Generalised fuzzy cognitive maps for modelling complex qualitative systems - the case of flood risk reduction planning in Kampala, Uganda
Abhishek Nair is a PhD student in the department of Urban and Regional Planning and Geo-Information Management (PGM). His supervisor is prof.dr.ir. M.F.A.M. van Maarseveen from the Faculty of Geo-Information Science and Earth Observation (ITC).
Climatic changes as a result of increasing greenhouse gases in the atmosphere is now evident and so are the consequences on human and natural system. Though, globally some regions benefit from climate change most regions, particular in the tropics and higher latitudes face adverse consequences. For cities in sub-Saharan Africa the expected 1.5oC increase in global temperature calls for an urgent reflexion towards low carbon, climate resilient development driven pathways in view of the rapid population growth, deep and persistent poverty, displacement, relocation (often forced), and emerging new climatic risks. System dynamic (SD) models are useful because they try to simulate behaviour of systems and test policy options. However, there is a need for local SD modelling exercises to understand location specific–climate risks and vulnerabilities to implement successful climate compatible development strategies.
System dynamic models usually rely on quantitative data sparing out interactions and complexities that are qualitative in nature. This is also evident with system dynamic models trying to simulate climate change impacts, adaptations and vulnerabilities. To be able to explain, predict and understand complexity qualitative phenomena–that can play a substantive role in systems–should be included. The exclusion of such qualitative concepts can bring into question the conclusion arrived at and the models relation to reality. Additionally, quantitative system dynamic models face the issues of low data availability, routine challenges of measurement and quantification, and the inclusion of uncertain information in the model. This limits their applicability in modelling complex system dynamics in data scarce regions such as emerging economies and low-income countries. Qualitative system modelling would be the ideal choice to overcome the aforementioned problems, however, no dynamical analysis of the system through time is possible due to the qualitative nature and uncertainty in numerical data obtained, which makes the formulation of mathematical models difficult. This limits qualitative system dynamic models in predicting the behaviour of systems as well as evaluating development strategies and policy options. Hence, there is a need for semi-quantitative SD modelling methods to address issues faced by quantitative and qualitative modelling methods. Semi-quantitative/mixed modelling methods such as FCMs, i) provide an approach for developing better more context specific instruments, ii) provide a holistic and comprehensive understanding of the research problem and iii) shows how causal processes propagate. In this regard, Fuzzy Cognitive Maps (FCMs) a powerful expert driven soft-computing method to model system dynamics (SD), and is increasingly being used in socio-ecological systems analysis, urban planning, and management, medicine, and business intelligence among others. However, FCMs suffer from several drawbacks and limitations thus rendering it not fully capable of modelling complex qualitative system dynamics.
The aim of this study was to design a flexible fuzzy cognitive mapping approach that is capable of modelling and analysing the behaviour of complex qualitative systems. The motivation for designing such a method is to help analyse the behaviour of complex social-environmental-technological systems, specifically, to understand the manifestation of climate risks and analyse possible policy options and development pathways. To achieve this objective an in-depth analysis of the capabilities of traditional FCMs and its advances in modelling complex qualitative systems were analysed. A flexible approach was designed based on cogitating complex system dynamic properties and exploring issues that need to be addressed in FCMs to be able to model complex qualitative SD. The efficacy of this approach, was tested in the light of a real world case of the socio-environmental-technological consequences of heavy rainfall in Kampala, Uganda.
Generalised Fuzzy Cognitive Maps (GFCMs) are expert driven that can capture and model complex causal reasoning. GFCMs use fuzzy rules to represent dynamics of concepts and relations (perceived), including the time dynamics of relations. GFCMs introduces several, single-layer perceptrons’ to simulate dynamics, i.e. temporally explicit development of concepts and relations. The results explains, the relative, and absolute change the system undergoes overtime and can also model time decayed and lagged dynamic influences. The results of the simulations using GFCMs explains the evolution of a system to a greater detail while overcoming most drawbacks. It is concluded that GFCMs are a possible improvement over most methods trying to model complex qualitative SD.
Finally, the outcome of this study is to help robust decision-making, in the light of uncertainty and complexity. This study can be useful two groups of stakeholders i) researchers that try model the behaviour of complex systems in data scarce environments/regions and ii) decision-makers that wish to ate several development strategies and policy options or pathways for a given system.