It is essential to understand how different modelling types are employed at various managerial levels in AFPSC. As mentioned earlier the tactical and operational planning models dominate the strategic models in the food industry (Table 2-4). From a distinct perspective, mathematical models have a clear dominance over the three decision levels compared to the other modelling types (Figure 2-7).
Figure 2-7: Decisions Level against Modelling Type
Single Decision Making/ Planning Level Analysis
The analysis here focuses on mapping the modelling types across a single decision- making level (e.g., operational level). Mathematical models have focused on various operational decisions related to product flow, resource hiring and allocation, planting and harvest quantities and truck/ vehicles routing. Most of these models employed either LP or MLIP techniques for formulating the mathematical relationships between system components under the deterministic assumptions of model parameters. However, few models have considered uncertainties related to demand, shelf-life decay, and resource productivity. Heuristic models are also used based on simple or meta-heuristic techniques to model
22% 17% 13% 2% 68% 58% 61% 5% 14% 11% 13% 5% 11% Operational Tactical Strategic
Decisions Level Vs. Modelling Type
decisions related to harvesting, inventory control, and vehicle routing. These models impose deterministic assumptions for exogenous variables, except a solo model that addresses the randomness behaviour in products delivery (Hsu, Hung, and Li 2007). At the operational level, only one study developed a simulation model to plan machinery usage in the harvesting operations to improve system’s efficiency and resources utilisation (Zhou, Jensen, et al. 2015). Two hybrid models (e.g. LP model and fuzzy model) are developed to relax some rigid constraints connected to uncertain factors such as costs and productivity elements in fresh produce distribution operations (Miller et al. 1997), (Broekmeulen 1998).
Figure 2-8: Modelling Type against Solo Decision-Making Level (i.e., ignore integrated decisions) From the tactical perspective, mathematical models have focused on seasonal planting and harvesting schedule in terms of crop selection (Sarker and Ray 2009), labour and resources planning (González-Araya, Soto-Silva, and Espejo 2015), and customer order planning (Grillo et al. 2016). Few of these models considered uncertainties connected to crop yield (Tan and Comden 2012) and market demand (Hu, Chen, and Huang 2014). While no studies
32% 27% 23% 3% 68% 45% 55% 5% 18% 18% 5% 3% Operational Tactical Strategic
Decisions Level (ignoring integrated decisions) Vs. Modelling Type
2013). Similarly, analytical models have focused on the coordination and cooperation decisions between supply chain parties based on game theory approach to preserve products quality and safety (Wang, Chen, and Wang 2015), costs sharing (Qi et al. 2017) and pricing products (Wang and Chen 2017). Other analytical models, such multi-criteria decision making and life cycle assessment, were used for decisions related to transportation planning (Marquez, Higgins, and Estrada-Flores 2015), packaging design (Manfredi and Vignali 2014), and inventory control (Kanchanasuntorn and Techanitisawad 2006). Some of these models considered uncertainty factors connected to product perishability and weather disruptions along with both demand and supply disruptions.
Contrary to the operational level, more simulation models are employed at the tactical level in AFPSC planning problems. In these models, decisions related to the transportation and storage conditions are modelled to assess their impact on product safety using either SD approach (McKellar et al. 2014) and DES (Rijgersberg et al. 2010). Other SD models are used to evaluate different packaging designs for fresh produce products (Orjuela-Castro, Herrera-Ramirez, and Adarme-Jaimes 2017) and explore product sourcing and imports policies (Teimoury et al. 2013). DES is also used for investigating ordering and replenishment policies for fresh lettuce retailers to reduce product loss and enhance customer satisfaction (Tromp et al. 2016). Single heuristic model is used at this decision-making level; the model is based on fuzzy sets for grading fresh fruit and segregating quantities valid for exports from entire yield (Lambert et al. 2014). One hybrid model, an evolutionary algorithm combined with an LP model, is employed to facilitate finding optimal crop planning on a macro level in order to maximise the return on investments (ROI) and secure country demand (Sarker and Ray 2009).
On the strategic level, mathematical models usually focus on decisions that are related to long-term capital investments such as food hub location and capacity design to optimise logistics costs (Etemadnia et al. 2015), planning growing areas on macro level to meet population demand and reduce water consumption (Atallah, Gomez, and Bjorkman 2014) and planning farms size and variety selection for perennial crops to optimise ROI (Catala et al. 2013). However, these models impose linear relationship assumptions between system components and ignore complexity and dynamism of planning problems. Analytical models employed at the strategic planning level are mainly used to assess the environmental (e.g., CO2 emissions) impact of AFPSCS when restructuring decisions, such as transformation from conventional to organic production (Falcone et al. 2016) and adopting recyclable packaging materials (Accorsi et al. 2014) are considered.
Although their ability to model complex and dynamic systems is apparent, only two articles have employed simulation models for strategic planning of AFPSC problems. A DES model was used to explore different configurations for fresh-cut pineapple SC between Ghana and Europe (van der Vorst, Tromp, and van der Zee 2009). One scenario is to locate the processing unit in Ghana and use air transportation for the processed products, while the other is to establish it in Europe and use sea transportation for unprocessed pineapples. Both scenarios are examined using DES against a set of environmental, economic, quality and safety measures. In the second article, an SD model was developed for a macro level planning of citrus production in Brazil (Ferreira, Batalha, and Domingos 2016). The objective of the
Integrated Decision Making/ Planning Analysis
Integrated decision making is presented in around 23% of the entire dataset (Figure 2-9). Models, that incorporate decisions at operational and tactical levels simultaneously received the highest attention (68% of integrated planning models). Most of these models have focused on harvesting, logistics functions. Models that integrate decisions at tactical and strategic levels simultaneously have received relatively less attention (22.7% of integrated planning models) and mainly focus on design and production functions. A solo application has integrated operational and strategic decisions, and another has combined the three levels simultaneously. Both applications have focused on design and logistics functions. Although the complexity increases when more than one planning level is modelled, mathematical modelling approaches are dominant models for integrated planning for AFPSC. Simulation modelling, which is a robust modelling approach suitable for complex system modelling, is employed only in 2% (just 2 papers) of the integrated planning models.
Figure 2-9: Integrated decision making in AFPSC models
At operational-tactical planning level, many mathematical models are employed. These are used to plan decisions relating to harvest, planting operations, resource recruiting and
23 Operational Tactical Strategic 38 11 1 15 1 5
use (Darby-Dowman et al. 2000) with growers’ outsourcing and cooperation decisions (Nagasawa, Kotani, and Morizawa 2009) at a tactical planning level. Other mathematical models are used to plan order quantities and product flows simultaneously with coordination between SC members (Su, Wu, and Liu 2014), supplier selection (Lin and Chen 2003) and cold storage design (Rong, Akkerman, and Grunow 2011). An Analytical model based on game theory is used to identify optimal ordering quantities between a grower and distributor and explore coordination scenarios for product pricing and sharing costs to keep the freshness of products (Cai et al. 2010). One simulation model is used at this integrated planning level. the model is used to study the dynamic behaviour of fresh produce supply chain in Netherlands using DES approach (van der Vorst, Beulens, and van Beek 2000). At an operational level, the model investigates decisions related to orders and deliveries between producer, distributor and retailer. Tactically speaking the model explores the efficacy in using an IT system to support ordering policies and allow real-time inventory management. The model is used to evaluate different scenarios regarding these decisions against a set of financial, operational, and quality indicators. A hybrid model is used at this integrated planning level for supplier selection and to optimise ordering quantities via a stochastic model complemented by an evolutionary algorithm to solve the mathematical model (Lin and Chen 2003). The objective of the model is to maximise the net profit while keeping supply and demand violations at the minimal level.
(Tsao 2013). Another LP model is used to macro level planning for cherry production in Argentina (Cittadini et al. 2008). Decisions related to Orchard design and variety selection are considered along with tactical decisions connected to labourers training programs and irrigations technologies. The objective of the model is to improve growers’ income and sustain labour workforce for this industry.
Only one application has integrated operational and strategic decisions (de Keizer et al. 2015). Design and logistics functions are modelled in this application to plan daily products flows and food hub location decision. Integration between the three levels is presented, also in one application only (Govindan et al. 2014). Similar to previous applications, designs and logistics, functions are modelled for planning facility location, the formation of transportation fleet and products flows.