The Governance of Climate Change Adaptation in Developing Countries: The Case of National Disaster Management in Bangladesh and Pakistan in Comparative Perspective

Index

Acronyms

List of Figures, Tables and Graphs

1 Introduction
1.1 Natural Disasters in the Scope of Social Sciences
1.2 Theoretical Framework and Conceptualization
1.2.1 Governance in Social-Ecological Systems
1.2.2 Vulnerability and Adaptation to External Stressors
1.2.3 Collective Learning Processes in Social-Ecological Systems
1.3 Measuring Collective Learning Processes and Vulnerability
1.4 Case Study Selection
1.4.1 Why Studying Floods in Developing Countries?
1.4.2 Comparative Method and Case-Studies

2 Bangladesh - Learning How to Life with Extreme Floods
2.1 Parameters of Flood and Disaster Management
2.1.1 Political Transition and the Governance of Flood Management
2.1.2 Complexity of Flood Disasters and Vulnerability
2.2 Learning Processes in the Governance of Flood Management before 1988 until Recently
2.2.1 Development Prior to1988
2.2.2 Changes after the Floods of 1988
2.2.3 Changes after the Floods of 1998
2.3 Collective Learning Processes in the Governance of Flood and Disaster Management in Bangladesh

3 Pakistan – First Steps towards Change
3.1 Parameters of Flood and Disaster Management
3.1.1 Political Transition and Flood Management
3.1.2 Complexity of Flood Disasters and Vulnerability
3.2 Learning Processes in the Governance of Flood Management after 1971
3.2.1 Development in the Flood and Disaster Management after 1971
3.2.2 The Floods of 2010 and their Aftermath
3.3 Collective Learning Processes in the Governance of Flood and Disaster Management in Pakistan

4 Conclusion
4.1 Collective Learning in Flood and Disaster Management: A Comparison
4.2 Summary of Findings
4.2.1 The Influence of Collective Learning Processes on Vulnerability
4.2.2 The Way Collective Learning Changes Vulnerability
4.3 Limitations and Outlook

Bibliography

Annex

Acronyms

illustration not visible in this excerpt

List of Figures, Tables and Graphs

Tables

Table 1 Definitions in the Concept of Triple-Loop Learning

Table 2 Institutional Changes in Governance Systems

Table 3 Changes in Actor Networks in Governance Systems

Table 4 Changes in Multi-Level Interactions in Governance Systems

Table 5 Changes in Governance Modes in Governance Systems

Table 6 Changing Concepts of Uncertainty

Table 7 Major Keywords Categorized by Stage of Research

Table 8 Mill’s Design of Difference

Table 9 Collective Learning Patterns in Bangladesh and Pakistan

Figures

Figure 1 Scheme of an Adaptation Process

Graphs

Graph 1 Physical Exposure and Relative Vulnerability to Floods in South Asia

Graph 2 HDI of Bangladesh and Pakistan by comparison from 1980-2010

1 Introduction

When exceptionally heavy monsoon rains led to rapidly increasing water levels of the Pakistani Indus River and its tributaries in late July 2010 not a single town hit by the torrents was adequately prepared or even warned despite the presence of information that was given to government officials (Webster et al. 2011; Vastag 2011). The severe unpreparedness of the affected areas is mainly reflected in the numbers of human losses of nearly two thousand but also the number of affected people with over twenty million being directly affected by the floodwater, including displaced and injured people (CRED 2009).

Natural disasters like the Indus River Floods are very complex phenomena that result from a variety of determinants including human as well as natural impacts (Wisner et al. 2003). As a catalyst, climate change is supposed to have a significant influence on such hydrological natural disasters in present and near future (IPCC 2007). Though climate change does not change the ways that natural disasters like floods occur, it acts as a catalyst aggravating existing problems (Ibid. 2007: 361). If climate change is not creating any unknown problems, one might ask, why should we start conducting research in fields that have been subject to detailed research for a long time previously? The answer can partly be found within the first paragraph of this introduction: Even though what can be observed stays the same, changes in the patterns and impact of climate change-related events occur at the present time and are predicted to intensify in the course of time. These changing patterns result in very high levels of uncertainty for policy-makers and societies, which potentially can lead to enormous devastation if they are being ignored.

General awareness about the severe impacts of disasters on human development has been raised especially on the international level within the framework of the Millennium Development Goals for instance. Based on the heightened awareness about risks, steps towards risk reduction have been initiated on an international and a national level. Consequently, in the year 2000 the United Nations International Strategy on Disaster Reduction (UNISDR) was established with the mandate to improve coordination of disaster reduction. On a national level, disaster risk was incorporated into Poverty Reduction Strategy Papers in order to ensure sustainable development (UNDP 2004). Also in academic circles for many years, research has been conducted on the subject of disaster risk reduction. This lead to an advanced understanding of how various types of natural hazards can interact with societies (Mercer 2010; Wisner et al. 2003). Or rather, how disasters are getting constructed by humans that modify their environment.

Within the past two decades, climate change became a broadly discussed issue within the discourse on disaster reduction of practitioners and academics (cf. UNISDR 2008; Mercer 2007). Besides efforts at mitigation, adaptation to climate change became a very relevant subject of concern. In order to adapt to the prospective and ongoing impacts of climate change at the international level, relevant instruments have been set up recently. The Adaptation Fund set up under the United Nations Framework Convention on Climate Change (UNFCCC) is one example of these instruments, whose major objective is to support programs and projects in member countries “that are particularly vulnerable to the adverse effects of climate change” (UNFCCC 2011). Vulnerability and risk assessments are therefore seen as important factors determining the prioritization of different countries or regions in order to support their adaptation measures (Adger et al. 2004:15)

Recently, studies in the climate change adaptation and disaster risk literature have been conducted in order to make countries comparable with regard to their vulnerability to external hazards (cf. Adger et al. 2004; UNDP 2004; Brooks et al. 2005). A variety of social, political, economic and hazard-specific indicators have been identified in these studies in order to build foundations of a vulnerability or risk index. Adger, Brooks, Bentham and Eriksen conclude in their assessment that the identified indicators of vulnerability “[…] might be of use to international agencies and donors wishing to prioritise adaptation assistance to the most vulnerable nations, but it tells us nothing about the structure and causes of vulnerability.” (Ibid. 2004: 93).

The need to develop and test further determinants of vulnerability, which are able to capture its complexity, was therefore addressed by some authors (e.g. Adger et al. 2004: 101; UNDP 2004: 115-16).

It is therefore the aim of the present study to address the currently existing gap in the knowledge about determinants of vulnerability. For this purpose, the influence of a more complex determinant on the vulnerability of a society on national level was tested. In order to deepen the insight into a specific field of disaster risk and delimit the scope of this study, a specific hazard was chosen. Thus, floods were chosen to be the major subject to this investigation because flood hazards are very complex phenomena that are largely influenced by the interaction of humans with their environment. Furthermore, floods are a typical hydrological hazard that is expected to be increased by climate change through extreme precipitation and stronger seasonal melt of glaciers (IPCC 2007).

In the course of investigating for a determinant that is eligible to capture broader complexity of vulnerability, a side observation gave the decisive impulse. A quantitative research study conducted by the United Nations Development Programme (UNDP) on the vulnerability of different countries to flood hazards, revealed a rather counter-intuitive result: In four world regions there exists a trend stating that the higher a society’s physical exposure^[1] to flood hazard is the lower tends to be its relative vulnerability^[2] (UNDP 2004; Graph 1). A similar trend is getting visible when holding the annual numbers of floods against the relative vulnerability between countries of one region. The hazard frequency and magnitude as external factors therefore seem to have an influence on the relative vulnerability to floods. The context described in this section reminds of learning patterns in such a way that increasing repetition and intensity of an external stress^[3] potentially advances the reaction of a system to this external stressor. Further support of the assumption that learning is an important determinant of vulnerability was found in theoretic approaches of complex adaptive systems and collective learning processes and their influence on adaptation (cf. Duit and Galaz 2008; Pahl-Wostl 2009; Duit et al. 2010; Loef 2010; Gerlak and Heikkila 2011). Even though there is a majority in the academic literature, which supports that learning processes in theory have a significant impact on the vulnerability of a society, empirical studies in this field are still rare or underway (cf. Pahl-Wostl 2009; Simonsen 2010). The present study therefore aims to diminish gaps within this field of study.

The determinant of vulnerability that was chosen for further investigation within the present study will therefore be collective learning processes. The guiding research questions are therefore: Do collective learning processes have a detectable influence on the vulnerability of a society to hazards? And subsequently: How did collective learning processes in Bangladesh and Pakistan influence the vulnerability of their respective societies to flood hazards? In order to find answers to these questions the present study will proceed in a four-step manner.

Firstly, a review of the current state of literature on climate adaptation and disaster risk builds a base in order to classify the present analysis and therefore defining its scope. In line with this classification a theoretical framework was developed in order to further enhance the research scope and to conceptualize and define important key components of adaptation and learning. This step comprises of the first two subsections.

Secondly, the conceptualization made was used in order to determine the methods of the core analysis. In order to assess the influence of learning processes on vulnerability a mixed research design was established. Collective learning processes contain very complex interactions and therefore it is not feasible to analyze them thoroughly based on sets of quantitative indicators. Vulnerability on the contrary can be measured by quantitative indicators human physical security. The present analysis is therefore composed out of a qualitative research about collective learning processes that is connected to a qualitative assessment of vulnerability (cf. Creswell 2009). In order to make informed statements about the influence of collective learning processes on vulnerability two case studies have been chosen that are compared following the example of Mill’s Design of Difference (Ibid. 2009 [1872]). The aim of this comparative method is to proof that higher vulnerability is connected to lower levels of collective learning. Major obstacle in this comparison was the identification of interfering variables, which had to be controlled. Collective learning processes on national level demand long periods of time and therefore can only be captured using information and data over extended time scales. In order to capture collective learning processes an ex-post research was conducted that covers the major developments in the governance of flood and disaster management of the two case studies from the early 1900s. This research was done based on a review of existing primary and secondary literature including majorly field studies, sector studies, policy studies, reports and legal documents. In order to systematize the information gathered and to identify learning processes within the case studies a research framework developed by Claudia Pahl Wostl (2010) in the water management literature was considered.

The third step includes the actual analysis based on the previously outlined research design. The case studies of Bangladesh and Pakistan were chosen under defined criterion oriented towards Mill’s Design of Difference (Ibid. 2009 [1872]) and firstly analyzed separately. Analyzing both case studies separately has the advantage that learning processes can be illustrated in a comprehensive manner. The disadvantage of this design draws clearly from the risk of falling into narrative analysis. In order to avoid this, the case studies include cross references.

Lastly, the knowledge acquired about collective learning processes in the flood and disaster development of Bangladesh and Pakistan will be used to do a final comparison in order to draw conclusions on whether learning processes influenced the vulnerability of the populations in both countries. Furthermore, the question of how vulnerability and collective learning processes are related to each other will be discussed.

1.1 Natural Disasters in the Scope of Social Sciences

Since the early seventies increasing attention is being focused on the subject of climate change in the scientific world, leading to the first World Climate Conference by the World Meteorological Organization (WMO) in 1979. A long period of controversy, primarily led by the IPCC and skeptics, about the existence of global climatic change and the degree of influence by human activities, started during the intervening years. In 2005 the Kyoto Protocol entered into force, indicating that a vast majority of countries consent on the acknowledgement of global climate change as an impact of human activities. The Fourth Assessment Report of the IPCC in 2007 reconfirms this consent and draws a sophisticated picture of observations of impacts that can be drawn back to climatic changes.

Parallel to a growing awareness and recognition of climate change and its mitigation, the problem of adaptation aroused attention since the early nineties. During this period of time, the IPCC started addressing the importance of adaptation until it was considered a priority area for research in the Third Assessment Report in 2001. Literature on climate change adaptation and its impact on vulnerability had been strongly increasing ever since.

The complexity of adaptation to climate change becomes apparent when scrutinizing its major fields of study like the physical science base of climate change, environmental governance systems, structures and determinants of adaptation processes and vulnerability or disaster risk management. Considering that this list is still only a rough picture of all components, clear systematizations and definitions are a crucial part of any study conducted in this field. The present study does not discuss the physical science base, but covers all of the other areas mentioned above to different extents. It will therefore be based on the theorization of adaptation and governance, vulnerability and adaptive capacity while applying these underlying concepts to hazard-specific disaster risk management.

Out of the variety of literature on adaptation, vulnerability, resilience or adaptive capacity^[4] derives a vast amount of definitions that often cover different concepts with similar terminology. Most of these differences in meanings exist because terms are used context-specific and sometimes authors do not explicitly refer to other uses of the term in order to differentiate their own concepts from others. If unnoticed by the reader this variety of meanings can cause confusion and misinterpretations.

An illustrative example for this ambiguity is the use of the term adaptive capacity, a very often used term within the adaptation literature. Whereas the practically-oriented adaptation literature^[5] (e.g. Smit and Wandel 2006; Handmer 2003; IPCC 2001) stresses materialistic capacities to adapt in their definition, theoretically-oriented adaptation literature^[6] (Pahl-Wostl 2009; Armitage et al. 2010; Diduck 2010) tends to stress the ability to adapt by itself. Also there exists a mélange out of both streams defining adaptive capacity as “ the ability or capacity of a system to modify or change its characteristics or behavior so as to cope better with existing or anticipated external stresses.” (Adger et al. 2004: 34). Again, the use of many terms in adaptation literature is very context-specific and should always be considered this way.

In order to reduce complexity and enhance transparency in the adaptation literature, a number of authors developed typologies sorting literature by purpose, research design or other criteria. Smit and Wandel (2006) distinguish four types of climate change adaptation literature according to their purpose: measuring the effectiveness of given adaptation measures (e.g. Fankhauser 1998; Parry 2002); comparing the utility of different adaptation measures to a specific system (e.g. Winters et al. 1998; Parry et al. 2001); prioritization of geographical areas by measuring vulnerability (e.g. Adger et al. 2004; Brooks et al. 2005) or investigating adaptive capacity or needs of a certain area in order to derive policy recommendations (e.g. Keskitalo 2004; Ibid. 2010). Even though this distinction provides a decent overview in the practical field, it leaves out literature that rather concentrates on the process of adaptation in a systemic context (overview in: Van Nieuwaal et al. 2009). In other words, parallel to the research on adaptation in practice, a section of literature is concerned with the human behavior that is the underlying driver of adaptation (Pelling and High 2005: 1).

Governance-centered studies on climate change adaptation are of importance for a better understanding of the process of adaptation within a governance system. Governance systems within most of this literature are treated as complex systems that are confronted with non-linear external stressors that create uncertainty (e.g. Levin 2003; Holland 2006; Bauer and Schneider 2007; Duit et al. 2010; Loef 2010). The major subject of study is a system’s response to stressors and the effectiveness of certain settings within a system like actor constellations.

Within the present study, the literature on governance and adaptation from a conceptual perspective is of special interest. General aim of literature on conceptual frameworks is “to converge the inherent complexity and unpredictability of ecosystem dynamics into new governance or management concepts” (Van Nieuwaal et al. 2009: 15). Van Nieuwaal, Driessen, Spit and Termeer (2009) attribute the concepts of adaptive governance; resilience management; adaptive management; adaptive co-management; adaptive collaborative management; environmental governance; and earth system governance to this section of literature. Focus of these analyses is the transformation in a system that is targeted on adapting it to an external stressor. Since this implies that a system recognizes a threat and actively tries to defend itself against it, these approaches impose high standards to the systems of concern.

In transformation studies, concepts on collective learning are an important field of analysis. Also different conceptual frameworks in the governance and adaptation literature identify learning processes as parts of adaptation processes (e.g. Armitage et al. 2008; Pahl-Wostl 2009; Loef 2010). Learning is assumed to be a social phenomenon occurring at multiple levels in such papers (Pahl-Wostl et al. 2007; Ibid. 2009; Diduck 2010). Even though it is widely recognized that learning processes are important for governance systems in order to sustain (Allen 2001; Armitage et al. 2008; Duit and Galaz 2008), only few studies exist on how these processes occur in practice and what factors foster them (Gerlak and Heikkila 2011: 2). The present study aims on further closing this gap and therefore relies on an existing approach.

The “Conceptual framework for analyzing adaptive capacity and multi-level learning processes in resource governance regimes” developed by Pahl-Wostl (2009) is a cornerstone for the empirical analysis of the present study. Major objective of Pahl-Wostl’s paper is to develop a framework allowing researchers to systematically analyze changes within resource governance systems as multi-level learning processes (Ibid. 2009: 355). The entire paper draws back to a preceding empirical study on water resource governance regimes in different industrial and developing countries on behalf of the European Commission (Huntjens et al. 2008). Thus, it is a very relevant framework of the present study for it focuses on a similar objective of study.

Resource governance systems are characterized by four major features:

i. “the influence of formal and informal institutions,
ii. the role of state and non-state actors,
iii. the nature of multi-level interactions and
iv. the relative importance of bureaucratic hierarchies, markets and networks.” (Pahl-Wostl 2009: 356).

Changes in resource governance regimes are conceived as societal and social learning processes (Ibid.:358). Societal learning is not understood as a change within an entire population (in contrast to Diduck 2010), but as a change taking place in the part of a society that experiences a common obstacle either as persons affected, organizations in charge or as decision-makers. Social learning refers to a way of learning that is particularly important in this framework because the involvement of various actors at multiple levels is supposed to lead to a higher adaptive capacity in a system (Folke et al. 2005; Pahl-Wostl et al. 2007; Ibid. 2009).

illustration not visible in this excerpt

Table 1: Definitions in the Concept of Triple-Loop Learning

Source: Author’s design derived from Argyris and Schön 1978; Pahl-Wostl 2009.

The concept of learning is based on a stepwise model that divides learning into three levels. This concept of multi-level learning derives from organizational theory and presumes that organizations undergo transformations of different qualities (Argyris and Schön 1978). The different learning levels an organization or society can undergo are characterized by their degree of change (Table 1).

In order to make learning patterns detectable within governance regimes, Pahl-Wostl (2009) creates a matrix that analyzes changes in governance regimes and attributes them to the three levels of learning. Within this matrix the four major features of governance and the capacity of dealing with uncertainty are focused upon. Additionally, the manner in which uncertainty is treated within these governance systems is considered. The indicators of change derive from the conceptualization of learning as well as the empirical background and thereby provide a balanced scope on learning processes. Additionally, a critical evaluation of Pahl-Wostl’s framework is integrated in the conclusion of the present study. This evaluation includes the underlying theoretical concepts as well as the empirical dimension.

Disaster risk management or in short disaster management is another relevant field of study to this study for it relates to the empirical case studies that include flood risk management. Disaster management is a considerably new area of study (GTZ 2002), which is mainly due to a paradigm shift away from reactive strategies towards more preventive approaches (Yodmani 2001). Literature on disaster risk management is often of a practical nature, concentrating on case studies and strategies towards a better management of risk (Mercer 2010:248). Besides empirically oriented studies there is also literature on theoretical aspects of disaster risk. Wisner, Blaikie, Cannon and Davis explain for instance how risk is constructed through environmental, economical, social and political influences (Wisner et al. 2003).

1.2 Theoretical Framework and Conceptualization

The present research question, in general, focuses on the interaction of a dynamic external stressor with a given society. Additionally, uncertainty about the external stressor is strengthened through projections of further uncertainty, in this case climate change. Since this requires the analysis of highly complex and non-linear interlinkages (cf. Duit and Galaz 2008: 312), a systemic approach is used as theoretical base throughout the analysis. The concepts of governance and governance systems, vulnerability and collective learning are in the centre of focus for they help explaining how collective learning processes can be detected and understood within the following case studies.

1.2.1 Governance in Social-Ecological Systems

The following subsection aims on highlighting factors within social-ecological systems that are relevant for this system’s ability to adapt to external shocks like natural disasters. It therefore starts with a brief introduction to important concepts in the governance research. Subsequently, specific components of governance systems are highlighted, which build the basis for the analysis of the two case studies.

From a systems theory perspective, the functioning of a social system cannot be solely seen as a consequence of the political system or state^[7] (Luhmann 1975; Ibid. 2000). The present study connects to this notion and draws back to a governance concept that considers hierarchical and non-hierarchical modes of social coordination, which may or may not include governmental involvement (Mayntz 2004; Risse and Lehmkuhl 2006; Börzel 2010). Throughout this study the term governance refers to “the entirety of all co-existing modes of collectively regulating social matters”^[8] (Mayntz 2004: 66). This definition includes different modes of social coordination, state and non-state actors and multi-level coordination that are going to be discussed further.

Governance is an often used, but nevertheless blurry term within the social sciences. Though, it is often confused with government, the actual meaning of both terms differ significantly. From a political science perspective, government and governance have similar outputs^[9], while differences derive from the mode of establishing rules or steering collective action. In contrast to hierarchical structures that are a defining feature of government, governance refers to less restrictive mechanisms of governing, which can also include private actors (Stoker 1998: 17). Governance perspective can therefore explain to certain extends how hierarchical and non-hierarchical patterns can be part of one system simultaneously.

For the past decades a paradigm shift from ‘government’ to ‘governance’ became apparent within social sciences (Van Nieuwaal 2009: 9), which is furthermore an indicator that a change in perceptions has occurred (Benz 2004: 13). This shift can mainly be explained by the observation that there exist spaces like the international system, weak states or Public-Private Partnerships that cannot be explained by state structures and hierarchical bureaucracies (Ibid. 2004). Especially when focusing on the developing world, this concept plays an important role because it provides explanations on how a weak state with limited steering capacity can still be able to rule a country.

The primary subject to this study will be the water- and flood management sector^[10] and the disaster management sector, whilst the influence of other relevant sectors like agriculture and forestry will be considered subsequently in the two case studies. Both these sectors form an environmental governance system that is a social-ecological system, whose purpose it is to govern human behavior towards a particular ecological system. In order to analyze the structures of environmental governance systems it is useful to divide them into their respective institutions, actors and their interactions and finally modes of governance (Pahl-Wostl 2009: 356).^[11]

Institutions refer to rules, which have been set up in order for a social entity to coordinate different kinds of interactions in order to reduce uncertainty for its members and also external actors (cf. North 1990: 1-6). These rules can either be established formally or practiced informally, whereas informal institutions can be transformed into formal institutions and both types influence each other. Most importantly, formal and informal institutions differ in the ways they can be enforced (cf. Ibid. 1990: 4). Legislative frameworks can help individuals to enforce their rights at higher levels like it was the case after the floods of 2010 in Pakistan when numerous petitioners held governmental agencies and authorities responsible for not fulfilling their duties which have been defined in the Disaster Management Ordinance and the respective Act (cf. section 3.2.2). Before these competencies were not formalized in a law, it was not possible for the aggrieved party to claim for compensation and responsibilities. Within the analysis of the case studies particular emphasis is placed on formal institutions and prevailing paradigms in flood- and disaster management. In contrast to Pahl-Wostl’s framework normative institutions are not considered within the present analysis, because these refer to informal habitual norms at the micro-level. The following analysis though, concentrates on macro-level developments with emphasis on the national level.

As environmental governance systems include a vast range of actors at different levels, competencies and responsibilities become blurred. The case study analysis indicates that international donors and agencies, the federal government, local governments, civil society and various kinds of stakeholders are key actors involved in these complex governance systems to different extents. Participatory approaches have become important for environmental governance, which has been recognized by academics as well as practitioners (cf. Ostrom 1999; Sultana et al. 2008; Pahl-Wostl 2009; Ibid. 2007). Especially in environmental governance systems knowledge of local individuals can contribute to a large extent to reduce uncertainty (Ostrom 1999: 520-21). Though participation is generally regarded as the second way besides national government especially in developing countries, it also includes weaknesses. Especially problems of legitimacy, lack of representing poor locals and the notion that what locals want does not necessarily mean that it is the solution benefiting the largest share of all stakeholders are hindering the positive effects of participation. Decisive for the impact of participatory processes are the degrees to which affected segments of society match with those who take action in building, maintaining or operating flood protection facilities for instance. In Bangladesh autonomous flood management by local stakeholders has been connected to some degree with formal governmental organizations (cf. Sultana et al. 2008). Diverse actor networks do not automatically generate broad benefits within societies but they are certainly a prerequisite for reducing uncertainty by ensuring multiple sources of knowledge and experience. Therefore, diversification of actors enhances the capacity of a system to adapt to external influences (Pahl-Wostl 2009; Folke et al. 2005).

Multi-level interactions are a forming characteristic of environmental governance systems since governance may include actors from supranational- down to individual level (cf. Hooghe and Marks 2003; Benz 2004). Interactions can emerge between actors of one level, which is often referred to as horizontal coordination, and also between actors of different levels, known as vertical coordination (van Nieuwaal et al. 2009; Scharpf 1997). Consequently, vertical coordination is the more decisive process when analyzing multi-level interactions. From a normative perspective a system’s capability to react to non-linear external shocks is getting increased when it includes more than one centre of governance that is able to operate independently (cf. Ostrom 2001; Duit and Galaz 2008; Pahl-Wostl 2009). The reason behind this suggested correlation is that these polycentric systems can counterbalance failures and provide multiple sources of knowledge and experience in order to mitigate, react and respond to external shocks (cf. Ostrom 2001; Pahl-Wostl 2009). Like Claudia Pahl Wostl points out: “Multi-level governance in polycentric systems implies that decision making authority is distributed in a nested hierarchy and does not reside at one single level” (Ibid. 2009: 357). Decentralization is therefore a crucial characteristic of systems that are more resilient towards external shocks. It is important to note in this context that whilst the different centers of governance are able to act independently from one another, they need to exercise vertical and horizontal coordination and cooperation in order to create a functioning system (cf. Kooiman 2000; Pahl-Wostl 2009; Ibid. 2008).

The major driver of governance systems is interactions among their respective actors. Out of the constellation of actors and the ways they interact, different modes of governance can be derived. Kooiman identifies three major modes of governance that is self-governance, co-governance and hierarchical governance (Ibid. 2007: 10). These three types relate to the degree of formality of institutions, whilst self-governance is the most informal arrangement. A second central dimension of governance is the involvement of state and non-state actors (Thompson et al. 1991: 228-29). By considering the two dimensions of governance modes like discussed above, networks, markets and bureaucratic hierarchies can be classified as the major modes of governance (Thompson et al. 1991). Within the two case studies bureaucratic hierarchies were the dominant mode of governance, whilst networks and markets existed but had only limited or no access to decision-making. Bureaucratic hierarchies are driven by state actors through formalized channels, while networks are organized in a very informal manner including only a limited proportion of state actors. Markets are dominated by non-state actors and regulated by formal as well as informal institutions (Thompson 1991; Pahl-Wostl 2009). Following Pahl-Wostl’s framework, equilibrium of all three modes of governance is the most desirable constellation for a system in order to adapt to sudden external changes (Ibid. 2009: 358).

All four elements are important to be considered in order to observe substantial changes in environmental governance systems. Within this section the basic components and concepts of governance and environmental governance systems were explained. The next subsection will address approaches that are concerned with how these systems interact with their complex environment.

1.2.2 Vulnerability and Adaptation to External Stressors

Environmental governance systems are considered to be complex adaptive systems. This is due to the extremely high levels of uncertainty and complexity that evolve while governing ecological systems. Changes in such systems are occurring on a frequent basis and are hardly predictable (Duit and Galaz 2008: 312-13). Human behavior has a great influence on ecological systems causing further changes. Whether a social system is vulnerable to the changes in its natural environment is not solely determined by exogenous factors. In order to fully understand vulnerability, endogenous factors like dependence of a society on agriculture play an important role (Adger and Vinecent 2004; Brooks 2003).

Each social entity has a certain coping range, which indicates to what degree it is resilient to external stresses (Smit and Pilifosova 2003: 12-14). This coping range is a product of factors like the sensitivity, the resilience and the adaptive capacity of a system. The following example shall clarify the relationships among these factors. A quarter of the Dutch territory lies beneath sea level making it extraordinarily sensitive to any rise of the sea water level. Planners were commissioned to build dykes and other flood protection structures over a thousand years ago in order to build up resilience. Sensitivity in this case emerges on one hand out of the geographical conditions on the territory and on the other hand out of the circumstance that humans settled within this particular location. The adaptive capacity of this system lies in the mental and physical capability of the planners and workers firstly to find a solution and secondly to construct proper structures in the right locations. The process of planning and constructing dykes is considered adaptation. Through the capability to adapt and the actual adaptation, the system improved its resilience to sea floods. As long as the coping range of a country to a specific stressor is not exceeded, meaning the resilience of a system is able to absorb the negative effects, a society is in balance to its environment (cf. figure 1).

illustration not visible in this excerpt

Figure 1: Scheme of an Adaptation Process

Source: Author’s design derived from Smit et al. 2000; Smit and Pilifosova 2003; Adger and Vinecent 2004; Smit and Wandel 2006.

Vulnerability emerges when the coping range of a system is exceeded meaning that the existing resilience is not sufficient in order to absorb the negative effects of the external stressor. This is a very much simplified view on vulnerability and how it develops. Vulnerability implies very many dimensions referring to the ways it develops and the ways it is reduced in societies. A more specified definition of vulnerability must therefore include what type of stressor is considered (Brooks 2003: 3). The focus of the present analysis lies on the external stressor of natural disasters, which is considered an extreme case. Following Wisner, Blaiki, Cannon and Davis vulnerability is defined as:

“ […]the characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist and recover from the impact of a natural hazard […]” (Ibid. 2003: 11).

This definition represents socially constructed vulnerability. Besides this type of vulnerability the present analysis also highlights biophysical determinants of vulnerability that derive from the intensity and frequency of a natural hazard (cf. Brooks 2003; Wisner et al. 2003). The assumption therefore is that high intensity and frequency of a hazard contribute positively to the adaptive capacity of an affected society. Improvements in adaptive capacity, in turn, decreases vulnerability of a system (cf. Armitage and Plummer 2010). Accordingly, vulnerability is by no means a static condition within any system.

The approach of complex adaptive systems assumes that external stress will at a certain point result in internal change of the affected system in order for it to adapt (cf. Duit and Galaz 2008). The point at which adaptation measures are taken up is hardly predictable, especially due to the complexity in the relations between key actors in highly differentiated governance systems. In these systems the group of people that are affected by a natural disaster for instance does not necessarily match the group that holds the powers and capacity of initiating adequate counter measures (Sultana et al. 2008: 360-61). The path from the perception of the risk to collective action can therefore be extremely challenging, especially under conditions of low compliance and communication among stakeholders and decision-makers (Ostrom 1998: 16-18). Sultana, Johnson and Thompson furthermore highlight how natural hazards combined with advanced media coverage can accelerate policy debates and therefore enhance communication about risk (Ibid. 2010).

Once a risk has been acknowledged and collective action was taken in order to adapt, the considered measure must not necessarily be an adequate solution to the problem. Adaptation measures can be harmful itself (cf. Fig. 1) like in the Bangladeshi case where large-scale embankment structures were viewed as final solution to the flood problem. In contrast, they turned out to be harmful due to their lack of flexible regulation of water flow (Section 2.2.2). The manner in which adaptation takes place influences the sustainability of the measures that are undertaken. Firstly, adaptation can take place as autonomous reaction to a certain event with limited long-term problem solving capacity. The present analysis, in turn, concentrates on adaptation as a consciously planned activity that is rather subject to collective action than autonomous adaptation. Planned adaptation activities can be either reactive or anticipatory (cf. Smit and Wandel 2006; Smit et al 2000; Nykvist and Hahn 2011). Anticipatory adaptation is the more advanced type and demands high levels of cooperation and knowledge generation among other factors in order to emerge (cf. Smit and Wandel 2006; Berkes et al. 2008: 11-12). In conclusion, for this type of adaptation a systemic structure is needed that has the capacity of overcoming reactive adaptation.

Adaptive capacity is considered to be the key concept in order to reduce vulnerability in a system. As already mentioned, adaptive capacity refers to all sorts of resources within a system that enable it to adapt to its environment (cf. Section 1.1). The example of the Netherlands reveals that adaptive capacity implies ‘physical’ and ‘mental’ capacities of a governance system.^[12] The present analysis stresses the importance of ‘mental’ capacities since they are needed to build up physical capacities like finances and infrastructure for instance. In accordance, the present analysis draws on Pahl-Wostl’s definition of adaptive capacity as a working definition, because it highlights skills. Adaptive capacity is therefore referred to as:

“ the ability of a resource governance system to […] alter […] and […] convert structural elements as response to experienced or expected changes in the societal or natural environment” (Pahl-Wostl 2009: 355).

Learning processes are closely related to the adaptive capacity of environmental governance systems, since they enable them to govern in an anticipatory manner. Through the collective acquisition of knowledge and experience, maladaptations are less likely to occur and easier to accommodate (Lebel et al. 2010).

1.2.3 Collective Learning Processes in Social-Ecological Systems

Collective learning is an important subject to the present analysis for it is presumed to be a very relevant determinant of vulnerability. This subsection seeks to highlight this nexus and establish a clear picture of what learning is in particular and what factors encourage it. It will therefore only focus on concepts of collective learning, which are particularly relevant for environmental governance systems. Theories and models on learning processes, also within collectives, are necessarily diverse whereas topics range from methodological aspects of learning processes to cognitive patterns of individual learners over to learning processes in international organizations.^[13] Out of these theories and models only a limited array is relevant to the present analysis.

Within the many concepts of learning, two dimensions of learning are especially highlighted (Gerlak and Heikkila 2011: 3). Firstly, collective learning implies a step-wise process, which mainly consists of the acquisition of knowledge, the dissemination and processing of information and the transformation of knowledge within an organizational framework (Argyris and Schön 1996: 2-3). This process has different facets depending on whether learning occurs in an experimental or in a targeted manner (Henry 2009). The process of collective learning is furthermore to be delineated from plain reaction to an external change (Löf 2010: 531-32). In contrast to reactive behavior, collective learning is considered more sustainable with regard to its time scale and the depth of change. In the present analysis the process described here is not regarded as being sufficient without transferring knowledge into substantial results in order to achieve learning. Collective learning is therefore a process in which knowledge is transferred into different kinds of changes, also called products of learning (Argyris and Schön 1996: 2-3). It is important to notice that learning does not necessarily lead to improved products, since experimental learning for instance even relies on trial and error schemes (Gerlak and Heikkila 2011: 3-4). Therefore failed trials can lead to new understanding and reconsiderations of strategies.

In order to decide which event is considered a collective learning process and which is not, the present analysis proceeds in a two-step fashion. Firstly, sustainable changes within the scope of the case studies are detected, which is followed by an analysis of the process that led to the observed change. For instance, a fundamental change in policies, which is not based on newly acquired knowledge or experiences^[14], is likely to be a plain product of political or economic interests but not learning. Also policy that lacks the attempt of its implementation is not considered learning according to the understanding of the present study. In summary, learning is expected to be a purposeful activity as opposed to fast and reactive decisions that are not based on past developments or lack any attempt to be put into practice.

Collective learning processes differ in their depth of change (Argyris and Schön 1978). The concept of multiple learning loops as illustrated in Table 1 characterizes the different stages of learning that ranges from single to triple-loop learning. Single-loop learning is merely an adaptation to external change that implies changes of existing structures but not underlying beliefs. As the level of learning increases, the extent of change also increases (Pahl-Wostl 2009: 359). It is assumed that the quality of learning has furthermore an impact on vulnerability. This assumption is based on the finding that higher levels of learning indicate that a system is able to encounter non-linear external changes in a more flexible manner (cf. Argyris and Schön 1978; Pahl-Wostl 2009; Löf 2009).

Finally, factors that supposedly support collective learning in environmental governance systems are summarized in order to explain how change in the highlighted components of these systems is interpreted within the framework of analysis. Major factors constraining collective learning in governance systems are therefore centralized systems, rigid bureaucracies, poor access to information by decision makers and the population and a lack of vertical integration (Pahl-Wostl et al. 2007; Mostert et al. 2007; Huntjens et al. 2008; Pahl-Wostl 2009). These assumptions are based on multiple empirical studies mainly in European countries and are an important input of Pahl-Wostl’s framework of analysis (Ibid 2009). Based on these assumptions important conclusions on how learning is shaped within environmental governance systems have been drawn.

A central weak point in organizational learning theory and in general learning theory is that concepts tend to be quite blurry and overlapping ( For the purpose of the present study a relatively limited range of concepts of organizational learning was considered in order to avoid over-complexity, which leads to blurred conceptualities.

It is important to notice that all of the theories, models and concepts introduced in this section are not entirely new and have been considered in the research of systems theory, organizational theory and other fields of research before. Climate change and more particularly adaptation to climate change opened new channels of applying this existing knowledge to a new context, whilst some areas have been highlighted or regarded under a new perspective. A reconsideration of existing knowledge and its application to new contexts can be an important way in order to achieve advanced systems and modes of governance.

1.3 Measuring Collective Learning Processes and Vulnerability

Inherently, this study seeks to substantiate the nexus between vulnerability and collective learning processes. It thereby lays open in what particular ways vulnerability is influenced by collective learning processes. Starting point of the analysis is the paradox that developing countries with low levels of physical exposure to flood hazards tend to be more vulnerable than developing countries suffering from higher levels of physical exposure. The underlying analysis goes beyond the plain assessment of social, economic and geographic indicators that may determine vulnerability. It does so by focusing on the actual processes that decide how vulnerable two different societies are to extreme flood events.

Before attempting an operationalization in the area of flood disaster management it is important to highlight some obstacles in this respect. Social-ecological systems are characterized by a high degree of complexity, which is even getting increased if it comes to natural disasters (Birkland 2006). This in turn means that each attempt to convert such systems into a model or theory has to accept a high degree of simplification. In order to avoid over-simplification though, it is important to reflect empiricism in the models and highlight limitations in order to further enhance their applicability. This study majorly works deductive by testing already achieved assumptions and theoretical frameworks in two empirical case studies using comparative method.

Applying learning theory to social-ecological governance systems implies the difficulties described in advance. The present study tries to provide a more holistic picture of collective learning by using an analytical framework that is based on theory as well as empiricism. An analysis of collective learning patterns in socio-ecologic systems may therefore not be able to cover aspects of the individual level. In the case of collective learning in flood management this means that not the single learning of a member in a community is in the centre of observation. In turn, informal networks are considered in this analysis and in this way can balance the fading-out of the individual level to some degree (cf. Pahl-Wostl 2009).

A key challenge of the present study is to make collective learning processes detectable because not every change means that collective learning processes have been taken place (Newig et al. 2010: 8). In order to meet this challenge, the following section aims to conceptualize learning processes in social systems following the conceptual framework of Pahl-Wostl (2009). The social systems of concern are resource governance systems which are characterized by their institutions, actors, multi-level interactions and governance modes (Ibid.: 356). In the scenario of reoccurring extreme weather events these systems are exposed to an unpredictable external stressor. Climate change further increases the uncertainty about this hazard for it causes increasing frequency and magnitude of extreme weather events. In such a scenario adaptation to the external stressor appears to be a logical reaction.

Adaptation is therefore closely interlinked with learning processes within the present framework for it analyzes learning processes that foster lower levels of vulnerability. Learning in the context of adaptation relates to a system’s ability to adapt to a changing environment, which is referred to as adaptive capacity in this conceptual framework. Adaptive capacity is understood as “[…] the ability of a resource governance system to first alter processes and if required convert structural elements as response to experienced or expected changes in the societal or natural environment.” (Pahl-Wostl 2009: 355).

However, adaptation cannot be regarded as an automatism that begins as an external stress starts interfering and causes capacity-building to adapt in future. Before planned adaptation can take place at national level, the interplay of a stimulus and how it is perceived and the political commitment (cf. Thompson and Gaviria 2004: 53) decide whether or not a system starts reacting as a whole. The nature of the stimulus can have some influence on the commitment of a system to react. Single natural disasters of high magnitude for instance tend to have higher influence on the agenda setting process than a series of similar events with lower magnitude (Birkland 2006: 19), which is a crucial factor in the Pakistan case study. Though in general it is almost impossible to predict at what point a political system will start reacting to an external stress or whether it reacts at all. Accordingly, adaptation and therefore learning can only be detected but not predicted in the current state of knowledge.

Uncertainty is a crucial factor within this analysis for it creates obstacles that can barely be overcome by traditional means and therefore demands rethinking existing forms of collective action (Duit and Galaz 2008). Complex phenomena like extreme weather events create high levels of uncertainty especially because they are assumed to increase in intensity through climate change. Under these prospects habitual means are hardly sufficient anymore in order to manage risk. In order to maintain its status quo or even benefit from changes, a governance system needs to change its own structures and modes of coordination. These particular changes are identified as collective learning processes within this conceptual framework.

So far, the environment of collective learning in resource government systems has been considered, which is now followed by its operationalization. Analyses of collective learning patterns can focus on the process as well as the products of learning (Gerlak and Heikkila 2011: 3). Accordingly, the present study uses the following strategies in order to “measure” collective learning processes:

i. Out of existing empirical and conceptual studies identifying factors that foster learning. Based on these factors drawing conclusions to the process of learning in the case studies chosen.
ii. Focusing on potential outcomes of learning processes^[15] that serve as indicators for successful completion of collective learning processes.

Outcomes of collective learning processes in governance regimes are not measured by their relative success since experimental learning like by trial and error can be an indicator for learning processes (cf. Fazey et al. 2005; Gerlak and Heikkila 2011). The present study supports the notion that collective learning processes, also if they imply changes that are not successful^[16], in long-term perspective support the process of adaptation and thereby make systems less vulnerable to hazards.

Changes in governance regimes are of different impact and can therefore be attributed to different levels of learning. Pahl-Wostl attributes changes in governance regimes to different levels of learning in a matrix. Therefore all four major features of governance regimes are taken into consideration. The present study will analyze the empirical case studies following the criteria established in this matrix.

Institutional changes are the first important indicator of learning processes since they imply that attention has been directed to a certain problem and new ways of handling it have been considered (cf. Table 2).

illustration not visible in this excerpt

Table 2: Institutional Changes in Governance Systems

Source: Pahl-Wostl 2009: 360

Participation of different stakeholders is a key feature in order to encourage collective learning processes that aim on increasing resilience in resource governance systems (cf. Kilvington 2005; Pahl-Wostl 2007; Newig et al. 2010). Even though broad participation in different stages of the policy process is not a guarantee for successful learning, it is a prerequisite for multiple resources of information and communication between key stakeholders (cf. Sabatier and Jenkins-Smith 1993; Cooney and Lang 2007; Pahl-Wostl 2009). Not only the variety of actors decides whether collective learning takes place but also the roles the actors and how they develop and transform over time. Accordingly, changes in actor constellations are considered another indicator for collective learning processes (Pahl-Wostl 2009: 357; cf. Table 3).

illustration not visible in this excerpt

Table 3: Changes in Actor Networks in Governance Systems

Source: Pahl-Wostl 2009: 360

Collective learning processes furthermore have impacts on how participation is formalized and established within a resource governance system. Changes towards a formalized participation of actors from multiple levels further indicate higher levels of collective learning like double- or triple-loop learning (Pahl-Wostl 2009: 357-8; cf. Table 4). The vertical coordination between national, local or international actors therefore stands in the focus of analysis.

illustration not visible in this excerpt

Table 4: Changes in Multi-Level Interactions in Governance Systems

Source: Pahl-Wostl 2009: 360

Collective learning processes influence the ways actors interact in governance systems. Modes of governance can be hierarchic like bureaucracies or non-hierarchic like markets or networks (cf. Mayntz 2004; Pahl-Wostl 2009; Börzel 2010). Within this conceptual framework diversity in governance modes indicates higher levels of collective learning (cf. Table 5).

illustration not visible in this excerpt

Table 5: Changes in Governance Modes in Governance Systems

Source: Pahl-Wostl 2009: 360

Besides changes in the major components of governance systems, it is also important to consider how uncertainty is treated within environmental governance systems in order to draw conclusions on the developments in the treatment of risk. High levels of learning are achieved when a governance system accepts uncertainty and starts managing risks in a more flexible manner (Table 6).

illustration not visible in this excerpt

Table 6: Changing Concepts of Uncertainty

Source: Pahl-Wostl 2009: 360

Single-loop learning is rather reactive and therefore less suitable for anticipatory adaptation to non-linear stressors like climate change-related natural hazards (cf. Duit and Galaz 2008). Still, single-loop learning is an important prerequisite in order to improve towards higher learning levels (Pahl-Wostl 2009: 359). Within the present study, multiple-loop learning is considered to be the most suitable in order to meet the challenges imposed by climate change-related natural disasters. Single-loop learning is regarded rather critically, since it does not change the ways a system operates even if it experienced frequent failures in the past.

Parallel to collective learning processes in resource governance systems multiple other processes take place within the society as a whole. Processes of socio-economic development and technological advancement are drivers of change in every society and therefore have to be considered in this analysis. However, they need to be distinguished from each other in relation to collective learning processes. Collective learning and socio-economic development are strongly interlinked, since learning can be either prerequisite or outcome of this process (cf. Cimoli and Dosi 1994). It is therefore considered an interfering variable within the present study. In contrast, the use of advanced technologies is only considered a result of collective learning processes in the sense that a system decides to develop or introduce new technologies. The introduction of new regulative river embankments, for instance, demands the reconsideration of old technologies like earthen embankments without regulative capacity (example derived from Wester and Bron 1998). Therefore technological advancement is not regarded as interfering variable but also as a product of successful collective learning.

The criteria summarized in the tables above will be applied to the two case studies based on a literature research. In order to analyze learning as a process, in both examples flood events, which had major impacts on learning processes will be considered. These flood events will be analyzed with regard to how the main five features of governance regimes have changed in the course of time. The two case studies will be compared diachronically so the processes of learning can be followed in each case study in a more consistent manner. Finally both learning processes will be compared and conclusions will be drawn how they influenced the vulnerability of each country.

Major source of information within this framework is a literature research that was conducted in printed literature via the Gemeinsamer Bibliotheksverbund (GBV) catalogue. Another major sources of information are internet resources like online journals on the online pages of Springer, ScienceDirect, Elsevier, SAGE Publications or Wiley Online Library; online pages of organizations like the IPCC, UNISDR, UNDP, OECD, World Bank, UNEP, WHO or the FAO; online pages of various institutes and research projects like the Tyndall Centre for Climate Change Research, the Stockholm Resilience Centre, PreventionWeb, ReliefWeb, Asian Disaster Preparedness Centre or the Sonderforschungbereich 700 of the Freie Universität Berlin; and finally online pages of NGOs working in the field of disaster relief in Asia like ActionAid, British Red Cross and The Red Crescent. A partial review of laws and policies was undertaken in order to achieve a complete picture of the developments in both case studies. Furthermore, databases like the International Disaster Database and the Human Development Index have been used.

In order to assure a balance in research results a set of keywords was used in each stage of the analysis (cf. Table 7). These keywords derive from a previous sample research and got adjusted permanently during the research process.

illustration not visible in this excerpt

Table 7: Major Keywords Categorized by Stage of Research

Source: Author’s design

The basic structure of the present study is oriented to a scheme deriving from of the concept of social vulnerability, namely the Pressure and Release Model (PAR) by Blaiki et al. (1994). This model composes out of the vulnerability of a system caused by socio-economic pressure and its exposure to an external hazard. The PAR is especially eligible because it regards a hazard as a result out of both the system itself and the actual hazard. Furthermore, the process of adaptation that can be regarded as a continuation of the PAR will be considered. Learning processes, which are the crucial subject to study in this study, are assigned within the process of adaptation within the concept of adaptive capacity. Hence, the basic structure of this study composes out of the governance regime, the actual hazard and their interaction in the form of adaptation.

1.4 Case Study Selection

In order to test the theoretical framework and to gain further knowledge about how collective learning structures and processes influence vulnerability, two empirical case studies will be compared. Beforehand, the choice of scenario will be laid open in greater detail in order to explain why developing countries and floods have been chosen in order to address the issue of adaptation to global climate change.

1.4.1 Why Studying Floods in Developing Countries?

The largest proportion of people that are affected by natural disasters lives in developing countries (UNFCCC 2007: 5-6). It is therefore of great importance to identify the challenges imposed by natural disasters to developing countries, especially the poorest ones. In order to prioritize countries according to their vulnerability to climate change, studies have been conducted that concentrate on predictive indicators of vulnerability mostly to natural disasters (cf. Blaikie 1994; Adger et al. 2004; Brooks et al. 2005). Indicators identified in these studies match with many indicators of common economical and social development (e.g. Adger et al. 2004: 92) meaning that developing countries, if compared to industrialized countries indicate higher levels of vulnerability. It is important to notice though, that large-n country comparisons on determinants of vulnerability among developing countries are rare in the current literature and therefore it is difficult to draw conclusions on what factors explain the differences among developing countries. Furthermore, role models like the Cuban hurricane disaster management^[17] compared to that of the USA in 2005 demonstrate that the degree of development does not always correlate positive to the levels of vulnerability (cf. Thompson and Gaviria 2004). In summary, research in disaster management tends to be biased towards a Western Universalism^[18] that fades out the particular circumstances in other countries, stating that being the way developing countries are makes them what they are. The present analysis will therefore focus on developing countries and their special challenges of adaptation.

Hydrological hazards account for more than half of all disaster occurrences on average in the past decade and in 2010 for 87 per cent of all disaster victims world-wide (Guha-Sapir et al. 2011: 22). Therefore, floods are particularly in the field of interest for further investigation. The present study exclusively analyzes flood events as examples of natural hazards also because they are very convenient to explain how complex governance systems in the field of disaster management function. The advantage of analyzing floods is that they require involvement at multiple levels in order to be managed successfully (cf. Pahl-Wostl et al. 2007: 10) and therefore can make adaptation and collective learning more likely to occur. Also, in environments with multiple stakeholders collective learning is more likely to evolve due to the influence of diverse information sources (cf. Haas 2004: 7).

The selection of the basic scenario in this analysis, namely extreme flood events in developing countries, was therefore done in accordance to conceptual and practical criteria. Climate change, even though it is a key issue within the scope of this study, will be addressed indirectly through the extreme weather event of flooding. It is therefore acknowledged that efforts have been recently made to address climate change risk reduction as a holistic concept like in UNDP’s National Adaptation Programme of Action. However, concentrating exclusively on policy on climate change adaptation would fade out the previous efforts, which have been made within the field of management of single hazards.

1.4.2 Comparative Method and Case-Studies

In order to illustrate differences in collective learning patterns among countries and explain how these differences have an effect on their vulnerability to flood hazards, the present study uses comparative method. This comparison is oriented to Mill’s Design of Difference (Mill 2009 [1872]), where one phenomenon y is tested in two cases that share similar features x1 and x2 excepting for one x3 (Table 8). This particular circumstance exists in case one while it is missing in the second case. If phenomenon y occurs in case one and does not occur (y0) in case two, this might be an indicator for a certain degree of correlation between y and x3.

illustration not visible in this excerpt

Table 8: Mill’s Design of Difference

Source: Author’s design deduced from Mill 2009 [1872])

The present comparison, however, is not a clear-cut example of the Most Similar Case Design because the phenomenon y cannot completely be absent in the present analysis. Within this design the phenomenon y equals a country’s vulnerability to floods. The absence of vulnerability to floods would imply that a society does not experience floods since a hundred percent resilience is hardly achievable. But societies without any exposure to floods are not considered within this framework. Hence, y here is considered a gradual phenomenon that cannot be completely absent, that is to say vulnerability is in first case high and in the second one lower. The flexible factor in this design (x3) shall be the presence and quality of collective learning processes that is expressed in the learning layers framework developed by Pahl-Wostl (2009). Similarly to phenomenon y, x3 is expected to be not totally absent but weaker^[19] in states with high relative vulnerability.

Vulnerability is measured and weighted under the main aspect of human physical security. Major measurement variable is therefore called relative vulnerability (UNDP 2004). This variable consists of the numbers of deaths directly caused by a specific hazard in relation to the numbers of people, who have been affected by that particular hazard. It therefore an outcome- oriented variable, which indicates how successful a country is adapting to this particular hazard.

The first case study was chosen by taking an extreme case of a flood-prone developing country (cf. George and Bennett 2005; Gerring 2009). It is thereby suggested that if the underlying approach of this study and the connected assumptions of learning are valid, collective learning processes are very likely to evolve in this case and should therefore be detectable. South Asia was chosen as a focus area, since this region inhabits the countries with largest populations that are exposed to flood in relation to their total population world-wide (UNDP 2004: 40-41). Floods are very typical phenomena in South Asia due to its very large and dispersed river systems that cross many borders in this region (Mirza et al. 2003). India and Bangladesh are the two countries experiencing exceptionally high levels of physical exposure to floods (Graph 1) within this region. For the present investigation Bangladesh was chosen as the most convenient case study, since it indicates the highest rate of physical exposure to floods per capita within the region while indicating low levels of relative vulnerability compared to other countries in the region (cf. Graph 1).

Taking Bangladesh as point of reference, another country in the South Asian region with comparably high relative vulnerability needed to be identified. It is necessary to look for a second case within the same region since the geographical and cultural background of the two cases needs to be similar for these are two determinants deciding the nature of the flood hazard and how a society responded to them in absence of a central authority. This in turn ensures that the two examples have a common starting point from where they commenced their development up to today. Furthermore, the two countries need to be similar in the interfering variables that have been identified by the UNDP`s Disaster Risk Index Project as common determinants of vulnerability to floods namely the GDP per capita, population density and physical exposure (UNDP 2004: 3). In addition the project suggests that high levels of corruption are an indicator of high levels of vulnerability. Socio-economic development is another very important variable to consider within the analysis of learning processes in disaster risk management (cf. Adger et al. 2004). Higher levels of development are thereafter linked to lower levels of vulnerability.

illustration not visible in this excerpt

As second case study Pakistan was chosen for several reasons. Firstly, Pakistan has a considerably high level of relative vulnerability compared to its low level of physical exposure to floods (Graph 1) and therefore applies as counterpart of Bangladesh concerning Mill’s Design of Difference. In terms of geography Bangladesh as well as Pakistan is part of an enormous stream system originating from the Himalayas. Annual flooding caused by snow-melt in spring and monsoon season are a typical phenomena in both countries, whereas ordinary floods have an important economical value for they leave behind fertile soil caused by sediments carried by the rivers (Ahmad et al. 2004: 45). Therefore, the hazard patterns in both countries are very even and like the South Asian Floods of 2007 there are certain flood events affecting both countries at a time.

A considerably important similarity between both countries lies in their cultural background that draws back to a shared history. Indigenous disaster management on community level plays an important role in the South Asian region and is supposed to be the dominating method from early history (SDMC 2008: Ch.1, 1-2). In British India rulers handed power over local water management to local elites in the provinces of Punjab and Bengal, the future territories of Pakistan and Bangladesh, whereas these structures prevailed even some time after independence (cf. Mustafa 2001: 821; Bildeng et al. 2008: 1; Chatterjee 2010: 133). From 1947 until 1971 Bangladesh was part of Pakistan called East Pakistan. The development of disaster risk strategies in both territories remained very insufficient up to the point where the considerably neglected Province of East Pakistan successfully fought for its independence. This is the starting point where the development of disaster risk management in both territories started to diverge.

The three major interfering ^[20] variables identified by the UNDP (2004) are assumed to have a causal direction as follows: The lower a country’s GDP per capita and local population density in flood affected areas; and the higher a population’s physical exposure the higher is its vulnerability or associated risk (Ibid. 2004: 3). Comparing Bangladesh and Pakistan this trend is not consistent. Whereas both indicate very high levels of population density in flood affected areas (UNDP 2004: 143-45), their GDP per capita and physical exposure differ. Pakistan’s GDP per capita is actually double as high as the one of Bangladesh, which has been a quite consistent trend during the past thirty years (The World Bank 2011a). This should imply that Bangladesh is more vulnerable to flood disasters. In reality though, Bangladesh’s relative vulnerability has been significantly lower than that of Pakistan as indicated before. Furthermore, the Pakistani population experiences significantly lower levels of annual exposure to floods than Bangladesh (UNDP 2004). Again, the causal direction of the indicators does not apply to this country comparison. The corruption levels of both countries are considered to be very high (Bertelsmann Stiftung 2009a, Bertelsmann Stiftung 2009b) the level of vulnerability to floods should therefore be equally affected by this factor. It is assumed here that the indicators of vulnerability to floods identified by the UNDP’s project are rather suitable for large-n country comparison than regional or interstate comparison.

In order to assess and compare the level of development of both countries, the Human Development Index (HDI) was taken into consideration. It is a three dimensional tool measuring a variety of indicators in the fields of education, health, and living standard. Currently, Bangladesh and Pakistan are at a similar level of human development taking in ranks 129 and 125 of the country ranking list. This connects to the trend in HDI of both countries of the past thirty years (Graph 2) indicating that the development of the two countries shows great similarities in improvement with a gradual assimilation. Accordingly, the process of socio-economic development is a controlled variable in the analysis of both countries.

illustration not visible in this excerpt

Graph 2: HDI of Bangladesh and Pakistan by comparison from 1980-2010

Data Source: UNDP 2010

According to most factors, which have been identified in the literature and their causal direction, Pakistan should be less vulnerable to floods than Bangladesh. But the relative vulnerability to floods in Pakistan is more than double as high as in Bangladesh^[21] (UNDP 2004). This indicates that in the case of Bangladesh and Pakistan there must be another factor, which has not been considered yet that plays a vital role. Collective learning patterns are the main suggestion of this study in this concern as it supposes that collective learning processes potentially decrease vulnerability to floods in a country. This case study selection is expected to make the learning processes in the field of flood disaster risk management visible and comparable and separate it from other influences that might be causes of change in the field.

2 Bangladesh - Learning How to Life with Extreme Floods

Bangladesh is one of the poorest and most disaster-prone countries in the world with an extremely high population density. As a result, it has comparably high numbers of people affected by disasters each year. Within media Bangladesh is often referred to as the most vulnerable country to natural disasters and climate change world-wide. The underlying measurement of vulnerability behind this statement is not always clarified and often the assessment of vulnerability does not differentiate between different types of hazards. Still, it is important to differentiate and relativize such statements because, as indicated before, vulnerability is measured through diverging scales.

Major reoccurring climatologic disasters in Bangladesh are floods, droughts, and cyclones whereas cyclones are the most destructive natural force in Bangladesh in terms of human losses. From 1980-2000 cyclones accounted for more than 7.450 annual deaths in Bangladesh compared to more than 450 annual deaths due to floods (UNDP 2004: Table 4 and 5). Even though there are severe problems in the cyclone risk management, major improvements have been made regarding early warning procedures (Paul 2009: 289-90). Drought and flood management, in turn, draw back to a comparably more successful history over the past thirty years. Disasters as a result of droughts have been successfully tackled by a combination of programs and legislature, which aimed on food security. The Famine Codes of 1880 introduced by the British colonialists are an important example in this respect (Bildeng et al. 2008: 14). The governance of flood management lies the centre of focus of this chapter for this case study is assumed to lay open characteristics of collective learning.

[...]

^[1] Physical exposure refers to the average number of people that have been exposed to floods form 1980-2000 (UNDP 2004).

^[2] Relative vulnerability refers to the number of fatalities (annual average from 1980 until 2000) in one million people exposed to floods (annual average from 1980 until 2000) in a country (UNDP 2004).

^[3] External stress is used here as a neutral term.

^[4] In this study, papers focusing on adaptation, vulnerability or resilience are referred to as adaptation literature. Other designations like resilience literature (eg. Loef 2010) for instance are also used, referring to a similar field of study.

^[5] Practically-oriented literature refers to papers that primarily focus on concrete solutions in order to adapt to climate change or measuring vulnerability.

^[6] Theoretically-oriented literature refers to literature that focuses on structures and “behavior” of systems that lead to adaptation or steer adaptation.

^[7] The term ‘state’ in this context describes the organization of a political system for the purpose of self-construction and self-sustaining in a society.

^[8] This translation was derived from Risse and Lehmkuhl (2006: 7).

^[9] The SFB 700 (2009) distinguishes between the provision of rules and collective goods as common outputs of governance, which coincides with outputs of government. Differences derive from the ways these goods are provided, which have nevertheless also influences on the actual outputs.

^[10] Flood management is part of the water sector. Therefore, the present analysis mainly focuses on a resource governance regime (cf. Pahl- Wostl 2009), which is referred to more broadly as environmental governance system in the following sections.

^[11] Based on Pahl-Wostl’s framework of analysis these four dimensions of governance systems will be the leading concept of analyzing structures and changes within environmental governance systems (Ibid. 2009).

^[12] Physical capacities here refer to technologies, manpower and the financial resources in order to employ these two. Mental capacities refer to managerial skills, knowledge generation and appliance and more specifically learning skills.

^[13] For a sophisticated overview of theories and models of learning Blackmore is very recommendable (Ibid. 2007:520-23).

^[14] New knowledge refers to newly induced studies or studies that have not been considered before for instance. New experiences can be previous failures of policies or external experiences like best practice examples that are relevant for the system of concern.

^[15] These outcomes are changes detected in the four major components of the environmental governance systems of Bangladesh and Pakistan.

^[16] Success in this study is measured by the ability of a country to reduce its relative vulnerability.

^[17] The Cuban government developed a disaster management plan, which proofed to be very successful even when facing high magnitudes of storm with only minor human losses and a fast recovery period (cf. Thompson and Gaviria 2004). In contrast, the great difficulties experienced by the USA in the Hurricane season of 2005 indicated that much more needs to be done in order to manage risk more effectively (BBC News Online 20.10.2005, available at http://news.bbc.co.uk/2/hi/americas/4360102.stm).

^[18] This term was used by Samuel P. Huntington in his work “Clash of Civilizations” (Ibid. 1996, 138) and defines here as a normative point of view that classifies a certain set of values attributed to a number of industrialized countries as being the most desirable world-wide.

^[19] Weaker therefore means learning is taking place at lower learning levels after Pahl-Wostl’s (2009) Model.

^[20] All values indicate the annual averages from 1980-2000. Pakistan’s relative vulnerability to floods (22,84) is therefore more than double as high as the Bangladeshi (10,96). Relative vulnerability indicates the number of fatalities per Millions exposed to a specific hazard.

^[21] From 1980 until 2000 in Bangladesh on average in every one million people affected by floods about 11 people died. In comparison in Pakistan it was about 23 people dying per one million affected (UNDP 2004: 144-46).