Conclusiones

The objective of this research is therefore to investigate the role of changes in rainfall variability, as a proxy for changes in climatic variability, in the migration decisions of people in Burkina Faso through the application of an agent-based modelling approach. Although both Warner (2010) and Tacoli (2011) highlight the existence of controversy over whether migration should be deemed an adaptation strategy or an example of failure to adapt in the face of climate change, for the purposes of this thesis migration is deemed one of a range of adaptation options that may be available to individuals affected by climate change. The ABM presented in Chapter 5 has therefore been developed using existing theoretical advances in the fields of both human migration and climate change adaptation. These theoretical foundations are combined with advances in the field of social psychology to develop a conceptual basis for agent cognition within the model. Agents in the modelled environment of Burkina Faso interact with one another and their environment to develop intentions to adapt to changes in rainfall through migration. The likelihood of an agent migrating is affected by both their individual attributes and their placement in a social network within which changes in rainfall are communicated.

Although recurrent reference to the existence of the impact of climate change upon migration within IPCC publications (Tegart, et al., 1990, Wilbanks et al., 2007) suggests relatively widespread acceptance of the notion, further understanding is required in order for policy makers to propose useful adaptation measures that can alleviate the impact of such change upon the livelihoods of vulnerable communities worldwide. As such, the ABM developed by this thesis aims to contribute to the growing body of literature that has emerged surrounding the notion of environmental migration. Although developed with the intention of quantifying the influence of rainfall-related environmental change upon migration within and from Burkina Faso towards 2050, results from the ABM are also intended to provide greater understanding of the multiple and interacting components that contribute to the migration decisions of individuals. As such, the overall impact of changes in rainfall variability upon the national-scale flow of migrants within and from Burkina Faso will be broken down into smaller scale sub flows in order to investigate the trends in behaviour at regional scales that make up the more generalised national trend.

From the proposed analysis of complexity at varying scales, the value of quantifying global environmental migration forecast to occur in the future will be assessed. Previous research by Acosta-Michlik and Espaldon (2008) suggested that very few studies had represented models of human behaviour or cognition in assessing vulnerability to global environmental change.

However, the authors found that social networks were an important source of adaptive capacity while cognition could be said to enable adaptation. The use of an agent-based modelling approach to considering the concept of environmental migration within and from Burkina Faso therefore incorporates both social networks and cognition in an effort to simulate the impact of rainfall variability upon migration decision-making.

1.3 Uncertainty

Inherently ‘wrong’ because by design they are a simplified view of the system being represented, models of any human or natural system suffer at the hands of considerable uncertainty. Such uncertainty is further increased when interactions between human and natural systems are additionally considered. An ABM intended to represent the interactions between human migrants and the climate conditions affecting them in the future therefore offers considerable potential for uncertainty. In any simulation that uses climate data, uncertainty results from the ambiguities in the extent and magnitude of the climate signals input into the model. This uncertainty is increased further when predictions of future climate are used, particularly when the proxy for climate is rainfall which shows considerable variation in both spatial and temporal terms. Furthermore, the basis upon which an individual’s attitude towards migration (or probability of migrating) is calculated involves a considerable degree of uncertainty. In modelling migration decision-making a number of assumptions must be made that categorise the migration decision into a format from which it can be expressed as a relatively simple algorithm. Creating such an algorithm permits the complexity of the migration decision to be defined by smaller components that require lower level assumptions and thus reduce uncertainty. Crucially therefore, the assumptions made in the development of a simulation of such a complex system are made at a lower level of analysis than have been undertaken in previous, more top-down, approaches to quantifying environmental migration and can act to reduce overall uncertainty. Furthermore, such assumptions can be subjected to a thorough process of sensitivity analysis in order to assess the range of impact that any adjustments to a relevant parameter may have on model outcomes.

In order to manage the extent of the uncertainty incorporated into the ABM presented by this thesis the notion of simplicity of design promoted by An (2011) in the field of complex human and natural systems (CHANS) is adopted. By researching theoretical advances made in the fields of migration decision-making, climate change adaptation and social psychology a conceptual model of migration decision-making was constructed. The conceptual model was then tested and further informed by a period of fieldwork in Burkina Faso during which thirty focus group interviews were undertaken across the country. Finally, survey data collected in 2000 was used to parameterise the ABM. Due to the nature of a model as a representation of a complex phenomenon, the outcome of any simulation can only be a descriptive output. As a result, any explanatory power is significantly constrained by the assumptions made during the process of model development. On this basis, the assumptions involved in the process of model construction were kept to a minimum and empirical parameterisation sought wherever possible.

1.4 Summary

When developed on the basis of observed empirical data that reflects a real-world situation, agent-based modelling provides a realistic and promising opportunity to integrate the multiple variables involved in migration decision-making and manipulate them in order to obtain simulations of future migration patterns. The influence of the unique responses and attitudes of individuals towards changes in rainfall variability, as a proxy for climatic variability and change, are of considerable importance in identifying the livelihood impact they perceive and the importance of this impact in their current and future existence. By developing and testing an ABM intended to investigate the role of changes in rainfall variability in the migration decision-making of Burkinabé people this research aims to quantify the likely influence of future changes in rainfall variability upon migration within and from Burkina Faso towards 2050. In achieving this goal, the ABM is anticipated to provide insight into the degree of complexity manifest within the context of a national-scale simulation of environmental migration.

Chapter 2 of this thesis provides a brief introduction to the significant body of literature that has led to this research. The convergence of the fields of climate change and migration through consideration of climate impacts and human adaptation is described before turning to the growth of the concept of environmental migration. Following a summary of the literature that highlights the nexus between climate change and migration the chapter turns to publications relevant to the case study location, Burkina Faso. Finally, previous approaches to conceptualising migration

decision-making and applications of agent-based modelling within related fields are documented.

Permitting further investigation into the literature surrounding previous models of migration and climate change, Chapter 3 outlines the process of developing a conceptual model of Migration Adaptation to Rainfall Change (MARC) upon which the ABM is based. Through consideration of reactive and proactive responses to climate stimuli and following significant contribution from the theory of planned behaviour, the full conceptual model (MARC) and its simplified version for appropriate ABM construction (SMARC) are described.

The methods and resources used by this research are documented in detail in Chapter 4. By drawing upon the extensive retrospective migration histories dataset made available to the project and the information gained during a two month period of fieldwork in Burkina Faso, the general approach used to construct the modelled rules of agent behaviour is described.

Furthermore, the processes of validating and testing the agent-based model and of interpreting the rainfall data are outlined.

Chapter 5 describes the process of constructing the Agent Migration Adaptation to Rainfall Change (AMARC) model. Following the construction of a basic model that depicts a simplistic form of demographic change within Burkina Faso, agents within the model are endowed with the capacity to undertake migration decisions on the basis of appropriate stimuli. With each agent forming intentions towards migratory behaviour as a result of their individual context and circumstances, the process of translating each component of the SMARC conceptual model into the ABM is documented in detail.

Following construction of the AMARC model, Chapter 6 outlines the extensive process of model validation and testing. The chapter describes how well the modelled migration flows simulated by two versions of the model, AMARC1 and AMARC2, correlate with relevant migration history data used for validation purposes. On the basis of these findings, the AMARC models are deemed to be an appropriate tool from which to investigate the role of changes in rainfall variability in historical migration decisions and to simulate future trends in migration as a result of different rainfall scenarios.

Using a final version of the validated ABM (AMARC3), the process of investigating the role of changes in rainfall variability in the modelled migration decision is documented in Chapter 7.

Within the 1970 to 1999 initial model timeframe, multiple and varied relationships between rainfall and modelled migration are identified at a variety of scales of analysis. The impact of rainfall upon modelled departures and arrivals from each zone are considered during dry, average and wet conditions using both an artificial scenario approach and an analysis of historically dry, average and wet years between 1970 and 1999.

Building upon the findings presented from analysis of model outputs that simulated migration flows between 1970 and 1999, Chapter 8 uses a range of future scenarios of changes in rainfall variability to further test the impact of changing rainfall conditions upon modelled migration. In parallel with the different rainfall scenarios, the additional impact of future scenarios of demographic change upon modelled flows is investigated. On the basis of the nature of modelled migrant flows seen under each combination of rainfall and demographic scenario tested, the potential for emergent outcomes is assessed. The final chapter of the thesis, Chapter 9, outlines the research that has been undertaken before documenting the conclusions, possible implications and recommendations for further work within the field.

Chapter 2