Crisis Governance: a Networked Enterprise? Case studies of crisis response in the Netherlands

(1)

Crisis Governance: a Networked Enterprise?

Case studies of crisis response in the Netherlands

Photo by Alina Grubnyak (Unsplash)

Jeff Luimes

Supervisor: Dr. J.J. Wolbers Second reader: Dr. L.D. Cabane Master thesis: MSc Crisis and Security Management Leiden University. Faculty of Governance and Global Affairs

(2)

2

One of the lesser known stories of September 11, 2001 is the evacuation of nearly half a million people from Lower Manhattan by boat. When the twin towers came down and concrete, ash, and smoke shrouded Lower Manhattan people fled in all directions. Many people found themselves stuck at the southern tip of Manhattan close to the water’s edge. It was impossible to leave the island because roadways, bridges, and tunnels were closed down. And then something remarkable happened. The boat operators who at that moment were close to lower Manhattan could see the people piling up at the dock areas. Not only the ferries that had just dropped of people from Staten Island, but also random boats turned around. They started to pick up as many people as possible. The United States Coast Guard made a call on the radio: “All available boats this is the United States Coast Guard for New York, anyone

wanting to help with the evacuation of Lower Manhattan report to Governors Island.” (Road

to Resilience, 2011). All kinds of boats started to arrive to rescue the citizens that were trapped. In less than nine hours the spontaneous and unplanned evacuation helped almost 500.000 people of Lower Manhattan. The story is an incredible example of the resilience of ordinary people during times of crises.

Not only is the story of the evacuation an example of resilience, but it also provided new insights into the practice of crisis response1. Public authorities are often held responsible for response efforts, but a crisis response operation commonly involves both public and private actors. When the water evacuation occurred spontaneously the U.S. Coast Guard, harbor pilots, and the harbor police allowed for a decentralized response. The Coast Guard could have tried to take control of and impose a command structure on the evacuation

operation. However, the Coast Guard did not try to take control but actually used its capacity to facilitate the emergent operation. Moreover, existing rules and procedures were bent to allow for improvisation and problem-solving. Kendra and Wachtendorf (2016) who studied the evacuation of Lower Manhattan challenged the command and control paradigm of crisis management that is still considered widespread in the world: “these observations point to a

particular concept of disaster management: one that sees plans as tools, not scripts; one that is respectful of new and unexpected resources; one tilted more toward effectiveness than efficiency” (pp. 138-139).

The command and control notion of crisis management is rooted in military thinking. The model assumes that a crisis is characterized by social chaos. In addition, it presumes that

1_{A common conceptualization of crisis management distinguishes between the activities related to four phases: preparedness,} mitigation, response, and recovery (Kapucu, 2008). The focus of this study is on the response phase.

(4)

4

citizens have a very limited ability to cope with disastrous events. Public authorities therefore have to establish an artificial authority structure to regain control and restore order. These assumptions of the command and control approach to crisis management have been

discredited as crisis myths by research (Drabek & McEntire, 2003). The alternative approach to crisis management that has been established by disaster sociology is the emergent or

societal model. The model emphasizes continuity, coordination, and cooperation. Moreover, it assumes that there is no breakdown of social structures and people do not engage in anti-social behavior. Research has demonstrated that people and groups are capable of self-organizing and responding to disasters2. Public authorities should not centralize authority but allow for decentralized and pluralistic decision making (Dynes, 1994).

However, some principles of the command and control model still exist in the practice of crisis management nowadays. The command and control approach is reflected by the organizational design of the Incident Command Systems (ICS). The ICS is a standardized approach that imposes a hierarchy – a form of centralized coordination – on emergency response operations (Kendra & Wachtendorf, 2016). It offers a framework for

multiorganizational crisis response operations (Moynihan, 2009). The ICS was developed in the United States (U.S.) in the 1970s and similar models have been adopted since in other countries throughout the world (Boersma, Comfort, Groenendaal & Wolbers, 2014). However, research has demonstrated that centralized coordination of the responders at the frontline is not a requirement for effective crisis response (Groenendaal, Helsloot & Scholtens, 2013; Wolbers, Boersma & Groenewegen, 2018).

The hierarchical and bureaucratic approach that remains embedded in contemporary crisis management has been challenged in the recent two decades from another perspective: network governance. This perspective has emerged from the field of public management, which offers a vast amount of research and insights on networks that can be applied to crisis management (Nohrstedt, Bynander, Parker & ‘t Hart, 2018). This perspective emphasizes the importance of collaboration across organizational and jurisdictional boundaries. For example, Waugh and Streib (2006) argue that: “Collaborative networks are a fundamental component

of any emergency response. It is also a mistake to assume that any individual or organization can manage all the relief and recovery eﬀorts during a catastrophic disaster.” (p. 134).

Similarly, Ansell, Boin, and Keller (2010) emphasize the transboundary dimensions of crises and disasters. They point out that response efforts to a transboundary crisis are complicated by

(5)

5

the challenges of inter-jurisdictional and inter-sectoral coordination. At the same time, however, the scholarly community has noted that relatively few studies have examined crisis management from a network governance perspective (Kuipers & Welsh, 2017). In particular, little effort has been made to conduct empirical analysis of the aspects of network governance that help to understand effective crisis response (Nohrstedt, Bynander, Parker & ‘t Hart, 2018). The thesis therefore aims to address the following research question: How does the

mode of network governance influence crisis response effectivity? It is hoped that this thesis

will make an important contribution to the field of collaborative crisis management by conducting empirical research.

This thesis uses a multiple case study approach to help understand how the mode of network governance influences crisis response effectivity. One major advantage of using a multiple case study approach is that it provides evidence from different empirical settings. A total of six crisis response operations in the Netherlands have been selected. This research therefore offers insight into the crisis response practice in the Netherlands. Three different types of crisis are studied: industrial incidents, shootings, and wildfires. These types of crisis have been recurrent phenomena in recent years. In April this year, a wildfire in the South of the Netherlands became the largest fire in the national history (AD, 2020). This wildfire did not stand on its own in the sense that it was one of many large wildfires in the recent decade (Donkers, 2019). Furthermore, the threat of terrorism has been and remains a serious topic of concern in the Netherlands (NCTV, n.d.). In addition, industrial incidents have been relatively common in recent years, which can for example be illustrated by the series of incidents at the industrial complex Chemelot that have been reported frequently by the media (AD, 2015; Ebisch, 2020; Peeters, 2019; Trepels, 2018). These crisis types are therefore much in focus of the Dutch authorities responsible for crisis management. For each crisis type two cases have been selected. The wildfires studied are the fire at the Strabrechtse Heide in 2010 and the fire at national park the Hoge Veluwe in 2014. The two fires were both exceptional in size. The industrials incidents that have been selected are the fire at Chemie-Pack in 2011 and the fire at Chemelot in 2015. In both cases the crisis was dealt with properly at the operational level, but there were several shortcomings at the strategic level. Finally, the cases of the shootings are Alphen aan den Rijn 2011 and Utrecht 2019. In both cases there was a single perpetrator and a significant period of uncertainty after the shooting.

Finally, this study is both scientifically and socially relevant. It aims to contribute to a particular knowledge gap in the literature, namely the lack of empirical research on

(6)

6

this research contributes to the understanding of network governance in different empirical settings. Importantly, it is also socially relevant because hierarchical and bureaucratic

approaches are still widely used throughout the world as frameworks for crisis response. The network perspective might yield new knowledge about response efforts to crises and disasters. In addition, the study can potentially identify shortcomings of and improvements to crisis response. The activities of response efforts to crises and disasters are crucial because they concern the protection of people, minimization of damage to property, and attempts to restore stability.

(7)

7

Chapter 2: Theoretical Framework 2.1 Introduction

A large and growing body of literature has investigated and tried to answer the question of what contributes to a successful crisis response. Research into this particular subject has a long history. Already in the 1960s Dynes challenged the conceptual roots of crisis

management, which are traced back to military thinking (Dynes, 1970; Quarantelli, Dynes & Haas, 1966). He argued that the assumptions embedded in the military models for disaster response did not adequately match the reality of a crisis. Rather than a situation of social chaos, Dynes explained that a crisis can be characterized by continuity. Accordingly, governments should not adopt policies based on command and control, but on coordination and collaboration (Dynes, 1994). Nowadays debates about these two competing schools of thought still dominate the scholarship on crisis management. However, over the past two decades research and practice has gradually shifted from an emphasis on centralized command towards a collaborative approach (Hu, Knox & Kapucu, 2014; Waugh & Streib, 2006).

This shift has been linked to the complex and transboundary nature of modern crises. A crisis can be understood as “a threat to core values, the safety of people, or the functioning

of critical infrastructures that must be urgently addressed under conditions of deep

uncertainty” (Boin & Bynander, 2015, p. 124). This definition combines the core conceptual

elements of a crisis: a severe threat, time pressure and high uncertainty (‘t Hart, Rosenthal, Kouzmin, 1993). Ansell, Boin, and Keller (2010) refer to a transboundary crisis as one that crosses political, functional, and time boundaries. Crises can cross borders and impact multiple cities, regions, and countries. In other words, they do often not fall neatly within the jurisdiction of a single political actor. Scholars distinguish between vertical and horizontal boundaries. The former refers different levels of government: local, regional, and national. The latter refers to a crisis which crosses boundaries between two political actors at the same level. For example, a large wildfire that affects multiple neighboring cities. A crisis can also cross functional borders, thereby threatening multiple policy areas. For example, the

Deepwater Horizon oil spill quickly became a disaster that crossed from the private to the public realm. And finally, crises often cross time boundaries. This means that they do not have a clear starting point and ending. Consequently, a transboundary crisis is more difficult to manage because more participants become involved. Naturally, crises can differ in the extent to which they affect multiple jurisdictions and sectors. However, it is clear that a disaster response requires the collaboration between a myriad of actors: national, regional,

(8)

8

and local departments; public, private, and non-profit organizations. For example, Bier (2006) describes that many of the problems in the aftermath of Katrina related to the coordination between multiple levels of organizations and multiple levels of government. It has thus been established that complex disasters cannot effectively be managed by a single organization that relies on command and control (Kapacu & Garayev, 2016; Nowell, Steelman, Velez, & Yang, 2018). Because crises require by their very nature collaboration between multiple actors, a crisis response can be understood as a networked enterprise (Boin, ‘t Hart & Kuipers, 2018; Kapucu, Arslan & Collins, 2010; Moynihan, 2009). More recent attention in the literature has therefore focused on network approaches towards crisis management.

In the following sections the theoretical framework will discuss what a network approach towards crisis management entails. The overall structure of the framework takes the form of three sections. The first section will briefly look at the origins of network theory in the public management literature. The second part includes a discussion of three aspects that are crucial in understanding the network governance of crisis response. The final section discusses how crisis response performance can be conceptualized.

2.2 Origins of network theory

In the field of public management, network theory has become a full-fledged theory (Klijn & Koppenjan, 2012). Networks can be defined as “structures of interdependence involving

multiple organizations or parts thereof, where one unit is not merely the formal subordinate of the others in some larger hierarchical arrangement” (O’Toole, 1997 p. 45). Networks are

distinguished from market and hierarchical structures. (Powell, 1990; Provan & Kenis, 2008). In hierarchical structures information and decision-making flow from the authoritative

positions at top down to the bottom of a system. Market structures promote participants to use their resources to achieve their self-interests (Fisk, Good & Nelson, 2019). It is important to introduce a caveat regarding these distinctions. These forms are ideal types that can be clearly distinguished in theory. However, empirical research has illustrated that this clear distinction is not always evident. In practice, there can be considerable overlap between forms of

hierarchy, market, and network (McGuire, 2006). Klijn and Koppenjan (2012) identify in a comprehensive literature review four common concepts and assumptions that underpin the theory on networks: (1) Actors, interdependency and frames: A network consists of multiple actors that mutually rely on each other for the delivery of a policy and service. However, actors have different perspectives on problems and solutions; (2) Interactions and complexity: The interdependencies between actors and the variety of frames lead to a complex process of

(9)

9

interactions and negotiations. The outcome of public policy and services are attributed to the interactions between multiple actors, rather that the work of a single actor; (3) Institutional

features: Institutional relations are created as a result of the interaction patterns. These

institutional relationships are characterized by two features: patterns of social relations and patterns of rules; (4) Network management: The complex processes require network management. This consists of facilitating interactions, exploring content, and organizing interactions between actors. The management is horizontal in nature (pp. 588- 591). Klijn and Koppenjan suggest that these generic concepts and assumptions can be applied in different sectors and to different issues. In the next paragraphs, the theoretical framework will discuss how the literature on crisis management has adopted and incorporated network theory.

2.3 The network governance of crisis response

Over the past two decades, a number of studies have begun to examine crisis management from a network perspective. As many of these authors point out, the usefulness of the network approach is that is acknowledges that crises “are outside the scope of any one given agency

and require multiorganizational and multisectoral interaction for successful policy solutions and implementations” (Kapucu et al., 2010, p. 223). In other words, during a crisis actors

cannot remain independent and have to collaborate with each other. Moreover, a network perspective emphasizes that collaboration is imperative for effective decision-making in the environment of disasters. Collaborative decision-making is defined as “the process of

decision-making by interdependent actors that are working toward a common goal” (Kapucu,

2012, p. 38). In contrast, hierarchical decision-making is deemed to be ineffective and

detrimental in crisis response operations. A lack of collaboration might therefore contribute to failures in crisis response operations. For example, Kapucu and Demiroz (2011) found that a significant gap between the planned collaboration and actual collaboration accounted for some of failures in the Hurricane Katrina response. The response phase of a crisis entails all activities that attempt to save lives and prevent further damage to property (Kapucu, 2008). A disaster response network consists of a variety of actors who engage in joint activities. The composition of a disaster response network is dependent on the nature of the disaster (Nolte & Boenigk, 2013). For example, Waugh and Streib (2006) observe that in response to the

Oklahoma City Bombing the response network involved hundreds of public, nonprofit, private organizations, and spontaneous volunteers. In doing network research, it is therefore important not to only focus on the official emergency agencies, but also to take into account the role of nonemergency organizations in a disaster response (Kapucu, 2006).

(10)

10

A network perspective then provides insight into the operations and processes of a crisis response by researching its network aspects. A number of studies has developed

theoretical frameworks on collaboration and networks (Ansell & Gash, 2007; Bryson, Crosby & Stone, 2006; Provan & Kenis, 2008; Thomson & Perry, 2006). Despite their particular approaches, these frameworks also share many assumptions (Bryson et al., 2015; Provan & Lemaire, 2012). Importantly, these studies suggest that there are several crucial elements that influence the functioning of networks. In what follows, the theoretical framework discusses three aspects that are key in understanding how networks function. These elements are governance, membership, and trust.

The first aspect that clarifies the functioning of disaster response networks is governance. Thomson and Perry (2006) explain that actors in a network need to establish rules that govern their behavior and relationships. Although networks are horizontal systems, some coordinating and monitoring activities are crucial for networks to survive (Bryson, Crosby, & Stone, 2006). It has been established by the literature on network governance that the design of a network affects its performance (Kenis & Provan, 2008). Networks can differ in terms of the type of actors involved, the boundaries of a network and the presence or absence of different types of links. Kenis and Provan (2008) identified three basic forms of network governance: shared governance, lead-organization governance, and network administrative organization (NAO) governance. However, in recent years more forms of networks have been introduced by studies, such as the combined lead/NAO governance form (Kenis, Schol, Kraaij-Dirkzwager & Timen, 2019). Moreover, a recent study in the crisis management literature identified a core-periphery form as an ideal-type for effective disaster response (Nowell et al., 2018). These different governance forms are depicted in Figure 1. First, a shared governance form consists of multiple organizations that work together as a network. Governance is performed by all participants rather than by a single entity. Second, in a lead organization network all actions and decisions are coordinated through and by one of the participants acting as lead organization. Network members will still hold and pursue individual and common goals. Third, the network administrative form consists of network members and a separate authority set up to manage and coordinate the network activities. The network administrative organization (NAO) differs from the lead organization because it is does not provide any services except for governing the network (Kenis & Provan, 2009, pp. 446-449). Fourth, the combined lead and NAO governance form represents a hybrid. In this case, the lead organization delegates some of its tasks to a third organization, but it retains overriding leadership capabilities. Fifth, the core-periphery form is characterized “by dense

(11)

11

connections among a central subgroup of actors at the core of the network surrounded by a peripheral set of actors with more sparse connections” (Nowell et al., 2018, p. 702). The lead

is taken by a number of organizations that will also have operational responsibilities. The network is most likely differentiated among functional subgroups, such as health, safety etc. The central subgroup may consist of one organization from each functional group (Kenis et al., 2019).

Figure 1. Forms of network governance. (Source: Kenis et al., 2019, p. 283)

This brief overview of governance types leads to an important observation about the form of crisis response networks. According to Kapucu (2012) “not all types of collaborative

networks and decision-making methods are suitable for managing complexities in disasters”

(p. 43). He and others find that decentralized networks are proper tools for dealing with complexities in crises (Kapucu & van Wart, 2006; Therrien, 1995). It is argued that decentralized networks are an effective form of governance, because of the breadth of

required capacities during a crisis response. However, Moynihan (2008b) observes that crises “also require traits unusual in networks: rapid and decisive coordination action” (p. 206). The asymmetric nature of a crisis thus leads towards a crisis management paradox. Crises require the abilities and resources that no single actor will have, because of their complex, unpredictable and infrequent nature. On the other hand, crises also require a rapid and coordinated response. He therefore argues that some form of centralization is required for effective response. In his study on the incident command system (ICS) in the U.S., Moynihan (2009) argues that the ideal types of hierarchy and network are portrayed as stark alternatives for policy tools by literature. However, the ICS has a hybrid character: it is mixture of

collaboration and hierarchy. These findings are reflected by McGuire (2006) who observes that emergency management networks often represent a blend of collaboration and hierarchy. In a subsequent study Wise and McGuire (2009) note that: “recent research suggests the need

(12)

12

networks that are completely governed by all their members (Provan & Kenis, 2008). Thus, theoretical, and empirical research both suggest that crisis response networks should be and are governed by a more brokered form of governance.

Two particular elements of governance form then explain the functioning of crisis response networks. First, the functioning of a crisis response network is dependent on the existence of collaborative linkages (Thomas & Perry, 2006; Wise & McGuire, 2009). These collaborative linkages are established and maintained by boundary spanners or brokers. Brokers are defined as “catalysts that enable and enhance the cooperation that builds and

sustains connections” (Comfort, Scheinert & Yeo, 2019 p. 16). Brokers connect the main

network actors and also the subgroups in a network to each other. Moreover, boundary spanners look out for joint interests and facilitate information exchange about constraints and resources (Nowell et al., 2018). Second, governance is made more difficult by the shared nature of authority in networks. Research notes that power either inhibits or facilitates action, because of its unequal distribution in networks. Organizations differ in the position they occupy in the network and the authority they hold. Some organizations will also be more dependent on other for resources. In other words, authority is a political and contested good in networks (Agranoff & McGuire, 2001). Kuipers and Swinkels (2019) refer to the occurrence of a fault line “when crises cross functional boundaries and tasks are not covered, and when

problems frames are set too narrow” (p. 30). This leads to discussions about responsibility in

networks or disputes between insiders and outsiders. Empirical research has found that shared authority can lead to ambiguity and disagreement in crisis management. Moynihan (2009), for example, observes that in response to the Oklahoma City Bombing there was disagreement about who was in charge and the transfer of authority. Similarly, Fisk et al. (2019) found that the recovery efforts following two large-scale water disasters benefited from clear lines of authority and clarity about the roles of stakeholders. These findings suggest that it is crucial to establish clarity about roles and responsibilities in crisis response networks.

The second element that several frameworks highlight relates to aspects of

membership in networks. Two issues are particularly important: network diversity and the network access. First, network participants from different background and cultures are confronted with uncertainty when they interact with one another. Each of the participants brings their own perspective into the network. Klijn and Koppenjan (2004) explain that institutional uncertainty arises from coordinating actors who come from different

backgrounds. Similarly, Kuipers and Swinkels (2019) maintain that clash of cultures between different organizational styles can occur in networks. Due to the different backgrounds and

(13)

13

cultures, there thus a greater risk of conflict between organizations (Moynihan, 2009; Fisk, Good & Nelson, 2019). For crisis response networks to be successful they have to overcome the structural and cultural differences of organizations (Kapucu et al., 2010). In addition, Kenis and Provan (2008) argue that a central challenge of a network is to coordinate the activities of multiple organizations. If the number of organizations in a network grows, the number of relationships grows exponentially. A crisis response network that requires a large number of responders will thus be confronted with a greater coordination challenge. Second, the literature emphasizes the importance of access to networks (Ansell & Gash, 2007). It has been established by the literature that the process of collaboration needs to be open and inclusive. This also means that troublesome, yet critical stakeholders must be included. Wise and McGuire (2009) argue that: “it is not sufficient merely to have a large number of

members in a network; the right ones must be in place” (p. 7). In the context of crisis response, a significant challenge is linked to the incorporation of new members during a crisis. Other organizations than then designated emergency agencies are often not considered in official planning. This can leave those responsible for coordination unsure how to

successfully integrate such actors in the official network (Boin & ‘t Hart, 2010; Dynes, 1994). The third aspect that clarifies the functioning of crisis response networks is trust. Trust can be understood as: “interpersonal behavior, confidence in organizational competence and

expected performance, and a common bond and sense of goodwill” (Bryson et al., 2006, p.

48). Trust building is a time-consuming process and requires long-term commitment. This is particularly true when stakeholders have poor historical relationships (Ansell & Gash 2007). It has been established that trust is relationship driven. These relationships are developed by mutual learning and action in a network (Kapucu et al., 2010). However, it is clear that the fundamental urgency of crisis leaves little time for relationship building. Trust and

relationships in emergency networks therefore need to be created before a crisis. In crisis management, preexisting trust is vital for an effective disaster response for a number of reasons. Kapucu et al. (2010) note that actors in a crisis network are willing to bypass procedures in response to unexpected needs if they have trusting relationships. Moreover, Moynihan (2009) observes that trust allows emergency responders to resolve conflict, for example by face-to-face meetings. In the response to Oklahoma City bombing, officials relied on existing social capital to facilitate working relationships. One of the officials of the city government was able to contact multiple actors in the network by calling them personally, which saved significant time. In addition, preexisting personal relationships also allowed responders to successfully incorporate emergent actors. One of the key responders was

(14)

14

contacted by a director of a telephone company who he knew personally. As a result, the responders used a location provided by the telephone company for the incident command and used several of its services. Finally, (in)effectiveness is not simply linked to the general need for trust in a network. Although it is clear that trust is a crucial for fostering an effective crisis response, it does not necessarily have to be widespread. According to Kenis and Provan (2008) the distribution – or density of trust – throughout the network is crucial. They maintain that when networks are governed by a brokered form, they can still be effective despite low-density trust. However, Moynihan (2009) maintains that in such a case there have to be positive working relationships between the core members.

2.4 Crisis response performance

The literature review set out to discuss the network governance of crisis response operations. However, one important aspect has been largely absent from this discussion so far: how do we conceptualize the performance of crisis response networks? Although there have been several studies in crisis management that focused on outcomes, there is still little agreement on how to measure performance (Nohrstedt et. al., 2018). A few studies have identified generic benchmarks that can be used to assess performance of crisis management. For example, Quarantelli (1997) identified ten criteria for good crisis management. Similarly, Harrald (2006) identified critical success factors of an organizational system that responds to a crisis. In a more recent study, McConnell (2011) notes that “there has been a lack of serious

academic debate on what constitutes crisis management success and failure” (p. 64). He

argues that success is made up of both facts and perceptions. A common approach for evaluation is goal attainment. However, this approach is problematic in the context of crises, because of vague, complex, and contradictory outcomes. Goals are not always explicitly formulated and might also be subject to change during a crisis response. Moreover,

effectiveness is often related to particular aspects of a crisis response, such as evacuation or search and rescue missions.

McConnell (2011) has developed a framework for evaluating success and failure of crisis management initiatives. He proposes a three-dimensional framework, which consists of different facets of crisis management: processes, decisions, and politics. The effectiveness of a crisis management initiative can be evaluated for each single facet. The theoretical

framework will not discuss the political dimension, because it is beyond the scope of this study. The first dimension refers to whether crisis management occurs according to the crisis-related procedures. It is assumed that these crisis-crisis-related procedures usually reflect the

(15)

15

processes that constitute ‘best practice’ for crisis management. Success is then based on adherence to processes relevant to resolving the crisis at hand. This implies that success not only refers to following the processes as specified in rules and procedures, or as McConnell puts it, sticking to the plan. The crisis at hand might actually require improvisation, which involves by-passing the emergency plans. The second dimension concerns decisions that are taken to intervene and stabilize the crisis. According to McConnell, decisions are analytically distinct from processes, because they are “separate interventions and may not conform to

what was expected or planned” (p. 67). Moreover, important decisions are not only taken at

the political-strategic level but also at the operational level. Three criteria can be used to evaluate whether crisis-related decisions have been successful and/or unsuccessful. The first concerns whether decisions help to contain or eradicate threats. Second, decisions are

successful if they help to minimize damage to people, property and any actors affected by the crisis. Third, decisions are successful if they help to restore order and stability.

It is important to address what constitutes ‘best practice’ for crisis management, because the framework of McConnell (2011) does not contain any key crisis response functions. The work of Boin, Kuipers and Overdijk (2013) provides direction in this regard. These authors have established a public crisis leadership framework, in which they identify several crucial tasks. For the acute crisis response phase, six tasks are crucial. For

convenience reasons, the six tasks are divided into the following three groups: early

recognition and sensemaking, decision making and coordination, meaning making and crisis communication. They can be grouped in this way, because the tasks in each group are closely connected to each other. The first task refers to the importance of the shared awareness that a threat has emerged. Closely related is the sensemaking task, which entails the challenge of arriving at a collective understanding of the situation. The decision making task refers to importance of crucial decisions taken at the strategic level. Importantly, these decisions as well as activities need to be coordinated with other actors across vertical and horizontal boundaries. Meaning making refers to the task of authorities to create a narrative about the crisis. In other words, the authorities need to present a convincing story about their handling of the crisis. Crisis communication refers to communication with citizens as well as between organizations. A key task is to provide the public with actionable advice.

Finally, it is important to recognize that the outcome of a crisis is not simply a success or a failure. McConnell (2011) portrays success and failure of crisis management along a continuum, which is depicted in Figure 2. Crisis management outcomes can be categorized as success, durable success, conflicted success, precarious success, and failure. There are three

(16)

16

outcomes that are in between outright success and failure. In the case of a durable success, the successes of the crisis management initiative outweigh the failures. In the case of a conflicted success there are more shortcomings, but these are fairly balanced by the achievements of a crisis management initiative. If an outcome is labeled as a precarious success, the handling of the crisis is perceived as a mishandling, which contributed to or even exacerbated the crisis. Although there are also some successes, they are outbalanced by the shortcomings. This five-fold classification clearly goes beyond the assessment of effectivity as a dichotomy. It is recognized that it is not an exact science to locate a crisis management initiative in any of the particular categories. Moreover, the dimensions discussed above can also be located

individually in one of the categories. For example, it is possible to judge the crisis

management initiative overall as a durable success, despite a particular process being labeled as a precarious success.

Figure 2. The crisis management success/failure spectrum. (Source: McConnell, 2011, p. 71)

In summary, it has been shown from this review that crisis response can be understood as a networked enterprise. By their very nature crises require collaboration between multiple actors. In general, crisis response networks are governed by a more brokered form of

governance. The studies reviewed indicate that three elements help understand the functioning of crisis networks: governance, membership, and trust. With regard to governance, the

literature emphasizes the importance of collaborative linkages. The role of brokers is to connect the network actors of the core with other actors. Moreover, the shared nature of authority within networks is likely to cause discussion and confusion about roles and responsibilities. In terms of membership, the interactions between actors from different institutional backgrounds and with different organizational styles can cause conflict. Crisis response networks have to overcome these structural and cultural differences between

organizations in order to be successful. In addition, integrating and incorporating actors other than the designated emergency agencies poses a significant challenge. These actors are often

(17)

17

not considered in official planning. In terms of trust, the literature emphasizes that trust and working relationships need to be established before a crisis. Finally, this theoretical

framework discussed the performance of crisis networks. The effectiveness of a crisis

response network can be evaluated based on processes and decisions, with a specific focus on six key crisis response functions. Finally, the outcome of a crisis is not simply a dichotomy but can be portrayed along a continuum from success to failure.

(18)

18

Chapter 3: Methodology 3.1 Research design

In order to explain how the mode of network governance influences crisis response

effectivity, I have adopted a multiple case study approach. According to Yin (2009), a case study is an empirical inquiry that: “investigates a contemporary phenomenon in depth and

within its real-life context” (p. 18). The multiple case study design is appropriate for two

reasons. First, it is useful for researching what aspects of network governance are visible in different empirical settings i.e. different types of crises. Second, a more in-depth investigation also allows to research whether aspects of network governance relate to crisis response

effectivity. The unit of analysis in this thesis is the network as a whole. The network as a ‘whole’ refers to interorganizational networks and is in contrast with the organizational level of analysis. This means that: “the focus is on the structures and processes of the entire

network rather than on the organizations that compose the network” (Provan, Fish, Sydow,

2007, p. 482). This study will particularly focus on crisis response networks. In the context of crisis management in the Netherlands, the focus is on the safety regions. A safety region is a geographical area in which actors are obliged to cooperate in the field of crisis management (Scholtens, 2008). A more detailed explanation of the safety regions and crisis management in the Netherlands can be found in the appendix.

3.2 Case selection and justification

The cases must serve a theoretical purpose and are therefore selected based on two important criteria (Toshkov, 2016). First, the cases represent different types of crises, ensuring that the findings are not tied to specific type of crisis. Second, the crisis management initiatives selected represent different outcomes of McConnell’s (2011) spectrum of success and failure. This allows to research whether and how the mode of network governance influences

outcomes of a crisis management initiatives. The crisis responses show similarities, because they operate according to the general framework of crisis management in the Netherlands. I have taken into account the implementation of the Dutch Safety Regions Act in 2010, which made changes to the crisis management system. Only crises that occurred after 2009 have therefore been selected.

Three different types of crises are studied: industrial incidents, shootings, and

wildfires. As stated in the introduction, these types have been selected because they represent types of crises that have occurred in Netherlands in recent years and that are in focus of the Dutch authorities responsible for crisis management. For each crisis type, two cases have been

(19)

19

selected. First, for the industrial incidents the cases are the fire at Chemie-Pack in 2011 and the fire at Chemelot in 2015. In both cases the fire caused the formation of large clouds of smoke, which potentially had the risk of spreading dangerous chemicals throughout the region. In response to this risk, the crisis networks in both cases scaled up to a similar level. Moreover, both cases also show similarities in the handling of the crisis. Whereas it is reported that the response networks at the operational level handled the crisis well, the political-strategic level has been criticized in both cases for a lack of collaboration. Second, for the shootings the cases are the shooting in Alphen aan den Rijn in 2011 and the shooting in Utrecht in 2019. In both cases there was an individual perpetrator who killed multiple people with a firearm. The response networks in the cases thus responded to a similar

situation, which is known as the active shooter situation. In both cases the response networks had to deal with a period of uncertainty after the attacks. In the case of Alphen aan de Rijn, there was uncertainty for several hours about the number of perpetrators and the potential of explosives in shopping malls. Similar, in Utrecht there was initially confusion about the number of perpetrators. Moreover, there was uncertainty due to the fact that the suspect was on the run. Third, for wildfires the cases are the fire at the Strabrechtste Heide in 2010 and the fire at national park de Hoge Veluwe in 2014. Although wildfires are relatively common in the Netherlands, these two wildfires were both exceptional in size. They affected an area of 150 and 320 hectares, respectively. In both cases the regional emergency services required significant assistance from outside the region. The emergency services received support from firefighters from multiple other regions and from the military. Moreover, in both cases there were also difficulties in establishing a common operational picture among the responders in the networks.

3.3 Data Collection

The data for the case studies is collected from a variety of sources, including governmental inquiries, evaluations by third parties, policy documents, and media reports. For all cases studied there are inquiries and evaluation reports available. Importantly, these reports discuss elaborately the effectiveness of the crisis responses. They do not only address the

shortcomings of responses, but also emphasize good practices. With regard to data sources, however, a limitation needs to be acknowledged. The analysis of each individual case was almost exclusively based on a single evaluation report. The reports came from a variety of sources, as indicated above. The disadvantage is that there are certain biases in these reports. For example, an evaluation report from one of the safety regions that has been used for one of

(20)

20

the cases has been criticized for being not critical enough. The reports of the Inspectorate of Security and Justice were clearly more in-depth and appeared to be more critical. A remedy to decrease this limitation was to conduct the analysis of the cases on the basis of more abstract themes, rather than in a detailed fashion.

3.4 Data analysis

The analysis of the cases has been done in four steps. First, an initial reading of the evaluation reports provided some clarity with respect to the outcomes of the crisis responses. In other words, the evaluation reports were scanned to get an idea of the success of the crisis response in each case. For example, the initial reading of the wildfire reports showed that the

Strabrechtse Heide case could be labeled as a conflicted success and the Hoge Veluwe case as a durable success. This initial categorization of outcomes was based on the statements made in the evaluation reports. Second, the evaluation reports were coded in ATLAS.ti based on the three groups of crisis response functions identified in the framework of Boin et al. (2013). These three groups of functions are: early recognition and sense making, decision making and coordination, crisis communication and meaning making. Next, I compared the cases of a crisis type based on this analysis. This allowed me to identify the key themes that were useful in explaining the effectivity in both cases. For example, in the cases of the industrial incidents it became clear that the measurement of hazardous materials, interregional coordination, and crisis communication were crucial themes in both cases. Third, the next step was then to discuss these themes more in detail and summarize the findings of the cases of a crisis type in a table. Fourth and finally, the last step of the analysis was to locate the crisis response in one of the five success/failure categories and discuss how and whether aspects of network

governance – governance, membership, and trust – related to these findings.

3.5 Reliability and validity

Toshkov notes that research is considered reliable if “the application of the same

measurement instrument or technique to the same data results in the same estimates” (2016,

p. 117). The use of the analytical framework by Boin et al. (2013) contributes to the reliability, because the data was analyzed in a systematic manner. Yin (2009) argues that a general way of approaching the reliability problem is to: “conduct research as if someone

were always looking over your shoulder” (p. 45). In order to increase the reliability of this

study, I accounted for all the steps taken in the analysis above. Another aspect that increases the reliability of this study is that all the data sources are publicly available. In terms of

(21)

21

validity, it is useful to distinguish between construct validity, internal validity, and external validity. According to Yin (2009), the first refers to whether correct operational measures have been established for the concepts that are being studies. The construct validity of this study is undermined because some of the cases represent more of a brushfire than the type of large-scale crisis referred to in the literature. Internal validity is increased through the

consistent use of six categories of crisis response functions in the analysis. Finally, external validity refers to whether the findings can be generalized to a larger population. The analysis of multiple cases might lead to findings that are relevant to similar types of crises in the Netherlands. However, because the cases studied are exclusive to the Netherlands, it is likely that some findings are specifically tied to the crisis management system of the Netherlands. At the same time, the six crisis response functions are relevant to crisis response operations in general and some findings might therefore be generalized to other countries with different crisis management systems.

(22)

22

Chapter 4: Findings industrial incidents

This chapter has been divided into five parts. The first part introduces the reader to the two cases. The subsequent sections discuss three groups of crisis response functions that account for the effectivity of the crisis operations. The last section provides an overview of the

analysis and reflects on the aspects of network governance of the cases. For readers unfamiliar with the crisis management system in the Netherlands, a brief explanation is offered in the appendix.

4.1 Chronological overview of the industrial incidents 4.1.1 Chemie-Pack

On Wednesday the 5th of January 2011, a fire broke out sometime past two in the afternoon at Chemie-Pack, a company that handled and processed chemicals. The company was located at an industrial park in Moerdijk, which is a municipality within the safety region Midden- en West-Brabant. The fire quickly spread to multiple containers that contained large amounts of flammable chemicals. Once the fire reached these containers they started to melt down and the fire became uncontrollable. At 14:26 Chemie-Pack called the fire services. In a short period of time the fire spread out over the property and caused a large, dark cloud of smoke. The cloud of smoke drifted northwards, as depicted in figure 4. By coincidence, the officer-in-charge (OIC) of the fire department was at his office, which is located at the industrial park. While on his way to Chemie-Pack, the OIC decided to scale up immediately to GRIP-2 and alert a large number of firefighters. At 14:50 the company commander of the fire brigade arrived. The commander and the OIC concluded that they were unable to extinguish the fire. They therefore decided to try preventing the fire from spreading and to let it burn out.

(23)

23

In multiple areas sirens were activated in order to warn citizens. Moreover, the

National Institute for Public Health and Environment (RIVM) and the fire services dispatched measurement teams. The measurement results that became available that day indicated that the hazardous materials were spread in such low quantities that they would not have adverse effects on health. At 16:52 Midden- en West-Brabant scaled up to GRIP-3. Although the fire intensified around 19:00 at the northwest side of the property, the firefighters expected to be in control around 21:00. However, the strategy to let the fire burn out appeared to be difficult to achieve, because of the nature and scope of the fire. At 19:30 Midden- en West-Brabant scaled up to GRIP-4. In the evening, the firefighters decided to extinguish the fire with foam. However, this plan needed to be communicated to all relevant parties, because of its potential to cause additional smoke. At 22:30 the firefighters carried out their plan. The plan was successful and around 00:15 the fire services reported that the fire was under control. Midden- en West-Brabant decided to scale down to GRIP-3 at 02:20. On Friday the 7th_{of January the}

fire was finally put out and the crisis organization scaled to GRIP-1. However, in the evening it was decided to scale up to GRIP-4 again. The incident had caused significant unrest among the public. It became clear that the aftermath of the incident still required coordination at the highest management level (DSB, 2012; POSI, 2011c).

.

Figure 4. Borders between the safety regions. (Source: DSB, 2012, p. 120)

4.1.2 Chemelot

On Monday the 9th of November 2015, a fire broke out in a warehouse at Chemelot, an industrial park in Zuid-Limburg. The industrial area is located in the municipality of Sittard-Geleen. The warehouse consisted of two compartments: one containing plastic pellets and the

(24)

24

other containing various chemical substances. The compartments were separated by a fire-resistant wall. Multiple chemicals were also stored next to the warehouse. At 11:45 the corporate fire services of Chemelot were alerted about a fire in the warehouse. The fire broke out in the compartment that contained the plastic pellets and immediately caused a large, dark cloud of smoke. The OIC of the Chemelot fire brigade scaled up to GRIP-1 at 12:02. In addition, the corporate fire services also notified the control room of the fire services of Zuid-Limburg. The first priority of the operational command team (COPI) was to prevent the spread of the fire. At 12:15 the safety region Zuid-Limburg scaled up to GRIP-2. About eight minutes later the fire services alerted three measurement teams. These teams took air samples and measured for toxic materials.

In the meantime, the cloud of smoke became visible in Belgium as well. The control room of Hasselt (Belgium) contacted the control room of the Dutch fire services at 12:27, because it received multiple calls from citizens. About three minutes later the leader of the COPI decided that the sirens needed to be activated in order to warn citizens. In accordance with the regional GRIP-procedure Zuid-Limburg scaled up to GRIP-3. In addition, an NL-Alert warning message was sent to citizens. About 80.000 citizens were advised to stay inside their homes. The area that was affected by the cloud of smoke is depicted in Figure 6. At 12:45 the RIVM was contacted as well and they sent two measurement teams to

Zuid-Limburg. The control room of Heinsberg, a district in Germany, was contacted at 12:50 about the fire. About five minutes later the control room of Zuid-Limburg called the control room of Noord-Limburg as well. At 13:00 the first measurements of the fire services indicated that the substances spread by the cloud of smoke did not exceed the applicable limits. However, Zuid-Limburg scaled up to GRIP-4 at 13:38, because the cloud of smoke affected multiple areas, including a municipality in Germany. Finally, around 15:35 the fire was under control. Zuid-Limburg scaled down to GRIP-1 at 17:54 and to GRIP-0 at 18:40 (VRZL, 2016).

(25)

25

(26)

26

4.2 Early recognition and sensemaking

The first theme that emerged from the analysis relates to the industrial nature of the incidents. In both emergencies the fire caused a large, dark cloud of smoke that spread over a substantial area. It was a crucial task for the crisis organization in both cases to assess whether the cloud of smoke contained a dangerous amount of hazardous materials. The cloud of smoke could have negative effects for the safety of responders, citizens, and the environment. The fire services and the RIVM therefore deployed measurement teams. However, this process encountered difficulties in both cases.

4.2.1 Chemie-Pack

Multiple parties took measurements of substances in the ground, water, and air. Among others the fire services of the safety regions Midden- en West-Brabant and Zuid-Holland Zuid took measurements of four toxic materials that were potentially spread by the fire. The RIVM dispatched a specialized team known as the Environmental Incident Service (MOD). However, poor communication and confusion about the lines of authority hindered the measurement process. There was confusion about the relationship between the MOD, the safety regions, and the national level.

The areas most impacted by the cloud of smoke fell within the boundaries of Holland Zuid. The MOD was deployed around 15:50, but only arrived around 19:30 in Zuid-Holland Zuid. By then the darkness of the night made it difficult for the team to carry out its tasks. The environmental incident service team reports to the policy support team for

environmental incidents (BOT-mi) of the RIVM. The leader of BOT-mi was initially unable to contact Midden-en West-Brabant because they did not answer his phone calls. Only around 18:00 he got a hold of the hazardous materials health advisor. Thus, the environmental

incident service team was already deployed long before any contact was made with any of the two safety regions. As a result, there was no consultation with the safety regions about the appropriate course of action. The team decided to travel to Midden- en West-Brabant, but Zuid-Holland Zuid was actually most affected by the cloud of smoke. Furthermore, it was unclear who held authority over the policy support team for environmental incidents. The Safety Regions Act stipulates that the management of this team falls under the command structure of the safety regions. However, the team itself believed that it operated under the authority of the ministry and therefore did not report its findings and advice directly to the safety regions. For example, Zuid-Holland Zuid did not receive the first advice and it only received the two subsequent advices from Midden- en West-Brabant. The confusion about the

(27)

27

lines of authority increased when the national level became involved. The national level exerted pressure on the policy support team to work faster. For the safety regions it now appeared as if the national level had the authority over the policy team. The confusion about the relationship between the BOT-mi and the safety regions was remarkable because there were cooperation agreements. Apparently, all stakeholders lacked knowledge about these agreements. However, it has also been acknowledged that the agreements lacked clarity. For example, the agreements did not incorporate the hazardous materials health advisors of the safety regions. Logically, these advisors should have an important role in collaboration between the BOT-mi and the safety regions (DSB, 2012).

4.2.2 Chemelot

In the Chemelot case the measurement of hazardous materials was hindered by two factors. First, the activities of the measurement team of Noord-Limburg were not monitored. Second, there was a lack of coordination between the operational and strategic teams of the safety region Zuid-Limburg, which made it difficult to act upon the measurement results.

The cloud of smoke drifted northeast across multiple areas. It affected not only areas in Limburg, but also areas in Noord-Limburg and Germany. The fire services of Zuid-Limburg quickly dispatched three measurement teams. The fire services were contacted around 12:00 by Chemelot about the fire and about 23 minutes later they dispatched these teams. Moreover, the hazardous materials health advisor of Zuid-Limburg contacted the RIVM for support around 12:45, who sent two measurement teams to Zuid-Limburg. However, these teams only took measurements in Zuid-Limburg. The advisor therefore contacted the fire services in Heinsberg (Germany) who dispatched a measurement team. Moreover, the advisor also contacted Noord-Limburg. Noord-Limburg agreed to dispatch two measurement teams and communicate their findings to Zuid-Limburg. Whereas collaboration between Heinsberg and Zuid-Limburg worked well, the collaboration with Noord-Limburg was unsuccessful. Noord-Limburg failed to communicate the measurement results to Zuid-Limburg.

Around 13:00 the first measurement results of the fire brigade indicated that the hazardous substances spread by the cloud of smoke did not exceed the applicable limits. However, Zuid-Limburg was unable to act adequately upon these findings, because of poor coordination between the regional operational team (ROT) and regional policy team (RBT). The ROT had to translate the results into actionable advice. The results were discussed for the first time within the meeting of the ROT at 14:00. However, the RBT only received the

(28)

28

actionable advice of the ROT one hour later at 15:00. The RBT found that the actionable advice from the ROT was not sufficient. It therefore requested the ROT to clarify some aspects. It takes another hour before the RBT received the additional information. The

foregoing illustrates how it took a significant amount of time for the safety region to act upon the first results. Zuid-Limburg only acted decisively upon the findings at 16:00, due to the cumbersome process of coordination between its teams (VRZL, 2016).

4.2.3 Comparison

In order to compare the two cases one issue requires further clarification. In the Chemie-Pack case there existed profound uncertainty with regard to the potential spread of chemicals by the cloud of smoke. It remained unclear what materials were burned by the fire. An official of Chemie-Pack handed over a USB-stick, which contained information about the materials stored at Chemie-Pack. However, the USB-stick contained 52 unstructured pages, which also included random information about the inventory such as the number of wooden pallets (POSI, 2011c). The crisis organization therefore had to assume that all chemical materials listed in the document were present. In contrast, in the Chemelot case the fire services had accurate information about the presence and details of the materials in the warehouse. In the Chemie-Pack case, the monitoring of the activities of the measurement teams of the safety regions Midden- en West-Brabant and Zuid-Holland Zuid was relatively successful. However, the measurement teams of the fire services had only limited capacity to take measurements. The role of the environmental incident service was crucial, due to the profound uncertainty about chemicals. As was illustrated above, it was precisely the

cooperation between the environmental incident service and the safety regions that fell short. In the Chemelot case, the activities of the measurement team of Noord-Limburg were not monitored. However, this had few consequences because the areas affected by the cloud of smoke fell primarily within Zuid-Limburg and Germany. Zuid-Limburg cooperated with Heinsberg to monitor the effects of the cloud of smoke in Germany. Overall, the measurement process in the Chemie-Pack was more problematic compared to the Chemelot case, due to confusion about roles and authority.

(29)

29

4.3 Decision making and coordination

The second theme concerns the decision making and coordination during the incidents. This theme has already been briefly touched upon in the previous section. In both cases the cloud of smoke drifted across multiple regions. Other regions tried to monitor the developments of the incidents closely, because of environmental and public health concerns. However, it was a real challenge for the professionals in both cases to share information timely with other regions.

4.3.1 Chemie-Pack

The cloud of smoke affected multiple other safety regions, including Zuid-Holland Zuid and Rotterdam-Roermond. Midden- en West-Brabant collaborated with Zuid-Holland Zuid, sharing information about the incident, its consequences, and the strategy of the fire fighters. At the same time, the slow upscaling process in Midden- en West-Brabant hindered

collaboration with Zuid-Holland Zuid. Moreover, the municipality Moerdijk and Midden- en West-Brabant failed to share information with other safety regions.

The officer-in-charge of the fire brigade recognized that the cloud of smoke drifted across the border with Zuid-Holland Zuid. He therefore instructed the control room to inform Zuid-Holland Zuid about the fire at Chemie-Pack. The control room contacted Zuid-Holland Zuid at 14:44, who alerted their hazardous materials health advisor and the fire services. Importantly, the leader of the measurement team of Zuid-Holland Zuid contacted the control room to establish contact with the measurement team of Midden- en West-Brabant. The control room enabled him access to the communication channels of the measurement teams of Midden- en West-Brabant. As a result, he could follow all communication of these teams and therefore had crucial information about their activities and findings. Moreover, the leaders of the measurement teams of the two safety regions consulted each other frequently. In general, collaboration between the two safety regions was primarily facilitated through

monodisciplinary channels. However, coordination between the strategic teams of the two safety regions was problematic, because Midden- en West-Brabant did not scale up to GRIP-4 for a long time. The safety region Zuid-Holland Zuid already scaled up to GRIP-4 at 15:45. In addition, it also made several requests to Midden- en West-Brabant to scale up to GRIP-4. However, the mayor of Moerdijk refused to scale up to GRIP-4, because he believed that from an operational point of view there was no need to scale up. He was right that the fire services handled the fire at Chemie-Pack relatively well. However, he failed to consider the effects of the fire beyond the borders of his own municipality. At the GRIP-4 level the regional policy

(30)

30

team (RBT) of Midden- en West-Brabant could have coordinated and enabled closer

collaboration with Zuid-Holland Zuid. Once Midden- en West-Brabant scaled up to GRIP-4 at 19:30, the RBT facilitated closer collaboration with Zuid-Holland Zuid. However, the slow the upscaling process in Midden-en West-Brabant had a significant impact on crisis

communication. This will be discussed elaborately in the next section.

Other regions than Zuid-Holland Zuid experienced many difficulties in contacting and receiving information from Moerdijk and Midden- en West-Brabant about the fire. They were left much in the dark and had to use the media as a source of information. For example, Rotterdam-Rijnmond tried to contact Midden- en West-Brabant multiple times throughout the day. However, no phone calls were answered. The safety regions could have used the national crisis management system (LCMS) as an alternative source of information. However,

Midden- en West-Brabant had not fully implemented LCMS and its officials were

unexperienced in using the system. As a result, they shared very little information through LCMS. In addition, multiple safety regions did not have access to the information that was shared by Midden- en West-Brabant. Only neighboring regions could access the information. Other regions needed to be given access manually by officials of Midden-en West Brabant. However, they did initially not authorize other regions to access the information in LCMS (POSI, 2011c).

4.3.2 Chemelot

In the Chemelot case the cloud of smoke did not only cross borders between safety regions, but also between countries. The cloud of smoke could be seen from Belgium crossed the border with Germany. In the aftermath of the incident, mayors from municipalities in Germany, Belgium, and Noord-Limburg expressed their concerns about the lack of

coordination. For example, the mayor of Maasmechelen complained that they were told not to worry because the direction of the wind was towards Germany and not Belgium: “The public

health in Maasmechelen should not be dependent on the wind direction” (VZRL, 2016, p. 19).

Many of the problems regarding interregional coordination have been traced back to the collaboration agreements on cross-border incidents and disasters. The agreements led to ambiguity and created different expectations among stakeholders.

The safety region Zuid-Limburg operated according to the blueprint for incidents at Chemelot. The plan stipulates which control rooms needs be contacted when the sirens are activated in certain areas. At that day, the sirens were activated in areas one and six. Zuid-Limburg therefore contacted the control rooms connected to these areas, Heinsberg

(31)

31

(Germany) and Noord-Limburg, respectively. Although Noord-Limburg was informed relatively late, 25 minutes after the sirens were activated, However, Zuid-Limburg had not incorporated other existing collaboration agreements into the blueprint. According to the blueprint, Zuid-Limburg did not have to inform Belgium, because the areas in which the sirens were activated did not border Belgium. However, the collaboration agreements with Belgium stipulated that Zuid-Limburg should have informed Belgium. The control room of Hasselt (Belgium) contacted Zuid-Limburg after it received messages from citizens about the cloud of smoke. In other words, the blueprint for incidents at Chemelot was followed suit, but was not in line with other existing agreements. Furthermore, these agreements also lacked clarity with regard to several aspects. This led to diverging expectations as who was to be informed by Zuid-Limburg and what information should have been passed on (VRZL, 2016).

4.3.3 Comparison

In both cases multiple areas were affected by the cloud of smoke, albeit that some areas were affected more than others. However, the regions more remotely also had serious concerns about the potential consequences of the incident for themselves. In the Chemie-Pack case it was difficult for those regions to contact the crisis organization. In contrast, in the Chemelot case the affected regions were able to establish contact with Zuid-Limburg. However, it became clear that these regions still believed that the information-sharing process fell short. It is worth noting here that in the Chemelot case the upscaling process has also been criticized. The upscaling from GRIP-3 to GRIP-4 took about one hour. It is argued that for such

incidents at Chemelot, a lot of time is lost by not immediately scaling up to GRIP-4. At the beginning of such incidents it can be anticipated that the cloud of smoke will most likely affect multiple areas. Overall, interregional coordination process was managed rather poorly in both cases.

Crisis Governance: a Networked Enterprise? Case studies of crisis response in the Netherlands