OPERATIONS RESEARCH TO IMPROVE DISASTER SUPPLY CHAIN MANAGEMENT
Ozlem Ergun, Gonca Karakus, Pinar Keskinocak, Julie Swann, and Monica
H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, Georgia
Disasters recently received the attention of the operations research community due to the great potential of improving disaster-related operations through the use of analytical tools, and the impact on people that this implies. In this introductory article, we describe the main characteristics of disaster supply chains, and we highlight the particular issues that are faced when managing these supply chains. We illustrate how operations research tools can be used to make better decisions, taking debris management operations as an example, and discuss potential general research directions in this area.
Keywords: humanitarian logistics; disasters; public sector OR
From improving the performance of fire and police departments, optimizing the public transportation system, programming delivery of blood to hospitals, planning housing projects, analyzing drug policies, to improving delivery of meals to senior citizens, there are many examples of how operations research (OR) addresses community needs (1). Public sector OR deals with solving public interest problems through the application of analytical tools. One of the public sector OR application areas is humanitarian OR, which particularly deals with the problem of delivering relief to people affected by a disaster. Hence, natural and man-made disasters are the most common subject of attention for humanitarian OR.
There were 6637 natural disasters between 1974 and 2003 worldwide, with more than 5.1 billion affected people, more than 182 million homeless, more than 2 million deaths, and with a reported damage of 1.38 trillion USD (Center for Research on the Epidemiology of the Disasters (CRED)) (2). In 2005 alone, over 180,000 deaths and 200 billion USD economic losses have occurred due to disasters according to the Disaster Resource Network Humanitarian Relief Initiative (HRI) (3). The September 11 attacks (2001), tsunami in South Asia (2004), Hurricane Katrina (2005), and earthquakes in Pakistan (2005) and Java (2006) are just some examples of the deadliest disasters witnessed by humankind in the past few years.
The consequences of these events are enormous, not only in the short term with injuries, loss of life, and damaged infrastructure but also in the long term with changes in social and economic conditions. Even though the occurrence of these events could not have been avoided, the impact could have been reduced by different means including humanitarian OR. For example, adequate warning systems could have prevented injuries and fatalities during the 2004 tsunami and help could not reach Pakistani earthquake victims due to logistics difficulties with limited infrastructure. Humanitarian OR differs from other OR applications because it deals with particularly unique and highly variable events, often in resource-poor and limited infrastructure environments, with multiple organizations trying to work together in response activities simultaneously. These factors increase the complexity of responding to these events.
The focus of this article is on supply chain-related issues in humanitarian operations, with a greater focus on “disasters” rather than ongoing conditions. We discuss differences between regular supply chains (SCs) and SCs used for disaster planning and response, “humanitarian SCs”. First, we describe characteristics of supply and demand of disaster SCs, followed by a discussion on the particularities of the execution and management of these SCs. Next, an application of OR in humanitarian operations is examined in more detail, and finally main challenges and future research directions are presented.
OVERVIEW OF DISASTERS
Disasters can be grouped into two main categories: natural and man-made disasters. Natural disasters are the consequences of natural hazards that affect people, whereas man-made disasters are caused by human actions. A more detailed categorization of disasters is shown in Table 1 along with examples. Also, disasters could be categorized in predictable timing (or seasonal) such as floods or unpredictable timing like earthquakes and predictable location such as hurricanes or unpredictable locations like tsunamis.
TABLE 1. Disaster Categorization and Examples
Political crisis, refugee crisis
Terrorist attacks, chemical leaks
Hurricanes, floods, earthquakes, tsunamis
No matter the type of disaster, the management of these events typically follows four sequential stages: mitigation, preparedness, response, and recovery. Mitigation is the application of actions to help prevent or reduce the hazards of the disaster. It differs from the other stages because it focuses on long-term measures for reducing or eliminating risk (4). Preparedness activities help prepare for response once a disaster occurs. The response phase covers activities for mobilizing emergency responders and services for the affected region. Recovery is the stabilization phase during which restoration of the disaster area is conducted in the long term. The disaster management timeline with the related operations can be seen in Fig. 1.
Figure 1. Disaster timeline and operations.
There has been some research on the use of OR techniques in disaster management. Altay and Green (5) survey the literature and summarize the research in this area. Most of the papers published deal with man-made disasters (47.6%) or general disaster operations management (40.5%). Among the papers reviewed, 61.1% is on the pre disaster phase (mitigation and preparedness) and mostly based on risk analysis, 23.9% on response, and only 11% about recovery. More than half of the research is based on model development, followed by theory and application development. Most of the application studies are made for the disaster phase, whereas the theory papers mostly focus on the pre disaster phase.
The primary role of a supplier in an SC is to source the required items downstream. The main sources in a humanitarian SC are vendors and donors. Vendors can be local to the region where the disaster occurred or global. Donors are the sources of donations of any type (financial, products, services, etc.).
Supplies consist of relief items, personnel/volunteers, and transportation and construction resources, among others. Most of the supplies fall into the relief items category. Figure 2 shows a categorization for relief items, as well as some examples. Consumable relief items require recurrent delivery to the affected community and nonconsumable relief items require a one-time delivery only. Nonconsumable operational relief items are required to set up an operation, while nonoperational are required to meet the essential needs of the affected population.
Figure 2. Relief supplies categorization and examples.
There are specific challenges related to supplies that come from in-kind donations. First, since the quantity and mix of the supplies depend at least to some degree on the donor, there is a high uncertainty of what is going to be received. Moreover, the timing of these supplies might not be appropriate: for example, consumables that arrive too early and cannot be stored for a long time or nonconsumables that arrive after the operation was set up are wasted. There are many cases in the recent history where donated items were not needed and were not deployed to people affected by the disaster. Autier et al. (6) discuss the case of drug supplies after the 1988 Armenian earthquake, when at least 5000 tons of drugs and consumable medical supplies were sent by international relief operations, but only 30% were immediately usable (sorted, relevant for the emergency situation, and easy to identify), and 20% of these supplies had to be destroyed by the end of 1989. Unsuitable donations caused bottlenecks in the SC, making storage and transportation processes more inefficient.
Donations place additional complications on the procurement process, since it is difficult to define what will come from donors and what will have to be sourced from vendors. But even if donations are not considered, the procurement process is by itself a challenging task. Supply availability is highly dependent on the location. Organizations often have a low visibility into existing inventory. Control of inventory is usually given to country offices, resulting in excess supplies in some locations and scarcity in others. The selection of suppliers during the procurement process includes not only total cost versus quality, response time, and reliability trade-offs but also considers less measurable factors like activating local economies by choosing a local provider. Developing contracts with suppliers is difficult given the uncertainty of the type, quantity and timing of items required, and the available budget. Also, there is competition for supply sources when there are local or international nongovernmental (nonprofit) organizations (NGOs) sharing similar relief objectives; further there is lack of coordination among them.
Challenges faced while sourcing from donors or vendors after the event occurs may affect supply availability. The shortage of supplies may cause emergency response to be ineffective and result in increased human suffering (7). Hence, it is important to develop strategies to accelerate supply response or deal with unpredictability of demand. One strategic initiative that has been recently implemented by several humanitarian organizations is the prepositioning of inventory instead of procurement after the fact. Prepositioning allows not only faster response but also better procurement planning and an improvement on distribution costs; however, it requires an additional investment before the event occurs, and funds are more difficult to obtain.
In summary, the supply process in a humanitarian SC is different from regular SCs. Supply in regular SCs follows a standardized order fulfillment process, while in the case of disasters a portion of supplies comes from voluntary donations. Also, since humanitarian SCs are often in resource-poor environments, supply may be especially variable. Greater certainty on the supply quantity, location and time; longer relationships with suppliers; and information about the suppliers available at the location of interest before decisions are made allow better procurement contracts for normal SCs. Finally, when common supply deals with routine events and not a one-time disaster event, developing a procurement “expertise” is easier.
The customers in a disaster SC include the population at the affected area, as well as intermediate customers at local or global storage facilities. Their needs change significantly according to disaster types and the phases in the disaster timeline. In the pre disaster phase, protection-based items such as batteries and flashlights are highly demanded both by the people and local stores for preparation, while immediately after the disaster, the high demand changes to first response items such as drugs, medicine, food, water, and shelter. Long-term recovery items such as infrastructure repair and construction equipment are among the items needed during the post disaster period.
The demand patterns are also different in each phase. The pre disaster stage consists of planning processes that are mostly based on forecasts, so the demand is not certain. Once a disaster hits, demand becomes complex: high and quickly changing, but more certain based on the reported needs that are sent eventually by the assessment team in the disaster area. During the post disaster period, demand again stabilizes and becomes more predictable with the real data from the region. Overall, demand structure of disasters is complicated and challenging because of the high unpredictability of its three main dimensions: time, location, and magnitude. Also, disaster demand has other drivers related to those dimensions such as population characteristics, economy, political conditions; these factors are also complex to formulate.
Dependency of demand in disasters on these hard-to-measure factors and its high uncertainty are the main differences from the demand in regular SCs. Unlike logisticians in the private sector, humanitarian workers are always faced with the unknown: when, where, what, how much, where from, and how many times; in short, the basic parameters needed for an efficient SC setup are highly uncertain (8). Disaster demand forecasting is also difficult due to the lack of historical data. Even though there do exist some databases from the past experiences prepared by both NGOs and governments such as the EMDAT: Emergency Events Database by the Center for Research on the Epidemiology of Disasters (9), they are occasionally inadequate because of inconsistent and/or insufficient data collection and reporting problems. Additionally, disasters are unique even if they occur in the exact same location, since other factors such as population structure or economic conditions could have changed since the previous occurrence. Hence, historical data is not always very useful for predicting future demand.
DISASTER SUPPLY CHAINS
SCs link the sources of “supply” (suppliers) to the owners of “demand” (end customers). In a typical humanitarian SC, governments and NGOs are the primary parties involved. Governments hold the main power with the control they have over political and economical conditions and directly affect SC processes with their decisions. After the 2004 tsunami, for instance, the Indian government did not invite international aid agencies to participate at all in the first 60 days of the relief effort, and functioned during that period with the local sources of supplies (10). Donors, military, and the media are the other significant players in the humanitarian SCs.
Disaster Supply Chain Management
Coordination and management of disaster SCs has challenging problems. The supply network is huge and complicated with numerous players (donors, NGOs, government, military, and suppliers), and it is hard to coordinate all of them along with all the items that need to be delivered. Despite the different cultural, political, geographical, and historical differences among them, collaboration and specialization of the tasks between NGOs, military, government, and private business are increasingly needed in the humanitarian SCs (8). Despite being experienced and aware of the key points in humanitarian SCs, people in charge of logistics and SC management in most NGOs or other humanitarian organizations are not often specialized in this area; thus, they are not experts in the tools for solving the problems that might occur during the operations. There could also be domestic barriers such as the need of excessive paper work and specific policies of the region that may cause additional delays, as well as external complications due to foreign relations.
Use of technology is essential in managing SC operations and maintaining coordination, but the organizations usually do not use specific software or other technology tools for SC management. The case study written for the El Salvador earthquake in 2001 (11) focuses on coordinating the logistics and SC operations using humanitarian relief supplies management software called SUMA. SUMA’s objective is “to develop a standardized methodology and operational capacity at the national or regional level to manage relief supplies and equipment efficiently” (12). The implementation of this software allowed the identification of urgent needs, helped prevent unsolicited donations that could upset the system, and created reports with centralized information to inform the population about the development of the operations, building visibility and transparency through the SC.
Finally, goals and performance metrics of humanitarian and regular SCs differ notably. Unlike the humanitarian SCs, which do not have any profit targets and rely heavily on volunteers and donors, in regular SCs, stakeholders are the “owners” of the chain. A company has to improve its profits and aims for making more money. It is easier to prove success using profit data in private SCs, but how to prove success is not clear for humanitarian SCs because cash flow data may not fully explain results. Nevertheless, the numerous models based on minimizing cost (or equivalently, maximizing profit) for building efficient SCs can be applied to the humanitarian SCs directly or with modifications. One example is an integrated multiobjective SC model that uses fill rate, cost, and flexibility as measurement factors for simultaneous strategic and operational SC planning (13). Lodree and Taskin (14) work on stochastic production/inventory control models for recovery planning, specifically for hurricanes, to determine how long to postpone decision making to optimize the trade-off between logistics cost efficiency and hurricane forecast accuracy.
Disaster Supply Chain Execution
The execution or delivery process in an SC consists of bringing supply and demand together. Delivery could mean different things at each disaster stage, such as setting up temporary warehouses or shelters during the pre-event stage or delivering relief to affected people during the response stage.
Distribution operations within a disaster framework are very challenging. Owing to the complexity of the sourcing process and the uncertainty of demand, there could be a big gap between supply and demand in terms of mix, quantity, timing, and location, and matching them becomes a hard task (15,16). Handling expertise becomes more challenging due to the large span of relief items and the use of ad hoc warehouses. Transportation infrastructure is highly dependent on the location; it may be damaged or disrupted and suffer from dynamically changing conditions. Additionally, communication infrastructure may be disrupted. Because of the sudden onset of most disasters, problems associated with inadequate distribution planning are common, including high expediting costs, choice of wrong transportation mode and/or provider, bottlenecks in the port of entry, and incomplete execution. Hence, preparation during the pre-event stage is vital, and strategies such as the use of staging areas for prepositioning and distribution of relief supplies (17) help overcome the difficulties in getting the right supplies to the right people, at the right time, at the right place.
Humanitarian SC delivery operations may have different targets than for-profit SCs. Usually, there is a trade-off between cost and responsiveness of the delivery process. In the case of disaster SCs, responsiveness concerns human welfare, and it has a higher priority compared to operational cost than regular SCs. Since every human life is equally valuable, fairness is a particularly important criterion for disaster SCs, and their responsiveness should not be based to the social or economical condition of the affected people.
AN ILLUSTRATIVE EXAMPLE—DEBRIS MANAGEMENT OPERATIONS
Depending on the nature and severity of the disaster, and the characteristics of the affected area, there could be massive amounts of debris. The resources required to collect the debris might be limited, debris could be blocking roads and obstructing aid supply, and some types of waste can endanger community health and safety; therefore, resources have to be allocated adequately to speed up the debris collection and disposal process, while minimizing its impact. Debris collection becomes a complex logistics issue, and a debris management plan should be developed in advance addressing each disaster stage. Figure 3 shows the main elements of such a plan based on the guidelines of the Federal Emergency Management Agency (FEMA) of the United States (18). While a management plan answers the question of what to do, OR answers the question of how to do it most efficiently and effectively (19).
Figure 3. Elements of debris collection management plan.
During the pre-event stage, forecasts are made to predict the quantities and location of debris prior to disasters. Debris forecasts are used to plan for the resources, to design debris management sites for temporary storage, and to develop adequate strategies for debris final disposal. An adequate debris forecast that considers the type and extent of the potential disaster and historical data is essential for effective preparedness. During the resource planning, debris collection authorities may realize that existing forces and equipment will not be sufficient and some services must be outsourced. Choosing both the right contract (unit price, lump sum, time-and-materials, etc.) and the right contractor is fundamental for controlling the costs, quality, and availability of the procured services. Finally, a facility location model could be used to select among potential debris management sites, such that the use of these locations expedites the collection process, while minimizing additional costs.
During the response stage, having access to injured people and to critical facilities such as hospitals or police stations is crucial. Roads and streets that are essential for the emergency operations might be obstructed, and debris must be removed to facilitate rescue efforts. A network model can determine how resources should be allocated and which roads should be cleared first. Such a model takes into account the priority of the facilities to be connected in order to maximize the supplied relief, as well as the available quantity of workforce and equipment.
The longest phase of the debris collection process occurs during the recovery stage of a disaster, once emergency and other major routes are cleared from debris. After the disaster, community residents begin to take debris to public rights-of-way, and this debris along with what was left after the response stage has to be collected. Debris such as white goods (containing refrigerants and other regulated machine fluids) and hazardous waste has to be handled with special care due to environmental and health safety impact. Scheduling and routing models for the debris collection resources can be developed to minimize the impact of the remaining debris while using the available resources efficiently to reduce costs. A model of this nature assigns each resource to loading tasks at the debris sites and unloading tasks at the debris management sites. After the debris is collected, it may go through some reduce or recycle processes, and then be taken to its final destination. Deciding the best policies to follow for each debris type is not a trivial problem when different constraints such as environmental regulations, available capacities, and budget are taken into account. A mathematical model can help the decision maker to find an optimal trade-off of cost and environmental impact.
CHALLENGES AND FUTURE RESEARCH DIRECTIONS
The use of OR techniques for disaster preparedness and response can help improve the efficiency and effectiveness of processes extensively. The challenges defined in this paper about the particular structure of disaster SCs demonstrate their differences from regular SCs. Since disaster SC management is a newly emerging research area, there are many problems to be solved and new directions to be discovered for future research beyond the limited literature so far.
Most of the current research on disaster operations management focuses on the pre-event phase, which covers planning, mitigation, and preparedness processes. There is limited research in recovery planning phases. Furthermore, when a specific problem within a disaster context is modeled, the model usually focuses on one specific stage (pre-event, response, and post-event). However, there is an undeniable interaction between the decisions made at different stages, that is, what is done today affects what can be done tomorrow. Therefore, more comprehensive models integrating multiple disaster stages are needed.
One of the main characteristics of disaster SCs is the presence of multiple stakeholders (NGOs, governments, local authorities, etc.). In this multifunctional environment, each of these players might have different objectives and priorities, leading to potential conflicts and inefficiencies in operations. Also, the best decision for each stakeholder alone is not the same when the entire SC is considered, that is, inefficiencies exist when individual incentives are not aligned. To handle the system as a whole, hierarchical planning, multiobjective models, centralized/decentralized system trade-offs, standard procedures, information sharing platforms, collaboration mechanisms, and incentive models need to be developed and analyzed.
When dealing with disasters, the human factor is always present. Even the best model that reflects the real conditions does not provide the perfect solution for disasters unless it takes human factors into account. Therefore, there is a need to incorporate behavioral aspects into the mathematical models. For example, in order to design an effective evacuation model, the model should be able to predict and incorporate the different reactions and attitudes of the people. There is a limited amount of work connecting mathematical representation and behavioral dynamics (20), and this multidisciplinary approach has a high potential for more efficient solutions for disaster SCs.
Finally, there is need for models that fully describe a disaster SC system within stochastic, dynamic, and adaptive settings. A stochastic model would be able to capture the high uncertainty in demand, supply, and delivery in these SCs. Even though there is some work done modeling disaster and emergency-related problems under uncertainty, many of the models in the literature are deterministic. Incorporating the dynamics of the choices made during each time period is crucial because of the strong interactions of these decisions. The proposed models should also be adaptive to incorporate new information when a change occurs. In a disaster framework, information is usually limited at the beginning, and as time passes, more and more accurate information becomes available. The model should capture and make use of the new pieces of information.
Disaster SC management is a rising area of OR, which has the potential to make a significant impact on the society. Defining the systems accurately and developing models that fit the structure of disasters are essential for applying OR techniques in disasters applications.
We gratefully acknowledge support of our work from the Harold R. and Mary Anne Nash Junior Faculty Endowment Fund and the Focused Research Program Grant from the College of Engineering at Georgia Tech.
- Kaplan E. “Adventures in policy modeling! Operations research in the community and beyond.” Omega—Int J Manage Sci 2008;36:1–9.
- Centre for Research on the Epidemiology of Disasters. “Thirty years of natural disasters 1974–2003: the numbers”. Brussels, Belgium: UCL Press; 2004. http://www.emdat.be/Documents/Publications/publication_2004_emdat.pdf. Accessed 2009 Jan 15.
- Disaster Resource Network Humanitarian Relief Initiative (HRI). http://www.weforum.org/en/initiatives/drn/index.htm. Accessed 2009 Jan 15.
- Haddow GD, Bullock JA, Coppola DP. Introduction to emergency management. Burlington, MA: Butterworth-Heinemann; 2008. Homeland security series.
- Altay N, Green WG. “OR/MS research in disaster operations management”. Eur J Oper Res 2006;175(1):475–493.
- Autier P, Ferir MC, et al. “Drug supply in the aftermath of the 1988 Armenian earthquake”. Lancet 1990; 335(8702):1388–1390.
- Knott R. “The logistics of bulk relief supplies”. Disasters 1987; 11(2):113–115.
- Van Wassenhove LN. “Blackett memorial lecture – humanitarian aid logistics: supply chain management in high gear”. J Oper Res Soc 2006; 57(5):475–489.
- Center for Research on the Epidemiology of Disasters (CRED). EMDAT: Emergency Events Database. http://www.bepress.com/jhsem/vol4/iss2/3. DOI: 10.2202/1547-7355.1249.
- Federal Emergency Management Agency (FEMA). Public assistance debris management guide. http://www.fema.gov/pdf/government/grant/pa/demagde.pdf. Accessed 2009 Jan 15.
- Carbajal JA, Ergun O, Keskinocak P, Siddhanthi A, Villarreal M. “Debris management operations”. Humanitarian Logistics Conference; 2009 Feb 19–20; Atlanta, Georgia. Available at http://www.scl.gatech.edu/research/humanitarian/projects.php. Accessed 2009 Jan 15.
- Dixit V, Pande A, Radwan E, Abdel-Aty M. “Understanding the impact of a recent hurricane on mobilization time during a subsequent hurricane”. Transport Res Rec: J Transport Res Board 2008; 2041:49–57.