Decision-making insights on the prepositioning of relief items: a systematic review Open Access

The frequency of disasters has increased over recent decades, causing significant loss of life, infrastructure damage and social and economic disruption, affecting millions of people each year (Tin et al., 2024). In recent decades, the world has experienced several catastrophic events, such as the 2011 Japan tsunami, the 2015 Nepal earthquake and the 2023 Turkey–Syria earthquake. Furthermore, as a consequence of global warming, more frequent extreme weather events are predicted (Caunhye et al., 2016).

Despite technological advancements, disasters still strike unexpectedly, leaving little or no time for preparation (Di Pasquale et al., 2020). In the immediate aftermath of a disaster, it is vital to organise relief operations and promptly mobilise and distribute essential items (e.g. food, water and medical supplies) to affected areas (L’Hermitte and Nair, 2021). Providing timely assistance relies heavily on the availability of items and can be a matter of life or death (Van Wassenhove, 2006; Overstreet et al., 2011). This highlights the need for proactive approaches through well-designed preparedness plans (Hu and Dong, 2019; Di Pasquale et al., 2020).

Prepositioning of relief items (hereafter referred to as “prepositioning”) is a key preparedness strategy in theory and practice (Van Wassenhove, 2006; Duran et al., 2011). It involves procuring and stockpiling relief items at strategic locations prior to disasters to enable swift distribution to affected communities in the post-disaster phase (Balcik and Beamon, 2008). Prepositioning was initially used by the military to deliver supplies during wartime, and the concept was later applied to emergency management. Prepositioning can significantly enhance response operations during the early stages of disasters by reducing the risk of shortages and delays (Celik et al., 2016).

The main challenges of prepositioning are the substantial investment required and the complex decision-making involved (Balcik and Beamon, 2008). As conceptualised in several studies (e.g. Turkeš et al., 2019; Rawls and Turnquist, 2010), prepositioning is multifaceted and involves decisions regarding facilities, inventory and distribution. It entails conflicting objectives related to logistics costs and service levels. Uncertainties, human suffering and long time horizons further add to its complexity (Balcik and Beamon, 2008; Bozorgi-Amiri et al., 2013). The literature involves multiple optimisation models that address this complexity and facilitate decision-making by providing insights into the costs and benefits of prepositioning and rational resource utilisation. Research in this area has increased since the seminal studies by Akkihal (2006), Balcik and Beamon (2008) and Rawls and Turnquist (2010).

While a variety of literature reviews has been published in the fields of emergency management and humanitarian logistics, dedicated review studies on prepositioning remain scarce. Prepositioning is often mentioned in broader review studies on emergency logistics (e.g. Caunhye et al., 2012; Grass and Fischer, 2016), humanitarian facility location models (e.g. Boonmee et al., 2017; Dönmez et al., 2021) and inventory management in humanitarian supply chains (e.g. Balcik et al., 2016; Ye et al., 2020). Notably, only two review studies have primarily focused on this research area, both covering research up to 2018. Sabbaghtorkan et al. (2020) concentrate on the prepositioning literature, offering a taxonomy of mathematical models and evaluating methodologies, while Sharifi-Sedeh et al. (2020) review 22 papers and identify some factors affecting prepositioning.

More recent review studies address topics such as innovations in humanitarian supply chain management (Altay et al., 2024), humanitarian aid distribution logistics and restoring access (Rojas Trejos et al., 2022), multicriteria models in humanitarian logistics (Alturki and Lee, 2024), optimisation models for response operations (Kamyabniya et al., 2024) and involvement of decision makers in humanitarian logistics modelling (Rodríguez-Espíndola et al., 2023). However, the focus of these studies is not specifically on prepositioning studies (see Table 1).

The practical applicability of prepositioning models is limited. A primary goal of research in emergency management is to provide theoretical support and tools that assist practitioners in decision-making (Kovács and Moshtari, 2019; Kovacs et al., 2019). However, quantitative models in this field are often criticised for offering limited insights into practice (Altay et al., 2021), stemming from misalignment with decision makers’ priorities and real-world situations (Rodríguez-Espíndola et al., 2023; Pedraza-Martinez and Van Wassenhove, 2016). This issue extends to prepositioning models, which have experienced limited real-world implementations (Sabbaghtorkan et al., 2020).

The effective design and implementation of quantitative models significantly depend on a qualitative foundation that outlines the essential elements to include (Keeney, 1999; Baharmand et al., 2022). Without a clear understanding of the decision problem and the priorities in emergency management, along with proper justification of assumptions and model design, these models are unlikely to be implemented in practice (Pedraza-Martinez and Van Wassenhove, 2016; Kovács and Moshtari, 2019). The current prepositioning literature and review studies predominantly focus on the modelling details while overlooking this qualitative dimension. Increased interest in this area, particularly over the past few years, and the limited applicability of prepositioning models underscore the need for a comprehensive review study on their qualitative foundations. To address this gap, it is essential not to focus exclusively on modelling details and technical jargon but to take a broader perspective on how models, as tools for supporting rational decision-making, are designed and intended to be used within the decision-making process. The aim of this study is to review the prepositioning literature and fill this gap.

Adopting an appropriate decision-making approach to guide reviewing the literature is crucial to effectively assess the underlying foundation of models and their impact on decisions. These foundations are rooted in values, which reflect decision makers’ priorities and denote what one desires to achieve by addressing a decision problem. The value-focused thinking approach has been widely applied in decision analysis to structure complex decision problems (Parnell et al., 2013). This approach emphasises that values are central to any decision-making process. The prior identification of these values is essential for defining the objectives and criteria, which then provide the structure for generating and evaluating alternatives. It ensures that decisions authentically reflect what matters most to decision makers (Françozo and Belderrain, 2022). The approach involves the following steps: recognise a decision problem, specify values, create alternatives, evaluate alternatives and select an alternative (Keeney, 1992). Using this approach, this study aims to review prepositioning literature and answer the following research questions:

By shifting the research focus from mathematical specifics to examining how decisions are defined, values are incorporated and models are developed and analysed, this study provides a foundation for researchers to align their modelling designs and assumptions with decision makers’ priorities. Furthermore, it aids practitioners in understanding the rationale behind the technical details of quantitative models. The outcomes of this research are expected to promote more evidence-based research, ultimately enhancing the practical application of prepositioning models in decision-making processes within the emergency management field.

The remainder of this paper is organised as follows: The next section describes the methodology for finding relevant literature and data extraction. The subsequent sections include the bibliographic and thematic analysis results. The Discussion section provides the discussion, contributions and implications. Finally, the last section involves conclusions and suggested avenues for further research.

A systematic literature review is a rigorous, systematic, explicit and reproducible approach for identifying, evaluating and synthesising literature related to specific research questions on a phenomenon or topic area (Okoli and Schabram, 2010; Snyder, 2019). This method has been successfully implemented across various topics within the field of emergency management (e.g. Altay et al., 2024; Rojas Trejos et al., 2022; Sabbaghtorkan et al., 2020). This research method was employed to systematically identify the most relevant literature on prepositioning for effectively addressing the research questions, synthesising the literature, drawing valid conclusions and outlining future research directions through an unbiased and methodologically sound approach.

The review was conducted by adopting the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines, which provide a structured approach for the review process, enhancing the transparency and reliability of the research (Page et al., 2021). ProQuest, Web of Science and Scopus databases were searched. These databases encompass a wide range of leading publishers, such as Elsevier, Emerald and Springer, and have been used in other reviews in this research area (e.g. Balcik et al., 2016; Sabbaghtorkan et al., 2020). The following syntax was used to search the databases within the title, abstract and keywords of papers:

The first line targets the primary subject, and the other terms refine the results within the context of emergency management. The syntax design was influenced by the most common keywords in the literature, keywords in other reviews and discussions among the authors. The database search was complemented by the snowballing method to enhance comprehensiveness.

This review focuses solely on the emergency management context, excluding business, military and crew prepositioning applications. Furthermore, since the prepositioning problem involves decisions regarding facilities, inventory and distribution, studies that exclusively focus on any single aspect are also excluded. The review is limited to peer-reviewed journals written in English and excludes book chapters, conference proceedings and reports. To ensure comprehensiveness, no publication period was specified. The database search was conducted on 01 September 2023 and repeated on 01 July 2024. Figure 1 presents the study selection process. After consolidating search results and removing duplicates, 371 papers were selected for screening. The screening was conducted in two stages: title and abstract evaluation and full-text assessment, resulting in 97 papers to review.

Bibliographic and thematic analyses were conducted on the selected papers. The thematic analysis process involves data preparation, initial code generation, theme identification, reviewing, defining themes and reporting the results (Braun and Clarke, 2006). Guided by the value-focused thinking approach and aligned with the research questions, four main themes were defined by the research team to analyse the decision-making process reflected in the models: decisions, values, alternatives creation and evaluation and final decision, while the codes evolved and were refined iteratively during the coding process. To enhance the consistency of the coding process, minimise overlaps between codes and improve reliability, the research team collaboratively discussed the coding framework and tested it on a sample data set, as recommended by Braun and Clarke (2022). The coding process was carried out using NVivo software. Table 2 shows the main themes and codes.

Figure 2 shows the annual number of published papers and the average trendline. The trend reflects the growing research interest in the area and highlights the need for literature reviews to understand the current domain and identify further research opportunities. Figure 3 categorises studies based on journal scope, illustrating that the literature primarily appears in journals focusing on modelling and optimisation across diverse fields. Managerial and emergency management-focused journals, however, have made limited contributions. The studies emphasise quantitative research and the development of optimisation models, highlighting a lack of qualitative and mixed-methods research in this area.

The research scope encompasses geographical area, disaster type and planning timeframe. The geographical scale of studies ranges from city to global levels (Figure 4). Most studies focus on local regions (59%) and cities (21%), reflecting the critical need for relief items near disaster-prone areas and contextual factors influencing decisions. Studies on a global or international scale tend to support organisations with broader scopes (e.g. Duran et al., 2011) or analyse multi-country cooperation in response to regional disasters (Rodríguez‐Pereira et al., 2021). Most studies focus on the USA, Iran and China, with some examining countries such as Mexico, Turkey, Brazil, Indonesia and Bangladesh. However, there is a lack of research focusing on prepositioning in other disaster-prone countries such as India, Japan, Pakistan, the Philippines and New Zealand.

Figure 5 illustrates the types of disasters studied, with hydrometeorological disasters, as the most prevalent type, receiving the greatest attention. Earthquakes also feature prominently due to their unpredictable characteristics and devastating impacts (Tin et al., 2024). Some studies consider multiple disaster types, such as earthquakes, hurricanes and tsunamis, occurring within the planning horizon (e.g. Duran et al., 2011) represented in the figure under the “multiple disasters” category. Other types of disasters, such as pandemics (Zhang et al., 2023a) and conflicts (Mpita et al., 2016), have received less attention.

The literature commonly considers a long-term timeframe for decision-making based on historical data and predictions about future disasters, known as long-term prepositioning (Balcik et al., 2016). However, a few studies focus on short-term prepositioning, involving operational decisions between receiving a disaster warning and the disaster occurring (Pacheco and Batta, 2016; Davis et al., 2013; Galindo and Batta, 2013; Marthak et al., 2021; Rezapour et al., 2021; Stauffer and Kumar, 2021). Short-term prepositioning is relevant for events such as hurricanes, as reliable forecasts are often available some days in advance, with more accurate updates provided at shorter intervals.

This section synthesises the thematic analysis results on the prepositioning decision-making process based on the value-focused thinking approach. It explores the key prepositioning decisions, the values that guide them, the methods to create and evaluate alternatives and make the final decision.

Facility decisions include selecting locations for establishing facilities, determining capacity levels and choosing storage methods. The models commonly assume that a set of candidate locations and possible capacity levels are exogenously given. While various attributes of possible locations, including political issues, transport network accessibility, risk exposure and human resources, are mentioned (Bozorgi-Amiri and Khorsi, 2016; Klibi et al., 2017), the selection process is rarely explained and detailed in only a few studies (e.g. Rodríguez-Espíndola and Gaytán, 2014). Having several potential locations offers more choices and enhances effective decision-making. Collaborating with other organisations and sharing facilities is one way to expand the options available. Using military storage for prepositioning and integrated facilities for civilian and military operations are proposed in some studies (Li et al., 2022; Zheng et al., 2017). Similarly, collaborative prepositioning between countries is examined in a few papers (Balcik et al., 2019; Rodríguez‐Pereira et al., 2021).

Location and construction methods influence a facility’s risk exposure. Most studies do not consider destruction risks, assuming that facilities are established in risk-free areas or designed to be resilient to disasters (e.g. Bemley et al., 2013). A few studies consider that facilities can be resilient to disasters but require higher establishment costs (e.g. Doodman et al., 2019). Alternatively, some argue that placing facilities far from demand points is impractical, and even remote areas could be vulnerable to disasters (Abazari et al., 2021b). Facilities near disaster-prone areas can offer more effective response operations, prompting a trade-off between risk exposure and operational effectiveness (Pacheco and Batta, 2016).

The number of facilities is a crucial decision that influences other key prepositioning decisions and the performance of relief operations (Balcik and Beamon, 2008). The prevalent approach does not explicitly limit facility numbers (e.g. Rodríguez-Espíndola and Gaytán, 2014). However, since establishment costs significantly impact overall expenses, some studies (e.g. Condeixa et al., 2017) limit the number of facilities, reflecting budget constraints.

Facilities can have different functions depending on distribution network design. While the literature predominantly focuses on the simplified two-tier distribution network, including facilities and demand points (e.g. Opit et al., 2014), considering additional layers could offer a more realistic depiction of the problem. One extension is to consider hierarchical multi-depots with various capacities and functions, such as central warehouses at a further distance and local facilities closer to demand points (e.g. Wang, 2024). In this setup, decisions about the location and capacity of facilities (as well as inventory and distribution decisions) should be made for different layers.

While most studies focus on permanent warehouses, some explore alternative storage methods. Using existing public facilities such as churches, schools or stadiums, as suggested in some studies (e.g. Rawls and Turnquist, 2012), incurs lower setup costs than establishing new facilities. Şahin et al. (2014) propose using freight containers to lower costs and enable faster response times by positioning items near demand points. Sharifyazdi et al. (2018) suggest that integrating onshore and offshore strategies can reduce costs without hindering response times. The mobility of shipping containers and vessels could offer unexplored advantages not covered in the above studies.

Inventory decisions involve determining the types and optimal quantities of relief items at each facility. Relief items are characterised by type, acquisition costs, required space, criticality and perishability. The items are categorised into food items (e.g. ready-to-eat meals and water), non-food items (e.g. tents, blankets and hygiene packages), medical supplies (e.g. medicine and protective material) and equipment (e.g. generators and telecommunication devices). While some studies focus on specific item types, such as food (Marthak et al., 2021) and medical supplies (e.g. Wang et al., 2022), the majority consider a combination of critical items, mainly water, food, medical items and tents. Equipment and some non-food items, such as hygiene packages, are often overlooked.

The literature considers both emergency kits (bundled items) and individual items. The former assumes homogeneous demand and ensures simultaneous delivery of specific items (e.g. beds and blankets) (Rawls and Turnquist, 2010). Nevertheless, contextual factors can lead to heterogeneous demand (Bozkurt and Duran, 2012); for example, the demand for hygiene kits after earthquakes is typically higher than after floods (Duran et al., 2011). The alternative method decides about items independently (Ni et al., 2018). This method helps to acknowledge demand diversity and takes into account features such as item criticality and perishability (Chen, 2020).

Item criticality is demonstrated by various methods. One approach categorises items into critical and non-critical, allowing for tailored decisions for each. For instance, Tofighi et al. (2016) propose central warehouses for stocking both groups, with local facilities dedicated to critical inventory. Other methods assign constant criticality weights (e.g. Balcik and Beamon, 2008) or consider various weights across probable disaster events (Rodríguez-Espíndola, 2022). The latter benefits affected individuals by prioritising needed items but adds complexity to inventory and dispatch planning.

Perishability has rarely been acknowledged. While items such as tents and equipment can be stored long-term, proper inventory planning for perishable items such as medical supplies and food is crucial to ensure the availability of usable stock (Abazari et al., 2021a). Most studies assume that relief items are non-perishable or long-lasting and that suppliers are responsible for replenishment (e.g. Shokr et al., 2022). In contrast, few studies investigate the influence of perishability on decisions, such as limited delivery timeframe (Khalili-Fard et al., 2024), specific vehicle requirements (e.g. Beiki Ashkezari et al., 2024) and return policies and periodical procurements (e.g. Akbarpour et al., 2020).

While distribution occurs post-disaster, planning ahead helps evaluate the effectiveness of prepositioning strategies for potential disasters (Zhang and Cui, 2021). Distribution decisions encompass how facilities cover the affected areas and how items are distributed to demand points following a disaster. Distribution decisions are approached in the literature through three main methods: assignment, network flow and routing. The assignment method specifies that each demand point can only be served by one facility (e.g. Şahin et al., 2014). In a network flow approach, multiple facilities can serve a demand point, requiring the determination of shipment volumes from various facilities to each demand point (e.g. Rawls and Turnquist, 2010). Routing decisions involve determining quantities and selecting distribution routes (e.g. Aslan and Çelik, 2019). To simplify classic routing problems, choices are usually limited to a predefined subset of possible routes.

The network flow approach theoretically offers cost reduction and demand coverage improvements over the assignment method by resource sharing and mitigating the shortage risks (Pradhananga et al., 2016). Moreover, routing integration could provide insights into vehicle requirements, allocation and delivery times (Aslan and Çelik, 2019). However, some authors (e.g. Elçi and Noyan, 2018; Turkeš et al., 2019) point out that prepositioning should involve strategic decisions about facilities and inventory, favouring the assignment method. They argue that detailed operational decisions are impractical at this stage because post-disaster situations can vary significantly from predictions.

The hierarchy of shipments is another aspect of distribution decisions. Lateral transhipment involves sharing and transferring inventory between facilities within the same layer to address demand and inventory mismatches (Anvari et al., 2023). It can enhance demand coverage, reduce inventory levels, and, in some studies, is proposed as an alternative to direct shipments from facilities to demand points (e.g. Wang et al., 2021a). In multi-layer distribution networks, shipment hierarchy involves either direct shipment from higher tiers to demand points (e.g. Galindo and Batta, 2013) or channelling the flow through local facilities (e.g. Ali Torabi et al., 2018). The former provides quicker deliveries but necessitates a higher level of coordination.

Transport mode choice is influenced by factors such as geographical characteristics, item type and mode attributes. Road transport is commonly considered due to its flexibility and the localised scope of most studies (e.g. Bai et al., 2018), while air transport is mainly considered in studies with a global perspective (e.g. Duran et al., 2011). Rail transport, despite its disadvantages of inflexibility and susceptibility to disruption, is examined by Yang et al. (2021). Multiple transport modes are acknowledged in a few studies. For instance, Baskaya et al. (2017) investigate the time benefits of integrating road and maritime transport, and Rodríguez‐Pereira et al. (2021) consider air and sea transport, balancing cost and response time. Some studies also consider helicopters as an alternative to ensuring access to isolated areas (e.g. Ni et al., 2018), but their availability is often limited in practice.

In summary, the Decisions subsection highlights that effective prepositioning relies on three interconnected key decision areas: facilities, inventory and distribution. Facility decisions focus on the number and location of facilities and their relevant capacities, alongside considerations for storage type, their function in the distribution network and risk exposure. Inventory decisions involve the type and quantity of items to be stockpiled, emphasising items (un)boundling and characteristics such as criticality and perishability. Distribution decisions involve planning the delivery of items to affected areas for future disasters. These decisions are illustrated by the assignment of demand locations to facilities, network flow calculations, integration of routing details into the problem and consideration of shipment hierarchy and transport mode selection.

Values are what a decision maker desires to achieve by addressing a decision problem. This subsection explores values considered in the models’ objective functions and constraints and their trade-offs. Objective functions specify evaluation criteria, whereas constraints reflect values through bounding resources and service levels. Appropriate objective functions and constraints ensure that model results align with decision makers’ priorities (Rodríguez-Espíndola et al., 2023).

Efficiency refers to the optimal utilisation of resources to achieve desired outcomes (Aslan and Çelik, 2019). In prepositioning literature, it primarily relates to monetary aspects (Rezaei-Malek et al., 2016a). Prepositioning can be costly, necessitating cost-efficiency considerations due to limited financial resources (Mohammadi et al., 2016). Models address these concerns by incorporating cost-related objectives and budget constraints.

Cost minimisation is the most common objective function (Turkeš et al., 2019), appearing as the sole objective (57 papers) or as one of multiple objectives (20 papers). The cost objective typically involves logistics costs (facility establishment, inventory and transport costs) and penalty costs linked to unmet demand (Celik et al., 2016). While some studies assume that demand should be fully met (e.g. Wang et al., 2024a), this approach can potentially lead to overly conservative and financially unfeasible decisions. A more realistic approach considers the possibility of shortages (e.g. Rawls and Turnquist, 2010). In such situations, focusing solely on minimising logistics costs can lead to the odd solution of providing no service, typically addressed by penalising unmet demand (Turkeš et al., 2019).

Shortage costs in emergency management are proxies for human suffering, making their definition complex and ethically sensitive. Some define shortage costs as high expenses to obtain resources using rapid transport modes such as helicopters (Pacheco and Batta, 2016), while the mainstream views them as prohibitive penalties for the inability to provide items, calculated by multiplying item acquisition costs (e.g. Hu and Dong, 2019). Quantifying shortage costs is criticised due to ethical concerns about assigning economic value to individuals’ suffering and the arbitrary nature of assessments (Cotes and Cantillo, 2019; Turkeš et al., 2019). Deprivation cost is another method of monetising human suffering based on deprivation time (e.g. Pradhananga et al., 2016). However, its foundations are on the concepts of willingness to pay and the statistical value of life, which, like penalty costs, remain controversial (Turkeš et al., 2019).

Monetary limits are incorporated in model constraints through various approaches. While some studies set a facility number threshold to reflect budget limitations (e.g. Balcik et al., 2019), the typical approach restricts the budget to demonstrate achievable benefits within a specified financial limit. The limited pre-disaster budget results in insufficient items, hindering response operations even with a sufficient post-disaster budget. Conversely, a deficient post-disaster budget results in the inability to distribute relief items promptly and ineffective operations (Aghajani et al., 2023).

Effectiveness reflects the ability to achieve desired results. In prepositioning studies, effectiveness refers to the speed of distribution and the extent to which demand can be fulfilled (Rodríguez-Espíndola and Gaytán, 2014), typically measured by response time, travel distance and demand coverage (Rezaei-Malek et al., 2016a).

The primary value of prepositioning is alleviating human suffering by providing relief to affected individuals as quickly as possible (Yang et al., 2021). Yet, response time is only occasionally considered a value in model design (20 papers). Some studies aim to minimise total travel time between facilities and demand points (e.g. Turkeš et al., 2019). A more realistic approach considers distributed quantities and minimises either the total weighted travel time (e.g. Tofighi et al., 2016) or the average travel time per item (e.g. Yakici, 2017). In routing problems, unlike other approaches that rely on the shortest path concept (Wang et al., 2021b), delivery times can be better illustrated by minimising the total tour time (e.g. Bozorgi-Amiri and Khorsi, 2016). As a constraint, response time appears as a coverage threshold, ensuring that demand points are covered by nearby facilities within a specific travel time (e.g. Mpita et al., 2016). Other approaches include limiting weighted travel time (e.g. Li et al., 2022) and tour time (Manopiniwes and Irohara, 2017).

Effectiveness can also be evaluated using travel distance criteria. While not explicitly stated, travel distance is often regarded as a proxy for travel time (e.g. Aghajani et al., 2023), with objectives and constraints resembling those of travel time criteria. Objective functions minimise the total distance between facilities and demand points (Şahin et al., 2014), quantity-weighted travel distance (Velasquez et al., 2019) or average travel distance per item (Baskaya et al., 2017). The constraints include the travel distance of items (e.g. Rezapour et al., 2021) and the service radius (e.g. Zheng et al., 2017). While the former helps in considering shorter travel distances for critical or perishable items, the latter encourages a more decentralised prepositioning network, ensuring quicker responses (Alem et al., 2021).

Another effectiveness-based value is providing service to as many individuals as possible, reflected in demand coverage (shortage level) criteria. In studies with demand coverage objectives, shortages are not penalised and monetary aspects are represented either through cost-efficiency objectives (e.g. Shokr et al., 2021) or budget constraints (e.g. Alem et al., 2021). Some studies focus on maximising expected demand coverage (Turkeš et al., 2021) or maximising the probability of fully meeting demand (e.g. Shu et al., 2022). Yet, incorporating items’ criticality (e.g. Bemley et al., 2013) and prioritising demand points (e.g. Rodríguez-Espíndola, 2023) can better reflect the needs of affected populations. Demand coverage appears as a threshold on the coverage level of demand points (e.g. Qi et al., 2023) or as a confidence level for full coverage (Bai, 2016). The former allows managing demand satisfaction for specific items and locations, while the latter enables defining a reliability level for the network (Rawls and Turnquist, 2011).

Equity is a crucial value because affected individuals expect fair treatment, irrespective of their background (Rezaei-Malek et al., 2016a). Demonstrated through two approaches, vertical equity involves providing resources based on varying needs and priorities, while horizontal equity focuses on equal resource allocation among all individuals (Tofighi et al., 2016). Vertical equity is mainly implied in the problem definition, seen through demand point priorities (Bozorgi-Amiri and Khorsi, 2016), item criticality weights (Rodríguez-Espíndola, 2022) and different shortage costs across demand points (Shokr et al., 2022).

Horizontal equity appears in objective functions and constraints. The objectives address disparities by minimising differences in demand coverage (e.g. Akbarpour et al., 2020) or response time (e.g. Manopiniwes and Irohara, 2017), typically observed in multi-objective models to balance equity with efficiency and effectiveness. Horizontal equity constraints define tolerable disparity of demand satisfaction (e.g. Rezaei-Malek and Tavakkoli-Moghaddam, 2014); however, setting such thresholds could be ethically challenging.

In the emergency context, similar to the commercial sector, a trade-off exists between costs and service level (Balcik et al., 2016). Yet, an ineffective response here can mean the difference between life and death (Van Wassenhove, 2006), underscoring the critical need to evaluate prepositioning costs and benefits thoroughly. Single-objective models provide insights using efficiency-based objectives by considering shortage/deprivation costs or effectiveness-based objectives with budget constraints. However, multi-objective models visualise trade-offs between conflicting objectives more clearly, offering deeper insights and assisting decision makers in balancing criteria based on their preferences (Rodríguez-Espíndola et al., 2018). Despite these advantages, single-objective models remain prevalent. Solving single-objective models is already computationally challenging for real-world instances (Turkeš et al., 2021), and multiple objectives increase this complexity.

Most multi-objective models (19 of 25 papers) are bi-objective, mainly aiming to balance costs against effectiveness or equity. In contrast, few choose coverage level and response time objectives, addressing the problem without shortage/deprivation costs inclusion (e.g. Turkeš et al., 2019). Studies with more than two objectives primarily analyse trade-offs between efficiency, effectiveness and equity, providing implications on the costs and benefits of different strategies (Aslan and Çelik, 2019). For instance, Bozorgi-Amiri and Khorsi (2016) illustrate the inverse relationship between costs and shortage levels and the conflict between travel time and shortages.

Incorporating decision makers’ preferences properly into multi-objective model designs ensure that trade-offs effectively reflect the decision makers’ values. The stage at which preferences are included divides these models into a priori, a posteriori and interactive (Akbarpour et al., 2020). The a priori methods assume that decision makers provide clear preference information in advance, either through assigning importance weights to objectives (weighted sum) (Wang et al., 2022) or providing ordinal preferences on objectives (lexicographic order) (Turkeš et al., 2021). Despite their computational simplicity, these approaches pose practical challenges because decision makers are unlikely to specify such information without detailed knowledge about alternatives and their probable payoffs.

In an a posteriori approach, preferences are not required in advance, and decision makers are provided with a set of best possible trade-offs (the Pareto set) for further analysis and choice (Bozorgi-Amiri and Khorsi, 2016). A commonly used method is defining various hypothetical weights for objectives (weighted sum) and solving the model multiple times (e.g. Manopiniwes and Irohara, 2017). Other prevalent methods, using similar iterative processes, aim to minimise the weighted deviation of objectives from a target level (Lp-metric) (e.g. Abazari et al., 2021a) or optimise one objective by representing others as constraints with pre-specified upper bounds (epsilon-constraint) (e.g. Rodríguez-Espíndola, 2023). The a posteriori approach, though offering a thorough view of trade-offs, can be computationally demanding and overwhelm decision makers with a large number of solutions requiring further analysis.

In the interactive approach, decision makers actively engage by providing preferences during iterative solving, guiding the process towards more desirable solutions or terminating it upon achieving satisfaction (e.g. Rezaei-Malek et al., 2016b). This approach is less computationally demanding; however, decision makers may not always be available to participate in such a typically time-consuming process. Another disadvantage compared to an a posteriori approach is that decision makers must make decisions with limited information about the alternatives and potential trade-offs.

The Values subsection identifies the key values influencing decision-making in prepositioning, including efficiency, effectiveness and equity, along with their trade-offs. Efficiency is primarily driven by cost considerations relevant to logistics and shortage costs, with the latter raising ethical concerns about monetising human suffering. Effectiveness focuses on ensuring timely and adequate relief delivery, evaluated through response time, travel distance and demand coverage. Equity highlights fair treatment through need-based (vertical) and equal allocation (horizontal) approaches. Analysing the trade-offs among these values is essential for informed decision-making. This requires incorporating decision makers’ preferences into modelling and analysis, whether before, during or after the solving process, with each approach presenting specific computational and practical challenges.

In prepositioning problems, alternatives refer to feasible configurations of decisions regarding facilities, inventory and distribution. The method for generating and evaluating alternatives involves formulating the problem into optimisation models and solving them numerically using context-specific data. This subsection analyses models based on uncertainty capturing and planning dynamics and discusses solution approaches regarding accuracy and computational efficiency.

Prepositioning decisions must be made under uncertainty about the timing, location and magnitude of disasters and their impact (Rawls and Turnquist, 2010), necessitating proper modelling approaches to incorporate uncertainty. While most studies acknowledge uncertainty, few papers use deterministic models (e.g. Lee et al., 2014). Although some of these deterministic models focus on unique variations of the problem, such as lateral transhipment (Baskaya et al., 2017), offshore prepositioning (Sharifyazdi et al., 2018) and using shipping containers (Şahin et al., 2014), the neglect of uncertainty hampers their applicability.

Common uncertain parameters include demand variability, transport network disruptions and supply availability. Demand estimation relies on demographic and historical disaster data sourced from reliable sources, e.g. EM-DAT (e.g. Hu et al., 2023), specialist reports (e.g. Ali Torabi et al., 2018) and field experts (e.g. Shehadeh and Tucker, 2022). Disasters may disrupt transport networks, leading to decreased capacity or longer alternative routes. This is reflected in inflated transport costs (e.g. Jiang et al., 2023), increased transport time (e.g. Davis et al., 2013), reduced capacity (e.g. Wang and Nie, 2019) and extended travel distances (Abazari et al., 2021a). Moreover, supply uncertainty due to facility destruction is illustrated as a partial (e.g. Condeixa et al., 2017) or complete loss of inventory (e.g. Shehadeh and Tucker, 2022). Some other parameters, such as shortage costs (Mete and Zabinsky, 2010), budget (Shehadeh and Tucker, 2022) and items’ criticality (Zheng et al., 2017), are also considered uncertain in a few studies.

Determining an uncertainty paradigm depends on information availability and the decision makers’ risk attitude (Che et al., 2024). Scenario-based stochastic programming is the prevalent method, which assumes sufficient information is available to model future disasters as scenarios, including their occurrence probabilities and impact on uncertain parameters (Wang et al., 2023b). Conventional stochastic programming is risk-neutral, aiming to optimise the average expected performance across future scenarios. However, it may perform poorly in case of less probable catastrophic events (Hong et al., 2015). To address this, some studies use risk-averse variations of stochastic programming (e.g. Das and Hanaoka, 2013), which are less sensitive to uncertain parameter fluctuations but may compromise on objective function optimally (Kelle et al., 2014).

The challenge in implementing stochastic programming lies in the difficulty of providing appropriate information due to data scarcity. Alternatively, robust optimisation assumes that future events cannot be explicitly represented as scenarios; instead, uncertain parameters are defined within bounded ranges (Jiang et al., 2023). Robust optimisation aims to find solutions that are resilient to all possible future events by optimising performance under the worst-case rather than under the expected outcomes. This approach is risk-averse but overly conservative, which is effective for extreme situations but can lead to high costs and sub-optimal outcomes for events with average or low impact (Wang et al., 2023b).

Distributionally robust optimisation, applied in some studies since 2021 (e.g. Wang et al., 2023a), bridges the gap between stochastic programming and robust optimisation (Zhang and Chen, 2023). It assumes partial information about uncertain parameters is available (Zhang et al., 2022), aiding in generating a set of potential probability distribution functions to find the optimal solution under the worst-case distribution (Zhang et al., 2023b). While risk-averse, this approach is less conservative than robust optimisation, which may neglect some available information.

Planning dynamics involve updating and adjusting decisions to adapt to changing circumstances. The commonly observed planning approach is the static single-period method (75 papers), which treats the pre-disaster and post-disaster phases as single periods. A few papers use dynamic multi-period planning to overcome this method’s limitations in capturing some aspects of time and periodic decisions.

Static pre-disaster and dynamic multiple post-disaster periods enable more precise decisions about distribution and post-disaster procurement (e.g. Klibi et al., 2017), capturing demand fluctuations (Wang et al., 2023b), differentiating early and late deliveries (Mahootchi and Golmohammadi, 2017) and distribution planning for primary and secondary disasters (Wang et al., 2024b). While theoretically addressing some drawbacks of static models, the availability of detailed demand information in different periods in advance is questionable (Turkeš et al., 2019). Assuming a multi-period pre-disaster horizon also helps in considering periodical pre-disaster decisions such as replenishing perishable items (e.g. Rabbani et al., 2015), expanding the transport network (e.g. Wang and Nie, 2022) and making long-term facility capacity adjustments (e.g. Alem et al., 2021).

Drawing from decision-making theory, the solution approach evaluates alternatives according to the decision makers’ values. In optimisation models, this step identifies and numerically compares feasible alternatives based on objective function(s) to find the optimal (or near-optimal) solution(s). The choice of solution approach is influenced by factors such as computational resources, complexity of models, solution time and the level of accuracy.

Regarding accuracy, solution methods in the literature are categorised into exact and approximate. Exact methods evaluate all possible solutions and ensure optimal results. Solver packages can find exact solutions within an acceptable time for small to medium-sized problems (Wang and Nie, 2022). However, for real-size instances and complex models, they may become ineffective (Hu et al., 2023), necessitating the use of alternative methods. While some use supplementary methods to enhance the time efficiency of exact approaches (e.g. Noyan, 2012), others use approximate methods to solve computationally challenging problems. These methods include heuristics and metaheuristics, which may not always yield the absolute best solution but can often achieve a near-optimal solution within a reasonable timeframe. Notably, while time pressure favours approximate methods for short-term prepositioning, current studies either use solver packages (e.g. Galindo and Batta, 2013) or do not provide solution approach details (e.g. Stauffer and Kumar, 2021).

In essence, this subsection discusses key aspects of optimisation models as the primary approach for creating and evaluating prepositioning alternatives in the literature. The primary factors in these models include uncertainty capturing, planning dynamics and the solution approach. Uncertainty is addressed through stochastic programming, robust optimisation or distributionally robust optimisation, depending on data availability and the attitude to risk. While static single-period models are favoured for their simplicity, dynamic multi-period models provide adaptability but rely on detailed information about future events, which is often scarce. The solution approach ranges from exact methods for problems with smaller sizes to heuristics and metaheuristics for more computationally challenging ones, balancing precision with computational time efficiency.

The final step in a decision-making process is to implement the results into real action. This subsection discusses the methods for analysing model outcomes to gain decision insights, identifies the practical applications of current studies and explores the extent of interaction with practitioners in different stages of the decision-making process.

Analysing model results is essential to derive managerial insights into prepositioning decision-making (Bemley et al., 2013). Sensitivity analysis is widely used to examine how parameter changes affect the optimal prepositioning strategy and to identify the most critical parameters. To illustrate, Balcik and Beamon (2008) show how increasing the pre-disaster budget decentralises operations while adequate post-disaster funding promotes a centralised strategy with fewer, larger facilities. Based on such analyses, decision makers can either confirm the optimised solution(s) or adjust input parameters and re-evaluate. Sensitivity analysis is conducted on various parameters based on problem specifications, modelling type and decision makers’ preferences. Budget levels (e.g. Yu, 2023), penalty costs (e.g. Rawls and Turnquist, 2012), response time thresholds (e.g. Mpita et al., 2016), demand coverage constraints (e.g. Klibi et al., 2017) and reliability levels (e.g. Rawls and Turnquist, 2011) are commonly analysed due to their high impact on results. Turkeš et al. (2022) argue that while conventional sensitivity analysis focuses on limited parameters in one context, more meaningful policy proposals could emerge by considering multiple case studies and exploring interactions between various factors.

For a posteriori models, a thorough analysis of the Pareto set by decision makers is essential to explore trade-offs between objectives and derive meaningful insights (Bozorgi-Amiri and Khorsi, 2016). Yet, analysing numerous solutions to select the preferred option can be overwhelming. Rezaei-Malek et al. (2016a) address this challenge by using a multi-criteria decision-making approach, assisting decision makers in the final choice. Similarly, Brito Junior et al. (2020) implement a multi-criteria decision-making approach to compare cost-optimal and suboptimal solutions by incorporating other attributes, e.g. human resource availability and infrastructure accessibility.

Only two studies report implementing their results in prepositioning practices (Duran et al., 2011; Brito Junior et al., 2020). This lack of practical applications underscores the gap between research and practice, echoing the findings of other studies in this area (e.g. Rodríguez-Espíndola et al., 2023; Kovács and Moshtari, 2019).

Efficient communication between researchers and practitioners, as highlighted by Kovacs et al. (2019), can enhance research applicability by incorporating practitioners’ requirements and realistic assumptions. Such interactions are reported by only one-third of the studies. Some papers refer to communications with authorities that resulted in defining decision problems (e.g. Mohammadi et al., 2016). Ten studies reflect interactions about values and their trade-offs. For instance, Rezaei-Malek et al. (2016b) mention that decision makers value cost-efficiency, demand satisfaction and response time, and the weights of the objectives are determined by practitioners in Tofighi et al. (2016).

Communications commonly involve practitioners assisting in alternative creations by providing input data such as scenario probabilities (e.g. Condeixa et al., 2017), demand levels (e.g. Shokr et al., 2021) and other details such as candidate locations, costs and items’ criticality. Despite its importance, presenting research results to decision makers and gathering their practical input on the applicability of outcomes is reported in only seven papers. For example, Shokr et al. (2022) confirm that their results are validated by relief managers, and Duran et al. (2011) mention that their results are used for fundraising.

In brief, the Final decision subsection examines the final decision-making phase in prepositioning problems, which focuses on analysing model outcomes to extract actionable insights. Sensitivity analysis and multi-criteria decision-making approaches are used to interpret results and balance trade-offs. A notable gap in the current literature is the limited interaction with practitioners and the lack of practical implementation of existing models.

Reviewing prepositioning literature within the value-focused thinking approach facilitates interpreting the problem from a decision-making perspective. The problem involves identifying the optimal prepositioning strategy within a specified research scope, where each strategy is characterised by multiple decisions. Based on the thematic analysis results, the main decision categories are identified: facilities, inventory and distribution. These are guided by three groups of values: efficiency, effectiveness and equity. The former relates to costs, while the others relate to prepositioning benefits.

Optimisation models are predominantly used to assist decision makers in creating, evaluating and choosing prepositioning strategies. Uncertainty is a main source of complexity in the emergency management context (Overstreet et al., 2011), and it is addressed using multiple approaches depending on the availability of information and the risk attitude. The primary approach to planning is static, but some studies develop dynamic models to better reflect adjustments needed based on evolving situations. The final decision is guided by sensitivity analysis, which informs how decisions change if assumptions vary, as well as objectives trade-offs.

Extensive quantitative literature on prepositioning exists, but unvalidated model designs and assumptions limit their applicability, in line with the observation of Altay et al. (2021) within the broader scope of humanitarian logistics. Our findings highlight that most studies lack explicit interactions with practitioners. We call for a greater emphasis on the applicability of studies. This necessitates more qualitative research and pragmatic approaches, such as mixed methods, a suggestion also made by other researchers (e.g. Maharjan et al., 2020) and developing models grounded in real-world contexts and practitioners’ values with clear justifications for underlying assumptions. As emphasised by Keeney (1999, p. 25): “The fact is that no quantitative model can be developed or used without a qualitative foundation that describes what is important to include in the quantitative model”.

Grounded in our findings and informed by the value-focused thinking approach, a framework for evidence-based research on prepositioning has been developed (Figure 6). This framework demonstrates how qualitative insights can guide quantitative research and inform decision-making. While the decision-making process inherently involves iterations and interconnections, the framework presents a linear approach to enhance clarity and understanding. This framework is expected to serve as a foundation for future research by outlining key steps and decisions at each stage, facilitating interactions between researchers and practitioners, and increasing the applicability of research models.

While we specifically targeted the prepositioning literature and reviewed it through a novel lens, comparing our findings with current review studies, whether within the same domain or on a broader scope, enhances our understanding of the commonalities, differences and overall validity of our conclusions. We investigated the qualitative foundations of prepositioning models and their implications for model design and results analysis, areas that had not been previously addressed in the literature. In terms of values in objective functions, our results resonate with Sabbaghtorkan et al. (2020), who noted that most prepositioning studies emphasise supply-side cost functions. This is further supported by Caunhye et al. (2012) and Rodríguez-Espíndola et al. (2023), who observed that the majority of emergency logistics models use only a single objective function, predominantly focused on cost, similar to the commercial sector. Although cost-efficiency and budget constraints are significant considerations, we argue that fundamental values related to effectiveness and equity should be more explicitly integrated into model design.

Notably, while no previous reviews reported real-world applications of prepositioning models, we identified two studies where their findings were applied in real-world cases. However, the gap between research and the practical applicability of prepositioning studies remains significant. We examined the involvement of practitioners in prepositioning studies, a crucial factor that significantly enhances the practicality of models and was previously unexplored. Our findings align with Kovács and Moshtari (2019) and Rodríguez-Espíndola et al. (2023), who cautions against the lack of collaboration between researchers and practitioners in the emergency management field, often leading to models disconnected from real-world situations. In particular, we found that only about 10% of prepositioning models are developed in consultation with practitioners, consistent with the results of Rodríguez-Espíndola et al. (2023) within the broader scope of humanitarian logistics models, confirming the lack of practical insights in model development. Additionally, our observations regarding some simplistic assumptions in models, such as the emphasis on two-tier network design and the oversight of more complex networks, as well as the scarcity of research on perishable items, echo the findings of Balcik et al. (2016) and Dönmez et al. (2021), respectively.

The increasing frequency and severity of disasters highlight the crucial role of disaster preparedness strategies, with prepositioning playing a pivotal role. Decision-making in prepositioning is complex due to its multifaceted nature, the inherent uncertainty of disasters, and its high dependence on the context. A key takeaway from our research is the need to enhance the practical application of prepositioning models by incorporating decision makers’ values and preferences, and more realistic assumptions to support rational decision-making. To achieve this, modellers must move beyond technical optimisations and adopt a broader approach that includes qualitative insights. This research contributes to this shift by identifying key decisions in prepositioning, the fundamental values involved, and how these values should be integrated into model design. Decisions, values, preferences and thresholds in model design should be informed by qualitative studies and practitioner collaboration, with well-justified choices for model components, solution approaches and result analysis methods. This study highlights the importance of grounding models in qualitative insights and working closely with practitioners to ensure their relevance and practical applicability.

This research makes several contributions to emergency management knowledge. Unlike previous reviews, this study adopts a broader perspective by reviewing papers through a decision-making lens to unveil the qualitative basis of current models, thereby addressing a gap in the literature. By reviewing various approaches for defining decisions, incorporating values, developing models, analysing results and explaining how they impact decision-making, this research enhances understanding of prepositioning strategies beyond their mathematical details. The proposed framework further supports this insight by outlining key decision-making steps and illustrating how the qualitative foundation can guide quantitative research and facilitate evidence-based research. This foundation is beneficial for researchers in rationalising their modelling design and assumptions, thereby prompting more evidence-based research in future.

This review also has significant practical implications. It helps practitioners better understand current quantitative models and reveals the rationale behind their mathematical intricacies. This knowledge is helpful in applying these tools to real-world decisions, contributing to the practical application of models. Additionally, the proposed framework establishes a common ground for more effective communication between researchers and practitioners, promoting closer collaboration and enabling more practical studies. This ultimately contributes to more effective emergency management practices and alleviates human suffering.

The literature on prepositioning has grown in recent years. Currently, many studies focus on developing optimisation models aimed at providing managerial insights relating to the implementation of prepositioning strategies. However, there is limited evidence of applying these models to real-world situations, highlighting the necessity for a review of the foundations used in their modelling design. In contrast to existing reviews, the objective of this research is not to develop taxonomies on modelling methods or influential factors. Instead, we aimed, via our research questions, to step back and gain a wider view of the decision-making process and the reasons behind model choices.

A systematic literature review was conducted to understand better what decisions are involved in the prepositioning problem, which values guide them, and how model design choices influence these decisions. The selected literature was analysed bibliographically and thematically guided by the value-focused thinking approach, shedding light on the fundamentals used to develop mathematical models. Based on our findings, we proposed a decision-making framework for prepositioning problems that establishes a foundation for future research by identifying key steps and decisions at each stage. This framework is expected to foster collaboration between researchers and practitioners, leading to the development of more evidence-based research.

This review excludes studies that concentrate on a single decision category: facility, inventory or distribution. Yet, those studies could provide insights into decision-making values and contribute to developing more practical models. Moreover, this study focuses on decisions about prepositioning and does not examine the impact of other decisions, such as framework agreements with suppliers, post-disaster procurement, road repairs and expansions and evacuation plans, on prepositioning strategies.

This subsection outlines our recommended directions for future research along with specific research question(s) tailored to each proposed avenue. One future research direction is to investigate current prepositioning plans and practices and the decision-making methods used by emergency organisations in formulating and implementing these strategies. Adopting methodologies such as action research or ethnography enables researchers to gain real-life insights and better understand mutual interactions, from problem recognition to a final decision. This would enhance understanding of actual emergency challenges and the specific needs of practitioners. Additionally, it enables practitioners to gain insights into academic approaches to emergency management problems, ultimately enhancing emergency operations, as emphasised by Kovács and Moshtari (2019). Consequently, the following research questions should be considered:

The literature primarily features single case studies in a few countries. Given the influence of contextual factors on prepositioning strategy, we recommend exploring other disaster-prone countries and incorporating multiple case studies resonating with Turkeš et al. (2022) and Turkeš and Sörensen (2019). Apart from their potential policy implications, such studies provide the grounds for comparing different strategies and analysing the impact of contextual issues on decisions. Accordingly, the following research question arises:

Despite their potential for addressing conflicting objectives, multi-objective models have rarely been used. Cost minimisation remains the most prevalent single objective, where human suffering is quantified as shortage/deprivation costs, making the implementation of these proxies controversial. While unavoidable, the trade-off between costs and human suffering can be more effectively illustrated and analysed through multi-objective models, avoiding the need to monetise human suffering and relevant ethical concerns. Thus, we suggest investigating the following question:

Establishing facilities involves expensive decisions with long-term consequences. The typical approach is to select some locations from a predefined set of candidates. Most studies assume this set is predetermined, overlooking the importance of the selection process. A future possibility is to integrate this step as a preliminary phase of studies. Moreover, although alternative storage methods are used in practice, they are less acknowledged in academia. Their unexplored advantages and disadvantages require deeper analysis. Therefore, the following research questions should be addressed:

Successful prepositioning practices depend on item availability, highlighting the importance of inventory decisions. Most studies focus on a limited range of life-saving items, often overlooking non-food items and equipment such as hygiene kits and generators, which are crucial in real-life situations. Moreover, despite the practical challenges posed by the perishability of food and medical items in implementing prepositioning, a common assumption is that stocked items are long-lasting or that their replenishment is the responsibility of suppliers without considering the associated costs for the organisation. A deeper investigation into the types of items and their characteristics would improve the applicability of future studies. Given these directions, the key research question is:

Another future research avenue is to review how prepositioning is integrated with other essential decisions in emergency management, such as framework agreements, post-disaster procurement, road repairs and expansions and evacuation plans. Examining these connections would provide valuable insights into how these decisions affect prepositioning strategies. Therefore, the following research question should be explored:

This study was supported by Te Hiranga Rū QuakeCoRE, a Centre of Research Excellence (CoRE) funded by the New Zealand Tertiary Education Commission. The QuakeCoRE publication number is 0982. QuakeCoRE’s contribution was limited to the financial support provided. QuakeCoRE did not play any role in the research process from study design to submission.