshold because of increasing population pressure (Andrieu et al., 2014). Finally, given that the availability of organic amendments is already limiting simply to sustain current yields (Valbuena et al., 2014), these organic resources cannot, by themselves, lead to the large-scale crop intensification that must be achieved to feed the fast growing population. As a result, the use of mineral fertilizers has been strongly encouraged in many places in SSA for increasing agricultural production (Chianu et al., 2012; Vanlauwe et al., 2014). Fertilizer application rates have generally been developed with a view to maximizing yields. However, these recommended rates are not affordable to most smallholder farmers and imply a high financial risk (Chianu et al., 2012). Consequently, adoption of mineral fertilization for food crops has been minimal. In places where fertilizers have been adopted, farmers have often reduced the quantity applied as compared to the rates recommended by extension services. For example, actual mineral fertilizer application rates in Northern Benin are generally smaller or equal to half the rate recommended by extension services (Kormawa et al., 2003). Given this situation, as for organic amendments, strategies have been developed to enhance the use efficiency and economic return from mineral fertilizers. One such successful technology is fertilizer microdosing (Buerkert et al., 2001; Muehlig-Versen et al., 2003). In this technique, a few grams of mineral fertilizer are hill-placed at sowing and/or within a few weeks after sowing. The technique has been shown to result in large yield increases with higher fertilizer use efficiencies and higher value-cost ratios than previous fertilizer recommendations (Sime and Aune, 2014). Although mineral fertilization is a key component for agricultural intensification (Vanlauwe et al., 2014), the use of mineral fertilizers alone cannot solve the problem of declining soil fertility and loss of productivity because fertilizers do not compensate for the many other environmental functions of

2.1 Introduction organic amendments (Vanlauwe et al., 2011). In addition, the sole use of some mineral fertilizers can enhance soil acidification and therefore lead to a decline in productivity (Adams et al., 2016). These considerations have led to the integrated soil fertility management paradigm, which advocates the combined use of organic and inorganic amendments for crop intensification purposes (Akponikpè et al., 2008; Vanlauwe et al., 2010). Numerous studies report substantial positive effects of the combined application of organic amendments and mineral fertilizers in addressing soil fertility depletion in the short and in the long-term by preventing soil acidification and improving soil functioning (Kihara et al., 2011; Bedada et al., 2014; Wei et al., 2016). A meta-analysis by Chivenge et al. (2011) concluded that, across SSA, the combined use of organic inputs and nitrogen fertilizers leads to a greater yield response than either input on its own. In Benin, agronomic researchers have long advocated an approach that combines mineral fertilizers with organic inputs (Vanlauwe et al., 2001a, b; Dagbénonbakin, 2005). Despite the fact that research trials have demonstrated short-term yield benefits from combining nutrient sources, integrated soil fertility management (ISFM) adoption is currently low, because of the above-mentioned constraints regarding the use of organic amendments and inorganic fertilizers at the rates recommended by extension services. There is thus a need to identify practices that better match farmers’ constraints. Although trial results combining mineral and organic fertilizers are fairly common (e.g., Vanlauwe et al., 2001a; Mando et al., 2005; Opala et al., 2010; Bedada et al., 2014), the present study addresses this question for the specific case when both the organic amendments and fertilizers are hill-placed. The increased resource use efficiency that results from hill-placement may lead to substantial economic returns albeit with lower application rates that more closely match the reality of farmer’s practices. Using the results from a 4-year maize cropping trial combining hill-placed manure and mineral fertilizer under rainfed conditions in northern Benin, the present study therefore aimed at identifying efficient and low-cost ISFM practices based on their productivity, resource use efficiency and economic viability.

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

2.2 Materials and methods 2.2.1 Experimental site description

From 2012 till 2015, a field experiment was carried out at the Agricultural Research Centre of Northern Benin (CRA-Nord) of the National Institute for Agricultural Research in Benin. The station is located at Ina village (Ina district, municipality of Bembèrèkè), Northern Benin (9°57’N and 2°42’E, 365 m a.s.l), 70 km northeast of Parakou. The average annual rainfall for the last 30 years at Ina was 1148 ± 184 mm (±SD) and the average daily temperature was 27.5°C (CRA-Nord Climate Database, 1982–2015). The climate is tropical sub-humid with a single rainy season that occurs between May and October. July and August are the wettest months. The soil is classified as ferruginous tropical soil in the French soil classification system and as Lixisols according to the FAO soil classification system. The soil is a loamy-sand with approx. 5% clay in the top 0.2 m, with low organic carbon and total nitrogen and medium phosphorus content (Table 2.1). Prior to the experiment, the experimental field had been cultivated with a continuous maize-sorghum rotation, manual tillage and without mineral fertilizer application. Maize residues were harvested each year for animal feeding as commonly practiced by most farmers in the study area.

Table 2.1. Initial soil chemical and physical properties of the experimental field Parameters 0-20 cm Soil texture Sand (%) 77.5 Silt (%) 17.2 Clay (%) 5.3 Texture Loamy-sand Soil chemical properties

pH-H2O 5.6

pH-KCl 5.3 Organic C (g kg-1) 4.5 Total N (mg kg-1) 320 P Bray-1 (mg kg-1) 9.3

2.2 Materials and methods

2.2.2 Experimental design and treatments

Twenty-seven experimental plots of 20 m2 (4 x 5 m) were arranged in a split-plot design (randomized blocks) with factorial combinations of farmyard manure (FYM; main plot) and mineral fertilizer (sub-plot) in three replicates. The three levels of FYM were no manure (NM) and an annual hill-placement of air-dried FYM at a rate of 3 t ha-1 (3M) and 6 t ha-1 (6M), corresponding to 96 and 192 g FYM hill-1, respectively. The 6M treatment corresponds to the rate recommended by the National Agricultural Research System of Bénin. The FYM was a mixture of cattle manure and crop residues collected from the barn of the Agricultural Research Centre of Northern Benin each year. The FYM was hill- placed 10 days after sowing (DAS) by digging small, 0.1-m diameter and 0.1-m deep holes at a distance of 0.1 m on both sides of each planting hole. The manure was then covered with soil with a long hoe. Unfortunately, because of an experimental error, the data from the 3M treatments could not be used in 2014 and 2015. The three levels of hill-placed mineral fertilizer were no fertilizer (NF), 50% (50F) and 100% (100F) of the rate recommended by the National Agricultural

Research System (200 kg ha-1 of NPK15-15-15 and 100 kg ha-1 of urea (46% N), equivalent to 76 kg N ha-1 + 13.1 kg P ha-1 + 24.9 kg K ha-1). For the 50F and 100F treatments, the corresponding application rates were 3.2 g NPK + 1.6 g urea hill−1 and 6.4 g NPK + 3.2 g urea hill-1, respectively. The 50F treatment was selected because it more closely matches farmer’s practice (Kormawa et al. 2003). NPK fertilizer was applied 15 DAS whereas the urea was applied 45 DAS. The mineral fertilizers were spot-applied at a distance of 10 cm on both sides of each planting hole. NPK fertilizer was not incorporated, whereas urea application was immediately followed by weeding-ridging in accordance with farmer’s practice in the study area. Note that in any given year, manure and fertilizer application spots were orthogonal with respect to the planting hole (Figure 2.1). The application spots of manure and fertilizers were alternated from one year to the next. Each year, a representative sample of the FYM was dried in the oven at 40 °C, ground to pass through a 1 mm sieve, and analyzed for organic carbon as well as total N, P and K. The FYM was composed of 14.8±5.6 % C, 1.4±0.5 % N, 0.3±0.2 % P, and 0.9±0.4 % K, on average. The corresponding mean annual application rates for the 6M treatment were 885±336 kg C, 84±30 kg N, 18±12 kg P and 54±24 kg K ha-1, and half of that for the 3M treatment.

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

Figure 2.1. Schematic representation of the FYM and fertilizer application spots with respect to maize plants. Application spots are alternated from one year to the next. 2.2.3 Trial installation and management

The experimental field was prepared for sowing using standard cultivation practices. At the onset of the experiment, land preparation was done uniformly across all plots by tractor disk- plowing to a depth of 0.2 m. At the time of planting, the experimental plots were leveled manually using rakes. The experimental plots were sown with maize, variety DMR-ESR-W (90 days to maturity). Sowing took place at the onset of the rainy season after the first rainfall event greater than 20 mm on 26 June in 2012, 28 June in 2013, 4 July in 2014 and 20 July in 2015. The planting hills were spaced 0.8 m × 0.4 m (Figure 2.1) with a total of six rows. Maize seedlings were thinned to two plants per hill 2 weeks after planting to attain a plant population of 62,500-plant ha-1 (currently recommended density). Plots were weeded thrice (15, 30 and 45 days after sowing) in each cropping year with a hand hoe. Harvesting took place on 18 October in 2012, 25 October in 2013, 20 October in 2014 and 5 November in 2015. Daily rainfall data was recorded each year with a rain gauge located at the experimental field. To measure total aboveground biomass and grain yields, three middle rows were cut at soil level in each plot. Grain, core and stover sub-samples were oven- dried to a constant mass at 65 °C for 48 h to determine moisture content. Total biomass and grain yields were expressed on a dry matter basis. In order to establish how much productivity could be gained by mineral fertilizer and/or FYM applications, the agronomic efficiency (AE-X) of each

2.2 Materials and methods nutrient X (X = N, P or K) was calculated as a proxy of nutrient use efficiency as follows:

AE-X (kg grain kg-1 X) = (Yt –Yc) / RX (Eq. 2.1) where Yt and Yc are the maize grain yield in treatment t and in the control (kg ha-1), respectively, and Rx is the application rate of nutrient X (kg ha-1). The quantity of nutrient applied was the sum of the nutrients applied through mineral fertilizer and FYM in each treatment.

2.2.4 Soil sampling and analysis

In 2012, prior to sowing, one composite soil sample (0.2 m depth) was taken using several randomly selected points from the entire experimental field. This sample reflects the status of the soil before treatment application. To assess the effects of the different treatments on soil properties, soil samples were taken to a depth of 0.2 m at harvest in 2013, 2014 and 2015 using a standard soil auger around 9-12 randomly selected planting hills from the three middle rows of each plot. For each planting hill, 4 samples were taken at the 4 cardinal locations where FYM and fertilizer had been applied (Figure 2.1). All the visible organic residues were removed by hand and then all samples from each plot were thoroughly mixed to obtain a composite sample before sub-sampling for analysis. Each sample was analyzed for pH (H2O) (soil/water ratio of 1:2.5), organic carbon (van Reeuwijk, 1993), available phosphorus Bray-1 (Van Reeuwijk, 1993) and exchangeable K (van Reeuwijk, 1993). All analyses were carried out at the soil and plant analysis laboratory of the International Crop Research Institute for the Semi-Arid Tropics (ICRISAT, Sadoré, Niger).

2.2.5 Economic analysis

Economic profitability of the different treatments was analyzed based on gross margin, benefit/cost ratio, and value/cost ratio. Fixed costs included the cost of seeds and all major labor charges (field preparation, seeding, weeding, ridging, harvesting, and threshing), whereas variable costs included the cost of fertilizers and FYM as well as labor charges for the application of the different fertilizer and manure rates. Input and output prices were taken as the average market price of the four years (2012-2015; Table 2.2). The price of seeds and fertilizers fixed by the Beninese government through the “Société Nationale pour la Promotion Agricole (SONAPRA)” were used. Since there is no market for manure in the study area and farmers consider it as a free input, the value of FYM was estimated to be equal to the costs required for collecting manure from homestead kraals and transporting it to the field during the experimental period. Labor costs for

Chapter 2. Combined application of hill-placed manure and mineral fertilizer land preparation, planting, manure/fertilizer application, weeding, and harvesting were collected during the experiments through farm diaries. Labor costs of hill application of manure included the costs required for digging holes and the labour costs involved in its application (Table 2.2). For the maize grain price, we used the official price of the “Office National pour la Sécurité Alimentaire du Bénin (ONASA)” database (http://www.onasa-benin.org/). Values fluctuate between a minimum of 120 FCFA kg−1 in September-October and a maximum of 200 FCFA kg-1 in June-July. In this study, we used the average yearly price of 150 FCFA kg-1 (1US$ = 500 FCFA). Total revenue was calculated by multiplying grain yield with the average price of grain. The gross margin (GM) was calculated by subtracting variable costs from total revenue. The net return (NR) was calculated by subtracting the sum of the fixed and variable costs from the revenue. The benefit/cost ratio (BCR) was obtained by dividing the net return by the total cost of cultivation (fixed and variable costs). The value-cost ratio (VCR) was computed as the difference in grain yield between the fertilized and/or manured plots and the control plot multiplied by the unit market price of grain, divided by the cost of applied fertilizer and/or manure. Some simulations were carried out to see how the VCR was affected by fluctuations (- 50% to +50%) in the price of fertilizer and maize grains.

Table 2.2. Input and output prices used in the economic analysis Unit Cost (US$) Inputs Maize seed US$ kg-1 0.70 NPK fertilizer US$ kg-1 0.50 Urea fertilizer US$ kg-1 0.50 Manure US$ t-1 4.00 Labor for maize cultivation Tillage US$ ha-1 60.00 Seeding US$ ha-1 14.00 Manure application US$ ha-1 36.00 Mineral fertilizer application US$ ha-1 12.00 Weeding US$ ha-1 56.00 Harvesting US$ ha-1 28.00 Output Maize grain US$ t-1 300.00

2.3 Results

2.2.6 Statistical analysis

Prior to the analysis, data were carefully checked for normal distribution using the Anderson-Darling test, and homogeneity of variance was assessed using Levene’s test. Agronomic efficiency data were log transformed before analysis of variance because of non-normality. First, a combined analysis of variance (ANOVA) over the four years was performed using a mixed-model, repeated measures analysis of variance. Fixed effects entered into the model included year (as a repeated variable), manure, fertilizer and the year by treatments interaction. Replicates were included as a random effect. Secondly, because there were significant year effects and year x treatment effects, an ANOVA was performed on a year-by-year basis for easier interpretation and comparison using a General Treatment Structure (with split-plot design). Mean separations were performed using the honestly significant difference (HSD)/Tukey’s test at an error probability < 0.05. To describe the relationship between the nutrient inputs, SOC or available P and maize grain yields, simple linear regression analysis was performed. In the absence of data about the rate of nutrient release during amendment decomposition, the total nutrient content of the organic amendments was used instead. We hypothesized that the total amount of nutrient inputs from manure will be make available to plant during the course of a single rainy season because of the hill-placement method and the small amount of manure used in the study. All analyses were done with GenStat Release 12.1 statistical software (GENSTAT, 2009). The post-harvest soil properties were compared to the initial nutrient status of the soils at the start of the experiment (2012).

2.3 Results 2.3.1 Rainfall and temperature distribution during the cropping periods

Rainfall patterns differed across the four years (July-October). The rains were more evenly distributed in 2013 and 2014 compared to 2012 and 2015, despite the greater number of short dry spells and the lower rainfall amount recorded in those two years (Figure 2.2). Rainfall was highest in 2012 (885 mm) followed by 2015 (797 mm), whereas 2014 and 2013 received the lowest rainfall of 694 mm and 650 mm, respectively. Of the total rainfall received in 2012, 52% occurred between the emergence and flowering stages, while in 2013 most of the rain

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

(57%) occurred from 48 till 90 DAS (flowering to maturity stage). In 2014, most of the rain occurred from 63 to 80 DAS, accounting for 38 % of the total rainfall recorded during the cropping period with three dry spells of 8-10 days at the beginning of the growing season (0 to 10 DAS, and 20 to 27 DAS) and at the flowering stage (54-63 DAS). Most of the rain in 2015 occurred between 15 and 40 DAS (the vegetative period), accounting for 55 % of the total rainfall recorded during the cropping period. The mean air temperature varied only slightly during the four maize cropping cycles. Mean daily temperature during the growing period ranged between 21.3 and 31.1, 21.1 and 33.0, 22.0 and 31.0, and 22.1 and 31.4 °C in 2012, 2013, 2014 and 2015, respectively.

Figure 2.2. Rainfall distribution from sowing to harvest in 2012, 2013, 2014 and 2015, with an indication of the major crop development stages. The sowing date corresponds to 0 DAS. 2.3.2 Maize grain yield

Overall, grain yields tended to increase over the 4-year period for all fertilized or manured treatments (Figure 2.3a). Depending on the treatment combinations, yields increased at an average rate comprised between 195 kg ha-1 yr-1 for NM+50F and 604 kg ha-1 yr-1 for 6M+50F. Only the NM+NF treatment showed a decreasing trend, from 1449 kg ha-1 in 2012 to 1073 kg ha-1 in 2015. Based on the mixed-model repeated measures analysis, there was a significant year effect 32

2.3 Results and significant year by treatment interactions (p < 0.001; Table 2.3), hence the results will hereafter be discussed on a year by year basis. The addition of manure significantly increased maize grain yield in all years (p < 0.01; Figure 2.3a). In 2012, grain yield was improved by 28% for 3M compared to NM, on average over all fertilizer application rates, but the application of 6M resulted in only a minor additional increase. On the contrary, in 2013 grain yield was improved on average by 64% and 97%, respectively, for 3M and 6M compared to NM. The response to manure in 2014 and 2015 was rather similar to 2013. Grain yields increased by 94% on average for 6M compared to NM. There was a significant fertilizer effect for maize grain yield in all years as well as a significant manure x fertilizer interaction in all years (p < 0.01) except in 2012 for which the effects of manure and fertilizer were additive (Figure 2.3a). In all years, the NM treatments responded well to the addition of 50F but adding 100F provided little additional gain compared to 50F. Compared to the NM+NF treatment, the response to fertilizer in the absence of manure tended to increase over time (from +821 kg ha-1 in 2012 to +1905 kg ha-1 in 2015), both as a result of decreasing yields in the NM+NF treatment and increasing yields in the fertilized treatments. In 2012, all manure treatments (3M and 6M) responded well to the addition of 50F, with an average yield increase of 878 kg ha-1 compared to NF. 100F provided only little additional gain in grain yield compared to 50F. In 2013, compared to NF the addition of 50F and 100F to the 3M treatment increased grain yield by 510 and 1458 kg ha-1, respectively, whereas the response to 50F and 100F was similar (on average +521 kg ha-1) in the presence of 6M. In 2014 and 2015, grain yields in the 6M treatment increased by 775 and 865 kg ha-1, respectively, after adding 50F (Figure 2.3). Adding 100F to 6M further increased yields by 561 kg ha-1 in 2014 compared to 50F, but no significant increase was observed in 2015.

2.3.3 Maize stover yield

As for grain yield, there was a significant year effect and significant year by treatment interactions on stover yield (p < 0.001; Table 2.3). Stover yields tended to increase over time for all treatments except for the NM+NF treatment for which yields remained fairly stable (Figure 2.3b). Significant effects of manure application were observed in all years except in 2012 (Table 2.3b). Compared to NM, stover yield in 2013 was improved by 61% and 75% for 3M and 6M, respectively, on average over all fertilizer treatments. In 2014 and 2015, stover yields increased by 60% and 74%, respectively, for 6M compared to NM. Significant effects of fertilizer application were observed in all years except in 2012 (Figure 2.3b). On average over all manure treatments, adding fertilizer

Chapter 2. Combined application of hill-placed manure and mineral fertilizer improved stover yields in the 50F and 100F by 39% and 53% in 2013, by 23% and 31% in 2014 and by 69% and 77% in 2015, compared to NF (Figure 2.3b). There was no interaction between any of the treatments in all years, except for 2013. In 2013, compared to NF the addition of 50F increased stover yield by 1697 and 2062 kg ha-1 in the NM and 3M treatment, respectively, while the addition of 100F provided only little additional gain in stover yield compared to 50F. Conversely, no significant increase in stover yield was observed after the addition of 50F in the 6M treatment, but adding 100F increased stover yields by 1147 kg ha-1 compared to NF (Figure 2.3b).

Table 2.3. Summary of the mixed-model repeated measures analysis of variance on all variables over the four cropping years (2012-2015). Source of variation (P-values) Variables Year (Y) Manure (M) Fertilizer (F) Y x M Y x F M x F Y x M x F Grain yield <0.001 <0.001 <0.001 <0.001 0.027 <0.001 0.014 Stover yield <0.001 <0.001 <0.001 <0.001 <0.001 0.002 0.015 AE-N <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.164 AE-P 0.197 <0.001 <0.001 <0.001 0.759 <0.001 0.508 AE-K <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.478 GM <0.001 <0.001 <0.001 <0.001 0.041 <0.001 0.014 BCR <0.001 <0.001 <0.001 <0.001 0.508 <0.001 0.026 VCR <0.001 <0.001 <0.001 0.006 <0.001 0.004 0.516 SOC 0.049 <0.001 0.918 0.063 0.577 0.913 0.469 P-Bray1 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 Exch-K 0.005 0.103 0.006 0.825 0.005 <0.001 0.033

2.3.4 Yield response to nutrient application

Maize grain yield was significantly correlated (p < 0.001) with the N, P and K application rates from the combined manure and fertilizer amendments in all years (Figure 2.4). Although there was some indication of a non-linear response for N in 2013 and 2015 and K in 2013, the nutrient response could be fairly well approximated in general by a linear regression (R² between 0.74 and 0.95), except for K in 2012. The slope of the relationship was year-dependent (Figure 2.4). For N and K, the response was lowest in 2012 and highest in 2015 (Figure 2.4). For P, the response was lowest in 2014 and similar for the other three years.

2.3 Results

Maximum grain yields of the order of 4800 kg ha-1 were achieved in 2015 and 2014, albeit with significantly lower inputs in 2015 (111 kg N, 31 kg P, 49 kg K ha-1) compared to 2014 (202 kg N, 49 kg P, 97 kg K ha-1) because of differences in manure composition (Figure 2.4).

Figure 2.3. Maize grain (a) and stover (b) yields following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years. Error bars represent standard error of the differences comparing means within a year. Note: Data from the 3M treatments are missing in 2014 and 2015. 35

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

Figure 2.4. Maize grain yield response to nitrogen (a), phosphorus (b) and potassium (c) as a result of the combined nutrient input from fertilizer and manure over four years (2012-2015).

2.3 Results

2.3.5 Agronomic efficiency (AE) of N, P and K

There was a highly significant effect of year on the AEs calculated on a grain yield basis, except for P (p < 0.001; Table 2.3). In all years, manure and fertilizer applications had a significant effect on the AE of N, P and K (p < 0.01), except in 2012. There was also a highly significant manure x fertilizer interaction in all years (p < 0.01; Figure 2.5), except in 2012. In general, the AE of N and P remained nearly stable or decreased with increasing rates of fertilizer, for each level of manure. However, this rate of decrease tended to be highest for the NM treatments and lowest for the 6M treatments. For K, the AE decreased with increasing rates of fertilizer application in the NM treatment. It remained nearly stable for the different fertilizer rates in the 3M treatment or slightly decreased in the 6M treatment.

Figure 2.5. Agronomic efficiency (AE) of N, P and K (in kg grain kg-1 N, P or K) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012-2015).

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

2.3.6 Post-harvest soil status

Soil organic carbon (SOC) content showed upward trends in all manured treatments (3M and 6M). On average across the fertilizer rates, SOC increased in 6M from 4.5 g C kg-1 in 2012 to approximately 10 g C kg-1 in 2015 compared to 6.1 g C kg-1 on average in the NM treatment (Figure 2.6a). Though not significant (Table 2.3), there was a tendency for SOC contents to increase with increasing levels of fertilizer (Figure 2.6a). As for SOC, available P tended to increase over time for the 3M and 6M treatments (Figure 2.6b). For NM, available P increased for 50F and 100F but decreased for NF. On average over 50F and 100F, the rate of increase in available P was higher for 6M than NM. The annual increase in available P ranged from 1.5 to 5 mg P kg-1 yr-1 in manured plots compared to 2 mg P kg-1 yr- 1 on average in the no manure treatment. At the end of the experiment, available P was significantly higher (p < 0.05) in the 100F and 50F treatments compared with the NF control treatment whether combined with 6M or NM. Compared with the initial values, soil exchangeable K increased significantly over the four years in all treatments except for the NM+NF treatment in which it declined (Figure 2.6c). On average over the 50F and 100F treatments, exchangeable K tended to increase more rapidly for 6M compared to NM. There was a significant manure x fertilizer interaction on soil exchangeable K in 2013 and 2015 (Figure 2.6c). Exchangeable K increased significantly with increasing rates of fertilizer for a given level of manure. Overall, grain yield was significantly related to SOC (p < 0.001; R² = 0.71; Figure 2.7a) and available P (p < 0.001; R² = 0.86; Figure 2.7b). There was also a significant relationship between available P and SOC (p < 0.01; R² = 0.83; Figure 2.7c).

2.3 Results

Figure 2.6. Changes in soil organic carbon (a), available P (b) and exchangeable K (c) in the 0–20 cm soil layer in the vicinity of the planting hills following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate after two (2013), three (2014) and four (2015) cropping years. Error bars represent standard error of the differences comparing means within a cropping year. Note: Data from the 3M treatments are missing in 2014 and 2015.

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

Figure 2.7. Relationship between maize yields, soil organic carbon (a) and available P (b) and between soil organic carbon and available P (c) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate.

2.3 Results

2.3.7 Economic performance indicators

Overall, there was a highly significant year effect on all the economic performance indicators (p < 0.001; Table 2.3). As for yields, the highest gross margin was observed in 2015 and the lowest in 2012 (Figure 2.8). In all years, increasing application rates of fertilizer (p < 0.01) and manure (p < 0.01) significantly increased gross margins (GM). On average, manure increased the GM by 26 and 32% in 2012 and by 67 and 99% in 2013 for the 3M and 6M treatments, respectively, compared to NM. The GM increased by 96% in 2014 and by 95% in 2015 for 6M, compared to NM (Figure 2.8). In 2012, similar GMs were achieved for 50F and 100F in combination with 3M or 6M (Figure 2.8). In 2013, the 3M+100F treatment was again comparable to the 6M+50F and 6M+100F treatments. In 2014 and 2015, the 6M+50F and 6M+100F achieved comparable GMs. Overall, benefit-cost-ratio (BCR) tended to increase with increasing application rates of farmyard manure. The BCR was generally highest for the 50F treatments irrespective of the manure application rate (p < 0.01; Figure 2.9a). BCR values < 1 were observed for the NM+NF treatment in all years except 2012. BCR of the 50F treatments were close to 2.0 or higher, reaching almost 4 in 2015 in combination with 6M (Figure 2.9a). Unlike the BCR, the value-cost ratio (VCR) generally and significantly (p < 0.01) increased with increasing application rates of manure and with decreasing application rates of fertilizer, except in 2012 where the manure effect was not significant and where NF had lower VCR than 50F and 100F (Figure 2.9b). VCR values of the 50F treatments were 1.5 to 1.6 times greater than those of the 100F treatment, irrespective of the years. In 2014 and 2015, NF and 50F had comparable VCRs when combined with 6M.

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

Figure 2.8. Gross margins (GM) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012-2015). Error bars represent standard error of the differences between means for each year. Note: Data from the 3M treatments are missing in 2014 and 2015

2.3 Results

Figure 2.9. a) Benefit cost-ratio (BCR) and b) value-cost ratio (VCR) following the application of 0 (NM), 3 (3M) or 6 (6M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate over four years (2012- 2015). Error bars represent standard error of the differences between means for each year. Note: Data from the 3M treatments are missing in 2014 and 2015.

As expected, VCR values are sensitive to fluctuations in fertilizer and maize grain prices. If the cost of fertilizers were to increase by 50%, the VCR values of all the treatments combining the 100F treatment would drop below the threshold line of 4 and even below 2 in the case of NM+100F (all other things remaining equal). For the treatments that include 50F, the VCR values always remain > 2

Chapter 2. Combined application of hill-placed manure and mineral fertilizer and even > 4 in the presence of 3M and 6M (Figure 2.10a). A rise in maize price by 25% would result in all treatments having a VCR > 4 except for the NM+100F treatment (Figure 2.10b).

Figure 2.10. Sensitivity analysis of the VCR for a ±50% fluctuation of fertilizer cost (a) and maize grain price (b) following the application of 3 (3M) or 6 (6M) t ha-1 of manure and 50 (50F) or 100% (100F) of the recommended mineral fertilization rate. Note: due to the lack of data for the 3M treatment in 2014 and 2015, only the data from the first two years (2012 and 2013) were used for the sensitivity analysis.

2.3 Results

2.4 Discussion 2.4.1 Soil fertility improvement

The continuous application of manure alone or in combination with mineral fertilizer increased the SOC content in the top 0.2 m of the soil in the neighborhood of the planting hills. This may be attributed to the direct additions of organic C through the manure, or indirectly through enhanced root biomass in the manured and fertilized treatments. Although the addition of manure and fertilizer both enhanced maize above ground biomass production (Figure 2.3), the increased C content in soils cannot be the result of crop residue additions since these were exported from the fields at the end of each cropping year, in accordance with farmers’ current practices. Several studies have indicated that an increase in soil carbon can be observed only if the dose of organic manure is sufficiently high (up to 25 t of manure ha-1) and applied for several years (Liu et al., 2010; Rusinamhodzi et al., 2013). In this study, the application rates were much smaller than these reported rates, yet significant improvements in SOC content were observed. This can be explained by the fact that both the manure application and soil sampling were performed in the vicinity of the plants. By concentrating the manure in pits around the planting hills, substantial soil improvements were achieved within a few years. In addition, the hill-placement of manure and fertilizer may have favored root development near the plant, which may explain the observed increase in SOC content in the no manure, fertilized treatments (Figure 2.6a). It must be remembered that the results of the soil analyses cannot be extrapolated to the entire plot surface because the sampling was restricted to the four cardinal locations were manure was hill-placed (Figure 2.1). In addition, even around the planting hills, some heterogeneity is expected since manure pits were alternated between years. A marked improvement in available P and exchangeable K contents in the topsoil (0–0.2 m) in the vicinity of the planting hills was observed in the plots that received manure compared with the NM plots (Figure 2.6b, c). Other studies have reported strong effects of manure on available P and exchangeable K (Zingore et al., 2008). The higher available P may be attributable to P released during manure decomposition on the manured plots compared to no manure plots. Unsurprisingly, mineral fertilization strongly increased available P and to a lesser extent exchangeable K (Figure 2.6b, c).

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

2.4.2 Maize productivity and resource use efficiency

Overall, there was a trend of increasing maize yields over the 4-year period for all fertilized or manured treatments (Figure 2.3a, b). In SSA, numerous studies reported that rainfall quantity and distribution are one of the most critical factors affecting growth and yield of rainfed crops (e.g., Traoré et al., 2013; Eyshi Rezaei et al., 2014; Ripoche et al., 2015). However, there was no indication from the rainfall records that climatic conditions improved during the course of the experiment, neither in terms of intra-annual rainfall distribution nor in terms of total seasonal rainfall (Figure 2.2). For example in 2012, when the rainfall was relatively high and evenly distributed, the grain yield response to fertilizer and manure applications was low. Early sowing and a longer growing season may favor higher yields. However, sowing dates tended to be increasingly delayed over the years (from 26 June in 2012 to 20 July in 2015), and the length of the growing period was shorter in 2014-2015 (108 days) than in 2012 (114 days) and 2013 (119 days). Finally, despite the occurrence of short dry periods, there was no indication of serious water stress for maize during the experiment, as plant-available water in soil did not reach critically low levels (data not shown). Hence, it seems likely that the increasing yields largely resulted from improved soil fertility (Figure 2.6) because of the beneficial cumulative effects of the manure and fertilizer on soil quality. This is apparent in the significant increase in SOC content near the maize plants. Manure and fertilizer additions also had a positive impact on the available P and exchangeable K levels in soil in the vicinity of the planting hills (Figure 2.6c). The role of soil fertility is further supported by the significant correlations between SOC (p < 0.001) or available P (p < 0.001) and maize grain yields (Figure 2.7). Furthermore, in the NM+NF treatment, the declining grain yields were accompanied by a decline in available P and exchangeable K contents, further emphasizing the strong link between yield and soil fertility. All this points to soil fertility as the main driver of maize grain yields during the experiment and certainly as the main determinant of the observed yield variations over time. Biotic constraints such as weeds, stem-borer, disease etc. were not assessed specifically but there was no indication that these constraints decreased over time and could have resulted in the observed yield trend. Maize responded positively to manure application rates (Figure 2.3). Maize yields were significantly larger for 3M and 6M compared to the NM plot. The lower response in 2012, despite the high seasonal rainfall, may have resulted from the post-emergence application of the manure. The latter most likely resulted in a delayed release of nutrients and possibly in an incomplete decomposition

Chapter 2. Combined application of hill-placed manure and mineral fertilizer during the growing season. This also applies to subsequent years, except that the maize may then have benefited from the residual effects of the amendments applied in the previous year(s). In the present study, the overall response to manure application may have been enhanced as a result of the hill-placement as compared to the usual broadcast application. Fatondji et al. (2009), Otinga et al. (2013) and Ibrahim et al. (2015a) reported benefits from manure hill-placement vs. broadcast application. Hill-placement of manure may result in a better uptake of the limited amount of nutrients by the roots possibly as a result of early root proliferation favored by this method. In addition, the hill-placed manure may favor moisture retention which would enhance microbial decomposition and nutrient release. The concentration of the manure around the planting hills may also explain the strength of the observed correlation between maize grain yield and SOC (Figure 2.7). Mineral fertilizer application significantly increased maize yields compared to the unfertilized control. On average across years and manure treatments, fertilizer application increased grain yields by 760-1320 kg ha-1 and 1069-1557 kg ha-1 for 50F and 100F, respectively, compared to the unfertilized control (NF). Likewise, stover yields increased by 568-2410 kg ha-1 and 616-2676 kg ha-1 for 50F and 100F, respectively, compared to the unfertilized control (NF). Since rainfall is identical in all treatments for a given year, these yield increases following mineral fertilization result in higher rainfall water productivity. In only a few instances (3M treatment in 2013, or 6M treatment in 2014) did the 100F treatments substantially increase grain yields as compared to 50F (Figure 2.3). Overall, the 100F treatment performed only marginally better than the 50F treatment. Consequently, the agronomic efficiency of the 50F treatment is higher on average than the 100F recommended rate (Figure 2.5). Similar trends have been reported by Akponikpe et al. (2008), Chivenge et al. (2011), Vanlauwe et al. (2011) and Kihara and Njoroge (2013). There was a significant interaction between manure and fertilizer for grain yield in all years except 2012 (Figure 2.3). Although synergetic effects have sometimes been reported (Chivenge et al., 2011), this was not observed in the present study. On the contrary, the response to fertilizer additions tended to be lower in the 6M treatments than in the NM treatments between 2013 and 2015, indicating a greater benefit from fertilizers in the absence of manure. In view of the low initial SOC contents at the experimental site (Figure 2.6a), this appears to contradict earlier findings in SSA that showed a positive response of crops to mineral fertilizer in soils rich in organic matter whereas the application of fertilizers in soils poor in organic matter led to no significant crop response (Wopereis et al., 2006; Rusinamhodzi et al., 2013; Kurwakumire et al., 2014).

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

When combining the nutrient inputs from manure and fertilizer, a rather linear relationship between grain yield and total N, P or K input was observed in most years (Figure 2.4). Strictly speaking, the response lines of Figure 2.4 do not reflect the response to single elements, as all three nutrients are present in both the fertilizer and the manure. They should, therefore, be interpreted as the response of maize to N, P or K inputs in the presence of proportional inputs of P+K, N+K, and N+P, respectively. The observed linearity emphasizes the strong dependence of yields on nutrient inputs and hence that soil fertility rather than rainfall was the main factor controlling yields in the present experiment. However, the slopes of the linear regressions are different across years, indicating that some inter-annual factors affected maize yield response to nutrient inputs. Most of the inter-annual effects can be attributed to the cumulative effects of the nutrient inputs as discussed earlier. Indeed, the slope of the regressions tends to increase from 2012 until 2015. In any given year, the response to the amendments therefore reflects the direct effect of the amendments as well as the residual effects of previously applied amendments, which includes residual soil nutrients (e.g. K; Figure 2.6c), previously undecomposed manure as well as improvements in soil properties (e.g., SOC; Figure 2.6a). Nevertheless, one observes that the response was on average better in 2013 than 2014, despite less favorable climatic conditions in 2013 (Figure 2.2). The reason for this discrepancy is unclear. Finally, in the case of P, with the exception of 2014 for which the yield response is clearly lower, the inter-annual effects are absent. This appears to indicate that plant available P achieved non-limiting levels as from the 1st year.

2.4.3 Implications for nutrient management by farmers

From an economic viewpoint, the VCR of 100F was always lower than that of 50F (Figure 2.9). This is because the application of 100F generally resulted in only marginal grain yield increases compared to 50F (Figure 2.3a). In addition, except for 3M in 2013, the VCRs of NF plots were lower or similar to the VCRs of 50F. Hence applying half the recommended rate appears to be an optimal choice in terms of value-cost ratio. A similar conclusion can be drawn from the BCR calculations, which indicate that 50F treatments always perform as well or better than NF and 100F treatments (Figure 2.9). Finally, the VCRs of the 50F treatments were always > 2, which is often considered a minimal condition for technology adoption in risky environments (Kihara et al., 2015). Consequently, any treatment relying on 50F sole or in combination with manure would appear to be an economically sensible choice. This is consistent with the current practice of many farmers who are using half the recommended rate of fertilizer for

2.4 Discussion economic reasons (Kormawa et al., 2003). In practice, many farmers apply 50F without any manure because of limited access to organic resources. Although this is not advisable for reasons explained below, it remains an economically sensible choice since the NM+50F treatment had BCR values close to 2.0 or higher, whereas NM-NF had BCR values of the order of 1 in the last three experimental years, i.e., no net benefit. Applying 50F without manure, though economically viable, should not be recommended in the long-term. Indeed, continuous cultivation of maize without organic amendment has been shown to lead to an increase of soil acidification and an overall decline in soil organic matter content and in the availability of other nutrients (e.g., Mando et al., 2005; Adams et al., 2016; Luo et al., 2017). Organic additions are essential for maintaining soil quality in the long-term and are an integral part of ISFM. Besides supplying micronutrients, organic amendments are also essential to sustain soil life (e.g., Liu et al., 2010; Opala et al., 2010; Chivenge et al., 2011; Kihara et al., 2011; Otinga et al., 2013; Agegnehu et al., 2016). However, the broadcast application of 3 t manure ha-1 is unlikely to substantially ameliorate soil quality. Hence, hill-placement of the manure appears to be a good alternative since it allows to substantially improve soil properties where it matters most, i.e., close to the plants. Given that most smallholder farmers cannot generate large quantities of manure due to the low number of livestock, relying on fertilizer to achieve acceptable yields (> 2000 kg ha-1) seems sensible. However, farmers should be encouraged to value the added biomass (Figure 2.4) in order to produce more manure and gradually substitute fertilizer by manure or complement the fertilizer with manure. As can be seen from Figure 2.9, the 3M+NF treatment provides returns on investments at least as good as the NM+50F treatment. The gross margin is, however, even better for the 3M+50F treatment than for the 3M+NF, and such that the former may be a suitable alternative in situations where labor is not a constraint. Given that the grain yields in the NM+NF treatment were fairly stable across the four years, it appears that the actual VCR values will strongly depend on the yield of the fertilized plots and on the fertilizer or grain prices. As expected, the VCR increases as fertilizer prices decrease or as maize prices increase (Figure 2.10). All treatments remain financially attractive (VCR > 2) even in the case of large fertilizer price increases (+50%) or a substantial drop in maize price (-25%), except for the NM+100F treatment. Hence, the results of the economic analysis will remain valid over a fairly broad range of fertilizer and maize prices. Nevertheless, supporting policies will be particularly required to keep mineral fertilizer affordable and support the internal maize market.

Chapter 2. Combined application of hill-placed manure and mineral fertilizer

The hill-application of low quantities of manure and fertilizer thus appears to be an effective technology in terms of absolute yields, soil improvement, and economic returns. However, hill-placement is labor intensive, which may constitute a limitation to its adoption. The main limitation is likely to be the manure application in pits. However, because the manure application was done after sowing, it does not interfere with crop sowing which is one of the bottlenecks in terms of labor requirement. Given the encouraging results of the on-station trial, large-scale testing of the technology in on-farm trials seems warranted.

2.5 Conclusions

The results of the current study show that soil quality can be significantly improved in the vicinity of the plants because of the hill-placement of limited quantities of manure and/or fertilizer. This increase in soil fertility resulted in a rapid upward trend in maize grain and stover yields. From an economic standpoint, hill-placement of half the currently recommended rate of fertilizer

(i.e 100 kg ha-1 NPK15-15-15 and 50 kg ha-1 of urea) appears sensible and may explain why many farmers already apply this rate in practice. However, although no detrimental effects were observed during the 4 years of the experiment, not applying organic amendments may prove unsustainable in the long-term. The results of the experiment indicate clearly that applying 3 t ha-1 of manure without fertilizer is economically at least as interesting as applying 50F without manure both in terms of gross margin and return on investment as off the second year. Hence, farmers should be encouraged to substitute fertilizer by manure gradually by valuing the increased maize stover production. They could also complement the manure with fertilizer which provides slightly higher gross margins than manure alone. However, measures have to be taken to provide farmers with more means of transportation of the manure from the homesteads to the fields. The economic results (BCR and VCR) of the present study remain valid over a rather wide range of fertilizer and maize prices. This warrants further testing of the technologies over a wider range of soil and climatic conditions through on- farm experiments.

Chapter 3

Fertilizer microdosing enhances maize yields but may exacerbate nutrient mining in maize cropping systems in northern Benin*

*This chapter has been published as: Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bertin, P., Bielders, C.L. 2017. Fertilizer microdosing enhances maize yields but may exacerbate nutrient mining in maize cropping systems in northern Benin. Field Crops Research 213: 130–142.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

Abstract

Fertilizer microdosing is a promising technology to complement traditional fertility management strategies, yet little is known regarding its performance in maize systems in western Africa. This study assessed to what extent the application of fertilizer microdosing to maize in northern Benin may contribute towards improved yields, greater nutrient use efficiency and reduced nutrient mining, in combination or not with various manure management practices. In a 2-year on-station experiment at Ina (northern Benin), four fertilizer options were tested [no fertilizer control, microdosing options 1 (MD1, 23.8 kg N ha-1, 4.1 kg P ha-1 and 7.8 kg K ha-1) and 2 (MD2, 33.1 kg N ha-1, 8.2 kg P ha-1 and 15.6 kg K ha-1), broadcast fertilizer at recommended rate (RR, 76 kg N ha-1, 13.1 kg P ha-1, and 24.9 kg K ha-1)] within five manure strata [manure applied through corralling in the same year (Cor-0) as well as one (Cor-1) and two years (Cor-2) before the experiment, transported manure (TM, 3 t ha-1), and no manure (NM)]. On average across all manure strata and years, fertilizer application significantly increased grain yields by 64% for MD1, 81% for MD2 and 93% for RR compared to the unfertilized control. Yields in MD2 were never different from those in RR. Across the manure strata, there was a general tendency for FUEs to decrease from MD1 (8.0 to 19.0 kg grain kg−1 fertilizer) to RR (4.6-8.0 kg grain kg-1 fertilizer). Maize response to fertilizer microdosing was best in the absence of organic amendments and tended to decrease with increasing fertility. Indeed, the greatest grain yield increases were observed for the NM (+1611 kg ha-1) and Cor-2 (+1468 kg ha-1) strata, followed by TM (+1258 kg ha-1), Cor-1 (+1183 kg ha-1) and Cor-0 (+1126 kg ha-1). Consequently, FUE was also best in NM plots and lowest in recently corralled plots (Cor-0). Fertilizer-induced yield increases resulted principally from larger numbers of grains per cob (+52% on average) and larger 1000-grain weights (+13%). For the NM and TM strata, the partial nutrient balances for the two cropping seasons ranged between -44 and +21, -24 and -9 and -78 and -45 kg ha-1 year-1 respectively for N, P and K depending on the mineral fertilization treatment. The balances ranged between -17 and +54, -77 and -50 and -345 and -228 kg ha-1 over a three-year corralling cycle, respectivel for N, P and K. Except for N applied at the recommended rate, partial nutrient balances were equally or more negative on fertilized plots than on the unfertilized controls, indicating that the tested fertilization options may exacerbate nutrient mining. This was particularly the case for P and K and suggests that microdosing should probably not be used for extended periods. Nevertheless, fertilizer microdosing appears better adapted to the realities of smallholder farmers than the RR while still ensuring very significant yield increases. There is a need to evaluate these nutrient management options in farmer’s fields, taking into account climatic, soil and management conditions to better assess and understand the maize response and the magnitude of nutrient mining and to evaluate its economic benefits and risk.

Keywords: Maize, fertilizer microdosing, manure, yield, fertilizer use efficiency, nutrient balances

3.1 Introduction

In Sub-Saharan Africa (SSA), low inherent soil fertility and soil degradation through nutrient and organic matter depletion remain major constraints to food production and contribute to chronic poverty (Vanlauwe et al., 2015; Barrett and Bevis, 2015). In the semi-arid zone of SSA, this situation is worsened by the recurrent droughts exacerbated by climate variability and change (Traoré et al., 2013). Traditional soil fertility management through vegetated fallows has become ineffective or even impossible to perform as a result of population pressure and rapidly decreasing fallow/cropland ratios (Andrieu et al., 2014). Although essential for the maintenance of soil health, organic amendments such as manure, compost or crop residues are in short supply, suffer from competition for usage and are not able to compensate fully for nutrient losses from cropland (Valbuena et al., 2014). The use of mineral fertilizers has therefore been advocated for a long time to complement organic amendments and as a means to boost productivity beyond what is achievable with the available organic resources. The fertilizer application rates recommended by agricultural research and extension services have generally proven too costly for smallholder farmers. In addition, they involve a high financial risk, the latter being a major factor driving decision making for smallholder farmers (Abdoulaye and Sanders, 2005). This situation calls for the development of innovative low-input technologies that can concurrently replenish soil nutrients as well as improve rainfall use efficiency in order to increase crop productivity. One such technology is the application of small quantities of hill-placed mineral fertilizers at the base of the plants either at sowing or shortly after planting, named fertilizer microdosing or microdose fertilization (Buerkert et al., 2001; Muehlig-Versen et al., 2003). This technique was developed in the Sahel in response to the limitations of conventional fertilizer management recommendations. Nutrient application rates under this technique may be as low as 10-30% of the rates recommended by research and extension for broadcast fertilization (Buerkert et al., 2001; Aune et al., 2007; Twomlow et al., 2010; Camara et al., 2013; Sime and Aune, 2014). It has been proven to be an effective technique to increase fertilizer use efficiency and crop yield (sorghum and millet) and to reduce investment costs and financial risk for smallholder farmers (Aune et al., 2007; Camara et al., 2013). Consequently, the microdose technique has been presented as a major step along the agricultural intensification pathway for SSA (Aune and Bationo, 2008). Although numerous studies have demonstrated the benefits of conventional fertilization practices to intensify maize-based systems (Wopereis et al., 2006; Vanlauwe et al., 2014; Naab et al., 2015), the ensuing 53

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin recommendations have seldom been implemented by smallholder farmers. Nevertheless, compared to other cereals (sorghum and millet), opportunities for intensification of maize production appear high given the high demand for maize in the fast-growing urban centers. This is, in particular, the case in northern Benin, where maize is increasingly replacing cotton as a cash crop. However, unlike for sorghum and millet, few studies have evaluated the performance of fertilizer microdosing on maize. Existing studies involving maize and fertilizer microdosing were so far performed in southern and eastern Africa under drier conditions than those prevailing in northern Benin (Twomlow et al., 2010; Sime and Aune, 2014). The results of these studies consistently showed that fertilizer microdosing can increase grain yields by 19–50% across a broad range of soil, farmer management, and climate conditions. Thus, it is considered to be a promising option for smallholder maize farmers, because of the high fertilizer use efficiency as well as favorable value-cost ratios and gross margins. However, given the high planting densities of maize compared to sorghum and millet, fertilizer microdosing on maize entails the use of much larger quantities of fertilizer which are no longer as different from the recommended rates as is the case for millet and sorghum. Consequently, the efficiency of the technique may be quite different for maize- based systems than what has been reported for millet- and sorghum-based systems. In addition, some authors have argued that fertilizer microdosing does not alleviate nutrient mining (Camara et al., 2013). Although some scientists consider this claim as an exaggeration under the marginal conditions of smallholder farmers in SSA (Buerkert and Schlecht, 2013; Aune and Coulibaly, 2015), Ibrahim et al. (2016) recently confirmed this statement by reporting negative partial nutrient balances of -37 kg N, -1 kg P and -34 kg K ha-1 under fertilizer microdosing in a low-input millet cropping system in Niger. In addition, these nutrient mining tendencies are likely to be more intense in maize cropping systems. Therefore, besides the agronomic evaluation of microdosing for maize systems, it is necessary to further evaluate its impact on the nutrient balance in order to develop mitigation strategies and enhance its sustainability. It is well established that nitrogen fertilization in the absence of organic amendment supply can enhance soil acidification (Liu et al., 2010; Zhou et al., 2014; Adams et al., 2016). Given the low application rates involved in microdosing, this acidification process will be slower than in the case of the high recommended fertilization rates (Adams et al., 2016), but it should nevertheless be avoided. It has been demonstrated repeatedly that the combined application of organic amendments and fertilizers helps buffer the acidification process (Bationo et al., 2000; Liu et al., 2010; Bado et al., 2012). In addition, the organic amendments contribute to a better nutrient availability, may improve water and

3.2 Materials and methods nutrient-holding capacity of the soil and, more generally, help maintain the soil’s ecosystem functions (Bationo et al., 2007; Rasool et al., 2010; Liu et al., 2010; Kihara et al., 2011). Whereas many studies have investigated the interaction between organic amendments and broadcast fertilization, to this date, the complementary effects between fertilizer microdosing and organic amendments have been minimally studied (Bielders and Gerard, 2015; Ibrahim et al., 2015a, b). The objective of the current study was therefore to evaluate the agronomic potential, nutrient efficiency and nutrient balance of different systems combining the application of fertilizer microdosing and organic manure for intensifying maize production in northern Benin.

3.2 Materials and methods 3.2.1 Description of study zone

The study took place in the northern region of the Republic of Benin (West Africa), the main production zone of food and cash crops. The predominant agricultural production systems are cotton and food crops such as maize, yam, sorghum, and millet. Maize is both a food and cash crop and is of strategic importance for smallholder’s farmers in Benin. As from 1961, the national maize production in Benin gradually increased from 219,593 to 1,354,344 tons (FAOSTAT, 2016). During the same period, the land allocated to maize production increased from 375,650 to 968,030 ha and yields grew from 584 and 1422 kg ha-1 (FAOSTAT, 2016). When used, manure is applied to cropland by farmers in either of two ways: (1) manure (dung plus bedding) is gathered from stalls, transported and hand spread onto the fields, and (2) cattle and/or small ruminants are corralled at night directly in the fields during the dry season. Animals are confined for a number of nights to a small part of the field by tying them to poles or by building enclosures of branches, after which they are moved to a different location in the field. Corralling is the preferred practice because it does not require carts or human labor to transport manure to the fields. In addition, the animals trample the manure during corralling, which ensures a partial mixing in the topsoil.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

3.2.2 Experimental site

The experiment was conducted at the Agricultural Research Station of Northern Benin (CRA- Nord) located at Ina village (Ina district, municipality of Bembèrèkè) (9°57’N and 2°42E, 365 m a.s.l), 70 km north-east of Parakou. Ina is located in the agro-ecological region III of Benin where annual rainfall ranges between 900 and 1200 mm. The average annual rainfall at Ina is 1148 ± 184 mm (mean ± SD) and the average daily temperature is 27.5°C (CRA-Nord Climate Database, 1982–2015). The climate is characterized by a single rainy season that occurs between May and October. The soil is classified as a ferruginous tropical soil in the French soil classification system with low inherent fertility, which corresponds to Acrisols or Lixisols according to the World Reference Base (Youssouf and Lawani, 2002).

3.2.3 Experimental design

The experiment was conducted during the rainy season in 2014 and 2015. In this study, the manure was applied either as transported manure or through corralling. Because of the limited availability of manure to farmers, the transported manure was applied at a rate 2 to 3 times lower than the rate recommended by extension services. To compensate for these lower application rates, and based on recent evidence (Ibrahim et al., 2015a; Tovihoudji et al., 2017a), the transported manure was hill-placed so as to increase its efficiency. Five manure treatments were considered: (i) no manure (NM); (ii) transported manure applied each year at a rate of 3 t ha-1 (TM); (iii) manure applied through corralling in the same year (Cor-0) or one (Cor-1) or two years (Cor-2) before. For practical reasons, each manure treatment was implemented as a single stratum, i.e., manured fields were not replicated. This prohibits a formal statistical comparison of the manure treatments but is not considered a problem since the emphasis of the study lies in the evaluation of fertilizer microdosing. Within each manure stratum, the experimental layout consisted in a randomized complete block design with three replications. The manure strata were separated by an alley of 3 m. For both the NM and TM treatments, different plots were used in 2014 and 2015 to avoid confounding from residual effects. For the TM treatment, small planting hills of 0.1-m diameter and 0.1-m depth were dug on both sides of each planting hole at an approximate distance of 7-10 cm of each planting hole. 3 t ha−1 of manure on oven-dry basis (corresponding to 96 g hill−1) were applied in equal parts on both sides at plant emergence 10 days after sowing (DAS). The rate of manure used was selected because it closely matches farmer’s practice.

3.2 Materials and methods

The TM was applied at the same time as the fertilizer and the holes were closed after application. The corralling practice was conducted by confining livestock at night in an enclosure to deposit the dungs and urine in the experimental plots directly (Figure B.2, Appendix B). Corralling was performed only once on a given plot. The corralling was done at the beginning of each season (May 10-20) by confining 38 to 44 cattle (approximately 35-38 tropical livestock units, TLU1) in 490 m² enclosures for 10 nights (6-7 pm to 8-9 am). The animals were attached so as to cover the surface as homogeneously as possible. The quantity of faeces voided per TLU per day was estimated at 1.1±0.1 kg DM by weighing the quantity of feces excreted by ten animals randomly selected during the corralling period (ten days). Hence, the total quantity of manure applied was on average 8.2 ± 0.7 t DM ha-1, equivalent to 1771±325 kg C, 144±29 kg N, 29±4 kg P and 21±13 kg K ha- 1 based on manure C and nutrient content (Table 3.1). The nutrients applied by urine were not quantified in this study. In order to have a Cor-1 and Cor-2 treatment as from 2014, part of the experimental site had been corralled in 2012, and another part in 2013. The entire experimental site had been cropped with maize in 2012-2013, but no fertilizer treatments were applied during those two years. Since there is no available data for nutrient uptake in 2012 and 2013, we approximated the three-year corralling cycle (Cor-0 + Cor-1 + Cor-2) by using the data from the Cor-0, Cor-1, and Cor-2 plots in 2014 or 2015. Farmyard and corralling manures were collected respectively from the barn of the Ina Agricultural Research Center and from the corralling plots each year during the same period (May) and air-dried before the beginning of the experiment. Each year, each sample was a composite of ten sub-samples. The chemical characteristics of these manures are presented in Table 3.1.

Table 3.1. Quality of the organic amendments. C N P K (%) 2014 Transported manure (TM) 13.2 0.9 0.6 1.2 Corralling manure (Cor-0) 24.4 2.1 0.3 0.1 2015 Transported manure (TM) 16.3 1.2 0.4 0.6 Corralling manure (Cor-0) 18.8 1.4 0.4 0.4

1One TLU = an animal ruminant of 250 kg liveweight

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

Four mineral fertilizer options were tested within each manure stratum: i) a control (no fertilizer); fertilizer microdosing at a rate of ii) 2 g of composite NPK (15-15-15) fertilizer per hill at 10 DAS + 1 g urea (46% N) per hill at 45 DAS (MD1) and iii) 4 g of NPK (15-15-15) fertilizer per hill at 10 DAS + 1 g urea per hill at 45 DAS (MD2); and iv) broadcast fertilizer at recommended rate (RR) of 200 kg NPK (15-15-15) ha-1 at 10 DAS + 100 kg urea ha-1 at 45 DAS. This is equivalent to 23.8 kg N ha-1, 4.1 kg P ha-1 and 7.8 kg K ha-1 for MD1, 33.1 kg N ha-1, 8.2 kg P ha-1 and 15.6 kg K ha-1 for MD2, and 76 kg N ha-1, 13.1 kg P ha-1, and 24.9 kg K ha-1 for RR. For fertilizer microdosing (both NPK and urea), small planting hills (0.05- m diameter and 0.05-m depth) were dug on one side of each planting hole and closed after application. The NPK and urea fertilizer in the RR treatment were spot-broadcasted at an approximate distance of 10 cm of each planting hole in accordance with farmer’s practice in the study area. In the RR treatment, NPK was not incorporated. For both the microdosing and RR treatments, urea application was immediately followed by weeding-ridging as done by farmers to limit volatilization losses. Fertilizers were added in the same plot each year in the corralling practice strata. At the onset of the experiment, land preparation was done uniformly across all plots by tractor disk-plowing to a depth of 0.2 m. Individual sub-plots measuring 4 x 5 m were delineated in the main experimental plot (manure stratum). Three to four seeds of the improved and early maturing maize variety DMR-ESR-W (Downy Mildew Resistant, Early-Streak Resistant -White) were sown on 4 July 2014 and 20 July 2015 after a major rainfall event greater than 20 mm. The planting hills were spaced 0.8 m x 0.4 m. Maize seedlings were thinned to 2 plants hill-1 two weeks after planting, giving a density of 62,500 plants ha-1 (currently recommended plant density). Plots were weeded twice (15 and 30 DAS) and ridged 45 DAS immediately after urea application with a hand hoe.

3.2.4 Measurements and calculations

Soil sampling and analysis. In 2014, before manure and fertilizer application, the experimental site was divided into three equal parts. At the center of each part, a soil profile was dug and sampled at depths 0-0.2, 0.2-0.4, 0.4-0.6, 0.6-0.8 and 0.8-1 m. For each profile and depth, 9 subsamples were randomly taken and mixed to form a composite sample. For the topsoil (0-0.2 and 0.2-0.4 m), the composite sample from each profile was then mixed with 9-10 samples taken at randomly selected points in each part of the field to form a new composite sample more representative of each part of the experimental site. In addition, prior to sowing, three composite soil samples were taken with a soil auger from

3.2 Materials and methods the top 0.2 m from the corralled block (only in Cor-0 plots). No samples were taken in the TM stratum after manure application because the manure hill- placement causes very high spatial heterogeneity over short distances. Particle size distribution was determined using the pipette method (Gee and Or, 2002). pH (H2O) was measured potentiometrically in a 1:2.5 soil:distilled water suspension (van Reeuwijk, 1993). Organic carbon was determined by the method described by Walkley and Black (1934), total N by the Kjeldahl method (Houba et al., 1995) and available phosphorus by the Bray-1 method (van Reeuwijk, 1993). Exchangeable K was determined after extraction by ammonium acetate

(NH4OAc) at pH 7 using the extraction method described by van Reeuwijk (1993).

Maize yields, yield components and nutrient uptake. Three central rows of each plot (8.6 m²) were harvested on 24 October 2014 and 31 October 2015 for yield determination. The cobs and stover were weighed in the field using a digital scale. The grain moisture content was then determined for each replication after oven-drying at 65 °C to a constant mass. Maize stover subsamples were taken to the laboratory for further drying (oven-drying at 65 °C to a constant mass) and moisture correction. The number of plants and the number of maize cobs per plant were counted. The dry weight of grains per cob was calculated by dividing the total dry grain weight per replicate by the number of cobs. The weight of 1000 grains was estimated by counting and weighing 200 randomly sampled grains from 3 replicates per plot. The number of grains per cob was calculated by dividing the weight of grains per cob by the average weight of one grain. The harvest index (HI) was calculated by dividing the dry weight of grains by the total dry biomass at harvest. Maize grain and stover yields were then calculated and expressed in kg ha-1 on a dry weight basis. Maize residues were removed from the fields after harvest as commonly practiced by farmers who use the residues as animal feed. In order to determine maize nutrient uptake, three whole plants were randomly sampled from two inner rows in each replication at harvest. The samples were separated into stover (stem, leaves, inflorescences, spars and cores) and grains, and oven-dried at 65 °C for 48 h. Sub-samples of the dried plant material were milled for total N, P and K analysis. Samples were digested with sulphuric acid (H2SO4) + salicylic acid + hydrogen peroxide (H2O2) + selenium. The quantitative determination for total N was carried out using an Auto-analyser (Pulse Instrumentation Ltd, Saskatoon, Saskatchewan, Canada) using a colorimetric method based on the Bertholet reaction (Houba et al., 1995). Total P was determined by the colorimetric method based on the phosphomolybdate complex, reduced with ascorbic acid (Houba et al., 1995), and total K was

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin determined by flame emission spectrophotometry. All chemical analyses (manure, soils and plants) were carried out at the ICRISAT laboratory (Sadoré, Niger).

Plant available water. Daily rainfall data was recorded with a rain gauge located in the experimental field. Access tubes were installed in the middle of each plot to monitor soil water content with a portable soil moisture meter (TRIME-PICO IPH/T3, IMKO Micromodultechnik GmbH) at approximately 14-day intervals in 2014 and 2015. The soil moisture meter had been calibrated in-situ using the gravimetric method. Measurements were taken every 0.1 m from 0 to 0.6 m depth in all the plots. The total stock of water between 0 and 0.4m depth (maximum root concentration zone) was calculated as described in Fatondji (2002). Moisture content was interpolated linearly between measurement depths. Readily available soil water for plants (PAW; mm) was calculated by subtracting the stock of water at the permanent wilting point (pF4.2) from the total soil water stock whenever the volumetric water content was greater than the water content at the wilting point.

Fertilizer use efficiency. Fertilizer use efficiency (FUE) was used as a proxy for agronomic efficiency and calculated for each year within each manure stratum based on Vanlauwe et al. (2010) (Eq. 3.1): FUE (kg grain kg-1 fertilizer) = (Yt -Yc) / Ft (Eq. 3.1) where Yt and Yc are the grain yields (kg ha-1) of the fertilized and control treatments in each manure stratum, respectively; and Ft is the total mass of fertilizer applied as NPK and urea fertilizer in treatment t per unit area (kg ha-1).

Partial nutrient balance. The partial nutrient balance metric was calculated by subtracting the quantity of nutrients (N, P or K) removed in the harvested products, grain (OUT1) and crop residue (OUT2) [Eq. 3.2] from the total quantities of nutrients applied through mineral fertilizer (IN1) and manure (IN2) [Eq. 3.3] for each plot as follows: OUT (kg nutrient ha-1) = OUT1 [Grain yield (kg ha-1) x nutrient content of grains (kg kg-1)] + OUT2 [Stover yield (kg ha-1) x nutrient content (kg kg-1)] (Eq. 3.2)

IN (kg nutrient ha-1) = IN1 [Nutrient applied by fertilizer (kg ha-1)] + IN2 [Manure amount (kg ha-1) x nutrient content of manure (kg kg-1)] (Eq. 3.3)

The balance was calculated on a yearly basis for the NM and TM plots. For the corralling practices, it was done on the basis of a three-year cycle to take into account the main effect (Cor-0) as well as the residual effects during the

3.2 Materials and methods subsequent two years (Cor-1, Cor-2). Since the residual effects are inherent to the corralling practice – corralling is never performed by farmers on a yearly basis at the same location – calculating a yearly balance would be meaningless. The balance for corralling thus considers the nutrient input from a single application of manure through corralling as well as the nutrient uptake during three years (Cor-0, Cor-1, Cor-2). Nutrients applied through urine could not be quantified and are excluded from the partial balance calculations.

Full balance. Generally, full nutrient balances at plot level consider five inputs (IN 1 to 5) and five outputs (OUT 1 to 5) (Stoorvogel and Smaling, 1990; Smaling et al., 1993; Table 3.2). The terms IN1 (mineral fertilizer), IN2 (manure except for urine), OUT1 and OUT2 have been accounted for in the partial balance calculation. The other terms were estimated using the transfer functions (developed and commonly used for full balance calculation in SSA) and secondary data from the literature because of the difficulty to measured them directly in this study.

Table 3.2. Calculated or estimated nutrient flows used for the full balance calculation. Code Flows Inputs IN1 Inorganic fertilizer IN2 Organic maendments IN3 Atmospheric depositions IN4 Biological nitrogen fixation IN5 Sedimentation Outputs OUT1 Harvested product (grain) OUT2 Crop residue OUT3 Leaching OUT4 Gaseous losses OUT5 Erosion

1. Inputs not directly measured in this study

a) Urine (IN2) This input applies only to the corralling treatment and is applied only once in the three-year corralling cycle. The amount of nutrients excreted in urine varies widely and ranges from 9.8 to 270 g for N (Schlecht et al., 1998), 0.01 to 0.5 g for P, and 0.03 to 125 g for K per animal per day (Sath et al., 2012; Nwite, 2015), equivalent to an average of 82, 0.2 and 36.5 g per animal per night (~14h), respectively for N, P and K. The total quantities of nutrients applied by urine can be estimated at 656, 2 and 292 kg ha-1 respectively for N, P and K.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

b) Atmospheric depositions (IN3) The input of nutrients by atmospheric depositions consists of two components: wet deposition associated with rainfall, and dry deposition related to Harmattan dust. These inputs are the same for all treatments. The wet depositions are calculated generally using the formulae developed by de Ridder et al. (1982) for N and P and by Roy and Misra (2003) for K as follows: Ndep = 0.0065 * Annual precipitation (mm) (Eq. 3.4) Pdep = 0.0007 * Annual precipitation (mm) (Eq. 3.5)

K2Odep = 0.011 *√퐴푛푛푢푎푙 푝푟푒푐푖푝푖푡푎푡푖표푛 (푚푚) (Eq. 3.6) where Ndep, Pdep and K2Odep (kg ha−1 yr−1) (to be multiplied by 0.83 to convert

K2O to K) are nitrogen, phosphorus and potassium from rainfall depositions. Estimated values were 7.8, 0.84 and 3.2 kg ha-1 yr-1, respectively, for N, P and K. The amount of nutrients deposited annually with the dust load in sub-humid zones of Benin is of the order of 1.1, 0.3 and 3.3 kg ha-1 yr-1 for N, P and K, respectively (Ramsperger et al., 1998; Herrmann et al., 2010). c) Biological nitrogen fixation (IN4) Nitrogen fixed from the atmosphere consists of symbiotic N fixation and non- symbiotic N fixation. Given the absence of legumes in the present study (maize mono-cropping), only non-symbiotic N fixation needs to be considered. Non- symbiotic N fixation is assumed to be the same for all treatments. It was estimated using the equation developed by Roy and Misra (2003) as follows:

N fixed = 0.5 + 0.1 *√퐴푛푛푢푎푙 푝푟푒푐푖푝푖푡푎푡푖표푛 (푚푚) (Eq. 3.7) Based on this equation, non-symbiotic N fixation results in a net input of 4.0 kg N ha-1 yr-1. d) Sedimentation (IN5) The input of nutrients by sedimentation, which refer to the nutrient inputs in irrigation water or in runon and input in sediment as a result of water erosion, are negligible considering that the experiment was conducted without an irrigation system and that the experimental site was not subject to runon and sediment inputs (flat terrain, rather sandy soil).

3.2 Materials and methods

2. Outputs directly and not directly measured in this study a) Leaching (OUT3)

Leaching losses apply to N and K. Phosphorus generally has low mobility in soils, especially in tropical soils as a result of P fixation on iron/aluminum (oxy-) hydroxides. In the absence of measurements, the quantities of N and K lost annually through leaching can be estimated from the transfer functions established by De Willigen (2000) for N and by Smaling et al. (1993) for K as follows:

OUT3 N = 21.37 + (P/C * L) * (0.0037*Nf + 0.0000601*OC – 0.00362* Nu) (Eq. 3.8) OUT3 K = (Ke + Kf) * (0.00029 * P + 0.41) (Eq. 3.9) where P is annual precipitation (mm yr−1), C is the clay content (%) of the topsoil, L is rooting depth (m), Nf is N applied through mineral fertilizer and/or organic fertilizer (kg ha-1), OC is organic carbon content (%) of the topsoil and Nu is N uptake by the crop (kg ha-1 yr-1), Ke is the exchangeable K (cmolc kg-1) in the topsoil and Kf is the amount of K derived from applied amendments. Regarding leaching losses from urine during corralling, see OUT4 (Eq. 3.10). b) Gaseous losses (OUT 4) Gaseous N losses consists of the losses through denitrification and a direct loss through volatilization of ammonia (NH3). Since the microdose fertilizer (NPK and urea), manure and urea in RR treatment were incorporated into the soil immediately after application, the gaseous losses are probably low. Nevertheless, we estimated the gaseous losses based on the amount of urea applied since it is the most affected by volatilization. The regression model developed by Roy and Misra (2003) was used to estimate this component as follows:

OUT4 = (0.025 + 0.000855 * P + 0.01725 * F + 0.0117 * OC) + 0.113 * F (Eq. 3.10) where P is the rainfall (mm), F is the amount of nitrogen in urea fertilizers (kg N ha-1) and OC is the organic carbon content (%).

Regarding the nutrients applied through urine, it has been reported that approx. 50% of N and K contained in urine are lost through leaching and/or volatilization (Brouwer and Powell, 1998). In the results section (Table 3.5), these combined losses through leaching and volatilization are reported as OUT 3+4_urine.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

c) Erosion (OUT5) Since the experiment was conducted in relatively flat terrain and on rather sandy soil, soil nutrient losses through water erosion and run-off at the plot level were considered negligible. This is consistent with the hypothesis made regarding IN5 (no input from runon and sedimentation).

The full nutrient balances and the partial balance including the contribution of urine (partial balance with urine) were quantified as follows: i) Partial balance with urine = (IN1 + IN2_manure + IN2_urine*) – (OUT1 + OUT2 + OUT3+4_urine) (Eq. 3.11) ii) Full balance = (IN1 + IN2_manure and urine* + IN3 + IN4 + IN5) – (OUT1 + OUT2 + OUT3+ OUT4 + OUT3+4_urine + OUT5) (Eq. 3.12) (*) when applicable

3.2.5 Statistical analysis

Prior to the analysis, data were carefully checked for normal distribution within each manure stratum using the Anderson-Darling test, and homogeneity of variance was assessed using Levene’s test. Nutrient uptake and FUE data were log transformed before analysis of variance because of non-normality. Fertilization and year effects on all variables were examined using analysis of variance (ANOVA) with the Genstat v.12 statistical package (GenStat, 2009). Because of the experimental design, manure x fertilizer interactions could not be evaluated. Based on the ANOVA, there was a strong year effect on the treatment responses (p < 0.001), hence the yields and FUE data were analyzed and discussed individually per manure stratum and per year. Separation of means was done using the honestly significant difference (HSD)/Tukey’s test at an error probability P < 0.05.

3.3 Results

3.3 Results 3.3.1 Soil properties of the experimental field

The texture of the soil was loamy-sand in the topsoil (0-0.4 m) and sandy-clay- loam in the lower layers (0.4-1 m) (Table 3.3). The soil was acidic at all depths. Bulk density was high throughout the profile. Organic carbon and total N levels were low and decreased with depth. The P-Bray content was fairly high in the top 0.4 m of the soil and decreased with depth. Following the application of manure through corralling, the pH was raised to 6.3 and SOC content in the top 0.2 m was doubled (Table 3.3). Corralling also caused an almost 5-fold increase in available P and a 2-fold increase in exchangeable K.

Table 3.3. Soil properties of the experimental fields on unamended plots (initial) as well as after corralling (mean ± SE; n =3). Initial After corraling

Depth (m) 0-0.2 0.2-0.4 0.4-0.6 0.6-0.8 0.8-1 0-0.2

Sand (%) 82.3±0.5 78.9±0.8 59.5±1.0 53.1±0.9 54.0±0.6 77.5±0.4 Silt (%) 13.8±0.3 13.1±0.2 15.0±0.1 15.2±0.1 15.5±0.1 17.2±0.2

Clay (%) 3.9±0.1 8.1±0.5 25.5±0.8 31.7±0.7 30.5±0.5 5.3±0.5

Bulk density (g cm-3) 1.62±0.03 1.73±0.03 1.70±0.01 1.65±0.04 1.71±0.02 nd

pH-H2O 5.7±0.2 5.7±0.3 5.4±0.1 5.1±0.2 5.1±0.2 6.3±0.3

Organic C (g kg-1) 4.5±1.0 1.9±0.1 1.8±0.3 1.3±0.3 1.2±0.2 10.7±0.8

Total N (mg kg-1) 417.0±67.3 194.0±10.2 191.7±6.7 153.0±3.8 148.7±9.2 1114.0±67.1

P-Bray 1(mg kg-1) 11.2±3.3 24.9±7.6 7.8±0.3 3.2±0.4 2.8±0.2 49.9±3.2

Exch. K (cmol+ kg-1) 0.24±0.02 nd nd nd nd 0.40±0.02 Exch. K: exchangeable K; nd: not determined

3.3.2 Rainfall characteristics and plant available water

Rainfall during the cropping period was 694 mm in 2014 (43 rainfall events) and 797 mm in 2015 (46 rainfall events) (Figure 3.1), whereas for the whole year it was 1142 mm in 2014 and 1085 mm in 2015. Rainfall was more evenly distributed in 2014 than in 2015 despite the greater number of short dry spells and the lower rainfall amount during the 2014 cropping cycle. Rainfall amount in 2015 was satisfactory but there were four heavy rainfall events ≥ 60 mm per day. Most of the rains occurred from 63 to 80 DAS, accounting for 38 % of the total rainfall recorded during the 2014 cropping period, while most of the rains in 2015 were concentrated from 15 to 40 DAS (the vegetative period), accounting for 55 % of the total rainfall recorded during the cropping period.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

There was no significant fertilizer effect on plant available water (PAW) in all measurement dates in both years (Figure 3.2). The maximum PAW in the top 40 cm (54 mm on average) was reached between 62 and 76 DAS in 2014, corresponding to the optimum maize silking period, whereas it was 35 mm at the flowering period. In 2015, the PAWs were always higher than in 2014 and ranged between 21 mm (at 14 DAS) to 73 mm (at 76 DAS) with small variation from 27 to 90 DAS. Since, PAWs did not reach critically low levels during the identified dry spells periods (Figure 3.2), no severe drought stress is expected in both years.

Figure 3.1. Rainfall distribution from sowing to harvest in 2014 and 2015, with an indication of the main crop development stages.

3.3 Results

PAW/Cumulative rainfall(mm)

Figure1 3.2 . Cumulative rainfall in-between soil moisture measurement dates, and plant available water (PAW) from sowing to physiological maturity in 2014 and 2015. Error bars are standard errors of differences between PAW means within each measurement date.

3.3.3 Maize grain and stover yields

Maize yields were significantly different across growing seasons, with mean yields larger in 2015 compared to 2014 (Figure 3.3a, b; p < 0.01). Manure application significantly improved maize yields compared to the no manure plots in both years (Figure 3.3a, b). In 2014, average grain yields increased in the following order: NM < TM < Cor-2 < Cor-0 < Cor-1, while the order was NM < TM < Cor-0 < Cor-2 < Cor-1 in 2015 (Figure 3.3a). In general, stover yields followed the same trend as grain yields in both years with a slight difference in 2014: NM < TM < Cor-2 < Cor-1 < Cor-0 in 2014 (Figure 3.3b). Across manure treatments, fertilization significantly increased maize yields compared to the unfertilized control in both years (Figure 3.3a, b; p < 0.01). In

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

2014, average maize grain yields ranged from 1124 to 2564 kg ha-1 for the unfertilized control and from 2311 to 3852 kg ha-1 for the microdose treatments (MD1 and MD2) depending on the manure strata (Figure 3.3a). In 2015, maize grain yields for the unfertilized control ranged from 1073 to 2329 kg ha-1, while those for the microdose treatments ranged from 2870 to 4050 kg ha-1 (Figure 3.3a). Stover yields followed generally the same trend with increases of 1338 and 2177 kg ha-1 in 2014, and 1841 and 2449 kg ha-1 in 2015, respectively for MD1 and MD2 across manure treatments (Figure 3.3b). Despite the higher fertilizer application rates in the RR treatment, yields in RR were similar to MD2 and greater than MD1 within most manure strata and for both years (Figure 3.3a, b).

Figure 3.3. Maize grain (a) and stover (b) yields as influenced by fertilizer application method across five manuring practices in 2014 and 2015. NM and TM refer to no manure and transported farmyard manure; and Cor-0, Cor-1 and Cor-2 refer to corralling in the same year, or 1 and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR mean fertilizer microdosing option 1 and 2 and the recommended fertilizer rate, respectively. Error bars are standard errors of differences between means within each manure stratum. Within each manure stratum bars with the same letter are not significantly different at p < 0.05.

3.3 Results

3.3.4 Yield components

For all years and manure strata, mineral fertilization generally had a significant effect (p < 0.05) on the number of grains per cob and the 1000-grain weight, with lowest values in the unfertilized control (Table 3.4D-E). Compared to the unfertilized treatment, and despite a high CV (21-29%), fertilizer application significantly increased the number of grains per cob within all manure strata, except in Cor-0 (in 2014) and in Cor-1 and Cor-2 (in 2015). On average across the manure strata and the three fertilization treatments, fertilization increased the number of grains per cob by 49% in 2014 and by 56% in 2015, compared to the unfertilized controls. Similarly, fertilizer application increased the 1000-grain weight by 9% in 2014 and by 17% in 2015, compared to the unfertilized controls. The RR treatment was found to have a similar number of grains per cob and 1000-grain weight as MD1 and MD2 for all manure strata except for Cor-1 and Cor-2 in 2014 for which the 1000-grain weight significantly increased with increasing fertilizer rates. Very few significant differences were found between treatments regarding the harvest index, the plant density at harvest and the number of cobs per plant (Table 3.4A-C). Whenever significant differences were observed, the differences between fertilized treatments were quantitatively low and displayed no clear trend. Unfertilized plots generally had significantly higher numbers of empty cobs per plant than fertilized treatments (data not shown).

3.3.5 Fertilizer and rainfall use efficiency

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

Table 3.4. Maize harvest index and yield components as influenced by mineral fertilization across five manuring practices in 2014 and 2015. 2014 2015 Contrl MD1 MD2 RR SED Contrl MD1 MD2 RR SED A) Harvest index (-)

NM 0.28a 0.39b 0.34ab 0.37b 0.02** 0.36 0.41 0.43 0.41 0.02 TM 0.33a 0.39ab 0.36ab 0.40b 0.01** 0.37 0.42 0.43 0.40 0.02 Cor-0 0.39 0.39 0.38 0.37 0.01 0.36 0.36 0.37 0.39 0.02 Cor-1 0.35 0.35 0.35 0.37 0.01 0.38 0.38 0.38 0.36 0.02 Cor-2 0.36 0.38 0.37 0.38 0.02 0.39 0.39 0.38 0.37 0.02 CV (%) 7.5 5.9 B) Stand count at harvest (%)

NM 76 85 86 84 4.10 73 85 81 89 5.50 TM 87 87 81 90 9.80 89a 86a 72b 83a 3.80** Cor-0 74a 81ab 89b 78ab 3.90* 87 74 79 86 4.60 Cor-1 72 86 85 88 7.10 72 83 87 72 4.90 Cor-2 82 94 84 78 5.70 87ab 74a 80ab 95b 5.50** CV (%) 6.7 8.5 C) Number of full cobs plant-1 NM 1.0 1.1 1.1 1.0 0.10 1.2 1.0 1.1 1.1 0.10 TM 1.1 1.1 1.2 1.0 0.14 1.0 1.0 1.2 1.1 0.11 Cor-0 1.0 1.0 1.0 1.2 0.10 1.0a 1.2b 1.1ab 1.0a 0.05** Cor-1 1.1 1.0 1.0 1.0 0.07 1.2 1.2 1.0 1.3 0.17 Cor-2 1.3 1.1 1.1 1.2 0.16 1.0 1.2 1.2 1.0 0.19 CV (%) 8.3 9.0 D) Number of grains cob-1 NM 78a 137b 137b 149b 15.10* 91a 186b 211b 169b 17.10** TM 121a 150ab 169ab 195b 18.10* 163a 231ab 257b 228ab 20.50* Cor-0 219 250 240 242 18.70 170a 194ab 226ab 239b 16.50* Cor-1 152a 154ab 207bc 213c 15.60* 176 211 245 244 39.00 Cor-2 83a 136b 164b 161b 12.80** 135 208 206 246 37.50 CV (%) 29.3 20.9 E) Thousand grain weight (g) NM 293a 300a 301a 320b 4.28** 227a 293b 290b 303b 12.84** TM 273a 315b 311b 315b 2.83*** 257 270 277 283 14.08 Cor-0 286 314 317 308 9.98 260a 277ab 303b 310b 11.94* Cor-1 299a 340b 315c 327d 2.04*** 250a 287b 293b 293b 7.70** Cor-2 287a 297b 311c 323d 2.28*** 260 287 307 300 16.72 CV (%) 5.1 7.7 NM and TM refer to no manure and transported farmyard manure; and Cor-0, Cor-1 and Cor-2 refer to corralling in the same year, or 1 and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR mean fertilizer microdosing option 1 and 2 and the recommended fertilizer rate, respectively; SED= standard errors of differences between means. CV= coefficient of variation. *, **, *** significant at 0.05, 0.01, 0.001 respectively.

3.3 Results

3.3.5 Fertilizer and rainfall use efficiency

The FUE values and statistics reflect the differences in grain yields between the treatments (Figure 3.4). As a result of the higher grain yields in 2015 than in 2014, average FUE was higher in 2015 (11.0 kg grain kg−1 fertilizer) than in 2014 (7.8 kg grain kg-1 fertilizer). In 2014, FUEs were highest in the NM stratum followed by Cor-0 and Cor-2, while in 2015, FUEs were highest for NM followed by Cor- 2 and Cor-1 (Figure 3.4). Generally, the higher the rate of manure (i.e. Cor-0 > TM > NM), and the more recent the manure application (i.e from Cor-0 to Cor- 2), the lower the fertilizer use efficiency. Across the manure strata, fertilization with MD1 and MD2 resulted in the highest FUEs across years, ranging from 8.0 to 19.0 and 7.6 to 14.0 kg grain kg-1 fertilizer, respectively, whereas the RR treatment had a FUE range of 4.6-8.0 kg grain kg-1 fertilizer. Since the same amount of rainfall reached all treatments in a given year, differences in rainfall use efficiency (RUE=grain yield/total seasonal rainfall) directly reflected differences in grain yields (data not shown). RUEs ranged between 1.6-3.4, 2.8-4.8 and 2.2-6.1 kg grain ha-1 mm-1 in 2014 and 1.3-4.1, 2.6- 4.7 and 2.2-5.4 kg grain ha-1 mm-1 in 2015, respectively for the NM, TM and corralling practices, on average across the fertilization treatments.

3.3.6 Nutrient inputs/uptakes and balances

Nutrient inputs and uptakes. On average over the three years of the corralling cycle (Cor-0, -1 and -2), mineral fertilizers supplied 33, 41 and 61% of all N inputs, 29, 45 and 57% of all P inputs and 53, 70 and 78% of all K inputs for MD1, MD2 and RR, respectively (Figure 3.5). For the TM practice, mineral fertilizers supplied 43, 51 and 70% of all N inputs, 21, 35 and 46% of all P inputs and 23, 37 and 48% of all K inputs for MD1, MD2 and RR, respectively (Figure 3.6). N, P and K uptakes by maize were significantly influenced by soil fertility management practices (p < 0.05) and generally exceeded application rates, leading to negative nutrient balances for most treatments (Figures 3.5 and 3.6). There was a significant year effect (p < 0.05; Table 3.5) for N and P uptake but not for K uptake. In general, N and P uptake tended to be higher in 2014 than in 2015 (Figure B.3., Appendix B). A significant fertilizer by year interaction was observed only for N in the NM stratum (Table 3.5). Uptake was higher in 2014 than in 2015 for MD1 and MD2, but it was lower or equal in 2014 compared to 2015 for the control and RR treatments.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

Figure 3.4. Fertilizer use efficiency (FUE) across five manuring practices in 2014 and 2015. NM and TM refer to no manure and transported farmyard manure; and Cor-0, Cor-1 and Cor-2 refer to corralling in the same year, or 1, and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR mean microdosing option 1 and 2 and the recommended fertilizer rate, respectively; Error bars are standard errors of differences between means within each manure stratum. Within each manure stratum bars with the same letter are not significantly different at p < 0.05.

3.3 Results

Table 3.5. Summary of the results of ANOVA for nutrient uptake (OUT) using log- transformed data over the two years of the trials (2014–2015). Nutrient Manure practices Year Fertilizer Year*Fertilizer

F Pr > F F Pr > F F Pr > F N NM 4.711 0.045 123.097 < 0.001 7.104 0.003 TM 26.272 <0.001 49.929 < 0.001 0.447 0.723 Corralling 33.996 <0.001 125.283 < 0.001 0.741 0.543 P NM 20.175 <0.001 17.714 < 0.001 0.461 0.713 TM 261.464 <0.001 24.633 < 0.001 1.415 0.275 Corralling 101.591 <0.001 28.025 < 0.001 0.723 0.553 K NM 0.368 0.553 16.501 < 0.001 1.849 0.179 TM 1.019 0.328 5.183 0.011 0.792 0.516 Corralling 4.057 0.061 11.574 <0.001 1.270 0.318 NM=no manure, TM=transported manure, Corralling = cumulative uptake over the three-year cycle (Cor-0 + Cor-1 + Cor-2).

Partial balances. On average across all fertilization treatments, nutrient balances were -24, -18 and -64 kg ha-1 year-1, respectively for N, P and K in the NM stratum (Figure 3.5a-c). Applying TM lead to less negative nutrient balances for N and P (-10 and -11 kg ha-1 year-1, respectively for N and P; Figure 3.5d-e), but not for K (-66 kg K ha-1 year-1; Figure 3.5f). In general, compared to the unfertilized treatment, applying fertilizers did not improve the partial nutrient balances. On the contrary, the partial nutrient balances were generally equally negative or even more negative in the fertilized treatments than in the unfertilized control, except for nitrogen in RR plots for which the balance was positive (+10 and +21 kg N ha-1 year-1 in NM and TM, respectively; Figure 3.5a, d). After summing the nutrient uptake over a three-year corralling cycle (i.e, in Cor-0, Cor-1 and Cor-2), nutrient uptakes ranged between 161 and 318 kg ha-1 for N, 79 and 143 kg ha-1 for P and 249 and 423 kg ha-1 for K, respectively (Figure 3.6a-c). N balances ranged between -17 and +54 kg ha-1 for the 3-year cycle depending on the fertilizer treatment. The N partial balances were still positive after the first two corralling cycles (Cor-0 and Cor-1) in all fertilizer treatments, whereas in the third cycle (Cor-2) N partial balances became negative, except for the RR treatment (Figure 3.6a). The combined application of fertilizer microdosing and manure through corralling generally reduced the extent of the negative N balance (-10 and -13 kg N ha-1 year-1 for MD1 and MD2 respectively; Figure 3.6a) compared to fertilizer microdosing alone (-44 and -34 kg N ha-1 year- 1 for MD1 and MD2 respectively; Figure 3.5a). The amount of P and K exported in the corralled strata was always higher than what was added by amendments, leading to negative partial balances ranging

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin between -50 and -77 kg ha-1 for P, and -228 and -345 kg ha-1 for K over the three- year cycle (Figure 3.6b, c). For P, the balances were near neutral after the first year, after which they became increasingly negative (Figure 3.6b). For K, balances were negative as from the first year of the corralling cycle (Figure 3.6c).

Figure 3.5. Annual N, P and K inputs (IN; transported manure in black and mineral fertilizer in gray) and uptakes (OUT; hatched pattern) in the no manure (NM) and transported manure (TM) strata as a function of the mineral fertilization treatment. Values are averages of 2014 and 2015. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars are standard errors of differences between means for uptakes. Note: Nutrient balances = IN – OUT.

3.3 Results

Full balances. The partial nutrient balances presented in Figures 3.5 and 3.6 consider that all crop residues were removed as commonly practiced by farmers and ignore nutrient inputs from urine in the corralling treatments. Also, the calculation excluded inputs from wet and dry depositions and sedimentation, and outputs by leaching, erosion, run-off, and gaseous losses. Thus, a full balance calculation was attempted. From the calculations of full balance, it appears that nutrient inputs such as atmospheric wet (7.8, 0.84 and 3.2 kg ha-1 year−1, respectively, for N, P and K) and dry deposition (1.1, 0.3 and 3.3 kg ha-1 year-1 N, P and K, respectively), and nutrient outputs by gaseous N losses (1 - 8 kg ha-1 year-1) are generally small and don’t vary greatly across treatments (Table 3.6). However, these terms may represent a substantial contribution to the total inputs for the control and microdose treatments in the absence of manure additions. The main sources of nutrient losses are the nutrients exported in the grain (OUT 1) and residues (OUT2) (Table 3.6). Leaching losses are more strongly dependent on the treatment (losses ranging from 0 to 53 and from 0 to 38 kg ha- 1 year−1 for N and K, respectively; Table 3.6) and can be a substantial source of loss in high input treatments, especially in the corralling system. Volatilization losses are generally low, except when urine is taken into account in the corralling practice. The average nutrient losses from urine is estimated at 328 kg N ha-1, 1 kg P ha-1 and 146 kg K ha-1 for the night corralling period (= 50% of input; Table 3.6). Regarding the NM and TM strata, both the partial and the full balances are negative and rather similar for all three nutrients in the control, MD1 and MD2 treatments (Table 3.6A, B). This is also true for P and K in the RR treatment. For N in the RR treatment, the partial balance was positive but the full balance became negative, mostly as a result of high leaching losses. Specifically regarding microdose fertilization, it can be seen that the conclusions drawn on the basis of the partial balance remain largely valid for the full balance. Full balances in MD treatments are systematically more negative than the control. This is most marked in the NM stratum. In the TM stratum, MD mostly enhanced the nutrient deficit for K and to a lesser extent for N as compared to the control, whereas for P differences are small between microdose and the control. Full balances in MD treatments are also more negative or similar to the full balances of the RR treatment (except sometimes for K) (Table 3.6A, B).

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

Figure 3.6. N (a), P (b) and K (c) inputs (IN; manure applied through corralling in black and mineral fertilizer in gray) and uptakes (OUT, dotted or hatched patterns) for a three-year corralling cycle as a function of the mineral fertilization treatment. Each compartment is the average of two years. Cor-0, Cor-1, and Cor-2 refer to corralling in the same year, or 1 and 2 years before. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars are standard errors of differences between means for uptakes. Note: Nutrient balances = IN – OUT.

3.3 Results

Table 3.6. Average partial and full balances calculated for the NM, TM and corralled treatments, for different mineral fertilization practices. Control MD1 MD2 RR Manure strata/components N P K N P K N P K N P K

A) No manure (NM) Inputs IN1 0 0 0 24 4 8 33 8 16 76 13 25 IN2 0 0 0 0 0 0 0 0 0 0 0 0 IN3 9 1 7 9 1 7 9 1 7 9 1 7 IN4 4 0 0 4 0 0 4 0 0 4 0 0 IN5 0 0 0 0 0 0 0 0 0 0 0 0 Total 13 1 7 37 5 15 46 9 23 89 14 32 Outputs OUT1 18 8 8 45 19 17 44 17 17 44 17 17 OUT2 10 5 39 23 9 56 23 11 88 22 9 63 OUT3 0 0 0 2 0 6 5 0 12 29 0 18 OUT4 0 0 0 3 0 0 3 0 0 7 0 0 OUT5 0 0 0 0 0 0 0 0 0 0 0 0 Total 27 13 47 73 28 79 75 28 117 102 26 98 Partial balance -27 -13 -47 -44 -24 -65 -34 -20 -89 10 -13 -55 Full balance -14 -12 -40 -37 -23 -65 -30 -19 -95 -13 -12 -67 B) Transported manure (TM) Inputs IN1 0 0 0 24 4 8 33 8 16 76 13 25 IN2 32 15 27 32 15 27 32 15 27 32 15 27 IN3 9 1 7 9 1 7 9 1 7 9 1 7 IN4 4 0 0 4 0 0 4 0 0 4 0 0 IN5 0 0 0 0 0 0 0 0 0 0 0 0 Total 45 16 34 69 20 42 78 24 50 121 29 59 Outputs OUT1 32 15 12 48 21 16 55 22 19 53 28 24 OUT2 19 9 60 28 9 85 30 12 98 34 11 106 OUT3 10 0 20 12 0 26 14 0 31 35 0 38 OUT4 1 0 0 3 0 0 3 0 0 7 0 0 OUT5 0 0 0 0 0 0 0 0 0 0 0 0 Total 62 24 92 91 30 127 102 34 148 129 39 168 Partial balance -19 -9 -45 -20 -11 -66 -20 -11 -74 21 -11 -78 Full balance -17 -8 -59 -22 -10 -86 -24 -9 -99 -8 -10 -110 C) Corralling practice (three-year cycle) Inputs IN1 0 0 0 72 12 24 99 24 48 228 39 75 IN2_manure 144 29 21 144 29 21 144 29 21 144 29 21 IN2_urine 656 2 292 656 2 292 656 2 292 656 2 292 IN3 27 3 20 27 3 20 27 3 20 27 3 20 IN4 12 0 0 12 0 0 12 0 0 12 0 0 IN5 0 0 0 0 0 0 0 0 0 0 0 0 Total 839 34 333 911 46 357 938 58 381 1067 73 408 Outputs OUT1 105 47 47 159 73 69 183 83 75 190 87 82 OUT2 56 32 202 86 43 277 100 47 339 128 56 341 OUT3 73 0 15 79 0 33 88 0 51 160 0 69 OUT4 3 0 0 12 0 0 12 0 0 24 0 0 OUT3+4_urine 328 0 146 328 0 146 328 0 146 328 0 146 OUT5 0 0 0 0 0 0 0 0 0 0 0 0 Total 565 79 410 664 116 525 711 130 611 830 143 638 Partial balance -17 -50 -228 -29 -75 -301 -40 -77 -345 54 -75 -327 Partial balance + urine 311 -48 -82 299 -73 -155 288 -75 -199 382 -73 -181 Full balance 274 -45 -77 247 -70 -168 227 -72 -230 237 -70 -230 Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Note: Nutrient balances = Total IN – Total OUT.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

For the corralling practice, Table 3.6C clearly shows that urine N inputs and volatilization are major contributors to the partial with urine and full nutrient balances. Including urine results in positive balances for N (balances ranging from +288 to +382 kg N ha-1 depending on the fertilizer treatment; Table 3.6), whereas otherwise the balances are negative. For P and K the balances are systematically negative after three years. Again, the balances are more negative for P and K in microdose than in the control treatment. Balances in MD and RR are rather comparable.

3.4 Discussion

The soil at the experimental site (Table 3.1) is representative of the ferruginous soils in northern Benin characterized by a loamy sand texture in the topsoil, acidic conditions and low levels of nutrients and organic matter (Youssouf and Lawani, 2002). Furthermore, seasonal rainfall (694 mm in 2014 and 797 mm in 2015 from sowing to harvest) is well within the 500 to 1000 mm range which ensures optimal development of maize plants under rainfed conditions in SSA (Folberth et al., 2013). In general, the higher yields recorded in 2015 compared to 2014 may have resulted from the larger amount of rainfall (Figure 3.1) and the smaller number of short dry spells in 2015 than in 2014, leading to higher levels of PAW in 2015 throughout much of the growing season (Figure 3.2). However, PAWs did not reach critically low levels in the top 0.4 m of the soil at any time during the two years, implying that no severe drought stress affected crop yields. Within the TM stratum, higher yileds in 2015 may also be attributed to the relatively higher nutrient content of the farmyard manure applied in 2015 (Table 3.1).

3.4.1 Maize response to manure application

The quantity of manure applied through corralling in this study is within the range of that reported in earlier studies in West Africa (1.5-15 t dry matter ha−1) (Powell et al., 1998; Brouwer and Powell, 1998; Schlecht et al., 2004). The manure applied through corralling significantly improved maize yields compared to the no manure plots (Figure 3.3). The results also highlight the substantial residual effect of organic manure applied through cattle corralling. On average over the two years, yields were 44, 45 and 33% higher in the Cor-0, Cor-1 and Cor-2 treatments, respectively, compared to the NM treatment. Significant residual effects over a period of 3 to 4 years were also reported by Gandah et al. (2003), Schlecht et al. (2004) and Bielders and Gérard (2015) albeit under the dryer, Sahelian conditions. Schlecht et al. (2004) further observed that the extent of the residual effect was linearly related to the amount of manure initially applied.

3.4 Discussion

The lasting residual effect observed in the present study can be attributed to the fairly high amount of manure (8.2±0.7 t DM ha-1) applied through corralling, which supplies major nutrients in excess of the requirements of maize for one to two years (Figure 3.6). In addition, consistent with previous studies in the Sahel (Powell et al., 1998; Suzuki et al., 2014), the combined application of manure and urine during corralling had a strong positive effect on the soil pH and other soil chemical properties (Table 3.2). In particular, the increase in pH resulting from the urine application favors phosphorus availability, which is a strongly limiting element in many tropical soils (Bationo et al., 2012; Nziguheba et al., 2016). The positive effect of corralling on maize yields can also be attributed to the supply of micronutrients such as Ca and Mg (Zingore et al., 2008) which are generally lacking in sufficient quantity in ferruginous soils (Youssouf and Lawani, 2002). Our results also show a significant increase in yields by 28% when small realistic quantities of transported manure (TM; 3000 kg ha-1) were hill-placed, compared to the NM treatment. This result is consistent with earlier reports by Otinga et al. (2013) and Ibrahim et al. (2015a) which indicated that hill-placement of small doses of manure can substantially increase maize and millet yields in smallholder cereal-based systems in sub-Sahara Africa. This positive effect of hill-placement of manure may result in a better uptake of nutrients by the roots due to the early root proliferation favored by this method (Ibrahim et al., 2015a). In addition, by concentrating the manure in pits around the planting hills, substantial soil improvement can be achieved in a shorter term (Tovihoudji et al., 2017a).

3.4.2 Maize response to fertilizer application

Fertilization significantly increased maize yields compared to the unfertilized control in both years. On average across both years and the 5 manure strata, fertilization increased grain yields by 64, 81 and 93% for MD1, MD2 and RR, respectively, compared to the unfertilized control (Figure 3.3a). Likewise, stover yields were increased by 47, 68 and 78% for MD1, MD2 and RR, respectively (Figure 3.3b). Yields in RR were not significantly different from yields in MD2. The present study thus demonstrates the potential of fertilizer microdosing to improve maize production in northern Benin and reinforces the earlier evidence from eastern and eastern Africa regarding its effectiveness at improving maize yields (Twomlow et al., 2010; Sime and Aune, 2014; Kisinyo et al., 2015). The harvest index was generally unaffected by the fertilizer treatments (Table 3.4), which implies that the different fertilizer management practices improved grain production to the same extent as it improved above-ground

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin biomass production. The increase in grain yield following fertilization was mainly a consequence of the larger number of grains per cob (+44% on average compared to the unfertilized control) and, to a lesser extent, an increase in 1000- grain weight (+13%; Table 3.4). Although the number of grains per cob and 1000-grain weight were slightly higher in 2015 than in 2014, the differences between the two years were small, indicating that climatic conditions were relatively similar during the various growth phases in both years. Better plant survival rates and increased numbers of cobs per plant contributed only 5% and 1% on average to the increased yields following fertilization. Since there was no indication that water stress may have differentially affected fertilizer treatments (Figure 3.2), the improvement in the number of grains per cob and in 1000-grain weight in fertilized treatments compared to the unfertilized treatments most likely resulted from the improved nutrition of the maize plants rather than water- related effects. This is supported by the general tendency for these two variables to increase along the following gradient: Control < MD1 < MD2 (Figure 3.7). Maize yields in RR plots were generally not significantly different from yields in MD2 plots (Figure 3.3), in spite of the fact that the fertilizer additions were about twice as high in RR than in MD2. On average across all manure strata, the number of grains per cob and the 1000-grain weight in RR were similar to MD2 (Figure 3.7). Since the PAW data did not reveal greater water stress in RR than in MD2, it may be concluded that the hill-placed fertilization resulted in a more efficient use of the applied nutrients. Jing et al. (2010) and Ibrahim et al. (2014) showed that fertilizer hill-placement promotes growth and proliferation of maize roots, leading to a better uptake of the small amount of applied nutrient by the roots. In the case of the broadcast fertilizer at recommended rate, nitrogen losses through leaching are likely to have been higher (Table 3.6). The high cumulative rainfall recorded after 45 DAS (the urea fertilizer application period) in 2014 and the high total seasonal rainfall and numerous heavy rainfall events which occurred after 10 DAS (the NPK fertilizer application period) (Figure 3.1) may have resulted in significant drainage.

3.4 Discussion

Figure 3.7. Evolution of the number of maize grains per cob and the 1000-grain weight as a function of the mineral fertilization treatment. Values are averages over all manure strata and both experimental years. Control refers to the absolute unfertilized treatment; MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars = standard deviation. 3.4.3 Efficiency of fertilizer microdosing across fertility levels

Maize response to fertilizer microdosing varied across manure strata. When comparing the microdose fertilizer treatments to the controls for the various manure strata, the greatest yield increases were observed for the NM (+1611 kg ha-1) and Cor-2 (+1468 kg ha-1) strata, followed by TM (+1258 kg ha-1), Cor-1 (+1183 kg ha-1) and Cor-0 (+1126 kg ha-1) (Figure 3.3). These results showed that maize response to fertilizer microdosing was best in the absence of organic amendments and tended to decrease with increasing fertility, from Cor-2 to NM. It is difficult to exactly position the TM treatment along this fertility scale but, after two years, the TM treatments still had received less manure compared to Cor-0. Hence, it is reasonable to expect the TM treatment to have a fertility level comparable to Cor-1 or Cor-2. The above ranking seems to confirm this. The lower maize yield response to fertilizer application in the recently corralled plots is consistent with earlier reports that yield response to fertilization decreases with increasing soil organic and nutrients contents. For millet in the Sahelian zone

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

(500 mm rainfall), Bielders and Gerard (2015) also observed that millet response to fertilizer microdosing was higher on average on lower productivity plots (recent corralled plots), where yields were doubled, whereas on high productivity plots no or even negative responses were observed. Recently, several studies in SSA have focused on trying to understand the conditions that govern the performance of integrated soil fertility management (ISFM) technologies. Vanlauwe et al. (2011) proposed that the response to mineral fertilization largely depends on the initial soil fertility level. These authors distinguished three soil categories based on their responsiveness to fertilizer additions: i) poor but responsive soils where crop yield responds well to fertilizers (e.g., Vanlauwe et al., 2011), ii) poor unresponsive soils, where crops respond poorly to nutrient additions (e.g., Vanlauwe et al., 2011), or “truly non-responsive fields” (e.g., Kurwakumire et al., 2014) and iii) rich and little-responsive soils, which are sufficiently fertile to supply most or all of the nutrients needed by a crop (e.g., Giller et al., 2011). In this study, it appears that the experimental site belonged to the first category of soils given the good response of NM plots to fertilizer additions. As more organic amendments were added, through transported manure or corralling, the response to fertilization decreased since much of the most limiting nutrients (N, P) were already supplied by the organic amendments. Hence, in terms of FUE (Figure 3.4), it is most efficient to apply fertilization on unmanured plots. It is worth noting, however, that this applies mostly to the MD1 treatment and to a lesser extent to MD2. For the higher fertilizer doses, FUE remained rather constant across manure strata (Figure 3.4).

3.4.4 Fertilizer microdosing may exacerbate nutrient mining

Overall, our results have shown that all fertilizer management strategies resulted in negative partial nutrient balances, with crop nutrient uptake exceeding additions through manure and/or fertilizers, except for N at the recommended rate across all manure strata (Figures 3.5 and 3.6). The nutrient balances of the current study are within the range of -56 to -13 kg N ha-1 reported by Ibrahim et al. (2014, 2016) for millet production under fertilizer microdosing, whereas for P and K, our values are much in excess of those obtained by these authors (-1 kg P ha-1 and -37 to -34 kg K ha-1). In the absence of manure, the nitrogen balance was similar for the control, MD1 and MD2, meaning that the N supplied by fertilizer microdosing only covered the N required for the yield increment but was insufficient to

3.4 Discussion compensate for the baseline deficit. For P and K, the negative balances were generally aggravated following fertilizer microdosing, compared to the unfertilized control. For the recommended fertilizer rate (RR), the negative balance was reduced (N) or stayed comparable (P, K) to the unfertilized control plots. Since the balances were more negative in microdose plots than in the control ones, it appears that microdosing fertilizer allowed plants to better exploit the soil reserves, most likely by boosting root development as reported by Ibrahim et al. (2014, 2015c). Hence, the long-term application of fertilizer microdosing without other sources of nutrients may enhance the rate of soil degradation through nutrient mining compared to unfertilized plots. For N and P, combining fertilizer microdosing with TM did not aggravate the negative balances compared to the unfertilized control, unlike what was observed for the NM strata (Figure 3.5). For K, microdosing in TM did result in larger nutrient mining compared to the unfertilized control. Hence the combined application of TM and fertilizer microdosing makes the latter more sustainable from a nutrient balance point of view, but particular attention should be paid to K. Combining corralling and fertilizer microdosing generally also reduced the negative N balances over a three-year cycle compared to no manure addition. For P, the balances were negative and of the same order as those observed on NM plots, whereas for K, the balances tended to be more negative in the corralled plots than in the NM plots after 3 years. As a matter of fact, the balances became negative as from the first year for K, the second year for P and the third year for N (Figure 3.6). Hence, fertilizer microdosing may also lead to soil degradation by nutrient mining in the long-term when combined with corralling. Considering a full balance rather than a partial balance (unsurprisingly) affects the outcome of the calculations, but the overall tendencies remain largely unchanged (Table 3.6). Full balances are systematically negative for the NM and TM strata for all fertilizer treatments, whereas a positive balance was reported for N in RR in the partial balance approach. As was observed for the partial balance, full balances calculated for microdose fertilization are more negative than in the unfertilized control except for N in the corralling stratum. This enhanced nutrient mining following microdose fertilization is most marked in the NM stratum. Addition of TM helps mitigate the issue but combining fertilizer microdosing with corralling does not alleviate nutrient mining, except when considering N inputs from urine in which case the N balances are systematically positive.

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

3.4.5 Opportunities for smallholder maize farming systems

Overall, our results demonstrate the potential of fertilizer microdosing for improving maize yields in northern Benin. Yield increases in MD2 and RR are comparable, yet FUE is much better for fertilizer microdosing (5.9 to 22.6 kg grain kg−1 fertilizer) than for the RR (4.6 to 8.7 kg grain kg−1 fertilizer) (Figure 3.3). Together with the higher financial investment required for the RR, this may explain why farmers are generally unwilling to apply fertilizers at the recommended rate. A fertilizer application technique that promotes a small realistic fertilizer amount and makes a more efficient use of it is, therefore, more suitable for smallholder farmers. In northern Benin in particular, cotton is often the sole entry point for mineral fertilizers into the farming system, not only for cotton but also for food crops. Indeed, farmers who agree to grow cotton receive an additional amount of fertilizer for food crops (Ripoche et al., 2015). In practice, the fertilizer amount (NPK + urea) applied by farmers to maize is generally around half the the broadcast fertilizer at recommended rate (100 ± 50 kg fertilizer ha-1), which is nearly equivalent to the microdose fertilizer rates tested in this study (94-156 kg fertilizer ha-1). On this basis, the fertilizer microdosing technique appears to be particularly interesting to maize growers who have access to only small quantities of fertilizer in the study area. It may be used to increase yield (this study) and improve farmer’s income (Sime and Aune, 2014). Although the response to fertilizer microdosing was best in no manure fields, fertilizer microdosing without manure application (or other sources of organic amendments) should not be recommended as a viable option for sustainable maize production in the long-term. Indeed, continuous cultivation without organic amendment has been shown to lead to an increase in soil acidification and an overall decline in soil organic matter and in the availability of other nutrients (Liu et al., 2010; Kintché at al., 2015; Adams et al., 2016). Organic additions are essential for maintaining soil quality in the long run and are an integral part of ISFM. Besides supplying micronutrients, organic amendments are also essential to sustain soil life (Vanlauwe et al., 2011; Opala et al., 2010; Kihara et al., 2011; Otinga et al., 2013). Currently, farmers export crop residues to be used as feed for livestock. It is essential to recycle some of this biomass and return it to the fields as manure, compost or through corralling. For those farmers who have access to little manure, hill placement of a realistic amount of

3.5 Conclusions

transported manure and microdosing fertilizer could be a viable option for sustainable maize production. However, the hill-placement of transported manure, which is similar to the “Zaï” technique albeit without the water harvesting component (Fatondji, 2002), demands additional labor for transporting and burying the manure. Hill-placement of manure and fertilizers may also complexify field management in the long run as a result of the spatial heterogeneity in nutrient distribution. Nevertheless, these constraints may be offset by shifting row positions and alternating application spots from year to year and through the higher income generated since maize is more and more grown as a cash crop in northern Benin to satisfy urban demand. For farmers who have the possibility to establish a stubble-grazing contract with Fulani herders, it may be of interest to combine corralling with fertilizer microdosing only as from the second or third year in order to achieve a higher fertilizer use efficiency. Recycling of crop residues and proper management of manure and urine to limit the nutrient losses under corralling could alleviate the potential soil mining effect of fertilizer microdosing technology. However, higher doses of fertilizers may be needed to ensure near neutral nutrient balances.

3.5 Conclusions

The results of the current study show that there is considerable potential for smallholder farmers of northern Benin to enhance maize productivity by means of fertilizer microdosing. With fertilizer microdosing, similar yields can be achieved as with the currently recommended fertilization rate, albeit with a much greater efficiency. The fertilizer application rates used in microdosing are also much closer to the rates currently applied by smallholder farmers in the study region through spot-broadcasting method, thereby making it more accessible to them. Despite observed benefits of fertilizer microdosing, based on partial nutrient balance calculations it appears that both microdose and recommended fertilization rates increased the risk of soil nutrient mining, especially regarding P and K, both in unmanured and manured plots. As an initial indication, for sandy loam soils such as those found in northern Benin, it may be most efficient and sustainable to apply microdose fertilization in combination with small amounts of manure or on plots that have not been corralled for one or two years. Recycling of crop residues and proper management of manure and urine which can limit the nutrient losses under corralling could, therefore, alleviate the potential soil mining effect of fertilizer microdosing technology. Since nutrient balance alone is not sufficient as an indicator of sustainability, it needs to be linked with soil nutrient stocks, either with the total stock or with the stock of available nutrients. Based on this, we suggest that the results are meant to alert 85

Chapter 3. Fertilizer microdosing in maize cropping systems in northern Benin

researchers, policy makers and other stakeholders, i.e., that nutrient mining under microdosing is a threat and needs more attention. Nevertheless, further long- term studies are required before precise advice can be given. There is also a need to evaluate these nutrient management options in farmer’s fields, considering climatic, soil and management conditions to better assess and understand the maize response and the magnitude of nutrient mining and to evaluate its economic benefits and risk.

Chapter 4

Variability in maize yield and profitability following hill-placement of mineral fertilizer and manure under smallholder farm conditions in northern Benin*

* This Chapter has been submitted to Field Crops Research as: Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Variability in maize yield and profitability following hill-placement of mineral fertilizer and manure under smallholder farm conditions in northern Benin.

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

Abstract

Whereas the decision to promote a given agricultural intensification technology has hitherto been largely based on its average agronomic or economic performance, it is increasingly being recognized that the variability in the performance must also be taken into account in order to develop more meaningful and flexible recommendations. This is true in particular for microdose fertilization which is being actively promoted in sub- Saharan Africa as a means to increase crop productivity, profitability and fertilizer use efficiency. To this end, a total of 51 on-farm maize trials were carried out in northern Benin in 2014 and 2015. The performance of two microdose fertilization options (MD1 = 23.8 kg N, 4.1 kg P, and 7.8 kg K ha-1; MD2 = 33.1 kg N, 8.2 kg P, and 15.6 kg K ha- 1) applied alone or combined with hill-placed manure (FYM) at 3 t ha-1 was compared to an unfertilized control and a broadcast fertilizer treatment at the recommended rate (RR; 76 kg N, 13.1 kg P, and 24.9 kg K ha-1). On average, microdose fertilization increased maize grain yields by 1090 kg ha-1 (99%) and 1201 kg ha-1 (110%) for MD1 and MD2, respectively, compared to the unfertilized control (1096 kg ha-1). There was no significant difference in yields between MD1, MD2 and RR in both years. Combining microdose fertilization with manure further increased yields by 848 kg ha-1 (40%) on average. There was a large variability in yields among farmers, from 420 to 1687 kg ha-1, 1419 to 3418 kg ha-1 and 1834 to 4475 kg ha-1 for the control, sole microdose (MD1 and MD2) and microdose + FYM treatments, respectively. Variability tended to be lowest in the control treatment. Absolute yield response to microdose fertilization tended to decrease with increasing yields in the control plots and was well explained by the combination of some measured soil parameters (clay and/or silt, total carbon, exch-Mg, pH) and weed pressure. Based on the value-cost ratio (VCR) the economic performance of the RR treatment was less than that of the microdose treatments (alone or combined with manure) despite the higher labor cost associated with the latter treatments. MD1 should be favored over MD2 because yields were not significantly different yet the risk of achieving low VCRs was lower in MD1. Despite the greater variability compared to the control, the risk of no return on investment was nearly nil for MD1 (6%) and MD1+FYM (2%) as a result of the strong increase in yield. Despite the overall good performance of fertilizer microdosing, more effort is needed to better understand crop response to microdose fertilization for a broader range of environmental conditions in Benin in order to fine tune recommendation domains.

Keywords: Microdose fertilization, Maize yield response, Management and environmental factors, Northern Benin, Profitability

4.1 Introduction

Microdose fertilization was introduced less than two decades ago as a means to increase crop productivity in the low-input, rainfed cropping systems of Sub- Saharan Africa (SSA) where farmers face severe challenges related to the low inherent soil fertility, limited availability of organic amendments, low capacity to invest in external inputs and high production risk (Buerkert et al., 2001; Tabo et al., 2007). As compared to the blanket fertilizer application rates hitherto recommended by agricultural research and extension services, microdose fertilization is characterized by a much higher fertilizer use efficiency and requires smaller financial investment, thereby making this technology more suitable to smallholder farmers. This has been well documented for sorghum and millet in the semi-arid zones of West Africa (e.g, Aune et al. 2007, 2012; Tabo et al., 2007; Hayashi et al., 2008; Palé et al., 2009; Bagayoko et al., 2011; Camara et al., 2013; Ibrahim et al., 2015a, b; Bielders and Gerard, 2015). More recently, on-station trials also demonstrated the potential benefits of microdose fertilization for maize in sub-humid climatic conditions in Benin (Tovihoudji et al., 2017b). This is of particular interest since, contrary to millet and sorghum for which market opportunities are limited, maize is increasingly grown as a cash crop in northern Benin which constitutes an additional incentive for farmers to invest in this technology. Though field studies have consistently established the benefits of microdose fertilization in low input farming systems when considering the average agronomic or economic performances, there is a growing concern that such average responses are insufficient to properly assess a technology considering the diversity of smallholder farming environments and practices (e.g., Tittonell et al., 2005, 2010, 2011; Zingore et al., 2011; Giller et al., 2011; Chikowo et al., 2014). Indeed, several studies in SSA have documented high variability in yield response to microdose fertilization, even within the same agro-ecological zone. For example, Buerkert et al. (2001), Bationo et al. (2005) and Tabo et al. (2011) reported a large variability in millet responses to microdose fertilization in Niger, with yield increases ranging from nil up to 2000, 1500 and 900 kg grain ha-1, respectively. Such variability is large considering that average millet yields in unfertilized farmer’s fields were of the order of 400-500 kg ha-1. Also in south- western Niger, Bielders and Gérard (2015) showed that a value-cost ratio (VCR) > 2 for microdose fertilization was achieved in only 59% of farmer’s plots (n=279) even though a VCR of 2 is often considered as a lower limit justifying investment in risky environments. Similarly, in southern Zimbabwe, Twomlow et al. (2010) reported a large variability in response to microdose fertilization, from slightly negative values to about +2000 kg grain ha-1 across a broad range 89

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields of soil, farmer management, and seasonal climatic conditions. According to these authors, 25% of farmers did not achieve a yield gain that would translate into acceptable net returns. The causes of the large variability in response to microdose fertilization are expected to relate both to environmental conditions (e.g., rainfall and soil) and crop management practices (e.g., planting density, weeding intensity) for a given cultivar within the same agro-ecological zone. For instance, Bielders and Gérard (2015) showed that sowing date, rainfall-related variables and planting density affected crop response to microdose fertilization in the Sahelian conditions of western Niger. These authors further demonstrated that microdose fertilization should be targeted preferentially to all fields or parts of fields where low yields are expected. Although the typology of factors affecting yield variability may be broadly similar across regions, the extent to which they affect yield response will undoubtedly differ depending on soil characteristics, climatic conditions, crop type and agricultural practices (e.g., Falconnier et al., 2016; Ronner et al., 2016). Characterizing the extent of the yield response variability and its causes thus remains an important step to develop meaningful and flexible recommendations that allow farmers to use scarce fertilizer and organic resources efficiently (Giller et al., 2011). With such understanding, site-specific recommendations with known levels of risk may be issued to smallholder farmers, which would greatly benefit the credibility of the technology and ultimately help its rapid diffusion. Because mineral fertilizers only supply macronutrients, it has been advocated to combine microdose fertilization with manure applications along the principles of integrated soil fertility management (ISFM) (Vanlauwe et al., 2011, 2015). Indeed, besides supplying macro and micronutrients, manure is essential for maintaining soil physical, chemical and biological quality in the long run (e.g. Bationo et al., 2007; Liu et al., 2010; Kihara et al., 2011). In environments where manure availability is limited, hill-placement of farmyard manure appears especially promising (e.g., Ibrahim et al., 2015a, b, 2016; Tovihoudji et al., 2017a). Fuerthemore, the manure may in some cases alleviate soil-related constraints (e.g., supply of micronutrients, improved soil water retention) that restricted crop response to microdose fertilization. Hence, combining microdose fertilization with manure may result in reduced variability in crop response, thereby making the technology more predictable and possibly less risky for smallholder farmers. The specific objectives of the present study were therefore (1) to quantify the variability in maize yield response to microdose fertilization alone or in combination with manure in smallholder farmers’ fields in northern Benin; (2) to identify the main agronomic management and environmental factors that govern such responses and (3) to evaluate the economic profitability and risk associated with each treatment.

4.2 Materials and methods

4.2 Materials and methods 4.2.1 Study sites and farm characteristics

During two consecutive rainy seasons (2014 and 2015), a series of demonstration trials were established in collaboration with farmers across an area of approx. 600 km² in the Ina district (North-eastern Benin), 70 km north of Parakou (Figure 4.1). Ina is located in the agro-ecological region III where annual rainfall ranges between 900 and 1200 mm. The annual rainfall for the last 30 years at Ina was 1148 ± 184 mm (mean ± standard deviation) and the average temperature was 27.5°C (CRA-Nord Climate Database, 1982–2015). The rainfall distribution is unimodal, characterized by a rainy season that occurs between May and October, and a dry season that prevails during the rest of the year. July and August are the wettest months. The soils have low inherent fertility and are classified as ferruginous tropical soils in the French soil classification system which corresponds to Acrisols or Lixisols according to the World Reference Base (Youssouf and Lawani, 2002).

4.2.2 Study design and management

During the 2014 and 2015 rainy season, a series of on-farm demonstration trials were established in collaboration with 18 farmers in 2014 and 33 in 2015 across five administrative villages (Table 4.1). Each farmer hosted one single, non- replicated trial with six treatments randomly arranged. Each farmer trial was considered as a replicate. The treatments consisted of: i) a control (no fertilizer, no manure), ii) a microdose option 1 (MD1): 2 g NPK15–15–15 per hill after plant emergence (10-14 days after sowing, DAS) + 1 g urea per hill at 45-50 DAS; iii) a microdose option 2 (MD2): 4 g NPK15–15–15 per hill after plant emergence + 1 g urea per hill at 45-50 DAS; iv) MD1 + farmyard manure 3t DM ha-1 after plant emergence (MD1+FYM), v) MD2 + farmyard manure 3t DM ha-1 after plant emergence (MD2+FYM) and vi) a recommended rate (RR): 200 kg ha-1 of

NPK15–15–15 after plant emergence + 100 kg urea ha-1 at 45-50 DAS. These treatments are equivalent to 23.8 kg N, 4.1 kg P and 7.8 kg K ha-1 for MD1; 33.1 kg N, 8.2 kg P and 15.6 kg K ha-1 for MD2; and 76 kg N, 13.1 kg P, and 24.9 kg K ha-1 for RR. The two microdose fertilization rates are identical to the rates tested previously in on-station experiment by Tovihoudji et al. (2017b). The RR treatment is the blanket fertilizer rate recommended by the National Agricultural Research System in the study area. The rate of manure used in the trials (3 t ha-1) is 2 to 3 times lower than the rate recommended by extension services because

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields of the limited availability of manure to farmers. To compensate for these lower application rates, and based on recent evidence (Ibrahim et al., 2015a, b; Tovihoudji et al., 2017a), the manure in all cases of these work was hill-placed at 96 g manure hill-1 to increase its efficiency. To apply microdose fertilization (both NPK and urea), small pits (0.05-m diameter and 0.05-m depth) were dug on one side of each planting hill and closed after application. The NPK and urea fertilizer in the RR treatment were spot- broadcasted at approximately 10 cm from each planting hill and not incorporated in accordance with farmer’s practice in the study area. For both the microdose fertilization and RR treatments, urea application was immediately followed by weeding-ridging as done by farmers. In case of combined application of microdose fertilization and manure (MD1+FYM and MD2+FYM), small pits (0.1-m diameter and 0.1-m depth) were dug on both sides of each planting hill and closed after the application of both amendments. In these trials, unwanted sources of variability were controlled by ensuring that all farmers participating in the trial used the same maize variety, planting density, and inorganic fertilizer type and manure source. Farmyard manure was taken from the barn of the Agricultural Research Centre of Northern Benin located in Ina district. Each season, the collected manure was thoroughly mixed, sampled for nutrient content, and measured and bagged for each plot before it was brought to the demonstration sites. A composite manure sample was analyzed for organic C as well as total N, P and K at the soil and plant analysis laboratory of the International Crop Research Institute for the Semi-Arid Tropics (ICRISAT, Sadoré, Niger). Manure characteristics were slightly different in 2014 (1.27% N; 0.48% P; 0.66% K; 14.3% organic C) and 2015 (1.30% N; 0.41% P; 0.62% K; 18.9% organic C), corresponding on average to 38.6±0.6 kg N, 13.4±1.5 kg P and 19.2±0.8 kg K ha-1 for the application rate of 3 t ha-1. The manure C/N ratio was 11.3 in 2014 and 14.5 in 2015. On each farmer's field, six contiguous plots measuring 4 m x 5 m were delimited, separated by a 1 m alley. Fields were ploughed by farmers, and planting was done under the control of technicians. Maize variety DMR-ESR (90 day-maturity) was planted at a spacing of 0.80 m × 0.40 m in all plots and thinned to two plants per hill at 10-14 DAS, giving a plant population density of 62,500 plant ha-1. Different plots were used in 2014 and 2015 to avoid interaction of residual effects. Participating farmers were identified through local farmer organizations and extension agents based on their experience in maize cultivation, willingness and consent to participate, and the accessibility of the field. The farmers were trained each year regarding the microdose application technique by research staff before the onset of the rainy seasons. The farmers managed fully their demonstrations from planting to harvesting, and the role of research was limited to the

4.2 Materials and methods monitoring of the management practices and the measurements. The choice of sowing, fertilizer application, weeding, thinning and harvest dates was left to each individual farmer. But sowing dates, weeding and harvest dates were identical across treatments at any given farmer field.

Table 4.1. Number of field demonstration sites per year in the five target villages. Villages Number of trials 2014 2015 Total Ina 4 9 13 Goua 3 4 7 Sikouro 4 4 8 Guessou 7 10 17 Konou - 6 6 Total 18 33 51

Figure 4.1. Location of Ina district (Municipality of Bembèrèkè) in northern Benin and distribution of rain gauges and demonstration sites.

4.2.3 Monitoring and measurements

Depending on the year, 7 to 11 geo-located manual rain gauges were installed from May to record individual rainfall events (Figure 4.1). Rainfall was measured by voluntary villagers recording water levels from gauges on paper tapes, which were later collected and encoded by a technician as rainfall amounts in a

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields spreadsheet. Rainfall from the nearest rain gauge was attributed to each demonstration site. From this data, total rainfall from the beginning to the end of the rainy season (May to October) and the cumulative rainfall from sowing to physiological maturity were calculated. For each experimental plot, field history was collected, including previous crops grown and the amount of mineral fertilizer or organic fertilizer applied. Also, the distance from the village to the trial field was calculated. The farmers received a manual (including record sheets) outlining the agreed experimental protocols. The farmers were assisted by a field assistant to fill in the record sheets. Information recorded included the timing of the different operations (sowing, fertilizer application, weeding, harvesting). Farmers could also record other observations such as the problems encountered (pests and diseases). Weed pressure was scored visually by the two field technicians who regularly visited the fields on a scale from 1 (< 25% of the plot surface covered with weeds) to 4 (> 75% of the plot surface covered with weeds) at 30 and 60 DAS. Scores were compared and debated until there was a consensus among the 2 technicians regarding each level. Farmers did not always remember exactly the amount of mineral fertilizer previously applied. Since we could not assign reliable quantities of fertilizer, it was used as a categorical factor (yes/no). At crop maturity, maize plants were sun dried in the plot for two weeks (farmers’ common practice). Thereafter, farmers and researchers jointly harvested maize cobs from the three middle rows of each plot. Cobs were weighed in the field, transported to the laboratory and oven-dried at 60 °C for 48 h to determine moisture content. After threshing, maize grain yields were then calculated and expressed in kg ha-1 on a dry weight basis. Yield response to any given fertilized treatment was calculated using the following ratio:

Yield response = (Yf – Yc) / Yc (Eq. 4.1) where, Yf and Yc are the grain yield (kg ha-1) from the fertilized and control treatments, respectively.

4.2.4 Baseline soil and manure analyses

To assess the nutrient status of the soils before sowing and amendment application, soil samples (0-20 cm soil depth) were taken at randomly selected points in each treatment plot and bulked as a composite sub-sample per plot. The sub-samples from each plot were then mixed and one composite sample per field was sent for analysis. The samples were air-dried, passed through a 2-mm sieve, and stored at room temperature prior to analysis. Particle size distribution was determined using the pipette method (Gee and Or, 2002). pH (H2O) was

4.2 Materials and methods measured potentiometrically in a 1:2.5 soil:water suspension (van Reeuwijk, 1993). Organic carbon was determined by the method described by Walkley and Black (1934), total N by the Kjeldahl method (Houba et al., 1995) and available phosphorus by the Bray 1 method (van Reeuwijk, 1993). Exchangeable bases (Ca2+, Mg2+, K+ and Na+) were determined after extraction by the ammonium acetate solution at pH 7 using the method described by van Reeuwijk (1993). All chemical analyses were carried out at the Sadore, ICRISAT laboratory in Niger.

4.2.5 Economic and risk analysis

An economic analysis of the profitability of microdose fertilization alone or in combination with hill-placed manure was conducted by calculating the gross margins (GM), the net returns (NR), and the value-cost ratios (VCR). Total revenue was calculated by multiplying grain yield with the average price of grain. The GM was calculated by subtracting variable costs (the sum of the costs of fertilizer and manure and the labor costs for application) from the total revenue. The NR was calculated by subtracting the total cost of cultivation (the sum of the fixed and variable costs) from the total revenue. Fixed costs included the cost of seeds and all major labor charges (field preparation, sowing, weeding, ridging, harvesting, and threshing), whereas variable costs included the cost of fertilizers and FYM as well as labor charges for the application of the different fertilizer and manure rates. VCR was computed as the difference in grain yield between the fertilized and the control plot multiplied by the unit market price of grain, divided by the variable costs. Prices of maize grain and mineral fertilizer were obtained from a market survey carried out in the study area in 2015. The fertilizer acquisition cost included the price of purchasing and the transportation cost (Table 4.2A). The maize grain prices fluctuate between 100 and 200 FCFA kg-1 during the year, with an average of 150 FCFA kg−1 (1 US$ = 500 FCFA; Table 4.2A). Since there is still no market for manure in the study area and farmers consider it as a free input, the value of farmyard manure was estimated to be equal to the costs required for collecting manure and transporting it to the fields and ranged between 200 and 600 FCFA per 100 kg bag, with an average of 400 FCFA (Table 4.2A). The cost of application of amendments included the labor for digging the holes, applying fertilizer or manure and incorporating into the soil. The time required for the remaining management practices (sowing, weeding or harvesting) did not vary significantly across farmer fields and treatments and were not included in the calculations. Labor requirements were estimated each year by direct observation for each whole treatment plot with 2-3 farmers per village. For each treatment and activity, the duration and the number of people were

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields recorded. The labor times for each task per treatment were calculated and converted into costs (Table 4.2A). Casual labor in the study area at the time of study was paid about 4 US$ per day, corresponding to 0.5 US$ h-1 on the basis of 8h work per day. For the combined fertility management treatments (MD1+FYM and MD2+FYM), labor costs for applying manure and fertilizer were summed. Risk was assessed on the basis of the probability of achieving a certain value-cost ratio (VCR) for a given treatment. In economic terms, a VCR value greater than 1 means that the cost of investment in fertilizer and additional labor costs are recovered, while a VCR of 2 represents 100% return on investment (Kihara et al., 2015; Ronner et al., 2016). In addition to calculations based on the mean cost of inputs and outputs, four scenarios were evaluated to assess the effects of fluctuations in prices on the VCR on the basis of the minimum and maximum values of inputs (fertilizer and/or manure and labor) and outputs (Table 4.2B).

Table 4.2. Input and output prices (A), and description of scenarios (B) used in the economic and risk analysis. (A) Input and output prices Unit Min Max Average Inputs (purchasing + transport) NPK and urea fertilizer US$ per 50 kg bag 24 37 28 Manure US$ per 100 kg bag 0.4 1.2 0.8 Fixed costs Tillage US$ ha-1 50 90 60 Sowing US$ ha-1 10 20 14 Weedings/ridging US$ ha-1 30 64 56 Harvesting and threshing US$ ha-1 20 40 28 Variable costs Hole digging US$ ha-1 14 28 20 Microdose fertilization (NPK + urea)* US$ ha-1 18 37 28 RR (NPK + urea) US$ ha-1 12 24 18 Manure application* US$ ha-1 24 48 36

Output prices Maize grain US$ kg-1 0.2 0.4 0.3

(B) Scenarios Codes Description Scenario 0 S0 Average grain and average fertilizer and/or manure+labor prices Scenario 1 S1 Minimum grain and minimum fertilizer and/or manure+labor prices Scenario 2 S2 Minimum grain and maximum fertilizer and/or manure+labor prices Scenario 3 S3 Maximum grain and minimum fertilizer and/or manure labor prices Scenario 4 S4 Maximum grain and maximum fertilizer and/or manure+labor prices *Applying fertilizer and closing the holes.

4.3 Results

4.2.6 Statistical analysis

Prior to analysis, data were carefully checked for outliers using descriptive statistics, boxplots and correlation analysis. Yield was square root transformed to ensure homoscedasticity of residuals, while soil Exch-Na and P-Bray 1 were log10 transformed. Pearson’s correlation was used to characterize the relationships among soil parameters on the one hand, and rainfall and related variables on the other hand. Scatter plots of the distributions of yields and VCRs for the different treatments were constructed for more informative understanding of the variability (Vanlauwe et al., 2016). Yield stability was analyzed by plotting treatment yields vs. the environmental mean, i.e., the mean yield of all treatments at a given site (Guertal et al., 1994). The slope of the linear regression was thereafter used to evaluate yield stability (smaller the slope, greater the yield stability; Guertal et al., 1994). The effects of treatment on maize grain yield, GM and VCR was first assessed with a linear mixed model (LMM) using the Restricted Maximum Likelihood (REML) for variance estimation of slope. Farm site and year were considered as random factors and treatment as a fixed factor. The Tukey-Kramer post-hoc test (p < 0.05) was used for mean separation when the analysis of variance showed a significant treatment effect. With the same LMM approach, the relation between treatment yields and different environmental and management variables, and their interactions by treatment were assessed. First, the strength of each variable in explaining yield and yield response variability were explored in separate analyses. Subsequently, a combined analysis over the two years was made. A final statistical model was obtained by backward selection of variables using the Akaïke Information Criterion (AIC). Since there were significant correlations between the cropping year and related factors (rainfall, sowing date) on the one hand and distance, soil total carbon and nitrogen on the other hand, they were not combined in the same model to avoid confounding effects. Hence, one combined model was computed for each of these variables, and the one with the lowest AIC was retained. The model performance was evaluated based on the significance level of the estimated coefficients, the coefficient of determination (R2), the root mean square error (RMSE), the plots of predicted vs. observed values, and the AIC value. High values of R2 and low values of RMSE and AIC indicate a better performance of the model. All analyses were performed using GenStat Release 12.1 statistical software (GENSTAT, 2009).

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

4.3 Results 4.3.1 Farmer-field sites characteristics

Soil texture was predominantly loamy-sand, with sand content above 70% in most fields. Most soil chemical characteristics varied widely across farm sites

(Table 4.3). Soil pH (H2O) ranged from acidic (4.8) to near neutral (6.8). Soil total N and P-Bray 1 contents at most sites were very low to moderate (range of 0.39- 1.40 g kg-1 and 1.60-22.43 mg kg-1, respectively) compared to optimal values of 2.5 g kg-1 and 17 mg kg-1, respectively (Hazelton and Murphy, 2007). In general, soil samples at most sites showed moderate to low limitations with respect to exchangeable bases (Ca, Mg, K and Na). Soil organic carbon was within the low to moderate range (3.40-15.40 g kg-1). Experimental farms were located between 0.1 and 6 km away from the nearest village (Table 4.3). There were significant correlations among a number of soil parameters. Total C and N were strongly correlated (r = 0.92; p < 0.001), and inversely correlated with distance from the village (r = -0.63 and -0.55 respectively; p < 0.001). This is due to the fact that farmer fields close to village homestead are generally better managed and are therefore more fertile than other remote fields. Except between Ca and Na, there were positive and significant correlations (p < 0.05) among all exchangeable cations with bivariate correlation coefficients ranging between 0.24 and 0.75. Soil pH, exch-K, -Ca and -Mg, and clay content were also positively correlated (p < 0.001). Total rainfall (from May to October) varied between sites and years, ranging from 819 to 1183 mm in 2014, and from 918 to 1103 mm in 2015. This range corresponds to the low to medium range of rainfall conditions observed in the region (long-term average = 1148 mm). Rains in 2014 started early with rainfall more evenly distributed than in 2015 despite a greater number of short dry spells. Rainfall in 2015 was poorly distributed, with more heavy rainfall events but also longer dry spells than in 2014. This caused delay in sowing, partial crop failure and major differences in crop establishment. Sowing occurred on average around mid-June, but ranged anywhere between mid-May (DOY 138) and mid-July (DOY 220) in 2014. In 2015, sowing occurred on average 20 days later than in 2014 (DOY 183). As for the total rainfall, the cumulative rainfall between sowing and plant physiological maturity also varied widely across sites and cropping years (Table 4.3) and showed a high degree of correlation with the sowing date (r = 0.49; p < 0.001). About 18 % of the experimental plots had less than 25% weed cover, while 26% had a cover greater than 75% when averaged over the two observation

4.3 Results

dates. The antecedent crops were mostly maize (51%), cotton (37%) and soybean (12%). Most sites (64%) had received fertilizers at least once during the last 3 years. As is common in the study area, mostly cotton and rarely maize received mineral fertilizer, on average for both crops about 150±50 kg NPK15-15-15 ha-1 and 100±50 kg urea ha-1, before the first and second weeding, respectively.

Table 4.3. Descriptive statistics of major farmer-field sites characteristics (2014-2015, all data). Variable Units Min Mean (SD) Median Max Soil and land characteristicspHH2O - 4.78 5.62 (0.43) 5.67 6.75 Total carbon g kg-1 3.40 8.20 (3.10) 8.14 15.40 Total nitrogen g kg-1 0.39 0.74 (0.26) 0.69 1.40 P-Bray1 mg kg-1 1.60 6.42 (4.68) 4.80 22.43 Exch-K cmol+ kg−1 0.17 0.35 (0.16) 0.31 0.78 Exch-Ca cmol+ kg−1 0.80 2.67 (1.38) 2.55 5.40 Exch-Mg cmol+ kg−1 0.26 0.67 (0.24) 0.64 1.19 Exch-Na cmol+ kg−1 0.01 0.08 (0.12) 0.03 0.54 Sand (50-2000 µm) % 64 77 (5.0) 78 85 Silt (2-50 µm) % 11 16 (3.0) 15 25 Clay (< 2 µm) % 3 7 (3.0) 6 15 Distance from village km 0.1 2.6 (1.8) 2.0 6.0 Rainfall and related factors Seasonal rainfall 2014 mm 819 985 (103) 1000 1183 2015 mm 918 1024 (69) 1002 1103 Cumulative rainfall between sowing and plant physiological maturity 2014 mm 389 520 (75) 520 638 2015 mm 477 602 (70) 596 752 Sowing date 2014 DOY 138 183 (27) 197 220 2015 DOY 181 202 (15) 201 226 DOY= Day of the year

4.3.2 Maize grain yields and response to treatments

The LMM analysis of maize grain yield revealed significant treatment effects within the two years (p < 0.001; Table 4.4). The average grain yield was slightly higher in the relatively wetter 2014 year (2432 kg ha-1) than in the drier 2015 year (2301 kg ha-1; p = 0.013). Across years, all fertilizer treatment yield means were significantly higher than those of the control (p < 0.001; Table 4.4). We observed a strong positive response at all sites to both MD1 and MD2 which significantly increased maize grain yields by 1090 kg ha-1 (99%) and 1201 kg ha-1 (110%), respectively, compared to the unfertilized control (1096 kg ha-1). Overall, there was no significant difference in yields between MD1, MD2 and RR in both years

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

(Table 4.4). On average, the addition of manure in microdose plots significantly increased grain yields by 848 kg ha-1 (40%) compared to microdose fertilization alone (p < 0.001; Table 4.4). There was no significant difference between the two manured treatments (MD1+FYM and MD2+FYM).

Table 4.4. Effect of treatments on maize grain yield and descriptive statistics across the two years of the trials. The statistical analysis was performed on square root transformed yield data.

a Mean Year Treatment Mean SD Min Max Median P90 – P10 response kg ha-1 2014 Control 1089a 264 617 1429 1127 734 - MD1 2240b 332 1602 2796 2246 768 1152 MD2 2330b 343 1653 2791 2412 896 1241 MD1+FYM 3072c 431 2443 3707 3159 1121 1983 MD2+FYM 3268c 206 2802 3596 3261 485 2179 RR 2590b 490 1767 3290 2633 1263 1502 p value <0.0001 2015 Control 1103a 364 420 1687 1170 981 - MD1 2127b 419 1419 2960 2010 1226 1027 MD2 2262b 506 1432 3418 2115 1237 1161 MD1+FYM 2931c 579 1834 4236 2819 1483 1831 MD2+FYM 3079c 717 1846 4475 3041 1780 1978 RR 2305b 573 1423 3815 2296 1411 1203 p value <0.0001 MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1. Means followed by the same letter in the same year are not significantly different at p = 0.05. aInterpercentile range

The range of environmental and management conditions encountered across the sites resulted in a high variability of yields within a given treatment (Figure 4.2 and Table 4.4). Yields in the control plots ranged from 420 kg ha-1 to 1687 kg ha-1 across sites and years. About 30% of the control plots yielded less than 1000 kg ha-1 in 2014 while in 2015 this was almost 42%. Across all fertilized treatments, maize yields varied from 1602 to 3707 kg ha-1 in 2014 and 1419 to 4475 kg ha-1 in 2015. Yield distributions differed among fertilizer treatments and years. On average across both years, the control treatment had the lowest yield variability based on the SD and interpercentile range (Table 4.4). With the noticeable exception of MD2+FYM in 2014 for which the SD was low, the SDs

100

4.3 Results in the combined MD+FYM treatments were higher than in the sole microdose treatments. The variability in the RR treatment was also higher than in the sole microdose treatments (Table 4.4).

Figure 4.2. Cumulative probability density function of grain yields (kg ha-1) for the different treatments across the two years of the trials. MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1.

101

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

Yield stability was highest for the control treatment (Figure 4.3). The treatments without manure (MD1, MD2 and RR) had intermediate responses in all environments whereas the combined treatments (MD+FYM) were the most responsive to improvement in environmental conditions.

Figure 4.3. Stability analysis of maize grain yields in the control and fertilized treatments across the two years of the trials (2014–2015). Environmental mean is the average yield of all treatments at a given farmer field site. MD1= microdose option 1, MD2=microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1.

As for the absolute yields, yield responses to microdose fertilization also varied widely from +157 to +2863 kg ha-1 and +739 to +3428 kg ha-1 for sole microdose and combined MD+FYM treatments, respectively (Figure 4.4). Yield response to microdose fertilization tended to decrease with increasing yields in the control plots, especially when microdosing was applied alone (Figure 4.4).

102

4.3 Results

4000

MD alone

) 1

- MD+FYM 3500 MD alone (y=-0.32x+1485.2) MD+FYM (y=-0.18x+2167.3)

3000

2500

2000

1500

1000 Absolute Absolute response to microdoseha (kg

500

0 400 600 800 1000 1200 1400 1600 1800 -1 Maize grain yield in control plot (kg ha ) Figure 4.4. Absolute response to microdose fertilization as a function of yield in control plots (2014-2015). Since yields in the two microdose rates were not significantly different, the average yield of these two treatments was used with a distinction between microdose (MD) alone and microdose + farmyard manure at 3 t ha-1 (MD+FYM). 4.3.3 Economic profitability and risk analysis

The economic analysis based on average prices of inputs and outputs (scenario S0; see Table 4.2B) revealed a larger benefit for microdose fertilization (alone or combined with manure) compared to the control and recommended fertilizer rate (Table 4.5). Overall, the gross margins (GM) and net returns (NR) of MD fertilization and RR treatments were statistically similar, despite the higher additional labor costs for microdose fertilization (Table 4.5). Combining hill- placed manure with microdose fertilization significantly increased GM and NR on average by 170 and 409 US$ ha-1, compared to sole microdose fertilization and the unfertilized control, respectively (p < 0.001; Table 4.5). Based on average costs, the VCR of the MD1 treatment was 1.3 and 2.2 times greater than the MD2 and RR treatments, respectively (p < 0.001; Table 4.5). Like grain yields, there was no significant difference between the two manured treatments (MD+FYM; Table 4.5). Despite the fairly high cost associated with microdose fertilization in scenario S0, applying MD1 or MD2 allowed 94 to 100% of farmers to break even (VCR ≥ 1). Most farmers (80 to

103

Chapter 4. Variability in maize yield and profitability following microdosing application on farmers’ fields

94%) applying microdose fertilization surpassed a VCR of 2, compared to only 47% for the RR treatment (Figure 4.5; Table 4.6). Combining microdose fertilization with manure was also highly profitable (VCR ≥ 2) for 88 to 98% of the farmers (Figure 4.5; Table 4.6). VCR values depended greatly on fluctuations in input (fertilizer and additional labor costs) and output prices (Figure 4.5; Table 4.6). For the worst- case scenario S2 (low prices of grain and high fertilizer and labor costs), 41% to 86% of the farmers had a VCR < 2 using microdose fertilization (alone or in combination with manure), while for the RR treatment 96% of farmers had VCR < 2 (Figure 4.5; Table 4.6). In contrast, following a reduction of fertilizer and labor costs and an increase in prices of grain (scenario S3), only 0 to 10% had a VCR < 2 when applying the microdose fertilization (alone and with manure) whereas 18% of the farmers had a VCR < 2 for the RR treatment (Figure 4.5; Table 4.6). For all treatments, S1 and S4 tended to yield similar VCR distributions to the S0 scenario, indicating that costs and income tended to compensate each other in these scenarios.

104

4.3 Results

Table 4.5. Economic analysis (US$ ha−1) following different treatments over the two years of trial (2014–2015). Gross margin (GM), net return (NR) and Value Cost Ratio (VCR) calculation was based on the average prices for inputs and outputs (Scenario S0; see Table 4.2B). GM (US$ ha-1) NR (US$ ha-1) VCR (-) Treatment Mean SD Min Max Mean SD Min Max Mean SD Min Max Control 329.6a 100.3 153.5 506.2 157.6a 97.9 -18.5 334.2 - - - - MD1 574.3b 117.7 191.5 932.4 402.3b 144.5 19.5 760.4 4.3a 1.4 1.3 6.9 MD2 565.0b 134.7 243.3 1065.4 393.0b 176.1 71.3 893.4 3.0b 1.2 0.4 5.1 MD1+FYM 735.3c 158.6 306.1 1166.1 563.3c 185.9 134.1 994.1 3.6b 0.9 1.9 6.0 MD2+FYM 742.8c 177.4 355.4 1223.6 570.8c 215.9 183.4 1051.6 3.1b 0.8 1.3 4.6 RR 533.1b 166.4 240.9 963.3 361.1b 184.4 68.9 791.3 2.1c 0.9 0.4 4.1 F statistic 54.972 39.110 29.675 p-Value <0.0001 <0.0001 <0.0001 MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1. Means followed by the same letter are not significantly different at p = 0.05 (REML analysis).

105

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

1.00 1.00

0.80 0.80 MD2 MD1

0.60 0.60

0.40 0.40 S0 S0 S1 S1 0.20 S2 0.20 S2 S3 S3 S4 S4 0.00 0.00 0.0 2.0 4.0 6.0 8.0 10.0 12.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0

1.00 1.00

) - 0.80 0.80 MD1+FYM MD2+FYM

0.60 0.60

0.40 0.40 S0 S0 S1 S1 S2 0.20 S2 0.20 S3 S3 S4 S4

0.00 0.00 Cumulativeprobability ( 0.0 2.0 4.0 6.0 8.0 10.0 12.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 1.00

0.80 RR

0.60

0.40 S0 S1

0.20 S2 S3 S4 0.00 0.0 2.0 4.0 6.0 8.0 10.0 12.0 VCR (-) Figure 4.5. Cumulative probability distributions of value-cost ratios (VCR) following different treatments and scenarios of input and output prices over the two years of the trials (2014–2015). Vertical dashed lines represent a VCR of 2. MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1. See Table 4.2b for an explanation of the different scenarios.

106

4.3 Results

Table 4.6. Proportion of fields (%) with value-cost ratios (VCR) <1 or <2 depending on the treatments and for different scenarios of input and output prices over the two years of the trials (2014–2015). See Table 4.2B for an explanation of the different scenarios. Value-cost ratio Treatment Scenarios S0 S1 S2 S3 S4 %fields with VCR<1 MD1 0 0 6 0 0 MD2 6 10 20 2 10 MD1+FYM 0 2 2 0 0 MD2+FYM 0 0 12 0 0 RR 10 18 37 2 10 %fields with VCR<2 MD1 6 8 41 0 6 MD2 20 27 86 10 20 MD1+FYM 2 2 71 0 2 MD2+FYM 12 16 84 0 12 RR 47 65 96 18 45 MD1= microdose option 1, MD2 = microdose option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1. 4.3.4 Effects of other management practices and environmental factors

4.3.4.1 Effect of individual factors on grain yield

On average over all treatments, late sowing resulted in a significant decrease in grain yield (p < 0.001; Table 4.7). Besides late sowing, high total rainfall was negatively correlated with grain yields (p = 0.008; Table 4.7). Among the management factors, weed pressure and previous crop significantly affected maize yield (p < 0.001; Table 4.7). As shown in Figure 4.6a, demonstration sites with higher weed pressure scores (>50%) had a lower mean yield (on average 2180 kg ha-1) while those with lower pressure (<25%) had a higher mean yield (2650 kg ha-1). Cotton and soybean as previous crop increased maize grain yield on average by 379 and 296 kg ha-1, respectively, compared to maize as previous crop (2171 kg ha-1) (Figure 4.6b). Differences in soil and land characteristics between sites (distance from village, soil clay+silt content, total carbon and nitrogen contents) were significantly associated with yield (Table 4.7). Yields tended to decrease with increasing distance from the village (p < 0.001), while it increased with increasing soil clay+silt (p = 0.010), total carbon (p < 0.001) and nitrogen content (p = 0.002; Table 4.7). There was no significant interaction between any of the explanatory variables and microdose and/or manure treatments regarding grain yield.

107

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

Table 4.7. Results of the linear mixed model analysis of individual variables to explain maize grain yields (square root transformed) over the two years of farmer field trials (2014–2015). Non-significant variables (p-values > 0.1) are not shown. There were no significant interactions with microdose and or manure treatments. Groups Explanatory variables Estimate Std. Pr Rainfall and related factors Year (2014 vs. 2015) -0.049 Error1.498 (>|t|)0.019 Total rainfall -0.008 0.008 0.008 Sowing date -0.110 0.031 <0.001 Management factors Previous crop Maize vs. Cotton 4.011 1.100 <0.001 Maize vs. Soybean 2.679 1.780 0.133 Weed pressure* 1 vs. 2 -2.400 1.506 0.111 1 vs. 3 -4.480 1.600 0.005 1 vs. 4 -5.710 1.581 <0.001 Soil and land characteristics Clay+silt content 0.1421 0.114 0.010 Distance from village -1.612 0.378 <0.001 Soil total carbon 1.078 0.235 <0.001 Soil total nitrogen 10.430 2.840 0.002 *Weed pressure: 1 (< 25%) to 4 (> 75% of the plot surface covered with weeds).

108

4.3 Results

)

Grain yield Grain (kg ha

Figure 4.6. Maize grain yields as influenced by different levels of weed pressure (a) and previous crop (b) over the two years of the trials (2014–2015).

4.3.4.2 Combined analysis

Table 4.8 shows the best regression model which contained all significant variables. In addition to the treatment effects (p < 0.001), all variables that were significant in the separate analyses (e.g., sowing date, weed pressure, previous crop, soil clay+silt and total carbon contents) were systematically retained by the final regression model (Table 4.8). The final model also retained soil P-Bray1 (p < 0.001), exch-Mg (p = 0.023) and pH (p < 0.001) despite not being significant in the separate analyses. This final model explained overall 80% of the yield

109

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields variability. Treatments effects explain the largest part of the variability (61%), while management and environmental factors explain 19% of the total variability of which 9%, 7% and 3% for rainfall and related variables (sowing date), management and edaphic factors, respectively. The plot of predicted vs. observed yields indicates good overall performance with a root mean square error (RMSE) of 416 kg ha-1 (Figure B.4, Appendix B). Based on the RMSE calculated for individual treatments, the model performance tended to be better (lower RMSE) for the control and worse (higher RMSE) for the high nutrient input treatments (Table A.1, Appendix A). However, the relative RMSE (RRMSE) was reasonably similar for all treatments (14-20%).

Table 4.8. Results of the multivariate linear mixed model analysis to explain the variability in maize grain yields (square root transformed) over the two years of farmer field trial (2014–2015). Estimate Std. Error Pr (>|t|) (Intercept) 58.640 6.540 <0.001 Treatment Control vs. MD1 13.570 0.876 <0.001 Control vs. MD2 14.697 0.876 <0.001 Control vs. MD1+FYM 21.101 0.876 <0.001 Control vs. MD2+FYM 22.304 0.876 <0.001 Control vs. RR 15.359 0.876 <0.001 Weed pressure 1 vs. 2 0.150 0.871 0.864 1 vs. 3 -3.165 0.910 <0.001 1 vs. 4 -4.190 1.120 <0.001 Previous crop Maize vs. Cotton 3.522 0.634 <0.001 Maize vs. Soybean 1.580 1.020 0.036 Sowing date -0.077 0.017 <0.001 Clay+silt content 0.131 0.089 0.034 Soil total carbon 0.532 0.097 <0.001 P_Bray1 0.235 0.059 <0.001 pH_H2O -2.904 0.810 <0.001 Exch. Mg -4.400 1.930 0.023

Adjusted R-squared: 0.80 F-value= 63.11 on 16 and 239 DF; p-value: < 0.001 Weed pressure: 1 (< 25%) to 4 (> 75% of the plot surface covered with weeds). MD1= microdosing option 1, MD2 = microdosing option 2, RR=recommended rate, FYM=farmyard manure at 3 t ha-1.

110

4.4 Discussion

To better understand which explanatory variables best explain maize yields in microdose fertilization treatments, a LMM analysis was performed using yield response (rather than absolute yields) as independent variable. For this analysis, we focused only on microdose treatments. Furthermore, since the LMM analysis revealed no significant difference between the two microdose rates (Table 4.4), the average yield of the treatments was used to analyze the yield response. Overall, the regression analysis revealed three parameters of influence: soil clay and total carbon content, and pH (Table 4.9). The response to microdose fertilization was negatively related total carbon and pH and positively related to soil clay content. In addition, soil silt content had a negative impact on yield response to microdose fertilization when applied without manure, whereas soil exch-Mg had a negative significant relationship with the response to MD+FYM (Table 4.9). Weed pressure had a negative impact on yield response but this effect was significant only for the MD+FYM treatment.

Table 4.9. Linear regression model using absolute yield response (kg ha-1) in microdose plots (mean of MD1 and MD2) as dependent variable over the two years of trial (2014–2015). MD alone MD+FYM Estimate Std. error Estimate Std. error (Intercept) 3.39e+03 6.84e+02*** 4.77e+03 8.78e+02*** Clay 7.50e+01 1.75e+01*** 1.45e+02 2.89e+01*** Silt -3.50e+01 1.64e+01* -3.51e+01 2.00e+01 Total carbon -3.58e+01 1.29e+01** -3.64e+01 1.64e+01* Exch. Mg -4.89e+02 3.83e+02 -1.16e+03 3.50e+02*** -3.43e+02 1.07e+02** -3.58e+02 1.30e+02** pH-H2O Weed pressure 1 vs. 2 -4.60e+01 1.66e+02 -1.88e+02 1.52e+02 1 vs. 3 -2.30e+01 1.71e+02 -9.70e+01 1.56e+02 1 vs. 4 -1.57e+02 1.82e+02 -6.31e+02 1.67e+02*** Adjusted R² =0.25; Adjusted R² =0.33;

F-value= 6.31; p-value:<0.001 F-value= 6.49; p-value: <0.001 Weed pressure: 1 (< 25%) to 4 (> 75% of the plot surface covered with weeds). Since yields in the two microdose rates were not significantly different, the average yield of these two treatments was used with a distinction between microdose (MD) alone and microdose + farmyard manure at 3 t ha-1 (MD+FYM). *, **, *** significant at 0.05, 0.01, 0.001 respectively.

111

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields

4.4 Discussion 4.4.1 Effect of treatments on grain yield and economic profitability

Irrespective of environmental and management conditions, all fertilized treatments increased yields compared to the unfertilized control in both years. On average across both years, microdose fertilization increased grain yields by 1090 kg ha-1 (99%), 1201 kg ha-1 (110%) and 1353 kg ha-1 (123%) for MD1, MD2, and RR, respectively, compared to the absolute control (1096 kg ha-1; Table 4.4). Our results also showed an average yield increase of 848 kg ha-1 (39%) following the addition of manure in microdose plots compared to microdose alone across both years (Table 4.4). The observed average yield response to microdose fertilization (alone or combined with manure) in farmer fields confirm the overall good performance of this technology as reported earlier on the basis of on- station experiments in northern Benin for the same treatments at the same rates (Tovihoudji et al., 2017b). The average yield increases are substantially higher than what has been reported previously from on-farm demonstrations with maize in Zimbabwe and Malawi (e.g, Twomlow et al., 2010; Kamanga et al., 2013; Mashingaidze et al., 2013). For instance, results of Twomlow et al. (2010) showed that fertilizer microdosing (17 kg N ha-1) consistently increased maize grain yields by 19–51% compared to the control plots (894-1546 kg ha-1), across a broad spectrum of soil, farmer management and seasonal climate conditions. Mashingaidze et al. (2013) reported that microdose application (28 kg N ha−1) significantly increased maize grain yield on average by +50 to +2000 kg grain ha−1 irrespective of N formulation compared to the control plots (591-2429 kg ha-1), across three seasons in farmers’ fields. Even though the application method of manure and fertilizer differed from what was done in the present study, Ncube et al. (2007) reported that maize grain yield was increased by 550 to 1810 kg ha-1 compared to the control plots (1260 kg ha-1) when small doses of manure and nitrogen (3 t manure + 12-19 kg N ha-1) were applied in combination on farmers’ fields in Zimbabwe. From an economic point of view, all treatments led to high average net returns and VCR values > 2 (which is generally considered as the lower threshold for adoption in smallholder, risk-averse farming systems; Table 4.5),. Hence all tested technologies may appear suitable for the conditions of northern Benin. Nevertheless, mean VCR values were notably higher for treatments involving the lower microdose rate compared to the higher microdose rate. This is a direct consequence of the fact that yields were similar for MD1 and MD2 at a given

112

4.4 Discussion rate of FYM, yet the cost of fertilizer is twice as high for MD2 compared to MD1. Similarly, although average yields in MD1 and RR were similar, microdose application was economically more profitable, with an average VCR two times larger for MD1 compared to RR (Table 4.5). In the microdose technology, the lower cost of fertilizer more than compensates for the higher labor cost, resulting in a technology that is much better suited to smallholder farmers than RR (Table 4.4). By combining MD1 with manure the average economic return still remained interesting, even though this technology demands even more labor than sole microdosing (Table 4.5). Indeed, by combining MD1 with manure the average net return was about 4 times higher than the unfertilized control (Table 4.5). MD1+FYM may thus be an optimal choice when considering economic but also edaphic (soil improvement) benefits in environments where organic resources are scarce and farmer’s capacity to invest is limited. The importance of taking into account labor costs in calculating VCRs for the fertilizer microdosing technology cannot be overstressed. In the present study, mean VCR values for MD1 and MD1+FYM were 6.5 and 7.4, respectively, when not considering labor requirements, as opposed to 4.3 and 3.6 when labor is included (Table 4.5). Similarly, on the basis of on-farm demonstrations in Malawi, Kamanga et al. (2013) reported VCRs of 1-2 and 2-10 with and without labor consideration, respectively, across N rates, price scenarios and weeding intensities. Labor is a major bottleneck for the microdosing technology and greater effort should be invested in alleviating the labor requirements of this technology.

4.4.2 Understanding variability in yields and responses

Average crop responses provide only partial insight into the agricultural intensification potential of a given technology. Greater insight can be achieved by considering yield variability (Sileshi et al., 2010; Vanlauwe et al., 2016). In the present study, we observed a high variability of yields or yield responses between farms and within a given treatment (Table 4.4, Figures 4.2 and 4.4). High variability in crop responses to microdose fertilization has been reported for various crops and environments (Buerkert et al., 2001; Bationo et al., 2005; Tabo et al., 2011; Bielders and Gérard, 2015), but only one study related to maize specifically addressed this issue (Twomlow et al., 2010). The yield response values but also the ranges observed in the present study (Figure 4.4) were substantial higher than what have been reported previously for maize by Twomlow et al. (2010) in the relatively dry areas of Zimbabwe (from 0 to about 2000 kg ha-1). As the variability increases, even greater care must be taken before

113

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields widespread diffusion of a technology because the mean yield increasingly becomes a less relevant indicator of performance for individual farmers. When considering the standard deviation (SD) or inter-percentile ranges, most treatments increased yield variability compared to the unfertilized control (Table 4.4). Furthermore, the yield variability in the combined treatments was higher than in microdose fertilization alone, except in 2014 when MD2+FYM had an unexpected very low variability (Table 4.4; Figure 4.2). Njoroge et al. (2017) also reported an increase of SD values for maize yields from 0.8-1.2 t ha- 1 in the control to 1.1-2.3 t ha-1 in the full NPK plots, across sites and seasons following sequential application of macronutrients (N, P and K) in nutrient omission trials. The present results also confirmed the overall trend reported by Kafesu et al. (2018), who showed that intensification strategies increased consistently maize yields, but also led to higher SD values (unfertilized control < full NPK < full NPK+Manure). A greater variability does not necessarily imply an increased economic risk of low return on investment for farmers as long as mean yields increases are high enough. What may be more important for smallholder farmers is to achieve an acceptable economic return regardless of yield variability. In the present study, based on average grain and fertilizer prices and labor costs (Table 4.2), the use of microdose fertilization was economically profitable (VCR ≥ 2) for more than 90% of farmers despite the greater yield variability compared to the control (Table 4.6). Although it increases the risk on average, combining microdose fertilization with manure still remains an attractive technology since more than 80% of the farmers achieved a VCR ≥ 2; Table 4.6). These levels of risk are much lower than those reported for millet under the dryer, Sahelian conditions of the Fakara region (Niger) by Bielders and Gérard (2015), especially on high productivity plots (yield > 400 kg ha-1) where as much as 56-58% of the demonstrations sites experienced VCR values < 2 (without consideration of labor costs). More favorable edaphic and climatic conditions in northern Benin but also differences in crop type may explain this discrepancy. As shown in Figure 4.5 and Table 4.6, reducing the cost of inputs and increased prices for outputs may further boost the attractiveness of the MD technology.

4.4.3 Explaining variability in yields and responses

Overall, the LMM analysis identified sowing date, distance from homestead as well as soil- (pH, clay+silt, total carbon, P-Bray1 and exch.-Mg contents) and management-related variables (previous crops and weed pressure) as the environmental and management variables that best explain yield variability across

114

4.4 Discussion all input treatments (Tables 4.7 and 4.8). Among other variables, better yields were associated with larger clay+silt and total carbon contents, and good weed management (Table 4.8). In contrast, the yields were negatively related to late sowing and high rainfall. Late sowing may reduce the length of the growing period and increase the risk of end-of-season drought-stress. Higher rainfall may have caused temporary waterlogging or increased losses of nutrients by leaching. The negative relationship between yield and Exch-Mg and between yield and pH (Table 4.8) is counterintuitive, and may have been the result of confounding with other variables or nutrient imbalances. Based on our dataset, 80% of the variability in maize grain yields could be explained. In itself this adjusted R² value is very good, but it is inflated by the contribution of the 'treatment' factor. When applying the model to estimate yields at the level of a given treatment, the model performance tended to be highest (low RMSE) for the control and lowest (high RMSE) for the high nutrient input treatments. The part of variability attributed to environmental and management factors (19%) is comparable to the results of Bielders and Gérard (2015) who found that management and environmental factors explained 20% of the variation in millet yield in Niger following microdose fertilization. Despite this overall good performance of the model, a non-negligible fraction of unexplained variation remains (20% in the present study). This could be related to biotic factors affecting maize yield (such as pests and diseases), climate factors (such as temperature, drought stress, etc) that are poorly considered by simple rainfall- related indices, edaphic factors (such as micronutrient deficiencies, soil structure or slope) but also management factors which were not well characterized here (land preparation, weeding quality, etc...). As previously reported by Sileshi et al. (2010), Bielders and Gerard (2015) and Kihara et al. (2017), a significant negative relationship was observed between yield response to mineral fertilization and yield in the control plots (Figure 4.4). Whereas yields increased by approx. 1300 kg ha-1 following microdosing on low yielding control plots, on high yielding control plots this increase was limited to 1000 kg ha-1 (Figure 4.4). This appears consistent with the variable responsiveness concept put forward by Kihara et al. (2016), which states that plots will be increasingly less responsive to additions of macronutrients as their initial fertility increases – especially if these additions are small. This negative relationship between yield response and yield in the control plot was less marked when fertilizer microdosing was combined with hill-placed manure (Figure 4.4). This may be because the addition of manure lifted some deficiencies not related to N and P (e.g., micronutrients), thereby allowing a stronger response to macronutrients. A regression analysis was performed to further explain the

115

Chapter 4. Variability in response and profitability following microdosing application on farmers’ fields relationship between response and environmental or management variables. With the variables included in the model, we could explain 25 to 33% of the yield response variability (Table 4.9), comparable to the results of Ronner et al. (2016) who found that environmental and management factors explained 45% of variability in soybean response to phosphorus fertilizer in farmers’ fields in northern Nigeria. Nevertheless, the explanatory power of the regression remains limited. In future trials, plant rather than soil analyses may help clarify the limiting nutrients and the resulting plant response to crop intensification practices.

4.4.4 Opportunities and implications for scaling out

In northern Benin, maize production plays an important role in the rural economy and livelihoods. However, the low inherent soil fertility, high intra- seasonal climate variability, and poor management of agricultural land result in low yields. With the fertilizer microdosing technology, there is a potential opportunity for smallholder farmers who are constrained in their capacity to invest in external inputs to increase maize productivity and resource use efficiency. Indeed, with this technology, the average yields were always close to or even outperform the targeted yield of 3 t ha-1 for achieving the African Green Revolution (Sanchez, 2010). The present study also contributes to the increasing recognition that average crop responses or economic indicators are insufficient to fully assess the performance of a technology, and that measures of variability (e.g. the frequency distributions) are also needed to assess the risks (e.g., Sileshi et al., 2010; Vanlauwe et al., 2016). While there are concerns regarding the extra labor required for hill-placement of manure or fertilizers, the present study established that fertilizer microdosing alone (preferably the MD1 option) or combined with hill-placed manure generally resulted in economic returns that are much higher than the recommended fertilization practice even in low productive fields. In addition, since manure or fertilizer application is done after sowing and before weeding at a period of greater labor availability, it does not interfere with crop sowing and weeding, which are critical labor bottlenecks in the study area. Moreover, the benefits reported here appear to be so economically attractive that they should draw farmers’ interest towards fertilizer microdosing. Nevertheless, adequate institutional support will be required particularly to develop labor- reducing equipment (e.g., mechanization), to make mineral fertilizer more affordable and to support the internal maize market, which may allow fertilizer microdosing to remain highly profitable and further motivate farmers to use this

116

5.5 Conclusion technology. These two latter actions seem to be more important in a context of unstable input and output markets which lead to reduction in prices of grain and an increase in fertilizer and labor costs and thereby to an increased economic risk (Figure 4.5; Table 4.6). The present on-farm experiment allowed testing microdosing technology across a range of climatic, soil and crop management conditions, but challenges remain as to how to improve the relevance of the recommendations regarding microdose fertilization. Although it is clear that maize response to microdosing may be linked to multiple factors, the plot’s productivity level can be used as a first approximation to make targeted recommendations. Based on Figure 4.4, microdose fertilization should be targeted preferentially to low productive plots. This is particularly true for MD alone since for MD+FYM this effect is much less pronounced. In addition, fertilizer microdosing may give its best potential under the overall best climatic conditions, appropriate sowing dates and good weed management particularly because of the small amount of nutrients applied (Figure 4.3).

5.5 Conclusion

Although microdose fertilization rates correspond to 31-44% of the rates recommended by research and extension for maize in northern Benin, similar average yields and lower financial risks were achieved. MD1 must be favored over MD2 because MD1 is associated with lower economic risk yet yields are similar to MD2. In addition, hill-placed manure application made the application of microdose fertilization more attractive and economically viable for a large proportion of farmers. The range of environmental and management conditions encountered across the sites resulted in a high variability of yields between farms. Sole microdose fertilization should be targeted preferentially to low productive plots, while yield response to MD+FYM treatments seem less affected by management and climatic conditions. However, further studies are needed across a broader range of locations in Benin and over several production seasons in order to better understand crop response to microdose fertilization. Such an endeavor would also be facilitated by the development of dedicated decision support tools.

117

Chapter 5

Using the DSSAT model to support decision making regarding fertilizer microdosing for maize production in the sub-humid region of Benin*

*This Chapter has been submitted to Frontiers in Plant Science/Section: Agroecology and Land Use Systems as: Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Using the DSSAT model to support decision making regarding fertilizer microdosing for maize production in the sub-humid region of Benin.

119

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Abstract

Fertilizer microdosing is being widely promoted across sub-Saharan Africa, yet all recommendations regarding this technology are derived from short-term studies. Such studies are insufficient to properly assess the production risk caused by climatic variability. To address this issue while avoiding costly long-term experiments, a common and well accepted strategy is to combine results from short-term experiments with validated dynamic crop models. However, there have been few documented attempts so far to model fertilizer microdosing under sub-humid tropical conditions. The objective was therefore to evaluate the potential of a commonly used crop simulation model (DSSAT) for simulating maize response to fertilizer microdosing, and to use the validated model to assess the effects of inter-annual rainfall variability on maize productivity and economic risk. The DSSAT model was calibrated and validated against data from a 2-year on-station experiment (2014 and 2015) with 2 levels of hill-placed manure and five mineral fertilization options including broadcast and fertilizer microdosing. Model simulations were in good agreement with the observed grain and biomass yields at harvest for conventional broadcast fertilization, with relative RMSE and d-values of 12% and 0.96 for grain and 8% and 0.97 for biomass, respectively. For fertilizer microdosing, the N stress coefficient (CTCNP2) needed to be adjusted to avoid occurrence of large N stresses during simulation. After optimization, the model could adequately reproduce grain yields for fertilizer microdosing, with relative RMSE of 10%. Considering the long-term scenario analysis, the use of the validated model showed that fertilizer microdosing (2 g of N-P-K15-15-15 fertilizer + 1 g urea per hill, equivalent to 23.8 kg N ha-1) improved both the long-term average and the minimum guaranteed yield without increasing inter-annual variability and the economic risk compared to unfertilized plots. Even though combining fertilizer microdosing with hill-placed manure (at least at 1 t ha-1) was economically slightly riskier than fertilizer microdosing alone, this risk remained low since a VCR of 2 could be achieved in almost 100% of the years. Furthermore, combined application consistently reduced the inter-annual yield variability. Considering this as well as the other benefits of manure for soil health, combining fertilizer microdosing with small quantities of manure would be recommended to increase the sustainability of the system.

Keywords: CERES-Maize, model, decision support, Fertilizer microdosing, Manure, Inter-annual yield variability, Economic risk.

120

5.1 Introduction

In Sub-Saharan Africa (SSA), low crop yields are a persistent concern because of their impact on food security, chronic poverty and hunger (Vanlauwe et al., 2015; Morris et al., 2007). A major cause for the low yields is the low inherent soil fertility as well as nutrient and organic matter depletion in smallholder farming systems (Henao and Baanante, 2006; Bationo and Waswa, 2011; Tittonell and Giller, 2013). Much effort has therefore been invested over the past decades to provide farmers with suitable nutrient management practices in view of raising crop yields and income. Fertilizer microdosing (or microdose fertilization) is a strategic adaptation of conventional fertilizer management and has recently been advocated as a way to increase crop productivity, profitability and resource use efficiency for smallholder farmers in SSA (Muehlig-Versen et al., 2003). Indeed, over the last two decades, a number of on-station and on-farm experiments have demonstrated a generally strong positive impact of this low-input technology on crop yields and farmer income (e.g., Aune et al., 2007; Camara et al., 2013; Sime and Aune, 2014; Ibrahim et al, 2015a, b, c; Okebalama et al., 2016; Tovihoudji et al., 2017b). However, all presently published data regarding microdose fertilization is derived from short-term studies. Such studies are insufficient to properly assess the agronomic and economic risk of agricultural technologies. One critical risk component is related to the as yet unpredictable temporal distribution of rainfall over the course of a season. This uncertainty regarding the temporal rainfall distribution drives much of the behavior of smallholder farmers (e.g., Akponikpè et al., 2011; Marteau et al., 2011; Comoé and Siegrist, 2015; Guan et al., 2017), since partial or total crop failure due to drought stress can strongly affect the profitability of crop intensification technologies. Thus, it is crucial to consider this long-term variability when evaluating new technologies. This is especially true for fertilizer management practices given that rainfed crop response to fertilizer inputs is strongly dependent on the amount and distribution of rainfall (e.g., Akponikpè et al., 2010; MacCarthy et al., 2010; Folberth et al., 2013). Moreover, farmers’ willingness to adopt the microdosing technology will depend not only on increased yields and profitability but also on yield stability. Indeed, microdosing is generally considered as a stopgap option for subsistence farmers for whom achieving some minimal yield every year to cover household food requirements is more important than maximizing yields in favorable years. To address the issue of long-term climatic variability while avoiding costly long-term experiments, a common and well-accepted strategy is to combine results from short-term experiments with robust and validated dynamic crop models (Jones et al., 2003; Rezzoug et al., 2008; Holzworth et al., 2014). Such 121

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing simulation models can be used to explore the impact of long-term climate variability on crop productivity and profitability for a range of soil and water management strategies. Among these models APSIM (Agricultural Production Systems Simulator; Keating et al., 2003) and DSSAT (Decision Support System for Agrotechnology Transfer; Jones et al., 2003) are the two most frequently and widely used in SSA. Unlike for conventional fertilization practices, only a few studies have so far attempted to model crop response to fertilizer microdosing. Modeling fertilizer microdosing using soil-plant-atmosphere models such as DSSAT and APSIM represents a specific challenge since such 1-D models are not well suited to deal with localized fertilizer placement. Previous attempts were all based on the use of APSIM in the context of southern and eastern Africa (Cooper et al., 2008; Twomlow et al., 2008; Turner and Rao, 2013). However, none of these studies actually demonstrate that APSIM was capable of properly reproducing crop response to microdose fertilization since simulation results are not compared to measured data. In the present study, we seek to model the response of maize to microdose fertilization under the sub-humid tropical conditions of northern Benin using DSSAT. The choice of DSSAT was based on the fact that the suitability of the CERES-Maize module implemented in DSSAT for simulating maize growth and yield has been successfully demonstrated for a broad range of soil, management and climatic conditions in smallholder farming systems in SSA (e.g., MacCarthy et al., 2012, 2017; Ngwira et al., 2014; Corbeels et al., 2016; Adnan et al., 2017a, b), including the sub-humid region of Benin (e.g., Igué et al., 2013; Saïdou et al., 2017; Amouzou et al., 2018). More specifically, the objectives of the study were to (1) assess the performance of DSSAT (calibration and validation) in simulating the effect of conventional nutrient management and fertilizer microdosing for maize in Northern Benin; (2) determine model sensitivity to key input and (3) use the validated model to assess the effects of seasonal climate variability on maize productivity and economic risk under fertilizer microdosing with or without manure through a long-term (32 years) scenario analysis.

122

5.2 Materials and methods

5.2 Materials and methods 5.2.1 Experimental data

5.2.1.1 Site description

The effect of fertilizer microdosing combined or not with hill-placed manure on maize yields was evaluated at the Agricultural Research Station of Northern Benin (CRA-Nord) located at Ina village (Ina district, municipality of Bembèrèkè) (9°57’N and 2°42’E, and altitude 365 m), 70 km north-east of Parakou. Ina is located in the agro-ecological region III in northern Benin, the main production zone of food and cash crops where annual rainfall ranges between 900 and 1200 mm. The average annual rainfall at Ina is 1148 ± 184 mm (mean ± SD) and the average daily temperature is 27.5°C (CRA-Nord Climate Database, 1982–2015). The climate is characterized by a single rainy season between May and October. The soil is classified as a ferruginous tropical soil in the French soil classification system with low inherent fertility, which corresponds to Lixisols according to the World Reference Base (Youssouf and Lawani, 2002). Maize (Zea mays L.) is the major staple crop in Benin. The total annual national production has increased from 219,593 tons in 1961 to about 1,376,683 tons in 2016 (FAOSTAT, 2016). In northern Benin, it is mainly produced under rainfed conditions on approximately 82-84% of the total area devoted to cereals crops (DPP/MAEP, 2010).

5.2.1.2 Experimental design and crop management

The experiment was conducted during two rainy seasons (2014-2015). Details of this experiment have been published elsewhere (Tovihoudji et al., 2017b) but are briefly described here. The experimental layout was a randomized complete block design with three replications within each manure stratum. Two manure treatments were considered: (i) no manure (NM) and (ii) hill-placement of farmyard manure applied each year at a rate of 3 t ha-1 (3M) at 10 days after sowing (DAS). Five mineral fertilizer options were tested within each manure stratum: i) a control (no fertilizer, NF); fertilizer microdosing at a rate of ii) 2 g of composite NPK (15-15-15) fertilizer per hill at 10 DAS + 1 g urea (46% N) per hill at 45 DAS (MD1), iii) 4 g of NPK (15-15-15) fertilizer per hill at 10 DAS + 1 g urea per hill at 45 DAS (MD2); iv) 50% (50F) and (v) 100% (100F) of the broadcasted recommended rate by the National Agricultural Research System (200 kg NPK (15-15-15) ha-1 at 10 DAS + 100 kg urea ha-1 at 45 DAS). This is equivalent to 23.8 kg N ha-1, 4.1 kg P ha-1 and 7.8 kg K ha-1 for MD1, 33.1 kg N

123

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing ha-1, 8.2 kg P ha-1 and 15.6 kg K ha-1 for MD2, 38 kg N ha-1, 6.5 kg P ha-1, and 12.5 kg K ha-1 for 50F and 76 kg N ha-1, 13.1 kg P ha-1, and 24.9 kg K ha-1 for 100F. Manure samples were taken in both years to determine chemical composition (Tovihoudji et al., 2017b). Different plots were used in both seasons, i.e., there was no cumulative effect of treatments. At the onset of the experiment, land preparation was done uniformly across all plots by tractor disk-plowing to a depth of 0.2 m. The improved and early maturing (90-days maturity) maize variety DMR-ESR-W (Downy Mildew Resistant, Early-Streak Resistant, White) was planted at a density of 31250 hills ha-1. Maize seedlings were thinned to 2 plants hill-1 two weeks after planting. Plots were weeded twice (15 and 30 DAS) and ridged with a hand hoe 45 DAS immediately after urea application.

5.2.1.3 Data collection

Different data were collected during the experiment including crop phenology (emergence day, date to anthesis and physiological maturity), leaf area and biomass time-series, grain and aboveground biomass yield, and N uptake at harvest. Pre-planting soil samples were analyzed for macronutrients (N, P and K), texture, pH, organic carbon and bulk density at various depths between 0 and 1 m. Periodic soil water content measurements were taken using in-situ calibrated portable soil moisture meter (TRIME-PICO IPH/T3, IMKO Micromodultechnik GmbH). Measurements were taken every 0.1 m from 0 to 0.6 m depth in all the plots. For the sake of clarity, water content data have aggregated over 0.2 m layers. Biomass was measured by destructively sampling whole plants from two planting holes every fortnight from 20 DAS until final harvest. Leaf area was recorded in each experimental plot by randomly tagging five plants from five different planting holes in the three middle rows reserved for final biomass measurement. The green leaf length and width were measured every fortnight with a ruler and the leaf area (LA) was calculated as : Leaf area (LA) = Leaf length * maximum width * k , where k is a shape factor with the value of 0.75 (Yi et al., 2006). The leaf area index (LAI) was calculated as the ratio of LA to the soil surface area of the hill. Final harvest was done by hand, and aboveground biomass and grain yields were recorded as described in Tovihoudji et al. (2017b).

124

5.2 Materials and methods

5.2.2 CERES-Maize model

5.2.2.1 Model description

In this study, we used DSSAT version 4.6, with CERES-Maize as the crop model (Hoogenboom et al., 2015). DSSAT-CERES-Maize has previously been shown to simulate accurately key soil and crop processes in the low input environments and low yielding maize systems in West Africa (e.g., Fosu-Mensah et al., 2012; MacCarthy et al., 2012, 2017; Adnan et al., 2017a, b) and in similar environments in the sub-humid zone of Benin (Igué et al., 2013; Saïdou et al., 2017; Amouzou et al., 2018). DSSAT is a complex non-linear dynamic model that simulates outputs such as crop development and yield as a function of a large number of input parameters and variables on a daily time step. It is informed about the speciﬁc weather (daily minimum and maximum temperatures, radiation and rainfall), crop and soil characteristics. The different input ﬁles comprising the modules for the field and cultivar characterization, the initial soil conditions and the management operations (planting, organic amendment or fertilizer application) are linked to the main structure of the model to simulate the case described in this paper. In this study, we used the ‘daily canopy photosynthesis method’ for maize photosynthesis (Jones and Kiniry, 1986), CENTURY to simulate soil carbon and nitrogen dynamics (Gijsman et al., 2002), the Priestly and Taylor method for evapotranspiration, and the Soil Conservation Service method (USDA Soil Conservation Service, 1972) for soil water infiltration. Below, a brief overview of the major processes simulated by the model with regard to the present study is provided. Further detailed descriptions are available in Jones and Kiniry (1986), Ritchie (1998) and Godwin and Singh (1998).

Phenological stages, growth rate and biomass partitioning. The different developmental stages, growth rate and biomass partitioning are affected by weather and soil conditions and are simulated based on thermal heat unit accretion in growing degree days (GDD) based on the daily maximum and minimum temperature (thermal time) and photoperiod (Jones et al., 2003). Maize potential growth in the model depends on the photosynthetically active radiation (PAR), and its interception is calculated as a function of leaf area index (LAI), plant population, and row spacing (Eq. 5.1):

푅푈퐸 푋 푃퐴푅 PGrate = [1 − 푒푥푝(푘 퐿퐴퐼)] 푥 퐶푂2 (Eq. 5.1) 푃푑푒푛푠푖푡푦 where PGrate is potential growth rate of maize biomass per day (g plant-1 d-1), PAR is photosynthetically active radiation (MJ m-2 d-1), Pdensity is plant

125

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing population density (plant m-2), k is a parameter that takes into account the effect of row spacing and plant population on light interception, LAI is leaf area index,

CO2 refers to the concentration of CO2 in the air that can be modified by the user. RUE is the radiation use efficiency (g DM MJ-1 PAR) defined as an input in the ecotype parameter file. Daily actual biomass growth rate (AGrate) is constrained by suboptimal temperatures, soil water deficits, and nitrogen deficiencies, as calculated below (Eq. 5.2):

AGrate = PGrate x min (Tfactor, Swfactor, Nfactor) x Soilfactor (Eq. 5.2) where AGrate is daily plant growth (g plant-1 d-1), min is a function that returns the minimum value, Tfactor is the temperature effect (0-1), Swfactor is the soil water stress factor (0-1), Nfactor is the nitrogen stress factor (0-1) and Soilfactor is the soil fertility factor (SLPF in the soils file, 0-1) that accounts for both biotic and abiotic conditions.

N demand, uptake and stress. In the CERES-Maize model as in most available crop growth models, the plant N budget is based on a comparison between nutrient availability and crop demand (Jones and Kiniry, 1986; van Oosterom et al., 2010a, b; Lizaso et al., 2011; Zhao et al., 2014). If demand exceeds supply, there is a deficiency, which modifies crop growth, thereby reducing subsequent crop N demand. If more N is available than is required by the crop for maximal growth, the surplus is absorbed by the plant up to a certain limit and stored in vegetative plant parts that is available for translocation later in the crop cycle. In CERES-Maize, to estimate crop N demand, the model compares the concentration of N in plant tissues with a target concentration called critical N

(Ncritical) by assuming that if the concentration of N in tissues is ≥ Ncritical, growth proceeds without any N limitation. But if the concentration of N is < Ncritical, then the crop experiences N deficit and growth is reduced accordingly. The daily nitrogen uptake depends on the N stress factor which also affects photosynthesis and leaf expansion. The N stress factor (Eq. 5.3) is calculated as a function of

Ncritical, Nactual (the actual vegetative N concentration) and Nmin (the minimum N concentration at which growth ceases) (Jones and Kiniry 1986; Godwin and

Singh 1998). Ncritical (Eq. 5.4) is determined as a function of the growth stage

(Gstage), Nmax (the maximum value for critical N concentration in the developing seed embryo, also named “CTCNP1”, default value= 1.52 in maize species file) and the CTCNP2 (the coefficient governing the rate of change in concentration as a function of growth stage; default value= 0.16 in maize species file). Nstress = 1.0- ((Ncritical-Nactual))/(Ncritical-Nmin)) (Eq. 5.3) Ncritical = 0.01 x exp (Nmax - CTCNP2 x Gstage) (Eq. 5.4)

126

5.2 Materials and methods

Mineral N from soil organic matter. Nitrogen availability comes from either organic matter mineralization or from mineral fertilizer application. The supply depends on the availability throughout the soil and the extent of soil exploration by the roots. Soil organic matter (SOM) mineralization and nutrient release is simulated using the Century model embedded in DSSAT (Gijsman et al. 2002) which divides SOM into three pools: SOM1, the active microbial pool with rapid turnover (days to months); SOM2, the intermediate pool with a slow turnover (years); and SOM3, the passive or recalcitrant pool which is very stable with turnover times of hundreds to thousands of years (Porter et al., 2010).

5.2.2.2 Model parameterization and calibration

Crop cultivar coefficients and management’s inputs. In the present study, the genetic coefficients for the maize cultivar used were calibrated based on the growth and development data recorded during the 2014 season (when the highest yields were observed) for the highest, broadcast mineral N treatment (NM-100F). To simulate the baseline soil water and N dynamics as well as the response to manure application, two additional treatments (the absolute control NM-NF and the manured treatment 3M-NF) were included in the calibration process. Values used for species-specific parameters were the default values for maize in the CERES-Maize model. Since the maize cultivar used in the experiment had not been previously modeled with DSSAT, the genetic coefficients were calibrated by choosing a default, medium-maturing cultivar from Ghana (Obatampa; identification code = GH0010 and ecotype IB0001) as a starting point and manually adjusting these parameters to minimize the root mean square error (RMSE) between simulated and measured data. First, the coefficients controlling phenology (P1, P2, P5 and PHINT; Table 5.1) were modified to match anthesis and maturity dates, and leaf number. Later, the G2 and G3 parameters were adjusted so as to minimize the RMSE between measured and modeled biomass and yield (Table 5.1).

127

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Table 5.1. Default and adjusted genetic coefficients of maize cv. DMR-ESR-W used in CERES-Maize. Parameters Default a Adjusted P1 (◦C day) 280 (130-380) 210.0 P2 (days) 0.000 (0-2) 0.000 P5 (◦C days) 750 (600-1100) 600 G2 (number) 540 (400-1100) 520.0 G3 (mg day-1) 7.50 (4-11.5) 9.50 PHINT (◦C day) 40.0 (30-90) 60.0 aA cultivar from DSSAT database (OBATAMPA); Values in parentheses are range of values from DSSAT/APSIM database for African cultivars and soils. P1: Thermal time from seedlings emergence to the end of the juvenile phase (expressed in ◦C day, above a base temperature of 8 ◦C) during which the plant is not responsive to changes in photoperiod; P2: Extent to which development (expressed as days) is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at a maximum rate (which is considered to be 12.5 h); P5: Thermal time from silking to physiological maturity (expressed in °C day above a base temperature of 8°C); G2: Maximum possible number of kernels per plant. G3: Kernel filling rate during the linear grain filling stage and under optimum conditions (mg day−1); PHINT: Phyllochron interval, i.e., the interval in thermal time (°C day) between successive leaf tip appearances.

Soil input data. The soil fertility factor (SLPF) was adjusted after adjusting species-specific parameters. It was modified manually and the value was set to 0.90 by minimizing the error between the observed and simulated total biomass. The root distribution weighing factor (SRGF) is an input for each soil layer and reflects physical or chemical constraints on root growth in certain soil layers (Ritchie, 1998). Its value ranges from 1 (indicating that the soil layer is most hospitable to root growth) to 0 (indicating that the soil is inhospitable for root growth). SRGF was estimated using a function in the DSSAT soil creation utility program based on soil texture, bulk density and soil organic carbon (Table 5.2). Measured soil properties including soil organic carbon, total nitrogen, soil bulk density (BD), pH and soil texture (percent silt, clay, and sand) were used as input (Table 5.2). Soil hydrological properties such as soil water content at field capacity (drained upper limit = DUL), at permanent wilting point (lower limit = LL) and at saturation (SAT) were taken from this thesis (Chapter 3). Soil hydraulic conductivity was estimated using pedotransfer functions available in DSSAT.

Initial mineral nitrogen content (NH4-H and NO3-N) was taken from unpublished data from the same experimental site. The runoff curve number (RCN) and drainage coefficient (SWCON) were set to 61 (default value) and 0.4 (adjusted), respectively, to simulate negligible runoff (accounting for the flat topography and good structure of the soil) and moderately well-draining soils which are characteristic of the soil at the experimental site. Measured field soil water content profiles were used as a basis for the calibration of the latter two parameters.

128

5.2 Materials and methods

Weather input data. Weather files for 2014 and 2015 were created using daily minimum and maximum air temperatures, rainfall, and solar radiation recorded at the experimental site and plotted using the WeatherMan utility program in DSSAT (Pickering et al., 1994). In addition, for the long-term analysis, a weather file was created for 32 years (1984–2015) of observed daily minimum and maximum temperature, solar radiation (collected at the nearest Meteo Benin synoptic weather station in Parakou, 70 km from the research site), and rainfall (collected from the experimental site in the Agricultural Research center).

Table 5.2. Soil physical and chemical characteristics at the experimental sites used for calibrating and evaluating the CERES-Maize model. Depth BD LL DUL SAT OC Clay Silt Sand pH SGRF cm g cm-1 m3 m-3 % - - 5-10 1.6 0.06 0.16 0.35 0.45 3.9 13.8 82.3 5.7 1.0 15-20 1.6 0.06 0.17 0.35 0.45 3.9 13.8 82.3 5.7 1.0 20-30 1.7 0.06 0.18 0.35 0.45 3.9 13.8 82.3 5.6 1.0 30-40 1.6 0.12 0.20 0.38 0.18 8.1 13.1 78.8 5.4 0.8 40-50 1.6 0.12 0.20 0.38 0.18 8.1 13.1 78.8 5.4 0.8 50-60 1.7 0.15 0.23 0.31 0.13 25.5 15.0 59.5 5.1 0.4 60-70 1.7 0.15 0.23 0.31 0.13 25.5 15.0 59.5 5.1 0.4 70-80 1.7 0.15 0.25 0.31 0.12 31.7 15.2 53.1 5.1 0.2 80-100 1.6 0.15 0.25 0.31 0.10 30.5 15.5 54.0 5.1 0.1 LL = lower limit, DUL = field capacity, SAT = saturated water content, BD = bulk density, OC = organic carbon, SGRF= Root growth factor

5.2.2.3 Sensitivity analysis

Firstly, a sensitivity analysis was carried out with respect to the 2014 grain yield data considering changes in selected model input variables within the range of ±75%. The variables considered were (1) soil chemical properties (SOC, total N and NO3 content in the top 0.40 m), and (2) physical properties (runoff curve number, drainage rate and water content at field capacity in the top 0.40 m) and (3) manure and fertilizer placement method (broadcasting without and with incorporation, or banding/hole placement at surface or 0.10 m beneath the surface). Each variable was varied individually, keeping all others constant. The sensitivity analysis of grain yield to the method and depth of application of fertilizer N rates using the initially calibrated model revealed in general an underprediction of growth and yield compared to observed values in the microdosing treatments (see section 5.3.4). In particular, for the low N rates used in fertilizer microdosing, the model simulated high nitrogen stress, thereby constraining the simulated crop growth to levels not consistent with observed growth data. As a result, the default value of the N stress coefficient CTCNP2 129

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

(Eq. 5.4) was selected for optimization. Indeed, higher CTCNP2 values lead to lower Ncritical values, thereby broadening the range of N concentrations which allow unrestricted growth. Based on a realistic range for this parameter (0.14 to 0.22), the value of 0.20 was retained. The performance of the modified model (called “optimized model”) was compared with the initial calibrated model using the default value of CTCNP2 (0.16) (called “baseline model”).

5.2.2.4 Model evaluation

The calibrated model was evaluated against the phenology, LAI and aboveground biomass time series, final grain and biomass yield data from the remaining treatments in 2014 and all treatments in 2015. The accuracy of model simulations was assessed based on the predicted deviation (PD, difference between the predicted and observed values in %), root mean-square error (RMSE) (Willmott, 1985), relative RMSE (RRMSE), index of agreement (d) and coefficient of efficiency (E1) (Liu et al., 2014). For time series data, the performance indicators were calculated across all measurement dates. A model is judged to simulate satisfactorily when PD, RMSE, RRMSE are close to zero, and “d” and “E1” are close to 1.0.

5.2.3. Model application: long-term simulation experiment

The effect of yearly climate variability on maize productivity and sustainability was simulated over a period of 32 years (1984–2015) using the calibrated model for microdosing (‘Seasonal analysis’ option in DSSAT; Hoogenboom et al., 2015). Factorial combination (4 x 3) of farmyard manure at 4 levels [0 (NM), 1 (1M), 2 (2M) and 3 (3M) t ha-1] and the 3 levels of microdosing tested in the experiment (NF, MD1 and MD2) was implemented. Manure and fertilizer microdosing were applied as in the experiment. During each simulation year, sowing was allowed when a total rainfall in excess of 20 mm occurred over three consecutive days between June 15 (DOY 165) and July 15 (DOY 195) which corresponds to the normal window for maize cultivation in the study area. Each year’s simulation was independent of the previous years as soil initial conditions were reinitialized 30 days prior to each planting. The seasonal analysis was evaluated by plotting the frequency distributions of maize yields and assessing stability of the model response using an agro- climatic risk indicator (INST, inter-annual standard deviation, Akponikpè et al., 2011):

130

5.3 Results

1 INST (kg ha-1) =√ ∑푛 (푌푠푖푚 – 푎푣푌푠푖푚)² (Eq. 5.5) 푛−1 푖=1 where Ysim is grain yield per year, avYsim is average grain yield over the number of cropping years and n the number of cropping years. The lower the value of INST, the more stable grain yield, hence the lower the probability of attaining extremely low (but also high) yields over the period (Akponikpe et al., 2011; MacCarthy and Vlek, 2012). Since greater temporal instability is not necessarily indicative of a higher risk provided that economic profitability is achieved, an economic profitability and risk assessment was performed. Economic risk was assessed and expressed in terms of probability of achieving a certain value-cost ratio (VCR) for a given treatment using the outputs of the simulation. VCR was computed as the difference in grain yield between the fertilized and the control plot multiplied by the unit market price of grain, divided by the variable costs (amendments + labor costs) using the 10-year average (2006–2015) market price of maize grain and fertilizer and labor cost estimated by direct observation in on-farm experiments (Tovihoudji et al., submitted). In the computation of VCR, the costs of the other inputs and management operations such as tillage, seed, planting, plant protection, and harvesting were assumed to be constant for all treatments. The variability of VCRs (median, min, max and CV) and the probability of achieving a VCR ≥ 2 or 4 (considered as a lower and upper limit justifying investment in risky environments) for a given management option were used as a measure of the long-term economic sustainability (Kihara et al., 2015; CIMMYT, 1988).

5.3 Results 5.3.1 Climatic conditions during the experimental years

Annual rainfall amounted to 1142 mm in 2014 and 1085 mm in 2015. From planting to harvest, the rainfall amounts of the two seasons were similar, but their distribution was slightly contrasted (Figure 5.1). Rainfall amount during the growing period was 694 mm in 2014 (43 rainfall events) and 797 mm in 2015 (46 rainfall events). Maximum daily temperature during the 2014 growing season ranged from 25°C to 37°C with an average value of 31°C, while minimum daily temperature ranged between 20°C and 24°C with an average value of 22°C (Figure 5.1). Average daily solar radiation was 18 MJ m-2 with a minimum value of 7 MJ m-2 and maximum value of 25 MJ m-2. In the 2015 season, maximum daily temperature ranged from 27°C to 38°C with an average value of 33°C, while minimum daily temperature ranged from 21°C to 24°C with an average of 22°C.

131

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Daily solar radiation ranged from 6 to 24 MJ m-2 with a daily average of 17 MJ m-2 (Figure 5.1).

)

Daily rainfall (mm day (mm rainfall Daily

Figure 5.1. Daily rainfall, solar radiation, maximum and minimum temperatures in 2014 and 2015. The horizontal red line shows the growing period (from planting to harvest). 5.3.2 Model calibration (manure and broadcast fertilizer)

The six genetic parameters adjusted in the present study are presented in Table 5.1. The range of these parameters were all close to the DSSAT “default” values for early to medium maturing maize varieties. After calibration, the model predicted the anthesis date and date of physiological maturity well (Table 5.3). Similarly, the predicted deviation for LAImax was low and ranged from -2 to 10% depending on the treatment (Table 5.3). The model accurately predicted the observed time course of LAI and aboveground biomass for the three treatments selected for calibration (Figure 5.2). Low RMSE values (0.07 m2 m-2 and 335 kg ha−1, respectively, for LAI and aboveground biomass across treatments and measurement dates) and high d-index values (0.99 and 0.98 for these two variables, respectively) were found (Figure 5.2a, b). Likewise, the simulation of

132

5.3 Results the soil water content (SWC) in the top 0.4 m of the soil profile was good as shown by low RMSEs (0.01-0.02 m3 m-3) and high d-values (0.79–0.84) (Figure 5.2c, d). In addition, there was generally good agreement between predicted and observed maize grain and above ground biomass yields at final harvest, with prediction deviations ranging between -1 to 8% and 5 to 16% for grain and aboveground biomass yields, respectively (Table 5.3).

Table 5.3. Results of model calibration for the 3 selected treatments in 2014. NM-NF NM-100F 3M-NF Obs. Sim. PD (%) Obs. Sim. PD (%) Obs. Sim. PD (%) Anthesis date (DAS) 56 55 -1.8 56 55 -1.8 56 55 -1.8 Maturity (DAS) 90 90 0.0 90 90 0.0 90 90 0.0 LAImax (m² m-2) 1.13 1.24 9.7 2.1 2 -4.8 1.45 1.42 -2.1 Grain yield (kg ha-1) 1099 1087 -1.1 2847 3068 7.8 1983 2039 2.8 Total biom. (kg ha-1) 3895 4071 4.5 7119 8272 16.2 5820 6121 5.2 NM = no manure; NF = no fertilizer; 100F = 100% of the recommended fertilizer rate; 3M= farmyard manure at 3 t ha-1; Obs. = observed; Sim. = simulated; PD = predicted deviation; Total biom. = Total biomass

)

2 2

LAI (m LAI

iomass(kg ha B

)

3 Soil Soil moisturecontent (m

Days after sowing (DAS) Figure 5.2. Comparison between observed and simulated time-series of maize LAI (a),

above ground biomass (b) and soil water content in the 0-0.2m (c) and 0.20-0.40 m layers (d) during model calibration in 2014. NM = no manure; NF = no fertilizer; 100F = 100% of the recommended fertilizer rate; 3M= farmyard manure at 3 t ha-1. Error bars =standard deviation (n = 3); RMSE = root mean- square error; E1 = coefficient of efficiency; d = index of agreement.

133

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

5.3.3 Model validation

After calibration, the model performance was checked with the remaining manure and broadcast fertilizer treatments in 2014 and all the manure and broadcast fertilizer treatments in 2015. The predicted days to anthesis were close to observations (55 and 56 days for simulated data against 56 and 54 days for observations during 2014 and 2015, respectively) for all treatments. Similarly for days to maturity, the deviation was 0 and -2 days between simulated and observed data during 2014 and 2015, respectively. The simulated leaf area index (LAI) values were very close to observed values (Figure 5.3a), with RMSE values varying between 0.04 m2 m-2 at 20 DAS and 0.9 m2 m-2 at 62 DAS across treatments and years. On average, the RMSE was 0.12 and 0.31 m² m-2 across treatments and measurement dates in 2014 and 2015, respectively. The corresponding d and E1 indices were 0.99 and 0.88, respectively, across treatments and measurement dates (Figure 5.3a). The model performance regarding above ground biomass was also good, with an absolute predicted deviation of 7-14%. RMSE values for TDM increased from 25 kg ha-1 at 20 DAS to 750 kg ha-1 at 90 DAS across treatments and years. The average RMSE was 304 and 378 kg ha-1 across treatments and measurement dates in 2014 and 2015, respectively, with average relative RMSE (RRMSE), d and E1 values of 9%, 0.99 and 0.99, respectively, across treatments, measurement dates and years (Figure 5.3b). Regarding final grain and biomass yields at harvest, the model performed well in simulating the response to combined application of manure and fertilizer across the two years (Figure 5.4). For grain yield, the RMSE and RRMSE were 327 kg ha-1 and 12%, respectively (Figure 5.4a). The d index of agreement and model efficiency E1 were 0.96 and 0.70, respectively (Figure 5.4a). The model performed also well in terms of biomass yield as indicated by the low RMSE and RRMSE (569 kg ha-1 and 8%, respectively) and high d and E1 values (0.97 and 0.69, respectively) (Figure 5.4b). The model estimated reasonably well the changes in soil water contents during the growth period as d-values were high (0.87 to 0.95) and RMSEs low (0.016–0.020 m3 m-3) in both soil layers (0–0.20 m and 0.20–0.40 m) across four contrasted treatments in 2015 (NM-NF, NM-100F, 3M-NF and 3M-100F), but tended to underpredict soil water in the 0.40–0.60 m layer towards the end of the growth period (Figure B.5, Appendixes).

134

5.3 Results

Figure 5.3. Model validation: comparison between observed and simulated maize LAI (a) and aboveground biomass (b) for all treatments not used during calibration (6 dates of measurement per treatment from 20 to 90 days after sowing) as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate (broadcast fertilizer) over two years. Error bars =standard deviation (n = 3); RMSE = root mean-square error; E1 = coefficient of efficiency; d = index of agreement.

135

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Figure 5.4. Model validation: comparison between observed and simulated maize grain (a) and aboveground biomass yield (b) at harvest as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and 0 (NF), 50 (50F) or 100% (100F) of the recommended mineral fertilization rate (broadcast fertilizer) over two years. Error bars =standard deviation (n = 3); RMSE = root mean-square error; E1 = coefficient of efficiency; d = index of agreement.

136

5.3 Results

Simulations showed that no treatment suffered from water deficit in both years despite some dry spells recorded in the two growing seasons, whereas excess water stress was simulated around 40-55 and 80 DAS in 2014 and 25-35 and 80 DAS in 2015 due to heavy rainfall events ≥ 40 mm per day recorded during those periods. The latter resulted in substantial simulated drainage and N leaching, especially in the high fertilized treatment (combined manure and fertilizer) in 2015 (data not shown). A short and long nitrogen stress period was simulated around 40 and 90 DAS, respectively, for the unfertilized treatments (NM-NF and 3M-NF) in both years (data not shown). The stress was more prominent in 2015 compared to 2014, especially at low levels of fertilization (NM-NF). The treatments with the highest fertilization rates were not affected by nitrogen stress but moderate N stress periods were simulated around 15 and 20 DAS in both years.

5.3.4 Sensitivity analysis

Soil chemical and physical properties. Grain yield was generally more sensitive to variations in the soil chemical properties in the low input treatments (NM-NF, NM-50F and 3M-NF) compared to the others (Figure 5.5a-d). This suggests that the amount of N applied was sufﬁcient in the latter treatments to mask the effect of the soil chemical properties on grain yield, as also reported by MacCarthy et al. (2012). SOC was the most sensitive chemical variable, followed by NO3. Increasing SOC and initial NO3 resulted in increased grain yields, with the highest sensitivity on NM-NF plots for SOC and on 3M-NF plots for NO3. Regarding the physical properties, grain yield was most sensitive to variations in DUL. A decrease in DUL resulted in lower yield, but this decrease was most marked in the fertilized treatments. In general, grain yield was least sensitive to changes in runoff curve number and drainage rate.

Fertilizer application rate and methods. The model was strongly sensitive to mineral fertilizer N application rate. Grain yields increased with increasing fertilizer rate, up to a maximum corresponding to 60 and 40 kg N ha-1 application under NM and 3M, respectively (Figure 5.5e, f). Crop response to N fertilizer inputs was much stronger for NM treatments than for 3M treatments. The model simulations were slightly sensitive to the method of fertilizer placement at low application rates, band placement leading to grain yield increases of +208 to +373 kg ha-1 compared to broadcast fertilizer application (Figure 5.5e, f). The model was more sensitive to depth of placement (0.10-m depth) when fertilizer was broadcasted (+40 to +745 kg grain ha-1 compared to surface broadcasting) than

137

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

when fertilizer was banded (no effect of depth of placement). There was no difference between the ‘fertilizer banding’ and ‘bottom-of-the-hole’ options (i.e. hole placement) in DSSAT (data not shown).

Sensitivity analysis of N demand under fertilizer microdosing. After selecting the bottom-of-the-hole fertilizer-placement method at 0.10 m depth as proxy to fertilizer microdosing, the sensitivity analysis showed that the model was quite sensitive to changes in the function controlling crop N demand for both years (Figure 5.6). Lowering the N stress coefficient CTCNP2 by 0.02 from the default value of 0.16 caused a higher N uptake (0 to +8% depending on N rate and year) but mostly a much stronger N stress and consequent reduction in growth and yield (Figure 5.6c-f). On the contrary, increasing the CTCNP2 value from 0.16 to 0.22 resulted in decreased N uptake (-19 to 0%; Figure 5.6a, b), and an increase in grain (0 to +33%) and biomass yields (0 to +21%; Figure 5.6c-f). Generally, these increases in yield led to predictions that approached those observed experimentally, with lower RMSEs (Figure 5.7). Overall, aboveground biomass and grain yield predictions were found to be sensitive to the CTCNP2 coefficient only for N rates below 55 and 70 kg ha-1 in 2014 and 2015, respectively.

138

5.3 Results

Changesin grain yield (%)

Change in parameter (%)

)

Grain yield Grain (kg ha

Fertilizer N rate (kg ha-1)

1 Figure 5.5. Sensitivity of grain yield to: (a-d) selected soil chemical and physical variables for four selected treatments and (e-f) application method and depth of fertilizer N rates under 0 (NM) or 3 (3M) t ha-1 of manure in the 2014 growing season. a- d: SOC (Soil organic carbon), TN (soil total N), and NO3 (Soil mineral NO3), RCN (Runoff curve number), DUL (ﬁeld capacity); e-f): BR-0=Broadcast, not incorporated; BR-10=Broadcast, incorporated at 0.10 m depth; BD-0=Banded on the surface; BD-10= Banded 0.10 m beneath surface or bottom of the hole.

139

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Figure 5.6. Sensitivity of maize aboveground N uptake (A, B), grain yield (C, D) and aboveground biomass (E, F) at harvest to incremental changes in the N stress coefficient from 0 to 90 kg ha-1 of N fertilizer rates using the hole-placement method at 0.10 m depth in 2014 and 2015. Error bars denote standard deviation (n = 3) for the two microdosing rates without manure (NM-MD1 and NM-MD2).

140

5.3 Results

Figure 5.7. Evolution of RMSE values of N uptake (principal axis), yield and aboveground biomass (secondary axis) at harvest to incremental changes in the N stress coefficient for MD1 and MD2 treatments over the two years during the optimization process. 5.3.5 Model testing for fertilizer microdosing

Following the sensitivity analysis, the CTCNP2 value was adjusted to 0.20 given the low RMSE values for the N uptake, grain and biomass yields over the two years (Figure 5.7). This led to a reduction of the N stress and to a satisfactory agreement and model efficiency between simulated and measured values (Figure 5.8; Table 5.4). From Figure 5.8c-d, it can be seen that the simulated grain yield showed globally good agreement with measured data in both years (d = 0.68 and RMSE = 323 kg ha-1 (10%) across years). The optimized model was also accurate for the aboveground biomass and N uptake as indicated by the performances indicators (RMSE and d values of 620 kg ha-1 (8%) and 0.79, and 7.1 kg ha-1 (10%) and 0.86 for aboveground biomass and N uptake, respectively, across years; Table 5.4). However, the optimized model generally simulated biomass yield better than the baseline model in 2015 compared to 2014 where the opposite was observed (Table A.2; Figure B.6, Appendixes). The latter will not affect the model use for scenario analyses since it is the grain yield data that will be used for this purpose.

141

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

2014 2015

)

N uptake N (kg ha

)

Grainyield (kg ha

)

Biomassyield (kg ha

1 Figure 5.8. Comparison between observed (box-and-whisker plots) and simulated (black points) N uptake (a, b) maize grain (c, d), and aboveground biomass yield (e, f) at harvest as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and two microdosing rates (MD1-MD2) in 2014 and 2015 using CTCNP2=0.20. MD1=microdosing option 1; MD2= microdosing option 2.

142

5.3 Results

Table 5.4. Statistical indicators showing the relationship between simulated and measured maize grain and biomass yield as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and two microdosing rates (MD1-MD2) over the two years (2014-2015). Baseline model Optimized model Indicators Grain yield Biomass yield N uptake Grain yield Biomass yield N uptake RMSE (kg ha-1) 494.5 782.5 12.9 322.5 620.0 7.1 RRMSE (%) 15.5 10.5 17.5 10.0 8.0 9.5 d (-) 0.59 0.76 0.55 0.64 0.79 0.86 RMSE = root mean-square error; d = index of agreement MD1=microdosing option 1; MD2= microdosing option 2. 5.3.6 Long-term scenario analysis regarding microdosing

The annual rainfall over the simulation period ranged between 843-1472 mm (average of 1182±171 mm) while the cumulative rainfall from sowing to harvest maturity ranged between 490-1030 mm (735±121 mm). The highest annual rainfall amounts were observed in 2002, 2009 and 2012, while the lowest were observed in 1986 and 1987 (Figure B.7, Appendix B). For the cumulative rainfall from June 15 to harvest maturity, the highest amounts were observed in 1988 and 2000, and the lowest in 1984, 1986 and 1987. Based on the long-term simulations using the optimized model, aboveground biomass and grain yields responded similarly to fertilizer and organic amendment inputs. Hence, only grain yield data was used to perform all analyses. The cumulative probability distribution of the simulated grain yields showed that yields were consistently higher with microdosing and further enhanced when combined with manure, compared to the no input treatment (Figure 5.9). For instance, average grain yield increased by 1272 kg ha-1 for MD1 without manure and by 1458-1647 kg ha-1 for MD1 with manure (1-3 t ha-1), respectively, compared to the no input treatment (1398 kg ha-1). On average, the yield difference between the NF and MD1 or MD2 treatments tended to decrease with increasing rates of manure. Finally, the yield difference between MD1 and MD2 tended to decrease with increasing manure additions, becoming negligible for 3M (Table 5.5 and Figure 5.10). Like the long-term average grain yields, the minimum yields also increase steadily with an increase in N inputs from manure and microdosing. The minimum grain yields increased from 411-1541 kg ha-1 in the no fertilizer treatment to 1800-2328 kg ha-1 in MD1 across manure rates (Table 5.5 and Figure 5.10). Thus, applying MD1 alone guarantees at least 1800 kg ha-1 every year. At the high microdosing rate (MD2), the minimum guaranteed yield was 1963 kg ha- 1.

143

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

Considering the inter-annual standard deviation (INST), the variability of grain yields following microdose fertilization alone was lower or similar to the no fertilizer treatment NF (Table 5.5 and Figure 5.10). Yield variability was little affected by manure application (1-3 t ha-1).

Table 5.5. Summary of simulated impact of yearly rainfall variations (32 years - 1984 to 2015) on maize grain yield (kg ha-1) in response to combined application of manure and fertilizer microdosing. Manure Fertilizer Mean INST Median Min Max NF 1398 360 1414 411 2238 NM MD1 2670 326 2607 1800 3400 MD2 2906 334 2875 1963 3651 NF 1910 380 1888 759 2770 1M MD1 2856 323 2807 1996 3570 MD2 3010 328 2984 2168 3769 NF 2285 350 2274 1166 3004 2M MD1 2976 321 2945 2172 3724 MD2 3062 323 3073 2335 3808 NF 2536 320 2504 1541 3215 3M MD1 3045 320 3022 2328 3800

MD2 3093 322 3097 2492 3813 NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure. INST = inter-annual standard deviation.

144

5.3 Results

Figure 5.9. Frequency distributions of maize grain yields as simulated by DSSAT over a 32-year period (1984 to 2015) in response to combined application of manure and fertilizer microdosing. NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure.

Based on average input and output prices, microdose fertilization (alone or combined with manure) appears to be economically profitable each year when considering a VCR threshold of 2 (Table 5.6). Under no manure application, the VCRs ranged from 2.3 to 4.9 irrespective of microdosing rates (Table 5.6). Combining hill-placed manure (1-3 t ha-1) with microdosing decreased the median VCRs by 0.8 (MD1) to 1.0 (MD2) point, compared to the sole microdosing application as a result of increased labour costs, but the probability to achieve VCR≥2 remained high (88-97%). Based on a VCR threshold of 4 required, sole microdose fertilization was noticeably less risky than microdose + manure (Table 5.6).

145

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing

)

- Grain yield Grain (kg ha

NM 1M 2M 3M NM 1M 2M 3M NM 1M 2M 3M INST 360 380 350 320 326 323 321 320 334 328 323 322 1 Figure 5.10. Maize yield variability as simulated by DSSAT over the study period (32 years—1984 to 2015) in response to combined application of manure and fertilizer microdosing. NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure. INST = inter-annual standard deviation.

Table 5.6. Summary of simulated impact of seasonal climate variations (32 years—1984 to 2015) on value cost ratio (VCR) in response to combined application of manure and fertilizer microdosing. Manure Fertilizer Mean SD Median Min Max %VCR≥2 %VCR≥3 %VCR≥4 NM MD1 3.7 0.58 3.8 2.3 4.9 100 91 32 MD2 3.7 0.58 3.8 2.3 4.9 100 88 28 1M MD1 3.0 0.47 3.0 1.9 4.0 97 56 3 MD2 2.9 0.53 2.9 1.7 4.0 97 44 3 2M MD1 3.0 0.53 3.1 1.9 4.2 97 63 6 MD2 2.9 0.58 2.9 1.7 4.2 97 47 3 3M MD1 3.0 0.59 3.0 1.8 4.4 97 63 6 MD2 2.8 0.61 2.8 1.6 4.3 88 34 3 NF= no fertilizer; MD1=microdosing option 1; MD2= microdosing option 2. 1M, 2M and 3M correspond to the application of 1, 2 and 3 t ha-1 of manure.

146

5.3 Results

5.4 Discussion 5.4.1 Model response to manure and conventional fertilization

The model calibration resulted in good predictions of phenological stages like anthesis and maturity as indicated by the different indicators (Table 5.3). DSSAT also accurately simulated the changes in soil water content in the various layers of the soil profile (Figure 5.2). Finally, model performance regarding biomass and grain yields at harvest was good across treatments and years (Figure 5.2). Such good performance was achieved in spite of the fact that maize response to P was not considered in the simulations yet P is often a strongly limiting factor in the West African semi-arid tropics (e.g, Nziguheba et al., 2016). This good model performance may be explained by the fact that all fertilizer and manure treatments considered here included proportional amounts of P. These P additions may have been sufficient to lift the P constraint and allow adequate model response to N. Alternatively, one may consider that the calibrated model response to N in fact reflects model response to the combined additions of N and P. The latter does not invalidate the model use for scenario analyses as long as treatments similar to those used for calibration are being investigated, which is the case in the present study.

5.4.2 DSSAT response to fertilizer microdosing practice

Using the default value of the N stress coefficient and the 0.10-m depth hole- placement setting, DSSAT tended to under-predict maize growth and yields in response to microdosing, especially for grain yield and in 2015. Fertilizer microdosing is known to increase fertilizer use efficiency compared to broadcast fertilization (Ibrahim et al., 2015). This positive effect has been attributed to faster early crop development (Hafner et al., 1993; Tabo et al., 2007; Aune and Bationo, 2008; Ibrahim et al., 2014, 2015a, c). In addition, localized application of nutrients may promote rapid fine root and root hair proliferation (Hodge, 2004; Ibrahim et al., 2014; 2015a, c). Ibrahim et al. (2014, 2015a, c) have reported an increase in root dry weight, total root length, and root length density, which may result in higher plant water and nutrient uptake and lower nutrient losses by leaching. In addition, an increase in lateral roots within the upper soil layers at early growth stages could stimulate the uptake of native nutrients. Without further adjustments, DSSAT did not seem able to correctly simulate crop

147

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing response to fertilizer microdosing. On the contrary, strong N stresses were simulated given the very low quantities of N supplied during microdosing. DSSAT and the CERES-Maize models have been extensively tested in the low-input cropping systems of West Africa under broadcast fertilization conditions (e.g., Fosu et al., 2012; Soler et al., 2011; Adnan et al., 2017a, b; Saïdou et al., 2017; Amouzou et al., 2018), but not under fertilizer microdosing conditions. A few studies have previously used APSIM to simulate fertilizer microdosing (Cooper et al., 2008; Twomlow et al., 2008; Turner and Rao, 2013), but none of these studies report on how microdosing was implemented in the model nor do they provide information regarding APSIM’s performance. Hence the present study appears to be the first to report on the performance of a conventional soil-plant-atmosphere model in the context of fertilizer microdosing. Following the sensitivity analysis, and given that N stress seemed to play a major role in the underestimation of maize yields in microdose treatments, the CTCNP2 value was adjusted in order to scale simulations of grain and biomass yields and N uptake to the observed values in both years (Figures 5.6 and 5.7). After optimization, model prediction at harvest matched satisfactorily with experimental observations, with a strong improvement in grain yield estimation (Figure 5.8; Table 5.4). Although adjusting the CTCNP2 factor does not capture the full complexity of the effects of fertilizer point-placement, other studies have previously had to adjust the N-stress factor in order to better model crop growth and N uptake under specific conditions (Liu et al., 2012; Yakoub et al., 2017). Adjusting the CTCNP2 factor offers a simple yet effective means of modeling the effect of microdose fertilization. Nevertheless, this approach should be confirmed in the future by testing the model against data from a wider variety environmental and management conditions. Furthermore, it may be worthwhile in the future to investigate in more detail the physiological response mechanisms of maize to fertilizer microdozing so as to better represent them in 1-D soil- plant-atmosphere models such as DSSAT.

5.4.3 Long-term scenario analysis and recommendations regarding fertilizer microdosing

Before widely promoting fertilizer microdosing in smallholder maize farming systems, one needs to evaluate two main issues. Firstly, the agronomic and economic benefits of the technique have to be demonstrated under on-farm conditions for a suitable range of environmental conditions. Furthermore, the associated short-term risks have to be low. These issues have been addressed

148

5.4 Discussion over the last decade through several on-farm experiments (e.g., Camara et al. 2013; Tabo et al., 2011; Bielders and Gérard, 2015; Twomlow et al., 2010), including the dry savanna region of northern Benin (Chapter 4). Secondly, given that smallholder farmers are usually risk-averse, the microdose technology should not substantially increase inter-annual variability in yield and income caused by variable rainfall conditions. Assessing the temporal yield stability, the economic risk and the capacity to achieve a minimum grain yield every each year are therefore a prerequisite before embarking on a large scale extension endeavor. As long-term field experimentation is expensive and time consuming, the use of calibrated and validated decision support systems like DSSAT constituted an attractive alternative to assess the long-term variability in maize yields following fertilizer microdosing. Overall, the predicted grain yields from the long-term simulations are within the ranges reported in on-farm experiments under the same microdosing treatments in northern Benin (Chapter 4). This suggest that the results can be confidently used to make recommendations regarding fertilizer microdosing. Considering the inter-annual standard deviation in grain yield (INST), applying microdose fertilization alone (MD1 or MD2) resulted in a lower variability compared to the no fertilizer input treatment (NF) (Figure 5.10). Interestingly, yield variability was little affected by manure application (1-3 t ha- 1). Hence, it may be concluded that intensification strategies combining manure and microdosing appear to be less instable (‘less risky’) for smallholder farmers compared to the sole application of manure or microdosing. Equally interesting, applying microdosing alone guarantees at least 1800 kg ha-1 every year without inducing additional inter-annual variability. However, in smallholder subsistence farming, the economic risk associated with the adoption of a new technology are more important than temporal yield stability. The results show that combining microdosing with manure increases the economic risk. Indeed, while SD values are similar, the mean VCR is lower for manure + microdose compared to microdose alone (Table 5.5). Nevertheless, a VCR of 2 is reached in almost 100% of the years, indicating that the level of risk associated with manure + microdose remains acceptable. Only if farmers have a very strong aversion for risk (e.g., require a VCR of 3 or 4; CIMMYT, 1988), does one see a notable short-term advantage for microdosing compared to the combined application of fertilizer and manure. This is because of the added labor costs related to hill-placed manure application. For instance, a VCR ≥ 3 is reached 88 to 91% of the time under sole microdose fertilization, compared to 34-63% when combined with manure (Table 5.5). However, model simulation do not take into account the long-term cumulative impact of the technologies on the soil. Complementing microdose fertilization with manure

149

Chapter 5. Using the DSSAT model to support decision making regarding fertilizer microdosing would be recommended for soil quality considerations and to increase the sustainability of the system even though it may compromise the short-term benefits. Moreover, because the VCR values can depend greatly on fluctuations in input and output market prices within and between seasons (Chapter 4), applying the lower microdosing rate (MD1) would be preferable for farmers irrespective of manure management.

5.5 Conclusion

In this study, we examined the ability of the DSSAT CERES-Maize model to accurately simulate maize response to fertilizer microdosing, and whether the validated model can be used to assess the effects of seasonal climate variability on maize productivity and economic risk. Using independent datasets for the calibration and validation, DSSAT exhibited good performance when simulating phenological stages, LAI, total biomass, grain yield, total N uptake, capturing the whole range of these variables, across conventional fertility management practices (broadcast fertilizer). For fertilizer microdosing, the N stress coefficient (CTCNP2) needed to be adjusted to avoid occurrence of large N stresses during simulation. After optimization, the model could adequately reproduce grain yields for fertilizer microdosing, indicating that it could be used as decision support tools through long-term scenario analysis. The 32-year, long-term simulation with the validated model showed that the application of 2 g of N-P- -1 K15-15-15 fertilizer + 1 g urea per hill (equivalent to 23.8 kg N ha ) improved both the long-term average and the minimum guaranteed yield without increasing inter-annual variability and the economic risk compared to unfertilized plots. Combining the application of fertilizer microdosing with hill-placed manure (at least at 1 t ha-1) consistently reduced the inter-annual yield variability. Even though combining the application of fertilizer microdosing with hill-placed manure (at least at 1 t ha-1) was economically slightly riskier than microdose alone, this risk remained low since a VCR of 2 could be achieved in almost 100% of the years. Considering this as well as the other benefits of manure for soil health, combining microdosing with small quantities of manure would be recommended to increase the sustainability of the system. Besides additional validation on the basis of a broader datasets, more in-depth investigations of the physiological response mechanisms of crops to fertilizer microdozing should be carried out so as to better represent them in 1-D soil-plant-atmosphere models such as DSSAT.

150

Chapter 6

Conclusions and perspectives

“It is not the quantity but the quality of knowledge which determines the mind's dignity.”

William Ellery Channing

151

Chapter 6. Conclusions and perspectives

6.1 Introduction

The research presented in this thesis aimed at assessing the agronomic and economic potential of localized application of organic amendments and fertilizers in maize-based cropping system in northern Benin, with a long-term goal of making recommendations, through field trials and decision support tools, for smallholder farmers in order to improve maize productivity, food security and household income. This general objective was achieved through (1) two on- station experiments to explore and understand the potential effect of localized application of manure and mineral fertilizer, (2) an on-farm experiment to evaluate in farmers’ fields the effects of promising treatments from the on-station experiments and (3) combining experimental and modeling approaches to support decision making regarding fertilizer microdosing for maize production through long-term crop model simulations. The implementation of the research activities towards achieving this goal led to some key findings and lessons discussed below. Besides these lessons, we reflected on the potential for extrapolating the results beyond the study zones, and on the perspectives for completing the development or improvement of a decision support tool for site- specific fertilizer microdosing recommendation for maize production in sub- Sahara West Africa.

6.2 Main findings 6.2.1 Potential of localized application of organic amendments and fertilizers in maize-based cropping system

The findings reported in Chapter 2 showed a significant increase in maize yields and economic profitability over a 4-year period when hill-placed manure and mineral fertilizer were combined and provided additional insight into the medium-term effects of the treatments. Soil fertility improvement in the top 0.2 m of the soil in the neighborhood of the planting hills was accompanied by an increase in maize grain and total biomass yields. The soil fertility improvement can be explained mainly by the fact that both the manure and fertilizer applications and the soil sampling were performed in the vicinity of the plants. In addition to this localized improvement in soil fertility, the hill-placement of manure and fertilizer may have resulted in a better uptake of the limited amount of nutrients by the roots possibly as a result of early root proliferation, moisture retention and enhancement of microbial decomposition and nutrient release near

152

6.2 Main findings the plant, leading to higher crop yields. Overall, there was a trend of increasing maize yields over the 4-year period for all fertilized or manured treatments which could not be linked to increasingly favorable climatic conditions. Hence, beneficial cumulative effects of the manure and fertilizer on soil quality in the vicinity of the plants and crop yields can be achieved with small quantities of inputs when these inputs are hill-placed. Besides the agronomic benefits, combined application of half the currently recommended rate of fertilizer (i.e 100 kg ha-1 NPK15-15-15 and 50 kg ha-1 of urea) and a realistic rate of manure (3 t ha-1) led to high economic returns (BCR and VCR) and appeared to be an economically sensible choice. As expected, VCR values were sensitive to fluctuations in fertilizer and maize grain prices in the market. Nevertheless, all treatments combining half the recommended rate remained financially attractive (VCR > 2; considered as a minimal condition for technology adoption in risky environments; Kihara et al., 2015) even in the case of large fertilizer price increases (+50%) or a substantial drop in maize price (-25%).One of the question that this study attempted to answer was whether the application of fertilizer microdosing to maize may contribute towards improved yields and greater nutrient use efficiency, alone or in combination with various manure management practices (Chapter 3). The present study demonstrates the potential of fertilizer microdosing to improve maize production in northern Benin and reinforces the earlier evidence from elsewhere regarding its effectiveness at improving maize yields (Twomlow et al., 2010; Sime and Aune, 2014; Kisinyo et al., 2015). The larger the application of manure and the more recent, the lower the response to microdose fertilization decreased since much of the most limiting nutrients (N, P) were already supplied by the organic amendments. Nevertheless, although it was shown efficient to apply microdose fertilization on unmanured plots (in terms of fertilizer use efficiency), it may be more efficient and sustainable to apply it in combination with small amounts of manure or on plots that have not been corralled for one or two years in the context of the sandy loam soils such as those found in northern Benin. The most noticeable finding that emerged from the Chapter 3 of this thesis was that the increase in total biomass yields with fertilizer microdosing technology was accompanied by a better uptake of nutrients by the roots, which resulted in negative nutrient balances. Indeed, overall, our results have shown that all fertilizer management strategies resulted in negative partial nutrient balances, with crop nutrient uptake exceeding additions through manure and/or fertilizers, except for N at the recommended fertilizer application rate across all manure strata. Since the balances were more negative in microdose plots than in the control ones, it appears that microdosing fertilizer allowed plants to better exploit the soil reserves, most likely by boosting root development as reported

153

Chapter 6. Conclusions and perspectives by Ibrahim et al. (2014, 2015c). Hence, the long-term application of fertilizer microdosing without other sources of nutrients may enhance the rate of soil degradation through nutrient mining in the long-term compared to unfertilized plots. Combining fertilizer microdosing with manure (as transported manure or through corralling) did not aggravate the negative balances compared to the unfertilized control, unlike what was observed for the NM strata. Hence the combined application of manure and fertilizer microdosing makes the latter more sustainable from a nutrient balance point of view, but particular attention should be paid to P and K. Since nutrient balance alone is not sufficient to evaluate sustainability, it needs to be linked with soil nutrient stocks, either with the total stock or with the stock of available nutrients. Based on this, we suggest that the results are meant to alert researchers, policy makers and other stakeholders, i.e., that nutrient mining under microdosing needs more attention. Recycling of crop residues and proper management of manure and urine which can limit the nutrient losses under corralling could, therefore, alleviate the potential soil mining effect of the fertilizer microdosing technology.

6.2.2 Variability in maize response and profitability following hill-placement of fertilizer and manure on farmers’ fields

Though Chapters 2 and 3 have consistently established the benefits of localized application of amendments and fertilizers when considering the average agronomic or economic performances, there is a growing concern that such average responses are insufficient to properly assess a technology considering the diversity of smallholder farming environments and practices. In reality, the larger the magnitude of the yield responses and the higher the probability of achieving a high response at a given site, the higher the visibility of the beneﬁt to farmers and the greater the chances for adoption of the technology. Our results show a strong variability in maize yields and response to microdose fertilization on farmers’ fields with a strong positive response at all sites. This contributes to the increasing recognition that measures of variability (such as the frequency and distribution of responses and economic benefits) are needed (Bielders and Gérard, 2015; Franke et al., 2016; Vanlauwe et al., 2016; Van Vugt et al., 2017). Overall, the causes of the large variability in response to fertilization were found to be related both to environmental conditions (e.g., rainfall, sowing date, soil) and crop management practices (e.g., antecedent management, weeding intensity). However, much of the variability remained unexplained (20%) and could be related to complex and non-linear interactions with biotic factors (such

154

6.2 Main findings as pests and diseases), climate factors (such as temperature, drought stress, etc) that are not or poorly taken into account by simple rainfall-related indices, edaphic factors (such as micronutrient deficiencies, soil structure) but also management (land preparation, weeding quality, etc...) and socio-economic characteristics (income, labour, training, size and intensity etc.) which were not well characterized here. As for the absolute yields, yield responses to microdose fertilization also varied widely. This variability is mainly due to the high variability of control yields resulting from variation in soil conditions and management which also resulted from variability in farmer’s resource endowments (e.g., Tittonell et al., 2008). Besides yields improvement and stability, understanding where the application of the technology is agronomically and economically appropriate are of paramount importance when defining the recommendation domain within a given agro- ecological zone. Should the microdosing technology be applied preferably to low or high fertility plots within a farm? Is microdosing technology economically viable and what is the level of the economic risk associated? Are there better, more profitable, or productive alternatives? This part of our research addressed these issues. As reported earlier also by Bielders and Gerard (2015), it has been observed in the present study that maize's yield response to microdose fertilization may be dependent on the plot's productivity level and could be well explained by the combination of some measured soil parameters (Clay/silt, total carbon, exch-Mg, pH) and weed pressure. Despite this variability observed in maize yields and response and by considering the additional labor costs related to microdose fertilization and hill- placed manure application, its use was economically profitable for a large proportion of farmers (net return about 4 times higher than the unfertilized control and more than 90% of the farmers achieved a VCR ≥ 2), demonstrating that there is considerable potential for smallholder farmers of northern Benin to increase both maize productivity and profitability by means of microdose fertilization. Furthermore, combining microdose fertilization with manure appeared still profitable and less risky (more than 80% of the farmers achieved a VCR ≥ 2). Thus, combined application of hill-placed manure and microdose fertilization (mainly the MD1 option) should be the preferred option as it is more agronomically viable (e.g., Ibrahim et al., 2016; Tovihoudji et al., 2017b) and less economically risky particularly for higher control yielding plots (>1200 kg ha-1) which exhibited generally low response to microdosing (Figure 4.4). Nevertheless, the reduction in input costs and increase in output prices appear to be the best scenario for farmers as it resulted to only 0 to 12% economic risk (i.e. VCR < 1) for the microdose fertilization (alone and with manure) against 37% risk for the RR treatment (Figure 4.6; Table 4.6).

155

Chapter 6. Conclusions and perspectives

6.2.3 Long-term scenario analysis regarding fertilizer microdosing

Following a first stage of calibration for broadcast fertilizer application, it was shown that the DSSAT model could adequately reproduce the maize phenology, growth and yields observed during the on-station experiments. To better take into account the effect of fertilizer microdosing on maize yields, the N stress coefficient was re-parameterized leading to satisfactory results. Using this validated model, it was shown that application of microdose alone (MD1 or MD2) resulted to a lower variability compared to the no fertilizer input (NF). Equally interesting, applying microdosing alone guarantees at least 1800 kg ha-1 every year without causing additional inter-annual variability. In smallholder subsistence farming the economic profitability and associated risk largely drive the adoption of a new technology. Our results showed that despite the higher yield stability, combining microdosing with manure increases the economic risk when considering a VCR of 4 (CIMMYT, 1988). Nevertheless, a VCR of 2 is reached in almost 100% of the years, indicating that the level of risk associated with manure + microdose remains acceptable. Since continuous use of sole fertilizer may enhance nutrient mining and soil acidification, complementing microdose fertilization with manure would be recommended to increase the sustainability of the system even though it may compromise the short-term benefits. Moreover, because the VCR values can depend greatly on fluctuations in input and output prices in the market within and between seasons (Chapters 3 and 4), applying the low microdosing rate (MD1) would be preferable for farmers irrespective of manure management.

6.3 Main implications and policy recommendations

The results of this thesis demonstrated that there is potential to improve livelihoods of the smallholder farmers in the study area through localized application of amendments. The results highlighted also the importance of adding organic amendments to enhance maize productivity and nutrient use efficiency (Chapters 2 and 3). In the study area, farmers (80%) have less than five heads of cattle and would have access to 2765±1827 kg ha-1 of manure (mean±standard deviation; data not shown) if all manure could be returned to the fields. In practice, incomplete collection of the manure and lack of means of transportation (carts)

156

6.3 Main implications and policy recommendations implies that many fields are left unmanured or insufficiently manured. In any case, the rate of 6 t manure ha-1 recommended by extension services seems highly unrealistic at present. Applying half the recommended rate of fertilizer (100 kg NPK ± 50 kg Urea ha-1) without manure, though economically viable, should not be recommended in the long-term. Also, the broadcast application of 3 t manure ha-1 is unlikely to substantially ameliorate soil quality. Hence, hill- placement of the manure appears to be a good alternative since it allows to substantially improve soil properties where it matters most, i.e., close to the plants. Given that most smallholder farmers cannot generate large quantities of manure due to the low number of livestock, relying on fertilizer to achieve acceptable yields (> 2000 kg ha-1) seems sensible. However, farmers should be encouraged to value the added biomass in order to produce more manure and gradually substitute fertilizer by manure or complement the fertilizer with manure. The gross margin is, however, even better for the 3M+50F treatment than for the 3M+NF, such that the former may be a suitable alternative in situations where money is not a constraint. For most subsistence maize growers who have access to only small quantities of fertilizer in the study area, the fertilizer microdosing technique appears to be particularly interesting since it promotes a small realistic fertilizer amount and makes a more efficient use of nutrients compared to the recommended fertilizer rate. It may be used to increase yield and improve farmer’s income. In practice, the fertilizer amount (NPK + urea) applied by farmers is nearly equivalent to the microdose fertilizer rates tested in this study (94-156 kg fertilizer ha-1). Although the response to fertilizer microdosing was best in no manure fields, fertilizer microdosing without manure application (or other sources of organic amendments) should not be recommended as a viable option for sustainable maize production in the long-term from a nutrient balance point of view. Currently, farmers export crop residues to be used as feed for livestock. It is essential to recycle some of this biomass and return it to the fields as manure, compost or through corralling. For those farmers who have access to little manure, hill-placement of a realistic amount of transported manure and microdosing fertilizer could be a viable option for sustainable maize production. For farmers who have the possibility to establish a stubble-grazing contract with Fulani herders, it may be of interest to combine corralling with fertilizer microdosing only as from the second or third year to achieve a higher fertilizer use efficiency. With the microdosing technology, there is a potential opportunity for smallholder maize farmers in northern Benin, who are constrained in their capacity to invest in external inputs, to increase maize productivity and resource use efficiency, close to the target yield of 3 t ha−1 set towards achieving the

157

Chapter 6. Conclusions and perspectives

African Green Revolution (Sánchez, 2010). While there are concerns regarding the extra labor associated with this technology, the present study established that fertilizer microdosing alone (preferably the MD1 option) or combined with hill- placed manure generally result in larger agronomic efficiency and economic returns and lesser risk compared to the conventional recommended fertilizer rate treatment. In addition, since manure/fertilizer application is done after sowing and the first weeding period - a period of greater labor availability - it does not interfere with crop sowing, which is one of the critical labor bottlenecks in the study area. From our results, it appears clearly that on low productive fields, i.e, where low yields are expected as a result of, e.g., low soil fertility or late sowing, microdose fertilization could prove to be a motivating cost-effective strategy. Smallholder farmers would be well recommended to apply microdosing in priority to low productive fields or plots. Adequate institutional support will be required particularly to develop labor-reducing equipment, make mineral fertilizer affordable and support the internal maize market, which may allow fertilizer microdosing use to remain highly profitable and further motivate farmers to use this technology. These two latter actions seem to be more important in a context of unstable input and output markets which lead to reduction in prices of grain and an increase in fertilizer and labor costs and thereby to an increased economic risk. Before widely promoting fertilizer microdosing in smallholder maize farming systems, one needs to evaluate two main issues. Firstly, the agronomic and economic benefits of the technique have to be demonstrated under on-farm conditions for a suitable range of environmental conditions. Furthermore, the associated short-term risks have to be low. These issues have been addressed through short-term on-station and on-farm experiments (Chapters 2, 3 and 4). Secondly, given that smallholder farmers are usually risk-averse, the microdose technology should not substantially increase inter-annual variability in yield and income caused by irregular rainfall distribution. Assessing the temporal yield stability, the economic risk and the capacity to achieve a minimum grain yield each year are therefore a prerequisite before embarking on large scale extension endeavor. As long-term field experimentation is expensive and time consuming, the use of calibrated and validated decision support systems like DSSAT and APSIM constituted an attractive alternative to assess the long-term variability in maize yields following fertilizer microdosing. The DSSAT modelling deployed in this thesis yielded a decision support system that can be used as a basis to develop site-specific fertilizer microdosing recommendation for maize production in northern Benin (Chapter 5). Combinations of treatments involving a modest fertilizer microdose rate (24-34 kg N ha-1) and manure rate (1-3 t ha-1) improves both the long-term average and the minimum guaranteed yield without increasing

158

6.3 Main implications and policy recommendations inter-annual variability compared to no N input and high rates of mineral fertilizer use. Such combined application appears more appropriate and sustainable in the long run for smallholder farmers, as it guarantees higher minimum yield (at least 1800 kg ha-1) and maximum return to investment irrespective of the years, thereby reducing their vulnerability. The research activities of this thesis were conducted in agro-ecological zones (AEZ) that cut across several countries in Sub-Sahara West Africa (SSWA). These AEZs include the Southern Guinean Savanna and Forest transition zones, which demarcate the maize belt of West Africa across Nigeria, Benin, Togo, Ghana, Côte d’Ivoire, Burkina-Faso, Mali, Guinea, Senegal and Gambia. This will facilitate the extrapolation of the results beyond the study sites and within the same AEZs. This extrapolation beyond the study zone will, hoxever, require further validation trials across a larger range of soils, climatic conditions, and maize cultivars.

6.4 Perspectives for futur research

Two main issues related to the prediction of maize growth and yields under microdose fertilization when using both linear mixed model (with data collected through on-farm demonstration trials) and dynamic crop models such as DSSAT have to be addressed in the future research.

1) How to improve the relevance of the recommendations regarding microdose fertilization when using linear mixed model? The study presented in Chapter 4 enabled the analysis of yield data collected from a range of environmental and management conditions which are representative for smallholder farmers’ conditions. Nevertheless, the approach also resulted in unbalanced data which can limit the number of variables that we could include in statistical analyses. Furthermore, many of the collected explanatory variables were often confounded, which may reduce the statistical power of the combined model and the capability to explain variability. Possibly, some potentially important variables and non-linear interactions that could contribute to increase the percentage of variability explained such as plant growth and nutrient data, temperature, drought stress at farm level, pest and disease incidence and severity, livestock damage, socio-economic profiles of farmers, and also differences in some management factors (e.g. land preparation, seed establishment, etc...) were not well apprehended. The limitations underlined in this study reflect a trade-off between on-farm evaluation of the performance of technologies under farmers’ conditions, and understanding the effects of particular factors on yield. Other

159

Chapter 6. Conclusions and perspectives on-farm trials over seasons and for new districts (larger sample size for a larger number of explanatory variables), would be most useful to validate the robustness of the recommendations through the cross-validation of model results (e.g. Ronner et al., 2016). However, though useful, replicating on-farm experiments over several locations and seasons while keeping more detailed monitoring is laborious and expensive and hardly ever feasible within the usual financial and organizational constraints imposed by donors.

2) How to improve the prediction of dynamic crop models such as DSSAT under microdose fertilization and the relevance of the recommendations? The present study provided empirical foundations of maize growth modelling under microdose fertilization using DSSAT CERES-Maize model based on N demand optimization approach. While the proposed model shows some promise for predicting optimum combinations of mineral N fertilizer and manure microdosing, it was not directly calibrated for P and K supply. Strictly speaking, the calibration process does not reflect the response to single elements, as all three nutrients are present in both the fertilizer and the manure. The calibration and validation results should, therefore, be interpreted as the response of maize to N inputs in the presence of proportional inputs of N-P and N-K. In situations where phosphorus is a limiting factor, accurate simulation of the response to fertilizer microdosing will also depend on an accurate representation of the soil P and its simulation by the model. The fact that the modelling approach did not consider the maize response to P (a main limiting factor of crop production in SSA in addition to N) may have resulted in the fact that it could not simulate accurately the maize response to high amounts of nutrients especially from combined manure and fertilizer. A soil-P module has been added recently in the DSSAT modelling framework and responds reasonably for maize (Dzotsi et al., 2010) but it needs accurate estimation of initial soil P pools sizes (inorganic and organic) with sufficient measures of soil and plant P characteristics. Therefore, there is a need for further research that should consider N and P simultaneously using experiments with different N and P combinations, specific characterization of soil and plant P and more detailed measurement of N and P dynamics in the soil and plant. Furthermore, although our modelling approach allowed to identify promising soil fertility management practices under variable weather conditions, we acknowledge that 2-year field experiments are insufficient to properly validate the model. Therefore, an extended calibration period, larger than two years, would warrant a better model calibration. Another future research should consider also the improvement of the sensitivity of the model to depth of

160

6.4 Perspectives for futur research fertilizer application, and further evaluation of the model for interactions between water and nutrients (N and P). There is also the need for more calibration and evaluation using additional data sets from different sites and more contrasted seasons to conﬁrm its applicability in the low-input cropping systems of SSA. The factors and variables considered in the seasonal analysis in this study also represented simplified conditions. For example, other crop management variables such as weed and pest management and tillage were assumed to be fixed and in standard conditions. For the same reason, the model considered each year’s simulation as independent of the other years since soil initial conditions were reinitialized at each planting. However, these conditions might vary over time and space. In the study area, most farmers grow maize continuously in the same plot for several years or in association/rotations with other crops and thus microdosing application may be applied continuously in the same plot. So, considering these conditions through a Sequential Analysis in DSSAT model can be a potential subject matter for a separate future study. In our study we used a non-mechanistic method to model maize response to microdose fertilization. This was based on several assumptions and simplifications that may affect the reliability of our predictions. In the future, it may be worthwhile to investigate in more detail the physiological response mechanisms of maize to fertilizer microdozing and how to best represent them in 1-D soil-plant-atmosphere models such as DSSAT in the physiologically dynamic manner. Also models that include the root physiological response to microdosing or the livestock/whole farm component are also essential. For this goal, the DSSAT model can be coupled with a 3D model for soil-root interactions such as R-SWMS (Javaux et al., 2008), and whole farm system models (such as the NUANCES framework using the FARMSIM model (Giller et al., 2006; Tittonell et al., 2007) for robust fertilizer microdosing recommendations. Another new research agenda may focus on the development of decision support tools regarding fertilizer microdosing under climate change scenarios. Climate variability and change could affect fertilizer microdosing performances and decision support system can contribute directly to the identification of more resilient coping and adaptive strategies.

161

References

Abdoulaye, T. and Sanders, J.H. 2005. Stages and determinants of fertilizer use in semiarid African agriculture: the Niger experience. Agricultural Economics, 32: 167-179. Adams, A.M., Gillespie, A.W., Kar, G., Koala, S., Ouattara, B., Kimaro, A.A., Peak, D. 2016. Long-term effects of reduced fertilizer rates on millet yields and soil properties in the West-African Sahel. Nutrient Cycling in Agroecosystems, 106:17-29. Adégbola, P.Y., Adékambi, S.A. 2006. Analyse de l’impact socio-économique de l ‘adoption des technologies nouvelles de gestion de fertilité du sol. Rapport de la phase exploratoire. PAPA/INRAB, p31. Adjei-Nsiah, S., Leeuwis, C., Sakyi-Dawson, O., Giller, K.E., Kuyper, T.W. 2007. Exploring diversity among farmers for orienting inter-disciplinary action research on cropping system management in Wenchi, Ghana: the significance of time horizons. International Journal of Agricultural Sustainability, 5(2-3): 176-194. Adnan, A.A., Jibrin, J.M., Kamara, A.Y., Abdulrahman, B.L., Shaibu, A.S. 2017a. Using CERES-Maize model to determine the nitrogen fertilization requirements of early maturing maize in the Sudan Savanna of Nigeria. Journal of Plant Nutrition, 40:1066–1082. Adnan, A.A., Jibrin, J.M., Abdulrahman, B.L., Shaibu, A.S., Garba, I.I. 2017b. CERES–Maize Model for Determining the Optimum Planting Dates of Early Maturing Maize Varieties in Northern Nigeria. Frontiers in Plant Science, 8: 1118. doi: 10.3389/fpls.2017.01118. Affholder, F., Poeydebat, C., Corbeels, M., Scopel, E., & Tittonell, P. (2013). The yield gap of major food crops in family agriculture in the tropics: Assessment and analysis through field surveys and modelling. Field Crops Research, 143, 106–118. Agegnehu, G., Nelson, P.N., Bird, M.I. 2016. Crop yield, plant nutrient uptake and soil physicochemical properties under organic soil amendments and nitrogen fertilization on Nitisols. Soil & Tillage Research, 160:1–13. Akossou, A.Y., Attakpa, E.Y., Fonton, N.H., Sinsin, B., Bosma, R.H. 2016. Spatial and temporal analysis of maize (Zea mays) crop yields in Benin from 1987 to 2007. Agricultural and Forest Meteorology, 220:177-189.

163

References

Akponikpè, P. B. I., Gerard, B., Bielders, C.L. 2014. Soil water crop modeling for decision support in millet-based systems in the Sahel: a challenge. African Journal of Agricultural Research, 9(22):1700-1713. Akponikpè, P.B.I., Gerard, B., Michels, K., Bielders, C.L. 2010. Use of the APSIM model in long-term simulation to support decision making regarding nitrogen management for pearl millet in the Sahel. European Journal of Agronomy, 32: 144-154. Akponikpè, P.B.I., Michels, K., Bielders, C.L. 2008. Integrated Nutrient Management of Pearl Millet in the Sahel Combining Cattle Manure, Crop Residue and Mineral Fertilizer. Experimental Agriculture, 44:453–472. Akponikpè, P.B.I., Minet, J., Gerard, B., Defourny, P., Bielders, C.L. 2011. Spatial fields' dispersion as a farmer strategy to reduce agro-climatic risk at the household level in pearl millet-based systems in the Sahel: A modeling perspective. Agricultural and Forest Meteorology, 151: 215-227. Amouzou, K.A., Naab, J.B., Lamers, J.P., Becker, M. 2018. CERES-Maize and CERES-Sorghum for modeling growth, nitrogen and phosphorus uptake, and soil moisture dynamics in the dry savanna of West Africa. Field Crops Research, 217:134-149. Andrieu, N., Vayssières, J., Corbeels, M., Blanchard, M., Vall, E., Tittonell, P., 2014. From farm scale synergies to village scale trade-offs: Cereal crop residues use in an agro-pastoral system of the Sudanian zone of Burkina Faso. Agricultural Systems, 134 :84–96. ANOPER Bénin, 2014. La situation actuelle de l’élevage et des éleveurs au Bénin : Analyse et perspectives. Annexe du document d’orientation stratégique de l’ANOPER. 68p. Anothai, J., Soler, C.M.T., Green, A., Trout, T.J., Hoogenboom, G. 2013. Evaluation of two evapotranspiration approaches simulated with the CSM– CERES–Maize model under different irrigation strategies and the impact on maize growth, development and soil moisture content for semi-arid conditions. Agricultural and Forest Meteorology, 176:64–76 Aune, J.B, Doumbia, M., Berthe, A. 2007. Microfertilizing sorghum and pearl millet in Mali – agronomic, economic and social feasibility. Outlook on Agriculture, 36: 199–203. Aune, J.B., Bationo, A. 2008. Agricultural intensification in the Sahel–the ladder approach. Agricultural Systems, 98(2): 119-125. Aune, J.B., Coulibaly, A. 2015. Microdosing of Mineral Fertilizer and Conservation Agriculture for Sustainable Agricultural Intensification in Sub-Saharan Africa. In Sustainable Intensification to Advance Food Security and Enhance Climate Resilience in Africa (pp. 223-234). Springer International Publishing.

164

References

Aune, J.B., Traore, C.O., Mamadou, S. 2012. Low-cost technologies for improved productivity of dryland farming in Mali. Outlook on Agriculture, 41(2): 103-108. Bado, B.V., Bationo, A., Lompo, F., Traore, K., Sedogo, M.P., Cescas, M.P. 2012. Long-term effects of crop rotations with fallow or groundnut on soil fertility and succeeding sorghum yields in the Guinea savannah of West Africa. In: Bationo, A., Waswa, B., Kihara, J., Adolwa, I., Vanlauwe, B., Saidou, K. (Eds.), Lessons learned from long-term soil fertility management experiments in Africa. Springer, Netherlands, pp. 27–40. Bagayoko, M., Maman, N., Palé, S., Sirifi, S., Taonda, S.J.B., Traore, S., Mason, S.C., 2011. Microdose and N and P fertilizer application rates for pearl millet in West Africa. African Journal of Agricultural Research, 6: 1141-1150. Barrett, C.B., Bevis, L.E.M., 2015. The self-reinforcing feedback between low soil fertility and chronic poverty. Nature Geoscience, 8: 907–912. Bationo, A., Buerkert, A. 2001. Soil organic carbon management for sustainable land use in Sudano-Sahelian West Africa. Nutrient Cycling in Agroecosystems, 61: 131-142. Bationo, A., Mokwunye, A.U. 1991. Role of manures and crop residue in alleviating soil fertility constraints to crop production: With special reference to the Sahelian and Sudanian zones of West Africa. Nutrient Cycling in Agroecosystems, 29: 117-125. Bationo, A., Waswa, B. 2011. New challenges and opportunities for integrated soil fertility management in Africa. In: Bationo A, Waswa BS, Okeyo JM, Kihara JM, Maina F (eds). Innovations as Key to the Green Revolution in Africa. Springer, Dordrecht, pp 3–17 Bationo, A., Hartemink, A., Lungu, O., Naimi, M., Pkth, P., Smaling, E., Thiombiano, L., Waswa, B., 2012. Knowing the African soils to improve fertilizer recommendations. In: Kihara, J., Fatonji, D., Jones, J.W., Hoogenboom, G., Tabo, R., Bationo, A. (Eds.), Improving soil fertility recommendations in Africa using the Decision Support System for Agrotechnology Transfer (DSSAT). Springer, Netherlands, pp. 19–42. Bationo, A., Kihara, J., Vanlauwe, B., Waswa, B., Kimetu, J. 2007. Soil organic carbon dynamics, functions and management in West African agroecosystems. Agricultural systems, 94: 13-25. Bationo, A., Ndjeunga, J., Bielders, C.L., Prabhakar, V.R., Buerkert, A., Koala, S. 1998. Soil fertility restoration options to enhance millet productivity on sandy Sahe-lian soils in south–west Niger. In: Lawrence, P., Renard, G., Von Oppen, M. (Eds.), Evaluation of Technical and Institutional Options for Small Farmers in WestAfrica. Margraf Verlag, Weikersheim, Germany, pp. 93–104.

165

References

Bationo, A., Tabo, R., Kihara, J., Kimetu, J., Adamou, A., Koala, S. 2005. Farming inthe drylands of West Africa: promising soil fertility restauration techniques. In: Omanya, G.O., Pasternak, D. (Eds.), Sustainable Agriculture Systems for theDrylands. Proceedings of the International Symposium for Sustainable Dryland Agriculture Systems, 2–5 December 2003, Niamey, Niger, ICRISAT, pp. 100–125. Bationo, A., Wani, S.P., Bielders, C.L., Vlek, P.L.G., Mokwunye, A.U. 2000. Crop residue management to improve soil organic carbon content, soil quality and productivity in the desert margins of West Africa. In: Lal, R., Kimble, J.M., Stewart, B.A. (Eds.), Global climate change and tropical ecosystems. CRC Press LLC, Florida, pp. 117–145. Bedada, W., Karltun, E., Lemenih, M., Tolera, M. 2014. Long-term addition of compost and NP fertilizer increases crop yield and improves soil quality in experiments on smallholder farms. Agriculture, ecosystems and environment, 195:193–201 Bielders, C.L., Gérard, B. 2015. Millet response to microdose fertilization in south–western Niger: effect of antecedent fertility management and environmental factors. Field Crops Research, 171: 165-175. Blackshaw, R.E., Molnar, L.J., Janzen, H.H. 2004. Nitrogen fertilizer timing and application method affect weed growth and competition with spring wheat. Weed Science, 52: 614-622. Bouman, B.A.M., van Laar, H.H., 2006. Description and evaluation of the rice growth model ORYZA2000 under nitrogen-limited conditions. Agricultural Systems, 87(3) : 249–273. Brisson, N., Mary, B., Ripoche, D., Jeuffroy, M.H., Ruget, F., Nicoullaud, B., Gate, P., Devienne-Barret, F., Antonioletti, R., Durr, C., Richard, G., Beaudoin, N., Recous, S., Tayot, X., Plenet, D., Cellier, P., Machet, J.-M., Meynard, J.M., Delécolle, R. 1998. STICS: a generic model for the simulation of crops and their water and nitrogen balances. I. Theory and parameterization applied to wheat and corn. Agronomie, 18 (5–6): 311–346. Brouwer, J., Powell, J. M. 1998. Increasing nutrient use efficiency in West-African agriculture: the impact of micro-topography on nutrient leaching from cattle and sheep manure. Agriculture, ecosystems and environment, 71(1-3): 229- 239. Buerkert, A., Bationo, A., Piepho, H. P. 2001. Efficient phosphorus application strategies for increased crop production in sub-Saharan West Africa. Field Crops Research, 72(1): 1-15. Buerkert, A., Schlecht, E. 2013. Agricultural innovations in smallholder farming systems of Sudano-Sahelian West Africa: Some prerequisites for success. Science et changements planétaires/Sécheresse, 24(4) : 322-329.

166

References

Camara, B.S., Camara, F., Berthe, A., Oswald, A. 2013. Microdosing of fertilizer - a technology for farmers' needs and resources. International Journal of AgriScience, 3: 387–399. Carsky, R.J., Becker, M., Hauser, S. 2001. Mucuna cover crop fallow systems: potential and limitations. In: Tian, G., o Ishida, F. and Keatinge, D. (Eds.), Sustaining Soil Fertility in West Africa. Soil Science Society of America and American Society of Agronomy, Special Publication, 58: 111-135. Castellanos-Navarrete, A., Tittonell, P., Rufino, M. C., Giller, K. E. 2015. Feeding, crop residue and manure management for integrated soil fertility management–A case study from Kenya. Agricultural Systems, 134: 24-35. Chianu, J.N., Chianu, J.N., Mairura, F. 2012. Mineral fertilizers in the farming systems of sub-Saharan Africa. A review. Agronomy for Sustainable Development, 32:545–566. Chikowo, R., Zingore, S., Snapp, S., Johnston, A., 2014. Farm typologies, soil fertility variability and nutrient management in smallholder farming in Sub- Saharan Africa. Nutrient cycling in agroecosystems, 100(1): 1-18. Chivenge, P., Vanlauwe, B., Six, J. 2011. Does the combined application of organic and mineral nutrient sources influence maize productivity? A meta- analysis. Plant and Soil, 342:1–30. CIMMYT, 1988. From Agronomic Data to Farmer Recommendations: An Economics Training Manual. CIMMYT, Mexico, D.F. Comoé, H. and Siegrist, M. 2015. Relevant drivers of farmers’ decision behavior regarding their adaptation to climate change: a case study of two regions in Côte d’Ivoire. Mitigation and adaptation strategies for global change, 20(2): 179- 199. Cooper, P.J.M., Dimes, J., Rao, K.P.C., Shapiro, B., Shiferaw, B., Twomlow, S. 2008. Coping better with current climatic variability in the rain-fed farming systems of sub-Saharan Africa: An essential first step in adapting to future climate change?. Agriculture, Ecosystems and Environment, 126(1-2): 24-35. Corbeels, M., Chirat, G., Messad, S., Thierfelder, C. 2016. Performance and sensitivity of the DSSAT crop growth model in simulating maize yield under conservation agriculture. European journal of agronomy, 76: 41-53. Dagbénonbakin, G.D. 2005. Productivity and water use efficiency of important crops in the Upper Oueme catchment: influence of nutrient limitations, nutrient balances and soil fertility. Dissertation, University of Bonn. 212p. de Ridder, N., Breman, H., Van Keulen, H., Stomph, T.J. 2004. Revisiting a ‘cure against land hunger’: soil fertility management and farming systems dynamics in the West African Sahel. Agricultural Systems, 80: 109–131.

167

References de Rouw, A., Rajot, J.-L. 2004. Nutrient availability and pearl millet production in Sahelian farming systems based on manuring or fallowing. Agriculture, ecosystems and environment, 104: 249-262. Delve, R.J., Cadisch, G., Tanner, J.C., Thorpe, W., Thorne, P.J., Giller, K.E. 2001. Implications of livestock feeding management on soil fertility in the smallholder farming systems of sub-Saharan Africa. Agriculture, ecosystems and environment, 84(3): 227-243. Diogo, R.V., Schlecht, E., Buerkert, A., Rufino, M.C., van Wijk, M.T. 2013. Increasing nutrient use efficiency through improved feeding and manure management in urban and peri-urban livestock units of a West African city: A scenario analysis. Agricultural systems, 114: 64-72. DPP/MAEP (Direction de la Planification et de la Prospective/Ministère de l’Agriculture, de l’Elevage et de la Pêche), 2010. Campagnes agricoles 2008 et 2009, Cotonou, Bénin. 90 p. Dzotsi, K.A., Jones, J.W., Adiku, S.G.K., Naab, J.B., Singh, U., Porter, C.H., Gijsman, A.J. 2010. Modeling soil and plant phosphorus within DSSAT. Ecological Modelling, 221:2839–2849 Eyshi Rezaei, E., Gaiser, T., Siebert, S., Sultan, B., Ewert, F. 2014. Combined impacts of climate and nutrient fertilization on yields of pearl millet in Niger. European journal of agronomy, 55:77–88. Falconnier, G.N., Descheemaeker, K., Van Mourik, T.A., Giller, K.E. 2016. Unravelling the causes of variability in crop yields and treatment responses for better tailoring of options for sustainable intensification in southern Mali. Field crops research, 187: 113-126. Falconnier, G.N., Descheemaeker, K., van Mourik, T.A., Sanogo, O.M., Giller, K.E. 2015. Understanding farm trajectories and development pathways: Two decades of change in southern Mali. Agricultural Systems, 139: 210–222. FAOSTAT, 2016. FAO Statistical Database: Production and Trade. Online: http://www.fao.org/faostat/en/#data/QC Fatondji, D. 2002. Organic amendment decomposition, nutrient release and nutrient uptake by millet (Pennisetum glaucum) in a traditional land rehabilitation technique (zaï) in the Sahel. PhD Thesis. Centre for Development Research, University of Bonn, Ecological and Development Series. Fatondji, D., Martius, C., Bielders, C.L., Vlek, P.L.G., Bationo, A., Gerard, B. 2006. Effect of planting technique and amendment type on pearl millet yield , nutrient uptake , and water use on degraded land in Niger. Nutrient Cycling in Agroecosystems, 76:203–217.

168

References

Fatondji, D., Martius, C., Zougmore, R., Vlek, P.L.G., Bielders, C.L., Koala, S. 2009. Decomposition of organic amendment and nutrient release under the zai technique in the Sahel. Nutrient Cycling in Agroecosystems, 85:225–239. Folberth, C., Yang, H., Gaiser, T., Abbaspour, K.C., Schulin, R. 2013. Modeling maize yield responses to improvement in nutrient, water and cultivar inputs in sub-Saharan Africa. Agricultural systems, 119: 22–34. Fosu, M., Buah, S.S., Kanton, R. a. L., Agyare, W.A., 2012. Modeling maize response to mineral fertilizer on silty clay loam in the Northern Savanna Zone of Ghana using DSSAT model. In: Kihara, J., Fatondji, D., Jones, J.W., Hoogenboom, G., Tabo, R., Bationo, A. (Eds.), Improving Soil Fertility Recommendations in Africa Using the Decision Support System for Agrotechnology Transfer (DSSAT). Springer, Netherlands, pp. 157– 168. Fosu-Mensah, B.Y., MacCarthy, D.S., Vlek, P.L.G., Safo, E.Y. 2012. Simulating impact of seasonal climatic variation on the response of maize (Zea mays L.) to inorganic fertilizer in sub-humid Ghana. Nutrient cycling in agroecosystems, 94(2-3): 255-271. Franke, A.C., Baijukya, F., Kantengwa, S., Reckling, M., Vanlauwe, B., Giller, K.E. 2016. Poor farmers–poor yields: Socio-economic, soil fertility and crop management indicators affecting climbing bean productivity in northern Rwanda. Experimental Agriculture, 1-21. Doi: 10.1017/S0014479716000028 Gaiser, T., Judex, M., Igué, A.M., Paeth, H., Hiepe, C. 2011. Future productivity of fallow systems in Sub-Saharan Africa: Is the effect of demographic pressure and fallow reduction more significant than climate change? Agricultural and Forest Meteorology, 151(8): 1120–1130. Gandah, M., Bouma, J., Brouwer, J., Hiernaux, P., Van Duivenbooden, N. 2003. Strategies to optimize allocation of limited nutrients to sandy soils of the Sahel: a case study from Niger, West Africa. Agriculture, ecosystems and environment, 94: 311-319. Gee, G.W., Or, D., 2002. Particle-size analysis. In: Dane, J.H., Topp, G.C. (Eds.), Methods of soil analysis: Part 4 Physical methods. Soil Science Society of America, Madison, WI, pp. 255–293. GENSTAT, 2009. Genstat. Lawes Agricultural Trust (Rothamsted Experimental Station), Rothamsted, UK. Gijsman, A.J., Hoogenboom, G., Parton, W.J., Kerridge, P.C. 2002. Modifying DSSAT crop models for low-input agricultural systems using a soil organic matter–residue module from CENTURY. Agronomy Journal, 94: 462–474.

169

References

Giller, K. E., Rowe, E. C., de Ridder, N., van Keulen, H. 2006. Resource use dynamics and interactions in the tropics: Scaling up in space and time. Agricultural Systems, 88:8-27. Giller, K.E., Tittonell, P., Rufino, M.C., van Wijk, M.T., Zingoré, S., Mapfumo, P., Adjei-Nsiah, S., Herrero, M., Chikowo, R., Corbeels, M., Rowe, E.C., Baijukya, F., Mwijage, A., Smith, J., Yeboah, E., van der Burg, W.J., Sanogo, O.M., Misiko, M., de Ridder, N., Karanja, S., Kaizzi, C., K’ungu, J., Mwale, M., Nwaga, D., Pacini, C., Vanlauwe, B. 2011. Communicating complexity: integrated assessment of trade-offs concerning soil fertility management within African farming systems to support innovation and development. Agricultural Systems, 104, 191–203. Gnanglè, C.P., Glèlè Kakaï, R., Assogbadjo, A.E., Vodounnon, S., Yabi, J.A., Sokpon, N. 2011. Tendances climatiques passées, modélisation, perceptions et adaptations locales au Bénin. Climatologie, 8: 27–40. Guan, K., Sultan, B., Biasutti, M., Baron, C., Lobell, D.B. 2017. Assessing climate adaptation options and uncertainties for cereal systems in West Africa. Agricultural and Forest Meteorology, 232: 291-305. Guertal, E.A., Raun, W.R., Westerman, R.L., Boman, R.K., 1994. Applications of stability analysis for single-site, long-term experiments. Agronomy Journal, 86: 1016–1019. Harris, F. 2002. Managment of manure in farming systems in Semi-Arid West Africa. Experimental Agriculture, 38: 131-148. Hayashi, K., Abdoulaye, T., Gerard, B., Bationo, A. 2008. Evaluation of application timing in fertilizer microdosing technology on millet production in Niger, West Africa. Nutrient Cycling in Agroecosystems, 80: 257- 265. Hazelton, P.A., Murphy, B.W. 2007. Interpreting Soil Test Results: What Do All the Numbers Mean. CSIRO Publishing, Melbourne. Henao, J. and Baanante, C. 2006. Agricultural production and soil nutrient mining in Africa: implications for resource conservation and policy development. International Fertilizer Development Center (IFDC), Muscle Shoals, Alabama Hodge, A. 2004. The plastic plant: root responses to heterogeneous supplies of nutrients. New Phytologist, 162: 9–34. Holzworth, D.P., Huth, N.I., Zurcher, E.J., Herrmann, N.I., McLean, G., Chenu, K., … Brown, H. 2014. APSIM–evolution towards a new generation of agricultural systems simulation. Environmental Modelling and Software, 6: 327- 350.

170

References

Honfoga, B.G. 2012. Segmentation du marché d’engrais minéraux pour répondre aux besoins des cotonculteurs au Bénin. Bulletin de la Recherche Agronomique du Bénin (BRAB), Numéro spécial Coton, 12-25. Hoogenboom, G., Jones, J.W., Wilkens, P.W., Porter, C.H., Boote, K.J., Hunt, L.A. 2015. Decision Support System For Agrotechnology Transfer (DSSAT). Version 4.6. DSSAT Foundation, Prosser, WA. Houba, V., Van der Lee, J., Novozamsky, I., 1995. Soil Analysis Procedures; Other Procedures (Soil and Plant Analysis, Part 5B). Department of Soil Science and Plant Nutrition, Wageningen Agricultural University, Wageningen, Netherlands, 217p. Ibrahim, A., Pasternak, D., Fatondji, D., 2014. Impact of depth of placement of mineral fertilizer microdosing on growth, yield and partial nutrient balance in pearl millet cropping system in the Sahel. Journal of Agricultural Science, 153: 1412–1421. Ibrahim, A., Abaidoo, R.C., Fatondji, D., Opoku, A. 2015a. Hill placement of manure and fertilizer microdosing improves yield and water use efficiency in the Sahelian low input millet-based cropping system. Field Crops Research, 180:29–36. Ibrahim, A., Abaidoo, R.C., Fatondji, D., Opoku, A. 2015b. Integrated use of fertilizer microdosing and Acacia tumida mulching increases millet yield and water use efficiency in Sahelian semi-arid environment. Nutrient Cycling in Agroecosystems, 103(3): 375-388. Ibrahim, A., Abaidoo, R.C, Fatondji, D., Opoku A., 2015c. Determinants of fertilizer microdosing-induced yield increment of pearl millet on an acid sandy soil. Experimental Agriculture, 52: 562–578. Ibrahim, A., Abaidoo, R.C., Fatondji, D., Opoku, A. 2016. Fertilizer microdosing increases crop yield in the Sahelian low-input cropping system: A success with a shadow. Soil science and plant nutrition, 62(3): 277-288. Igué, A.M., Adjanohoun, A., Saidou, A., Ezui, G., Attiogbe, P., Kpagbin, G., Gotoechan-Hodonou, H., Youl, S., Pare, T., Balogoun, I., Ouedraogo, J., Dossa, E., Mando, A., Sogbedji, J.M. 2013. Application et adaptation de l’approche intégrée DSSAT-SIG à la formulation des doses d’engrais pour la culture du maïs au Sud et au centre du Benin. Bulletin de la Recherche Agronomique du Bénin (BRAB), Numéro spécial Fertilité du maïs, pp 24–33. INSAE. 2013. Recensement Général de la Population et de l’Habitat. Résultats provisoires du RGPH4, MPPD, Cotonou, Bénin, 7 p. IPCC. 2007. Summary for policymakers. In M. L. Parry, O. F. Canziani, J. P. Palutikof, P. J. van der Linden, & C. E. Hanson (Eds.), Climate Change 2007 : Impacts, Adaptation and Vulnerability. Contribution of Working

171

References

Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. Godwin, D.C., Singh, U. 1998. Nitrogen balance and crop response to nitrogen in upland and lowland cropping systems. In: Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (Eds.), Understanding Options for Agricultural Production, Systems Approaches for Sustainable Agricultural Development. Springer, Netherlands, pp. 55–77. Jamieson, P.D., Semenov, M.A., Brooking, I.R., Francis, G.S., 1998. Sirius: a mechanistic model of wheat response to environmental variation. European Journal of Agronomy, 8 (3–4): 161–179. Jamil, M., Charnikhova, T., Jamil, T., Ali, Z., Mohamed, N.E.M.A., Van Mourik, T., Bouwmeester, H.J. 2014. Influence of fertilizer microdosing on strigolactone production and Striga hermonthica parasitism in pearl millet. International Journal of Agriculture and Biology, 16: 935‒940. Jamil, M., van Mourik, T., van Charnikhova, T., Bouwmeester, H.J. 2012. Effect of diammonium phosphate application on strigolactone production and Striga hermonthica infection in three sorghum cultivars. Weed Research, 53:121–130. Javaux, M., Schroder, T., Vanderborght, J., Vereecken, H. 2008. Use of a three- dimensional detailed modeling approach for predicting root water uptake. Vadose Zone Journal, 7: 1079-1088. Jing, J., Rui, Y., Zhang, F., Rengel, Z., Shen, J. 2010. Localized application of phosphorus and ammonium improves growth of maize seedlings by stimulating root proliferation and rhizosphere acidification. Field Crops Research, 119: 355–364. Jing, J., Zhang, F., Rengel, Z., Shen, J. 2012. Localized fertilization with P plus N elicits an ammonium-dependent enhancement of maize root growth and nutrient uptake. Field Crops Research, 133: 176-185. Jones, C.A., Kiniry, J.R., 1986. CERES-Maize. A Simulation Model of Maize Growth and Development. Texas A&M Univ. Press Jones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Batchelor, L. Hunt, P.W. Wilkens, U. Singh, A.J. Gijsman, J.T. Ritchie. 2003. The DSSAT cropping system model. European Journal of Agronomy, 18 (3-4): 235–265 Kafesu, N., Chikowo, R., Mazarura, U., Gwenzi, W., Snapp, S., Zingore, S., 2018. Comparative fertilization effects on maize productivity under conservation and conventional tillage on sandy soils in a smallholder cropping system in Zimbabwe. Field Crops Research, 218: 106-114. Kamanga, B.C.G., Waddington, S.R., Whitbread, A.M., Almekinders, C.J.M., Giller, K.E. 2013. Improving the efficiency of use of small amounts of

172

References

nitrogen and phosphorus fertiliser on smallholder maize in central Malawi. Experimental Agriculture, 50(2): 229-249. Kayombo, B., Lal, R. 1993. Tillage systems and soil compaction in Africa, Soil Tillage Research, 27: 35–72. Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J., Holzworth, D., Huth, N.I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., Mclean, G., Verburg, K., Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L., Asseng, S., Chapman, S., Mccown, R.L., Freebairn, D.M., Smith, C.J. 2003. An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy, 18:267-288. Kihara, J., Bationo, A., Mugendi, D.N., Martius, C., Vlek, P.L. 2011. Conservation tillage, local organic resources and nitrogen fertilizer combinations affect maize productivity, soil structure and nutrient balances in semi-arid Kenya. Nutrient Cycling in Agroecosystems, 90:213–225. Kihara, J., Huising, J., Nziguheba, G., Waswa, B. S., Njoroge, S., Kabambe, V., Iwuafor, E., Kibunja, C., Esilaba, A.O., Coulibaly, A. 2015. Maize response to macronutrients and potential for profitability in sub-Saharan Africa. Nutrient cycling in agroecosystems, 105: 171-181. Kihara, J., Njoroge, S. 2013. Phosphorus agronomic efficiency in maize-based cropping systems: A focus on western Kenya. Field Crops Research, 150:1–8. Kihara, J., Nziguheba, G., Zingore, S., Coulibaly, A., Esilaba, A., Kabambe, V., Njoroge, S., Palm, C., Huising, J. 2016. Understanding variability in crop response to fertilizer and amendments in sub-Saharan Africa. Agriculture, Ecosystems and Environment, 229: 1-12. Kihara, J., Sileshi, G.W., Nziguheba, G., Kinyua, M., Zingore, S., Sommer, R. 2017. Application of secondary nutrients and micronutrients increases crop yields in sub-Saharan Africa. Agronomy for Sustainable Development, 37: 1-14. Kintché, K., Guibert, H., Sogbedji, J.M., Lévêque, J., Bonfoh, B., Tittonell, P. 2015. Long-term mineral fertiliser use and maize residue incorporation do not compensate for carbon and nutrient losses from a Ferralsol under continuous maize–cotton cropping. Field Crops Research, 184: 192–200. Kisinyo, P.O., Opala, P.A., Palapala, V., Gudu, S.O., Othieno, C.O., Ouma, E. 2015. Microdosing of lime, phosphorus and nitrogen fertilizers effect on maize performance on an acid soil in Kenya. Sustainable Agriculture Research, 4: 21–30. Kormawa, P., Munyemana, A., Soule, B. 2003. Fertilizer market reforms and factors influencing fertilizer use by smallholder farmers in Bénin. Agriculture, Ecosystems and Environment, 100:129–136.

173

References

Kurwakumire, N., Chikowo, R., Mtambanengwe, F., Mapfumo, P., Snapp, S., Johnston, A., Zingore, S. 2014. Maize productivity and nutrient and water use efficiencies across soil fertility domains on smallholder farms in Zimbabwe. Field Crops Research, 164: 136–147. Lal, R. 1995. Tillage systems in the tropics. Management options and sustainable implications. FAO, Rome, Italy. 70 p. Lemaire, G., van Oosterom, E., Jeuffroy, M.H., Gastal, F., Massignam, A. 2008. Crop species present different qualitative types of response to N deficiency during their vegetative growth. Field Crops Research, 105 (3): 253–265. Li, F.Y., Jamieson, P.D., Johnstone, P.R., Pearson, A.J. 2009. Mechanisms of nitrogen limitation affecting maize growth: a comparison of different modelling hypotheses. Crop and Pasture Science, 60 (8): 738–752. Li, H., Ma, Q., Li, H., Zhang, F., Rengel, Z., Shen, J. 2014. Root morphological responses to localized nutrient supply differ among crop species with contrasting root traits. Plant and soil, 376(1-2): 151-163. Li, L., Tang, C., Rengel, Z., Zhang, F.S. 2004. Calcium, magnesium and microelement uptake as affected by phosphorus sources and interspecific root interactions between wheat and chickpea. Plant and Soil, 261: 29-37. Liu, E., Yan, C., Mei, X., He, W., Hwat, S., Ding, L., Liu, Q., Liu, S., Fan, T. 2010. Long-term effect of chemical fertilizer, straw, and manure on soil chemical and biological properties in northwest China. Geoderma, 158:173– 180. Liu, H.L., Yang, J., He, P., Bai, Y., Jin, J., Drury, C.F., Zhu, Y., Yang, X., Li, W., Xie, J., Yang, J., Hoogenboom, G. 2012. Optimizing parameters of CSM- CERES-maize model to improve simulation performance of maize growth and nitrogen uptake in Northeast China. Journal of Integrative Agriculture, 11(11): 1898-1913. Liu, S., Yang, J.Y., Drury, C. F., Liu, H.L., Reynolds, W.D. 2014. Simulating maize (Zea mays L.) growth and yield, soil nitrogen concentration, and soil water content for a long-term cropping experiment in Ontario, Canada. Canadian Journal of Soil Science, 94(3): 435-452. Livingston, G., Schonberger, S., Delaney, S. 2011. Sub-Saharan Africa: the state of smallholders in agriculture. In IFAD Conference on New Directions for Smallholder Agriculture. Rome (Available at http://www.ifad.org/events/agriculture/doc/papers/ livingston.pdf.). Lizaso, J.I., Boote, K.J., Jones, J.W., Porter, C.H., Echarte, L., Westgate, M.E., Sonohat, G., 2011. CSM-IXIM: a new maize simulation model for DSSAT version 4.5. Agronomy journal, 103: 766–779. Luo, L., Ma, Y., Sanders, R.L., Xu, C., Li, J., Myneni, S.C. 2017. Phosphorus speciation and transformation in long-term fertilized soil: evidence from

174

References

chemical fractionation and P K-edge XANES spectroscopy. Nutrient cycling in agroecosystems, 107:215–226. Ma, Q., Zhang, F., Rengel, Z., Shen, J. 2013. Localized application of NH4+-N plus P at the seedling and later growth stages enhances nutrient uptake and maize yield by inducing lateral root proliferation. Plant and Soil, 372: 65-80. MacCarthy, D.S. and Vlek, P.L.G. 2012. Impact of climate change on sorghum production under different nutrient and crop residue management in semi- arid region of Ghana: a modeling perspective. African Crop Science Journal, 20(2):275–291. MacCarthy, D.S. and Vlek, P.L.G., Fosu-Mensah, B.Y. 2012. The response of maize to N fertilization in a sub-humid region of Ghana; understanding the processes using a crop simulation model. In: Kihara, J., Hoogenboom, G., Bationo, A. (Eds.), Improving soil fertility recommendation in Africa using Decision Support for Agro-technological Transfers (DSSAT). Springer. doi:10.1007/978-94-007-2960-5-5. MacCarthy, D.S., Adiku, S.G.K., Freduah, B.S., Gbefo, F., Kamara, A.Y. 2017. Using CERESMaize and ENSO as decision support tools to evaluate climate-sensitive farm management practices for maize production in the northern regions of Ghana. Frontiers in Plant Science, 8: 31. MacCarthy, D.S., Vlek, P.L.G., Bationo, A., Tabo, R., Fosu, M. 2010. Modeling nutrient and water productivity of sorghum in smallholder farming systems in a semi-arid region of Ghana. Field Crops Research, 118, 251–258. Mando, A., Bonzi, M., Woperei, M.C.S., Lompo, F., Stroosnijder, L. 2005. Long‐ term effects of mineral and organic fertilization on soil organic matter fractions and sorghum yield under Sudano‐Sahelian conditions. Soil Use and Management, 21:396–401. Manyame, C. 2006. On-farm yield and water use response of pearl millet to different management practices in Niger. Ph.D Dissertation, Texas A & M University, 138p. Mapfumo, P., Giller, K.E. 2001. Soil fertility management strategies and practices by smallholder farmers in semi-arid areas of Zimbabwe. International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) and Food and Agricultural Organization (FAO), Bulawayo, Zimbabwe and Rome, Italy, 60 p. Marteau, R., Sultan, B., Moron, V., Alhassane, A., Baron, C., Traoré, S.B. 2011. The onset of the rainy season and farmers’ sowing strategy for pearl millet cultivation in Southwest Niger. Agricultural and forest meteorology, 151(10): 1356-1369. Mashingaidze, N., Belder, P., Twomlow, S., Hove, L., Moyo, M., 2013. Improving maize (Zea mays L.) performance in semi-arid Zimbabwe 175

References

through microdosing with ammonium nitrate tablets. Experimental agriculture, 49: 179-196. McIntire, J., Powell, J. M. 1995. African semi-arid tropical agriculture cannot grow without external inputs. In International Conference on Livestock and Sustainable Nutrient Cycling in Mixed Farming Systems of Sub- Saharan Africa, Addis Ababa (Ethiopia), 22-26 Nov 1993. ILCA. McIntire, J., Powell, J., Fernández-Rivera, S., Williams, T. 1993. African semi- arid tropical agriculture cannot grow without external inputs. In International Conference on Livestock and Sustainable Nutrient Cycling in Mixed Farming Systems of Sub-Saharan Africa, Addis Ababa (Ethiopia) 22-26 Nov 1993. MEHU. 2011. Deuxième Communication Nationale de la République du Bénin sur les Changements Climatique (p. 168). Cotonou. http://www.agrhymet.ne/portailCC/images/pdf/nc2benin.pdf Morris, M., Kelley, V.A., Kopicki, R.J., Byerlee, D. 2007. Fertilizer Use in sub- Sahelian regions. Journal of Agricultural Science, 88: 12-16. Muehlig-Versen, B., Buerkert, A., Bationo, A., Roemheld, V. 2003. Phosphorus placement on acid arenosols of the west African Sahel. Experimental agriculture, 39: 307–325. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A. 2012. Closing yield gaps through nutrient and water management. Nature, 490: 254-257. Naab, J.B., Mahama, G.Y., Koo, J., Jones, J.W., Boote, K.J. 2015. Nitrogen and phosphorus fertilization with crop residue retention enhances crop productivity, soil organic carbon, and total soil nitrogen concentrations in sandy-loam soils in Ghana. Nutrient cycling in agroecosystems, 102 : 33–43. Nago, C.M. 1989. Technologies traditionnelles et alimentation au Bénin : aspects techniques, biochimiques et nutritionnels. Identification et caractérisation des principales filières et technologies du secteur traditionnel de transformation alimentaire. Abomey-Calavi (Bénin) FSA/UNB. Ncube, B., Dimes, J.P., Twomlow, S.J., Mupangwa, W., Giller, K.E. 2007. Raising the productivity of smallholder farms under semi-arid conditions by use of small doses of manure and nitrogen: a case of participatory research. Nutrient Cycling in Agroecosystems, 77: 53-67. Nederlof, E.S., Dangbégnon, C. 2007. Lessons for farmer-oriented research: experiences from a West African soil fertility management project. Agriculture and human values, 24(3): 369-387. Nezomba, H., Mtambanengwe, F., Chikowo, R., Mapfumo, P. 2015. Sequencing integrated soil fertility management options for sustainable crop

176

References

intensification by different categories of smallholder farmers in Zimbabwe. Experimental agriculture, 51:17–41 Ngwira, A.R., Aune, J.B., Thierfelder, C., 2014. DSSAT modelling of conservation agriculture maize response to climate change in Malawi. Soil and Tillage Research, 143: 85–94. Njoroge, R., Otinga, A.N., Okalebo, J.R., Pepela, M., Merckx, R. 2017. Occurrence of poorly responsive soils in western Kenya and associated nutrient imbalances in maize (Zea mays L.). Field Crops Research, 210: 162- 174. Nwite, J.N., 2015. Effect of different urine sources on soil chemical properties and maize yield in Abakaliki, Southeastern Nigeria. International Journal of Advance Agricultural Research, 3: 31–36. Nziguheba, G., Zingore, S., Kihara, J., Merckx, R., Njoroge, S., Otinga, A., Vanlauwe, B. 2016. Phosphorus in smallholder farming systems of sub- Saharan Africa: implications for agricultural intensification. Nutrient Cycling in Agroecosystems, 104: 321–340. Okebalama, C.B., Safo, E.Y., Yeboah, E., Abaidoo, R.C., Logah, V. 2016. Fertilizer Microdosing in the Humid Forest Zone of Ghana: An Efficient Strategy for Increasing Maize Yield and Income in Smallholder Farming. Soil Science Society of America Journal, 80: 1254-1261. Opala, P.A., Othieno, C.O., Okalebo, J.R., Kisinyo, P.O. 2010. Effects of combining organic materials with inorganic phosphorus sources on maize yield and financial benefits in western Kenya. Experimental agriculture, 46: 23–34. Otinga, A.N., Pypers, P., Okalebo, J.R., Njoroge, R., Emong’ole, M., Six, L., Vanlauwe, B., Merckx, R. 2013. Partial substitution of phosphorus fertiliser by farmyard manure and its localised application increases agronomic efficiency and profitability of maize production. Field Crops Research, 140: 32–43. Palé, S., Mason, S.C., Taonda, S.J.B., 2009. Water and fertilizer influence on yield of grain sorghum varieties produced in Burkina Faso. South African journal of plant and soil, 26: 91-97. Pascual, U., Barbier E.B. 2006. Deprived land-use intensification in shifting cultivation: the population pressure hypothesis revisited. Agricultural Economics, 34:155–165. Pickering, N.B., Hansen, J.W., Jones, J.W., Wells, C.M., Chan, V.K., Godwin, D.C. 1994. WeatherMan: a utility for managing and generating daily weather data. Agronomy Journal, 86(2): 332-337.

177

References

Porter, C.H., Jones, J.W., Adiku, S., Gijsman, A.J., Gargiulo, O., Naab, J.B. 2010. Modeling organic carbon and carbon-mediated soil processes in DSSAT v4.5. Operational Research, 10: 247–278. Powell, J.M., Williams, T.O. 1993. Livestock, nutrient cycling and sustainable agriculture in the West African Sahel. Sustainable Agriculture Programme, International Institute for Environment and Development. London, UK Powell, J.M., Ikpe, F.N., Somda, Z.C., Fernandez-Rivera, S., 1998. Urine effects on soil chemical properties and the impact of urine and dung on pearl millet yield. Experimental Agriculture, 34: 259–276. PSDSA, 2017. Plan Stratégique de Développement du Secteur Agricole (PSDSA) 2025 et Plan National d’Investissements Agricoles et de Sécurité Alimentaire et Nutritionnelle PNIASAN (2017 – 2021). 91 p. http://www.agriculture.bj/IMG/pdf/psdsa_2025_et_pniasan_2017_- _2021_version_finale_adoptee.pdf Raab, T.K., Terry, N. 1994. Nitrogen source regulation of growth and photosynthesis in Beta vulgaris L. Plant Physiology, 105: 1159–1166. Rahman, N.A., Fosu, M., Tetteh, F.M. 2014. Effect of soil system and soil amendments on growth and yield of maize in northern Ghana. Journal of Agronomy, 13:79–84 Ramsperger, B., Herrmann, L., Stahr, K. 1998. Dust characteristics and source‐ sink relations in eastern West Africa (SW‐Niger and Benin) and South America (Argentinean Pampas). Journal of Plant Nutrition and Soil Science, 161: 357–363. Rasool, R., Kukal, S.S., Hira, G.S. 2010. Root Growth and Soil Water Dynamics in Relation to Inorganic and Organic Fertilization in Maize–Wheat. Communications in soil science and plant analysis, 41(20): 2478-2490. Ratjen, A. M., Kage, H. 2016. Nitrogen-limited light use efficiency in wheat crop simulators: comparing three model approaches. The Journal of Agricultural Science, 154(6): 1090-1101. Rezzoug, W., Gabrielle, B., Suleiman, A., Benabdeli, K. 2008. Application and evaluation of the DSSAT-wheat in the Tiaret region of Algeria. African Journal of Agricultural Research, 3(4): 284-296. Ripoche, A., Crétenet, M., Corbeels, M., Affholder, F., Naudin, K., Sissoko, F., Douzet, J.M., Tittonell, P. 2015. Cotton as an entry point for soil fertility maintenance and food crop productivity in savannah agroecosystems– Evidence from a long-term experiment in southern Mali. Field crops research, 177: 37-48. Ritchie, J.T. 1998. Soil water balance and plant water stress. In: Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (Eds.), Understanding Options for

178

References

Agricultural Production. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 41–54. Robinson, D. 2001. Root proliferation, nitrate inflow and their carbon costs during nitrogen capture by competing plants in patchy soil. Plant and Soil, 232, 41–50. Ronner, E., Franke, A.C., Vanlauwe, B., Dianda, M., Edeh, E., Ukem, B., Bala, A., van Heerwaarden, J. Giller, K.E. 2016. Understanding variability in soybean yield and response to P-fertilizer and rhizobium inoculants on farmers’ fields in northern Nigeria. Field Crops Research, 186: 133-145. Rufino, M.C., Tittonell, P., Van Wijk, M.T., Castellanos-Navarrete, A., Delve, R. J., De Ridder, N., Giller, K.E. 2007. Manure as a key resource within smallholder farming systems: analysing farm-scale nutrient cycling efficiencies with the NUANCES framework. Livestock Science, 112(3): 273- 287. Rusinamhodzi, L., Corbeels, M., Zingore, S., Nyamangara, J., Giller, K.E. 2013. Pushing the envelope? Maize production intensification and the role of cattle manure in recovery of degraded soils in smallholder farming areas of Zimbabwe. Field Crops Research, 147:40–53 Saïdou, A. 2006. Converging strategies by farmers and scientists to improve soil fertility and enhance crop production in Benin. PhD Thesis dissertation, Wageningen University, The Netherlands. 232p. Saïdou, A., Balogoun, I., Ahoton, E. L., Igué, A. M., Youl, S., Ezui, G., Mando, A. 2017. Fertilizer recommendations for maize production in the South Sudan and Sudano-Guinean zones of Benin. Nutrient Cycling in Agroecosystems, 110:361–373. Saïdou, A., Kuyper, T. W., Kossou, D. K., Tossou, R., Richards, P. 2004. Sustainable soil fertility management in Benin: learning from farmers. NJAS-Wageningen Journal of Life Sciences, 52(3): 349-369. Samaké, O., Smaling, E.M.A., Kropff, M.J., Stomph, T.J., Kodio, A. 2005. Effects of cultivation practices on spatial variation of soil fertility and millet yields in the Sahel of Mali. Agriculture, Ecosystem and Environment, 109: 335- 345. Sanchez, A.C., Fandohan, B., Assogbadjo, A. E., Sinsin, B. 2012. A countrywide multiethnic assessment of local communities’ perception of climate change in Benin (West Africa). Climate and Development, 4(2): 114–128. Sanchez, P., 2010. Tripling yields in tropical Africa. Nature Geoscience, 3: 299–300. Sangaré, M., Fernández-Rivera, S., Hiernaux, P., Bationo, A., Pandey, V. 2002. Influence of dry season supplementation for cattle on soil fertility and millet (Pennisetum glaucum L.) yield in a mixed crop/livestock production system of the Sahel. Nutrient cycling in agroecosystems, 62(3): 209-217.

179

References

Sanginga, N., Woomer, P.L. 2009. Integrated soil fertility management in Africa: Principles, prcatices and developmental process. Nairobi: Tropical Soil Biology and Fertility Institute of the International Centre for Tropical Agriculture. Sath, K., Pauly, T., Holtenius, K., 2012. Mineral balance of Cambodian cattle based on their faecal and urinary excretion. Journal of Animal and Veterinary Advances, 11: 4221–4225. Schlecht, E., Buerkert, A. 2004. Organic inputs on millet fields in western Niger: the implications of farmers’ practices for sustainable agricultural production. Geoderma, 121: 271–289. Schlecht, E., Fernandez-Rivera, S., Hiernaux, P. 1998. Timing, size and N concentration of faecal and urinary excretions in cattle sheep and goats— can they be used for better manuring of cropland? In: Renard, G., Neef, A., Becker, K., von Oppen, M. (Eds.), Soil fertility management in West African land use systems. Margraf Verlag, Weikersheim, Germany, pp. 361–368. Schlecht, E., Hiernaux, P., Achard, F., Turner, M.D. 2004. Livestock related nutrientbudgets within village territories in western Niger. Nutrient Cycling in Agroecosystems, 68: 199–211. Schlecht, E., Hiernaux, P., Kadaouré, I., Hülsebusch, C., Mahler, F., 2006. A spatio-temporal analysis of forage availability and grazing and excretion behaviour of herded and free grazing cattle, sheep and goats in Western Niger. Agriculture, Ecosystem and Environment, 113: 226–242. Sileshi, G., Akinnifesi, F.K., Debusho, L.K., Beedy, T., Ajayi, O.C. Mong’omba, S. 2010. Variation in maize yield gaps with plant nutrient inputs, soil type and climate across sub-Saharan Africa. Field Crops Research, 116: 1-13. Sime, G., Aune, J.B., 2014. Maize response to fertilizer dosing at three sites in the Central Rift Valley of Ethiopia. Agronomy, 4: 436–451. Sinclair, T.R., Amir, J. 1991. A model to assess nitrogen limitations on the growth and yield of spring wheat. Field Crops Research, 30: 63–78. Smale, M., Byerlee, D., Jayne, T. 2011. Maize revolutions in Sub-Saharan Africa. Policy Research Working Paper Series. Washington D.C.:. The World Bank. Smale, M., Byerlee, D., Jayne, T. 2013. Maize revolutions in Sub-Saharan Africa. In: Otsuka, K., Larson, D.F. (Eds.), An african green revolution: finding ways to boost productivity on small farms. Springer, Dordrecht, pp 165– 195. Smit, A.L., Blom-Zandstra, M., van der Werf, A., Bindraban, P.S. 2013. Enhancing early root growth to exploit indigenous soil P and fertilizer P.

180

References

VFRC Report 2013/4. Virtual Fertilizer Research Center, Washington, D.C. 36pp. Soler, C.T., Bado, V.B., Traore, K., Bostick, W.M., Jones, J.W., Hoogenboom, G. 2011. Soil organic carbon dynamics and crop yield for different crop rotations in a degraded ferruginous tropical soil in a semi-arid region: a simulation approach. The Journal of agricultural science, 149(5): 579-593. Srivastava, A.K., Mboh, C.M., Gaiser, T., Ewert, F. 2017. Impact of climatic variables on the spatial and temporal variability of crop yield and biomass gap in Sub-Saharan Africa-a case study in Central Ghana. Field Crops Research, 203: 33-46. Srivastava, A.K. 2010. Management Effects on Yam Production in Benin Republic-Experimental Analysis and Modeling. PhD Dissertation, University of Bonn. 87p. Suzuki, K., Matsunaga, R., Hayashi, K., Matsumoto, N., Tabo, R., Tobita, S., Okada, K. 2014. Effects of traditional soil management practices on the nutrient status in Sahelian sandy soils of Niger, West Africa. Geoderma, 223: 1-8. Tabo, R., Bationo, A, Amadou, B., Marchal, D., Lompo, F., Gandah, M., Hassane, O., Diallo, M.K., Ndjeunga, J., Fatondji, D., Gerard, B., Sogodogo, D., Taonda, J.B.S., Sako, K., Boubacar, S., Abdou, A., Koala, S. 2011. Fertilizer microdosing and ‘‘warrantage’’ or inventory credit system to improve food security and farmers’ income in West Africa. In: Bationo, A., Waswa, B., Okeyo, J.M., Maina, F., Kihara, J.M. (Eds.), Innovations as key to the green revolution in Africa. Springer, Netherlands, pp. 113–121. Tabo, R., Bationo, A., Gerard, B., Ndjeunga, J., Marchal, D., Amadou, B., Diallo, M.K. 2007. Improving cereal productivity and farmers’ income using a strategic application of fertilizers in West Africa. In: Bationo, A., Waswa, B., Kihara, J., Kimetu, J. (Eds.), Advances in integrated soil fertility management in sub-Saharan Africa: Challenges and opportunities. Springer, Netherlands, pp. 201–208. Tidjani, M.A., Akponikpè, P.B.I. 2012. Evaluation des stratégies paysannes d’adaptation aux changements Climatiques : Cas de la production du maïs au Nord-Bénin. African Crop Science Journal, 20(2): 425-441. Tittonell, P., Giller, K.E. 2013. When yield gaps are poverty traps: The paradigm of ecological intensification in African smallholder agriculture. Field Crops Research, 143, 76-90. Tittonell, P., Corbeels, M., Van Wijk, M.T., Vanlauwe, B., Giller, K.E. 2008a. Combining organic and mineral fertilizers for integrated soil fertility management in smallholder farming systems of Kenya: explorations using the crop-soil model FIELD. Agronomy Journal, 100: 1511–1526.

181

References

Tittonell, P., Vanlauwe, B., Corbeels, M., Giller, K.E. 2008b. Yield gaps, nutrient use efficiencies and response to fertilisers by maize across heterogeneous smallholder farms of western Kenya. Plant and Soil, 313: 19–37. Tittonell, P., Muriuki, A., Shepherd, K.D., Mugendi, D., Kaizzi, K.C., Okeyo, J., Verchot, L., Coe, R., Vanlauwe, B. 2010. The diversity of rural livelihoods and their influence on soil fertility in agricultural systems of East Africa - A typology of smallholder farms. Agricultural Systems, 103: 83–97. Tittonell, P., Vanlauwe, B., Leffelaar, P.A., Shepherd, K.D., Giller, K.E. 2005. Exploring diversity in soil fertility management of smallholder farms in western Kenya. II. Within farm variability in resource allocation, nutrient flows and soil fertility status. Agriculture, Ecosystems and Environment, 110: 166–184. Tittonell, P., Vanlauwe, B., Misiko, M. and Giller, K. E. 2011. Targeting resources within diverse, heterogeneous and dynamic farming systems: Towards a ‘Uniquely African green revolution’. In: Bationo, A., Waswa, B., Okeyo, J.M., Maina, F., Kihara, J.M. (Eds), Innovations as key to the green revolution in Africa. Springer, Netherlands, pp. 747–758. Tovihoudji, P.G., Akponikpè, P.B.I., Adjogboto, A., Djenontin, J.A., Agbossou, E.K., Bielders, C.L. 2017a. Combining hill-placed manure and mineral fertilizer enhances maize productivity and profitability in northern Benin. Nutrient Cycling in Agroecosystems. https://doi.org/10.1007/s10705-017- 9872-8 Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bertin, P., Bielders, C.L. 2017b. Fertilizer microdosing enhances maize yields but may exacerbate nutrient mining in maize cropping systems in northern Benin. Field Crops Research, 213: 130–142. Traoré, B., Corbeels, M., Van Wijk, M.T., Rufino, M.C., Giller, K. E. 2013. Effects of climate variability and climate change on crop production in southern Mali. European Journal of Agronomy, 49:115-125 Turner, N. C., Rao, K.P.C. 2013. Simulation analysis of factors affecting sorghum yield at selected sites in eastern and southern Africa, with emphasis on increasing temperatures. Agricultural Systems, 121: 53-62. Twomlow, S., Mugabe, F. T., Mwale, M., Delve, R., Nanja, D., Carberry, P., Howden, M. 2008. Building adaptive capacity to cope with increasing vulnerability due to climatic change in Africa–A new approach. Physics and Chemistry of the Earth, Parts A/B/C, 33(8-13): 780-787. Twomlow, S., Rohrbach, D., Dimes, J., Rusike, J., Mupangwa, W., Ncube, B., Hove, L., Moyo, M., Mashingaidze, N., Mahposa, P. 2010. Microdosing as a pathway to Africa’s Green Revolution: evidence from broad-scale on- farm trials. Nutrient Cycling in Agroecosystems, 88: 3-15.

182

References

Valbuena, D., Erenstein, O., Tui, S.H.K., Abdoulaye, T., Claessens, L., Duncan, A.J., Gérard, B., Rufino, M.C., Teufel, N., van Rooyen, A., van Wijk, M.T. 2012. Conservation agriculture in mixed crop–livestock systems: scoping crop residue trade-offs in Sub-Saharan Africa and South Asia. Field Crops Research, 132: 175–184. Valbuena, D., Tui, S.H.K., Erenstein, O., Teufel, N., Duncan, A., Abdoulaye, T., Swain, B., Mekonnen, K., Germaine, I., Gérard, B. 2014. Identifying determinants, pressures and trade-offs of crop residue use in mixed smallholder farms in Sub-Saharan Africa and South Asia. Agricultural Systems, 134: 107-118. van Reeuwijk, L.P. 1993. Procedures for soil analysis. Technical paper No 9, Fourth Edition edited by the International Soil Reference and Information Center (ISRIC). Wageningen, Netherlands. van Vugt, D., Franke, A.C., Giller, K.E. 2017. Understanding variability in the benefits of N2-fixation in soybean-maize rotations on smallholder farmers’ fields in Malawi. Agriculture, Ecosystems and Environment, https://doi.org/10.1016/j.agee.2017.05.008 Vanlauwe, B., Giller, K.E. 2006. Popular myths around soil fertility management in sub-Saharan Africa. Agriculture, ecosystems andenvironment, 116(1-2): 34-46. Vanlauwe, B., Aihou, K., Aman, S., Iwuafor, E. N., Tossah, B.K., Diels, J., Sanginga, N., Lyasse, O., Merckx, R., Deckers, J. 2001a. Maize yield as affected by organic inputs and urea in the West African moist savanna. Agronomy journal, 93(6): 1191-1199. Vanlauwe, B., Aihou, K., Houngnandan, P., Diels, J., Sanginga, N., Merckx, R. 2001b. Nitrogen management inadequate'input maize-based agriculture in the derived savanna benchmark zone of Benin Republic. Plant and Soil, 228(1): 61-71. Vanlauwe, B., Bationo, A., Chianu, J., Giller, K.E., Merckx, R., Mokwunye, U., Pypers, P., Shepherd, K., Smaling, E.M.A., Woomer P.L., Sanginga N. 2010. Integrated soil fertility management operational definition and consequences for implementation and dissemination. Outlook on agriculture, 39(1): 17-24. Vanlauwe, B., Coe, R., Giller, K.E., 2016. Beyond averages: new approaches to understand heterogeneity and risk of technology success or failure in smallholder farming. Experimental Agriculture, 1–23. http://dx.doi.org/10.1017/S00144797160001931. Vanlauwe, B., Descheemaeker, K., Giller, K.E., Huising, J., Merckx, R., Nziguheba, G., Wendt, J., Zingore, S. 2015. Integrated soil fertility management in sub-Saharan Africa: unravelling local adaptation. Soil 1: 491–508. http://dx. doi.org/10.5194/soil-1-491-2015.

183

References

Vanlauwe, B., Kihara, J., Chivenge, P., Pypers, P., Coe, R., Six, J. 2011. Agronomic use efficiency of N fertilizer in maize-based systems in sub- Saharan Africa within the context of integrated soil fertility management. Plant and soil, 339(1-2): 35-50. Vanlauwe, B., Six, J., Sanginga, N., Adesina, A.A., 2015. Soil fertility decline at the base of rural poverty in sub-Saharan Africa. Nature Plants, 1: 15101, doi:10.1038/nplants.2015.101. Vanlauwe, B., Wendt, J., Giller, K. E., Corbeels, M., Gerard, B., Nolte, C. 2014. A fourth principle is required to define conservation agriculture in sub- Saharan Africa: the appropriate use of fertilizer to enhance crop productivity. Field Crops Research, 155: 10-13. Versteeg, M. N., Amadji, F., Eteka, A., Gogan, A., Koudokpon, V. 1998. Farmers' adoptability of Mucuna fallowing and agroforestry technologies in the coastal savanna of Benin. Agricultural systems, 56(3): 269-287. Vos, J., Van Der Putten, P.E.L., Birch, C.J. 2005. Effect of nitrogen supply on leaf appearance, leaf growth, leaf nitrogen economy and photosynthetic capacity in maize (Zea mays L.). Field Crops Research, 93: 64–73. Walkley, A., Black, I.A. 1934. An examination of the Degtjareff method for determining soil organic matter, and a proposed modification of the chromic acid titration method. Soil Science, 37: 29–38. Wei, W., Yan, Y., Cao, J., Christie, P., Zhang, F., Fan, M. 2016. Effects of combined application of organic amendments and fertilizers on crop yield and soil organic matter: An integrated analysis of long-term experiments. Agriculture, Ecosystems and Environment, 225: 86-92. Wildemeersch, J.C.J., Timmerman, E., Mazijn, B., Sabiou, M., Ibro, G., Garba, M., Cornelis, W. 2015. Assessing the constaints to adopt water and soil conservation tecnhique in Tillabery, Niger. Land Degradation and Development, 26(5): 491-501. Williams, T.O. 1999. Factors influencing manure application by farmers in semi- arid west Africa. Nutrient Cycling in Agroecosystems, 55: 15-22. Williams, T.O., Powell, J.M., Fernandez-Rivera, S. 1995. Manure utilisation, drought cycles and herd dynamics in the Sahel: implications for cropland productivity. In: Powell, J.M., Fernandez-Rivera, S., Williams, T.O., Renard, C. ((Eds.), Livestock and Sustainable Nutrient Cycling in Mixed Farming Systems of sub-Saharan Africa, Vol. Vol. II: Technical Papers., 393-409 Addis Ababa, Ethiopia. Willmott, C.J., Ackleson, G.S., Davis, R.E., Feddema, J.J., Klink, K.M., Legates, D.R., O’donnell, J., Rowe, C.M., 1985. Statistics for the evaluation and comparison of models. Journal of Geophysical Research, 90: 8995–9005.

184

References

Wopereis, M.C.S., Tamélokpo, A., Ezui, K., Gnakpénou, D., Fofana, B., Breman, H. 2006. Mineral fertilizer management of maize on farmer fields differing in organic inputs in the West African savanna. Field Crops Research, 96: 355– 362. Yakoub, A., Lloveras, J., Biau, A., Lindquist, J.L., Lizaso, J.I. 2017. Testing and improving the maize models in DSSAT: Development, growth, yield, and N uptake. Field Crops Research, 212: 95-106. Yegbemey, R.N. 2014. Adaptation to climate change and sustainable agriculture in West Africa: a case study of maize farming in Benin. PhD Thesis. Justus- Liebig University Giessen, Germany, Margraf Publishers. Yegbemey, R.N., Kabir, H., Awoye, O. H., Yabi, J.A., Paraïso, A.A. 2014. Managing the agricultural calendar as coping mechanism to climate variability: A case study of maize farming in northern Benin, West Africa. Climate Risk Management, 3: 13-23. Yemadjè, R.H., Crane, T.A., Vissoh, P.V., Mongbo, R.L., Richards, P., Kossou, D. K., Kuyper, T.W. 2012. The political ecology of land management in the oil palm based cropping system on the Adja plateau in Benin. NJAS- Wageningen Journal of Life Sciences, 60: 91-99. Yi, Z.X., Wang, P., Zhang, H.F., Shen, L.X., Liu, M., Dai, M.H. 2006. Effects of type and application rate of nitrogen fertilizer on source-sink relationship in summer maize in North China Plain. Plant Nutrition and Fertilizer Science, 12 : 294–300. Youssouf, I., Lawani, M. 2002. Les sols béninois : classification dans la base de référence mondiale. Quatorzième réunion du sous-comité ouest et centre africain de corrélation des sols pour la mise en valeur des terres, 9-13 octobre 2000, Abomey, Bénin. In: Rapport sur les Ressources en Sols du Monde n° 98, FAO, Rome, pp. 29–50. Zhang, H.M., Forde, B.G. 1998. An ArabidopsisMADS box gene that controls nutrientinduced changes in root architecture. Science, 279: 407–409. Zhou, J., Xia, F., Liu, X., He, Y., Xu, J., Brookes, P.C. 2014. Effects of nitrogen fertilizer on the acidification of two typical acid soils in South China. Journal of soils and sediments, 14(2): 415-422. Zingore, S., Delve, R. J., Nyamangara, J., Giller, K.E. 2008. Multiple benefits of manure: the key to maintenance of soil fertility and restoration of depleted sandy soils on African smallholder farms. Nutrient Cycling in Agroecosystems, 80(3): 267-282. Zingore, S., Tittonell, P., Corbeels, M., Van Wijk, M.T., Giller, K.E. 2011. Managing soil fertility diversity to enhance resource use efficiencies in smallholder farming systems: a case from Murewa District, Zimbabwe. Nutrient Cycling in Agroecosystems, 90(1): 87-103.

185

Appendices

187

Appendices

Appendix A

Table A.1. Average observed grain yield/standard deviation and statistical performances of model outputs [root mean square error (RMSE) and relative RMSE (RRMSE)] across the two years of the trials (2014–2015). Mean SD RMSE RRMSE Treatment kg ha-1 % Control 1069 308 212 20 MD1 2212 437 352 16 MD2 2289 510 470 20 MD1+FYM 2981 559 443 14 MD2+FYM 3135 534 490 16 RR 2455 568 507 20

188

Appendices

Table A.2. Statistical indicators showing the relationship between simulated and measured maize grain and biomass yield for combined application of 0 (NM) or 3 (3M) t ha-1 of manure and the two levels of microdosing (MD1 and MD2) in 2014 and 2015. Number between brackets shows the relative RMSE (%) Baseline model Optimized model

Grain yield Biomass yield N uptake Grain yield Biomass yield N uptake

Indicators 2014 2015 2014 2015 2014 2015 2014 2015 2014 2015 2014 2015 239 750 745 820 10.2 15.6 250 395 1010 230 10.0 4.2 RMSE (kg ha-1) (8%) (23%) (10%) (11%) (14%) (21%) (8%) (12%) (13%) (3%) (13%) (6%) d (-) 0.73 0.45 0.75 0.77 0.75 0.34 0.60 0.68 0.60 0.97 0.78 0.93

E1 (-) 0.26 -2.51 0.10 -0.43 0.19 -3.20 0.25 -0.55 -0.26 0.60 0.27 0.35

189

Appendices

Appendix B

Figure B.1. Maize root growth and proliferation in field experiment following localized application of P plus N in North China (from Jing et al., 2012). Root proliferation was stimulated in the nutrient-rich patches with localized application of P plus ammonium (a) or nitrate (b); Maize roots (e) and root hairs (f) have greater responses to localized application of P plus ammonium.

A B

Figure B.2. Experimental site before (A) and after (B) corralling practice

190

Appendices

Figure B.3. N, P and K uptake by above-ground biomass within different manure strata as a function of mineral fertilization in 2014 and 2015. Control refers to the unfertilized treatment. MD1, MD2 and RR refer to microdosing option 1 and 2, and the recommended fertilizer rate, respectively. Error bars =standard errors.

191

Appendices

Figure B.4. (a) Predicted vs. Observed grain yields and (b) relation between residuals and predicted values.

192

Appendices

Figure B.5. Comparison between field measured and DSSAT simulated soil water in 0– 0.20 m (upper panel) and 0.20–0.40 m (lower panel) layers for four selected treatments in 2015. Error bars=standard deviation (n = 3).

193

Appendices

2014 2015

)

N uptake N (kg ha

)

Grainyield (kg ha

)

Biomassyield (kg ha

1 Figure B.6. Comparison between observed (box-and-whisker plots) and simulated (black points) N uptake (a, b) maize grain (c, d), and aboveground biomass yield (e, f) at harvest as affected by combined application of 0 (NM) or 3 (3M) t ha-1 of manure and two microdosing rates (MD1-MD2) in 2014 and 2015 using the default value of the N stress coefficient (CTCNP2=0.160). MD1=microdosing option 1; MD2= microdosing option 2.

194

Appendices

1600 Events ≥ 40mm Rain events Cumul rain Sow-Mat. Annual rainfall 100 90 1400 82 83 80 70 81 81 78 77 1200 74 77 76 75 74 72 71 72 75 80 70 68 70 68 69 62 76 1000 64 72 70 67 64 61 60 63 61 69 800 50

40 600 30 400 20 200 10 4 4 1 3 11 4 5 6 7 5 9 3 6 5 8 10 9 6 9 6 5 6 4 8 5 10 5 5 11 3 8 7 0 0

Figure B.7. Rainfall characteristics of the simulation period (32 years; 1984-2015)

195

Publications and Conferences

Publications in international peer-reviewed journals

1. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bertin, P., Bielders, C.L. 2017. Fertilizer microdosing enhances maize yields but may exacerbate nutrient mining in maize cropping systems in northern Benin. Field Crops Research, 213: 130-142. 2. Tovihoudji, P.G., Akponikpè, P.B.I., Adjogboto, A., Djenontin, J.A., Agbossou, E.K., Bielders, C.L. 2017. Combining hill-placed manure and mineral fertilizer enhances maize productivity and profitability in northern Benin. Nutrient Cycling in Agroecosystems, 110(3): 375-393. 3. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Variability in maize yield and profitability following hill-placement of mineral fertilizer and manure under smallholder farm conditions in northern Benin. Submitted to Field Crops Research. 4. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Using the DSSAT model to support decision making regarding fertilizer microdosing for maize production in the sub-humid region of Benin. Submitted to Frontiers in Plant Science/Section: Agroecology and Land Use Systems. 5. Agbangba, C.E, Sossa, E.L., Dagbenonbakin, G.D., Tovihoudji, P.G., Kindomihou, V. 2017. Modelisation de la réponse de l’ananas Cayenne lisse à l’azote, au phosphore et au potassium sur sols ferralitiques au Bénin. Revue CAMES, Sciences de la vie, de la terre et agronomie, 4(2): 12-25. 6. Yolou, I., Tovihoudji, P.G., Batamoussi Hermann, M., Yabi, I., Paraïso, A. A., Akiyo, R., Afouda, F. 2015. Short-term effects of conjunctive use of municipal solid waste compost and inorganic fertilizer on soil properties and maize productivity in Northern Benin. International Research Journal of Agricultural Science and Soil Science, 5(5): 137-149. 7. Tovihoudji, P.G., Djogbenou, C.P., Akponikpe, P.B.I., Kpadonou, E., Agbangba, C.E., Dagbenonbakin, D.G. 2015. Response of Jute Mallow (Corchorus olitorius L.) to organic manure and inorganic fertilizer on a ferruginous soil in North-eastern Benin. Journal of Applied Biosciences, 92(1): 8610-8619. 8. Agossou, J., Afouda, L., Adédémy, J.D., Noudamadjo, A., N’da Tido, C., Tovihoudji, P.G.,……. Ayivi, B. 2014. Risques de fièvres typhoïdes et paratyphoïdes liés à l’utilisation des eaux usées en agriculture urbaine et

197

Publications and Conferences

périurbaine: cas du maraîchage dans la ville de Parakou. Environnement, Risques & Santé, 13(5): 405-416. 9. Tovihoudji, P.G., Yegbemey, R.N., Akponikpè, P.B.I,. Yabi, J.A., Bielders, C.L. Resource endowment and soil fertility management strategies in maize farming systems in Northern Benin. To be submitted 10. Tovihoudji, P.G., Akponikpè P.B.I. Growth, yield and nutrient content of lettuce (Lactuca sativa L.) and amaranth (Amaranthus cruentus L.) grown under wastewater irrigation and inorganic fertilizer in northern Benin. To be submitted

National and international conference proceedings

1. Takpa, G.M.M.O., Tovihoudji, P.G., Ollabodé, N., Labiyi, I.A., Yabi, J. A. Fertilizer microdosing application as a climate change mitigation strategy in maize farming in northern Benin: Farmers’ perception and willingness to be informed and to apply. Colloque UAC 2017, Abomey Calavi,Benin, 25 au 30 septembre 2017. 2. Abiola, W.A., Tovihoudji, P.G., Sekloka, E., Moumouni, I.M., Afouda, L. Performance and farmers’ preference of different drought tolerant maize varieties under fertilizer microdosing in northern Benin. Colloque UAC 2017, Abomey Calavi,Benin, 25 au 30 septembre 2017. 3. Tovihoudji, P.G., Defourny, M., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Fertilizer microdosing adoption in maize farming in northern Benin: Understanding farmer decision-making and potential constraints. Colloque UAC 2017, Abomey Calavi,Benin, 25 au 30 septembre 2017. 4. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Understanding Variability in Maize Yield and Profitability under Fertiliser Microdosing Technology in Farmers’ Fields in Northern Benin. Poster presentation at International conference TROPENTAG, September 20-22, 2017, Bonn, Germany. 5. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. Microdose fertilization increases yields but exacerbates nutrient mining in maize farming systems in Benin”, oral and poster presentation at International conference on “Plant Nutrition, Growth & Environment Interactions III (Session: Plant Nutrition: Fertilizers and Soil), February 20-21, 2017, Vienna, Austria. 6. Akponikpè, P.B.I., Tovihoudji, P.G., Agbossou, E.K., Bielders, C.L. Efficient use of nutrients and water through combination of manure and

198

Publications and Conferences

fertilizer in smallholder maize farming system in northern Benin”. Présentation orale à l’atelier de recommandation des engrais organisé par IFDC à Lomé (Togo) le 14 juin 2016. 7. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. On- farm evaluation of fertilizer microdosing technology in smallholder maize farming system of Northern Benin », présentation orale au 5ème Colloque « Recherche Scientifique Face aux Nouveaux Défis de Développement en Afrique», Université d'Abomey Calavi (Bénin) du 28 Septembre au 03 Octobre 2015. 8. Tovihoudji, P.G., Akponikpè, P.B.I., Agbossou, E.K., Bielders, C.L. 2015. Integrated fertiliser microdosing and organic manure to adapt to climate variability and change in Northern Benin. Global Science Conference/Climart Smart Agriculture 2015, Montpellier, France; 03/2015.

199

About the author

Pierre G. Tovihoudji was born on 1st January 1984 in Adjarra, southern Benin. He pursued his primary and secondary educations in Avrankou and Porto-Novo from 1990 to 2001 before joining the “Lycéé Agricole Medji de Sékou” in 2002. He graduated from this school with the “Dipôme d’Etude Agricole Tropicale”, DEAT, in November 2005. In the same period, he prepared the scientific baccalaureate degree (Série D: Mathematics and Biology) in self-taught and graduated in August 2004. He won in December 2005 the National College Entrance Examination (Agricultural Sciences) and enter the Faculty of Agronomy of the “Université de Parakou” and graduated in 2011 in Crop Production Sciences. As the best Engineer student, he obtains a scholarship of distinction from the “Université de Parakou” to pursue a MSc. programme on Agricultural Engineering and Environmental Sciences and graduated in April 2013. He started his PhD research in 2014 on soil fertility management under fertilizer microdosing technology with a part on soil-crop modelling in a joint doctoral programme between the Université catholique de Louvain (UCL- Belgium) and the Université d’Abomey-Calavi (UAC, Benin) funded by the West Africa Agricultural Productivity Program (WAAPP-Benin) and UCL. He has been involved, as research assistant and freelance consultant, in many research and development projects in Agricultural Engineering since 2012. His stay at the Ministry of Agriculture as rural extension agent (2012-2015) and at the Agricultural Research Centre of Northern Benin as Research assistant (2015- 2016), allowed him to work closely with farmers, rural extension agents, researchers scientists with various profiles (agronomy, soil, social, rural economy) for the dissemination of good crop production practices. He is interested in soil fertility management and dynamic modelling of the soil-plant system.

201