People, Infrastructure, and Conflict: Analyzing the Dynamics of Infrastructure Disruption and Community Response

People, Infrastructure, and Conflict: Analyzing the Dynamics of Infrastructure Disruption and Community Response

Natalie Myers, Jeanne Roningen, Ellen Hartman, Tina Hurt, Scott Tweddale, and Patrick Edwards

Introduction

Human well-being is heavily reliant on infrastructure systems to provide food, water, power, communication, shelter, and transportation. In 2050, the world will have welcomed a net 2.5 billion additional people in areas characterized as urban, with most of this increase concentrated in Asia and Africa. In this urbanized future, societal dependencies on built infrastructure will remain, but it is highly likely that the organization, construction, maintenance, and governance of those infrastructures will have changed in response to increasing environmental, economic, and political pressures. Therefore, the ability to characterize the changing connections, additions, and disruptions between physical infrastructures and the people who depend on them is essential to fully understand future operational environments.

Military operations inevitably encounter and interact with complex infrastructure systems associated with an equally complex array of users. Infrastructure systems that distribute centrally-generated resources often coexist with infrastructure that is installed at the household or community level, but whose functioning in many cases has dynamic impacts on both supply of and demand for resources; infrastructure interdependencies therefore can occur across scales and are subject to dynamic changes and constitute complex adaptive systems (Rinaldi et al. 2001). The reaction of a population to a disruption to the infrastructure system due to conflict, terrorism, or natural disaster can transfer and aggravate the burden on surviving infrastructures, which may cause cascading secondary impacts. A fuller understanding of the function of infrastructure in the operational environment will enable mission planners to anticipate the effects of operations across all phases of conflict and potentially mitigate some of the effects of military interventions in urban areas where a heterogeneous mix of friendly, hostile, and neutral actors are in close proximity and share infrastructure networks.

Currently, resources for addressing doctrinal requirements in the area of infrastructure are insufficient. U.S. Army doctrine requires that commanders understand, visualize, and describe the infrastructure component of the Joint Operating Environment to accomplish the Army’s missions of protecting, restoring, and developing infrastructure (Hart et al. 2014). The Army’s current doctrinal tool (FM 3-34.170) for in-theater assessment of the operating status of infrastructure presents a collection of smartcards (i.e., checklists) on the following topics: (S)ewage, (W)ater, (E)lectricity, (A)cademics, (T)rash, (M)edical, (S)afety, and (O)ther Considerations, which include transportation networks, fuel distribution, housing, explosive hazards, environmental hazards, communications, places of worship, and attitude. SWEAT-MSO assessment is concerned with evaluating the operational status of essential services and the staffing level of critical positions; with securing infrastructure and the populace; and with ensuring civil order. However, tools are not available to help assess the dynamic interactions between these infrastructure services or between a population and the infrastructure network; or to integrate infrastructure into planning throughout all phases of conflict. Some people and groups are impacted differentially by disruptive events, react differently in the aftermath, adjust to circumstances in dissimilar ways, and recover in a different manner. Such differences are rooted in the societal characteristics of communities. The ultimate impact of disruptive events is the product of dynamic interactions between the built-environment (e.g., civil infrastructures) and the societal characteristics of the community, and these interactions are not adequately addressed in current methods of analysis (Myers et al. 2016). The consequences of disruptive events often extend beyond the geographic boundaries of the physically impacted region with the impacts spreading through multiple social systems and scales affecting governments, institutions, economic sectors, livelihoods, and people. Past experience in both conflict and natural disaster situations highlights the importance of accounting for the far-reaching societal impacts that are crucial for effective pre-event mitigation planning and optimal post-event resource allocation.

Desired outcomes of this research are improved abilities to 1) map, conceptualize, and model interconnected infrastructure networks, 2) quantify the effects of cascading disruptions across networked systems, 3) quantify and visualize the projected reactions of a population to a disruption, including switching resource providers and changing patterns of use of the infrastructure network, 4) define acceptable and tolerable levels of actual and perceived consequences to heterogeneous communities, 5) visualize the effects of disruptions on a population at a neighborhood level, and 6) assess social impacts of decisions to protect, destroy, or restore infrastructure assets across multiple operational phases.

Methods

Human-Infrastructure Systems Analysis (HISA) is an analysis method that accounts for both the physical/functional and social effects of infrastructure changes on society and local populations. A holistic analysis framework was developed that integrates infrastructure interdependencies and human community behaviors to evaluate a city’s vulnerability to disruptions and to assess the impact of potential disruptions. To accomplish this, a game-theoretical equilibrium model was developed in a multilayer infrastructure network, to systematically investigate the mutual influence between infrastructures and communities. In this model, two types of infrastructure failure patterns were formulated to capture general network interdependencies; network equilibrium was extended into infrastructure and community systems to address redistribution of demand for life-supporting resources; and the societal impact of disasters was estimated based on resource demand loss, cost increase, and total infrastructure failure.

To further quantify the societal impacts of infrastructural disruptions, the HISA framework adopted the Capability Approach (CA) (Sen 1985). In CA, the impact of disruptions on individuals’ well-being is evaluated in terms of genuine opportunities open to individuals to do or be things of value, like being sheltered, nourished, or mobile. The opportunity to achieve different functionings collectively gives rise to a certain capability state that could be classified as acceptable, tolerable, or intolerable. To operationalize the CA, we identified 10 capabilities that capture various aspects of individuals’ well-being, and developed indicators for those capabilities based on census data and resource access cost values derived from the physical infrastructure model. For illustration purposes, the HISA framework is used to visualize the changes in access costs and capabilities of communities within the city of Maiduguri, Nigeria in the aftermath of disruptive event scenarios. The final product of the analysis is a series of maps that represent the spatial distribution of both access cost changes and capability states (from acceptable to intolerable) as a function of changes to an infrastructure network.

The Network Model

As disruption takes place in an infrastructure system, cascading failure begins to develop following certain failure propagation patterns that are inherent in the network, and people start to change the way they access resources. Eventually, the disrupted system arrives at a state where the remaining infrastructures and the impacted population behaviors fall into new equilibria. Modeling and querying the structure of these new equilibria aims to understand and visualize how local populations might be affected by particular infrastructure changes.

The framework therefore (a) generalizes various types of interdependencies among infrastructures with a layered network model, (b) estimates entangled system failure and equilibrium community behavior as part of a game-theoretical model, and (c) evaluates the cascading propagation of disruptions (due to interdependencies) and the consequential societal impacts (such as demand loss, cost increase, and capability state). We developed a heuristic algorithm to calculate the system equilibrium (Buldyrev et al. 2010; Lu et al. 2016). We performed numerical comparisons based on real-world data to examine the impact of cascading infrastructure failures on the population, and explored model sensitivities.

Typically, especially after disasters, people access life-supporting resources by utilizing the transportation layer. The trivial case that people get resources within the community, e.g., tap water at each household, can be incorporated as traveling a distance of zero. Under disruption, fewer resource nodes remain functioning, which can lead to overwhelming demand concentration and associated long queuing, longer travel detours, and associated traffic congestion. This situation significantly increases people’s resource access costs and may further lead to demand loss (i.e., people giving up service) if the costs exceed affordability for a given community. A variant of the traffic equilibrium model was implemented to capture this issue.

Two modes of failure propagation were represented in the network model. To maintain working status, some infrastructure nodes need resource supply, which are also transported via the transportation layer (e.g., a diesel generator accesses fuel from diesel tanks filled by tanker trucks). Therefore, the traffic congestion under the community behavior may significantly delay or block resource procurement as well, or even cause failure to the infrastructure. This type of failure is referred to as a resource failure. Secondly, infrastructure disruption could also be due to direct and obvious dependencies based on physical connections. For example, a water tank with an electrical pump must be supported by a nearby power source (e.g., a power grid). Such a dependency is normally established for the long run, at a large setup cost, and hence difficult to modify. This type of failure, caused by supporting infrastructure disruption, is referred to as support failure. Comparing the two types of failure, support failure is found to be one major cause of cascading disruptions. In particular, a network with a very dense presence of links (e.g., a tree) can be very vulnerable—the failure of the root node will propagate (cascade) and disrupt the entire system. In contrast, resource failure normally poses a potential risk, but one that is less obvious to foresee in advance. However, resource failure can also bring devastating damage to the entire system, especially when it happens at some critical node that can cause consequential cascading support failures.

Quantifying Social Effects

To develop the Capability Approach, we derived 16 indicators obtainable from census and survey data to quantify 10 capabilities, some of whose inputs are proposed to be analogous to infrastructure model outputs of access time to various resources (Table 1). A probabilistic predictive model was developed to determine the sets of indicators available from census data that were most correlated with capability scores derived from household surveys. Then, a fault-tree was developed that schematically illustrates how the combinations of different indicators collectively give rise to the capability states. The methods of system reliability analysis were used to obtain the corresponding probability of each capability state. (Gardoni and Murphy 2010). For illustration purposes, the proposed framework was used to visualize the capabilities of the communities in the city of Maiduguri in the aftermath of a disruptive event. The final product of the analyses is a series of maps that represent the spatial distribution of the well-being of the communities in terms of each capability as well as an overall capability score. While the physical network model output and associated resource access cost calculations are explicitly spatial in nature and limited by the fidelity of the infrastructure network, the spatial representation of the CA approach is limited by the smallest geographic area for which indicators derived from census data can be disaggregated.

Figure 1: Selected Capabilities and Indicators for the Maiduguri Case Study

End-User Applications

An end-user might want to know how to represent a case where there is incomplete information about the true fragility of infrastructure under the given failure propagation mechanisms. We therefore explored model formulations where, rather than presuming that infrastructure failure occurs completely and consistently when certain predefined failure conditions are met, instead a Monte-Carlo probabilistic failure analysis is used, whereby each failure propagation occurs based on a certain predefined probability of failure. This technique can help model results under incomplete information because the mean, range, and distribution of the results of multiple model runs can be analyzed to understand the implications of uncertainties in input failure probabilities.

In another “Restore-Recover-Rebuild” case study, the team was given an already-disrupted system and a set of nodes that could be restored or rebuilt in new locations. The infrastructure network model was viewed as a function that maps each subset of possible combinations of nodes to be restored to a set of social welfare indicators, such as the cumulative time for all communities to access water, markets, healthcare, etc. A method can then be set up to explore the best possible combination of nodes to restore, given that the aim is to maximize social benefits while only restoring/rebuilding a certain fixed number of nodes at any point in time. This method translates mathematically to a nonlinear non-convex optimization problem, which can be solved approximately but efficiently by using a genetic algorithm.

Additional modifications allowed the model to represent partial rather than complete failure; to represent resource limits or service capacities of a given infrastructure node; and to represent demand elasticity whereby demand decreases as a function of increasing cost to account for a community’s tendency to make do with less as costs rise rather than using arbitrary cost ceilings at which demand collapses.

Case Study: Maiduguri, Nigeria

Maiduguri is the capital city of Borno State in northeastern Nigeria and has an estimated population of 1.2 million. Concurrent with rapid urban growth, the local government has been facing additional severe challenges. Maiduguri is located in the heart of the rebel activity of Boko Haram, experiencing frequent attacks on its infrastructure.

Boko Haram has, in the past three years, vandalized public infrastructure like telecommunication masts that had hitherto cut off various parts of Borno State from the rest of the world. The group had also used bombs and fire to destroy schools, hospitals, police offices, barracks and even cratering of roads. Their most recent attack on public infrastructure was the bombing of an ultra-modern drilling rig procured by Borno State at the cost of over N300 million at the site where it was mobilized to drill water for rural dwellers of the northern part of Borno State. (All Africa 2014)

Both active military events and terrorist attacks threaten people’s daily life and the security of urban infrastructure (Ibeh 2015). Because of the ongoing conflict, large numbers of internally displaced persons (IDPs) flee from the surrounding countryside into Maiduguri after terrorist attacks. That influx in population further stresses the city’s resources, resulting in increased demands on infrastructural systems (Haruna 2015).

A geospatially-referenced network model of the critical infrastructures in the city was developed using a combination of existing data and an understanding of how different infrastructure components are used in this location. In the network model of Maiduguri’s infrastructure, the following utility, institutional, and community sectors were represented: Fresh Water, Fuel, Electricity, Transportation, Food, Schools, Medical Facilities, and Community. Subsequently, cascading infrastructure failures were propagated under different scenarios throughout the network, creating geospatially-referenced output representing before-and-after costs to access those different resources. The results of different failure scenarios were then assessed and several types of sensitivity analysis were performed to better understand both system and model behavior. Finally, this information was included in the CA method to derive spatially-distributed maps that relate infrastructure changes to changes in social well-being.

Figure 2: Change in Water Accessibility After Disrupting the Main Power Substation

An example scenario was considered where disruption initiates at the main power substation in the center of the city, and the failure propagation and social impacts were then investigated. After propagating the network disruption, the system converges to equilibrium. As expected, all electricity transformers are shut down, the electricity network is disabled, while only local electricity generators can work based on fuel, providing limited power supply to nearby communities. Furthermore, the water network is also disrupted, while only local commercial water vendors that can pump water from wells using generators still function. As a result, educational institutions and healthcare facilities are disrupted due to the shortage of power and water. In Figure 1, the spatial distribution of changes in communities’ water access costs shows that costs increased significantly around several of the commercial water providers as competition increases for those limited resources.

A set of simulations for this case study provided the means to explore potential system behavior. Predictably, resilience to disruptions increased with increased resource capacity. However, disruption that occurred at some apparently critical infrastructures (such as a water treatment plant) did not always produce catastrophic capacity reductions because the network of independent suppliers (e.g., commercially-run water wells) was able to replace that supply, although the spatial distribution of the attendant increased access costs varied. Under some scenarios, we observed that taking some infrastructure components offline actually produced overall benefits by supplying necessary resources or reduced congestion to other underserved parts of the system. Under other scenarios, dynamic interactions among infrastructures and communities led to new equilibria showing high-vulnerability risks to the system, with implications for operational planning. Our findings emphasize the importance of using a holistic model with dynamic community response rather than a one-way failure propagation model to correctly capture the interdependent relationships between the physical infrastructure systems and the population when evaluating system reliability and societal impacts.

Ongoing Research

Ongoing research aims to modify the model toward specific end-user objectives, for example, identifying the most critical infrastructure components in a given system. Based on this, geospatially-informed reinforcement or interdiction strategies can be sought that will best protect an urban system as a whole or certain vulnerable communities within it. Currently, the model’s framework only uses modeled values of access time to measure a community’s generalized resource cost; this can be naturally expanded by considering additional cost components (e.g., volumes and/or prices of resources). Additional applications include optimization of an infrastructure system in order to produce maximum social benefit with a limited budget as well as derivation of the set of configurations that would restore system function to some threshold capability distribution. Finally, for effective use of this type of model for planning purposes, it will be critical to better understand the sensitivity of model outputs to the level of detail and structural differences in the input data that define the joint human-infrastructure network.

Conclusion

Accurately anticipating the sociocultural impacts of military interventions on infrastructure is particularly important to security and stability operations, in which appropriate engagement with the local population is essential for mission success. However, any human-infrastructure system is comprised of both centrally-planned utilities and community-level installations and infrastructure use patterns that are generated by efforts of the local population to provide for their own resource needs. As the number of expanding urban populations that outpace planned infrastructure is expected to grow dramatically in coming decades, models of the dynamics of infrastructure development, use, and failure propagation need to account for these realities. HISA enriches the understanding of the future battlespace through better visualizations of cascading failures of physical infrastructures and the resulting effects on populations, accounting for dynamic human responses to those failures. Military planners and strategists will benefit from human-infrastructure analysis when the results are integrated across tactical, operational, and strategic military planning processes.

References

All Africa. (2014). Nigeria: Boko Haram Attacks Power Station, Plunges Maiduguri into Darkness. http://allafrica.com/stories/201407210823.html.

Sen, Amartya. (1985). Commodities and Capabilities. Amsterdam: North-Holland.

Buldyrev, S. V., R. Parshani, G. Paul, H. E. Stanley, and S. Havlin. (2010). Catastrophic Cascade of Failures in Interdependent Networks. Nature 464 (7291): 1025-1028.

FM 3-34.170. (2008). Appendix C, Infrastructure Reconnaissance, to FM 3-34.170, Engineer Reconnaissance.

Ibeh, N. (18 February 2015). Nigerian Military Kills Over 300 Terrorists, Captures Many Others. http://www.premiumtimesng.com/news/top-news/177064-nigerian-military-kills-over-300-terrorists-captures-many-others.html.

Hart, Steven D., J. Ledie Klosky, Scott Katalenich, Berndt Spittka, and Erik Wright. (2014). Infrastructure and the Operational Art: A Handbook for Understanding, Visualizing, and Describing Infrastructure Systems. ERDC/CERL TR-14-14.

Haruna, K. (January 2015). Maiduguri: A Troubled Capital City Overtaken By IDPs. http://leadership.ng/features/401050/maiduguri-troubled-capital-city-overtaken-idps.

Lu, Liqun, Xin, Wang, Zhaodong Wang, Yanfeng Ouyang, Jeanne Roningen, Scott Tweddale, Patrick Edwards, and Natalie Myers. (2016). Assessing Socioeconomic Impacts of Cascading Infrastructure Disruptions in a Dynamic Human-Infrastructure Network. ERDC/CERL TR-16-XX.

Myers, N., A. Rhodes, L. Whalley, G. Al-Chaar, J. Roningen, G. Calfas, T. Bozada, T. Hurt, D. Krooks, and D. Morrison. (2016). Understanding the Effects of Infrastructure Changes on Sub-Populations. ERDC/CERL TR-16-3.

Rinaldi, S. M., J. P. Peerenboom and T. K. Kelly. (2001). Identifying, Understanding, and Analyzing Critical Infrastructure Interdependencies. Control Systems, IEEE 21 (6): 11–25.

Gardoni, Paolo and Colleen Murphy. (2010). Gauging the Societal Impacts of Natural Disasters Using a Capability Approach. Disasters: The Journal of Disaster Studies, Policy and Management, 34(3), 619−636.