MODELING PARASITIC WEED EMERGENCE ACROSS SMALLHOLDER FARMING SYSTEMS: THE CASE OF CENTRAL MALAWI By Timothy Robert Silberg A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Community Sustainability Doctor of Philosophy 201 9 ABSTRACT MODELING PARASITIC WEED EMERGENCE ACROSS SMALLHOLDER FARMING SYSTEMS: THE CASE OF CENTRAL MALAWI By Timothy Robert Silberg Four out of five households in Malawi rely on farming as a primary source of income , most of whom cultivate maize ( Zea mays ) . Disconcertingly, 63 - 80% of maize yield losses among these households are attributed to the emergence of invasive and parasitic weeds such as Striga ( Striga spp. ). A plethora of Striga - control practices (SCPs) have been developed and disseminated to smallholder farmers (cultivating < 2 ha). These SCPs are commonly evaluated at agricultural research stations prior to disse mination. Mixed results often arise later when they are implemented across the diverse agroecological and socioeconomic landscapes of smallholders. Many agree research will need to assess how SCPs perform under smallholder - conditions, and ultimately, how t heir uptake will affect emergence. The following dissertation is divided into three empirical studies. In the first essay, discrete choice experiments (DCEs) are used to estimate the percent of maize yield farmers are willing to sacrifice for different SCP attributes (e.g., labor, soil fertility). In the second essay, a seed bank stock and flow model (SB - SFM) is developed to assess emergence rates across different SCPs. In the final essay, results from the DCEs and SB - SFM are integrated within a system dyna mics model (SDM) to simulate how environmental and socioeconomic parameters affect emergence across space and time. DCE findings highlight farmers are willing to sacrifice significant tradeoffs to implement SCPs that increase soil fertility and provide leg umes . SB - SFM findings indicate the attachment phase and seed bank must simultaneously be addressed with multiple SCPs to suppress emergence over three to five years . Finally, alteration of different climate, farm - management and adoption parameters in the SDM underline that nutrient input subsidies and agricultural extension must be included in an aggregated effort to suppress the spread of Striga across the region . Copyright by TIMOTHY ROBERT SILBERG 201 9 v To Mom. The woman that put me in the garden first. I love you. vi ACKNOWLEDGEMENTS This work would have not been possible without the support, both academic ally and personal ly , from countless people and organizations. First and foremost, I would like to thank my advisor, mentor and friend Dr. Robert B. Richardson. No person has transformed my view of international agricultural development and taught me more about honest researc h, always acknowledging the strength s and limitations of methods . I am most grateful for his unwavering support during my academic and research setbacks. His knowledge of when and when not to step in at these challenging times facilitated exponential profe ssional and emotional growth. This was most evident when he critiqued my work, offering comments that never discouraged my progress, but rather, pushed me to think and write beyond a capability I knew ( or believed ) I could reach . Robby, knowing and working with you has been the highlight of my doctoral tenure. I hope I can offer someone a fraction of the support and knowledge you have given me these past four and a half years. Thank you. I would also like to express my gratitude for my advisory committee: Dr. Karen Renner, thank you for welcoming an outsider and teaching me about the intricacies of weed persistence and prevalence. The university could use more professors that are as dedicated to teaching as you are . Dr. Laura Sch mi t t Olabisi, thank you for introducing me to the world of system dynamics. I hope you will continue to transform student view s about global issues as you have mine. Dr. Maria Claudia Lopez, thank you for always pushing me to create my own theoretical fr amework . I look forward to our future collaborations researching farmer - decisions and the institutions that guide them. Dr. Bruno Basso, thank you for opening the door to crop modeling vii and bestowing upon me your breadth of experience in agroecology and cli mate variability. Your practical approach to building models and adapting them to real - word outcomes is something I hope to instill in my work for years to come. Quite often graduate students have a so - called secondary researcher committee. While these in dividuals may not be listed as co authors in their dissertation articles, they are the unsung heroes of academia. I would be remiss . Dr. Vimbayi Grace Petrova Chimonyo, thank you for your countless ad - hoc crop modeling and smallhold er agriculture lessons. You picked me up during my darkest hours. I wonder where smallholder agriculture would be today if it was filled with more individuals who are as honest, brilliant and selfless as you. Dr. Andrew Ford, thank you for your unwavering patience with my modeling. My crop model and dissertation would be nowhere near where it is today without your plethora of suggestions and comments . Drs. Vincenzina C aputo and Aniseh Sjona Bro, thank you for assisting me with my quant it ative analyse s. Chun - Lung Lee and Awa Sanou, thank you for discussing these findings with me so I could understand them . Drs. Erin Anders and Phil Grabowski , thank you for assisting m e w ith my instrument development and field research plans. And finally, but not least, Drs. John Kerr and Phil Howard. Thank you for your thoughtful critiques of my writing early in the doctoral program. You set the bar for how to become a concise and clear w riter . I am sure Robby appreciated it. My family has been the rock and foundation for this dissertation, and with it, the doctoral program . Their love never faltered when I needed it most. A doctoral pursuit is sometimes the most selfish time for many , opt ing to finish assignments instead of coming home to celebrate holidays, birthdays and other important family gatherings. Even when these individuals are viii physically present at these gatherings , they are absent mentally, preoccupied with looming deadlines and exams. Not once did my family guilt me for these preoccupations, but only offered empathy. Mom and Dad, thank you for always supporting my passion, even if it took me thousands of miles away from you for ex tended periods of time. Few can say they do what they love and I know, without reservation, I can say this because of your support. You have sacrificed so much for my happiness. Please know that this dissertation is a product of that sacrifice, but more so , your lessons of love and compassion for each other. Thank you to my three older brothers Jonny, Nicky and Justin. Even if I do not say it, I have always looked up to you, whether it was watching you become a department chair and an incredible father, a l eader and respected mentor in your profession, or a successful lawyer that defied countless setbacks. These journeys and tribulations showed me how to approach my challenges previously, now and in the future . I am a better man for it. I love all of you ver y much. This dissertation is also a product of my Michigan family, the Department of Community Sustainability and wider Michigan State University community. Thank you, Kyle Metta , Udita Sanga, Brockton Feltman, Adebiyi (Gana) Jelili Adegboyega, Obafemi El egbede, Katherine Groble, Alison Singer, Eva Kasara, Payam Aminpour, Nathan Brugnone, Laura Castro Diez, Aldo Gonzalez, Andrew Gerard, Drs. Kim Chung, Chewe Nkonde, Jenny Hodbod, Celina Wile and Steven Gray (just to name a few). In some shape or form , you offered a space for interdisciplinary scholarship and lent an ear to a struggling student who needed to think out loud. Many of these models and explanations are a product of our conversations in front of a white board, in your office , or just in passing. Thank you sacrificing your time to speak with me about my struggles, academically and personally. There are few organizations I have worked ix with which practice what they preach. I believe this department embodies the very idea of social justice and am proud to be a part of it. I must thank and recognize the farmers affiliated with the Africa RISING (Research for Sustainable Intensification for the Next Generation) and Total Land C are (TLC) programs. Without them, there would be no dissertation. Their honest testimonies and warm invitations to their fields allowed for a holistic analysis of Striga spp . emergence. More times than none, re search is one - sided, benefiting the investigator more than the participants of the study . While I do not believe research should always be paternalistic or solely intended to assist its participants, if that is its intent however , the instrumentation and results of the research should be ev aluated by its participants as much as possible when appropriate . That being said, I hope the findings of this dissertation will contribute to wider effort to address Striga spp . emergence not because of like, but rather, its p ractical discussion and inclusion of local knowledge. These discussions in no way would have been possible without the assistance of my colleague Cyprian Mwale, research assistant Chilungamo Banda, the agricultural extension development officers of t he c en tral region of Malawi, and the Chitedze Agricultural Research Station community. Zikomo kwam biri . Finally, I would like to acknowledge the financial support of the United States Agency for International Development (USAID) through the Borlaug Fellows in Global Food Security at Purdue University. In addition, the farmers surveyed in this dissertation were affiliated with the USAID - funded proje the Graduate School at Michigan State University by means of the Dissertation Completion Fellowship, x Critical Needs Fellowship, Graduate Research Enhancement Grant and the Gender, Justice and Environmental Change Fellowship. Lastly, outside financial support was received from the MAXQDA Research Software Company and the Claffey - Meyer Foundation. Thank you for your generous support. xi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ................... xiii LIST OF FIGURES ................................ ................................ ................................ .................. xv CHAPTER 1: INTRODUCTION ................................ ................................ ................................ .. 1 REFERENCES ................................ ................................ ................................ .......................... 7 CHAPTER 2: MAIZE FARMER PREFERENCES FOR STRIGA CONTROL PRACTICES IN MALAWI .... 11 2.1 Introduction ................................ ................................ ................................ ........................... 11 2.2 Background ................................ ................................ ................................ ............................ 14 2.2.1 Context specific weeding practices in Malawi ................................ ................................ .......................... 15 2.2.2 Determinants of practice ................................ ................................ ................................ ........................... 16 2.2.3 Understanding smallholder decisions under resource - poor conditions ................................ ................... 22 2 .3. Empirical Model ................................ ................................ ................................ .................... 24 2.3.1 Random utility theory ................................ ................................ ................................ ................................ 25 2.3.2 Random parameter logistic regression ................................ ................................ ................................ ..... 26 2.4 Methods ................................ ................................ ................................ ................................ 29 2.4.1 Discrete choice experiments ................................ ................................ ................................ ..................... 29 2.4.2 Choice design ................................ ................................ ................................ ................................ ............. 31 2.4.3 Site description ................................ ................................ ................................ ................................ .......... 46 2.4.4 Sampling procedures ................................ ................................ ................................ ................................ . 49 2.4.5 Instrument calibration and protocol ................................ ................................ ................................ ......... 52 2.5 Results & Discussion ................................ ................................ ................................ ............... 53 2.5.1 Descriptive statistics ................................ ................................ ................................ ................................ .. 53 2.5.2 Marginal value of Striga control attributes ................................ ................................ ............................... 55 2.5.3 Gender level differences ................................ ................................ ................................ ........................... 60 2.6 Conclusions ................................ ................................ ................................ ............................ 64 APPENDICES ................................ ................................ ................................ ........................ 68 APPENDIX 1. Focus Group Instrument ................................ ................................ ............................ 69 APPENDIX 2. Survey Instrument ................................ ................................ ................................ .... 73 APPENDIX 3. Key for Survey Questionnaire ................................ ................................ ..................... 92 REFERENCES ................................ ................................ ................................ ........................ 99 CHAPTER 3: CROP MODELING: AN INTEGRATED APPROACH TO SIMULATING EMERGENCE ....... AND PERSISTENCE OF STRIGA ASIATICA ................................ ................................ ............. 112 3.1 Introduction ................................ ................................ ................................ ......................... 112 3.2 Background ................................ ................................ ................................ .......................... 114 3.2.1 Crop Models, Their Required Inputs & Selection Considerations ................................ ........................... 114 3.2.2 Factors to Consider When Modeling Weed Emergence ................................ ................................ ......... 116 3.3 Methods ................................ ................................ ................................ .............................. 128 3.3.1 Model review ................................ ................................ ................................ ................................ ........... 128 3.3.2 Model development ................................ ................................ ................................ ................................ 136 3.3.3 Greenhouse experiment ................................ ................................ ................................ .......................... 141 3.3.4 Analysis ................................ ................................ ................................ ................................ .................... 150 xii 3.4 Results & Discussion ................................ ................................ ................................ ............. 154 3.4.1 Soil ................................ ................................ ................................ ................................ ........................... 154 3.4.2 Emergence ................................ ................................ ................................ ................................ ............... 157 3.4.3 Model runs ................................ ................................ ................................ ................................ ............... 159 3.5 Conclusions ................................ ................................ ................................ .......................... 172 APPENDICES ................................ ................................ ................................ ...................... 177 APPENDIX 1. Justification off Values/Equations Applied to Parameters of Cropping Sy stem Model .. 178 APPENDIX 2. Seedbank Calculation ................................ ................................ .............................. 198 APPENDIX 3. Root Canopy Calculation ................................ ................................ ......................... 199 REFERENCES ................................ ................................ ................................ ...................... 200 CHAPTER 4: SYSTEM DYNAMICS - COMBINING CHOICE EXPERIMENTS AND CROP SIMULATION TO MODEL PARASITIC WEED EMERGENCE ................................ ................................ .......... 215 4.1 Introduction ................................ ................................ ................................ ......................... 215 4.2 Background ................................ ................................ ................................ .......................... 217 4.3 Methods ................................ ................................ ................................ .............................. 221 4.3.1 Theoretical framework ................................ ................................ ................................ ............................ 221 4.3.2 Methodological framework ................................ ................................ ................................ ..................... 223 4.3.3 Data collection ................................ ................................ ................................ ................................ ......... 235 4.3.4 Data analysis ................................ ................................ ................................ ................................ ............ 239 4.4 Findings and discussion ................................ ................................ ................................ ........ 240 4.4.1 Summary statistics ................................ ................................ ................................ ................................ ... 241 4.4.2 Participant causal loop diagrams ................................ ................................ ................................ ............ 251 4.4.3 Participant system dynamics models ................................ ................................ ................................ ...... 252 4.4.4 System dynamics model ................................ ................................ ................................ .......................... 254 4.4 Conclusions ................................ ................................ ................................ .......................... 267 APPENDICES ................................ ................................ ................................ ...................... 271 APPENDIX 1. Enumerator Demonstrating Choice Tasks ................................ ................................ . 272 APPENDIX 2. Justification of Values/Equations Applied to Parameters for Syste m Dynamics Model Of Striga Emergence and Control Practice Implementation ................................ ................................ 273 REFERENCE S ................................ ................................ ................................ ...................... 280 CHAPTER 5: CONCLUSIONS ................................ ................................ ................................ 289 xiii LIST OF TABLES Table 1 - Striga control attributes used in choice experiments ................................ ................... 41 Table 2 - Definition of variables used in choice model ................................ ................................ . 43 Table 3 - Example of dummy and effect coding schemes for labor requirement attributes ....... 44 Table 4 - Farmers expressed Striga as a primary challenge in 2016 (out of a list of 15 productivity challenges) ................................ ................................ ................................ ........ 50 Table 5 - Sample characteristics ................................ ................................ ................................ ... 53 Table 6 - Random parameters logit model and willingness to pay space for Striga control practices ................................ ................................ ................................ ................................ 55 Table 7 - Correlation matrix for random parameters logit model in Table 6 ............................... 58 Table 8 - Willingness to pay space for Striga control practices across gender ............................ 61 Table 9 - Participant quotes related to food security and financial costs and preference of Striga control practices ................................ ................................ ................................ ................... 62 Table 10 - Participant quotes related to gender roles and preference for Striga control practices ................................ ................................ ................................ ................................ ............... 63 Table 11 - Participant quotes related to labor and preference for Striga control practices ....... 63 Table 12 - Details of Plot Management ................................ ................................ ...................... 144 Table 13 - Number of soils samples per treatment, depth and field location (ridge vs furrow) 145 Table 14 - Details of greenhouse sample ................................ ................................ ................... 147 Table 15 - Location in greenhouse based on sampling location ................................ ................ 148 Table 16 - Soil analyses of practices by length of implementation and soil depth .................... 154 Table 17 S. asiatica emergence by practice, length of implementation and soil depth ......... 157 Table 18 - Characterization of Striga control metho ds for sub - Saharan African smallholders .. 220 Table 19 - Methodological framework ................................ ................................ ....................... 224 xiv Table 20 - Source used for parameter equation/values. ................................ ............................ 239 Table 21 - Participant characteristics according to Striga spp. prevalence on HH farm ............ 242 Table 22 - Sources SCPs were heard and implemented from ................................ .................... 244 Table 23 - Popular SCPs implemented across information sources ................................ ........... 246 Table 24 - Common outcomes received by implementing SCPs ................................ ................ 248 Table 25 - Extent of positive and negative outcomes shared with peers ................................ .. 250 xv LIST OF FIGURES Figure 1 - Control trait preferences related to control method preferences across gender ....... 35 Figure 2 - Data collection sites ................................ ................................ ................................ ...... 46 Figure 3 - Sample choice task ................................ ................................ ................................ ....... 48 Figure 4 - Lifecycle of Striga asiatica ................................ ................................ ........................... 120 Figure 5 - Practices that address Striga emergence based on the stage of the weed lifecycle . 126 Figure 6 - Structure and flow between state variables of Striga spp. ................................ ........ 129 Figure 7 - Life cycle diagram of Striga hermonthica ................................ ................................ ... 131 Figure 8 - Flow diagram of parasitic weed crenate broomrape (Orobanche crenala Forsk.) in APSIM ................................ ................................ ................................ ................................ .. 133 Figure 9 - Density - dependent feedback model (DDFM): Striga hermonthica emergence in sorghum - based system ................................ ................................ ................................ ....... 135 Figure 10 - Cropping Systems Model (CSM): Emergence of Striga asiatica in maize - based system ................................ ................................ ................................ ................................ ............. 137 Figure 11 - Sampling procedure conduct in farmer plot ................................ ............................ 145 Figure 12 - Greenhouse experiment at Chitedze Agricultural Research Center ........................ 147 Figure 13 - Greenhouse average daily temperature during experiment ................................ ... 158 Figure 14 - Base case runs: seedbank (a), germination (b), attachment - emergence - flowering (c) ................................ ................................ ................................ ................................ ............. 160 Figure 15 - Monthly unattached seedlings in association with varying germination rates ....... 162 Figure 16 - Monthly flowers in association with varying attachment rates ............................... 163 Figure 17 - Monthly surface seeds in association with varying successful emergence fractions ................................ ................................ ................................ ................................ ............. 164 Figure 18 - Monthly attachments in association with varying flowering success fractions ....... 164 xvi Figure 19 - Reductions in monthly Mature Striga (a) and surface seed (b) in response to weeding ................................ ................................ ................................ ................................ ............. 165 Figure 20 - Monthly unattached seedlings in response to basal manure application ............... 166 Figure 21 - Monthly seedbank (a), unattached seedling s (b) and mature Striga (c) reductions in response to cowpea intercropping ................................ ................................ ..................... 168 Figure 22 - Monthly attachment (a), seedling/flow ering (b) and seedbank (c) reductions in response to N - based basal fertilizer application at planting ................................ .............. 169 Figure 23 - Monthly surface seed (a) and attachment (b) reductions in response to weeding + manure application ................................ ................................ ................................ ............. 170 Figure 24 - Monthly surface seed (a) and mature Striga (b) reductions in response to weeding + manure application + legume intercropping VS weeding + manure application + fertilizer application ................................ ................................ ................................ ........................... 171 Figure 25 - Monthly surface seed (a), attachment (b) and flowering (c) reductions in response to weeding + manure application + legume intercropping VS wee ding + manure application + fertilizer application ................................ ................................ ................................ ............ 172 Figure 26 - Bass adoption model of cereal - legume intercropping ................................ ............. 222 Figure 27 - Reduced view of CSM (attached seedlings stock circled in purple) ......................... 230 Figure 28 - Individual causal loops diagrams digitized in Vensim ................................ .............. 232 Figure 29 - CA indicates conservation agriculture practices and CV indicates conventional farming practice ................................ ................................ ................................ .................. 235 Figure 30 - Causal loop diagrams created by farmers and AEDOs (a), private sector participants (b) and policy makers (c) ................................ ................................ ................................ ..... 252 Figure 31 - System dynamics models created by farmers and AEDOs (a), private sector participants (b) and policy makers (c) ................................ ................................ ................. 254 Figure 32 - System dynamics model of Striga emergence and control practice implementation ................................ ................................ ................................ ................................ ............. 256 Figure 33 - Implementation (a) and Striga emergence (b) base on rainfall (c) .......................... 258 Figure 34 - Sensi tivity of implementer (a), discontinuer (b), abandoner (c) and parasitized field (d) populations to seed spread rates of Striga spp. ................................ ............................ 260 xvii Figure 35 - Sensitivity of dis - implementer population to observation rate ............................... 261 Figure 36 - Implementer (a) and emergence (b) population response to favorable rainfall ..... 263 Figure 37 - Emergence (a) and implementer population that receives a positive outcome (b) in response to fertilizer subsidy ................................ ................................ .............................. 264 Figure 38 - Emergence (a) and implementer (b) population response to increased agricultural extension ................................ ................................ ................................ ............................. 266 Figure 39 - Emergence (a), impl ementer (b) and abandoner (c) response to fertilizer subsidy & increased agricultural extension ................................ ................................ ......................... 267 1 CHAPTER 1: INTRODUCTION In much of s ub - Saharan Africa (SSA) , farming is a primary livelihood for rural society (Garrity et al., 2010). The agrarian population is mainly comprised of smallholders (those cultivating less than two hectares), representing 80% of the far ms across the region (Altieri et al., 2012). Approximately 33 million farmers commonly cultivate crops such as maize, millet and sorghum (Tafirenyika, 2014). Thus, cereal production is commonly viewed as an indicator for rural food security and wealth, especially in countries like Malawi (UNICEF, 2013) . Cereal production has been supported by a numbe r of policies and institutions in SSA, particularly in Malawi. For example, many times fertilizer subsidies are made available to farmers who cultivate hybrid maize as opposed to other food crops (Garrity et al., 2010). In addition, dietary norms have long - encouraged the cultivation of soil - erosive crops like maize. In conjunction with these policies and institutions, population growth and unequal distribution of land have obligated smallholders to intensify their monocultures of maize (Bezner Kerr, 2005; G ilbert, 2004; Hockett & Richardson, 2016). As these maize - based systems are intensified, application rates of synthetic fertilizers and sowing rates per hectare are increased. As a result, more soil organic matter and soil - N ar e removed than can be replace d ( Heinrichs et al. 1995). Under such conditions, competition is increased for nutrients and maize is susceptible to invasion by weeds (Gigou, 1992). In southern Africa, it is estimated that 63 to 80% of maize yields are lost due to competition for nutri ents and the removal of water by parasitic weeds (Parker, 2012). One of the most prevalent parasitics in the region is commonly known as witchweed or Striga ( Striga spp. ). As an 2 obligate hermiparasitic angiosperm, witchweed is unable to fully access minera ls, photosynthates and water by individual growth, therefore, requiring a host (e.g., maize) to obtain these resources ( Midega et al., 2013). After maize develops a well - established root system (4 - 6 weeks after sowing) , witchweed will attach to the rootstock and cause a phytotoxic effect, removing nutrients and water taken up by maize. In addition, Striga spp . will compete for nutrients in the soil later when fully grown (30 - 40 days after emerging from the soil) . As a c onsequence, maize plant height, biomass and grain yield are drastically reduced ( Frost et al., 1997; Gurney et al., 1999 ). Copious seed production and a long - lived seed - bank allows witchweed to rapidly invade and iods of time. Seeds can remain viable in the soil for ten years, waiting for sorghum or maize to be planted under favorable soil conditions for germination (e.g., sandy acidic soil, 30 - 35C o ) (Khan et al., 2010). After emerging from the soil, one plant can produce thousands of seeds, spreading by wind, water, and/or cultural practices (Khan et al., 2002). Much of these infestations can be deterred when lands are left to fallow, but arable - land scarcity and a long history of cultivating and consuming maize ha s made such methods impractical to smallholders ( Bezner Kerr, 2005; Kureh et al., 2006). O ther agricultural practices such as crop rotation with green manure legumes have been proposed to reduce Striga spp. infestations . Unfortunately, some of these legumes can attract pests that consume maize leaves or are associated with bad luck, making them difficult to adopt (Forsythe et al., 2015; Sileshi et al., 2000). Based on the aforementioned cases, it appears then, without in cluding input from smallholders to develop weed control strategies, parasitic weed emergence and land abandonment will likely ensue (Berner, 1995; Connelly, 1994) . In 3 addition to gaining smallholder - input to reduce witchweed emergence, the social, environm ent and financial context parasitic weeds proliferate in must be understood. Given the vulnerability of smallholder farms to witchweed, a large consensus agrees research is needed to develop adaptive strategies that provide food and/or revenue to farmers a nd control parasitic weeds under resource - limited conditions ( Debra, 1994; Johnson, 1996; Khan et al., 2010; Orr et al., 2002; Orr et al., 2009; Riches et al., 2005 ). The dissertation explores the implications technological attributes have on farmer choic es for practices and how the choice of implementing one or several practices a ffect the lifecycle of Striga asiatica in a Malawian smallholder setting. The study is carried out in a consecutive manner, investigating Striga emergence across three successive dimensions: control preferences, control simulation (via crop modeling ) and control diffusion based on the two aforementioned dimensions. The dissertation is split into three empirical essays . E ach essay investigates parasitic weed emergence differently a nd are intended to be published as three separate manuscripts. Findings from one essay often informs the instrumentation or findings of the following essay. For these reasons, the description of one study or its data may appear partially repetitive across chapters. Chapter 2 , Maize Farmer Preferences for Striga Control Practices in Malawi, explores the primary traits of Striga control practices (SCPs) smallholders consider prior to implementation. Thereafter, the study determines which tradeoffs smallhold ers are willing (or not willing) to make to implement a SCP (e.g., increased labor for reduced Striga emergence). While numerous studies have documented the inputs required to execute several parasitic controls strategies, little research has investigated the socioeconomic drivers behind their use or the barriers that 4 impede their implementation. The study employs focus groups to identify SCP attributes (e.g., labor days, maize yield) and conducts discrete choice experiments (DCEs) to quantify the percent o f maize yield farmers are willing sacrifice for these attributes. Findings indicate that lower Striga emergence and labor requirements as well as increased soil fertility and legume yield significantly influenced the decision to select a SCP across 215 par ticipants. Female and male far m ers were willing to sacrifice different percentages of their maize yield for higher legume yield and increased soil fertility. Understanding these tradeoffs informs researchers how to better align SCPs with desired outcomes a nd ensure they are implemented once they are disseminated. Chapter 3 , Systems Modeling: An Integrated Approach to Simulating Emergence and Persistence of S. asiatica , investigates the underlying feedback behavior in the S. asiatica lifecycle. Uncovering which stages of the lifecycle drive emergence and the accumulation of the seed bank inform when, where and how to address the weed with various farming practices. The study develops a cropping systems model (CSM) from previous Striga spp . models found in the literature. I nterviews with Malawian scientists confirm the parameterization of the model. Local climatic data and findings from previous S. asiatica studies apply values and equ ations to model parameters. Emergence rates in farmer soils are used to calibrate the output of the model. Results from model runs reveal that an integrated approach is needed to manage the parasitic weed under smallholder conditions. In addition, the bott leneck behavior in the model highlight s the importance of focusing control ef forts on attachment rather than germination, emergence or flowering. Given the devastating effects witchweed has had in Malawi, it is imperative to develop parasitic weed modules for low - cost crop simulators to better evaluate 5 smallholder technologies (Ejeta, 2007). Models that do not capture underlying mechanisms in the weed lifecycle, risk informing extension agents with potentially misleading or ineffective practices to deliver to farmers. Chapter 4 , System Dynamics: Combining Choice Experiments and Crop Simulation to Model Parasitic Weed Emergence , studies the dynamic behavior of S. asiatica emergence based on the implementation of its control strategies. The implementation of SCPs is influenced by an interlinked natural, financial and social environment, making them dynamic as well (Debra, 1994). Feedback behavior between SCP implementation and S. asiatica emergence is studied using survey questionnaires, mediated modeling and system dynamics. A system dynamics model (SDM) is parameterized from several adoption models found in the literature as well as input from various stakeholders collecte d at a mediated modeling workshop (Bass et al., 2000; Kopainsky et al., 2012). Parameters in the SDM are applied with values and equations from summary statistics gathered from survey questionnaires, utility coefficients calculated from DCM data and weed e mergence readings from the CSM. The potential for SCP implementation is reduced largely by the stochasticity of maize yields across seasons combined with significant social pressure to abandon these practices. Low yields suppress implementation and increas e abandonment due to the dynamics of utility in SCPs. A critical factor in explaining low implementation rates of agricultural technologies is the stochasticity in their performance (and in this case - yield) (Bahmanziari et al., 2003) . Understanding how th at stochasticity interacts with the social dynamics of learning and communicating about the performance among users and protentional users is critical to successfully disseminate SCPs. 6 The objective of this dissertation is to highlight the processes behind parasitic weed emergence in Malawi based on interlinked biophysical and socioeconomic factors. These processes and other findings generated from the dissertation are valuable to farmers, development and extension practitioner s, policy makers and other stakeholders in the smallholder cereal production sector. The findings are intended to highlight critical areas to address the growing problem of Striga spp. as well as guide policy of how to do so. More specifically, findings fr om Chapter 2 provide s technology disseminators knowledge regarding tradeoffs farmers are willing (or not willing) to make to implement various Striga control practices. With t his information, extension agents can better diffuse agricultural practices to fa rmers. Second, Chapter 3 findings will highlight the degree of each practice required to suppress Striga. The results will be most valuable to farmers given the limited capital they have to carry out a limited number of practices. In addition, the developm ent of a weed module will be the first of its kind, benefiting crop modelers across the globe that use systems models . Third, results from various climatic, farmer management and policy scenarios shown in Chapter 4 may shed light on when, where and how lon g interventions will need to be implemented to significantly reduce Striga. Such information will offer key insights to policy makers wishing to know which programs to fund in order reduce weed prevalence and subsequent food insecurity. 7 REFERENCES 8 REFERENCES Altieri, M. A. (2002). Agroecology: the science of natural resource management for poor farmers in marginal environments. Agriculture, Ecosystems & Environment , 93 (1), 1 - 24. Bahmanziari, T., Pearson, J.M., Crosby, L., 2003. Is trust important in technology adoption? A policy capturing approach. The Journal of Computer Information Systems 43, 46. Bass, F.M., Jain, D., Krishnan, T. (2000). Modeling the marketing - mix influence in new - product Diffusion. In Mahajan, V., Muller, E., & Wind, Y. (Eds.). New - Product Diffusion Models . Boston, MA: Kluwer Academic Publishers. Berner, D. K., Kling, J. G., & Singh, B. B. (1995). Striga research and control. A perspective from Africa. Plant Disease , 79 (7), 652 - 660 . Bezner Kerr, R. (2005). Food security in northern Malawi: Gender, kinship relations and entitlements in historical context . Journal of Southern African Studies, 31 (1), 53 74. Conelly, W. T. (1994). Population pressure, labor availability, and agricultural disintensification: the decline of farming on Rusinga Island, Kenya. Human Ecology , 22 (2), 145 - 170. Debra, S.K. (1994). Socio - economic constraints to the adoption of weed control technologies: The case of Str iga control in West African Semi Arid Tropics. International Journal of Pest Management, 40, 153 - 158. Ejeta, G., (2007). The Striga Scourge in Africa: A Growing Pandemic. In Ejeta, G., Gressel, J. (Eds), Integrating New Technologies for Striga Control . Singapore: World Scientific Publishing. Forsythe, L., Nyamanda, M., Mbachi Mwangwela, A., & Bennett, B. (2015). Beliefs, taboos and minor crop value chains: the case of Bambara Groundnut in Malawi. Food, Culture & Society , 18 (3), 501 - 517. Frost , D.L., Gumey, A.L., Press, M.C. & Scholes, J.D. (1997). Striga hermonthica reduces photosynthesis in sorghum: the importance of stomatal limitations and a potential role for ABA?. Plant Cell and Environment, 20 , 483 - 492. Garrity, D. P., Akinnifesi, F. K. , Ajayi, O. C., Weldesemayat, S. G., Mowo, J. G., Kalinganire, A., & Bayala, J. (2010). Evergreen Agriculture: a robust approach to sustainable food security in Africa. Food Security , 2 , 197 214. Gigou, J. J. (1992). L'azote dans les systèmes de culture d u nord et du centre de la Côte d'Ivoire. Biological Nitrogen Fixation and Sustainability of Tropical Agriculture , 377 - 394. 9 Gilbert, R. A. (2004). Best - bet legumes for smallholder maize - based cropping systems of Malawi. In Eilittä, M., Mureithi, J., and De rpsch, R. (Eds), Green Manure/Cover Crop Systems of Smallholder Farmers: Experiences from Tropical and Subtropical Regions , (pp. 153 174). Dordrecht: Kluwer Academic Publishers. Grenz, J. H., Manschadi, A. M., DeVoil, P., Meinke, H., & Sauerborn, J. (2005). Assessing Strategies for sp. Control Using a Combined Seedbank and Competition Model. Agronomy Journal , 97 (6), 1551 - 1559. Gurney, A. L., Press, M. C., & Scholes, J. D. (1999). Infection time and density influence the response of sorg hum to the parasitic angiosperm Striga hermonthica. New Phytologist , 143 (3), 573 - 580. Heinrichs, E. A., Johnson, D. E., Afun, J. K., & Robison, D. J. (1995). Rice pests of shifting cultivation in Côte d'Ivoire, West Africa. In International Rice Research Conference on Fragile Lives in Fragile Ecosystems. Los Banos, Laguna (Philippines) : IRRI. Hockett, M., & Rich ardson, R.B. (2016). Examining the Drivers o f Agricultural Experimenta tion Among Smallholder Farmers i n Malawi. Experimental Agriculture , 1 - 21. Johnson, D. E. (1996). Weed management in small holder rice production in the tropics. IPM Word Textbook, University of Minnesota, St. Paul, MN. Retrieved from http://ipmworld .umm.edu/chapters/johnson.htm . Khan, Z. R., Hassanali, A., Overholt, W., Khamis, T. M., Hooper, A. M., Pickett, J. A., ... & Woodcock, C. M. (2002). Control of witchweed Striga hermonthica by intercropping with Desmodium spp., and the mechanism defined as allelopathic. Journal of Chemical Ecology , 28 (9), 1871 - 1885 . Khan, Z. R., Midega, C. A., Bruce, T. J., Hooper, A. M., & Pickett, J. A. (2010). Exploiting in Africa. Journal of Experimental Bot any, 61 (15), 4185 - 4196 . Utility Evaluations for Good and Bad. Systems Research an d Behavioral Science , 29 (6), 575 - 589. Kureh, I., Kamara, A. Y., & Tarfa, B. D. (2006). Influence of cereal - legume rotation on Striga Nigeria. Journal of Agriculture and Rur al Development in the Tropics and Subtropics (JARTS) , 107 (1), 41 - 54. 10 Midega, C. A., Pittchar, J., Salifu, D., Pickett, J. A., & Khan, Z. R. (2013). Effects of mulching, N - fertilization and intercropping with Desmodium uncinatum on Striga hermonthica infestation in maize. Crop Protection , 44 , 44 - 49. Orr, A., Mwale, B., & Sa iti, D. (2002). Modelling agricultural 'performance': smallholder weed management in Southern Malawi. International Journal of Pest Management , 48 (4), 265 - 278. Orr, A., Mwale, B., & Saiti - - we ek The Journal of Development Studies , 45 (2), 227 - 255. Parker, C. (2012). Parasitic weeds: a world challenge. Weed Science , 60 (2), 269 - 276. Riches, C. R., Mbwaga, A. M., Mbapila, J., Ahmed, G. J. U., Harris, D., Richards, J. I. , ...& Wircombe, J. R. (2005). Improved weed management delivers increased productivity and farm incomes from rice in Bangladesh and Tanzania. Aspects of Applied Biology , 75 , 127 - 138. Sileshi, G., Maghembe, J. A., Rao, M. R., Ogol, C. K. P. O., & Sithanantham, S. (2000). Insects feeding on Sesbania species in natural stands and agroforestry systems in southern Malawi. Agroforestry Systems , 49 (1), 41 - 52. Africa Ren ewal. Retrieved from http://www.un.org/africarenewal/magazine/january - 2013/what - went - wrong - lessons - malawi%E2%80%99s - food - crisis UNIC http://www.unicef.org/infobycountry/malawi_statistics.html. 11 CHAPTER 2: MAIZE FARMER PREFERENCES FOR STRIGA CONTROL PRACTICES IN MALAWI 2.1 Introduction In southern Africa, it is estimated that 63 - 80% of maize ( Zea mays ) cropping systems are parasitized by witchweed ( Striga spp. ) (Parker, 2012). As a hemiparasite 1 , the weed attaches to the maize rootstock, removing water and nutrients, and consequently devastating yields ( Frost et al., 1997). In Malawi, maize is widely cultivated by smallholder farmers (cultivating less than two hectares), many of whom rely on the crop as a staple food and primary source of income (Garrity et al., 2010). Over time, the repeated cultivation of this c ereal as a monoculture can reduce soil organic matter (SOM), augment soil - nitrogen (N) loss and create conditions for parasitic weeds to proliferate (Hakansson, 1982; Gigou, 1992). In addition, rapid population growth, abandonment of traditional fallow per iods and minimal organic inputs application have exacerbated soil erosion, allowing parasitic weeds to become ubiquitous (Franke et al., 2004; Kureh et al., 2006; United Nations, 2014). In the savannas of Southern Africa, Striga spp . is found in association with sorghum ( Sorghum bicolor ), millet ( Pennisetum glaucum ), and maize ( Akobundu, 1980). As an obligate parasite 2 , the weed is unable to fully access minerals, photosynthates and water from individual growth, requiring a host to facilitate its development . There is no light requirement for the plant, but germination is more prevalent in less - fertile sandy acidic soils, hence their omnipresence across the intensely cultivated soils of Af rica (Singh et al., 1997). Seeds generally require a ~2 - week 1 A parasite that is capable of photosynthesis, but relies on host plants for a significant portion of their carbon supply, sequestering water and nutrients (Rich & Ejeta, 2008) 2 An organism that cannot complete its life - cycle without exploiting a suitable host 12 wet - conditioning period. O ptimum day/night temperatures for germination and attachment are 15 and 20 ° C, respectively (Baskin & Baskin, 1998). According to temperature and water conditions, the we ed can thrive in vast number of low - altitude agricultural ecosystems , for instance in rainfed fields or in rice paddies. As the soils of such regions become more degraded, cereals excrete leachates, signaling mycorrhizal fungi to assimilate phosphorous in exchange for carbohydrates (Hudu & Gworgwor, 1998). Unfortunately, these leachates also catalyze Striga germination. Thus, as fields become more degraded, more leachates are excreted, increasing parasitization. Numerous parasitic control practices have been developed and disseminated to smallholders cultivating cereal - cropping systems (CCS) (Dugje et al., 2008). Some of these practices include applying pre - emergent herbicides, planting weed - resistant crop varieti es and/or micro - dosing crops with fertilizer (Oswald, 2002). Wealthier farmers tend to benefit from these strategies given their capital, yet poorer farmers who make up the primary population in Malawi, have little opportunity to benefit. When parasitic we eds germinate in high - input systems, their effects are less devastating (Doggett, 1984). In low - input systems however, 30 - 100% loss can occur, leaving 50 million hectares and 300 million African farmers with annual losses of $7USD billion per year (Parker, 2009). Given the vulnerability of such a large population to these losses, research will need to develop more adaptive strategies to control weeds under resource - limited conditions while simultaneously providing food and/or revenue to the smallholder (Deb rah, 1994). One Striga control strategy that has been proposed to address these needs is the incorporation of legumes in combination with mulching and/or minimum tillage. The combination of mulching, minimum tillage and intercropping/rotating legumes in CC S has 13 not only drawn attention as a Striga control strategy, but also as a means to improve soil fertility, provide food or fodder and supplement farm income (Kumwenda et al., 1996). Employing these practices controls Striga in several ways. First, certain legumes have the ability to chemically inhibit germination by exuding substances from their roots. In the presence of legumes, some parasitic seeds will germinate absent of a host, dying and consequently depleting their soil seed bank over several growing seasons (Khan et al., 2010). Second, the seeds of parasitic weeds lose viability each season by rotating cereals with non - hosts (e.g., legumes) (Ransom 2000). S ome rotation crops and their mulches immobilize soil - phosphorous (P), reduce soil erosion and i ncrease overall soil fertility, all of which are negatively associated with Striga emergence (Cechin & Press, 1993; Schultz et al., 2003). Finally, minimum tillage decreases incidences of bringing dormant seeds to soil depths where they can germinate. Des pite the numerous benefits parasitic weed controls provide, their use has been minimal in Malawi. Many were once widely practiced across Malawi in the past. For instance, maize was commonly intercropped with pulses until the late 1960s (Heisey & Smale, 199 5). Reasons for why farmers have abandoned or continued traditional practices involving legumes are difficult to ascertain, as they have rarely been studied. Therefore, it is essential to understand what practice or practices farmers prefer which can contr ol Striga and why they would implement them (if at all). Prior to disseminating parasitic weed control practices to farmers, researchers must first consider number of questions. First, what attributes are farmers most concerned with when selecting a paras itic weed control practice? Then, which attributes and levels most significantly influence their selection of a control practice against others (particularly thos e without 14 legumes)? More specifically, among the significant attributes, which ones are associ ated with specific farmer types (e.g., wealthy, larger landholders)? Based on these questions, researchers can assess an unbiased estimation of individual preferences while enhancing the accuracy of farmer needs to implement these practices. B y understandi ng heterogeneous preferences among heterogeneous farmers, better recommendations can be made to policy makers regarding which attributes to invest in for purposes of encouraging Striga control. This study employs discrete choice experiments to assess the a ttributes specific farmers are most concerned with when implementing a Striga control practice. The objective of the study, therefore, is to determine which tradeoffs farmers were willing (or not willing) to make to implement a Striga control practice versus continuing their current practices. 2. 2 Background To explai n the drivers behind implementing Striga control practices, current practices and their decision - making contexts must be identified, especially barriers to their implement ation. Before discussing these three points, it should be noted that long - term Striga control practices, in many cases, entail one or a combination of soil fertility management (SFM) practices (Ransom, 2000). SFM practice s aim to improve soil structure and input use efficiency. In the process, they reduce soil erosion and improve soil structure, which, in turn, create less - favorable conditions for Striga ( Esila ba et al., 2000). Conversely, conditions favoring Striga germination are characterized as nutrient poor soils exhibiting low productivity, many of which receive low inputs of fertilizer and/or improved management practices (Oswald & Ransom, 2001) materials and mine 15 incorporation of crop residues, natural fallowing, improved fallows, relay or intercropping of integration of legumes , keep seed production from increasing significantly over a four - year time period (Ransom, 2000; Reda et al., 2005). A vast body of literature has discussed the drivers and barriers of these practices. Likewise, the drivers and barriers of Striga control pr actices, while not documented at great length, are assumed to be similar to that of SFM practices. 2.2.1 Context specific weeding p ractices in Malawi Fairly little is known about smallholder weeding practices and their efficacy across sub - Saharan Africa ( Dimes et al., 2004). Without understanding the farming systems managed by farmers, the technologies they implement, soil conditions they cultivate under, and/or recurring weed populations (just to list a few), it is difficult to provide agricultural extens ion recommendations to control Striga (Collinson, 1997). Many weeding recommendations are based on fixed or predetermined designs conducted at agricultural experiment stations (Orr et al., 2002). In addition, management at these stations is relatively unaf fected by the financial and labor constraints smallholders face at the field level. Therefore , the utility and relevance of these fixed weeding recommendations is limited and potentially erroneous. Rural farmers are well aware of the repercussions of not weeding effectively, but much of the challenge to weed completely or in a timely manner arises from labor and financial constraints (Kumwenda, 1997). One of the most critical times for smallholders t o weed their cereal cropping systems is three to four weeks after sowing. Unfortunately, this period coincides with the time when food supply and finances have dwindled from last season. During that time, 16 available labor is typically allocated for off - farm employment (e.g., wage labor) to supplement finances for food before harvest (Giller et al., 2011). Thus, farmers become stuck in a vicious rather than their own. Malawi is one of the few African countries where weeding practices have been documented in great detail (e.g., illustrations, applications) (Orr and Ritchie, 2004). In several studies, researchers have found Malawians employ a complex set of weeding pract ices 3 according to specific contexts (Orr et al., 2002). The local language, Chichewa, has no fewer than 36 different attributed to the natural, financial and/or soci al circumstances farmers face (Tafirenyika, 2014). Several researchers have compiled reports detailing descriptions of these manual techniques, many of which agree are highly advanced, but lack punctuality, allow ing weeds like Striga to flower, reproduce, and reemerge (Orr et al., 2002; Orr et al., 2009; Riches et al., 1993; Sileshi et al., 2008). Based on these findings, it would seem frivolous to advise farmers with limited time, income and labor to weed timelier and more often. Rat her, it may be more useful to offer alternative techniques that require less labor and coincide with current management practices, crops and agroecological/financial conditions. 2.2.2 Determinants of practice The decision to implement a n agricultural pr actice is contingent upon the social, physical and financial resources available to a farmer (Mugwe et al., 2009). Resources such as food, land, 3 Kupalira is exclusively for first weeding and kubandira (i.e., banking) for second weeding. Kukwazira (or kupala) is used on compacted soils instead of kubandira. Following kubandira kukwazir is conducted on fields where weeds have re - established. In addition, kukwazira is used for relay - crops (mbwera) such as beans ( Phaseolus vulgaris ), green pea ( Pisum sativum ) and sweet potato ( Solanum tuberosum ) (Orr et al., 2002) 17 labor and cash are constrained at different times of the year for smallholder farming households. Household de cisions to allocate these limited resources, therefore, will be influenced by their resilience against risk and the costs and benefits a technology offers (Ajayi et al., 2003). Smallholders are often considered to be risk - averse, but when the incentives of a practice reduce risk, empirical evidence has shown they will increase expenditures and time devoted to such technologies as a strategy to cope with climatic shocks ( Shiferaw & Bantilan, 2004 ). Social scientists must also consider the context of when and where a practice is implemented; thus, farming decisions are often time and space - specific (Feder, 1993). In addition, scientists must consider the motivations behind these decisions, including preferences for one or multiple attributes of a practice such as legume intercropping (e.g., weed control, soil nitrogen additions, provision of protein rich food) ( Silberg et al., 2017; Waldman et al., 2016). There are several socioeconomic, institutional and cultural factors that affect the implementation of Str iga control practices commonly mentioned in literature. 2.2.2.1 Socioeconomic Food security is often considered as a primary driver (or hindrance) of farming practice implementation. Literature suggests that households with fewer members are unable to grow enough food to satisfy caloric needs, making them more likely to seek out pra ctices to improve food security (Mugwe et al., 2009). Larger households, on the other hand, are sometimes less likely to implement new agricultural practices (e.g., a Striga control) for several reasons. For example, in less productive agricultural ecosyst ems, household labor and finances tend to be allocated for supplementing caloric needs, rather than for investing in new practices (Ajayi et al., 2007 ). Less food - secure households have shown a reluctance to employ new agricultural 18 practices as well when t hey believe they will negatively affect staple food crop yields (Adato & Meinzen - Dick, 2002) . Field size or total land holdings are typically used to estimate the determinants of SFM decisions (Marenya & Barrett, 2007). It has been argued that farmers who cultivate larger areas of land are able to experiment with new cropping systems and integrate them later if positive outcomes transpire ( Feder et al., 1985; Feder & Umali , 1993). In Malawi specifically, farmers cite that larger landholdings (or more field s) are needed to experiment with new SFM practices (Hockett & Richardson, 2016). Others add that smallholders with more land to cultivate crops will experiment with new agricultural practices frequently on marginalized or highly degraded lands, resulting i n modest short - term yield improvements (Oluoch - Kosura et al., 2001). With little improvement , larger landowners are less likely to continue using these practices. Decisions made by farmers with other trait preferences may be less affected by the timeliness of a benefit being received . For instance, farmer - decisions for perennial legume technologies in Malawi were found to be driven by long - term objectives such as higher soil fertility (Waldman et al., 2017). Income streams that support household wealth, i ncluding off - farm income, affect farming decisions. If a smallholder household receives their primary income from off - farm activities, farm - level decisions can be influenced in several ways. Some researchers argue that off - farm incomes encourage implementa tion of SFM technologies such as the integration of leguminous hedgerow species in cereal systems ( Adesina et al., 2000 ). Implementation of these technologies are encouraged by the ability to purchase seed from supplementary incomes and experiment ing with the new technology. Still , without off - farm income, households may be 19 motivated to diversify their farms with new practice s , such as maize - legume intercropping, to reduce risk. W ithout knowing the motivation of smallholder farmers, it is difficult to assess how off - farm earnings affect their decision s to implement a yield - maximizing technology. This is one factor that seems to be missing in many quantitative assessments about Striga control implementation. Beyo nd off - farm income streams, overall wealth index scores are often used to assess practice implementation. In lower - income households, family labor fulfills much of the on - farm tasks because little, if any, contract labor can be hired (Marenya & Barrett, 20 07). Thus, when a new practice demands more labor relative to current practices, uptake rates usually remain low among poorer or labor - constrained families. When labor markets are available, wealthier families are able to practice more labor - intensive agri cultural technologies by hiring contract labor as needed (Pender & Kerr, 1998). Distance to towns or urban centers providing agricultural extension and markets are often considered as an important factor in practice implementation. Exposure to agricultural extension and subsequent farmer training has shown to increase the speed at which new practices are learned and implemented (Nkonya et al., 1997). In addition, the existence of markets near communities can affect their access and sales of production from new technologies (Place et al., 2003). Much literature has covered distance to or contact with extension, but little has studied farmer perceptions or trust with recommendations made by extension. Some researchers claim that this is an important factor to consider when assessing practice implementation because farmers will be less likely to integrate new innovations in 20 their fields diffused by extension agents they believe are not - versed in farming or knowledgeable about their circumstances (Anderson & Fede r, 2004) 2.2.2.2 Institutional Institutions are defined as prevalent social rules that structure social interactions (Hodgson, 1989). Hodgson (1989) adds that institutions are often referred to as the rules of the game in society that structure incentives in human exchange. In addition, organizations of people are often considered as institutions because they are groups of individuals bound together by some common purpose to achieve a given set of objectives. As such, these rules and organizations are impor tant factors to consider when assessing practice implementation. Practices aimed to improve soil fertility, for example, are affected by land tenure and/or property rights. In African rural communities, many times village headsmen allocate land to smallholders but do not offer formal ownership (Otsuka & Place, 2001). Wi thout ownership, there is little incentive to implement technologies that will improve the fertility and monetary value of these lands (Kalaba et al., 2010). This finding highlights the importance of not only considering land tenure institutions but gender institutions as well. Others argue, however, (Adesina et al., 2000). Other gendered institutions such as the markets each sex is permitted (or not) to parti cipate in must be considered. In East Africa several studies have found that women have more control over profits gained from selling milk in the evening compared to the morning because morning milk is often sold to cooperatives and chilling plants where m en are registered members (Njuke et al., 2011). Female participation was excluded from these cooperatives. Therefore, women 21 will sell milk in the evening to neighbors and local traders. Thus, these studies conclude that the longer the distance between the output of a new technology (e.g., legume) and market for its output (e.g., grain) , the less control women have over the income generated from the technology (Njuke et al., 2011). Researchers add though, the further women participate in the supply chain wit h the yield gained from a new technology , the more likely they are to receive profit from and implement the technology . Blackie (1994) discovered that farmer coops increased access to inputs, markets and extension agents for new technologies and practices. Affiliation with such institutions might then influence practice implementation. In addition, he highlighted the fact that better - off famers preferred more independent modes of operation rather than joining groups. This may be attributed in some part due to credit access. For these reasons, institutional support for certain practices needs considerable attention when assessing drivers and barriers to implementing Striga control practices. 2.2.2.3 Cultural Household head or field manager characteristics such as gender, are often emphasized as determinants of agricultural practices. For example, in Uganda and Malawi, many times men are more knowledgeable about cash or commodity crops while women have more experience with low market value cr ops ( Njuki et al., 2011) . In this respect, tobacco or cotton are often - headed households are more likely to implement practices such as legume rotation that provide ample and diverse diets for their families (Ferguson & Mkandawire, 1993). 22 2.2.3 Understanding smallholder decisions under resource - poor conditions Researchers have identified numerous barriers to conduct Striga control practices. Drechsel et al. (2005) note the biophysical barriers that limit integrating legumes are not as great as the barriers presented by poor socioeconomic conditions. Perceptions about the costs and benefits associated with a certain practice must first be understood before assessing why households choose to invest their scarce resources into one method over another. Smallholders tend to the year. As such, different amounts of land, labor and finances are more constrained during specific times of the year (Kunze, 2000). Given the fluctuation of a given resource (e.g., labor), farmers freque ntly are unable to dedicate a sufficient amount of one resource (in a timely manner) when it is most needed (Mbaga - Semgalawe & Folmer, 2000). Many times, it is assumed that household labor is readily available during critical times of the growing season; however, its availability hinges on a range of factors. For instance, during planting and weeding times, household members will often leave the farm to seek temporary contract work (Graves et al ., 2004). While this may seem counterintuitive, income from of f - et al ., 2001). When labor is available, decisions to implement new practices are further truncated when members are not skilled enough to fulfill the complex tasks of a Striga control practice (Bartel & Lichtenberg, 1987; Bonaban - Wabbi & Taylor, 2012). While highly beneficial, many times the tradeoffs between a Striga - term losses and long - term benefits are too great to bear. To elaborate, one of the largest barriers for smallholders to implementing an SFM practice lies in its delayed returns on investment 23 crop in Malawi) cannot be delivered during the first or second year, farmers that rely on short - term gains (from annual cropping systems) are less likely to implement SFM practices (Nowak, 1987). Barriers to implement Striga practices can also be compounded when rights to land ownership are restricted (Fenske, 2011). For example, household heads often show little interest to invest in practices promising long - term soil fertility benefits when their children are not permitted to inherit matrilineal lands (Amsalu & De Graaff, 2007). In some instances, re searchers have observed village headsman appropriating lands from widows once they were improved (Bezner Kerr et al., 2007). Not only does the timing of benefits place barriers for some smallholders to implement Striga control practices, but also the exte nt to which they provide these benefits. For example, some suggest that an SFM method would need to provide at least 50 - 100% higher yields relative to current practices for smallholders to consider adoption (Baum et al., 1999). Unfortunately, not all Striga control practices can deliver these outcomes, and for the ones that do, access to agricultural extension (e.g., external agent visits, demonstration trials) must be provided as well (Ntege - Nanyeenya et al., 1997). Although, due to inconsistent fundi ng and training, developing countries face difficulties in providing effective extension about these practices (Kassie et al., 2013). Many Striga control practices involve rotating or intercropping legumes. When new plants such as legumes are introduced to traditional cropping systems, food security can be reduced in several ways. Farmers who believe that legumes will reduce their staple crop yields are often - averse tendency will also deter him or her 24 from investing in unfamiliar green manure legumes where no food is provided for one or more seasons (Pengelly et al ., 2003). Legumes (e.g., Vigna uguiculate , or cowpea) that do provide food still run the risk of attracting new pests to staple crops, further red ucing chances of adoption (Ndove et al ., 2004). Often funding and support for the agricultural sector is limited in developing countries. When support is given, all too often subsidies and export markets are allocated to crops which erode soils such as ma ize. These subsidies and/or markets contradict the promotion of practices and their associated crops that rehabilitate unproductive soils (FAO Land and Water Development Division , 2001). When fertilizer and markets are provided environmentally beneficial c rops, they still may not be taken - up by households when they do not coincide with taste or cooking norms (Drechsel et al ., 2005). This study contributes to the determinants of Striga - control practices in Malawi. There are numerous gaps mentioned in this su bsection, such as where farmers receive information about Striga and the extent they trust this information, from an agricultural extension officer, for example. Results from this study address these gaps using a number of analyses. In addition, the study contributes to the body of research evaluating tradeoffs farmers are willing to make for more Striga control. Finally, findings reveal which attributes farmers are most concerned with when implementing the Striga control practices in their fields. 2.3. Emp irical Model It is unlikely that any one Striga control practice will work effectively across the diverse biophysical and socio - economic landscape of Malawi. Instead, it may be more beneficial to offer farmers a basket of choices to choose from, allowing them to select a practice adapted to 25 their local conditions and livelihood strategies (Orr et al., 2002). To examine these choices, I employ a theoretical framework that is grounded in choice modeling, which is based on consumer theory (Lancaster, 1966). T he study employs discrete choice experiments to estimate the marginal value of various attributes for agricultural practices. 2.3.1 Random utility t heory Discrete choice experiments entail a controlled experiment where hypothetical scenarios are constructed and respondents choose one out of two or more alternatives. In each scenario, a respondent will choose an alternative that is characterized in terms of the levels of several attributes . By presenting multiple attributes that comprise the alternative, researchers can understand how respondents value certain attributes and confront tradeoffs between their levels . For example, instead of presenting several v arieties of maize for a respondent to choose from, a researcher may present the crop in a picture indicating the price of the seed, if it can be purchased using credit and whether the variety is resistant to pests or not (Birol et al., 2012). In this respe ct, respondents are obligated to make a choice based on their valuation of the three specified attributes. Valuation of attributes is consistent with choice theory, whereby farmers do not select the agricultural technologies themselves, but the characteris tics they embody (Ortega et al., 2014). Since there is uncertainty about which alternative will be chosen by an individual from a sample, researchers can assess the probability of him or her choosing a specific alternative (Lancsar & Savage, 2003). Hence, discrete choice experiments are rooted in random utility theory (RUT) because of its probabilistic nature. The framework proposes that utility is divided 26 into two components - an explainable (or rather observable) and a stochastic component. That is, (1) where U ij is the utility derived from choice j chosen by individual i , V ij is the observable component and ij is the random component. Eq. (2) explains the assumption that an individual would select alternative j if the utility derived from that alternative is greater than the utility derived from another alternative in choice set j. Such that, (2) In this equation, smallholder n will choose alternative j so long as Vnjs*>Vnks* Eq. (2) , actual utility ( V njs ) is observed, but indirect utility ( ) is not. In this study, a farmer is assumed to maximize his or her utility derived from cho osing a Striga control practice. In econometric terms, farmer n faces K alternatives contained in choice set s . I define an underlying latent variable Vnjs that denotes the value function associated with smallholder n choosing option j in a given choice task (Waldman et al. 2017). 2.3.2 Random parameter logistic regression Given that smallholders are socioeconomically heterogeneous, their preferences for Str iga control practices may be as well. One analysis often employed to evaluate preference heterogeneity is random parameters logistic regression (RPL), commonly referred to as mixed logistic regression. In this regression, indirect utility is assumed to be linear whereas marginal utility is monotonic (i.e., not increasing nor decreasing), yielding corner solutions where one choice is selected (Useche et al., 2013). Based on this assumption, farmer i written as 27 (3) where X ij represents the vector of attributes for the j th choice observed by the i th individual; Z i represents the vector of personal characteristics for that individual (i); and are vectors of servable component (e.g., choosing a corvette); and ij is the unobserved (or rather stochastic) component of utility, independent from the observed components (i.e., X and Z ) and equally distributed across individuals and alternative choices. The unobserv ed component acknowledges that unobserved variations and errors are present in farmer preferences for a given alternative in a scenario. As Train (2009) outlines, the probability that a smallholder n chooses alternative j in choice task s is assumed to b e - ( 4 ) where njs the parameters characterizing the distribution of random parameters such as mean and , the attributes of Striga control practices) and their respective attribute levels (e.g., low, medium and high labor). In Eq. (4) , the probability is approximated numerically through maximum likelihood simulation. In the analysis, we allow coefficients co rresponding to each attribute take a normal distribution. In doing so, their sign can either be positive or negative, indicating preferences for each of the attributes. 28 Due to the non - cardinal nature of utility, the coefficients generated by an RPL regres sion have limited economic interpretation. To gain insights about the behavior of a given sample of individuals, economic tradeoffs are calculated by dividing attributes that do not necessarily have monetary values (e.g., soil fertility) with ones that do (e.g., maize yield). Discrete choice experiments can explicitly account for zero, positive and negative willingness to pay (WTP) ratios. As Train (2009) explains, rather than assigning individuals with the same value associated across different attributes, RPLs indicate whether a statistically significant distribution exists between coefficients across individuals. In the sample, the sign of the random coefficient can be positive or negative. Nahuelhual et al. (2004) estimates - (5) where MU is the marginal utility gained from a various productive attribute and MUI is the marginal utility of income gained from a monetary attribute (i.e., profit). MUI is used as a proxy for the premium/discount coefficient. When there is a negati ve ratio for an attribute parameter, it is not strictly correct, but indicates the amount individuals are willing to accept in compensation to suffer a utility reducing attribute change (Rigby & Burton, 2005). In this study, a negative and statistically si gnificant WTP ratio indicates individuals would demand a certain amount of maize grain for higher soil fertility, for example. Oppositely, if the sign is positive and statistically significant, the individuals would be willing to sacrifice or accept (WTA) maize grain for higher soil fertility. 29 2.4 Methods 2.4.1 Discrete choice experiments Discrete choice experiments have been used as a method in research to assess tradeoffs farmers face when choosing alternative practices (Vaiknoras et al., 2014). Discre te choice experiments present respondents with scenarios having two or more alternatives to choose from. In the case of Striga weed control, these alternatives may be different practices (e.g., hand - weeding, herbicide) with the same attributes (e.g., time in field, yield received) but at different levels (e.g., 4 hours/day + 500 kg/ha vs 8 hours/day + 1000 kg/ha). In these scenarios, many times, respondents may pick an alternative they are not familiar with or opt - out (i.e., continue what they were already doing). By selecting one alternative over another, farmers reveal what tradeoffs they are (or not) willing to make to make. Then, by estimating marginal values of attributes, researchers can quantify these tradeoffs (Knowler et al., 2009). In this section , the strengths and limitations of discrete choice experiments are disused as well as the choice design and its implementation. 2.4.1.1 Strengths and limitations of method There are several advantages and limitations to using discrete choice experiments in the context of examining the implementation of an agricultural practice. First, RPL has moved beyond earlier methods of analysis (e.g., conjoint analysis) that assessed practice implementation by assuming homogenous preference across respondents (Birol et al., 2009). By assuming heterogeneity of preferences, discrete choice experiments enable unbiased estimation of individual preferences to accurately assess their needs for implementing a farming practice. In accounting for heterogeneous preferences across a population, better 30 policy recommendations can be made for which attributes to invest in for which groups to encourage implementation (Boxall & Adamowicz, 2002). Second, discrete choice experiments allow researchers to examine the willingness of farmers to confront tradeoffs among technologies with which they may not be familiar . Third, following mixed logistic regression, matrix correlations can show which attributes are correlated and affect implementation (Ortega et al., 2016). These correlatio ns are important to be cognizant of when assessing which attributes are of most concern to the smallholder. For example, a positive correlation found between two attributes of an SFM technology, such as soil fertility and time in field, would indicate that respondents were motivated by increased soil fertility were also motived by increased time in the field (Waldman et al., 2016). Discrete choice experiments assume several limitations. One of the largest drawbacks of using discrete choice experiments is that they are susceptible to hypothetical bias (Hensher, 2010). That is, the stated responses of farmers in the experiment may not reflect their actual behavior in the field. Also, discrete choice experiments are prone to researcher bias. In this instance, farmers could be selecting alternatives in the choice sets they believe researchers want them to make in hopes of receiving compen sation (e.g., seeds, extension). Second, farmers anchor their choice base on only one attribute rather than all attributes of the technology presented in the experiment (Árvai et al., 2014) . If they do not confront tradeoffs, the experiment has little proc ess validity. Third, attributes may have weights applied to them that are not congruent with the realities of smallholders. Later, when results are analyzed, conclusions made about tradeoffs do not reflect tradeoffs made in reality (Á rvai & Gregory, 2003). To address the three aforementioned concerns , researchers can first explain to participants that their decisions will 31 have no influence over inputs disseminated by the organization their affiliated with. Second, researchers can survey participants prior t o the experiment to determine what is the average time they spend fulfilling a SCP, for example, and apply these averages as appropriate weights. Third, researchers can ask participants during each choice scenario for participants to explain what the trade offs are between each choice set. 2.4.2 Choice d esign management practices against hypothetical Striga control practices. To make this comparison, control (i.e., alternat ive) attributes were discussed with farmers, but more specifically, their corresponding levels according to each practice. Literature and supporting data were later used to confirm the levels identified by farmers for each attribute. 2.4.2.1 Identificatio n of choice attributes via focus group discussion Three focus group discussions were held in May - June 2017 across three EPAs to determine the practices (i.e., alternatives) farmers were aware of that control Striga and the attributes they were most concerned with when implementing them. The manner in which focus groups were conducted such as the number of participants selected per discussion, participant recruitment and settings where focus groups took place are exp lained in section 2. 4.4 (Sampling procedures). The study took a Feminist - Political Ecology (FPE) perspective to inform its methodology for analyzing focus group data. There are many definitions of FPE, but this study drew from , whereby the perspective views gendered experiences as a result of political - economic environments. In turn, these environments govern how livelihoods are 32 affected in terms of institutions of property , social relations, etc. Livelihoods also inform how fa rmers value different attributes of new technologies. The valuation of attributes is influenced by the control that farmers have over resources in their households , thus, gendering his or her choices about new technologies (Adato & Meinzen - Dick, 2002; Din h et al., 2014). Prior to asking questions about Striga , results from a preliminary study conducted in 2013 were reported back to farmers at the beginning of the focus group discussions. Results from the study estimated the drivers of legume - maize intercro pping and described the implications for AEDOs and policy makers. Afterward, the researcher stated that the objective of the study was to identify which Striga control practices participants had heard about or used, and what attributes they considered befo re implementing them. The reason for reporting results back to farmers was to gain trust prior to data collection (Creswell & Miller, 2000). For example, the researcher explained that in 2015 - 16 , over 50% of the participants in the study had reported Stri ga as a primary challenge to production; hence, the study was being conducted to address their voiced concerns. Afterwards a series of open - ended questions were asked in a specific sequence so that attributes of locally implemented Striga control practices emerged (refer to Appendix 1 for further detail). First, participants were asked about their familiarity with S triga ( lifecycle, identification, effect on yield, seed transport), then about the history and extent of its effects in their field (e.g., when Striga first appeared in their fields, what yield losses occurred). After the preliminary questions were asked, farmers were asked to state any treatment and/or preventative practices they had heard of. Treatment practices are employed when Striga is observed in the field and removed by a famer. In some instance s , after the weed is removed, a 33 treatment practice may also entail applying an input where it emerged. Oppositely, preventative practices are employed before Striga is observed in the field in an effort to create less favorable conditions for germination. The treatment practices mentioned by focus group parti cipants included timely weeding, disposal in a deep pit and micro - dosing the affected area with maize bran, ash, fertilizer and/or manure. The preventative practices mentioned by focus group participants included mulching, minimum tillage and/or crop rotat ion/intercropping with legumes. The participants from each EPA mentioned different treatment and preventative control practices, but the intention of the focus groups was to gather and compose a list of all practices farmers had heard about or implemented. Afterward, participants were asked to identify the source they learned or heard about the practice. Then, what was required to carry out the practice (e.g., timing, required inputs) if they had implemented it in their field/s. The third part of the focus group discussion inquired about the goals or objectives farmers took into account before choosing and implementing a Striga control practice. Before they yield, c Striga control, participants were asked what were the short - and long - term objectives they took into account before implementing the practice as well as the primary/secondary benefits they aimed to receive. By identifying these objectives and benefits, attributes of Striga control practices were revealed. At the closing of each focus group discussion, participants were asked to rank the attributes from most to least important as well as the practices they believe were most to least effective in controlling Striga . Finally, participants were asked to rank which practices they preferred (from most and least) considering the attributes mentioned. 34 To assess other factors that affect ed preferences for various Striga control attr ibutes, focus group discussions were first recorded and transcribed from Chichewa to English. The enumerator who conducted the interviews assisted the researcher with translating each data from each focus group discussion . Then, transcriptions were uploaded into the qualitative data analysis software MAXQDA to analyze the data . Data was coded into nodes and sub - nodes. Nodes included farmer knowledge about the Striga lifecycle (e.g., germination, attachment), the type of pr actices mentioned (e.g., preventative, treatment), their understanding of the control mechanisms employed by each practice (e.g., suicidal germination catalyzed by legumes), the attributes they consider ed before implementing a practice (e.g., labor) and th eir preferences for each practice. To determine the valuation of attributes between farmers, Striga knowledge, practice preference and practice attribute preference nodes were applied to different participant quotes . In addition, q uotes were applied with gender and location nodes . Then knowledge and practice/attribute preferences were compared across gender and location. A concept map was made (see Figure 1 ) to assess if there was a qualitative relationship between gendered concerns for Striga control attr ibutes, and to what extent these concerns informed preferences for Striga control practices. In the concept map, l ine thickness represented the frequency of statements according to an attribute or practice. Each attribute or practice node could be opened t o view statements made about preferences for attributes or practices. 35 Figure 1 - C ontrol trait preferences related to control method preferences across gender 2.4.2.1 Selection of attributes Across eight themes, five common attributes were selected across both genders. They were later confirmed with literature. These attributes included Soil Fertility Improvement, Labor Requirement, Striga Emergence, Legume Yield and Maize Yield. They are reviewed below. The inputs and benefits required and received from each practice were quantified. Informal discussions with Striga experts and stakeholders were asked to verify these quantifications. 36 Literature was then referred to confirm the quantifications. From this review, many of the practices and their required inputs were based on the extent of Striga emergence in a field as well as number of other factors. 2.4.2.1.1 Soil fertility i mprovement Many Striga preventative control practices have shown to increase soil fertility and long - term yields, while Striga treatment control practices often increase same - season yield. Thus, increased soil fertility can be a secondary benefit received from implementing many Striga control practices. As a result, these benefits can positively affect farmer decisions to implement a practice such as maize - legume intercropping (Place et al., 2003). Other researchers have found in choice - experiment research that farmers are willi ng to sacrifice maize yields for soil fertility improvements (Waldman et al., 2017). Lastly, evidence has shown farmers are more likely implement SFM practices (e.g., Mucuna pruriens , or velvet bean, a cover crop) when they perceive soil fertility as a pri mary problem to their productivity (Versteeg et al., 1998). In this study, farmers would have to perceive soil fertility as a determinant of Striga emergence. Many control practices aim to alter the soil fertility by increasing soil pH and soil - P, given th at there is a strong correlation between higher soil - P and lower Striga emergence (Abdul et al., 2012). Soil fertility improvement was applied with three levels: low (sandy soil), medium (sandy - clay soil) and (dark loamy - clay) soil fertility. Smallholders were asked which fields in their community had these types of soils, and thereafter, were collected and put in sacks for them to see during the discrete choice experiment. Farmers were then asked to identify the difference between soils to ensure they agre ed with the assessment that sandy soils were assumed as low soil fertility, sandy - clay soils were assumed as medium fertility, and dark - loamy - clay soils were 37 assumed as high fertility soils. Soils in Dedza are dominated by coarse, well - dr ained A l fisols and a mixture of eutric Cambisols and eutric F luvisols (pr imarily in Golomoti) or ferric L uvisols (primarily in Linthipe) (Lowole, 1983). Whereas in Ntcheu, fields are dominated by mixed chromic Luvisols and orthic F erralsols (Mungai et al., 2016). Three leve ls of soil fertility were used: low, medium/current and high, which corresponds to a decrease in current soil fertility, no increase or decrease from current practices and a 50% increase in soil fertility. 2.4.2.1.2 Labor r equirement Different practices r equire different inputs to carry out in the field. For instance, some mulching practices call for 0.5 - 2 tons of maize stover to be applied per hectare to significantly change a soil organic matter , macronutrients) (Giller et al. , 2009) . This requires farmers to lay stover, and in many instances, harvest supplementary biomass to reach a 0.5 - 2 ton/ha threshold. Other control methods, such as minimum tillage, restrict soil disturbance (e.g., making ridges, use of a plough). In doing so, farmers can reduce the time they spend preparing their fields before planting, but extend their weeding labor given that some annual weeds were not buried from tilling at sowing. Preventative methods such as cereal - legume intercropping can reduce weed ing labor by shading low emerging annuals. Apart from intercropping, rotating cereals with legumes can spread labor to sow, weed and harvest to off - peak times, relieving labor burdens for households with smaller labor pools ( Thierfelder & Wall, 2010). On t he other hand, perennial legumes, particularly pigeon pea ( Cajanus cajun ), sometimes demand pruning during and after the growing season. In addition to these pruning activities, farmers have to de - shell pulses, adding to postharvest activities. These activities (among many others) have shown to affect farmer decisions when imp lementing Striga control 38 practices. Three levels of labor requirement were used: low, medium/current and high, which corresponds to a 50% decrease, no increase or decrease from current practices and a 50% increase in person - day farm activities. 2.4.2.1.3 S triga e mergence Many farmers will determine how effective a control method is by how much Striga emerges the same or following season. Thus, Striga can be controlled in terms of prevalence and persistence. Prevalence refers to the extent a weed emerges ac ross a given area whereas persistence refers to the extent a weed emerges across consecutive seasons. Emergence is typically not uniform across a field. Rather, weeds will emerge at various densities in different areas across a field . In other instances, u nder heavy infestations, a weed like Striga can have uniform emergence across an entire field. One factor that is consistent across all farmer settings is the number of individual Striga plants that can parasitize and be supported by a single maize plant. Only a maximum of 9 juvenile parasites can attach to one maize plant, and after underground attachment, a maximum of 8 flowers can emerge given the percentage of juveniles that make it to adulthood (Kunisch et al., 1991). It is clear that Striga emergence , in terms of its prevalence and persistence plays a highly farmers consider this attribute. Three levels of Striga emergence are specified in the choice experim ent: low emergence (0 - 1 flowers per plant), mild/current emergence (3 - 4 flowers per plant), to high emergence (6 - 7 flowers per plant). Flower numbers were based on average field observation and maximum attachment factors found in the literature (Kunisch et al., 1991; Smith et al., 1993) . 39 2.4.2.1.4 Legume y ield Several preventative Striga control practices involve the use or integration of legumes within CCS. Apart from improving soil fertility and reducing attachment, these crops provide a protein - rich food source during the interim (Place et al., 2003). Grains delivered by the legume not only can increase food security, but also provide an alternative income source for farming households. Researchers posit that farmers who are knowledgeable about the benefits legume grain provides, as well as the secondary benefits they offer (e.g., prote in - rich fodder), are more likely to implement them within their cropping systems (Bezner Kerr et al., 2007). In this study, the researchers used groundnuts ( Arachis hypogaea ) as the representative of legumes given that farmers were more familiar with thei r production, processing and market price. It is important to note that not all preventative controls, such as mulching, involve legumes. Many, in fact, are absent of legumes given they can be parasitized by Striga spp . and/or Alectra vogelii (another para sitic weed). Hence, the soil is blanketed with residues completely , absent of any legumes . With these considerations , the legume yield attribute was applied with three values - no yield (i.e., removal of any legume from field to conduct mulching), average y ield (300kg/ha) and high yield (600kg/ha). Farmers were presented with a smaller version of 50kg sacks filled with the aforementioned amounts during the discrete choice experiment. Thus, no sacks corresponded to the yield received from current practices, t hree sacks corresponded to low yield and six sacks corresponded to high yield. The attribute was applied with the aforementioned yields based on previous cereal - legume intercropping studies conducted by Kamanga et al. (2002) in Malawi. The researchers agre ed, however, groundnut yields were quite variable across EPAs. 40 2.4.2.1.5 Maize y ield Maize is the primary crop cultiv , and 51 % of total farmed 8). Hence, maize yield is a critical attribute considered by farmers when implementing any new agricultural technologies. Many times, preventative Striga control strategies and their related SFM practices have delayed returns on investment. Hence, the short - term and long - term maize yields such as mulching (Nowa practices (e.g., legume intercropping) stems from the fear that increased crop diversity will increase competition for resources, and consequently, reduce maize yield (Gliessman, 1992; Wa ldman et al., 2017). Based on the aforementioned points, maize yield was included as an attribute. Maize is often considered as a currency in rural areas where the study was conducted. The attribute, therefore, serves as a substitute for a cost/price va riable in order to evaluate the tradeoffs (Ortega et al., 2016). As Birol et al. (2009) explains, an indirect measure of cost (as opposed to a direct monetary variable) is more suited for discrete choice experiments conducted with subsistence farmers given that they may not be able to accurately assess the true value of their currency. In addition, financially insecure smallholders may not be familiar with the true value of cash given their limited access to it, making it an ineffective measure of currency for them. Maize yield (without fertilizer application) per hectare within the specified EPAs ranged from 500 - 2000kg/ha. Hence, the researchers agreed maize yield was quite variable across the EPAs. Four levels were applied to the maize attribute; a 50% los s (approximately 500 kg/ha); a 25% 41 loss (approximately 750 kg/ha); average yield (approximately 1000 kg/ha); and a 25% gain (approximately 1250 kg/ha). The following percentages were applied as values to the maize attribute based on observations in Malawi and supporting literature (Ngwira et al., 2013). In the discrete choice experiment, farmers were presented with a smaller version of 50kg sacks filled with the maize. Hence, 5 sacks corresponded to a 50% loss, 6 ½ sacks corresponded to a 25% loss, 10 sacks corresponded to current yield and 12 ½ sacks corresponded to a 25% gain. Detailed information on the selected attributes and their levels is presented in Table 1. Table 1 - Striga c ontr ol attributes used in choice e xperiments Attribute Variable form in equation Levels Definition Soil Fertility Improvement Hi_Soil_Fert, Low_Soil_Fert Less, current, more Soil fertility improvement received for applying the method. Less (sandy soil), current (sandy - clay soil) and more (dark loamy - clay) soil fertility. Labor Requirement Hi_Lab_Req, Low_Lab_Req Less, current, more Labor requirement defined as a 50% increase in labor (more), current labor or a 50% decrease in labor (less). Striga Emergence Hi_Strig_Emerg, Low_Strig_Emerg Less, current, more The extent Striga emerges per maize plant. Less emergence (0 - 1 flowers per plant), mild/current emergence (3 - 4 flowers per plant), to more emergence (6 - 7 flowers per plant). Legume Yield Hi_Leg_Yield, L ow_Leg_Yield None (i.e., current), low, high Legume harvest received from Striga control practice. The following yields were 0kg/ha, 200kg/ha and 400kg/ha Maize Yield Hi_Maiz_Yield, Low_Maiz_Yield 50% loss, 25% loss, average yield, 25% gain Maize harvest received from Striga control practice. The following yields were 500kg/ha, 750kg/ha, 1000kg/ha and 1250kg/ha. Based on the attributes selected above, a choice model is regressed. The derivatives of the likelihood estimates of the coefficien ts yield the probability of selecting one alternative over 42 two others. This gives a measure of explanatory power for all independent variables included in the equation. The equation used to estimate the parameters of a choice model is: (6) In the choice model ( Equation 6 ), Y ijs is the choice as a function of the various attributes and their respective levels, Hi_Soil_Fert/Low_Soil_Fert are variables indicating lower/higher soil fertility received from choosing a practice relative to the status quo, Hi_Lab_Req/Low_Lab_Req are var iables indicating the required labor required from the chosen practice relative to the status quo, Hi_Strig_Emerg/Low_Strig_Emerg are variables indicating lower or higher Striga emergence surrounding a maize plant received from a chosen practice relative t o the status quo, Hi _ Leg_Yield/Low _ Leg_Yield is the amount of legumes received from a chosen practice relative to the status quo and Maiz_Yield is the percent of maize yield received from a chosen practice relative to the status quo. The indices i, j and s represent the farmer, the choice and the scenario, respectively; whereas is the coefficient associated with each attribute and ij is the random component, which is assumed to be equally distributed across individuals and choices. The description , cod ing scheme and unit of measurement for each explanatory variable are listed in Table 2 . Different coding schemes were applied to each variable, but an effect coding scheme was applied to the non - monetary random parameters. Dummy or effect coding schemes could have be en applied to these attribute s . Neither type of coding scheme is necessarily better , but yield different interpretation s of the estimated effect an attribute has on the choice (if significant) (Kugler et al., 2012). Base d on the effect found in the results, the 43 researcher can decide whether one coding scheme yields more meaningful results to explain farmer decisions. The effects being estimated with an effect coding scheme are generally y codes do not estimate main effects (Rodgers et al., 1984). Main effects are defined as the difference between the mean response at one level of a particular attribute and the mean response at the other level, collapsing over the levels of all remaining a ttributes (Montgomery, 2009). Table 2 - Definition of variables used in choice m odel Short Form Description Coding * Values Random parameter Hi_Soil_Fert Higher soil fertility received from a Striga control practice Effect 0,0,1 Low_Soil_Fert Lower or higher soil fertility received from a Striga control practice Effect 1,0,0 Hi_Lab_Req More labor required to carry out Striga control practice as compared to status quo Effect 0,0,1 Low_Lab_Req Less labor required to carry out Striga control practice as compared to status quo Effect 1,0,0 Hi_Strig_Emerg More or more Striga emergence per maize plant by carrying out a Striga control practice Effect 0,0,1 Low_Strig_Emerg Less Striga emergence per maize plant by carrying out a Striga control practice Effect 1,0,0 Hi_Leg_Yield High (~600kg) legume yield received from an intercropping Striga control practice Effect 0,0,1 Low_Leg_Yield Low (~300kg) legume yield received from an intercropping Striga control practice Effect 1,0,0 Maiz_Yield Percent of current maize yield received from carrying out current Striga control practice Ordinal 50, 75, 100, 125 Non - random parameter Opt_Out Whether or not to continue status quo practices for controlling Striga Dummy 0,0,1 Note: If an attribute had had high soil fertility, then the code three cells in the column would read, 0 - 0 - 1. 44 When using a dummy coding scheme, row values correspond to the alternative and column values correspond to the individual making the choice (see Table 3 ). Values for the coding scheme are either 0 or 1 (Hardy, 1993). On the contrary, with an effect coding scheme, column values correspond to the alternative and row values correspond to the individual making the choice. Values for the coding scheme are either - 1 or 1 , with 0 referring to the status quo. When running a logistic regression, the status quo attri bute level (e.g., Med_Lab_Req) is not included with the higher/more and lower/less attribute levels in the model . The status quo attribute level acts as a reference point to interpret the results from its two counterparts . Table 3 - Example of dummy and effect coding schemes for labor requirement attributes Alternative Dummy Coding Effect Coding Low_ Lab_ Req Med_ Lab_ Req Hi_ Lab_ Req Low_ Lab_ Req Med_ Lab_ Req Hi_ Lab_ Req 1 1 0 0 - > 1 0 0 2 0 1 0 - > - 1 - 1 - 1 3 0 0 1 - > 0 0 1 A near - orthogonal design was made with the aforementioned attributes and levels using NGENE software. In perfectly efficient designs, each level would appear as equally often in each attribute, but in this design, each pair of levels appears equally often across all pairs of attributes with the design (Johnson et al., 2013). NGENE generated 36 choice sets blocked into 6 groups of 6 choice scenarios. Each scenario respondents were provided with three alternatives to choose from: two Striga control practices and an opt - out option. The opt - out option allowed participants to select neither of the two alternatives, inferring they would continue current practices (i.e., the status quo). Louviere et al. (2000) postulate that it is important to include an opt - out ch oice so that respondents can compare and infer what 45 tradeoffs they will make by selecting one or neither of the alternatives. The random parameter logistic regression (outlined in Equation 7 ) was estimated using the statistical software package Stata 15.0 as well as parameter logistic regression for the willingness to pay space (outlined in Equation 8 ) . The following commands were imputed in Stata 15.0: mixlogit Farmer_Choice Opt_Out, rand( Hi _Soil_Fert Low_Soil_Fert Hi_Lab_Req Low_Lab_Req Hi_Strig_Emerg Low_Strig_Emerg Hi_Leg_Yield Low_Leg_Yield Maiz_Yield ) group(Group_ID) id(Hh_ID) (7) and mixlogitwtp Farmer_Choice Opt_Out, rand( Hi_Soil_Fert Low_Soil_Fert Hi_Lab_Req Low_Lab_Req Hi _Strig_Emerg Low_Strig_Emerg Hi_Leg_Yield Low_Leg_Yield ) price( Maiz_Yield ) group(Group_ID) id(Hh_ID) (8) where Farmer_Choice is specified as the dependent variable, Opt_Out is specified as a non - random parameter, Hi_Soil_Fert Low_Soil_Fert Hi_Lab_Req Low_Lab_Req Hi_Strig_Emerg Low_Strig_Emerg Hi_Leg_Yield Low_Leg_Yield Maiz_Yield are specified as random parameters , Group_ID specifies the choice set where an alternative (out of three) was selected and Hh_ID specifies who made the decision (i.e., the participant identification number). In equation/ command (8) Maiz_Yield is specified as the price variable to estimate how attributes are valued by participants in terms of percent maize yield. In some instances, random parameter logistic regressions specify the price attribute as a non - random parameter. Meijer and Rouwendal (2006) postulate though, it is difficult to assume all individuals receive the same marginal utility from the monetary attribute. Alternatively, researchers can specify preference for Maiz_Yield to be heterogeneous and argue the coefficient for this monetary attribute is log - normally distributed; hence, the Maiz_Yield is specified as a random parameter ( Hole & Kol stad, 46 2012). Similar to price, Opt_Out can be specified as a random parameter, but other regression logistic regressions have specified Opt_Out as a non - random parameter (Waldman & Richardson, 2018). 2.4.3 Site d escription Household surveys were conducted over a 3 - week period from August - September 2017 using questionnaires and discrete choice experiments in two central districts of Malawi - Dedza and Ntcheu. Dedza and Ntcheu are located in the Kasungu Lilongwe Plain (14.1667° S, 34.3333°E) and Rift Valley Escarpment (14.7500°S, 34.7500°E), respectively. Within these districts, four extension - planning areas (EPAs) were selected for data collection, namely Linthipe, Kandeu, Nsipe and Golomoti ( See Figure 2 ). These EPAs were speci fically chosen based on the growing challenge of Striga reported by farmers in recent years (Atera et al., 2012). Hence, the study was highly relevant to the region and its current farming population. Figure 2 - Data c ollecti on s ites Malawi has a unimodal rainy season occurring from November to April, and a dry season from 47 May to October (Jury & Mwafulirwa, 2002). The sites in this study provide a gradient of biophysical potential as described in depth by Mungai et al. (2016). The marg inal environment of Golomoti has a high evapotranspiration and erratic rainfall, compared to the medium potential sites of Kandeu and Nsipe. The high potential agricultural site of Linthipe has a medium - high elevation and generally receives well - distribute d rainfall (Smith et al., 2016; Tamene et al., 2015). 2.4.3.1 Choice experiment setting Demonstrative choice sets were created for each block, each containing 6 choice sets (see Figure 3 ). To increase comprehension of each choice task and reduce cognitive burdens, actual soil, maize grain and unshelled - peanuts were used. In addition, actual h and - hoes and life - sized photos of Striga flowers were used. This design was pre - tested in the field to ensure the props were relevant. Each choice task AEDOs asked participants to indicate what difference/s in attribute levels were present between the alte rnatives and the status quo (i.e., opt - out option). Stating differences between alternatives for each choice task ensured that farmers understood the tradeoffs they were making by selecting one alternative over another. 48 Figure 3 - Sample choice t ask Attribute Option A Option C Status Quo Striga emergence High Low Neither Labor requirement Soil fertility benefits Medium High Legume yield Maize yield (kg/ha) Enumerators introduced one block of the discrete choice experiment to 10 respondents at a time. Each group of respondents had an enumerator explain the purpose of the discrete choice experiment and clarify any questions respondents had afterwards. Responde nts were then given a card with a blue, green and orange circle corresponding to alternative 1, alternative 2 and opt out, respectively. Each choice task respondents would indicate which choice participants made by pointing to the circle behind their back decisions. 49 2.4.4 Sampling procedures Data used to run the mixed logistic regression was collected at two different time - periods. First, in March of 2016 (Maize Harvest Questionnaire) then June - September 2017 (Striga Question naire Survey). The program Africa Research In Sustainable Intensification for the Next Generation (RISING) conducted the Maize Harvest Questionnaire using questionnaires to investigate how various sustainable intensification methods affect food security, farming livelihoods and agroecological system health. The program has been conducting action - based research with the farmers since 2013. Farmers who participated or were surveyed by the Africa RISING program (treatment, local - control and distant - control) were selected to participate in the Striga Emergence Questionnaire. Socioeconomic data from the Maize Harvest Questionnaire were used to avoid re - collecting the same data and prevent respondent fatigue. The data from this survey were not collected by the researcher, but he had access to the instrument, data, its investigators and enumerators to clarify any questions regarding clarity and validity. A stratified sample of 21 (N = 298) across four EPAs (Linthipe, Golomoti, Nsipe, Kandeu) to determine who would be surveyed in the Striga Emergence Questionnaire and discrete choice experiments. First, a stratum of 125 participants were purposefully selected consisting of households that expressed Striga as primary challenge to productivity (see Table 4 ). After these households were removed from the roster, famer names were segregated into their respective EPAs. G iven the budget constraints of this study, only 50 - 60 Striga Emergence Questionnaires and discrete choice experiments were carried out per EPA. Taking this budget into account, the first names of the 50 household heads were put in ascending order alphabetical ly and the remaining balance was taken to fill a quota of 50 - 60 questionnaires per EPA. For example, in Linthipe, 36 farmers were and the first 24 names were select ed to make a total of 60 famers. After eliminating households for which data were missing or incomplete, 51 households were selected from Linthipe, 59 from Golomoti, 52 from Nsipe and 53 from Kandeu. Table 4 - Farmers expressed S triga as a primary c hallenge in 2016 (out of a list of 15 productivity challenges) EPA No Striga challenge Striga challenge TOTAL Farming HHs Linthipe 31 36 (54%) 67 Golomoti 32 26 (45%) 58 Kandeu 51 27 (35%) 78 Nsipe 39 36 (48%) 75 TOTAL 153 125 (45%) 278 *Striga frequencies were sourced from an Africa RISING database where a maize harvest - questionnaire collected information about productivity challenges in the 2016 growing season. No manuscript has been published with this data. Data was availa ble from a Maize Harvest Questionnaire conducted in March 2017, but the roster could not be used to inform which participants to sample for several reasons. A question What challenges did you face in production ON THIS PLO T this The 2017 questionnaire assessed Striga challenges based on field observations. That is, enumerators would randomly select seven areas in two primary fields of production to see if Striga was present. This question may have not yielded a good representative of the farmers facing Striga challenges for two reasons. First, enumerators may have missed identifying juvenile plants given their small nature, and second, farmers could have removed the weed prior to observation. 51 The plot mentioned in the 2016 questionnaire was referring to one of two fields the Africa RISING program has been monitoring over the course of five years. These two plots are considered as the two primary fields of producti on for the household. Thus, if the farmer indicated that Striga was a challenge in one or both plots, they were considered as a household Striga Striga out of a list of 16 c hoices related to primary challenges (e.g., not enough fertilizer, drought, low soil fertility). Hence, the data gathered from this question (as indicated by Table 4 ) was a strong indicator that a household was facing Striga challenges. The researcher ackn owledges farmers may have been suffering yield losses unknowingly to Striga parasitism, thus table frequencies may be underestimated. Apart from using the 2016 Maize Harvest Questionnaire to indicate which farmers to participate in the study, no other data was used. For focus group selection, Agriculture Extension Development Officers (AEDOs) were given lists of farmers from the Africa RISING program and participants were randomly selected from each gender. AEDOs purposefully selected 6 - 8 men and 6 - 8 women per focus g roup to avoid one gender from dominating the discussion and to capture a diverse dialogue. The focus group quota was set to 12 - 15 participants to ensure each participant had ample opportunity to share his or her opinion (Fern, 1982). Each discussion lasted between 60 - 80 minutes, was recorded and transcribed after. Discussions took place in or near an extension office. Participants were compensated with a soda and bread to discuss Striga control practices with the researcher and his enumerator. 52 2.4.5 Inst rument calibration and p rotocol Three enumerators and the researcher held focus groups, conducted surveys and administered discrete choice experiments (i.e., the second time - period). All instruments were pretested to assess the comprehension and suitabilit y of questions used in the focus groups/questionnaires as well as the attributes used in the discrete choice experiments. The researcher trained all enumerators prior to data collection. For example, enumerators would hold mock - focus groups or conduct mock - questionnaires/choice experiments with non - participants prior to collecting data for the study. During each of these instances, the researcher observed how questions were asked, respondents answered and made critiques to instruments as well as suggestions to enumerators about their data collection techniques (e.g., probing). The March 2017 questionnaire went through a similar process whereby the instrument was tested in the field prior to data collection. All enumerators that participated in the researcher - September 2017) had also fielded questionnaires in March 2016. Striga knowledge, current and past agricultural practices, and the effects the weed has had on their farms. Plea se refer to the instrument in Appendix 2 and 3 for further clarification about questions and responses . Once the questionnaire was completed, farmers participated in a discrete choice experiment. The survey and discrete choice experiment took approximately 40 to 70 minutes. Farmers were compensated with 5 blocks of laundry soap for completing the questionnaire and one bottle of soda for completing the discrete choice experiment. Survey - respondents primarily consisted of those charged with makin g farm decisions for their household. Focus group questions were discussed in the previous sub - Identification of choice 53 attributes via focus group discussion R efer to Appendix 1 for further clarification about questions. The discussion took appr oximately 40 to 70 minutes. Farmers were compensated with one loaf of bread and one bottle of soda for participating in the discussion. Survey - respondents primarily consisted of those charged with making farm decisions for their household. 2.5 Results & Di scussion 2.5.1 Descriptive statistics Table 5 explains the sample characteristics by gender using summary statistics. Almost three quarters of the participants were female. The average respondent age was 45 years old with a little more than some secondary education completed (Mean Education = 2.65). Male participants were slightly older and received more education than female participants. Households had about two members participating on their farms regularly, varying little across gender. As an indicator of wealth, annual income ranged between 33,000 MKW to 55,000 MKW, and among that income, 10.7% was derived from casual labor (i.e., ganyu labor). Male participants received slightly higher incomes (21.5% more) than female participants. Women, however, seem ed to engage in more ganyu labor than men. Table 5 - Sample c haracteristics Full sample Women Men n = 215 n = 160 n = 55 Mean Age 45.62 45.30 46.53 Mean Education (1 - 7 levels) 1 2.21 2.06 2.65 Mean Maize Yield/Ha (2016) 1,523 1,471 1,574 Mean Household Labor 2.82 2.79 2.88 Mean Income 27,583 24,264 30,902 Mean Total Land Ownership 2.33 2.17 2.79 Mean # of SFM Practices Employed 2.33 2.2 2.38 54 Table 5 % Intercrop Maize in Past or Currently 79.17 79.14 79.17 % Engaged in Ganyu Labor During the 2016 - 17 Season 46.98 56.9 35.6 % Consider Ganyu Labor as Primary Income 10.72 12.50 5.54 % Have One or More Fields with Low Soil Fertility 53.86 52.72 55.00 % Mention Soil Fertility as a Primary Production Challenge 19.63 15.00 14.50 % Mention Striga as a Primary Production Challenge 53.87 55.00 52.73 Mean % Fields with Striga 57.71 57.48 58.49 Mean Striga Knowledge (0 - 11) 5.10 5.14 5.00 1 Schooling levels: 1 - No School, 2 - some primary, 3 - complete primary, 4 - some secondary, 5 - complete secondary, 6 some post - secondary, 7 complete post - secondary) As an indicator of household food security, maize yield (per hectare) varied dramatically, ranging from 400 kg/ha to 4000 kg/ ha, but the average was approximately 1500kg/ha. The mean yield seemed to exceed farmer yields (approximately 1000 kg/ha) from the 2016 season. This commonly occurs when yield cuts are extrapolated into kg/ha, which is what the study did. Still, yield cuts are effective in determining which participants received higher or lower yields, and consequently, higher or lower food security. Male and female participants received very similar yields. Participants owned on average 2.33 ha of land and cultivated 76% ( 1.76ha) of their farms with maize as a primary crop. In terms of Striga prevention, both male and female participants overwhelmingly (approximately 80%) had cultivated legumes as an intercrop or rotator crop with their maize in the past or currently. Exclu ding legumes from SFM practices, both men and women conducted between 1 and 3 soil SFM practices that directly or indirectly prevented Striga (e.g., minimum tillage, manure application). We asked farmers a series of questions related to their knowledge of Striga . Approximately 53% of the sample (115 farmers) had expressed Striga as a primary challenge to their farm production (see Table 5). Women expressed slightly (3%) more concern over S triga than men as 55 a primary challenge to production. Women also seemed to know slightly more about Striga Striga questions farmers answered correctly regarding the identit y o f Striga and the mechanisms behind its parasitism (Refer to Appendix 2 , Questions 1 - 5, 10 for further clarification). Both genders assumed virtually the same percentage of fields having Striga (57% vs 58%). Apart from Striga , equivalent percentages of men and women mentioned soil fertility as a primary challenge to production; however, men characterized marginally more fields as having low soil fertility than women (55% vs 53%). 2.5.2 Marginal value of Striga control attributes Results from the discrete choice experiments are displayed in Table 6 . Being that Striga emergence, labor requirements and soil fertility improvement are all coded either as a - 1, 1 or 0, a positive coefficient indicates decisions to select a Striga control practice were e ncouraged by the attribute. A negative coefficient indicates decisions to choose a Striga control practice were discouraged by the attribute. Since legume yield is coded with a zero, one or two, a positive coefficient indicates the valuation from a low to high level. The same assumption can be made for the maize yield attribute. Table 6 - Random p aramet ers logit model and willingness to pay s pace for Striga control p ractices Variable Preference Space WTP - space Coefficient Std. error Coefficient Std. error Random parameter means Hi_Soil_Fert 0.166*** 0.055 11.158*** 4.200 Low_Soil_Fert - 0.152** 0.063 - 9.134* 5.037 Hi_Lab_Req - 0.122* 0.065 - 9.273* 4.955 Low_Lab_Req 0.091 0.067 7.844 5.292 56 Table 6 Hi_Strig_Emerg - 0.229*** 0.061 - 16.844*** 4.790 Low_Strig_Emerg 0.141*** 0.063 10.288** 4.887 Hi_Leg_Yield 0.367*** 0.066 27.792*** 5.841 Low_Leg_Yield 0.173*** 0.061 13.165*** 4.961 Maiz_Yield 0.012*** 0.002 - 4.378*** 0.131 Non - random parameter means Opt_Out - 0.126 0.174 8.779 11.934 Random parameter standard deviations Hi_Soil_Fert 0.022 0.146 5.123 8.381 Low_Soil_Fert 0.195 0.140 8.662 8.862 Hi_Lab_Req 0.234** 0.117 6.429 9.303 Low_Lab_Req 0.085 0.196 18.696** 9.294 Hi_Strig_Emerg 0.085 0.093 7.162 6.517 Low_Strig_Emerg 0.101 0.119 12.019 8.698 Hi_Leg_Yield 0.261*** 0.119 30.206*** 7.351 Low_Leg_Yield 0.066 0.090 7.890 7.998 Maiz_Yield 0.011*** 0.002 0.311*** 0.098 N 3870 3870 LR chi2(9) 42.75 4200.150 Log - Likelihood - 1221.2861 - 1232.357 Prob > chi2 <0.01 <0.01 Note: ***, **, * represent significance at the 1%, 5%, and 10% levels. Random parameters logit model estimated using Stata 15. The coefficients of high soil fertility improvement, Low_Strig_Emerg, Hi_Leg_Yield and Low_Leg_Yield are positively significan t at the 1% level in the RPL. Hi_Leg_Yield was valued twice as much as any of the other positively significant attribute. These findings suggest that all attributes were strong determinants of farmer decisions, but Hi_Leg_Yield may be an overarching factor . As expected, Maiz_Yield coefficient (i.e., the monetary attribute) was both positive and highly significant. Low_Soil_Fert, Low_Lab_Req and Hi_Strig_Emerg were negatively significant, implying all attributes deterred the selection of Striga control pract ices with the following attributes . Hi_Strig_Emerg assumed the highest coefficient and was 57 significant at the 1% level (as compared to Low_Soil_Fert [5%] and Hi_Lab_Req [10%] ) . While these values are relative, they indicat e practices that require higher la bor requirements and result in lower soil fertility or higher Striga are unlikely to be implemented. Low_Lab_Req was negative but not significant, indicating decreases in labor requirements were not necessarily an important determinant of choice for a Striga control practice. Only 5% (200/3870) of decisions were Opt_Out; hence, the non - random parameter was not found to be significant, but the coefficient was negative. The Wald chi - square statistic allows us to reject the null hypothesis whereby no attributes significantly influence respondent decisions at the 1% level. Significant standard deviation estimates indicate preference heterogeneity for high labor requirements, high legume yield and maize yield across participants. Their significance suggests that a subset of farmers value attributes differently when compared to their counterparts. Results from the estimation in WTP - Striga control attributes. The maize yield attribute is used to calculate the marginal rat e of substitution; thus, the coefficient of an attribute can be interpreted as the percent of maize yield a participant is willing to sacrifice or be compensated for by choosing a Striga control practice. Positive valuation (i.e., a positive coefficient) s hould be interpreted as how much yield a participant is willing to sacrifice to receive or have higher levels of that attribute. Negative valuation (i.e., a negative coefficient) should be interpreted as how much yield a participant must be compensated to receive or have higher levels of that attribute. The coefficient of willingness to pay estimates should be considered as relative values, not exact magnitudes (Rocker et al., 2012). The coefficient is negative because in a willingness to pay space attribut e 58 values for the price attribute (i.e., Maiz_Yield) are multiplied by - 1 to run the regression in Stata 15 (Hole & Kolstad, 2011). In a willingness to pay space, participants were willing to sacrifice the largest percentage of maize to receive a high legu me yield (27.8%) or low legume yield (13.2%), followed by high soil fertility improvement (11.2%) and low Striga emergence (10.3%). Based on the value of the negatively significant coefficients, participants would need to be compensated with the highest pe rcent of maize yield for Striga control practices that were associated with high Striga emergence (16.8%), followed by high labor requirements (9.3%) and lower soil fertility improvement (9.1%). The marginal utility of the monetary attribute (Maiz_Yield) i s negative and significant, as expected, because the attribute was multiplied by - 1. Table 7 - Correlation matrix for random parameters logit m odel in Table 6 (1) (2) (3) (4) (5) (6) (7) (8) (9) Maize (1) 1.000 Hi_Soil_Fert (2) - 0.004 1.000 Low_Soil_Fert (3) - 0.007 0.496 1.000 Hi_Lab_Req (4) - 0.010 0.025 - 0.005 1.000 Low_Lab_Req (5) - 0.019 - 0.006 - 0.019 0.502 1.000 Hi_Strig_Emerg (6) 0.011 0.017 0.016 - 0.029 0.028 1.000 Low_Strig_Emerg (7) 0.014 - 0.001 0.013 0.013 0.014 0.496 1.000 Hi_Leg_Yield (8) 0.001 - 0.052 - 0.004 0.015 0.001 0.012 - 0.012 1.000 Low_Leg_Yield (9) - 0.016 - 0.016 - 0.014 - 0.012 - 0.004 0.001 0.011 0.507 1.000 Table 7 is a matrix correlation which indicated whether respondents were motivated by the choice of certain attribute based on the value another. Under this premise, the attributes can be positively or negatively correlated. No negative or positive correlation co uld be found 59 between two attributes ( Table 7 ), implying respondents were not motivated by an increase of legume yield and an increase in soil fertility improvement, for example. These assumptions were based on conventional levels of statistical significanc e. The highest negative correlation found among the attributes was between high and low labor requirements as well as high and low Striga emergence whereas the highest positive correlations were found between high and low legume yield. This analysis indica tes farmers do not necessarily find an association between attributes when choosing a Striga control practice. Results from the experiment agree with findings from similar studies that evaluated soil fertility management practices using choice experiments (Silberg et al., 2017). For example, the negative association found between Striga control practices and high labor requirements coincides with the counterfactual finding presented by Vaiknoras et al. (2015), where Ugandan smallholder preferences for soil conservation practices were positively associated with lower labor requirements. Choice experiments that evaluated legume intercropping decisions in Malawi (Waldman et al., 2017) analogously found smallholders were willing to sacrifice a large percent of their maize yield (36.5%) for soil fertility improvement. Unlike Waldman et al. (2017), labor requirements and soil fertility improvement attributes were not correlated. Their study, however, included two different attributes (e.g., biomass, pigeon pea yie ld) in their choice experiment, which may be a reason why no correlation was found in our study. Farmers were only willing to sacrifice a marginal percent of their maize yield for legumes (perennial pigeon pea grain) in the Waldman et al. (2017) experiment . Our experiment conversely found farmers were willing to sacrifice willing to sacrifice large percentages of maize yield for 60 legumes. The species of legume (e.g., annual groundnut vs perennial pigeon pea); therefore, may be a large reason why farmer - deci sions were vastly different in this experiment. 2.5.3 Gender level differences We also estimated RPL models in a willingness to pay space for each gender (see Table 8 ). The analysis revealed that men and women farmers valued attributes differently. Female participants were not willing to pay for higher soil fertility improvement while their male counterparts were (17.12% loss in maize yield). Men were willing to sacrif ice 7.57% more losses in maize yield for lower labor requirements than women would need to be compensated for control practices with higher labor. Both genders would need to be compensated for control practices that received higher Striga emergence, but women would need to receive more than a 20% increase in current maize yields for this burden. Furthermore, women were willing to sacrifice maize yield losses for lower Striga emergence while men were not. Women were also willing to sacrific e 39.4% more losses in maize yield for higher legume yield when compared to men. This is not to say men did not value legumes, but women were willing to sacrifice more maize yield losses for higher legume yield. The opt out dummy for male participants was very large and significant, indicating that they derived more utility selecting Striga control alternatives as opposed to continuing status quo practices. The price attribute was negatively significant for both genders. No correlation was found between att ributes among either gender. 61 Table 8 - Willingness to pay s pace for Striga c o ntrol practices across g ender Female WTP space Male WTP space Coefficient Std. error Coefficient Std. error Random parameter means Hi_Soil_Fert 7.726 0.147 17.124** 7.191 Low_Soil_Fert - 9.637 0.151 - 9.077 7.303 Hi_Lab_Req - 12.264* 0.161 - 6.894 7.626 Low_Lab_Req 7.548 0.149 19.833*** 7.374 Hi_Strig_Emerg - 21.103*** 0.138 - 16.209** 7.706 Low_Strig_Emerg 15.276** 0.149 3.433 8.155 Hi_Leg_Yield 36.380*** 0.149 22.436*** 6.799 Low_Leg_Yield 7.296 0.140 14.929** 7.702 Maiz_Yield - 4.613*** 0.184 - 3.978*** 0.207 Non - random parameter means Opt_Out - 8.359 0.402 27.792* 14.449 Random parameter standard deviations Hi_Soil_Fert 11.036 0.179 26.240*** 8.116 Low_Soil_Fert 12.801 0.201 19.313** 8.958 Hi_Lab_Req 21.896** 0.234 22.943** 9.417 Low_Lab_Req 9.004 0.202 4.175 6.236 Hi_Strig_Emerg 3.946 0.167 1.080 7.362 Low_Strig_Emerg 10.069 0.235 33.504*** 12.315 Hi_Leg_Yield 26.497*** 0.464 11.313 8.417 Low_Leg_Yield 1.533 0.255 6.556 10.224 Maiz_Yield 0.610*** 0.125 0.662*** 0.260 N 2880 990 LR chi2(9) 2116.01 986.730 Log - Likelihood 293.849 290.970 Prob > chi2 0.156 <0.001 Note: ***, **, * represent significance at the 1%, 5%, and 10% levels. Women participated and derived more of their primary income from ganyu labor, supporting the finding whereby women were deterred by control practices with higher labor requirements whereas men were not. Opposing results between men and women in the willing ness to pay space did not appear to be supported by what little socioeconomic differences were found 62 between genders (seen in Table 5 ). Qualitative findings seemed to shed more light between these differences however. For example, focus group transcription (see Table 9 ) highlights an underlying concern for hunger and food provided by Striga control practices, whereas men seemed to deliberate more about the monetary costs of a practice before implementing it. These considerations are supported by the literat ure as well (Njuki et al., 2011). In addition, men appeared to have preferences for control practices that used synthetic inputs given their preferential access and use of them (Bezner Kerr et al., 2007). These findings may support why women were willing t o pay more for legumes than men. Table 9 - Participant quotes related to food security and financial costs and preference of Striga control practices Female Participant Male Participant harvest anything. It brings hunger G money in your pocket , you just buy herbicide and spray. That would lessen your - N crop rotation more maize the following season. With manure application, the maize will still grow well with Striga, but not as well with crop - L fertilizer , for you to apply three times, you need to have enough money - N With respect to soil fertility improvement , men were willing to sacrifice while women were not. Women may have been less concerned about soil fertility improvement being that they generally do not own land nor are they permitted sell it once they have increased the value (Pircher et al., 2013). Rather they are willing to sacrifice more of their maize yield for lower Striga emergence perhaps because they are more familiar with the weed and its effects. Literature explains women may be more familiar with weeds given that they are charged with 63 the task of hand - weeding (Andersson & Giller, 2012). Men do weed, but primarily with the use of a hoe (via scraping), as they are charged w ith preparing the land. Men may only see the negative effects of witchweed at harvest, which may be a reason their willingness to pay is smaller. Gendered roles, thus, may be a driver of attribute preferences (e.g., men and soil fertility improvement ) (see Table 10 ). Table 10 - Participant quotes related to gender roles and preference for Striga control practices Female Participant Male Participant scrape , you leave a cutting in the ground to grow again. So I think uprooting would G weeding in the field, but you will not produce anything if M1 - N Each gender was concerned about labor requirements differently. Male focus group participants were willing to pay for lower labor requirements to control Striga whereas female participants mentioned multiple times they would only implement a practice if th ey could manage it themselves. When asked which practices they preferred, the female narrative (as shown in Table 11 ) seemed to avoid any practice associated with intensive labor requirements. Concern for labor requirements may stem from previous experienc es when new technologies shifted the burden of increased weeding labor to them (Giller et al., 2009). Table 11 - Participant quotes related to labor and preference for S triga control p ractices Golomoti Female Participant Linthipe F emale Participant Nsipe Female Participant takes a lot of hands - G we look at labor about what we think we can manage in term of labor. If we can manage it, then we will choose - L will not consume too much of our time F2 - N 64 2.6 Conclusions The objective of the study was to learn: 1) when given a choice among Striga control practices, which alternative would participants choose (or remain with the status quo); 2) which attributes most significantly influenced the selection of a Striga control practice; and finally, 3) what tradeoffs were male and female participants willing to accept. Few farmers opted out (5%, 200/3870), even when faced with selecting an alternative with lower maize yield. A strong inclination to select one of the two alternatives reveals one or a combination of motives. First, given the widespread emergence of Striga across the central region of Malawi, policies providing legumes may encourage Striga reduction. Second, farmers may have selected alternatives they believed researchers wanted them to select (i.e., hypothetical bias) in hopes of receivi ng compensation later (Hensher, 2010). Third, farmers may have been willing to sacrifice larger percentages for maize yield for various attributes given the choice experiment was conducted one month after maize harvest. Had the experiment been conducted on e month prior to harvest season (when maize foodstuffs are most scarce), farmers may have been more reluctant to choose any alternative where maize yield was less than average. Significant attributes in the RPL suggest the correct characteristics and appr opriate levels were applied to hypothetical alternatives for the choice experiment. As expected, participants chose scenarios with lower Striga emergence as well as higher maize yield, legume yield and soil fertility improvement. Participant decisions were not influenced by scenarios that had lower labor requirements, but were negatively and significantly influenced by higher labor 65 requirements. These findings suggest farmers will implement Striga control practices that integrate legumes within their maize - based systems that offer soil fertility improvements, but more conce rned with a Striga control practices that offers more food security, as opposed to ones that provide soil fertility improvement or reductions in labor requirements. As much as farmers expressed their concerns about Striga and the negative effect it has on their maize yield, the willingness to pay space suggests otherwise. Farmers were willing to sacrifice the highest maize yield loss for high legume yield (>27%), followed by low legume yield, high soil fertility improvement, and finally, lower Striga emergence. Strangely, farmers would need to be compensated for practices that received higher Striga more than any other attribute. Concern about higher Striga emergence may be out of fear of maize yield loss. Less concern for decreasing emergence may be that Striga is not viewed as a limiting factor to production as compared to input availability (e.g., fertilizer) or rainfall. That is, farmers do not believe Striga emergence has passed a limit (i.e., economic threshold level) where it should be controlle d (Debrah, 1994). This view about the effect of Striga on maize may be attributed to two issues. First, Striga knowledge among farmers may be low. While the assessment of knowledge in the study was subjective, the majority of farmers scored very low ( = 5 . 1 out of 11). Second, farmers may not believe they have the means to control a stubborn weed such as Striga . In Ethiopia, smallholders abandoned fields long after they had discovered Striga and claimed they were well aware of the effect they pest had on th eir cereal crops (Tamado & Millberg, 2000). Prior to halting cultivation, numerous control practices had been disseminated 66 for several years. The researchers found that abandonment was attributed to input availability and economic feasibility, not necessar ily neglect. Malawian farmers may be faced with the same dilemma. Further research is needed to assess barriers to Striga control implementation and drivers behind their decisions. These barriers may be due to any number of issues besides the obvious soci oeconomic challenges cited in literature. As it became evident in this study, very few farmers were able to make the connection between the mechanism behind what increased or decreased Striga emergence. That is not to say farmers were not aware of Striga c ontrol practices, but they were unaware of why Striga emergence decreased when legumes were planted. Many times, creating the connection between a practice and its effect on a pest has shown to increase uptake of technologies (Oswald, 2005). The study als o encourages future choice experiment research to confirm findings and inform instruments with qualitative inquiry. Summary statistics and parametric tests can be limited in explaining the difference between male and female participant decisions, especiall y in the willingness to pay space. Many times, when qualitative and quantitative methods are used separately to analyze farmer decisions, findings are not generalizable or do not highlight the context - specific nuances, respectively. The addition of qualita tive methods in the analysis unveiled different preferences between different farmer - types. Furthermore, qualitative inquiry helped determine whether quantitative findings had any internal validity (Barbour, 2001). Without focus groups and a thorough liter ature review, the results of this study may have been confounding. Consecutive mixed method approaches, therefore, may be valuable in 67 explaining the tradeoffs farmers are willing to make to implement future agricultural technologies (Morse, 2005). The f ollowing DCE allowed opportunities to present technologies to farmers they may not have seen, heard or used (e.g., Striga - resistant maize). Demonstration trials were not neede d to e licit decisions about these relatively new or unknown control technologies. Choice experiments offer more economical avenues for researchers to evaluate technological preferences. The study also attempted to move beyond earlier methods (e.g., latent class models) that assess farming practice decisions which assume heterogeneous p references across heterogeneous respondents (Birol et al., 2009). By assuming heterogeneity, the study enabled unbiased estimation of individual preferences while enhancing the accuracy of smallholder needs. Country - wide maize yields can be increased when the development and dissemination of agricultural practices are informed by a better understanding of smallholder preferences for specific attributes. Malawian farmers are unlikely to employ Striga control practices which exceed yield losses for attribute s (e.g., lower labor requirements) they do not desire. With little uptake, Striga will likely continue to emerge and reduce maize yield in Malawi . Curtailing practices for specific smallholder groups encourages implem entation, increasing maize yield and consequent food security (Boxall & Adamowicz, 2002). 68 APPENDICE S 69 APPENDICES APPENDIX 1. Focus Group Instrument Objective: Discover traits of Striga control methods that are of most concern to smallholder farmers Introduction: Today we are going to ask you about your knowledge and practices regarding weed management. The intention of this interview is to first, gather informa tion so researchers can better understand Striga management strategies conducted by farmers like yourself. Then according to your description of these weed management strategies, determine reasons for your implementation and/or preference. 1. Are you famil iar with Striga? If yes, what do you know about Striga (e.g., lifecycle, identification, effect on yield, seed transport)? 2. Have you ever faced challenges with Striga? When do you first notice Striga in your field (specifically at what physiological stage/ height)? 3. 4. 5. Are you aware of any practices used to treat Striga (e.g., manual pulling)? If so, please describe (e.g., timing, required in puts, etc.). 6. Are you aware of any practices used to prevent Striga (e.g., soil fertility techniques)? If so, please describe (e.g., timing, required inputs, etc.). 7. How did you hear about these practices? How did you learn about them (e.g., experimentation, extension, NGO, etc.)? 8. Would you consider any of these practices traditional? That is, agricultural extension, an NGO or an outside party did not promote them to you. They were passed on from generation to generation. Please indicate whic h ones. 9. Among the treatment practices you mentioned, do you implement any of them? If so, why not? 70 10. Among the preventative practices you mentioned, do you implement an y of them? If implement, why not? 11. What are some of the control methods you used because inputs were subsidized (or given free) to you? What were the inputs? Describe why you did them because the 12. How do the treatment and preventative practices control Striga (e.g., reduce emergence the next season, remove before seeds are mature, etc.)? Essentially, what are th e mechanisms or processes behind them that control Striga? 13. What goals or objectives do you take into account before choosing/implementing a Striga control practice? (Give the example of choosing a legume seed. You would perhaps look at taste, yield, cooki ng time, etc.) 14. Rank these traits from most important to least according to importance. Then, rank the practices you mentioned from most to least effective in controlling Striga. 15. Rank the methods you mentioned from most to least preferred. While they may be effective, we are trying to determine which ones are the practiced among farmers. Date:_____________________________________________________________ Beginning Time:____________________________________________________ Ending Time:_____________________ __________________________________ Location:__________________________________________________________ Cholinga: amagwiritsa ntchito Mawu oyambirira: Lero timafuna tikambilane z omwe mumadziwa ndi kuchita zokhuzana ndi kasamalilidwe ka zomera zosafunika mmunda. Cholinga cha kucheza kwathu ndikufuna kupeza uthenga ofunika kuti anthu a kafufuku ngati ineyo amvesese mmene kaufiti amasamalilidwa ndi alimi ngati inuyo. Kutengera ndi zo mwe mutafotokoze, ndifunsaso zifukwa zomwe munasankhira njira zomwe mukugwiritsa ntchito posamala ndi kuthana ndi zomera zomera zokha mmunda. 1. Mukudziwapo kalikonse kokhudzana ndi kaufiti? Ngati eya, mukudziwa chani zokhuzana ndi kaufiti? (monga mayendedwe a moyo, maonekedwe ake, mmene zimakhuzira zokolola). 2. Mwakumanako ndi mavuto ena liwonse ndi kaufiti? Munamuzindikila ali potani mmunda mwanu? (kakulidwe, katalikidwe) 3. Ngati mwakumanako ndi mavuti ndi kaufiti, zachitika kwa zaka zingati? 71 4. Ngat i munakhalako ndi kaufiti mmunda mwanu, anakhuzako zokolola zani? Zinakhuzika bwanji? 5. Mukudziwapo ndondomeko/njira ina iliyonse yomwe mungathe kuthana ndi kaufiti (monga kuzula pamanja)? ngati ilipo ifotokozeni (monga nthawi yoyenera, zipangizo zofunika)? 6. Mukudziwapo ndondomeko/njira ina iliyonse yomwe mumatsata poteteza kaufiti (monga njira zobwezeretsa nthaka)? ngati ilipo ifotokozeni? (monga nthawi yoyenera, zipangizo zofunika)? 7. Munadziwa bwanji za ndondomeko zimenezi? Munaziphunzira bwanji? (monga ku yesela, alangizi, mabungwe ndi ena otero) 8. Pa ndondomeko/njira zimenezi, ndi ziti zomwe zili zamakolo? Kutanthauza kuti alangizi, mabungwe kapena anthu ena obwera sanazakuphunzitseni. Izi ndi njira zomwe zakhala zikutsatidwa ndi mibadwa yonse. 9. Pa ndondomeko/njira zothana ndi kaufiti zomwe mwatchulazi, mukugwiritsa ntchito ziti? Chonde fotokozani zomwe mukugwiritsa ntchito (monga nthawi yoyenera, zipangizo zofunika ndi zina zotero). Pa zomwe simukugwiritsa ntchito, ndi chifukwa chani simukuzigwirits a ntchito? 10. Pa ndondomeko/njira zoteteza kaufiti zomwe mwatchulazi, mukugwiritsa ntchito ziti? Chonde fotokozani zomwe mukugwiritsa ntchito (monga nthawi yoyenera, zipangizo zofunika ndi zina zotero). Pa zomwe simukugwiritsa ntchito, ndi chifukwa chani sim ukuzigwiritsa ntchito? 11. Ndi ndondomeko ziti zothana ndi kaufiti zomwe munagwiritsa ntchito chifukwa zipangizo zinali zotsika mtengo mokuthanidzani kapena zinapatsidwa mwaulele? Zinali zipangizo zanji? Fotokozni chifukwa chimene munagwiritsa ntchito njirazi chifukwa zipangizo munapatsidwa mwa ulele? Ndi chifukwa chani simukanatsatila ndondomeko/njira zi popanda zipangizo zimenezi? 12. Ndondomekozi/ njirazi zinathana kapena zinateteza bwanji kaufiti? (monga kuchepesa kumera kwa kaufiti mu chaka china, kuthana naz o zisanayambe njere) 13. Ndi zinthu ziti zomwe mumaona musane sankhe ndondomeko/ njira yothana ndi kaufiti? (pelekani chitsanzo: mukamasankha mbewu ya mtundu wa nyemba mumaona kakomedwe kake, zokolola komanso nthawi yomwe zimatenga kuti zipsye) 14. Ikani zinthu zomwe mumaganizira musanasankhe ndondomeko/ njira yothana ndi kaufiti mu dongosolo kuyambira yofunika kwambiri kumalizira yosafunika. Mukatero, ikani ndondomeko/ njira zomwe munatchula zothana ndi kaufiti mu dongosolo kuyambila yomwe imagwira kwambiri kum alizila yosagwira bwino 72 15. Ikani ndondomeko/ njira zothana ndi kaufiti mu dongosolo kuyambira zomwe zimakondedwa pakati pa alimi kumalizira ndi zomwe sizikondwedwa. Njira zina zitha kukhala zogwira kwambiri koma tikufuna tidziwe zomwe alimi ambiri amakonda k ugwiritsa ntchito Tsiku: Nthawi yoyambila: Nthawi yomalizira: Malo: 73 APPENDIX 2. Survey Instrument Informed Consent Enumerator (say to respondent): Today we will be asking you about your knowledge about Striga and its control practices. You can contact Timothy Silberg or the Institutional Review Board at Michigan State University and/or withdraw from the study without penalty at any time. Wofunsa (nenani kwa wofunsidwa): Lero tikufunsani za mmene mumadziwira kaufiti komanso njira Michigan State University komanso muli ndi ufulu wosiya kuyankha mafunso ndipo palibe chilango chilichonse pochita izi. Section A. Basic memb er & household characteristics 1. District:______________________________________________________________________ 2. EPA: ________________________________________________________________________ 3. Respondent Name, Age, Gender & Education: _________________________ ______________ 4. Village: ______________________________________________________________________ 5. Date: ________________________________________________________________________ 6. Enumerator: __________________________________________________________________ 7. Household identification (HHID):__________________________________________________ Section B. Perceptions and awareness of Striga 8. Are you familiar with Striga? Y N Mumadziwa tchire la mmunda lotchedwa kaufiti? 9. Can you visibly tell the difference between annual weeds and Striga? Y N Mungathe kusiyanitsa pakati pa tchire lina (udzu wa mmunda) ndi kaufiti? a. If YES, how can you identify Striga among annual weeds? (List up to three) ( Co de A)______________________________________________ Ngati ndi choncho, mungamudziwe bwanji kaufiti pakati pa tchire la mtundu wina uliwonse? 10. Do you find there is a difference between the ways annual weeds affect your maize yield versus the way Striga affec ts your maize yield? Y N Mumatha kusiyanitsa mmene tchire lina lonse limavutitsira chimanga kuyerekeza ndi mmene kaufiti amachitira? a. If YES, how so? (List up to three) ( Code B)_________________________________ Ngati ndi choncho, zimasiyana bwanji? 11. Ca n you visibly tell where Striga is in your field before it emerges? Y N Muli ndi kuthekera kodziwa kuti pamalo pali kaufiti ngakhale asanamere? a. If YES, how can you tell Striga is present before it emerges from the soil? (List up to three) ( Code C)_________________________________________ Ngati ndi choncho, mumadziwa bwanji? 12. Are you aware how Striga attacks maize? Y N Mumadziwa mmene kaufiti amaonongera chimanga chathu? a. If YES, please describe (List up to three) ( Code D)______________________ Ngati ndi choncho, fotokozerani 13. What does the enumerator consider their general knowledge of Striga is? (Circle one) Wofunsa akuwona kuti woyankha akum udziwa bwanji kaufiti? (zungulizani chimodzi) 74 0 1 2 3 (Unaware) (Little Knowledge) (Some Knowledge) (Very Knowledgeable) Sakumudziwa Akumudziwa ndithu Akumudziwa kwambiri Section C. Striga history & impact 14. Please indicate the number of fields you cultivated this past season where maize was the primary crop (#)_____________________________________________________________ a chapitachi? a. How many fields had Striga? (#)____________________________________ Ndi minda ingati imene munamera kaufiti? High Fertility Plot 15. Did this plot have striga? Munganene kuti kaufiti ndi vuto lalikulu pa ulimi wanu? Y N *Note - If farmer responds NO, please go to #20) *Ngati mlimi ayankha kuti ayi, pitani ku funso #20) a. If YES , what year did these challenges begin?_________________________ 16. Comment on Striga emergence on the plot. Mu ndiuzeko za kameredwe ka kaufiti mu minda imene ili ndi vutoli? a. Striga emergence was patchy Y N Kaufiti anamera patalipatali i. Please comment on the soil conditions of this plot (List up to four) ( Code E )___________________________________ Nthaka ndiyotani mmindayi? b. Striga emergence extended across the entire plot Y N i. Please comment on the soil conditions of this plot (List up to four) ( Code E )___________________________ ________ Nthaka ndiyotani mmindayi? 17. Now think about this season. Mwa minda mwatchulayi, ndi chaka chiti chimene mudayamba kuonamo kaufiti? a. What month and week did you plant maize?_________________________ Tchulani mwezi komanso sabata imene munadzala chimanga? b. Did the maize plant begin to wilt before you saw Striga? Mbewu yanu ya chimanga inayamba kufota kaufiti asanamere? Y N i. What month and week did the maize begin to wilt? Ndi mwezi komanso sabata iti mmene chimanga chanu chinayamba kufo ta? ________________week #/month #) ii. At what physiological stage did the maize begin to wilt? (List 1) ( Code F )__________________________________________ Chimanga chinayamba kufota chitakula bwanji/motani? 18. Now think about once Striga emerged. Taganizani mme ne kaufiti anamera. a. What week (#) and month (#) did you begin seeing Striga emerge from the soil?___________________________________________________________________ Unali mwezi uti komanso sabata iti mmene kaufiti anayamba kumela? 75 b. At what physiological sta ge was the maize when Striga emerged? (List 1) ( Code F) _________________________________________________ Chimanga chinali chitakula bwanji/motani mmene kaufiti amaonekera? c. Please comment on the health of maize once Striga emerged (List up to three) ( Code G) _________________________________________ Thanzi la chimanga linali bwanji mmene kaufiti amamera? 19. Was there a cob at harvest? Y N Chimanga chinali chili ndi tiana mmene kaufiti amamera? a. If YES, please comment on cob size at harvest (List up to 3) ( Code G )_____________________________________ Ngati ndi choncho, Zisononkho zinali zazikulu bwanji pa nthawi yokolola? b. That season, what yield did you receive compared to others that did not have Striga (List 1) ( Code H )_________________________ Munakolola zochuluka bwanji poyerekezera ndi zaka mmbuyomu (mmene munalibe kaufiti)? *Note If farmer states that all fields had Striga, ask them how their yields compared to those of their neighbors * Ngati mlimi wanena kuti minda yonse inali ndi kaufiti, afunseni za zokolola zawo poyerekeza ndi minda yoyandikana nayo. c. In addition to Striga, did you face any other challenges on this plot relat ed to productivity? Panalinso mavuto ena omwe anakhudza ulimi wanu kupatulapo vuto la kaufiti? Y N I. List up to 3 ( Code I )_________________________ Low Fertility Plot 20. Did this plot have striga? Munganene kuti kaufiti ndi vuto lalikulu pa ulim i wanu? Y N *Note - If farmer responds NO, please go to #25) *Ngati mlimi ayankha kuti ayi, pitani ku funso #25) a. If YES , what year did these challenges begin?_________________________ 21. Comment on Striga emergence on the plot. Mundiuzeko za kameredwe ka kaufiti mu minda imene ili ndi vutoli? a. Striga emergence was patchy Y N Kaufiti anamera patalipatali i. Please comment on the soil conditions of this plot (List up to four) ( Code E )__ _________________________________ Nthaka ndiyotani mmindayi? b. Striga emergence extended across the entire plot Y N i. Please comment on the soil conditions of this plot (List up to four) ( Code E )______________________________ _____ Nthaka ndiyotani mmindayi? 22. Now think about this season. Mwa minda mwatchulayi, ndi chaka chiti chimene mudayamba kuonamo kaufiti? a. What month and week did you plant maize?_________________________ 76 Tchulani mwezi komanso sabata imene munadzala chimanga ? b. Did the maize plant begin to wilt before you saw Striga? Mbewu yanu ya chimanga inayamba kufota kaufiti asanamere? Y N i. What month and week did the maize begin to wilt? Ndi mwezi komanso sabata iti mmene chimanga chanu chinayamba kufota? ________________week #/month #) ii. At what physiological stage did the maize begin to wilt? (List 1) ( Code F )__________________________________________ Chimanga chinayamba kufota chitakula bwanji/motani? 23. Now think about once Striga emerged. Taganizani mmene kaufiti anamera. a. What week (#) and month (#) did you begin seeing Striga emerge from the soil?___________________________________________________________________ Unali mwezi uti komanso sabata iti mmene kaufiti anayamba kumela? b. At what physiological stage was the maize when Striga emerged? (List 1) ( Code F) _________________________________________________ Chimanga chinali chitakula bwanji/motani mmene kaufiti amaonekera? c. Please comment on the health of maize once Striga emerged (List up to three) ( Code G) _________________________________________ Thanzi la chimanga linali bwanji mmene kaufiti amamera? 24. Was there a cob at harvest? Y N Chimanga chinali chili ndi tiana mmene kaufiti amamera? a. If YES, please comment on cob size at harvest (List up to 3) ( Code G )_____________________________________ Ngati ndi choncho, Zisononkho zinali zazikulu bwanji pa nthawi yokolola? b. That season, what yield did you receive compared to others that did not have Striga (List 1) ( Code H )_________________________ Munakolola zochuluka bwanji poyerekezera ndi zaka mmbuyomu (mmene munalibe kaufiti)? *Note If farmer states that all fields had Striga, ask them how their yields compared to those of their neighbors * Ngati mlimi wanena kuti minda yonse inali ndi kaufiti, afunseni z a zokolola zawo poyerekeza ndi minda yoyandikana nayo. c. In addition to Striga, did you face any other challenges on this plot related to productivity? Panalinso mavuto ena omwe anakhudza ulimi wanu kupatulapo vuto la kaufiti? Y N i. List up to 3 ( Code I )_________________________ Section D. Methods used to address Striga 25. practices you have heard about that control Striga. A treatment practice involves the removal of Striga once it has emerged from the soil. 25a. What treatment practice s have you heard? ( Ndi njira ziti zothana ndi kaufiti zomwe munamvapo?) 25b.When did you hear about them? ( Munazimva liti?) 25c. How did you hear about them? ( Munazimva bwanji?) 25d. What were the benefits you heard about? ( Munamva kuti ubwino wake ndiwotani?) 77 Wofunsa (nenani kwa wofunsidwa): Tsopano, ndikufuna ndikufunseni mafunso okhudzana ndi njira zothana ndi kaufiti zomwe munamvapo. Njira yothana ndi kaufiti ndi iyo yomwe imatengera mlimi kuchotsa kaufiti akamera mmunda mwake. CODE J CODE K CODE L CODE M i. 1 st Practice i. Years ago i. (List up to 3) i. (List up to 3) ii. 2 nd Practice ii. Years ago ii. (List up to 3) ii. (List up to 3) iii. 3 rd Practice iii. Years ago iii. (List up to 3) iii. (List up to 3) iv. 4 th Practice iv. Years ago iv. (List up to 3) iv. (List up to 3) 78 26. Enumerator (says to respondent): Next, I would like to ask some questions about the treatment fi eld. Wofunsa (Nenani kwa wofunsidwa): Tsopano, ndikufuna ndikufunseni mafunso okhudzana ndi njira zothana ndi kaufiti zomwe munagwiritsapo ntchito. Muonenetsetse kuti mwatidziwitsa njira zomwe 26a. What treatment practices did you implement? Ndi njira ziti zothana ndi kaufiti zomwe munazitsatira? CODE N 26b. When did you first begin implementing them? Munayamba kuzitsatira liti? 26c. What did you do with the Striga? Kaufiti amene munamuc hotsayo munapanga naye chiyani? CODE O 26d. What happened (in terms of Striga control and secondary benefits)? Chinachitika (kumbali yoteteza kaufiti komanso ubwino wake kuposera apo)? CODE P 26e. How many seasons did it take for you to s ee these results? Panapita zaka zingati kuti inu muyambe kuona zotsatira? CODE Q 26f.Did you stop or continue the practice after seeing these results? Munasiya kapena kupitiriza njirazo mutaona zotsatira zakezo? 26g. Why did you stop or continue the practice? CODE P i. 1 st Practice i. Year i. (List up to 2) i. (List up to 3) i. i. Stop (0) / Continue (1) i. (List up to 3) ii. 2 nd Practice ii. Year ii. (List up to 2) ii. (List up to 3) ii. ii. Stop (0) / Continue (1) i. (List up to 3) iii. 3 rd Practice iii. Year iii. (List up to 2) iii. (List up to 3) iii. iii. Stop (0) / Continue (1) i. (List up to 3) iv. 4 th Practice iv. Year iv. (List up to 2) iv. (List up to 3) iv. iv. Stop (0) / Continue (1) i. (List up to 3) 79 27. Enumerator (says to respondent): Also, I would like to know whom you shared the positive/negative results from implementing these practices. Wofunsa (Nenani kwa wofunsidwa): komanso, ndimafuna nditadziwa kuti munauza ndani za zotsatira zabwino/zoipa za njira zimene munatsatirazo. *Note - Before asking farmers who they shared results with, transcribe the control practices they mentioned and their respective outcomes from question 26a and 26b in column 27a and 27b, respectively. *Musanawa funse alimi za amene anawauza za zotsatira, akumbutseni za njira zothana ndi kaufiti ndi zotsatira zake zimene azitchula mu mafunso 26a ndi 26b mu ndandanda 27a ndi 27b. 27a. Treatment practice (Refer to 26a) Njira yothana ndi kaufiti (muonere funso 26a) 27b. List of outcomes (Refer to 26d) Zotsatira (muonere funso 26b) 27c. Who did you share these results with? Munauzako ndani za zotsatira? CODE R 27d. How many? Munawauza zotsatira zingati? CODE S i. 1 st Practice i.1. i.1. i.1. i.2. i.2. i.2. i.3. i.3. i.3. ii. 2 nd Practice ii.1. ii.1. ii.1. ii.2. ii.2. ii.2. ii.3. ii.3. ii.3. iii. 3 rd Practice iii.1. iii.1. iii.1. iii.2. iii.2. iii.2. iii.3. iii.3. iii.3. iv. 4 th Practice iv.1. iv.1. iv.1. iv.2. iv.2. iv.2. iv.3. iv.3. iv.3. 80 28. Enumerator (says to respondent): implemented, but would have liked to in the past or in the future. Wofunsa (Nenani kwa wofunsidwa): Pomaliza, ndimafuna nditadziwa za njira zomwe simunathe *Note Make sure no practice listed in column 28a was listed in 27a. *Wonetsetsani kuti njira zotchulidwa mu mndandanda wa mayankho a 28a zisafanane ndi njira zomwe zatchulid wa kale mu mdandanda wa mayankho a 27a . 28a. What treatment practices would you have like to Ndi njira ziti zothana ndi kaufiti mukadakonda mukadatsatira mmbuyomu koma simunathe kutero? CODE T 28b. What was the reason you could not implement the treatment practice? kaufitiyi? CODE U i. 1 st Practice i. (List up to 3) ii. 2 nd Practice ii. (List up to 3) iii. 3 rd Practice iii. (List up to 3) iv. 4 th Practice iv. (List up to 3) 81 29. Enumerator (says to respondent): Now I would like to ask you about some practices you have heard about that prevent Striga. These practices are different from the previous ones you mentioned earlier. These practices would be implemented before you see Striga so it will not emerge from the soil in the future. There are multiple ways you can prevent Striga. Some practices you may have heard of, but are not aware of or consider them as preventati ve practices. These include soil fertility management practices, which improve soil texture, decrease acidity and increase nitrogen/phosphorous in the soil. Essentially, these practices enhance soil fertility. In doing so, these practices make less favorab le soil conditions for Striga to spread. Wofunsa (Nenani kwa wofunsidwa): Tsopano ndikufunsani za njira zina zomwe munamva zomwe zimateteza kaufiti mminda mwanu. Njirazi ndi zosiyana ndi zomwe mwatchula kale. Njirazi zingatsatidwe kaufiti asanamere mmunda mwathu ndi cholinga chokuti kaufitiyo asamere mtsogolomu. Pali njira zosiyanasiyana zomwe mungapewere kaufiti. Pali njira zina zoti munazimvapo, koma simukuzidziwa mmene zimatsadwira kapena kuzitenga ngati njira zopewera kaufiti. Izi ndi monga kupititsa ch onde patsogolo, zomwe zimathandizira kuti nthaka isakanikilike bwino, kuchepetsa michere yowononga komanso kuonjezera Michele yomwe ili yofunikira pa kakulidwe ka mbeu zathu.chachikulu njirazi zimapanga nthaka yathu kuti isalore kaufiti kuti afalikire mmunda mwathu. 29a. What preventative practices have you heard? Ndi njira ziti zopewera kaufiti zomwe munazimvapo? CODE V 29b. When did you hear about them? Munazimva liti? CODE W 29c. How did you hear about them? Munazimva kudzera mu njira yanji? CODE X 29d. What were the benefits you heard about? Ndi ubwino wanji wa njirazi umene munamva? CODE Y i. 1 st Practice i. Years ago i. (List up to 3) i. (List up to 3) ii. 2 nd Practice ii. Years ago ii. (List up to 3) ii. (List up to 3) iii. 3 rd Practice iii. Years ago iii. (List up to 3) iii. (List up to 3) iv. 4 th Practice iv. Years ago iv. (List up to 3) iv. (List up to 3) v. 5 th Practice v. Years ago v. (List up to 3) v. (List up to 3) vi. 6 th Practice vi. Years ago vi. (List up to 3) vi. (List up to 3) 82 30. Enumerator (says to respondent): Next, I would like to ask some questions about some preventative practices you have implemented across an entire field. Wofunsa (Nenani kwa wofunsidwa): Tsopano ndikufuna ndifunse mafunso okhudzana ndi zina mwa njira zopewera kaufiti zomwe munagwiritsapo ntchito. Chonde tidziwitseni njira zom we 30a. What preventative practices did you implement? Ndi njira ziti zopewera kaufiti zomwe munagwiritsapo ntchito? CODE Z 30b. When did you first begin implementing them? Munayamba kuzigwiritsa ntchito liti? 30c. What happened (in terms of Striga control and secondary benefits)? Chinachitika yoteteza kaufiti komanso ubwino wake kuposera apo)? CODE AA 30d. How many seasons did it take for you to see these results? Zinatengera zaka zingati ku ti inu muyambe kuona zotsatira? CODE BB 30e.Did you stop or continue the practice after seeing these results? Munasiya kapena kupitiriza njirazo mutaona zotsatira zakezo? 30f. Why did you stop or continue? Chinachitika yoteteza kaufiti komanso ubwino wake kuposera apo)? CODE AA i. 1 st Practice i. Year i. (List up to 3) i. (List up to 1) i. Stop (0) / Continue (1) i. (List up to 3) ii. 2 nd Practice ii. Year ii. (List up to 3) ii. (List up to 1) ii. Stop (0) / Continue (1) ii. (List up to 3) iii. 3 rd Practice iii. Year iii. (List up to 3) iii. (List up to 1) iii. Stop (0) / Continue (1) iii. (List up to 3) iv. 4 th Practice iv. Year iv. (List up to 3) iv. (List up to 1) iv. Stop (0) / Continue (1) iv. (List up to 3) v. 5 th Practice v. Year v. (List up to 3) v. (List up to 1) v. Stop (0) / Continue (1) v. (List up to 3) vi. 6 th Practice vi. Year vi. (List up to 3) vi. (List up to 1) vi. Stop (0) / Continue (1) vi. (List up to 3) 83 31. Enumerator (says to respondent): Also, I would like to know whom you shared the positive/negative results from implementing these practices. Wofunsa (nenani kwa wofunsidwa): Komanso, ndimafuna nditadziwa kuti munauzako ndani za zotsatira zabwino/zoipa kuch okera njira zimenezi. *Note - Before asking farmers who they shared results with, transcribe the control practices the mentioned and their respective outcomes from question 30a and 30b in column 31a and 31b, respectively. *Musanafunse alimi za omwe anawauza za zotsatira, akumbutseni za njira zothana ndi kaufiti ndi zotsatira zake zomwe anatchula kale mu mafunso 30a ndi 30b komanso mu ndandanda wa 31a ndi 31b. 31a. Preventative practice (Refer to 30a) Njira zopewera kaufiti (Onerani funso 30a) 31b. List of outcomes (Refer to 30c) Ndandanda wa zotsatira (onerani funso 30b) 31c. Who did you share these results with? Munauza ndani za zotsatirazi? CODE CC 31d. How many? zingati? CODE DD i. 1 st Practice i.1. i.1. i.1. i.2. i.2. i.2. i.3. i.3. i.3. ii. 2 nd Practice ii.1. ii.1. ii.1. ii.2. ii.2. ii.2. ii.3. ii.3. ii.3. iii. 3 rd Practice iii.1. iii.1. iii.1. iii.2. iii.2. iii.2. iii.3. iii.3. iii.3. iv. 4 th Practice iv.1. iv.1. iv.1. iv.2. iv.2. iv.2. iv.3. iv.3. iv.3. v. 5 th Practice v.1. v.1. v.1. v.2. v.2. v.2. v.3. v.3. v.3. vi. 6 th Practice vi.1. vi.1. vi.1. vi.2. vi.2. vi.2. vi.3. vi.3. vi.3. 84 32. More specifically, I would like to know where you receive your inputs from to complete the preventative practices you mentioned. Makamaka, ndimafuna nditadziwa kuti zipangizo za ulimi zomwe munagwiritsa ntchito popewa kaufiti munazipeza kuti? *Note - Before asking farmers about the sources of their inputs, transcribe the control practices they m entioned from question 31a in column 32a. *Musanafunse alimi za kumene anapeza zipangizo zawo, akumbutseni za njira zopewera zimene anatchula mu mafunso 31a mu ndandanda was 32a. 32a. Preventative practice (Refer to 31a) Njira zopewera Kaufiti 31a 32b. Week & Month Sabata ndi mwezi 32c. Specify crop if rotation/ intercropping listed Tchulani mbeu ngati mwayidzala mwa kasinthasintha kapena kasakaniza CODE EE 32d. Specify seed source if rotation/ intercropping listed Mundiuze za kumene munapeza mbeu (ngati kasinthasintha/ kasakaniza zatchulidwa) CODE FF 32e. Specify type of fertilizer or manure if fertilizer/ manure application was listed Mundiuze mtundu wa fetereza kapena manyowa (ngati fetereza/manyo wa zatchulidwa) CODE GG 32f. Specify source of fertilizer or manure if fertilizer/ manure application was listed Mundiuze za kumene munapeza fetereza/manyo wa (ngati zinatchulidwa) CODE HH 32g. How was the fe rtilizer/ manure incorporated? Fetereza/ manyowa anathiridwa motani? CODE II i. 1 st practice i.1. Week # i.2. Month # i. (List up to 3) i. (List up to 3) i. (List up to 3) i. (List up to 3) i.1. Degraded Y N i.2. (List up to 2) ii. 2 nd practice ii.1. Week # ii.2. Month # ii. (List up to 3) ii. (List up to 3) ii. (List up to 3) ii. (List up to 3) ii.1. Degraded? Y N ii.2. (List up to 2) iii. 3 rd practice iii.1. Week # iii.2. Month # iii. (List up to 3) iii. (List up to 3) iii. (List up to 3) iii. (List up to 3) iii.1. Degraded? Y N iii.2. (List up to 2) iv. 4 th practice iv.1. Week # iv.2. Month # iv. (List up to 3) iv. (List up to 3) iv. (List up to 3) iv. (List up to 3) iv.1. Degraded? Y N iv2. (List up to 2) v. 5 th practice v.1. Week # v.2. Month # v. (List up to 3) v. (List up to 3) v. (List up to 3) v. (List up to 3) v.1. Degraded? Y N v.2. (List up to 2) vi. 6 th practice vi.1. Week # vi.2. Month # vi. (List up to 2) vi. (List up to 3) vi. (List up to 3) vi. (List up to 3) vi.1. Degraded? Y N vi.2. (List up to 2) 85 33. Enumerator (says to respondent): implemented, but would have liked to in the past or in the future. Wofunsa (Nenani kwa wofunsidwa): Pomaliza, ndimafuna nditadziwa za njira zimene simunathe kuzitsatira, koma mukanakonda mutazitsatira. *Note Make sure no practice listed in column 32a was listed in 33a. *Wonetsetsani kuti njira zatchulidwa mu ndandanda 32a zisafanane ndi zimene zatchulidwa mu 33a . 33a. What preventative practices would you have like to have implemented in the past, bu Ndi njira ziti zopewera kaufiti zomwe mukadakonda simunathe kutero? CODE JJ 33b. Specify type of manure or fertilizer if farmer mentioned manure/ fertilizer application Mundiuze mtundu wa fetereza kapena manyowa (ngati fetereza/manyo wa zatchulidwa) CODE KK 33c. Specify seed if farmer mentioned intercropping/crop rotation Tchulani mtundu wa mbeu (ngati mlimi watchula mbeu) CODE LL 33d. What was the reason you could not implement the preventative practice? Ndi chifukwa chiyani munakanika kutsatira njira yopewera kaufiti? CODE MM i. 1 st Practice i. (List up to 2) i. (List up to 2) i. (List up to 3) ii. 2 nd Practice ii. (List up to 2) ii. (List up to 2) ii. (List up to 3) iii. 3 rd Practice iii. (List up to 2) iii. (List up to 2) iii. (List up to 3) iv. 4 th Practice iv. (List up to 2) iv. (List up to 2) iv. (List up to 3) v. 5 th Practice v. (List up to 2) v. (List up to 2) v. (List up to 3) vi. 6 th Practice vi. (List up to 2) vi. (List up to 2) vi. (List up to 3) 86 34. Enumerator (says to respondent): Finally, I would like to know about some soil fertility impl emented across an entire field. Wofunsa (Nenani kwa wofunsidwa): Tsopano ndikufuna ndifunse mafunso okhudzana ndi zina mwa njira zopewera kaufiti zomwe munagwiritsapo ntchito. Chonde tidziwitseni njira zomwe 34a. What soil fertility practices did you implement? Ndi njira ziti zopewera kaufiti zomwe munagwiritsapo ntchito? CODE JJ 34b. How did you hear about them? Munazimva kudzera mu njira yanji? CODE X 34c. When did you hear about them Sabata ndi mwezi? 34d. After hearing about them, how long did it take until you fully implemented them across an entire field? i. 1 st Practice i. (List up to 3) i. Year i. Years ii. 2 nd Practice ii. (List up to 3) ii. Year ii. Years iii. 3 rd Practice iii. (List up to 3) iii. Year iii. Years iv. 4 th Practice iv. (List up to 3) iv. Year iv. Years v. 5 th Practice v. (List up to 3) v. Year v. Years vi. 6 th Practice vi. (List up to 3) vi. Year vi. Years Section E. Food and labor preferences Enumerator (says to respondent): choose between cultivating a monoculture of maize or another cropping system across 1 ha . Wofunsa (Nenani kwa wofunsidwa): Tsopano ndikuyerekezerani njira zing apo zosiyanasiyana ndipo mukuyenera kusankhapo imodzi pakati pa kalimidwe ka chimanga pachokha kapena kalimidwe ka mtundu wina pa munda wokwana hekitala imodzi. 35. Enumerator (says to respondent): Suppose you have the choice of accepting 20 (50kg) bags of tr aditional maize (Option A) or a less amount of early - maturing maize (Option B). There is a chance you may not be able to receive 20 bags of traditional maize. With the early maturing maize, however, you will receive the specified amount (e.g., 15 bags) wit hout a chance of losing it to Striga. Wofunsa (Nenani kwa wofunsidwa): Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50kg) a chimanga cha makolo (chisankho A) kapena chimanga chocheperapo koma chocha msanga (chisankho B). *Note - Do NOT show this table to the farmer. You will present the choices in an iterative manner. - yield of 15 bags of early maturing maize/ha ice, then present T2. * Musaonetse zimene mukufunsazi kwa mlimi. Mudzifunsa zisankhozi polankhulana basi. Tchulani kasinthanitsa woyamba (mwachitsanzo, Mungatenge zokolola zotsimikizika zokwana matumba 87 15 a chimanga chocha msanga pa hekitala kapena mungafune matumba 20 pa hekitala a chimanga cha makolo?). Ngati sasankha chisankho choyamba, afunseni kasinthanitsa T2 Choice Option A Chisankho A (Traditional Maize) ( Chimanga cha makolo) Option B Chisankho B ( Early - maturing maize ) (Chimanga chocha msanga) T1 20 Bags 15 Bags T2 20 Bags 16 Bags T3 20 Bags 17 Bags T4 20 Bags 18 Bags T5 20 Bags 19 Bags T6 20 Bags 20 Bags If Option B was not chosen, how many bags would it take for them to switch:________ maganizo:______ 88 36. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize for each season for two years (Option A) or a larger amount of food in if you intercrop pigeon pea and maize within the same field (Option B). For each of the six choice sets presented, check the box for the option you prefer? Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha chakudya mu chaka choyamba kapena chachiwiri ngati mungalime chimanga ndi nandolo nse mwa * Note Emphasize that both crops are planted simultaneously, but pigeon pea matures later into the season once the farmers have harvested maize. Tsimikirani mfu ndo yokuti mbeu zonsezo zimadzalidwa pa kamodzi, koma kaufiti amakhwima mochedwerapo alimi atakolora kale chimanga. Choice Option A (Maize Monoculture) Option B (Maize - Pigeon Pea Intercrop ) T1 20 Maize Bags 15 Maize Bags + 7 Pigeon Pea Bags T2 20 Maize Bags 16 Maize Bags + 7 Pigeon Pea Bags T3 20 Maize Bags 17 Maize Bags + 7 Pigeon Pea Bags T4 20 Maize Bags 18 Maize Bags + 7 Pigeon Pea Bags T5 20 Maize Bags 19 Maize Bags + 7 Pigeon Pea Bags T6 20 Maize Bags 20 Maize Bags + 7 Pigeon Pea Bags If Option B was chosen, how many bags in year one, would it take for them to switch:___________ maganizo:______ 37. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize for each season for two years (Option A) or a larger amount of maize in a second season if you cultivate sole soybean the first year (Option B). For each of the six choice sets presented, check the box for the option you prefer? Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha chaka chachiwiri ngati mungalima soy chisankho cha mlimi. Choice Option A Chisankho A (Maize 1 st Season - > Maize 2 nd Season) Option B Chisankho B ( Soybean 1 st Season - > Maize 2 nd season ) T1 20 Maize - > 20 Maize Bags 30 Soybean - > 26 Maize Bags T2 20 Maize - > 20 Maize Bags 30 Soybean - > 28 Maize Bags T3 20 Maize - > 20 Maize Bags 30 Soybean - > 30 Maize Bags T4 20 Maize - > 20 Maize Bags 30 Soybean - > 32 Maize Bags T5 20 Maize - > 20 Maize Bags 30 Soybean - > 34 Maize Bags T6 20 Maize - > 20 Maize Bags 30 Soybean - > 36 Maize Bags If Option B was not chosen, how many bags in year two, would it take for them to switch:________ 89 maganizo:______ 38. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize for a single season (Option A) or a larger amount of food in one season if you intercrop cowpea and maize within the same field (Option B). For each of the six choice sets present ed, check the box for the option you prefer? Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha makolo kwa chaka chimodzi (chisankho A) kapena matumba ochulukirapo a chakudya ngati unda womwewo (Chisankho B). Pa likusonyeza chisankho cha mlimi. If Option B was not chosen, how many bags in year one, would it take for them to switch:________ maganizo:______ 39. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize (Option A) from weeding twice or a larger amount for more labor, not tilling the land and reserving maize residues for mulching the next season (Option B). More labor would entail three weedings as well as cutting and applying crop residues prior to sowing maize. For each of the six choice sets presented, check the box for the option you prefer? Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha makolo ch ifukwa chopalira kawiri (chisankho A) kapena matumba ochulukirapo chifukwa cha ntchito yochulukirapo (chisankho B). Ntchito yochulukirapo ikutanthauza kupalira katatu komanso kuthira manyowa mmudza tisanadzale chimanga. Pa chisankho chilichonse mwa zisankh mlimi. If Option B was not chosen, how many bags in year two, would it take for them to switch:_______ maganizo:______ Choice Option A (Maize 1 st Season - > Maize 2 nd Season) Option B ( Intercrop 1 st Season T1 20 Maize Bags 15 Maize + 6 Cowpea Bags T2 20 Maize Bags 16 Maize + 6 Cowpea Bags T3 20 Maize Bags 17 Maize + 6 Cowpea Bags T4 20 Maize Bags 18 Maize + 6 Cowpea Bags T5 20 Maize Bags 19 Maize + 6 Cowpea Bags T6 20 Maize Bags 20 Maize + 6 Cowpea Bags Choice Option A Chisankho A (2 Weedings) Option B Chisankho B ( 3 Weedings + No Till + Crop Residue App.) T1 20 Bags - > 20 Maize Bags 20 Bags - > 21 Maize Bags T2 20 Bags - > 20 Maize Bags 20 Bags - > 22 Maize Bags T3 20 Bags - > 20 Maize Bags 20 Bags - > 23 Maize Bags T4 20 Bags - > 20 Maize Bags 20 Bags - > 24 Maize Bags T5 20 Bags - > 20 Maize Bags 20 Bags - > 25 Maize Bags T6 20 Bags - > 20 Maize Bags 20 Bags - > 26 Maize Bags 90 40. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize (Option A) or a larger amount by applying herbicide (Option B). Option A comes with enough fertilizer for a 1 - acre field to receive 20 bags. Option B comes with a sprayer and enough herbicide to apply across a 1 - acre field. For each of the six choice sets presented, check the box for the option you prefer? Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha makolo (Chisankho A) kapena matumba ochulukirapo pothira mankhwala opha tchire mmunda (Chisankho B). Chisankho A chikubwera ndi matumba a fetereza wokwanira kuthira munda wa 1 acre kuti mudzapate matumba 20. Chisankho B chikubwera ndi sprayer komanso mankhwala opha tchire okwa chisankho cha mlimi. If Option B was not chosen, how many bags would it take for them to switch:________ Ngati B wa maganizo:______ Choice Option A Chisankho A (2 Weedings + 1 Fertilizer) Option B Chisankho B (No Weeding + Herbicide) T1 20 Bags 21 Bags T2 20 Bags 22 Bags T3 20 Bags 23 Bags T4 20 Bags 24 Bags T5 20 Bags 25 Bags T6 20 Bags 26 Bags 91 41. Suppose you have the choice of accepting 20 (50kg) bags of traditional maize (Optio A) or larger amount by applying herbicide (Option B). Option A will cost 31,250MKW for 25kgs of seed. You will not ba able to apply fertilizer. Optoin B will cost 31,250 fo r seed, 20,000 for a sprayer and 3,650 for herbicide. So Option A will cost 31,250 and Option B will cost 54,900MW. For each of the six choice sets presented, chock the box for the option you prefer. Tingoyerekeza mwapatsidwa mwayi wolandira matumba 20 (a 50 kg) a chimanga cha makolo (Chisankho A) kapena matumab ochulukirapo a chimanga pothira mankhwala opha tchire mmunda (Chisankho B). Chisankho A chikutengerani MK 31,250 pogula mbeu yokwana 25kg. Simukyenera kuthira fetereza. Chisankho B mugwiritsa Ntchi to MK31,250 kugulira mbeu, MK20,000 kugulira sprayer komanso MK3,650 kugulira mankhwala okupha tchire ndipo chisankho B ndalama YOnse pamodzi ikuwana MK54,900. Pa chisankho chi sankho cha mlimi. Choice Option A Chisankho A (No Herbicide) Option B Chisankho B ( Herbicide) T1 20 Bags 22 Bags T2 20 Bags 24 Bags T3 20 Bags 26 Bags T4 20 Bags 28 Bags T5 20 Bags 30 Bags T6 20 Bags 32 Bags If Option B was not chosen, how many bags in would it take for them to switch:________ maganizo:______ Section F. Debriefing Question Answer (CODE NN) Enumerator Assessment of Data Quality/Farmers Ability to Recall Information Time to complete questionnaire 92 APPENDIX 3. K ey for Survey Questionnaire Code A 1. Red flowers (maluwa ofiila) 2. Yellow flowers (maluwa a chikasu) 3. Pink flowers (maluwa ofiilirako) 4. Small red roots (timizu tofiila) 5. It grows underground unlike an annual (maka uyo yekhayo amakulira pansi pa nthaka osaonekera) 6. Tiny thin leaves (spike leaf arrangement) 7. White roots 8. Grows on maize plant 9. Smaller/thinner than annual weeds 99. Other (specify) ( Zina [tchulani]) Code B 1. No yield vs some yield (osakolora kalikonse kapena kukolora zochepa kwambiri) 2. ¼ of what you would receive ( limodzi mwa ma gawo anayi a zimene mumayembekezera [quarter]) 3. ½ of what you would receive ( theka la zimene mumayembekezera) ode* 4. ¾ of what you would receive ( magawo atatu mwa anayi a zomwe mumayembekezera) 5. Short/stunted ( chachifupi/chokwinimbira) 6. Skinny/thin ( choonda/toonda) 7. Wilting 8. Poor germination 9. Cob formed early (before maize plant was fully grown) chimanga chimabereka mwamsanga (cgisanakule) 99. Other (specify) ( Zina [tchulani]) Code C 1. Maize wilts even though there is water (Chimanga chimafota ngakhale pamakhala pali chinyontho) 2. Maize wilts even though there is fertilizer (chimanga chimafota ngakhala pamakhala pathiridwa fetereza) 3. Maize wilts before 2 nd weeding (chimanga chomafota tisanapalire kachiwiri) 4. Tassle forms early (before maize plant is fully grown) c himanga chimamasula mwamsanga (chisanakule) Code* Code* 5. Cob forms early (before maize plant is fully grown) chimanga chimabereka mwamsanga (cgisanakule) 6. Yellowing of leaves 7. Stunted growth/early maturity 8. Thin maize stalk 9. Poor germination 99. Other (specify) ( Zina [tchulani]) Code D 1. Removes nutrients from soil (amachotsa chakudya cha mu nthaka) 2. Removes water from soil (amachotsa madzi mu nthaka) 3. Poisons roots (mizu yake ndi poizoni/chiphe) 4. Attaches to roots (amamera pa mizu inzake) Code* 5. Removes water from plant (amamwa madzi mu zomera zathu) 6. Removes nutrients from plant. (amayamwa chakudya kuchoka mu zomera zathu) 7. Harbors pests 99. Other (specify) ( Zina [tchulani]) CODE E 1. Heavy infestation of non - Striga weeds ( lochuluka chire losakhala kaufiti) 2. Acidic soil ( la mchere wa acid) 3. Sandy soil ( la mchenga) 4. Little to no soil organic matter (manure) ( popanda chonde chokwanira [manyowa]) 5. Eroded soil ( nthaka yokokololoka) 6. Low fertility ( F etereza wosakwanira) 7. Water logging/hard pan ( la madzi ochuluka) 8. Iron/red soil 9. Hard pan (clay soil) 99. Other (specify) ( Zina [tchulani]) CODE F 1. Before tasseling (chisanamasule) 2. At tasseling (chitamasula) 3. Before you see the cob but after tasseling (chitamasula 4. Once cob appears (chitabereka tiana) 5. Once cobs have reached full maturity (chimanga chitakhwima) 99. Other (specify) ( Zina [tchulani])koma chisanabereke ana) CODE G 1. Short/stunted plant ( c hachifupi/chokwinimbira) | Short/stunted cob 2. Skinny/thin ( choonda/toonda) | Skinny/thin cob 3. Poor Germination ( sizinamere bwino) 4. Maize leaves turn purple ( Masamba a chimanga amasanduka mtundu wa purple) 5. Maize leaves turn yellow ( Masamba a chimanga amasanduka mtundu wa chikasu) 6. Maize leaves turn brown before harvest (early leaf senescence) ( Masamba a chimanga amasanduka mtundu wotuwa chisanakoloredwe [masamba amauma msanga]) 7. Early maturity (bearing cob before fully grown) 99. Other (s pecify) ( Zina [tchulani]) 93 CODE H 1. Normal/no difference ( zabwinobwino) 2. ¼ of what you would receive ( limodzi mwa ma gawo anayi a zimene mumayembekezera [quarter]) 3. ½ of what you would receive ( theka la zimene mumayembekezera) 4. ¾ of what you would receive ( magawo atatu mwa anayi a zomwe mumayembekezera) 5. Basically nothing 6. 1/3 of what you would receive 99. Other (specify) ( Zina [tchulani]) CODE I 0. None 1. Erosion Kukokoloka kwa nthaka 2. Drought 3. Lack of fertilizer kusowa kwa fetereza 4. Lack of compost, manure, mulch, etc., 5. Soil acidity mchere wa mu dothi 6. Low soil fertility kuchepa kwa chonde mu nthaka CODE I 7. Waterlogging 8. Pests, disease tizilombo, matenda 9. Lack of seed 10. Annua l weed pressure; too many weeds 11. Flooding 12. Termites 13. Lodging 14. Lack of labor 15. Illness 99. Other (specify) ( Zina [tchulani]) CODE J 0. Nothing 1. Kupalira (scraping) 2. Kuzulira atapanga maluwa (uprooting after flowering) 3. Kuzulira asanapange maluwa (before flowering) 4. Kusenda/Kuojekera (cover weeds with soil) 5. Kubandira (banking) 6. Herbicide ( Mankhwala wopha tchire) 7. Deep tillage ( kulima mozama/mwakuya) 8. Point manure applicat ion ( Kuika manyowa pa phando lodzalira) 9. Burning affected area 10. Point fertilizer application 11. Point manure application 12. Point maize bran application 13. Point ash application 14. Remove and bury in a deep pit 99. Other (specify) ( Zina [tchulani]) CODE K 1. 0 - 2 2. 3 - 5 3. 6 - 10 4. 10+ CODE L 1. NGO/Inter. Org. ( Mabungwe wosakhala a boma) 2. Radio ( Wailesi) 3. Poster/Hand Out ( Postala/zojambulidwa pa pepala) 4. Demonstration trial ( Munda wachionetsero) 5. Extension agent ( Alangizi) 6. Mar ket ( Ku msika) 7. Experimentation ( Kuyeselera/kafukufuku ) 8. Neighbor/Farmer ( Wokhala moyandikana naye/mlimi) 9. Farmer group/coop ( ku gulu/bungwe la alimi) 10/14. Family, relatives 11. Tradition Za makolo 12. Agro - dealer ( Wogulitsa zipangizo za ulimi) 13. Intuition ( Kungopanga poganiza kuti ndizotheka) 15. School 99. Other (specify) ( Zina [tchulani]) CODE M 1. Striga would come back less in short term (same season) ( Kaufiti anameranso pasanadutse nthawi yaitali) 2. Striga would come back less in the long term (next season) ( 3. Striga would not come back at all ( Kaufiti sanamerenso) 4. Soil fertility would increased ( Chonde chimaonjezereka) 5. Biomass for fuel o r fodder would increase ( Mapesi amachuluka) 6. Maize yield would increase in the short term (same season) ( 7. Maize yield would increase in the long term (next season) ( 8. Ag gregate food production would increase ( Chakudya chimachuluka) 9. Pest incidence would decrease ( Tizilombo toononga mbeu timachepa) 10. Profit (from on - farm production) would increase ( Phindu [purofiti] lochokera ku zokolola zathu limachuluka) 11. On - farm labor would decrease ( Ntchito yogwira pamundapo imachepa) 12. Overall weed pressure decreased 13. Delayed Striga emergence 14. Improved water retention/soil moisture holding capacity 15. Reduced erosion; improved soil structure/texture 16. Sustained Striga control; prevented Striga problem from getting worse 99. Other (specify) ( Zina [tchulani]) 94 CODE R 0. None/nobody 1. Neighbor/Farmer ( Wokhala moyandikana naye/mlimi) 2. Farmer group/coop ( ku gulu/bungwe la alimi) specify) ( Zina [tchulani]) 3. NGO/Inter. Org. ( Mabungwe wosakhala a boma) 4. Agro - dealer ( Wogulitsa zipangizo za ulimi) 5. Family, relatives 99. Other (specify) ( Zina [tchulani]) CODE S 0. None ( Palibe) 1. 0 - 5 2. 5 - 10 3. 10 - 15 4. 15 - 20 5. +20 CODE N 0. None 1. Kupalira (scraping) 2. Kuzulira atapanga maluwa (uprooting after flowering) 3. Kuzulira asanapange maluwa (before flowering) 4. Kusenda/Kuojekera (cover weeds with soil) 5. Kubandira (banking) 6. Herbicide ( Mankhwala wopha tchire) 7. Deep tillage ( kulima mozama/mwakuya) CODE N 8. Point manure application ( Kuika manyowa pa phando lodzalira) 9. Burning 10. Point fertilizer application 11. Point manure/fertilizer mix application 12. Point maize bran application 13. Point ash application 14. Remove and bury in a deep pit 99. Other (specify) ( Zina [tchulani]) CODE Q 0. The same season ( Chaka chomwecho) 1. The following season ( Chaka chotsatira) 2. The following 2 seasons ( Patatha zaka ziwiri) 3. The following 3 - 5 seasons ( patatha zaka zitatu kufikira zisanu) CODE N 4. The following 6 - 10 seasons ( ku fikira khumi) 5. The following 10+ seasons ( Patatha zaka zoposera khumi) 99. Other (specify) ( Zina [tchulani]) CODE O 0. Nothing ( Palibe) 1. Burn ( Kuyatsa) 2. Incorporate into ridge ( Kumukwilira mu mzere) 3. Put in furrow ( Kumuika mu khwawa) CODE P 4. Remove from field ( Kumutaya kunja kwa munda) 5. Consumed/Fed to livestock ( Kudya/kudyetsera ku ziweto) 6. Bury in a deep pit 99. Other (specify) ( Zina [tchulani]) CODE P 0. None/no change (positive) Ubwino wake 1. Striga came back less in short term (same season) ( Kaufiti 2. Striga came back less in the long term (next/multiple season/s) ( Kaufiti anameranso koma wocheperapo patapita nthawi [chaka/zaka] zotsatira) 3. Striga did not come back at all that s eason (same season) ( 4. Soil fertility increased ( chonde chinaonjezereka mu nthaka) 5. Increased biomass for fuel or fodder ( mapesi wochuluka omwe anagwira ntchito ngati nkhuni kapena chakudya cha ziweto) 6. Maize yi eld increased in the short term (same season) ( 7. Maize yield increased in the long term (next season) ( 8. Aggregate food production increased Chakudya (chonse tikach iphatikiza chinachuluka) 9. Pest incidence reduced ( tizilombo toononga mbeu tinachepa) 10. Profit (from on - farm production) increased (Phindu [purofiti] lochokera ku zokolola zathu linachuluka) 11. On - farm labor decreased ( Ntchito yogwira pamundapo inachep a) (negative) kuipa kwake 12. Striga came back more in short term (same season) ( Kaufiti CODE BB 13. Striga came back more in the long term (next season) ( Kaufiti chotsatira) 14. Maize yield decreased in the short term; harvested little (same season) ( 15. Maize yield decreased in the long term; harvested little (next season) ( zokolola 16. Aggregate food production decreased ( Chakudya chonse tikachiphatikiza chinachepa) 17. Pest incidence increased; harbored pests ( Tizilombo toononga mbeu tinachuluka) 18 Profit (from on - farm production) decreased ( Phindu [purofiti] lochokera ku zokolola zath u linachepa) 19. On - farm labor increased ( Ntchito yogwira pamundapo inachuluka) 20. Inputs became unavailable ( zipangizo zogwilira ntchito ya ulimi zinasowa) 21. Inputs became too expensive ( zipangizo zogwilira ntchito ya ulimi zinakwera mtengo) 22. Delaye d Striga emergence 23. Reduced overall weed pressure 24. Improved water retention/soil moisture holding capacity 25. Reduced erosion/Improved soil structure/texture 26. Sustained Striga emergence; prevented Striga problem from getting worse 27. Damaged soi l 99. Other (specify) ( Zina [tchulani]) 95 CODE T 1. Kupalira (scraping) 2. Kuzulira atapanga maluwa (uprooting after flowering) 3. Kuzulira asanapange maluwa (before flowering) 4. Kusenda/Kuojekera (cover weeds with soil) 5. Kubandira (banking) tchire) 7. Deep tillage ( kulima mozama/mwakuya) 6. Herbicide ( Mankhwala wopha 8. Point manure appli cation ( Kuika manyowa pa phando lodzalira) 9. Burning 10. Point fertilizer application 11. Point manure/fertilizer mix application 12. Point maize bran application 13. Point ash application 14. Remove and bury in a deep pit 99. Other (specify) ( Zina [tchulani]) [tchulani]) CODE U 1. No time Analibe mpata (nthawi) 2. Shortage of household labor ( Kuchepa kwa wogwira ntchito panyumba) 3. Could not hire outside labor ( Sakanakwanitsa kulemba a ganyu) 4. Illness/death in family ( Matenda/Maliro wokhudza banja) 5. Rain ( Mvula) 6. Market price for outputs (e.g., legume grain) was too low ( Mitengo ya zokolola (mwachitsanzo, mbeu za gulu la nyemba) inali yotsika kwambiri) 7. Ganyu labor took away from practice ( Aganyu anatsata njira ina osakhala ime ne anauzidwa) 8. Could not afford inputs ( Sindikadakwanitsa kupeza zipangizo zotsatilira njirayi) 9. Input availability ( Kapezekedwe ka zipangizo) 10. Did not know; Not enough information/training ( samadziwa za mmene ndingatsatire njirayi; sanaphunzitsidwe /sanapatsidwe upangiri) 99. Other (specify) ( Zina [tchulani]) CODE V 1. Crop rotation ( Kulima mwa kasinthasintha) 2. Manure application ( Kuthira Manyowa) 3. Early yielding variety ( Kubzala mbeu zocha msanga) 4. Fertilizer application ( Kuthira fetereza) 5. Legume crop residue mulch ( Kuphimbira ndi masangwi a mbeu zathu za mgulu la nyemba) 6. Maize crop residue mulch (Kuphimbira ndi mapesi a chimanga) 7. Intercropping ( Kulima mwa kasakaniza) 8. Minimum tillage ( mtayakhasu) 9. Deep tillage ( Kulima mozama/mokuya) 10. Pre - emergence herbicide ( kuthira mankhwala okupha tchire [wothira mbeu zisanamere] 11. Maize residue incorporation 12. Legume residue incorporation 13. Maize + legume residue mix incorporation 14. Tobacco pellet application 15. M aize bran application 16. Ash application 17. Planting leguminous trees (e.g., tephrosia) 99. Other (specify) ( Zina [tchulani]) CODE W 1. 0 - 2 2. 3 - 5 3. 6 - 10 4. 10+ CODE X 1. NGO/Inter. Org. ( Mabungwe wosakhala a boma) 2. Radio ( Wailesi) 3. Poster/Hand Out ( Postala/zojambulidwa pa pepala) 4. Demonstration trial ( Munda wachionetsero) 5. Extension agent ( Alangizi) 6. Market ( Ku msika) 7. Experimentation ( Kuyeselera/kafukufuku ) 8. Neighbor/Farmer ( Wokhala moyandikana naye/mlimi) 9. Farmer group/coop ( ku gulu/bungwe la alimi) 10 or 14. Family, relatives 11. Tradition Za makolo 12. Agro - dealer ( Wogulitsa zipangizo za ulimi) 13. Intuition ( Kungopanga poganiza kuti ndizotheka) 15. School 99. Other (specify) ( Zina [tchulani]) CODE Y 1. Striga came back less in short term (same season) ( Kaufiti 2. Striga came back less in the long term (next/multiple season/s) ( Kaufiti anameranso koma wocheperapo patapita nthawi (chaka/zaka zotsatira) 3. S triga did not come back at all that season (same season) ( 4. Soil fertility increased ( chonde chinaonjezereka mu nthaka) 5. Increased biomass for fuel or fodder ( mapesi wochuluka omwe anagwira ntchito ngati nkhuni kapena chakudya cha ziweto) 6. Maize yield increased in the short term (same season) ( 7. Maize yield increased in the long term (next season) ( CODE Y 8. Aggr egate food production increased Chakudya (chonse tikachiphatikiza) chinachuluka 9. Pest incidence reduced ( tizilombo toononga mbeu tinachepa) 10. Profit (from on - farm production) increased (Phindu [purofiti] lochokera ku zokolola zathu linachuluka) 11. On - farm labor decreased ( Ntchito yogwira pamundapo inachepa) 12. Delayed Striga emergence 13. Decreased overall weed pressure 14. Improved water retention/soil moisture holding capacity 15. Reduced erosion; improved soil structure/texture 16. Sustained Striga emergence; prevented Striga problem from getting worse 99. Other (specify) ( Zina [tchulani]) 96 CODE Z 0. None 1. Crop rotation ( Kulima mwa kasinthasintha) 2. Manure application ( Kuthira Manyowa) 3. Early yielding variety ( Kubzala mbeu zocha msanga) 4. Fertilizer application ( Kuthira fetereza) 5. Legume crop residue mulch ( Kuphimbira ndi masangwi a mbeu zathu za mgulu la nyem 6. Maize crop residue mulch (Kuphimbira ndi mapesi a chimanga) 7. Intercropping ( Kulima mwa kasakaniza) ba) 8. Minimum tillage ( mtayakhasu) 9. Deep tillage ( Kulima mozama/mokuya) 10. Pre - emergence herbicide ( kuthira mankhwala okupha tchire [wothira mbeu zisanamere]) 11. Maize residue incorporation 12. Legume residue incorporation 13. Maize + legume residue mix incorporation 14. Tobacco pellet application 15. Maize bran application 16. Ash application 17. Planting leguminous trees (e.g., tephrosia) 99. Other (specify) ( Zina [tchulani]) CODE AA 0. N one/no change (positive) Ubwino wake 1. Striga came back less in short term (same season) ( Kaufiti 2. Striga came back less in the long term (next/multiple season/s) ( Kaufiti anameranso koma wocheperapo patapita nthawi [chaka/zaka] zotsatira) 3. Striga did not come back at all that season (same season) ( 4. Soil fertility increased ( chonde chinaonjezereka mu nthaka) 5. Increased biomass for fuel or fodder ( mapesi wochuluka omwe anagwira ntchito ngati nkhuni kapena chakudya cha ziweto) 6. Maize yield increased in the short term (same season) ( 7. Maize yield increased in the long term (next season) ( 8. Aggregate food production increased Chakudya (chonse tikachiphatikiza chinachuluka) 9. Pest incidence reduced ( tizilombo toononga mbeu tinachepa) 10. Profit (from on - farm production) increased (Ph indu [purofiti] lochokera ku zokolola zathu linachuluka) 11. On - farm labor decreased ( Ntchito yogwira pamundapo inachepa) CODE BB (negative) kuipa kwake 12. Striga came back more in short term (same season) ( Kaufiti wochulukirapo) 13. Striga came back more in the long term (next season) ( Kaufiti 14. Maize yield decreased in the short term; harvested little (same season) ( 15. Maize yield de creased in the long term; harvested little (next season) ( zokolola 16. Aggregate food production decreased ( Chakudya chonse tikachiphatikiza chinachepa) 17. Pest incidence increased; harbored pests ( Tizilombo toononga mbeu tinachuluka) 18 Profit (from on - farm production) decreased ( Phindu [purofiti] lochokera ku zokolola zathu linachepa) 19. On - farm labor increased ( Ntchito yogwira pamundapo inachuluka) 20. Inputs became unavailable ( zipangizo zogwilira ntchito ya ulim i zinasowa) 21. Inputs became too expensive ( zipangizo zogwilira ntchito ya ulimi zinakwera mtengo) 22. Delayed Striga emergence 23. Reduced overall weed pressure 24. Improved water retention/soil moisture holding capacity 25. Reduced erosion/Improved soil structure/texture 26. Sustained Striga emergence; prevented Striga problem from getting worse 27. Damaged soil 99. Other (specify) ( Zina [tchulani]) CODE BB 0. The same season 1. The following season ( Chaka chotsatira) 2. The following 2 seasons ( Patadutsa zaka ziwiri) 3. The following 3 - 5 seasons ( Patatha zaka za pakati pa zitatu mpaka zisanu) 4. The following 6 - 10 seasons ( Patatha zaka za pakati pa chisanu 5. The following 10+ seasons ( Patatha zaka zoposera khumi) CODE CC 0. None/nobody 1. Neighbor/Farmer ( Wokhala moyandikana naye/mlimi) 2. Farmer group/coop ( ku gulu/bungwe la alimi) specify) ( Zina [tchulani]) 3. NGO/Inter. Org. ( Mabungwe wosakhala a boma) 4. Agro - dealer ( Wogulitsa zipangizo za ulimi) 5. Family, relatives 99. Other (specify) ( Zina [tchulani]) CODE DD 0. None ( Palibe) 1. 0 - 5 2. 5 - 10 3. 10 - 15 4. 15 - 20 5. +20 97 CODE EE 1. Common bean ( Nyemba) 2. Soybean ( Soya) 3. Cowpea Khobwe 4. Groundnut (peanut) ( Mtedza) 5. Pigeon Pea ( Nandolo) 6. Cereal (sorghum, millet) ( mawere, mapira) 7. Cash Crop (Tobacco, Cotton) ( Fodya, Thonje) 8. Root tuber/starch (Cassava, Pumpkin, Sweet Potato, Irish Potato) ( Chinangwa, mawungu, mbatata, mbatata ya kac hewere) 99. Other (specify) ( Zina [tchulani]) CODE FF 1. Saved seed ( Mbeu yosungidwa) 2. Purchased ( Mbeu yochita kugula) 3. Subsidy (Mbeu yotsika mtengo) 4. Private trader ( Mbeu yogula kwa a ma bizinesi) 5. NGO (free) ( Kuchoka ku mabungwe osakhala a boma [ yaulere]) 99. Other (specify) ( Zina [tchulani]) CODE GG 1. Cow ( 2. Goat ( Mbuzi) 3. Chicken or other poultry ( Nkhuku kapena zina za gulu la nkhuku) 4. Pig ( Nkhumba) 5. Compost mixed with manure ( Manyowa) 6. NPK ( Fetereza wa chitowe) 7. Urea ( Fetereza wa Urea) 8. CAN ( Fetereza wa CAN) 9. Maize residues 10. Legume residues 11. Maize + legume residue mix 12. Maize bran 99. Other (specify) ( Zina [tchulani]) CODE HH 1. From own production/livestock ( Kuchokera ku ziweto zathu) 2. Purchased ( Kugula) 3. Paid ganyu to collect/apply ( Ndinalemba waganyu kuti akatenge/athire) 4. Given by other farmer ( Ndinapatsidwa ndi mzanga) 5. Gathered in village ( Ndinasonkhanitsa a mmudzi) 6. Subsidy ( Ndinapeza wotsika mtengo) 7. Received from NGO ( Ndinalandir a kuchokera ku mabungwe omwe si aboma) 8. Agro - dealer ( Wogulitsa zipangizo za ulimi) 9. Family, relatives 99. Other (specify) ( Zina [tchulani]) CODE II 1. When turning ridges early right after harvest ( Popanga mizere moyambilira tikangomaliza kukolora) 2. When turning ridges just before first planting rains ( Popanga mizere mvula ikayandikira) 3. When turning ridges after first planting rains ( Popanga mizere mvula yoyamba ikangogwa) 4. After ridge turning but before planting (incorporate) ( Mizer e itapangidwa koma tisanadzale) 5. After ridge turning but before planting (planting station application) ( Kuthira pa phando [tisanadzale]) 6. At planting station with seed Kuthira pa phando (nthawi yodzala) 7. About 10 days after planting with first weeding ( Patadutsa masiku kupalira koyamba) 8. About 30 days after planting with second weeding ( Patadutsa masiku ya kupalira kachiwiri) 9. Dig basin/zaii pit 99. Other (specify) ( Zina [tchu lani]) 98 CODE JJ 1. Crop rotation ( Kulima mwa kasinthasintha) 2. Manure application ( Kuthira Manyowa) 3. Early yielding variety ( Kubzala mbeu zocha msanga) 4. Fertilizer application ( Kuthira fetereza) 5. Legume crop residue mulch ( Kuphimbira ndi masang wi a mbeu zathu za mgulu la nyemba) 6. Maize crop residue mulch (Kuphimbira ndi mapesi a chimanga) 7. Intercropping ( Kulima mwa kasakaniza) 8. Minimum tillage ( mtayakhasu) 9. Deep tillage ( Kulima mozama/mokuya) 10. Pre - emergence herbicide ( kuthira mankhwala okupha tchire [wothira mbeu zisanamere]) 11. Maize residue Incorporation 12. Legume residue Incorporation 13. Maize + legume residue mix 14. Maize bran Application 15. Leguminous trees (e.g., tephrosia) 99. Other (specify) ( Zina [tchulani]) CODE KK 1. Cow ( 2. Goat ( Mbuzi) 3. Chicken or other poultry ( Nkhuku kapena zina za gulu la nkhuku) 4. Pig ( Nkhumba) 5. Compost mixed with manure ( Manyowa) 6. NPK ( Fetereza wa chitowe) 7. Urea ( Fetereza wa Urea) 8. CAN ( Fetereza wa CAN) 9. Maize Bran 99. Other (specify) ( Zina [tchulani]) CODE LL 1. Common bean ( Nyemba) 2. Soybean ( Soya) 3. Cowpea ( Khobwe) 4. Groundnut (peanut) ( Mtedza) 5. Pigeon Pea ( Nandolo) 6. Cereal (sorghum, millet) ( mawere, mapira) 7. Cash Crop (Tobacco, Cotton) ( Fodya, Thonje) 8. Root tuber/starch (Cassava, Pumpkin, Sweet Potato, Irish Potato) ( Chinangwa, mawungu, mbatata, mbatata ya kachewere) 99. Other (specify) ( Zina [tchulani]) CODE MM 1. No time Analibe mpata (nthawi) 2. Shortage of household labor ( Kuchepa kwa wogw ira ntchito panyumba) 3. Could not hire outside labor ( Sakanakwanitsa kulemba a ganyu) 4. Illness/death in family ( Matenda/Maliro wokhudza banja) 5. Rain ( Mvula) 6. Market price for outputs (e.g., legume grain) was too low ( Mitengo ya zokolola (mwachitsanzo, mbeu za gulu la nyemba) inali yotsika kwambiri) 7. Ganyu labor took away from practice ( Aganyu anatsata njira ina osakhala imene anauzidwa) 8. Could not afford inputs ( Sindikadakwanitsa kupeza zipangizo zotsatilira njirayi) 9. Input availabi lity ( Kapezekedwe ka zipangizo) 10. Did not know; Not enough information/training ( samadziwa za mmene ndingatsatire njirayi; sanaphunzitsidwe/sanapatsidwe upangiri) 99. Other (specify) ( Zina [tchulani]) CODE NN 1. Poor 2. Fair 3. Good 4. Very Good 5. Excellent 99 REFERENCES 100 REFERENCES Abdul, K., Chemining'wa, G. N., & Onwonga, R. N. (2012). Relationships between agronomic practices, soil chemical characteristics and Striga reproduction in Dryland areas of Tanzania. Journal of Agricultural Science and Technology. A , 2 (10A), 1134. Adato, M., & Meinzen - Dick, R. (2002). Assessing the Impact of Agricultural Research on Poverty Using the Sustainable Livelihoods Framework. Washing ton DC, USA: International Food Policy Research Institute Adesina, A. A., Mbila, D., Nkamleu, G. B., & Endamana, D. (2000). Econometric analysis of the determinants of adoption of alley farming by farmers in the forest zone of southwest Cameroon. Agriculture, Ecosystems & Environment, 80 (3), 255 265. doi:10.1016/ S0167 - 8809(00)00152 - 3 Ajayi, O. C., Franzel, S., Kuntashula, E., & Kwesiga, F. (2003). Adoption of improved fallow technology for soil fertility management in Zambia: Empirical studies and emerging issues. Agroforestry Systems , 59 , 317 - 326. doi:10.1023/B:AGFO.0000005232.87048.03 Akobundu, I. O. (1980). Weed science research at the International Institute of Tropical Agriculture and research needs in Africa. Weed Science , 28 (4), 439 - 445. Amsalu, A., & De Graaff, J. (2007). Determinants of ad option and continued use of stone terraces for soil and water conservation in an Ethiopian highland watershed. Ecological Economics , 61 (2), 294 - 302. Anderson, J. R., & Feder, G. (2004). Agricultural extension: Good intentions and hard realities. The World Bank Research Observer , 19 (1), 41 - 60. Andersson, J. A., & D'Souza, S. (2014). From adoption claims to understanding farmers and contexts: A literature review of Conservation Agriculture (CA) adoption among smallholder farmers in southern Africa. Agricult ure, Ecosystems & Environment , 187 , 116 - 132. for conservation agriculture and the politics of its promotion in African smallholder farming. In Sumberg, J., Thompson, J. (Eds.), Contested Agronomy: Agricultu ral Research in a Changing World . London: Earthscan. Á rvai, J.L., & Gregory, R. (2003). Testing alternative decision approaches for identifying cleanup priorities at contaminated sites. Environmental Science Technology , 37 , 1469 1476 101 Árvai, J., Kellon, D., León, R., Gregory, R., & Richardson, R. (2014). Structuring international development decisions: confronting trade - offs between land use and community development in Costa Rica. Environment Systems and Decisions , 34 (2), 224 - 236. Atera, E. A., Itoh, K. adoption of weed control options: the case of Striga asiatica in Malawi. Journal of Agricultural Science , 4 (5), 41. Barbour, R. S. (2001). Checklists for improving rigour in quali tative research: a case of the tail wagging the dog? British Medical Journal , 322 (7294), 1115. Barrett, C. B., Reardon, T., & Webb, P. (2001). Nonfarm income diversification and household livelihood strategies in rural Africa: Concepts, dynamics, and policy implications . Food Policy , 26, 315 331. doi:10.1016/S0306 - 9192(01)00014 - 8 Bartel, A. P., & Lichtenberg, F. R. (1987). The comparative advantage of educated workers in implementing new technology. The Review of Economics and Statistics , 1 - 11. Baskin, C. C., & Baskin, J. M. (1998). Seeds: ecology, biogeography, and evolution of dormancy and germination . Elsevier. Baum, E., Gyiele, L. A., Drechsel, P., & Nurah, G. K. (1999). Tools for the economic analysis and evaluation of on - farm trials (pp. 3 1 32, IBSRAM global tool kit series No. 1), Bangkok: International Board for Soil Research and Management . Bennett, J., & Birol, E. (2010). Choice Experiments in Developing Countries . Cheltenham, UK: Edward Elgar Publishing. Bezner Kerr, R., Snapp, S., Chirwa, M., Shumba, L., & Msachi, R. (2007). Participatory research on legume diversification with Malawian smallholder farmers for improved human nutrition and soil fertility. Experimental Agriculture , 43(4), 437 453. doi:10.10 17/S0014479707005339 Birol, E., Smale, M., & Yorobe Jr, J. M. (2012). Bi - modal preferences for bt maize in the Philippines: A latent class model. AgBioForum, 15 (2), 175 - 190. Birol, E., Villalba, E. R., & Smale, M. (2009). Farmer preferences for milpa d iversity and genetically modified maize in Mexico: a latent class approach. Environment and Development Economics , 14 (04), 521 - 540. Blackie, M.J. (1994). Maize productivity for the 21 st Century: The African Challenge. Outlook in Agriculture, 23: 189 - 195. 102 Bonabana - Wabbi, J., & Taylor, D. B. (2012). A limited dependent variable analysis of integrated pest management adoption in Uganda. Journal of Agricultural Science and Technology, 2 (10A), 1162. Boxall, P.C. & Adamowic z, W.L. (2002). Understanding heterogeneous preferences in random utility models: A latent class approach. Environmental and Resource Economics, 23 , 421 - 446. Cechin, C. , & Press , M. C. ( 1993) . Nitrogen relations of the sorghum Striga hermonthica host parasite association: germination, attachment and early growth . New Phytology, 124 , 681 687 . Collinson, S. T., Clawson, E. J., Azam - Ali, S. N., & Black, C. R. (1997). Effects of soil moisture deficits on the water relations of bambara gro undnut (Vigna subterranea L. Verdc.). Journal of Experimental Botany , 48 (4), 877 - 884. Creswell, J. W., & Miller, D. L. (2000). Determining validity in qualitative inquiry. Theory into Practice , 39 (3), 124 - 130. Debrah, S.K. (1994). Socio - economic constraints to the adoption of weed control technologies: The case of Striga control in West African Semi Arid Tropics. International Journal of Pest Management, 40, 153 - 158. Dimes, J., Muza, L., Malunga, G., & Snapp, S. (2004 ). Trade - offs between investments in nitrogen and weeding: On - farm experimentation and simulation analysis in Malawi and Zimbawbwe. In Integrated Approaches to Higher Maize Productivity in the New Millennium, (p. 5 - 11). Proceedings of the Se venth Eastern and Southern Africa Regional Maize Conference. Nairobi, Kenya: CIMMYT. Dinh, T. P., Cameron, D., & Nguyen, X. T. (2015). Rural livelihood adoption framework: A conceptual and analytical framework for studying adoption of new activities by sm all farmers. International Journal of Agricultural Management , 4 (4), 163 - 172. Doggett, H. J. (1984) Striga - its biology and control - an overview. In: Striga, Biology and Control (Ed. by E. S. Ayensu, H. Doggett, R. D. Keynes, J. Marton - Lefevre, L. J. M usselman, C, Parker and A. Pickering) pp. 27 - 36, ICSU Press, Paris and The International Development Research Centre , Ottawa, Canada Drechsel, P., Olalaye, A., Adeoti, A., Thiombiano, L., Barry, B., & Vohland, K. (2005). Adoption, Driver and Constraints o f Resource conservation technologies in Sub - Saharan Africa. Berlin : FAO, IWMI, Humbold University. Retrieved from http://westafrica.iwmi.org/Data/Sites/17/Documents/PDFs/AdoptionConstraints - Overview.pdf 103 Dugje, I. Y., Kamara, A. Y., & Omoigui, L. O. (2008). Influence of Farmers Crop Management Practices on Striga hermonthica Infestation and Grain Yield of Maize (Zea mays L.) in the Savanna Zones of Northeast Nigeria. Journal of Agronomy , 7 (1), 597 - 600. Esilaba, A. O., Reda, F., Ransom, J. K., Bayu, W., Woldewahid, G. , & Zemichael, B. (2000). Integrated nutrient management strategies for soil fertility improvement and Striga control on Northern Ethiopia. http://repository.cimmyt.org/handle/10883 /1421 FAO Land and Water Development Division. (2001). Conservation agriculture and the role of policy, (p. 31 - 41). In The economics of conservation agriculture . Rome: Food and Agriculture Organization of the United Nations. Retrieved from ftp://ftp.fao.org/agl/agll/docs/ecconsagr.pdf Feder, G., Just, R. E., & Zilberman, D. (1985). Adoption of agricul tural innovations in developing countries: A survey. Economic Development and Cultural Change , 33 (2), 255 - 298. Feder, G., & Umali, D. L. (1993). The adoption of agricultural inno vations: A review. Technological Forecasting and Social Change, 43 (3), 215 23 9. doi:10.1016/0040 - 1625(93)90053 - A Fenske, J. (2011). Land tenure and investment incentives: Evidence from West Africa. Journal of Development Economics , 95 (2), 137 - 156. Ferguson, A. E., & Mkandawire, R. M. (1993). Common Beans and Farmer Managed Diversity: Regional Variations in Malawi. Culture & Agriculture, 13 : 14 - 17. doi: 10.1525/cuag.1993.13.45 - 46.14 Fern, E. F. (1982). The use of focus groups for idea generation: th e effects of group size, acquaintanceship, and moderator on response quantity and quality. Journal of Marketing Research , 1 - 13. Franke, A. C., Schulz, S., Oyewole, B. D., & Bako, S. (2004). Incorporating short - season legumes and green manure crops into ma ize - based systems in the moist Guinea savanna of West Africa. Experimental Agriculture , 40, 463 479. doi:10.1017/S001447970400211X Frost, D.L., Gumey, A.L., Press, M.C. & Scholes, J.D. (1997). Striga hermonthica reduces photosynthesis in sorghum: the importance of stomatal limitations and a potentia l role for ABA? Plant Cell and Environment, 20 , 483 - 492. Garrity, D. P., Akinnifesi, F. K., Ajayi, O. C., Weldesemayat, S. G., Mowo, J. G., Kalinganire, A., & Bayala, J. (2010). Evergreen Agriculture: a robust approach to sustainable food security in Africa. Food Security , 2 , 197 214. doi: 10.1007/s12571 - 010 - 0070 - 7 104 Gigou, J. J. (1992). L'azote dans les systèmes de culture du nord et du centre de la Côte d'Ivoire. Biological Nitrogen Fixation and Sustainability of Tropical Agriculture , 377 - 394. Giller, K . E., Corbeels, M., Nyamangara, J., Triomphe, B., Affholder, F., Scopel, E., & Tittonell, P. (2011). A research agenda to explore the role of conservation agriculture in African smallholder farming systems. Field Crops Research, 124 (3), 468 - 472. Giller, K . E., Witter, E., Corbeels, M., & Tittonell, P. (2009). Conservation agriculture and Field Crops Research , 114 (1), 23 - 34. Graves, A., Matthews, R., & Waldie, K. (2004). Low external input technologies for livelihood improvement in subsistence agriculture. Advances in Agronomy , 82 , 473 - 557. Hakansson, S. (1982). Multiplication, growth and persistence of perennial weeds. Pages 123 - 135 in W. Holzner and M. Numata, editors. Biology and ecology of weeds . Dr. W. Junk. The Hague. The Netherlands Heisey, P. W., & Smale, M. (1995). Maize technology in Malawi: A green revolution in the making? (Research Report No. 4). Mexico: CIMMYT. Retrieved from http://repository.cimmyt.org/xmlui/bitstream/handle/10883/737/59230.pdf?sequence =1 Hensher, D. A. (2010). Hypothetical bias, choice experiments and willingness to pay. Transportation Research Part B: Methodological , 4 4 (6), 735 - 752. Hockett, M., & Richardson, R. B. (2016). Examining the drivers of agricultural experimentation among smallholder farmers in Malawi. Experimental Agriculture , 1 21. doi:10.1017/ S0014479716000673 Hodgson, G.M. (1989). Economics and institutions. Cheltenham: Edward Elgar. Hole, A. R., & Kolstad, J. R. (2012). Mixed logit estimation of willingness to pay distributions: a comparison of models in preference and WTP space using data from a health - relat ed choice experiment. Empirical Economics , 42 (2), 445 - 469. Hovorka, A. J. (2006). The No. 1 Ladies' Poultry Farm: A feminist political ecology of urban agriculture in Botswana. Gender, Place & Culture , 13 (3), 207 - 225. Hudu, A. I., & Gworgwor, N. A. (1998). Preliminary results on evaluation of trap crops for Striga hermonthica (Del.) Benth. control in sorghum. International Millet Newsletter , 39. 118 121. Hardy, M.A. (1993). Regression with dummy variables . Newbury Park, CA: Sage. 105 Johnson, F. R., Lancsar, E., Marshall, D., Kilambi, V., Mühlbacher, A., Regier, D. A., ... & Bridges, J. F. (2013). Constructing experimental designs for discrete - choice experiments: report of the ISPOR conjoint analysis experimental design good research p ractices task force. Value in Health , 16 (1), 3 - 13. Jury, M. R., & Mwafulirwa, N. D. (2002). Climate variability in Malawi, Part 1: Dry summers, statistical associations and predictability. International Journal of Climatology, 22 , 1289 1302. doi:10.1002/ joc.771 Kalaba, K. F., Chirwa, P., Syampungani, S., & Ajayi, C. O. (2010). Contribution of agroforestry to biodiversity and livelihoods improvement in rural communities of southern African regions. Tropical Rainforests and Agroforests under Global Change , 461 476. doi:10.1007/978 - 3 - 642 - 00493 - 3_22 Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., & Mekuria, M. (2013). Adoption of interrelated sustainable agricultural practices in smallholder systems: Evidence from rural Tanzania. Technological Forecasti ng and Social Change, 80 (3), 525 540. doi:10.1016/j.techfore.2012.08.007 Khan, Z. R., Midega, C. A., Bruce, T. J., Hooper, A. M., & Pickett, J. A. (2010). Exploiting in Africa. Journal o f Experimental Botany, 61 (15), 4185 - 4196. Knowler, D., Philcox, N., Nathan, S., Delamare, W., Haider, W., & Gupta, K. (2009). Assessing prospects for shrimp culture in the Indian Sundarbans: a combined simulation modelling and choice e xperiment approach. Marine Policy , 33 (4), 613 - 623. Kugler, K. C., Trail, J. B., Dziak, J. J., & Collins, L. M. (2012). Effect coding versus dummy coding in analysis of data from factorial experiments. University Park, PA: The Methodology Center, Pennsylvania State University . Technical Report Series #12 - 120. Kumwenda, J. D. T., Waddington, S. R., , S. S., Jones, R. B., & Blackie, M. J. (1996). Soil fertility management research for the maize cropping systems of smallholders in southern Africa: A review. Natural Resources Group (p aper 96 - 02 ) . International Maize and Wheat Improvement Center, Mexico. Kumwenda, J. D., Waddington, S. R., Snapp, S. S., Jones, R. B., & Blackie, M. J. (1997). Soil fertility management in Southern Africa. , 157 - 172. Kunze, D. (2000). Economic assessment of water harvesting techniques: A demonstration of various methods. Quarterly Journal of International Agriculture , 39 (1), 69 - 91. Kureh, I., Kamara, A. Y., & Tarfa, B. D. (2006). Influence of cereal - legume rotation on Striga 106 Nigeria. Journal of Agriculture and Rural Development in the Tropics and Subtropics (JARTS) , 107 (1), 41 - 54. Lancaster, K.J. (1966). A new approach to consumer theory." The Journal of Political Economy, 132 - 157. Lancsar, E., & Savage, E. (2004). Deriving welfare measures from discrete choice experiments: inconsistency between current methods and random utility and welfare theory. Health Economics , 13 (9), 901 - 907. Marenya, P. P., & Barrett, C. B. (2007). Household - level determinants of adoption of improved natural resources management practices among smallholder farmers in western Kenya. Food Policy, 32 , 515 536. doi:10.1016/j.foodpol.2006. 10.002 Mbaga - Semgalawe, Z., & Folmer, H. (2000). Household adoption behaviour of improved soil conservation: the case of the North Pare and West Usambara Mountains of Tanzania. Land Use Policy , 17 (4), 321 - 336. Meij er, E., & Rouwendal, J. (2006) Measuring welfare effects in models with random coefficients. Journal of Applied Economics , 21 (2), 227 244. Montgomery, D.C. (2009). Design and analysis of experiments. Hoboken, NJ: John Wiley & sons, Inc. Morse, J. M. (20 05). Evolving trends in qualitative research: advances in mixed - method design. Qualitative Health Research , 15 (5), 583 - 585. Mugwe, J., Mugendi, D., Mucheru - Muna, M., Merckx, R., Chianu, J., & Vanlauwe, B. (2009). Determinants of the decision to adopt inte grated soil fertility management practices by smallholder farmers in the central highlands of Kenya. Experimental Agriculture, 45 (1), 61 75. doi:10.1017/S0014479708007072 Mungai, L. M., Snapp, S. S., Messina, J. P., Chikowo, R., Smith, A., Anders, E., .. . Li, G. (2016). Smallholder farms and the potential for sustainable intensification. Frontiers in Plant Science , 7, 1 17. doi:10.3389/fpls.2016.01729 Nahuelhual, L., Loureiro, M.L. & Loomis, J. (2004) Using random parameters to account for heterogeneous preferences in contingent valuation of public open space. Journal of Agricultural and Resource Economics , 537 552. Ndove, T. S., Whitbread, A. M., Clar k, R. A., & Pengelly, B. C. (2004). Identifying the factors that contribute to the successful adoption of improved farming practices in the smallholder sector of Limpopo Province, South Africa. In Whitbreak, A.M. & Pengelly, B.C., (Eds.), Tropical Legumes for Sustainable Farming Systems in Southern Africa and Australia . 107 Paper presented at Australia Centre for International Agricultural Research Proceedings, (pp. 146 - 154). Canberra, Australia: Australia Centre for International Agricultural Research . Ngwira , A. R., Thierfelder, C., & Lambert, D. M. (2013). Conservation agriculture systems for Malawian smallholder farmers: long - term effects on crop productivity, profitability and soil quality. Renewable Agriculture and Food Systems , 28 (4), 350 - 363. Njuki, J. , Kaaria, S., Chamunorwa, A., & Chiuri, W. (2011). Linking Smallholder Farmers to Markets, Gender and Intra - Household Dynamics: Does the Choice of Commodity Matter?. European Journal of Development Research , 23 (3), 426 - 443. Nkonya, E. M., Schroeder, T., & Norman, D. (1997) Factors affecting adoption of improved maize seed and fertiliser in northern Tanzania. Journal of Agricultural Economics, 48 (1 - 3), 1 - 12. Nowak, P. J. (1987). The adoption of agricultural conservation technologies: economic and diffusion explanations. Rural Sociology , 52 , 208 - 220. Retrieved from http://search.proquest.com/docview/1290948481?pq - origsite=gscholar Ntege - Nanyeenya, W., Mugisa - Mutetikka, M., Mwangi, W. M., & Verkuijl, H. (1997). An Assessment of Factors Affecting Adoption of Maize Production Technologies in Iganga District, Uganda . Addis Ababa, Ethiopia: National Agricultural Research Organization (NARO) and International Maize and Wheat Improvement Center (CIMMYT). Oluoch - Kosura, W. A., Marenya, P. P., & Nzuma, M. J. (2001, February). Soil fertility management in maize - based production systems in Kenya: Current options and future strategies. In Friesen, D.K., & Palmer A. F.E. (Eds.), Integrated Approaches to Higher Maize Productivity in the New Millennium. Paper presented at 7 th Eastern and Southern Africa Regional Maize Conference, (pp. 350 - 355), Nairobi, Kenya: CIMMYT and KARI. Orr, A., & Ritchie, J. M. (2004). Learning from failure: smallholder farming systems and IPM in Malawi. Agricultural Systems , 79 (1), 31 - 54. Orr, A., M wale, B., & Saiti, D. (2002). Modelling agricultural 'performance': smallholder weed management in Southern Malawi. International Journal of Pest Management , 48 (4), 265 - 278. Orr, A., Mwale, B., & Saiti - Chitsonga, D. (2009). Exploring seasonal poverty trap - week The Journal of Development Studies , 45 (2), 227 - 255. Ortega, D. L., Waldman, K. B., Richardson, R. B., Clay, D. C., & Snapp, S. (2016). Sustainable Intensification and Farmer Preferences for Crop System Attribut es: Evidence from Malawi's Central and Southern Regions. World Development , 87 , 139 - 151. 108 Ortega, D. L., Wang, H. H., Widmar, N. J. O., & Wu, L. (2013). Chinese Producer Behavior: Aquaculture Farmers in Southern China. China Economic Review, 30 , 540 - 547. Oswald, A. (2005). Striga control technologies and their dissemination. Crop Protection , 2 4 (4), 333 - 342. Oswald, A., & Ransom, J. K. (2001). Striga control and improved farm productivity using crop rotation. Crop Protection , 20 (2), 113 - 120. https://doi.org/10.1016/S0261 - 2194(00)00063 - 6 Oswald, A., Ransom, J. K., Kroschel, J., & Sauerborn, J. (2002). Intercropping controls Striga in maize based farming systems. Crop Protection , 21 (5), 367 - 374. Otsuka, K and Place, F. (2001). Land Tenure and Natural Resource Management: A comparative Study of Agrarian Communities in Asia and Africa. Baltimore, MD: Johns Hopkins UP. Print. Parker, C. (2009). Observations on the current status of Orobanche and Striga problems worldwide. Pest Man agement Science , 65 (5), 453 - 459. Parker, C. (2012). Parasitic weeds: a world challenge. Weed Science , 60 (2), 269 - 276. conservation investments in semi - arid India. Agricultural Economics , 19 (1), 113 125. Pengelly B.C., Whitbread A., Maziawana P.R. & Mukombe N. (2003). Tropical forage r esearch for the future better use of research resources to deliver adoption and benefits to farmers. Tropical Grasslands , 37 , 207 216. Pircher, T., Almekinders, C. J., & Kamanga, B. C. (2013). Participatory trials and farmers' social realities: understanding the adoption of legume technologies in a Malawian farmer community. International Journal of Agricultural Sustainability , 11 (3), 252 - 263. Place, F., Barrett, C. B., Freeman, H. A., Ramisch, J. J., & Vanlauwe, B. (2003). Prospects for integrated soil fertility management using organic and inorganic inputs: evidence from smallholder African agricultural systems. Food Policy , 28 (4), 365 - 378. Ransom, J.K. (1990). Weed control in maize/legume intercrops. In: Eds S.R. Waddington, A.F.E. Palmer and O.T. Edje, Research methods for cereal/legume intercropping. Proceedings of a workshop on Research methods for cereal and legume intercropping in Eastern and Southern Africa . Jan 23 - 27, 1989, Lilongwe, Malawi. CIMMYT, Mexico DF, pp. 41 - 44. 109 Ran som, J. K. (2000). Long - term approaches for the control of Striga in cereals: field management options. Crop Protection , 19 (8 - 10), 759 - 763. Reda, F., Verkleij, J. A. C., & Ernst, W. H. O. (2005). Intercropping for the improvement of sorghum yield, soil fe rtility and Striga control in the subsistence agriculture region of Tigray (northern Ethiopia). Journal of Agronomy and Crop Science , 191 (1), 10 - 19. Rich, P. J., & Ejeta, G. (2008). Towards effective resistance to Striga in African maize. Plant Signaling & Behavior , 3 (9), 618 - 621. Richardson, R. B., Kellon, D., Leon, R. G., & Arvai, J. (2013). Using choice experiments to understand household tradeoffs regarding pineapple production and environmental management in Costa Rica. Journal of Environme ntal Management , 127 , 308 - 316. Rigby, D., & Burton, M. (2005) Preference heterogeneity and GM food in the UK. European Review of Agricultural Economics, 32 (2), 269 288. Rockers, P. C., Jaskiewicz, W., Wurts, L., Kruk, M. E., Mgomella, G. S., Ntalazi, F. , & Tulenko, K. (2012). Preferences for working in rural clinics among trainee health professionals in Uganda: a discrete choice experiment. BMC Health Services Research , 12 (1), 212. Rogers, J.L., Nicewander, W.A., & Toothaker, L. (1984). Linearly indepen dent, orthogonal, and uncorrelated variables. The American Statistician, 38 (2), 133 - 134. Schulz, S., Honlonkou, A. N., Carsky, R. J., Manyong, V. M., & Oyewole, B. D. (2003). Alternatives to mucuna for soil fertility management in southern Benin: Farmer perception and use of traditional and exotic grain legumes. Experimental Agriculture , 39 (3) , 267 - 278 . Shiferaw, B., & Bantilan, C. (2004). Agriculture, rural poverty and natural resource management in less - favored areas: Revisiting challenges and conceptual issues. Journal of Food, Agriculture and Environment, 2 (1), 328 339. Retrieved from http://oar.icrisat.org/1278/1/FoodAgricEnviron2_1_328 - 339_2004.pdf Silberg, T. R., Richardson, R. B., Hockett, M., & Snapp, S. S. (2017). Maize - legume intercropping in central Malawi: determinants of practice. International Journal of Agricultural Sustainability , 15 (6), 662 - 680. Sileshi, G., Schroth, G., Rao, M. R., & Gi rma, H. (2008). Weeds, diseases, insect pests and tri - trophic interactions in tropical agroforestry. In Ecological Basis of Agroforestry, (pp., 73 - 94 ) . Boca Raton: CRC Press . Singh, B. B., Chambliss, O. L., & Sharma, B. (1997). Recent advances in cowpea b reeding. In Singh, B.B, Mohan, R., Dashiell, K.E., & Jackai, L.E.N., Advances in Cowpea Research , (p. 3049 ). IITA: Ibadan, Nigeria. 110 Smith, A., Snapp, S., Dimes, J., Gwenambira, C., & Chikowo, R. (2016). Doubled - up legume rotations improve soil fertility and maintain productivity under variable conditions in maize - based cropping systems in Malawi. Agricultural Systems , 145 , 139 - 149. Africa Renewal. Retrieved fro m http://www.un.org/africarenewal/magazine/january - 2013/what - went - wrong - lessons - malawi%E2%80%99s - food - crisis Tamado, T., & Milberg, P. (2000). Weed flora in arable fields of eastern Ethiopia with emphasis on the occurrence of Parthenium hysterophorus. Weed Research , 40 (6), 507 - 521. Tamene, L., Mponela, P., Ndengu, G., & Kihara, J. (2015). Assessment of maize yield gap and maj or determinant factors between smallholder farmers in the Dedza district of Malawi. Nutrient Cycling in Agroecosystems, 105 , 1 18. doi:10.1007/s10705 - 015 - 9692 - 7 Thierfelder, C., & Wall, P. C. (2010). Rotation in conservation agriculture systems of Zambia : effects on soil quality and water relations. Experimental Agriculture , 46 (3), 309 - 325. Train, K. (2009). Discrete choice methods with simulation . Cambridge University Press, New York. United Nations. (2014). . 11 December 2014. http://www.mw.one.un.org/country - profile/ Useche, P., Barham, B. L., & Foltz, J. D. (2012). Trait - based adoption models using ex - ante and ex - post approaches. American Journal of Agr icultural Economics , 95 (2), 332 - 338. Vaiknoras, K., Norton, G., & Alwang, J. (2014). Farmer Preferences for Attributes of Conservation Agriculture in Eastern Uganda . African Journal of Agricultural and Resource Economics, 10 (2), 158 - 173. Vanlauwe, B., Bationo, A., Chianu, J., Giller, K. E., Merckx, R., Mokwunye, U., ... & Smaling, E. M. A. (2010). Integrated soil fertility management: operational definition and consequences for implementation and dissemination. Outlook On Agriculture , 39 (1 ), 17 - 24. of Mucuna fallowing and agrofor - estry technologies in the coastal savanna of Benin. Agricultural Systems, 56 , 269 287. doi:10.1016/S0308 - 521X(97)00 041 - 3 Waldman, K.B. & Ortega, D.L., Richardson, R.B., Clay, D.C., & Snapp, S. (2016). Preferences for legume attributes in maize - legume cropping systems in Malawi. Food Security , 8 (6) , 1087 1099. doi: 10.1007/s12571 - 016 - 0616 - 4 111 Waldman, K.B., Ortega, D. L., Richardson, R. B., & Snapp, S. S. (2017). Estimating demand for perennial pigeon pea in Malawi using choice experiments. Ecological Economics , 131 , 222 - 230. Waldman, K. B., & Richardson, R. B. (2018). Confronting Tradeoffs Between Agricultu ral Ecosystem Services and Adaptation to Climate Change in Mali. Ecological Economics , 150 , 184 - 193. 112 CHAPTER 3: CROP MODELING: AN INTEGRATED APPROACH TO SIMULATING EMERGENCE AND PERSISTENCE OF STRIGA ASIATICA 3.1 Introduction Smallholder farmers (culti vating <2ha) comprise the majority of the agrarian population in Malawi ( Ricker - Gilbert et al., 2014 ) . Among the smallholder population, maize ( Zea mays ) is the most commonly cultivated food crop and its production is often used as an indicator for rural food security and wealth ( Dorward & Chirwa, 2011 ) . In recent years, production has faced numerous challenges, one being the emergence and parasitism of w itchweed ( Striga spp .). As a hermiparasitic angiosperm, witchweed attaches to maize underground and extracts minerals, photosynthates and water ( Khan et al., 2010) . Attachment is visible several days after, creating a pathogenic effect, as underground juve niles disrupt the hormonal balance and reduce photosynthetic processes (Watling & Press, 2001). Once the weed emerges, more pronounced effects can be seen. The leaf tissue of Striga has greater osmotic pressure than maize, and being that its leaves have lo wer stomatal resistance, a higher transpiration rate renders Striga a stronger sink for the solutes and water than maize (Musambasi et al., 2002) . Growth and development are severely affected by this sink, resulting in yield losses between 30 - 100% ( Parker, 2012). Striga is one of the most widely studied weeds in the world. Countless short - term (1 - 3 year) studies conducted at research stations have investigated the drivers of germination, attachment and emergence. Germination, for example, is largely guided by the seed so wing depth, cereal root canopy and conjugated forms of flavonoids leached by this canopy (among many others) (Chaboud & Rougier, 1991; Doggett, 1994; Ndakidemi & Dakora, 2003). Less 113 research, however, has focused on the dynamic behavior between the three a forementioned stages with respect to time and practice (e.g., manual removal, crop rotation). More specifically, less research has investigated the fluctuation between juvenile, dormant seed and viable seed populations across an extended period of time (>5 years). Understanding feedback behavior between these populations is critical for two reasons. One, different processes leading up to seed production need to be quantified to determine what stages of the lifecycle should be intervened with different contr ols (Van Mourik et al., 2007). Second, quantification of populations at these stages informs farmers how long and to what extent a practice (or practices) should be applied to control Striga emergence at a given threshold. One way to assess interactions b etween weeds and their agricultural systems is to use process - based models. These models simulate competing and synergetic relationships between weeds and crops for light, water, and macronutrients (Keating et al., 2003). The objective of this study was to develop a process - based model that could simulate the accumulation and dissipation of Striga asiatica seeds, juveniles and flowers in a one - hectare field cultivated by a Malawian smallholder. The model addresses three primary questions: (1) how do differe nt seed, underground - juvenile, seedling and flowering populations of S. asiatica fluctuate in response to one another (if at all)?; (2) what overriding interactions or feedback behavior (if any) influence the S . asiatica seedbank?; (3) how does cowpea - ( Vi gna unguiculata ) maize intercropping, mulching and/or ridging influence the emergence and fecundity of S. asiatica in smallholder farming systems? Findings from the study inform which practice or a combination of practices are more effective to control S. asiatica in a Malawian smallholder context. 114 3.2 Background 3. 2.1 Crop Models, Their Required Inputs & Selection Considerations A crop model is a quantitative scheme for predicting growth, development and yield of a plant. Thus, crop models simulate events that have already occurred to inform future decisions about farm management. To simulate weed growth, models typically require four inputs: crop selection, weather data (e.g., rainfall), soil base (e.g., water balance, nutrient balance), an d management specifications (e.g., sowing density, tillage). Many of these data can be collected via household farm surveys. If data is not available for the necessary parameters, literature may be sourced. In the arena of agricultural simulation, there ar e generally three types of models. These include empirical, stochastic or deterministic models. An empirical model is based on observed quantitative relationships between parameters without any insight into the functional or causal operation of the system. A stochastic model uses one or more functional relationships that depends on random parameters, and are thus, related to a probability distribution. Deterministic models are non - stochastic in nature, t hat is, no random variables are recognized. Exact rela tionships are postulated, and the output is predicted by the input with complete certainty. Prior to selecting or developing a crop modeling system, researchers must first consider how the model will be applied. A strategic model focuses on a long - term obj ective whereby inter - year analyses are conducted. A tactical model addresses a within - season decision whereby intra - year analyses are conducted. Then literature suggests modelers should reflect on three questions: (i) what is the intended use of the model? ( e.g., scientific understanding, 115 decision/policy support); (ii) what approaches must researchers take to modify the model (if needed)?; and finally, (iii) what are the target scales for the model? (e.g., field, landscape) (Jones et al., 2016). In this study a dynam ic cropping systems model (CSM) is developed by using Vensim. Vensim is an industrial - strength simulation software used to develop models for analyzing dynamic feedback between stocks (e.g., emerging seedlings, dormant seeds). The CSM is composed of severa l stocks that interact directly or indirectly with one another to demonstrate the fluctuating behavior of a S. asiatica seedbank when one or several control practices are applied. In response to the three questions posed by Jones et al. (2016) - (i) The intention of CSM is not to account for every single component that drives emergence or fecundity, but rather to expand an unde rstanding about the behavior between S . asiatica seedbanks and their relative emergence. In addition, the CSM was developed to serve as a decision - support tool for selecting and determining how long one or several practices should be implemented. (ii) To impro ve model performance, the CSM will combine findings from previous studies and a greenhouse trial to calibrate each component. (iii) The CSM is scaled to a one - hectare field, aimed to conduct inter - year analyses for informing long - term strategies to reduce the S . asiatica seedbank and subsequent emergence. 116 3.2.2 Factors to Consider When Modeling Weed Emergence Several components of the lifecycle of a weed should be considered before developing or modifying a crop model to simulate Striga emergence. These facto rs include allelopathy, conditions required for germination, attachment, emergence, flowering, seed dispersal, seed predation, seed dormancy and control practice. Each factor is included in the proceeding paragraphs of this section. Many times, not all of these components can be included in a model for several reasons. First, data availability and accuracy are two of the largest limitations to calibrate parameters. Second, available data may have been collected at different scales (e.g., kg/ha), creating ch allenges to scale them equally and upload into a model. Finally, adding an exhaustive list of parameters can create more room for error in model outputs (Jakeman & Hornberger, 1993). 3. 2.2.1 Allelopathy Striga spp . germination is triggered by allelopathy. That is, seed conditioning, germination, parasitic contact (attachment) and penetration are mediated by chemical communication between host (or false host) and parasite (Maass, 1999). Once seeds are ripened and exposed to warm moist conditions for several days, exogenous chemical signals produced by a cereal or legume root system can stimulate germination (Worsham, 1987). Elevated levels of soil - phosphorous (P) has shown to reduce the production of these signals, or rather simulants, thus limit ing attachment by underground juveniles (Hearne, 2009). Upon germination, a germ tube, which is in close proximity to the host roots, elongates towards the root of the host, haustorium develop to create a bridge between the parasite and its host. The bridg e then acts as a one - way pump, depriving the host of its water, mineral nutrients and 117 carbohydrates (Frost et al. 1997). Hydrolytic enzymes carry out the penetration of the xylem and/or phloem. Still, allelopathy is a complicated process which is not compl etely understood in the research community. Ndakidemi and Dakora (2003) explain though, when conjugated forms of flavonoids and nitrogenous metabolites (e.g., alkaloids, amino acids) solubilize and enter the soil, they suppress weed seed germination in the Scrophulariaceae plant family (i.e., the plant family of parasitic weeds). 3. 2.2.2 Conditions needed for germination The optimum day/night temperatures for germination and attachment are 15 and 20 ° C, respectively (Baskin & Baskin, 1998). In terms of soil - water content, seeds persist in free drainage environments (e.g., sandy). If exposed to moist conditions for a prolonged amount of time, the seed can enter a state of wet dormancy (Mohamed et al., 1 998). The osmotic potential of seeds requires a preconditioning period at - 1.2 and - 1.5 MPa 4 . Generally, S. asiatica seeds must be exposed to moist conditions for 2 - 3 weeks at warm (26 o C) temperatures prior to germination (Song et al., 2005). There is no light requirement for the plant, but seeds thrive in less - fertile acidic soils, hence their omnipresence across intensely mined soils (Singh et al., 1997). The gaseous environment of the soil can affect germination as well (e.g., ethylene enhan ces germination). Other compounds such as gibberellic acid, strigol analogues and hypochlorite can trigger germination (Visser, 1989). Limited literature is available which explains the extent seeds are susceptible to microbial activity; however, seeds of a similar species ( Alectra vogelii ) will fail to germinate 4 Megapascals (MPa) are units used to measure internal pressure 118 when colonized by Fusarium oxysporum and Fusarium solani (Riches, 1989). Thus, if their maternal environment is housed by these fungi, seed persistence may negatively be affected. Apart from water and light, germination is heavily dependent on the chemical substances produced by the roots of maize that Striga spp . parasitizes (Visser et al., 1987). A number of non - host plants can trigger germination (e.g., Desmodium spp). Strong evidence shows that high production of these stimulants is found in fields with low soil - P. Cereals secrete leachates in these soils to assimilate phosphorous from mycorrhizal fungi in exchange for carbohydrates (Hudu & Gworgwor, 1998). Unfortunately, the very leachates that initiate this symbiotic relationship also catalyze Striga spp . germination. In addition to low soil - P, high soil - potassium (K) increase s germination (Abdul et al., 2012). It is difficult to assess exactly when Striga spp . will germinate because seed development is contingent upon the quantity and quality of root exudates produced by a host or false host (Li et al., 2013) . Research has yet to discover the specific quantities required, but much of the literature suggests a we ll - established maize plant (4 - 6 weeks) can secrete enough leachates to trigger germination. Once these leachates are exuded, germination will occur in approximately five days (Ejeta & Butler, 1993). The radicle can only grow 5 mm so a host root must be loc ated 3 - 4 mm away (Ramaiah et al., 1991). Once a seed germinates, the radicle of parasite must attach to a cereal root within three to five days to survive (Matusova et al., 2005). Otherwise, seed reserves are depleted and root penetration is impossible (Ch ang & Lyn, 1986). Therefore, root architecture of the host drives attachment in the upper soil area (0 - 10 mm) (Gurney et al., 1999). 119 3. 2.2.3 Attachment, emergence and flowering Given the parasitic nature of Striga spp ., its attachment, emergence and arrival to maturity (i.e., flowering) are dependent on resources accessible to maize, allowing access to nutrients/water and photosynthates for the parasite. The timing and extent of erudites leached will determine germinat ion and subsequent attachment. Parasitism of S. asiatica begins approximately 2 - 3 weeks before the weed emerges from the soil. The leaching of erudites can be delayed when there is a soil - P pool available (e.g., manure) for maize to uptake. Once this pool is depleted, timing of leaching is contingent upon the physiological stage of a cereal plant and its P - demands (Yoneyama et al., 2007). After germination, host root length and density (i.e., canopy) will determine how many underground seedlings can success fully attach and emerge from the soil (Cherif - Ari et al., 1990). A healthy host can support between 14 - 17 underground seedlings (Smith et al., 1993). Only 10 - 30% of underground seedlings that attach to the host will emerge from the soil (Doggett 1965). See dlings will emerge from the soil between three to six weeks after attachment (Olivier, 1991). Then, after a period of one to two months, a seedling will mature into a flower (see Figure 4 ) (Parker & Riches, 1993). A ripe seed capsule is dropped one week af ter formation (Webb & Smith, 1996). The number of seeds produced per mature plant vary widely depending on growing conditions, host vigor and host variety (Rodenburg et al., 2006). 120 Figure 4 - Lifecycle of Striga asiatica 3. 2.2.4 Seed dispersal Extreme estimates posit that one S. asiatica plant can produce 400,000 to 600,000 seeds (Visser, 1978). More conservative estimates have found that mature flowers produce between 36,308 and 45,729 seeds/plant in Malawi (Abdul et al., 2012). Microscopic seeds are easily spread by wind and surface water flow. Controlled experiments have shown that S. asiatica can set seed as a result of either self - or cross - pollination. The reticulated surface of the minute seeds trap pockets of ai r when they float on water, making the seed buoyant and easily dispersed at least for short distances on rainwater run - off. The trumpet - like structure of the outer seed coat makes the seeds aerodynamically suited to for wind transfer even in the lightest b reeze. 121 Farmers are the primary dispersal agent through harvest and transfer to un - infested stands. Seeds of the parasite have also been found on contaminated maize grain during threshing and transported to markets or neighboring farms during local sales. Although S. asiatica is widespread across semi - arid agroecologies in Africa, further spread is possible as contaminated maize shipments are distributed throughout the continent. Thus, the introduction of biotypes with differential host specificity from one area to another has caused many problems in sub - Saharan Africa (CABI, 2014). 3. 2.2.5 Predation (at pre - and post - dispersal) Pre - dispersal predators include, Smicronyx, S. albovariegatus , along with a noctuid moth ( Eulocastra argentisparsa) . These pests we re imported from India and released in Ethiopia for Striga hermonthica. S. hermonthica often has similar pests to that of S. asiatica . Agromyzids ( Ophiomyia strigalis ) have also been found to mine the stems of Striga spp. in East Africa but have yet to be evaluated in terms of S. asiatica management. Galling weevils ( Smicronyx spp .) have been found extensively across western Africa predating Striga spp. seed capsules (Pronier et al., 1998). The weevil s either tunne l into the stems, causing galls to develop and disrupt v egetative growth, or, penetrate the seed capsules, negatively impacting seed production. In West African countries such as Ghana, Smicronyx spp . is found in 22.5% to 50% of Striga spp. plants (Kroschel et al., 1995). 3. 2.2.6 Seed dormancy Primary dormancy is broken when ripened seeds are exposed to warm moist conditions ( at 28 - 30°C) for 6 - 10 days followed by the exogenous chemical signals produced by host roots (Elzein & Kroschel, 2003 ). A prolonged period of imbibition by water, in the absence of a stimulant, 122 does not induce wet dormancy. With regard to secondary dormancy, s eeds in a dry state can remain viable for up to 10 years until a host is planted (Bebawi et al., 1984) . Dormant seeds generally lie at >15 cm depth given that 0 - 15 cm is ideal for germination (Baskin & Baskin, 1998; Doggett, 1984). Striga spp. have a type IV persistent seed bank. That is, the seeds remain viable for more than one year and have a large persistent germination rate year - round (pending that the host is growing nearby). The weed does not differentiate between seasons, but rather waits for a host to be planted. Some might argue that if a host, such as maize, is plante d off - season, then this would cl assify Striga spp . as having a type III persistent seed bank. 3. 2.2.7 Controls There are several controls practiced in Malawi which affect Striga spp. pre valence and persistence. These involve removing the weed physically, sowing maize at various depths, altering the soil profile (making less favorable conditions for germination), inducing suicidal germination and predating seeds. Hand - pulling is an effecti ve method to control Striga spp. if the plant is removed prior to the flowering stage. If weeds cannot be removed prior to maturity, flowers can be buried at deep soil depths, so their seed cannot attach to maize between 1 - 30 mm of soil. Also, deep plantin g maize on raised beds can reduce root length in the upper soil layers where Striga spp. seeds are predominantly found. In doing so, less underground seedlin gs can attach, and those that do, may die overtime (Van Delft et al., 2000). Under rain fed conditi ons, Elzein and Kroschel (2003) note that underground development of S. hermonthica was lower when sorghum ( Sorghum bicolor ) was sown in holes 30 cm lower than the surface of the ridge compared to sorghum directly sown in the ridge. In addition to planting 123 at lower depths, transplanting maize has also been found to reduce parasitic attachment (Oswald et al., 2001). The practic e allows maize to develop a juvenile root system before being exposed to parasites, which is less prone to the phytotoxic effect (Graves et al., 1989; Ransom et al. 1996). Intercropping and the rotation of legumes has shown to reduce parasitic weeds in deg raded environments. Certain legumes have the ability to chemically inhibit weed growth by exuding substances from their roots. Legumes with these capacities are often referred to as trap - crops. In the presence of legumes, absent of cereals, parasitic seeds will germinate, transpire and deplete their soil seedbank over several growing seasons (Khan et al., 2010). The extent or rate of suicidal germination is contingent upon legume type, sowing density and planting date. Legume species, such as cowpeas, have been found to induce germination by 60% (Carsky et al., 1994). Others, such as silverleaf ( Desmodium spp.), induce >90% germination (Khan et al., 2010). Much of their success is contingent upon their placement (e.g., in - row) and sowing time (e.g., relay cr opping 10 weeks after sowing) (Oswald et al., 2002). Underground seedlings that do survive may be less effective in attaching to the host, as haustorium development is truncated (Oswald et al., 2002). It is difficult to quantify the rate of suicidal germin ation induced by living roots versus the rate induced by decomposing leaves and roots. Suicidal germination is primarily attributed to living roots that secrete leachates, but a smaller percentage is attributed to decomposing leaf and root tissues (Sanging a et al., 2003). Legume residues left to decompose after harvest (as opposed to being burned or consumed by livestock), therefore, should not be discounted in the control of Striga spp. 124 Several authors posit that the incorporation of legumes and their mul ches in cereal - based systems not only boosts soil fertility, but alters soil conditions, creating less favorable conditions for weed growth. For instance, pigeon pea can utilize iron - bound phosphorous in alfisols, thereby increasing total P availability an d reducing parasitic - potential (Ae et al., 1993). An increase in soil - P via legumes or manure application delays the secretion of strigolactones employed by maize to signal mycorrhizal fungi for assimilating P (Kanampiu et al., 2003). When secretion is del ayed, the time - window for Striga spp. germination is shortened (e.g., 4 - month growing season vs 3 - month growing season). With a shorter time - window, less seeds are able to germinate, leaving them in the soil to decay or become predated. Some argue that or ganic fertilizer application significantly reduces the density of the soil weed seed - bank as well when legumes are incorporated with cereals ( Jiang et al., 2014) . Increased soil - nitrogen (N) is associated with higher N - concentration in maize roots, increas ing their cell - wall, and therefore, reducing cell - wall degradation by enzymes via haustorium attachment (Cechin & Press, 1993). Legume canopies also make less conducive environments for parasites by reducing soil temperature. For example, intercropping gro undnuts in the same row as sorghum has been observed to decrease soil temperature by 2°C at a 10cm depth (Matthews et al., 1991). Such reductions decrease the viability of seeds, thus, decreasing their chance to germinate the following season (Carsky et al ., 1994; Carson, 1989). Viability can also be decreased by increasing the pH of the soil. One practice commonly implemented by smallholders is the application of ash on maize planting stations (Netzly, 1988). The application of biocontrol agents has also been shown to reduce emergence and overall seedbank of Striga. Local insects that damage tobacco ( Spodoptera litura, Heliothes armigera, 125 Myzus persicae ) can infect Orobanche plants, perhaps as a result of tobacco compounds in the parasite. These insects s eem to be a more appropriate control for the weed since they are natural across the sub - Saharan African landscape (SP - IPM, 2003). Little literature cites natural predators of Striga spp. seed once the seed has been dropped on the soil. In terms of post dis persal, the introduction of Agromyzid flies ( Phytomyza orobanchia ) have been found to destroy species in the Orobanche genus though. The rate of attack seems to depend on the relative timing of parasite emergence and insect arrival, but in Morocco, Orobanche seed production was reportedly reduced by 95% (Musselman, 1980 ) . Smith et al. (1993) determined the use of some species ( Sm. Umbrinus ) as a biocontrol agent would need to destroy 95% of the seeds each year to reduce Striga density by 50%. Since the seedbank increases markedly from very low densities in just a few sea sons, weevils may not be able to effectively control Striga alone. As mentioned, there are numerous practices that control Striga spp . Researchers argue these practices must be used to address the weed at specific points or stages of its lifecycle (see Figure 5 ) (Hearne, 2009). For example, legumes can be used to decrease the soil seedbank by inducing suicidal germination. Fertilizer application can be used to reduce attachment. Weeding can be employed to remove emerging seedlings before they flower. An aggregated approach, therefore, is argued as the most effective manner in controlling the weed (Westerman et al., 2007). 126 Figure 5 - Practices that address Striga emergence based on the stage of the weed lifecycle 3. 2.3 Model justification Much of the Striga spp . literature discusses either the effect the weed has on yield or how effective a control practice is by using cereal yield as an indicator. Less literature, however, focuses on emergence of Striga spp . or its subsequent seedbank. Even fewer, study or quan tify attachment. There are several reasons why emergence or seedbanks are less studied. First, Striga spp . seeds are microscopic and difficult to monitor and/or quantify (Van Mourik et al., 2008). Like seeds, emerging seedlings are difficult to identify an d many transpire quickly after emerging from the soil. Second, cereal yields are less affected by Striga spp . if ample resources (e.g., fertilizer) are available for production (Doggett, 1975). Unfortunately, the majority of Malawians cultivate low - input s ystems, augmenting the effect parasitism has on their maize yield (Parker, 1991; Ransom et al., 1990). Thus, the assessment of Striga spp . emergence is becoming ever more critical to determine maize yield production in smallholder fields. 127 There is a consi derable amount of deterministic or stochastic Striga spp . models found in the literature (Abdul et al., 2012; Chikoye et al., 2011; Ekeleme et al., 2014; Tarfa et al., 2006). The following models often use statistics and/or econometrics as a tool to predic t emergence, attachment and flowering. Application of these models offers opportunities to identify certain interactions (e.g., soil acidity: Striga spp . emergence) that do not occur out of coincidence. Identification of these processes or interactions imp roves the understanding about determinants of Striga spp different processes interact across different scales in a non - linear way, and such interactions are poorly iang et al., 2004, p. 298). In addition, field experimental data is needed to calibrate numerous parameters needed to model underlying processes. Many times, these data are not available, inaccurate and/or expensive to collect. Some researchers have devel oped process - based simulators in an effort to address these challenges ( Kunisch et al., 1991; Van Mourik et al., 2008 ). Through a systems approach, researchers simulate the fluctuation of different stocks (e.g., soil moisture) relative to outside parameter s (e.g., root growth, evapotranspiration rates). These parameters do not behave in a linear fashion, but rather, change according to the fluctuating stock they feed into or pull from (Kopainsky et al., 2012). It is important to account for the fluctuation or plasticity of parameters. While quantitative analyses such as econometrics can illustrate the elasticity or sensitivity certain variables have upon Striga spp . emergence, system dynamics allows the observation of both the elasticity and plasticity of pa rameters. In addition to this flexibility, parameters in systems models can be applied with values or equations (provided by the literature) when there is incomplete or missing data. While this is not advisable, researchers 128 can develop and run systems mode ls without being confined to data availability. As a final note, one of the most the most significant contributions of the method is its ability to model causal relationships, and therefore, test hypotheses about causation. 3.3 Methods The manner in which the CSM was parameterized is explained in the following section. After outlining the development of the model, study area and protocol of a greenhouse experiment are defined. Finally, the methods used to analyze model outputs and greenh ouse results are explained. 3. 3.1 Model review Different stages of the Striga spp . lifecycle were used to inform how to construct the CSM. Supporting literature was then sourced to confirm the structure and apply values or equations to its parameters (e.g., seedbank, attachment, germination and flowering parameters) . Several reviews and short - term studies provided starting values and equations for parameters. To explain the manner in which the CSM was developed, the structures of several models are presented in this subsection. Then, the parameterization of the CSM is described based on t he strength and limitations of each model. 3. 3.1.2 Model 1 3. 3.1.2.1 Structure and Objective In the first model ( Figure 6 ), Smith et al. (1993) developed a deterministic biocontrol simulator (DBS), with transition probabilities defining the proportion of S. hermonthica plants surviving from one life stage to the next. The DBS employs an annual time - step, evaluating seed populations in a one square meter area where millet ( Pennisetum glaucum ) is sown. The model 129 evaluates the potential of gall - forming weevils ( Smicronyx umbrinus Hustache) as a biocontrol agent of S. hermonthica . The structure of the DBS was developed from the earlier work of Kunisch et al. (1991 ) where Striga spp . develops at five defined stages, beginning from an existing seed bank and ending with the dispersal of new seeds from flowers. In between these two stages, germination (i.e., the preconditioning and stimulation by root exudates), attach ment, underground growth, emergence and maturity occur. This model appears to be one of the earlier Striga spp . emergence simulators developed and its structure has since been used to inform several others since. Figure 6 - Structure and flow between state variables of Striga spp . Note: Structure of the model showing flow between the state variables [ Striga seed bank (m - 2 ), X j ; simulated Striga seeds (m - 2 ), X e ; Striga seeds produced (m - 2 ), X f ] according to the annual transition rates (probability of stimulation, p 1 ; probability of emergence, p 2 ; average seed production per emerged plant, p 3 ; proportion of seeds destroyed by Smicronyx, s ; seed viability, p 4 ) Source: Figure modified from Kunisch et al. (1991) X i X s X e X f P i P 2 P 3 1 - s 1 - P 1 P 4 130 3. 3.1.2.2 Strengths and limitations There were several strengths and limitations of the DBS, particularly if it is used as a tool for evaluating control strategies to reduce the S. hermonthica seedbank. One of the primary strengths lies in its ability to show the persistent seedbank of S. he rmonthica, even from only several flowers emerge in a given field. By accurately reflecting this high fecundity, the model is able to postulate whether weevils are an in/effective biocontrol or not. There were several limitations to the model. First, much of the parameters were applied from fragmented data collected from farmer fields or pot experiments (Van Mourik et al., 2008). Fragmented data limits how much results can be extrapolated to other settings. Second, the model was not sensitive to changes in key parameters such as the control agent population. Predator populations should fluctuate according to seed availability and vis - à - vis. Given this non - dynamic nature, model results do little to inform how much (or how little) a control should be administe red seasonally. 3. 3.1.3 Model 2 3. 3.1.3.1 Structure and objective In the second model ( Figure 7 ), was developed by Van Mourik et al. (2008). The stochastic model assessed the probability of successful establishment of S. hermonthica in millet - ( Penniset um glaucum ) based systems. In this model, seeds dispersed by flowers are added to the soil seedbank at the end of each annual time step. Seven stages are outlined in the structure, including viable seeds, germinated seeds, attached seedlings, emerged plant s, mature reproductive plants, seeds on reproductive plants, seed shed by flowers, and viable seed added to the seedbank. The model evaluates different controls of S. hermonthica based 131 on seed populations across a one - hectare field. These practices include planting long and short duration millet varieties, weeding of S. hermonthica flowers at different times, intercropping cowpea and sesame ( Sesamum indicum ) or planting them as fallow crops. Each practice was evaluated separately. Figure 7 - Life cycle diagram of Striga hermonthica Source: Van Mourik et al. (2008) 3. 3.1.3.2 Strengths and limitations The stochastic model is powerful in the sense that it assesses the probability of attachment in un - (Van Mourik et al., 2008, p.,84). In addition to highlighting this vulnerability, the model informs readers that intercrops or rotator crops must be employed for at least 3 years to reduce the seedbank to a significant 132 threshold (90%). Practically speakin g, many experiments do not show this reduction in three years, regardless of setting (e.g., farm, experiment) (Abunyewa & Padi, 2003; Franke et al ., 2006; Murdoch & Kunjo, 2003; Oswald & Ransom, 2001; Schulz et al ., 2003). A 90% reduction in a seedbank or emergence is seldom reported in three years, and the studies that do report such reductions are typically conducted under strict controlled conditions (Khan et al., 2010). The projection limits how much model results can be extrapolated to field outcomes. In addition, model developers admit that more life cycle processes should be added to their . Finally, model simulations indicate that planting a cowpea intercrop with millet simultaneously should suppress S. hermonthica emergence. In reality though, smallholders sow cowpea two to four weeks later and do not have as high of a sowing density used in the model. 3. 3.1.4 Model 3 3. 3.1.4.1 Structure and objective In the third model ( Figure 8 ), Grenz et al. (2005) developed a model from APSIM software (Agricultural Production systems Simulation) to simulate the parasitism of broad bean ( Vicia faba L.) by broomrape ( Orobanche crenata Forsk.). Broomrape is in the same genus (Orabanche) as Striga spp . The objective of the model was to quantify the effect crop rotation, tillage, hand - pulling and combined strategies on a dynamic seedbank. As seen in the parameters of the model, seed production is measured by parasite dry weight. Seeds fluctuate at three different levels in th e soil. Seed viability follows a negative exponential function given that seed decay is driven by soil moisture. Effect of external factors such as temperature are 133 not included. The model is calibrated with data that reflects daily changes in soil moisture (0 - 15 and 15 - 30 cm), precipitation, broad bean root length density and dry weight of emerged parasites. The model is event - based whereby events occur over crop development stages, catalyzing changes in the broomrape seedbank. Figure 8 - Flow diagram of parasitic weed crenate broomrape ( Orobanche crenala Forsk .) in APSIM Note: Dashed arrows represent input of information form APSIM modules (module names printed in bold). Simulated crop phenolog y and rules defined in the APSIM - Mana broad bean (Vicia faba L.) . For further information on APSIM framework, see Keating et al., (2003). Source: Grenz et al. (2005) 3. 3.1.4.2 Strengths and limitations Calibration of parameters at daily time - steps allows more accurate interactions to occur between the host, parasite, environment and imposed management practices. In doing so, the model is ab le to make more precise assessments and conclude that the parasite population can only be contained by combining several management approaches. As detailed as the model is, there are several limitations. First, seed decay is only driven by soil moisture, r egardless of soil temperature, seed predation, microbial activity and soil chemical p roperties (e.g., pH), which are considered as important determinants of seed decay (Grenz et al., 2005). Another limitation 134 is that seeds are assumed to be distributed hom ogenously across horizontal layers of soil. In smallholder f ields, parasite seed distributions are usually patchy (González - Adujár et al., 2001). This assumption can cause an overestimation of parasitism. Finally, the model only specifies one event /stock w here seeds germinate and attach to millet. In fact, this event should be separated into three different events: germination, seedling formation without attachment, and attachment to the host ( t er Bor & van Ast, 1991). 3. 3.1.5 Model 4 3. 3.1.5.1 Structure and objective In the fourth model ( Figure 9 ), Westerman et al . (2007) developed a density - dependent feedback model (DDFM) to simulate sorghum parasitism by S. hermonthica. The researchers developed the model to assess how attachment was affected by host varieties that emitted fewer exudates from their root systems. The objective of the model was to identify key points in the S. hermonthica life cycle where inter vention strategies could be applied. The model sourced literature to apply values/equations for parameters and was run using existing and earlier published data. The model is divided into 10 steps, each assuming different transition probabilities (except f or seed production). The processes and probabilities associated with the - host cues (n), (3) attachment (a), (4) establishment (b), (5) subsurface growth until seedling emergence (e), (6) the proportion that develops into above - ground vegetative plants (v), (7) the proportion that becomes reproductive (r), (8) seed production (s), (9) vi ability of newly produced seeds (l), 135 and (10) survival of non - germinated seeds into the next season (1m; with m mortality of seeds Figure 9 - Density - dependent feedback model (DDFM): Striga hermonthica emergence in sorghum - based system Source: Westerman et al. (2007) 3. 3.1.5.2 Strengths and limitations Unlike the previous three models, the DDFM emphasizes the importance of feedback behavior. Also, the model highlights the importance of parameter, which strongly influences the populations of the other nine steps regardless of number of seedlings that ca n attach at specific time points during the sorghum lifecycle. Although the model emphasizes how important feedback behavior is in the lifecycle, only two - rest of th e parameters operate in a consecutive linear fashion (i.e., one after the other). Another limitation of the model is that it operates in annual timesteps, constricting the effect of 136 any inter - seasonal events (e.g., rainfall, fertilizer application). Lastly , the model only includes one parameter that influences non - Controls germination may be trigged by several factors, including host exudates, false - host exudates (e.g., legumes) and false - host residues. Each o f these factors induces germination at different rates (e.g., 95% vs 10%). 3. 3.2 Model development The objectives, structure, strengths and limitations of each model were used to inform the development of a single CSM (see Figure 10 ) to simulate S. asiatica emergence in maize - based systems across Malawi at monthly intervals. Explanations regarding these considerations are outlined in the subsection below. 137 Figure 10 - Cropping Systems Model (CSM): Emergence of Striga asiatica in maize - based system 138 Figure 10 Note: Secondary view of parameters embedded within initial structure 139 3. 3.2.1 Parameter considerations The general structure of model 1 was used as a jumping point to initially parameterize the CSM. After which, other parameters were added. In model 1, however, predation of Striga spp . seed was included, which were not included in any of the other models di scussed. Second, the application of an equation for the flowering parameter with a high seed - rain value was used in the CSM to accurately reflect high fecundity. Third, the emergence factor in the CSM was validated sourcing a long - term S. asiatica control experiment in Malawi. The factor was validated in this manner to avoid relying fragmented data which model 1 had done. Fourth, the CSM connects multiple parameters (e.g., attachment population relative to germination rate) to reflect the dynamic nature of a parasitic seedbank. This type of parameterization was carried out so seedbank populations were sensitive to the intensity of one or more control practices. Model 1 did not account for these control practices. - maize intercrop parameter was integrated into the CSM structure. Van Mourik et al. (2008) mentioned that calibration of certain management p arameters (e.g., intercrop sowing date) in model 2 did not align with farmer practices, rendering an overestimation of S. asiatica control in their results. To avoid inaccurate outputs, the CSM calibrated its emergent rate factors with results from a green house experiment using farmer - managed soils. This experiment will be explained later in the study. In addition, the greenhouse experiment followed a protocol that attempted to mimic farmer practices (e.g., fertilizing maize two weeks after sowing as oppose d to at sowing) in an effort to more accurately reflect emergence in the field. 140 Viable seed depth (0 - 15 cm) and dormant seed depth (15 - 30 cm) were two key components integrated from model 3 into the CSM structure. In addition, a tillage parameter was conn ected to these two components to guide seedbank fluctuation. This soil disturbance parameter was added since dormant seeds can be shifted from lower depths (15 - 30 cm) to higher ones (0 - 15 cm) (Van Delft et al., 2000). Root density and root depth could not be added to the CSM, but the number of parasites that can attach per maize plant were included. Grenz et al. (2008) advised future modelers to consider the use of multiple parameters to model the efficacy of controls, above and below ground. Hence, the con trol parameters (e.g., weeding, intercropping) were adjustable in the CSM so weed emergence could be evaluated in the event a farmer used one or several control practices. Model 3 developers also admitted that key external factors such as temperature and m oisture were not included in the calibration of seed decay and germination. The CSM included these factors. Two critical strengths of model 4 lay in its ability to simulate feedback and specifying a threshold of attachment by S. hermonthica onto sorghum. T hese two specifications in model 4 informed the parameterization of a maize root canopy with max attachment converter in the CSM as well as a reinforcing loop between germinating seed stock and the attachment outflow. Both specifications allowed the CSM to illustrate how the flooding of several thousand germinating seeds (as compared to hundreds of thousands) can still drive max emergence and flowering rates. The developers of the DDFM mentioned that the model did not necessarily demonstrate the relationshi ps between increased soil fertility and consequent crop growth affected parasitism (Westerman et al., 2007). This statement encouraged the inclusion of these three parameters and their respective converters to reflect this relationship. First, manure 141 appli cation was attached to germination to delay germination based on soil - P. Second, legume intercropping was connected to the attachment rate to decrease successful attachments made between the parasite and the host due to reduced appendage growth. Lastly, fe rtilizer application was connected to the attachment stock to illustrate how fertilizer application and thicker root tissues allowed attachment, but not the siphoning of photosynthates. As a final note, none of the four models simulated the parasitization of maize by S . asiatica . The CSM modeled this specific host and parasite species given their cultivation and emergence, respectively, across Malawi . The parasitization of millet and sorghum by S. hermonthica is more characteristic of Western and Eastern A frica (Kim et al., 2002). Among the 42 Striga species, asiatica is the most widespread across Africa (Nail et al., 2014). 3. 3.2.2 Application of values/equations to parameters The table below outlines the application of various parameters in the model. The values and/or equations applied to each parameter are explained in the Appendix 1 . Each explanation is supported by literature and/or previous studies that investigated the germination, attachment, emergence and flowering of S. asiatica in the presence of maize across southern Africa. If no values or equations could be found in studies with the aforementioned context, other sources were used with species akin to S. asiatica (e.g., S. hermonthica ) under semi - arid conditions (i.e., Benin). 3. 3.3 Greenhouse experiment A greenhouse experiment was conducted from December 2017 to March 2018 at the Chitedze Research Station in Malawi in order to evaluate the effects of tillage and cowpea - intercropping had on S. asiatica emergence in maize based systems. Soils used for the experiment were 142 collected from farmer - managed plots from July 20 - 22, 2017. Farmers were affiliated with the Conservation Agriculture project funded by Total Land Care (TLC) and CIMMYT - Harare in partnership with the Ministry of Agriculture in Malawi. The project funded by TLC advocates for minimal disturbance of the soil, retaining crop residues on the soil surface during - and off - season and/or rotating legumes. After forming a cooperative (10 - 12 farmers), cooperatives in communi ties are given several inputs (fertilizer, maize seed, cowpea seed) by TLC to assist with beginning one or all farming practices. 3. 3.3.1 Site Description Soil sampling took place in the Zidyana and Mwansambo extension planning areas (EPA) which are part o f the Salima Agricultural Development Division (ADD), central Malawi. The EPAs are located in the southern region of the Nkhotakota district along the lakeshore plain. Altitude ranges between 200 - 500 meters above sea level and receive a mean annual rainfal l between 600 - 800 mm. Rainfall commences in November and generally concludes in April. The EPA is generally comprised of alkaline Lithosols having a pH of 6.1. The texture of soils generally ranges between loamy - sand (upper region) to sandy - loam (lower reg ion) given that they are situated near the lake (Kanyama - Phiri et al., 2000). Farmers primarily cultivate maize, cotton ( Gossypium ) and cassava ( Manihot escuelenta ), but rice ( Oryza sativa ) along river valleys. 3. 3.3.2 Sampling Two phases of sampling occurred during the study. The first sampling phase consisted of selecting farmers to collect soils from for the greenhouse experiment. The second sampling phase took place after soils were transferred into pots. During this phase, S . asiatica emergent rates were observed in a greenhouse . Each sampling phase is outlined below. 143 3. 3.3.2.1 Farmer field soils Soils were collected from farmer plots that administered one or all three practices over different periods of time. The practices ranged from - 1.) soil disturbance (ridging [via hand hoe] or zero tillage carried out during November 2016 to prepare fields for the 2016/17 growing season); 2.) soil cover (applied with maize residues or removed post - harvesting); and 3.) crop diversity (cultivated w ith continuous - maize or intercropped with cowpea). According to these practices, plots in farmer fields were first segregated into three strata - 1.) minimum tillage + mulching with sole maize (treatment 1 [T1]) 2.) minimum tillage + mulching with maize - cowpea in tercrop (treatment 2 [T2]) 3.) conventional tillage with sole maize (treatment 3 [T3]) Then, plots were further segregated into three sub - strata according to length of practice - 1.) <4 years 2.) 4 - 7 years 3.) >7 years In total, 15 farmers were selected which cultivated al l three treatments. Five farmers were selected who had been cultivating the plots for <4 years, five who had been cultivating the plots for 4 - 7 years and five who had been cultivating the plots for >7 years. Each field was approximately 0.25 ha of which wa s sub - divided into 3 equal plots (0.08 ac each) for each practice. In T1, farmers administered minimal soil disturbance at time of sowing, crop residues were retained for mulch and maize was planted on a flat plain with rows spaced at 75 cm apart and 25 cm between stations with one seed per station. In T2, the same tillage and residue 144 management were administered with cowpea planted alongside maize, but in between the stations. This equates to 25 kg of maize seed per hectare, equating to 53,000 plants. For cowpea, 35 - 40 kg seed were sown per ha, equating to 27,000 plants. Treatment 1 and 2 rendered 45 - 60% groundcover. In T3, no crop residues were retained, ridges were approximately 22 - 30 cm wide and 75 cm apart. Maize spacing was the same as the two previous treatments (refer to Table 1 2 for further clarification about practices associated with treatments). The maize variety used in these plots were hybrid, DKC8033. The cowpea variety used was IT1833. Four bags of fertilizer were applied; two for basal dressing (23 - 21 - 0+4S) and two bags o f urea (46%) for top - dressing, giving the recommended rate of 69 kg/ha. Table 12 - Details of Plot Management Minimum tillage Ridging Residue applied Cowpea intercrop Treatment 1 x x Treatment 2 x x x Treatment 3 x All samples were collected according to grids defined in sketch below ( Figure 11 ). Samples were taken from alternate grids on rows. Second grid sampling points were made half way inside from the first and third grid points. Each sampling point was separate d by 20 meters vertically and 10 meters horizontally. 145 Figure 11 - Sampling procedure conduct in farmer plot Using a hand hoe, five samples were collected at two sampling depths (0 - 10 cm and 10 - 20 cm) in each plot. For T1 and T2, one sample was taken from the 0 - 10 cm depth, and then in the same hole, another sample was taken at the 10 - 20 cm depth. For T3, one sam ple was taken from the ridge (for the 0 - 10 cm depth) and one sample was taken from the furrow (for the 10 - 20 cm depth). Therefore, for T1 and T2, 30 samples were taken from the 0 - 10 cm depth and 30 samples were taken from the 10 - 20 cm depth. For T3, 30 sam ples were taken from the ridge and 30 samples were taken from the furrow. A total of 60 samples were taken per field, 180 samples per farmer and 900 soil samples for the entire experiment (see Table 1 3 ). Each sample consisted of 2 kg soil samples in 5 kg blue polysacks. Table 13 - Number of soils samples per treatment, depth and field location (ridge vs furrow) Treatment Depth Ridge Furrow Total Treatment A 0 - 10cm 75 75 300 10 - 20cm 75 75 Treatment B 0 - 10cm 75 75 300 10 - 20cm 75 75 Treatment C 0 - 10cm 75 75 300 10 - 20cm 75 75 TOTAL 450 450 900 146 3. 3.3.2.2 Sampling Greenhouse data collection and protocol Two types of data were collected from the greenhouse experiment. First, S. asiatica emergent rates, and second, soils. Six weeks after sowing (September 9, 2017), emerged seedlings were counted daily. Emergence was observed twelve weeks after sowing (Janua ry 12, 2018). Aggregate samples from each farmer according to soil depth and treatment were sampled for soil analysis. In total, 90 samples were collected (15 Farmers x 2 depths x 3 treatments), 6 from each farmer. 3. 3.3.3 Greenhouse design Observation of S. asiatica emergent rates were conducted in a greenhouse at Chitedze Research Station. Chitedze is located on latitude 13 0 0 Malawi. The site is 1146 meters above sea level, has a mean annual temperat ure of 20 o C and mean annual rainfall of 892 mm. Mean maximum and minimum temperatures are 24 o C and 16 o C respectively (MoAFS, 2007). 3. 3.3.3.1 Transfer from field to greenhouse One composite sample was taken for each depth at each plot to analyze the pH, soil texture, NPK and organic matter. The five samples taken at each plot for each depth were mixed, totaling to approximately 10 kg (5 sampling points x 2 kg) composite sample a vailable for the soil analysis and the greenhouse experiment. Given that 5 kg of soil were needed for filling the 6L pots to analyze emergence in the greenhouse , the remaining approximate 5 kg were available for analysis. Approximately 1 kg was needed for each composite for soil analysis, leaving approximately 4 kg to supplement any 6 - liter pot that had a deficit depth after watering. The researchers had to account for deficits, given that some soils were loamier than 147 others. Hence, when watering, the soil depth would decrease, requiring more soil. One 6 - liter pot was filled with five samples taken for each plot at one depth. Each pot was sown with a susceptible host (8338 - 1) (see Figure 12 ) . In all, there were 360 samples (i.e., 60 samples per treatment) us ed for soil analysis; 180 samples for the top layer (0 - 10 cm) and 180 samples for the bottom layer (10 - 20 cm) (see Table 1 4 ) . Figure 12 - Greenhouse experiment at Chitedze Agricultural Research Center Table 14 - Details of greenhouse sample Treatment Depth Ridge Furrow Total Treatment A 0 - 10cm 30 30 120 10 - 20cm 30 30 Treatment B 0 - 10cm 30 30 120 10 - 20cm 30 30 Treatment C 0 - 10cm 30 30 120 10 - 20cm 30 30 TOTAL 180 180 360 148 3. 3.3.3.2 Labeling & pot organization Each pot was labeled according to treatment, depth and farmer. As mentioned previously, treatments were labeled as T1, T2 and T3. Depths were labels as A (0 - 10 cm) or B (10 - 20 cm). Farmers were coded with the values 1 through 15. For example, Treatment A, taken at 0 - 10 cm, from farmer 1 would be labeled as T1A1. For the bottom layer (10 - 20 cm) of the same farmer, the label would read T1B1. Farmer codes were loaded in GenStat (Discovery 18th Edition) 24 times (360 pots / 15 farmers = 24) in chronological order. Then the program randomized the order in one column. In the parallel column, 1 through 360 were inputted. With thi s order, the pots w ere placed in the greenhouse in chronological order (see Table 1 5) . In the greenhouse , four blocks were created, equating to 90 pots per block. Each block was comprised of pots with different sampling locations, soil depths and treatments. Table 15 - Location in greenhouse based on sampling location 1 Random # (Pot Placement) Block Treatment Soil Depth Sample Location (Farmer #) 2 Corresponding Code 1 1 Continuous CA + Sole maize (T1) 0 - 10 cm (A) 2 T1A2 57 1 Sole maize conventional tillage ( T3) 10 - 20 cm (B) 14 T3B14 91 2 CA + Maize/legume intercrop 2 (T2) 0 - 10 cm (A) 2 T2A2 116 2 CA + Maize/legume intercrop (T2) 10 - 20 cm (B) 14 T2B14 188 3 Sole maize conventional tillage ( T3) 0 - 10 cm (A) 10 T3A10 226 3 Continuous CA + Sole maize (T1) 10 - 20 cm (B) 3 T1B3 310 4 CA + Maize/legume intercrop (T2) 0 - 10 cm (A) 8 T2A8 352 4 Sole maize conventional tillage ( T3 ) 10 - 20 cm (B) 7 T1B7 1 Written on tag in pot 2 Written on pot *Note: Farmers 1 through 5 corresponded to 0 - 4 years of practice; farmers 6 through 10 corresponded to 4 - 7 years of practice; farmers 11 through 15 corresponded to >7 years of practice 149 3. 3.3.3.3 Management protocol On July 15 th (2017) all pots were transferred to the greenhouse . For a two - week period, pots were watered each morning using a watering can with a fine rose top to keep the soil moist and avoid hard pan forming on the surface. This action was carried out to mimic the two - week conditioning period where S. asiatica received rainfall in farmer fields from light rains in November. After two weeks of conditions, all pots were sown with maize on July 29 th , 2017. After which, watering was continued daily for 3 months to mimic the rainy season. Watering was based on physical observation of whether the pot was dry or not to avoid overwatering of the seeds. Pots were not perforated to avoid the leaching of nut rients and reduced moisture stress. In previous experiments, perforated pots dried quickly and required researchers to water plants frequently (3 times per day). In addition, dry pots increased plant stress (Mwale, 2009). Pots were applied two weeks after sowing (August 17 th , 2017) with a basal application of the same fertilizer (23 - 21 - 0+4S) used in farmer fields. Pots were applied with the equivalent amount of 33 kg NPK/ha. Application was only made on pots where emergence occurred so as to mimic fertiliz er application practices employed by farmers. Fertilizer is applied in this manner to avoid application to seeds that will not emerge. No side dress of Urea was made given that few farmers in the sample had done so in the past nor was it considered as a co mmon practice countrywide given the state of poverty. 150 3. 3.4 Analysis 3. 3.4.1 Greenhouse 3. 3.4.1.1 Soil At the Chitedze Agricultural Research Center, samples were analyzed for pH, % organic matter, % nitrogen ( NO 3 - N ), phosphorous (ug/g), potassium (Cmol/Kg), % sand, % silt and texture class. Analyses were conducted by six soil technicians in the Malawian Ministry of Agriculture Soil Analysis Lab. The lab was located at the Chitedze Agricultural Research Center. Soil was passed through a 2 mm screen after being air - dried. Soil - P and K were determined using an ammonium fluoride (NH 4 F) and an ethylenediaminetetraacetic acid (EDTA) based extradant that is associated with extractants used in the semi - arid tropics (Wendt, 1 995). Soil pH was determined in 1:2.5 soil/water ratio (Snapp et al., 1998). Texture was assessed by dispersal and hydrometric readings (i.e., a measure of density) (Anderson & Ingram, 1989). As discussed in previous sections, Striga spp . emergence is asso ciated with different levels of nitrogen, phosphorous, potassium and soil texture. The connection between organic matter (OM) and Striga emergence is more abstract than nitrogen, phosphorous, potassium and soil texture. While abstract, OM was collected for several reasons. OM is composed of living and decomposing plant and animal remains. Different percentages of OM are associated with different practices and lengths of implementation. For example, consecutive annual applications of mulch can increase OM. I n addition to sand, silt and clay, soils are composed of OM particles. Clay and OM have net negative charges. That is, they are composed of more negatively charged ions (i.e., anions) than positively charged ions (i.e., cations). The net charge of soils de termines their anion exchange capacity (A EC). A EC influences the ability of soil 151 particles to absorb negatively charged minerals , such as inorganic forms of phosphorous (i.e., orthophosphate [H 2 PO 4 - ] ) . AEC is affected by amount of clay particles it is comprised of as well as t he addition of OM. More clay particles and OM allows for more inorganic - P to be capture d and later assimilation by pla n ts such as maize. As the quantity of clay particles increases, t he P - sorption capacity increases (Cordell et al. , 2009) . Clay particles have relatively larger surface area than sand particles which affords for more phosphate sorption. The addition of OM provides organic phosphate which can be converted to inorganic forms of phosphate later . Also, organic anions from OM can displace sorbed phosphate that are tightly bound to positively charged particles, liberating them for plant uptake (Buresh & Tian, 1997) . Furthermore, humus in OM coats aluminum and iron oxides, which redu ce P sorption by positively char ged soil particles (Cordell et al., 2009) . As more phosphorous become available, secrete less strigolactones. Reduced secretion of strigolactones decreases and/or delays Striga germination rates. 3. 3.4.1.2 Emergent rates Parametric tests were planned to b e used to analyze emergent rates across practice and/or length of implementation. More specifically an analysis of variance (ANOVA) was going be used to determine if there are any statistical differences between practices and/or length of implementation. I f there were, then t - tests would be used to analyze if there are any positively or negatively significant relationships between S . asiatica emergence and practice. These tests were not used, reasons for why they were not used are explained in the results s ection. 152 3. 3.4.2 Model runs 3. 3.4.2.1 Sensitivity analysis A sensitivity analysis was used to evaluate how subtle changes in parameters of the CSM affected the seedbank population as well as attachment, emergence and flowering rates of S. asiatica . In crop modeling, sensitivity analyses are used to investigate the resistance and resilience of simulated outcomes against certain events (Patten, 2013). Previous studies have evaluated how different climate parameters affected germination (Elzein & Kroschel, 2003). In this study, lifecycle stage parameters, such as mont hly germination rate and monthly successful attachment rate were adjusted 25 - 50% from their specified averages in the CSM to determine their effects upon subsequent populations, such as emergent seedlings and mature adults. 3. 3.4.2.2 Scenario analysis Scenario analyses are typically conducted in dynamic crop models to investigate the relationship between model outcomes and parameters guided by manager modules (Patten, 2013). For example, the integration of legumes in cereal systems are initiate d in manager modules to determine the volume of soil - N associated with intercropping (Waddington, 2002). A scenario analysis, therefore, was conducted to determine which stage or stages of the S. asiatica lifecycle were notably affected by a single managem ent parameter (e.g., weeding, intercropping). In addition, the extent to which each stage or stages was affected by a management parameter was assessed. Thereafter, different management practices were initiated together to determine which combination reduc ed one or multiple stages in the least number of months. Such analyses are generally conducted in a consecutive manner, adding one practice after another, to assess the number and type of practices required to bring a lifecycle 153 stage to an acceptable thres hold (University of Honhenheim , 2015 ). These analyses entail running the model with status quo practices (e.g., hand pulling) first and then adding newer practices such as fertilizer application and/or intercropping. 3. 3.4.2.3 Validation When testing the accuracy of process - based models, it is important to compare model output Kobayashi & Salam, 2000, p. 345). In this manner, field observations (i.e., S. asiatica counts) are plotted against simulated outcomes, a correlation coefficient is calculated and regression lines are fitted. Unfortunately, there were not enough emergent s eedling observations to compare model behavior across different soils, practices and lengths of implementation. Another avenue used to validate model behavior is by comparing runs against the results of other peer - reviewed articles. For instance, the use o f legume intercrops in combination with other SFM practices (e.g., fertilizer application) should take no less than 3 years to significantly reduce the soil seedbank (Abunyewa & Padi, 2003; Franke et al ., 2006; Murdoch & Kunjo, 2003; Oswald & Ransom, 2001; Schulz et al ., 2003). This behavior is exemplified in Model 3 by Van Mourik et al. (2008). Model runs can also be validated if the behavior of specific stocks respond to specific controls. For example, Model 4 exemplified drastic differences in attachment between traditional sorghum and S. hermonthica - resistant sorghum varieties (Westerman et al., 2007). If a control, such as weeding is loaded in the CSM, drastic drops in mature S. asiatica populations should be reflected in the results. 154 3. 4 Results & Dis cussion 3. 4.1 Soil Soil analyses revealed subtle differences in soil - acidity, organic matter (OM), nitrogen (N), phosphorous (P), potassium (K), percent clay, percent silt and overall texture between S. asiatica control practices, sampling depth and length of practice. These differences are presented in Table 1 6 . Table 16 - Soil analyses of practices by length of implementation and soil depth Continuous Maize + Minimum Tillage + Mulch Maize - Cowpea Intercrop + Minimum Tillage + Mulch Continuous Maize + Tillage All Practices 0 - 4 Years 4 - 7 Years >7 Years All Years 0 - 4 Years 4 - 7 Years >7 Years All Years 0 - 4 Years 4 - 7 Years >7 Years All Years Avg. 1 Text Class 0 - 10cm 2.40 2.60 2.60 2.53 4.10 3.00 2.80 2.97 2.90 3.00 2.50 2.80 2.71 10 - 20cm 3.70 2.50 2.70 2.73 2.90 2.80 2.60 2.77 2.40 2.60 2.40 2.47 Both Depths 2.70 2.55 2.65 2.63 3.00 2.90 2.70 2.87 2.65 2.80 2.45 2.63 2 OM 0 - 10cm 2.35 1.98 2.19 2.18 2.68 2.48 2.32 2.49 1.97 2.09 2.22 2.09 2.24 10 - 20cm 1.88 1.78 2.46 2.04 2.27 2.10 2.87 2.41 2.00 1.88 2.85 2.24 Both Depths 2.12 1.88 2.32 2.11 2.48 2.29 2.59 2.45 1.98 1.98 2.54 2.17 155 Table 16 1 Text Class represents an ordinal value of soil texture where 1 = sand, 2 = sandy loam, 3 = loam, 4 = silty loam, 5 = clayey loam, 6 = silty clay and 7 = heavy clay 2 OM represents total percent organic matter in soil 3 Clay represents total percent clay in soil 4 Sand represents total percent sand in soil 5 N represents total percent nitrogen in soil 6 P represents soil - phosphorous measured in units of micrograms by gram (ug/g) 7 K represents soil - potassium measured in units of centimoles by kilograms Cmol/Kg The physical characteristics of each soil across practices, sampling depth and length of practice varied little. All samples were characterized as primarily sandy loams (40) or loamy sands (42), aligning with regional classifications of soil within the cen tral region of Malawi (Snapp et al., 1998). Soil texture varied little across practices, especially with regard to clay content 3 Clay 0 - 10cm 15.04 12.46 17.94 15.15 17.84 12.40 18.34 16.19 19.04 11.60 16.74 15.79 16.58 10 - 20cm 20.24 13.77 20.34 18.11 19.84 13.20 19.54 17.53 15.84 14.00 20.34 16.73 Both Depths 17.64 13.12 19.14 16.63 18.84 12.80 18.94 16.86 17.44 12.80 18.54 16.26 4 Sand 0 - 10cm 78.56 80.74 77.66 78.98 79.36 84.80 76.86 80.34 77.36 85.60 78.06 80.34 79.25 10 - 20cm 76.16 80.8 74.86 77.28 76.16 84.80 76.46 79.14 80.56 82.80 74.86 79.4 Both Depths 76.86 80.77 76.26 78.13 77.76 84.80 76.66 79.74 78.96 84.20 76.46 79.87 pH 0 - 10cm 6.12 6.20 5.76 6.02 6.08 6.30 5.94 6.12 5.99 6.21 5.95 6.05 6.04 10 - 20cm 6.06 6.06 5.86 5.99 5.96 6.23 5.95 6.05 6.06 6.05 5.93 6.01 Both Depths 6.09 6.13 5.81 6.03 6.02 6.27 5.95 6.07 6.03 6.13 5.94 6.01 5 N 0 - 10cm 0.12 0.10 0.11 0.11 0.13 0.12 0.12 0.12 0.10 0.10 0.11 0.10 0.11 10 - 20cm 0.09 0.09 0.12 0.10 0.11 0.11 0.14 0.12 0.10 0.09 0.14 0.11 Both Depths 0.11 0.09 0.12 0.11 0.12 0.11 0.13 0.12 0.10 0.1 0.13 0.11 6 P 0 - 10cm 73.06 32.70 25.33 43.70 62.37 56.30 26.40 48.36 59.88 26.22 33.83 39.98 39.69 10 - 20cm 47.24 28.16 25.47 33.62 66.77 32.50 25.81 41.69 50.71 15.91 25.81 30.82 Both Depths 60.15 30.43 25.50 38.66 64.57 44.40 26.11 45.02 55.30 21.06 29.83 35.50 7 K 0 - 10cm 0.09 0.06 0.12 0.09 0.10 0.08 0.14 0.11 0.09 0.08 0.18 0.12 0.10 10 - 20cm 0.07 0.05 0.13 0.08 0.07 0.06 0.14 0.09 0.08 0.07 0.13 0.09 Both Depths 0.08 0.06 0.13 0.09 0.09 0.07 0.14 0.10 0.09 0.07 0.16 0.11 156 same can be said for OM, where the s ample average (2.24) differed slightly from the regional average (2.10%) (Kamanga, 2011). Intercropping assumed the highest percentage as compared to mulching (2.11%) and tillage (2.17%). The higher average among intercropping practices may be due to the a pplication of both maize and legume residues. Chemical composition varied little as well. All practices were found to have slightly acidic more alkaline than soils cultivated without legumes. Overtime, maize cultivation and fertilizer application practices may have decreased the pH, if only slightly, across years. Percent nitrogen s, sampling depths and lengths of practice. The regional average of soil - N (0.06%) explains that nitrogen, while higher as a sample average, is considered as one of the most limiting factors to production among smallholders (Snapp et al., 2014). More noti ceable differences were found between soil - P and treatments. For example, intercropping illustrated a 16 - 22% difference in available phosphorous compared to its counterparts. Higher levels may be attributed to the liberation of phosphorous in the legume rh izosphere and increased microbial biomass (Tang et al., 2014). Attributing higher phosphorous availability between treatments should be cautioned however, as the sample average (39.69 ug/g ) was larger than the regional average (24.2 ug/g) (Snapp et al., 19 98). In addition, some treatment plots were not located in a single field, but separate fields. Dissimilar soil - P levels, therefore, may be attributed to field variability. The physical and macronutrient soil properties provide optimal conditions for S. as iatica germination and attachment being that 157 soil is slightly acidic and has low CEC (as evidenced by low OC/OM). Under such conditions, it is difficult for maize to assimilate what little micronutrients are available, obligating the crop to release high c oncentrations of strigolactones (Gebreslasie et al., 2018). 3. 4.2 Emergence Very few Striga successfully emerged from pots in the greenhouse (see Table 1 7 ). No statistical inferences could be made from the small number of observations across practices, depth of sampling or length of practice. The highest frequency was found among practices that excluded mulching and intercropping. In addition, more emergence was observed at shallower soil depths. These obse rvations concur with previous studies suggesting that legumes, minimum tillage and mulching are associated with lower emergence rates when compared to continuous maize cultivation and tillage practices (Thierfelder et al., 2015). Table 17 S. asiatica emergence by practice, length of implementation and soil depth Failed emergence may have been attributed to a number of factors. Germination and attachment may have occurred, but hosts may have not been healthy enough to support one or more parasites. The health of maize plants was most likely compromised by repeated exposure to maximum temperatures (>35C o ) in the greenhouse (see Figure 1 3 ). The ideal, or rather, base Continuous Maize + Minimum Tillage + Mulch Maize - Cowpea Intercrop + Minimum Tillage + Mulch Continuous Maize + Tillage All Practices 0 - 4 Years 4 - 7 Years >7 Years All Years 0 - 4 Years 4 - 7 Years >7 Years All Years 0 - 4 Years 4 - 7 Years >7 Years All Years TOTAL Striga Counts 0 - 10cm 3 0 0 3 3 0 0 3 2 2 0 4 10 10 - 20cm 2 1 1 3 0 0 0 0 2 2 1 4 7 Both Depths 4 1 1 6 3 0 0 3 3 4 1 8 17 158 temperature of maize is approximately 26 o C (Sánchez et al., 2014). Exceeding base temperatures, shortened the thermal time it took maize to reach full maturity. Figure 13 - Greenhouse average daily temperature during experiment Plants progress through different phenological stages (e.g., vegetative, reproductive) based on degree - days or heat units, as opposed to calendar - days or hour units (Ritchie & NeSmith, 1991). Their lifecycle or progression of through phenological stages are closely related to their thermal environment. The photoperiod is also used to modify thermal time. When plants are exposed to higher base temperatures under adequate soil - moisture and nutrient co nditions, they progress through various vegetative stages at the expense of biomass production. With little biomass, organs (e.g., roots) are not adequately developed to support critical phenological stages (e.g., grain filling) , and in the case of this ex periment, expire prematurely (Tebaldi & Lobell, 2018). Dec 1 st 2017 Jan 1 st 2018 Jan 15 th 2018 Feb 1 st 2018 159 In the greenhouse experiment, maize had reached the VT - R1 stage in week 7 (44 - 49 days), which is 2 - 3 weeks earlier than normal. Hastened growth may have jeopardized the emergence of S. asiatica in sev eral ways. First, dwarfed root systems may have not reached certain areas of the soil where seeds were present. Observation revealed many of plants did not have a fully developed root system that reached the wider and lower soil depths of th eir 6 - liter pot s. Second, moisture is critical at the R1 stage. Even with daily water application, heat stress and transpiration during this critical period may have caused the plants to die . Increased transpiration from leaves may have left little water to secrete leach ates to trigger germination and/or supply a vigorous root system to support parasites. Third, pots were not perforated to prevent seed loss. Without drainage, fertilizer losses (with ample phosphorous) were probably minimal, leaving a pool for maize to ass imilate phosphorous from. With this available pool, maize may have secreted little strigolactones to signal fungi to assist with the assimilation of phosphorus . Minimal secretion of strigolactones may have hindered S. asiatica germination. 3. 4.3 Mod el runs 3. 4.3 .1 Sensitivity analysis The behavior of the model will first be explained absent of any control prior to assessing its sensitivity to variations in individual parameters. In Figure 14 (a) , there is an initial decrease in surface seeds (0 - 10 cm) in month 11/23/35/47 given that 50% of the surface level seed population is circulated via ridging in November to the subsurface seed stock. After which, a second reduction occurs in the surface seed population whereby seeds germinate from month 12/24/36/48/ 60 to 16/28/40/52/64. Then, there is a sharp spike because seeds are being dropped by mature flowers. In the offseason (month 17/29/41/53 to 23/35/47/59), there is a 160 slow increase because seeds are still settling from transpiring flowers. Across 60 months, the surface level seed population quickly reaches a peak, leveling out at 700 million seeds. The subsurface seed (0 - 10 cm) population increases in a stepwise manner, spiking in months 11/23/35/47 from the transfer of seeds from 0 - 10 cm depth via ridging. Between months 12/24/36/48/60 and 23/35/47/59 there is marginal decline in subsurface seeds due to decay and predation. Figure 14 - Base cas e runs: seedbank (a), germination (b ), a ttachment - emergence - flowering (c ) a b c *Note: The Y - axis of graph A is scaled in millions. Graph B and C are scaled in thousands In Figure 14 (b and c) , between months 12/24/36/48/60 and 16/28/40/52/64, seeds germinate, attach to the host, emerge from the soil and flower. In Figure 14 (b) , the model demonstrates an increasing S - curve, whereby the germinating seed population reaches a maximum population of 950,000. In Figure 14 ( c ) , the large number of germinating seeds occupies any and 161 all available attachments; hence max attachment (775,00 0), emergence (425,000) and mature flowering (50,000) populations are reached in the first cropping season and persist into the proceeding seasons. Base case runs demonstrate how quickly a 1 ha field can become inundated with S. asiatica seed, seedlings an d flowers. With no controls, max saturation in almost all stocks is reached in 65 months. The behavior of the model (e.g., exponential growth) reflects what practitioners may observe in the subsurface seed and unattached seedling stocks across a 1 ha Malaw ian smallholder field if no controls are implemented (Ejeta, 2007). One result which validates the model is that max flowering reaches approximately 70,000 flowers. Being that there are 54,000 hosts ( which can potentially support 918,000 parasites ) , one ma y expect more than 1 - 2 flower per maize plant (70,000 flowers/ 54,000 maize plants = 1.30 flowers per maize plant) . This was observed in a field cultivated by one of the participants of the study. 3. 4.1 .3 .1 Germination rate In the base case run, the mont hly germination rate ranged between 60% to 80% based on monthly rainfall and temperature. For the sensitivity analysis, the germination rate was altered between 40% to 80%. As the germination rate increased by 25%, the CSM only reflected noticeable changes in surface/subsurface seed stocks. This finding validates the behavior the CSM based on earlier studies which found varying germination rates affected the seedbank, but not necessarily attachment (Vallance, 1950). The CSM did not reflect any moderations i n attachment, emergence or flowering if the germination rate was altered. The CSM did demonstrate, however, more noticeable differences between unattached seedlings with 162 varying germination rates (see Figure 15 ). Each 25% increment in germination (e.g., 40% to 60%) reflected approximately a 17% increase in germination (e.g., 500,000 to 600,000). Figure 15 - Monthly unattached seedlings in association with varying germination rates *Note: The Y - axis of graph is scaled in thousands 3. 4.1 .3 .2 Attachment rate In the base case run, the monthly attachment rate ranged between 13 and 17 parasites per maize plant based on the basal application of N - based fertilizer. For the sensitivity anal ysis, the attachment rate was altered between 7 and 17 parasites per maize plant. As the attachment rate was altered, noticeable changes were seen across all stocks, particularly the flowering (see Figure 16 ). The flowering population associated with 17 at tachments reached 70,000 by the second season and sustained this population into the following seasons. The flowering populations associated with 7 and 12 attachment rates were sustained at 9,000 and 29,000, respectively. Based on these rates, there was ap proximately an 86% difference between the lowest and highest attachment rates, indicating a strong reduction in seeds dropped by flowers, unattached seedlings, and consequently, emerged seedlings. It appears then, monthly attachment rate is one of the (if not the most) sensitive parameters of the CSM. Previous studies validate this behavior, arguing that Striga spp . resistant varieties drastically reduce the parasite seedbank via reduced attachment within three seasons (Westerman et al., 2007) . 163 Figure 16 - Monthly flowers in association with varying attachment rates *Note: The Y - axis of graph is scaled in thousands 3. 4.1 .3 .3 Emergence and flowering rate In the base case run, the successful emergence fraction (i.e., monthly emergence rate) did not change from its base value (70%) because no other parameters were attached to this fraction. For the sensitivity analysis, the germination rate was altered between 34% and 70%. A 25% decrease in the emergence fraction equated to a 33% decrease in the surface and subsurface seed populations (see Figure 17 ). Similar behavior of these stocks was seen when the flowering rate decreased by 25% (e.g., 25%, 38%, 50% flowering rate). No changes were reflected in attached seedlings or mature Striga in the proceeding seasons when emergence varied. When flowering rat es were varied between 25% and 50%, attachment remained at the same population as its base case (see Figure 18 ). Attachment did not fluctuate due to the inundation of unattached seedlings occupying potential attachment sites. Even with little emergence or flowering, the fecundity of S. asiatica (34,000 seeds per flower) provided over one million seeds with just 30 flowers. The behavior in the CSM concurs with previous studies that contend reductions in seedling emergence or flowering rates are ineffective i n controlling S. asiatica due to its overwhelming fecundity (Khan et al., 2002). 164 Figure 17 - Monthly surface seeds in association with varying successful emergence fractions Figure 18 - Monthly attachments in association with varying flowering success fractions *Note: The Y - axis of graph A is scaled in millions. Graph B is scaled in thousands 3. 4.4 .2 Scenario analysis Practices were individually run in the CSM to highlight any notable behavior at certain lifecycle stages across time. Identifying this behavior informed which practices to combine and how their aggregation addressed emergence (if at all) in a specific timespan. Weeding and manure application are some of the most wi dely implemented control practices implemented by smallholders in the region (Orr et al., 2002). Legume intercropping or mulching (e.g., conservation agriculture) are relatively newer and costlier (e.g., additional seed cost) when compared to the aforement ioned practices ( Giller et al., 2009 ). The combination of new and traditional practices ran in the CSM were done so based on their practicality to smallholder settings (e.g., labor constraints, financial scarcity, compliments already - implemented practices) . 165 3. 4.4 .2.1 Weeding Weeding in January and March illustrated a sharp decline in the mature Striga , yet the population remained at approximately 25,000 in the CSM (see Figure 19 [a] ). The sharp decline does little to re - emerging seedling and mature flower populations. To elaborate, the slope of mature Striga in a weeding scenario was similar to the slope of the base case, illustrating that timely weeding was ineffective to controlling S. asiatica emergence. This may be attributed to S. asiatica emerg ing in areas where smallholders began weeding before they finish weeding an entire 1 ha field. Sharp declines in the Mature Striga stock led to minor dips in the surface seed population three months after weeding months (see Figure 19 [b] ). There were notic eable differences in surface seedbank populations between the base case and weeding runs (350 million vs 590 million), but with a high seedbank feeding into the unattached seedling stock, the attached seedling population is unaffected by weeding. Disconcer tingly, this is the stage maize yield is negatively affected the most by parasitism (Frost et al., 1997). The finding agrees with the argument that timely weeding throughout the season is ineffective in reducing Striga spp . emergence (Joel, 2000). Figure 19 - Reductions in monthly Mature Striga (a) and surface seed (b) in response to weeding a - > b *Note: The Y - axis of graph A is scaled in thousands. Graph B is scaled in millions 166 3. 4.4 .2.2 Manure application The CSM demonstrated that manure application delayed the onset of germination. As evidenced by multiple studies, basal applications of well - decomposed manure (which at that point is considered compost) can delay germination when applied at planting (Yoneya ma et al., 2007; Sherif, 1986) . Soil - P availability is contingent upon several conditions (e.g., pH, soil moisture), but the CSM demonstrated a delay in germination by one month, reducing the unattached seedling population marginally (850,000 vs 800,000) ( see Figure 20 ). Soil - P pools that remain longer around the root canopy (e.g., January) can decrease window of germination (e.g., 4 months vs 3 months), equating to more evident reductions in the unattached seedling population. Larger applications can achie ve this objective, but damage maize seedlings if the manure is not fully decomposed (i.e., nutrient burning) or applied under dry conditions ( Materechera, 2010 ). This finding suggests the repeated application of manure could significantly reduce emergence, which seems to be little studied. Smallholders may not have access to an extensive amount of well - decomposed manure, or the labor to repeatedly apply the input at an effective rate (Schulz et al., 2003). Figure 20 - Monthly unatta ched seedlings in response to basal manure application 167 3. 4.4 .2.3 Intercropping The CSM indicated legume intercrops were associated with reductions in the S. asiatica seedbank, unattached seedlings and mature flowers as a result of suicidal germination, reduced haustorium growth and shading, respectively (Sanginga et al., 2003; Oswald et al., 2002). In Figure 2 1 (a) , the surface seed stock did not accumulate into Februar y (month 14/26/38/50/62), but rather, abruptly decreased as additional germination was triggered by legume leachates, countering an increase in the population from flowering. In the offseason (month 17/29/41/53 to 23/35/47/59), there was a less steep decli ne in surface seeds, but a decline none the less, due to continued suicidal germination from decomposing legume residues. In Figure 2 1 (b) , an influx of unattached seedlings with truncated haustoria did not lead to a reduction in attachment or emergence. Certain legumes (e.g., Desmodium spp .) have shown to significantly reduce Striga spp . seedbank and its related emergence in several seasons (Khan et al., 2002). The reduction was not only attributed to suicidal germination, reduced haustoria growth, but also soil - N contributions. The CSM demonstrated with cowpea could not decrease emergence given that little soil - N was contributed. In compari son, cowpea is a poor N - fixer relative to Desmodium spp. (Piha & Munns, 1987). Had there been more N contributions, perhaps the CSM would have indicated reduced attachment and consequent emergence in this scenario. Shading from the legume canopy did reduce mature flowering population slightly ( Figure 2 1 [c] ). Reductions via legume canopy is noted in the literature (Kureh et al., 2006). In a separate scenario, mulching demonstrated the same effect legumes had on the mature Striga stock. The CSM demonstrated t hat legumes do not necessarily reduce attachment, but reduce S. asiatica emergence by keeping the seed bank and flowering population below a certain 168 threshold. The finding implies maize - cowpea intercropping alone will not successfully reduce S. asiatica em ergence in the long - term. Figure 21 - Monthly seedbank (a), unattached seedlings (b) and mature Striga (c ) reductions in response to cowpea intercropping a - > b - > C *Note: The Y - axis of graph A is scaled in millions. Graph B and C are scaled in thousands 3. 4.4 .2.3 Fertilizer Application The CSM illustrated a stark reduction in attachments (see Figure 2 2 [a] ) in response to a basal fertilizer application at planting, equating to reductions in emerged/mature Striga ( Figure 2 2 [b] ) and the seedbank ( Figure 2 2 [c] ). The reduction in successful attachment corroborated findings whereby N - based fertilizer application s were associated with >40% reductions in S. asiatica emergence (Dugje et al., 2006). Like legume - intercropping, the CSM demonstrated fertilizer applications reduced flowering and seed stock population to a point, but not enough to eradicate the weed in a 5 - year timespan. Seedbank behavior did not demonstrate a decreasing trend in the off - season as evidenced by the previous intercropping scenario. Fertilizer application reduced emergent seedling and mature S. asiatica rates by approximately 169 96% (50,000 vs 4 25,000) and 97% (2,000 vs 70,000), respectively, relative to the base case. These notable changes were attributed to the voluminous starting seedbank. Had it been smaller, the changes would not have been noteworthy. Previous studies argue the effect of a c ontrol practice, such as N - applications, is more evident in fields with higher infestations (Carson, 1989). Figure 22 - Monthly attachment (a ), seedl ing/flowering (b) and seedbank (c ) reductions in response to N - based basal fertilizer application at planting a - > b - > c *Note: The Y - axis of graph A and B are scaled in thousands. Graph C is scaled in millions 3. 4.4 .2.6 Aggregated practices The application of manure at planting in coordination with weeding lowered the surface seed bank to a threshold (375 million) (see Figure 2 3 [a] ). The decrease had no effect on attachment (see Figure 2 3 [b] ). The delay in germination ( Figure 2 3 [a] ), deduction of emerged seedlings and removal of flowers twice did little to decrease emergence. The finding suggests two weedings and one basal application of manure at planting will not address attachment and subsequent 170 seedling emergence. The finding disagrees with previous findings suggesting this strategy will control Striga spp . within 3 - 4 seasons (Oswald, 2005). The behavior of the CSM illustrated weeding and manure application are either not addressing enough stages in weed lifecycle or not substantially decreasing stocks to reduce emergence. This control strategy is typically implemented by smallholders who cannot afford additional inputs, and without them , the weed will likely persist . The model was not parameteriz ed to simulate the S. asiatica lifecycle , not maize yield, but if it was, we would expect 30 - 100% losses across 3 - 5 growing seasons ( Oswald, 2005 ). Figure 23 - Monthly surface seed (a) and attachment (b ) reductions in response to weeding + manure application a - > b *Note: The Y - axis of graph A is scaled in millions. Graph B is scaled in thousands The addition of intercropping or fertilizer to traditional smallholder practices in the CSM illustrated different outcomes. In Figure 2 4 ( a ) , the addition of intercropping to traditional practices kept subsurface seeds under a 175 million thresho ld. The addition of fertilizer application kept subsurface seeds under a 20 million threshold. While there were stark differences between the two control strategies, 20 million seeds feeding into the unattached seedling stock of the CSM allowed approximate ly 40,000 attachments. In Figure 2 4 ( b ) , fertilizer application, appeared to be a more effective addition to manure application and weeding as 171 opposed to intercropping. Still, several thousand flowers survived in response to manure application, weeding and fertilizer application. This finding indicates fertilizer appears more effective in managing emergence, but the survival of flowers allowed seeds to be dropped, germinate and attach to maize. By allowing monthly attachment to reach a small threshold each s eason, even just several thousand across one hectare, S. asiatica still remains a significant threat to maize yield (Kim et al., 2002). Figure 24 - Monthly surface seed (a) and mature Striga (b ) reductions in response to weeding + manure application + legume intercropping VS weeding + manure application + fertilizer application a - > b *Note: The Y - axis of graph A is scaled in millions. Graph B is scaled in thousands The combination of intercropping, fertilizer application, manure application and weeding in the CSM denoted an eradication of weed across five seasons (see Figure 2 5 ). Unlike previous scenarios with different combinations of practices, the aforementioned s trategy does not permit any stage of the lifecycle to return to a threshold in one season. Instead, surface seeds and subsurface seeds ( Figure 2 5 [a] ), attachment ( Figure 2 5 [b] ) and mature Striga ( Figure 2 5 [c] ) illustrated a consistent downward trend across consecutive seasons in the CSM. The dynamic behavior of the CSM highlights how all major stocks must decrease in a coordinated effort to decrease emergence. Permitting one stock (e.g., mature Striga) to return to the same equilibrium each season, allowed weed emergence to persist. This finding agrees with literature 172 which argues the weed must be managed at all stages of the lifecycle to address its persistent seedbank and emergence (Westerman et al., 2007; Joel et al., 2000). The finding also argues under the specified average seedbank and controls, conservatively, S. asiatica will not be eradicated unless all practices are employed in a consistent manner across five seasons. That is not to say the weed will devastate yields if it is not completely eradicat ed, but will continue to be a significant pest until it is lowered to particular threshold. Figure 25 - Monthly surface seed (a), attachment (b ) a nd flowering (c ) reductions in response to weeding + manure application + legume intercropping VS weeding + manure application + fertilizer application a - > b - > c *Note: The Y - axis of graph A is scaled in millions. Graph B and C are scaled in thousands 3.5 Conclusions The following study developed a cropping systems model to simulate the dynamic behavior within the S. asiatica lifecycle under various smallholder management strategies. The CSM differed from previous models in that it was adapted to smallholder settings (e.g., Striga seed density, marginalized soils) and practices (e.g., ridging). In addition, the model evaluated the efficacy of practices smallholders were already using or strategies which could be adapted to 173 traditional practices and farming systems. Many models have evaluated practices that may be effective in addressing Striga spp . emergence (e.g., transplantin g of maize seedling, no till), but such practices are unlikely to be implemented in Malawi given their labor demands and/or incompatibility with smallholder farming systems (Giller et al., 2009; Orr et al., 2002). Apart from these practice considerations, the CSM also included a parameter that was either absent or limited in previous models - attachment. As evidenced in this CSM, attachment was quintessential to evaluating emergence in response to one practice or many. Without attachment, a parasitic weed ca nnot survive. The CSM highlighted several critical findings for Striga spp . researchers and practitioners. Without the use of hybrid seed (e.g., Striga - resistant maize), an aggregated effort must be made to address emergence (Grenz et al., 2005). The CSM h ighlighted the dynamic nature of S. asiatica . In allowing one stage to return to a certain threshold, such as attachment, emergence and flowering will persist. Even though attachment was the most sensitive parameter in the CSM, as evidenced by model behavi or, all stages must be addressed to eradicate the weed. In addition to these findings, the CSM suggests smallholders will likely need to modify traditional practices to address S. asiatica emergence (i.e., manure application and weeding). Highly infested f ields are likely to be colonized without the addition of fertilizer and intercropping or Striga resistant seed. While the combination fertilizer application and traditional practices appeared to be more effective than intercropping, induced suicidal germin ation and shading provided by legumes completed the suppression of emergence. There are human and financial tradeoffs for implementing all practices (if possible) (Oswald, 2005). These tradeoffs must be 174 considered when developing and disseminating new stra tegies to address S. asiatica emergence across Malawi. There were several limitations to the study and its subsequent findings. Given what little emergence occurred in the greenhouse trial, the model could not be validated from emergent and flowering rates observed across different practices and periods of implementation. Model behavior was largely validated by emergent and flowering studies found in the literature. In addition, there was no stock to account for N - accumulation contributed by N - fixation and N - residues via legume intercropping. These contributions can affect cell wall thickness and associated attachment ( Cechin & Press, 1993) . Furthermore, the CSM does not account for the Striga spp . seeds are in the soil (i.e., a sink), a lower percentage can successfully attach because there are less resources (e.g., healthy maize roots) available to support their growth. When this occurs, emergence and flower development are decelerated as compare d to a field with less sinks (Hearne, 2009). The last limitation of the model is that several parameters (e.g., mulching) were informed by West African studies investigating S. hermonthica emergence. While both weeds are similar, their fecundity and emerge nt rates are different. These four limitations warrant the need to study maize parasitism in southern Africa further. A replication of the trials conducted in the study will assist in calibrating CSM and validating outputs. Replicated greenhouse experimen ts must take into consideration the methodological errors observed in this study. No S . asiatica had flowered and very few seedlings had emerged in the experiment. Germination and attachment were not monitored during the experiment until the plants were di sposed in the compost pit once after four months. Maize transitioned through its 175 vegetative phases rapidly in the greenhouse , affecting one, if not more, of the growth stages of S . asiatica . Hastened growth was most likely attributed to maize being exposed to extended hours of heat above the outside temperature during the day and evening (Fanadzo et al., 2010) . Hastened growth , first, was responsible for retard ing root growth s , shortening the window of strigolactone secretion which trigger s S. asiatica germination. Second, s maller root canopies reduces the likelihood a root will access an area where S. asiatica are, reducing their chances to germinate and attach later . Another factor to consider is that fertilizer (primarily N) was applied. The application is associated with thicker root tissues and prevented the germ tube of the parasites to access the phloem. A dult leaves from V5 - VT exemplified a dark green pigment , illustrating they were not N - deficient and most likely had a thick enough root system to reduce the rate of successful attachment. Rather limited soil - P may have been the primary culprit. The macronutrient is responsible for reduced root volume, and consequently, l ess germination (Yao et al., 2007). A combination of low germination r ates and f ew successful attachment s will ultimately lead to fewer observations of emergence and flowering. Findings from the study, specifically model behavior, have several implications for practitioners and researchers in S. asiatica emergence in Malawi an smallholder systems. Given the extent attachment has on the system, delaying its onset with repeated micro - dosing of manure at planting stations could assist the financially - constrained farmer. In addition, modifying planting station depth (e.g., poking holes in ridges and sowing >30 cm) provides an economic avenue to controlling the weed (in addition to other practices) ( Elzein & Kroschel, 2003) . Being that fertilizer addressed the attachment stage in the CSM, other avenues, such as coating seeds with herbicide may be a more economic avenue if fertilizer is more expensive than herbicide 176 ( Kanampiu et al., 2003) . These practices must be evaluated f urther under smallholder conditions, which can be difficult to control from outside factors. Attachment is difficult to monitor in experiments given its sporadic nature. Perhaps one reason why a disproportionate amount of practices that address emergence and flowering have been disseminated to smallholders is due to the fact that emergent counts and seed rates are relatively more feasible to quantify than attachment. By not modeling the dynamic behavior of this stage or parasitic weeds in general, research will be limited in informing effective policy to address this ever - growing agricultural issue. Given the devastating effects Striga spp . has had in Malawi and sub - Saharan Africa, it is imperative to develop parasitic weed modules for crop models to better evaluate smallholder practices (Ejeta, 2007). Models that do not capture underlying biological mechanisms presented in the CSM, risk informing extension with potentially misleading information regarding the efficacy of control practices. 177 APPENDICES 178 APPENDICES APPENDIX 1. Justification off Values/Equations Applied to Parameters of Cropping System Model Parameter Value or Equation Explanation Source Initial Values Surface Seeds (iSs) 15 million (292,511,700.47) 1 12 - 17.9 seeds per 100g of soil. Extrapolate density based on bulk density of seed Baskin & Baskin, 1998; Hartman & Tanimonure, 1991; Ngwira et al., 2013; Visser & Wentzel, 1980 Subsurface Seeds (iSSs) 7.5 million Half the amount of Surface Seeds due to ridging. See source/s of Surface Seeds (iSs) Unattached Seedlings (iUs) 0 No seedlings germinate before a host is present N/A Attached Seedlings (iAs) 0 No seedlings can attach before seeds germinate N/A Emerged Seedlings (iEs) 0 No seedlings can emerge until underground seedlings have attached N/A Mature Striga (iMs) 0 No Striga has emerged yet N/A Host Plants (iHP) 0 Land has not been prepared yet for sowing N/A Intercrop Plants (iIP) 0 Land has not been prepared yet for sowing N/A Stocks Surface Seeds (monthly production of viable seeds +seed resurfacing) - (monthly burial + regular germination + suicidal germination + surface monthly mortality) Seeds are added to the soil surface from flowers dropping seed or farmers bringing seed to 0 - 15cm from ridging. Surface seeds decrease as they germinate, becoming underground seedlings. Also, seeds can transpire via suicidal germination, predation and deca y. Babiker et al. (1987); Behawi et al. (1984); Musambasi et al. (2002) Subsurface Seeds monthly burial (seed resurfacing + subsurface monthly mortality) The dormant seed population (below 15cm of soil) is increased from seeds buried during ridging. The population decreases as seeds decay or brought to the top of the ridge (0 - 15cm) during land preparation Behawi et al. (1984); Benvenuti (2007) Unattached Seedlings germination in thousands - (seedlings The surface seedbank provides a population of unattached Jamil et al. (2012); Kunisch et al. 179 removal + monthly attachments + Unable to attach) seedlings which can either attach to the maize root system (triggered by strigolactones leached by hosts). The unattached seedlings population decreases from those that successfully or unsuccessfully attach. Unsuccessful attachment (i.e., seedling removal) can be attributed to the shortening of the haustoria by legumes, preventing timely attachment as reserves are depleted (1991); Makoi & Ndakidemi (2012) Attached Seedlings monthly attachments - (attached seedlings removed + Attached without sequestering photosynthates) More seedlings are able to attach monthly as the maize root system develops and enters new areas of the soil. Approximately four weeks after attachme nt, seedlings emerge from the soil and are no longer considered as attached seedlings. A percentage of attached seedlings never emerge from the soil given that they cannot successfully penetrate the cell wall of the maize root systems and siphon nutrients. A portion that do successfully siphon nutrients transpire before emerging from the soil. Makoi & Ndakidemi (2012) Emerged Seedling monthly emergence - (monthly mortality emerged seedlings + emergent weed removal + monthly maturation + emergent weeding) Att ached seedlings that emerge from the soil add to the population. As emerged Striga matures into flowering adults, are weeded or die due to unfavorable conditions, the population decreases. Cechin & Press (1993) Mature Striga monthly maturation - (mature weed removal + monthly mortality mature plants + mature weeding) Emerged seedlings add to the mature Striga population. The population decreases as flowers are removed or transpire before developing seed. After April, no mature Striga is present because there is no longer maize in the field Kabambe et al. (2008) Host Plants in Field host sowing - host removal 53,000 maize plants are sown late November which are all expected to germinate and mature. In April, the entire population is harv ested and Kabambe et al. (2008) 180 removed from the field via burning or livestock Intercrop Plants in Field Planting the Legumes - intercrop removal 27,000 maize plants are sown late February which are all expected to germinate and mature. In May, the entire population is harvested and removed from the field via burning or livestock Kermah et al. (2017); Ngwira et al. (2013) Flows Connected to Surface and/or Subsurface Seeds Stocks Monthly Production of Viable Seeds Monthly seed production * Seed viability fraction surface from all the mature Striga plants in the field See source/s of Germination in Millions potential germination*(1 - suicidal fraction) Under speci fic soil conditions (e.g., pH, texture) surface seeds germinate which become underground seedlings. A portion of these seedlings are removed IF legumes are planted and induce germination without a host to attach to (i.e., suicidal germination) See source/ s of Seed Resurfacing If Ridges were made THEN Seeds resurfaced from burying ridges*Subsurface Seeds. OTHERWISE seeds do not resurface As farmers create ridges, they transfer subsurface seeds to a depth where they are able to germinate See source/s of from burying Surface Monthly Mortality surface mortality rate*Surface Seeds Seeds from the previous seasons that do not germinate decay over time See source/ s of Monthly Burial If Ridges were made THEN burial fraction*Surface Seeds OTHERWISE a portion of the surface seeds are not transferred to a lower depth IF ridges are made, farmers commonly pull the top of the ridge from the previous season into the furrow, circulating bottom soil to the top of the ridge. In the process, surface seeds that did not germinate are moved to the bottom layer of the soil See source/s of Subs urface Monthly Mortality subsurface mortality rate*Subsurface Seeds Seeds from the previous seasons that were not circulated to the top layer for possible germination, decay over time See source/s of subsurface 181 Surf ace Suicidal Germination Surface Seeds*(suicidal fraction*(1 - potential germination))/TIME STEP Under specific soil conditions (e.g., pH, texture) surface seeds germinate if legumes are present. Different legumes induce different percentages of germination See source/s of Connected to Unattached and/or Attached Seedlings Stocks Germination in thousands If Host Plants are present THEN germination in millions*units conversion OTHERWISE seeds do not germinate Converts millions of surface level seeds to thousands of seeds germinating if Mai ze is in the field See source/s of Seedlings removal If Host Plants are present THEN Unattached Seedlings*Maize Removal OTHERWISE all underground seedlings are removed Once maize is removed in May, all unattached seedlings will transpire because no host is supporting their development See source/s of Monthly attachment s If N - Fert Application=1 THEN MIN(indicated attachments from vacant sites, indicated attachments from exposed unattached seedlings)*.70 OTHERWISE attachment is not reduced by 30% (Attachment only occurs only during December - April) Maize can only support a certain number of parasites. Each month, as the root system develops, a certain numbe r of sites are available for attachment, reaching a threshold in April. When N - based fertilizer (33kg) is applied (basal), the maize root system can develop a thicker cell wall around the phloem, reducing successful attachment. Cechin & Press (1993); Mume ra & Below (1993) See source/s of attachments from vacant sites & attachments from exposed unattached Unable to attach (Haustorium Attachment Factor*Unattached Seedlings)/TIME STEP In the presence of legumes, a percentage of germinated seedlings will unsuccessfully develop haustoria each month Ejeta 2001; Serghini et al. 2001; Tsanuo et al. 2003 See source/s of Attachment Attached without sequesterin g photo - synthates IF THEN ELSE(""=1,(Attached Seedlings*Root Cell Wall Factor)/TIME STEP,0) A lower percentage of germinated seeds can break the cell wall of a maize root system which has a high N - concentration. Higher concentrations develop thicker cell walls, red ucing the ability of Jamil et al., 2012; Chechin & Price, 1993 182 breakdown the cell wall and access photosynthates Connected to Emerged Seedlings and/or Mature Striga Stocks Monthly emergence Attached Seedlings*successful emergenc e fraction The number of seedlings that emerge from the soil is based on the number of underground seedlings that have successfully attached See source/s of Monthly mortality of emerged seedlings Emerged Seedlings*monthly emergents mortality rate A certain number of emerged seedlings are not supported well enough by maize, transpire and do not develop into mature flowers See source/s of Emerged emergents Emergent weed removal When the 1 st weeding takes place and/or maize is removed THEN Striga removal fraction*Emerged Seedlings Once maize is removed in May, all emerged seedlings are removed from the field because they die since maize is n o longer present See source/s of Striga removal Emergent Weeding IF THEN ELSE(month of year="1st Weed Month":AND:"<1st Weeding?>"=1:OR:mo nth of year="2nd Weed Month":AND:"<2nd Weeding?>"=1,(0.9*E merged Seedlings),0)/TIME STEP Emergent seedlings can be removed 2 weeks and/or 6 weeks after sowing See source/s of st nd Monthly maturation potential monthly survivals*flowering success fraction After seedlings emerge from the soil, a certain percentage mature in to flowers after a specified time period See source/s of potential monthly Mature weed removal When the 2 nd weeding takes place and/or maize is removed THEN Striga removal fraction*Mature Striga*Maize Removal Once maize is removed in May, all mature plants transpire because there is no host is supporting their development See source/s of Striga removal Mature Weeding IF THEN ELSE(month of year="1st Weed Month":AND:"<1st Weeding?>"=1:OR:mo nth of year="2nd Weed Month":AND:"<2nd Weeding?>"=1,(0.9*M ature Striga),0)/TIME STEP Emergent seedlings can be removed 2 weeks and/or 6 weeks after sowing See source/s of st nd 183 Monthly mortality of mature plants Mature Striga/Striga maturation interval A certain percentage of mature flowers do not survive to produce seed. See source/s of Connected to Host Plants Stock Planting the maize In the month of December Thousands of plants per hectare*Planting The specified sowing density across a hectare of land on the specified date See source/s of Thousands of Host removal In the month of April Host Plants*Maize Removal The specified date when all maize is removed from the field See source/s of Connected to Intercrop Plants Stock Planting the Legumes (Thousands of Leg plants per hectare*Planting Legumes/TIME STEP)*Decision to Intercrop The specified plant population (27,000) relay intercropped in January (one month after sowing maize). Relay intercropping with cowpea is a common practice in Malawi Ngwira et al. (2013); Silberg et al. (2017) Intercrop Removal Intercrop Plants*Legume Removal/TIME STEP The specified date of removal in June (one month after maize removal) Ngwira et al. (2013) Converters/Factors Attached to Flows Connected to Surface monthly mortality outflow Surface mortality rate 0.02 Seed loss predated by fungus, bacteria and/or organisms (e.g., earthworms). Note The microscopic size of seeds presents difficulties to assess the exact percent predated; thus, a percent was created to not di scount this factor in seed loss Ciotola et al. (1995); Lendzemo et al. (2006)n Connected to Monthly production of viable seeds inflow Monthly seed production Mature Striga*Monthly seeds produced per plant Total seed production is calculated by multiplying the amount of emerged flowers by the number of seeds produced by one flower See source/s of produced per Seed viability fraction 0.325 32.5% of the seeds produced by one mature Striga asiatica flower have the capacity to germinate at a 0 - 15cm soil depth. The percent was based on taking the average of the range (20 - 45%) presented in the paper Smith et al. (1993) Connected to Germination in millions & suicidal germination outflows Suicidal fraction suicidal germ ination from leg residue + suicidal germination from living legumes Different rates of suicidal germination occur based on the leachates that are secreted by roots or decomposing foliage (usually assisted by rainfall) See source/s of germination from germination from 184 Potential germinatio n IF THEN ELSE(""=1,(monthly germination rate*("Host Plants Present?" - Delayed Germination)),monthl y germination rate*"Host Plants Present?") Germination is largely driven by soil moisture and soil temperature. Generally, higher germination rates are found in acidic sandy moist soils that have low phosphorous (without manure applied) and receive an average daily temperature of 29 - 33 o C (equating to 24 - 26 o C soil temperature). Lower germination percentages are a result of less favorable conditions. Aflakpui et al. (1998); Bationo & Mokwunye (1991); Jamil et al. (1998); Netzly et al. (1988); Osman et al. (1991) Connected to seed resurfacing & monthly burial outflow/inflow s Bury the ridges 0,1 (User option) A common practice among farmers, also known as ridging Kabambe et al. (2008) Resurface fraction with burying ridges .50 In preparation of growing season, farmers pull the bott om top of the mound, circulating lower level seeds to a higher soil depth (0 - 15cm) Benvenuti (2007) Burial fraction .50 In preparation of growing season, farmers pull the top of the furrow, circulating 50% of the seeds to a lower soil depth (15 - 30cm) Benvenuti (2007) Connected to Subsurface monthly mortality outflow Subsurface mortality rate 0.034 20% of seeds are viable after 4 years but none are viable after 9 years, which corresponds to a viability of ~70%. 20% indicates 80% decay over 4 years corresponding to a 41% decay rate/year. 40%/12 months = 3.42% decay rate/month Behawi et al. (1984) Connected to Germina tion in thousands inflow Units conversion 1000 Multiplies surface seeds by 1000 to converts into units of 1000 N/A IF THEN ELSE(Host Plants>0, 1 , 0 ) If maize is in field, germination occurs, otherwise, germination is not possible N/A Connected Unable to Attach outlfow Unattached Seedlings See stocks for further information See source/s of .48/.70 Perfect attachment allows 70% of seedlings to attach, but in the presence of legumes, 48% can only attach because haustoria is reduce. Ejeta 2001; Oswald et al., 2002 ; Sanginga et al., 2003 ; Serghini et 185 al. 2001; Tsanuo et al. 2003