Adoption of Multi-Modal Transportation for Configuring Sustainable Agri-Food Supply Chains in Constrained Environments

Abstract

Agri-food supply chains have the potential to make a significant contribution to achieving sustainable development goals through ongoing improvements in their configurations. A range of strategic, tactical, and operational level decisions pertaining to the design and operation of sustainable supply chains have been studied in the extant literature. However, investigations into the adoption of multi-modal transportation as a strategic decision in the context of agri-food supply chains operating in constrained environments are limited. As such, in this study, the adoption of bi-modal transportation for the domestic vegetable supply chain in a developing country context under certain constraints was examined. A mixed-integer linear programming model was developed to determine the volume and direction of the product flow to achieve the minimum total food-miles and smallest emissions footprint. As a case study, a Sri Lankan mainstream vegetable supply chain was used to investigate the applicability of a combination of truck and railway modes to transport vegetables from farms to retailer locations via economic (consolidation) centers. The adoption of a bi-modal transportation structure demonstrated the potential to reduce food miles by 32%, transportation costs by 36%, contributions to global warming potential by 35%, and empty truck hauls by 38%, compared to a structure with truck-based, uni-modal transportation.

1. Introduction

Agri-food systems are faced with the ongoing challenge of meeting the dietary needs of a growing population while mitigating negative impacts on the environment. Towards addressing this challenge, the adoption of a range of innovative practices, including technology implementation for smart farming, post-harvest management through effective logistics operations, and improved traceability and transparency, aligned with sustainable development goals, has been proposed in the literature [1,2]. Logistics related costs, carbon emissions, wastage, and losses are among the main challenges that need to be addressed in designing sustainable agri-food supply chains [2]. Reducing food miles across the entire agri-food supply chain, a key measure pursued in addressing these challenges, has the potential to make a significant contribution to achieving sustainability goals as it affects product quality, shelf-life, and environmental impact [3,4]. The use of alternative transportation modes or multi-modal transportation is a widely practiced approach for minimizing food miles while maximizing resource utilization through economies of scale, subject to practical constraints [5].

Multi-modal transportation solutions, which combine different transportation modes such as road, rail, canal, ocean, and air, seek to address the trade-offs among the cost, speed, and volume associated with the transportation of the cargo involved. As such, they allow for the timely and efficient distribution of goods to meet customer demand, while minimizing the carbon footprint [3]. However, multi-modal transportation systems need to be designed and implemented to suit the specific attributes of the agri-food system concerned and its environment. For instance, the characteristics and requirements of different supply chains, the infrastructure conditions and economic status of different countries, the nature of trade arrangements, and industry regulations across regions could all impact the choice among alternative logistics systems for agri-food supply chains [6]. In general, supply chains can benefit from multi-modal transportation by way of improving the speed of delivery, flexibility, and reliability, while reducing transportation costs and environmental impact [7]. Researchers have further noted that inter-modal transportation services and multi-modal transportation infrastructure play central roles in the logistics systems of businesses competing in global markets [8]. Despite these advantages, there are a number of challenges, such as assessing benefits and supporting infrastructure and associated costs, as well as effective coordination across the supply chain that need to be addressed when adopting multi-modality [6].

In large geographical territories such as the USA, the EU, India, and China, and where there are well-developed infrastructure settings, the potential for developing multi-modal transportation systems to serve domestic markets is greater than in small countries with limited infrastructure options. However, compared to the research effort directed at investigating multi-modal systems in developed counties, only a few detailed investigations have been carried out in developing counties and/or smaller geographical contexts. Therefore, this study aims to fill a significant gap in the existing literature by exploring the use of multi-modal systems for transporting agri-food products in a constrained environment [9] using Sri Lanka, a developing country with a small geographical area and limited transportation infrastructure options, as a case study. Post-harvest losses associated with packing, loading, unloading, and transportation within the Sri Lankan vegetable supply chain have been estimated at 20–40% [10]. Furthermore, a significant portion (20%) of the post-harvest losses occur during transportation, which utilizes small trucks with a capacity of 5–10 Ton [10]. The current practices also result in high transportation costs due to the large number of trips involved and empty return trips, as well as wastage due to excessive handling with multiple instances of loading and unloading. Comparatively, rail transportation is a cost-effective and environmentally friendly option, which has a greater transportation capacity and the ability to reduce empty hauls. Even though all major economic centers can be accessed via the existing railway network, the Sri Lankan agri-food supply chain has not benefited from using the railway transportation system so far, primarily due to the lack of recognition of its potential contribution.

This study proposes a bi-modal transportation strategy as a solution to overcome certain challenges faced by the current truck-based uni-modal transportation system, by eliminating unnecessary truck transportation through combined truck and rail transportation. Current truck transportation and the proposed bi-modal transportation have been modeled as a mixed-integer linear programming problem for comparing transportation costs, travel distances, and emissions intensity. As such, the study aims to address the following research questions towards enhancing the sustainability of the agri-food supply chain.

RQ1: Which segments of the supply chain can be switched to a bi-modal structure?

RQ2: What proportion of the total crop volume can use the bi-modal structure?

The rest of this paper is organized as follows. Section 2 outlines the literature review, informing the key decisions pertaining to sustainable agri-food supply chains and the use of multi-modal transportation in agri-food supply chains, highlighting the research gaps. The methodology followed in the study is presented in Section 3. Section 4 discusses the resultant vegetable supply chain configuration performance and the practical implications of the study. Section 5 presents the overall findings of the study along with its contributions, limitations, and the directions for future research on the topic.

2. Literature Review

In comparison to other types of supply chains, such as for manufactured goods, decisions related to agri-food supply chains possess certain attributes concerning the nature of the cargo being handled, as well as the structural and behavioral dimensions of the network, which have been paid comparatively less attention in the literature [11]. With reference to the different stages of agri-food supply chains, the decisions addressed can be categorized into the following areas: (i) upstream—planting (e.g., crop type selection [12,13], area allocation [14], rotation [15], harvesting (e.g., harvest amount planning [15], and scheduling [16]); (ii) mid-stream—processing (e.g., facility location, e.g., [17,18], capacity addition [18], and inventory planning [13]); and (iii) downstream—distribution (vehicle selection and routing [19,20], technology adaptation for logistics [21], and transportation mode selection [22,23]). Depending on the applicable planning horizon, these decisions may belong to either strategic, tactical, or operational levels. Many studies have focused on the decisions at the single planning level, mostly on strategic level decisions, and noticeably limited attention has been paid to the integrated decisions, particularly configuration-related decisions including entity selection across the entire supply chain accounting for the volume of goods handled [19,23,24]. Since ”food miles” are a critical aspect of agri-food supply chains due to the perishable nature of the products involved, there is considerable attention towards decisions such as vehicle routing and transportation mode selection; however, the literature on these topics is limited to uni-modal transportation [25].

Multi-modality is considered to be an essential element of transportation networks. Simply stated, multi-modal transportation refers to using a mix of transportation modes for moving goods, which involves additional handling, transit, and consolidation and bulk-break activities compared to uni-modal transportation. With the increasing volumes of global trade, multi-modal transportation provides greener transportation options to logistics companies, partly due to the more efficient operations and smaller carbon footprints involved [26]. The potential for developing multi-modal transportation systems to serve domestic markets in large countries, such as the US, India, and China where transportation services are efficient with no intermediate handling, is greater than in small countries like Sri Lanka where most goods are transported by short hauls [6]. According to a study undertaken in the Arctic region, it was illustrated that supply chains can benefit from multi-modal transportation by improving the speed of delivery, flexibility, and reliability while reducing transportation costs and environmental impact [7]. The majority of multi-modal-related studies have focused on freight transportation between supply chain stages. For example, ref. [27] investigated the viability of road, rail, and short-sea shipping to transporting cargo in a three-echelon supply chain minimizing transportation costs and emissions. In comparison, ref. [28] addressed strategic decisions including supplier selection, product allocation, and network configuration through multi-modal transportation to design a carbon-market-sensitive green supply chain network. In [29], supply chain formation in the context of multi-modal transportation has been examined by leveraging multi-agent systems and dynamic negotiation processes. This study has included carrier and transshipment mode selection using a temporal constrained shortest path to minimize the total transportation and transshipment costs [29]. Overall, multi-modal transportation has been identified as an effective approach to enhance supply chain performance.

In the supply chain management literature, the adoption of multi-modal transportation has been investigated for both perishable and non-perishable food products. Under the non-perishable food category, grain supply chains have been examined, specifically in developing countries such as India and Ukraine. For example, a study by [30] on the Ukrainian grain supply chain has emphasized the capability to reduce the total supply chain cost by 60%, through measures to improve the designing processes of the rail-road multi-modal routes and to redesign the task sequence. Another study [31] has investigated the Ukrainian grain supply chain and developed a mathematical model using graph theory and fuzzy logic to cater to transportation service providers in order to minimize the coinciding paths simultaneously, as well as the economic costs under time and risk constraints [31]. In this particular study, the authors considered multiple combinations of river, sea, and railway transportation modes in the automobile, aviation, and pipeline sectors. The Indian grain supply chain has been investigated in order to integrate a distribution system incorporating multiple echelons (farmer, procurement centers, central, state/district level warehouses, and fair price shops), transportation modes (truck, rail, and rake), and capacitated vehicles, by formulating a bi-objective mathematical model [32].

Multi-modal transportation has also been applied in perishable agri-food supply chains, although its applications are rather limited. Studies such as [6] presented the challenges (e.g., coordination, government regulations) associated with implementing multi-modal transportation for perishable products. Another study has investigated the poultry supply chain in Mississippi to optimize distribution by minimizing both direct and indirect transportation between suppliers and customers and establishing distribution hub locations [33]. In this particular study, the authors demonstrated a 30% cost-saving amount by deciding optimal hub locations and allocating farmers and customers to the established distribution hubs, as well as allocating transportation capacities to different conveyors (i.e., links), transportation mode combinations (truck–ship–rail), and deciding shortage/surplus demands using two mixed-integer linear programming models. Initially, the model established the distribution hubs in optimal locations in a way that reduced the establishment cost and the expected cost of the second model. The second model determined the direct and indirect product flow, in alignment with farmer production capacity, customer demand, and flow balance constraints. An Italian supply chain study has also focused on perishable food products, including meat, fruit and vegetables, cold cuts, and dairy products, to introduce a modular passively refrigerated food-carrying unit. This particular study found that this unit, which is compatible with rail and road transportation modes, is profitable under a minimum 20% increase in the amount of goods for the route from Bologna to Catania in Italy [34]. The Indonesian mango supply chain has been investigated to propose a logistics network configuration to optimize material flow balance, reducing postharvest losses and costs, and enhancing responsiveness [35]. Goal programming was used in this particular study to determine product flow, direction, and transportation mode combination, and the number of IOT devices needed, in a way to maximize customers and minimize post-harvest losses and transportation costs. Qualitative studies such as [36] have also been undertaken on the Italian dairy supply chain to identify enabling factors for the adoption of inter-modal transportation (rail, road) in the dairy industry, considering the needs of different actors in the supply chain. The vegetable supply chain in Spain has been investigated to promote the use of inter-modal transportation (a combination of truck and short-sea shipping), for the distribution of perishable products to enhance efficiency, sustainability, and cost-effectiveness by introducing a p-median multi-criteria model to analyze the location of coordination centers between customers and suppliers [37]. This study has resulted in cost savings (around 15%) and a reduction in externalities (around 35%), despite an increase in transit time (by 2.3 times), compared to the uni-modal (i.e., road) transportation option.

The vast majority of the studies on adopting multi-modality in the context of perishable food supply chain reconfiguration have mainly focused on establishing intermediate hubs, introducing cost-effective food-carrying units, and reconfiguring the product flow and flow direction. Out of the remaining studies that have focused on reconfiguring product flows and flow directions, one [33] has looked into the poultry supply chain in Mississippi, introducing two mixed integer linear programming models to establish distribution hubs and to decide transportation modes and shortage/surplus demands along with product flow and direction. Another study [35] on the Indonesian mango supply chain has identified the transportation modes along with the product flow and direction using a multi-objective mixed integer linear programming model. The Mississippian study developed a two-phase MILP model incorporating a hybrid solution approach based on a branch and cut algorithm to address multiple decision criteria and objectives. Similarly, in the Indonesian mango supply chain study, a multi-objective mixed-integer linear programming model was solved using goal programming to meet multiple goals. However, none of these studies applied multi-modality in the context of the Asian region and in constrained environments for three-echelon domestic vegetable supply chains. Furthermore, they did not focus on aligning the vegetable supply and demand while minimizing the total transportation cost, along with empty truck haul distance, emissions, and transportation distances. To address these gaps, this study examines the adoption of multi-modal transportation in the vegetable supply chain of an Asian country, considering transportation costs, the distance (food miles) traveled, emissions intensity, and the empty truck haul distance. Further, the suggested multi-modal configuration has been evaluated using selected economic and environmental dimensions.

3. Materials and Methods

The adaptability of bi-modal transportation was investigated for the vegetable supply chain aided by a mixed-integer linear programming formulation to derive the variables of product flow volume and the direction of flow, which resulted in the minimum total travel distance and emissions. Initially, a mixed-integer linear programming model (see Psuedo Code S1 in Supplementary Materials) was formulated to model a three-echelon supply chain (as shown in Figure 1a) without considering multi-modal transportation. Then the model was extended (see Psuedo Code S2 in Supplementary Materials) to a four-echelon supply chain (as shown in Figure 1b) considering bi-modal transportation (truck-rail). The vegetable supply chain in Sri Lanka was used as a case study (see Section 3.1) to demonstrate the utility and efficacy of bi-modal transportation. A multi-stage sampling method was used for the collection of empirical data, as outlined in the following section.

3.1. Model Formulation

This section explains the mathematical formulation of the problem. The three main assumptions used in building the mathematical problem are listed below.

• The vegetable transportation between farmer to economic centers, economic centers to the nearest outbound railway station, and from the inbound railway station to the retailer DS divisions were assumed to occur via trucks with a capacity of 10 tons, which represents the current dominant scenario.
• Since most of the wholesalers in the economic centers sell the stock they receive within the same day, the economic centers were assumed to hold no inventory.
• The nearest stations chosen for the transportation of vegetables were considered to have parking facilities and operating space available in their vicinity.

Indices

i	Production AI regions
j, j’	Economic centers
m, m’	Railway stations
k	Retailer AI regions

Sets

F	{Set of production AI regions} = {1,2,3, …, I} where I is maximum number of production AI regions
E	{Set of all economic center} = {1,2,3, …, J} where J is maximum number of economic centers
St	{Set of all stations} = {1,2,3, …, M} where M is maximum number of railway stations
R	{Set of all retailer locations} = {1,2,3, …, K} where K is maximum number of retailer DS divisions

Parameters

Cl	Per km fuel cost to transport one ton of vegetables from production AI regions to retailer locations
Ct	Per km rail cost to transport one ton of vegetables between stations
Si	Production/Supply at production AI region i
Dk	Demand at retailer location k
dPEij	Distance from the production AI region i to economic center j
dESm	Distances from the economic center m to the nearest railway station
dEEjj’	Distance from the economic center j to the economic center j’, where j’≠j
dSSmm’	Distance from the railway station m to the station m’, where m ≠ m’
dSRmk	Distances from the railway station m to retailer DS division k
dERjk	Distances from the economic center j to retailer DS division k
Dl	Maximum capacity of the lorry (10 ton)
Dt	Maximum capacity of the train (100 ton)
M ̂	A very large number
ε	A very small number

Decision Variables

xPEij	Vegetable quantities delivered from production AI region i to economic center j
xESm	Vegetable quantities delivered from an economic center tothe nearest station m
xEEjj’	Vegetable quantities delivered from economic center j to j’ where, j ≠ j’
xSEm	Vegetable quantities received to an economic center from the nearest station m
xSRmk	Vegetable quantities delivered from railway station m to retailer location k
xERjk	Vegetable quantities delivered from economic center j to retailer DS division k
xSSmm’	Vegetable quantities delivered from the railway station m to m’, where m ≠ m’
zPEij	Binary variable indicating whether vegetables are flowing from supplier i to economic center j
zESm	Binary variable indicating whether vegetables are flowing from the nearest economic center to station m
zSEm	Binary variable indicating whether vegetables are flowing from station m to the nearest economic center
zSRmk	Binary variable indicating whether vegetables are flowing from station m to retailer k
zSSmm’	Binary variable indicating whether vegetables are flowing from station m to m’, where m ≠ m’
zSRm’k	Binary variable indicating whether vegetables are flowing from station m’ to retailer DS division k

The objective function of the MILP model (given in Equation (1)) is to minimize the total transportation cost, which consists of seven cost components relating to transportation between the multiple supply chain entities.

Objective Function:

$$\begin{array}{cc}Min\left(cost\right)& =C_l({\displaystyle \sum _{i=1}^I}{\displaystyle \sum _{j=1}^J}{\displaystyle \frac{{xPE}_{ij}}{D_l}}{dPE}_{ij}+{\displaystyle \sum _{j=1}^J}{\displaystyle \sum _{j^{\prime }\ne j}^J}{\displaystyle \frac{{xEE}_{j{j}^{\prime }}}{D_l}}{dEE}_{j{j}^{\prime }}+{\displaystyle \sum _{m=1}^M}{\displaystyle \frac{{xES}_m}{D_l}}{dES}_m\\ & +{\displaystyle \sum _{m=1}^M}{\displaystyle \frac{{xSE}_m}{D_l}}{dSE}_m+{\displaystyle \sum _{m=1}^M}{\displaystyle \sum _{k=1}^K}{\displaystyle \frac{{xSR}_{mk}}{D_l}}{dSR}_{mk}+{\displaystyle \sum _{j=1}^J}{\displaystyle \sum _{k=1}^K}{\displaystyle \frac{{xER}_{jk}}{D_l}}{dER}_{jk})\\ & +C_t\left({\displaystyle \sum _{m=1}^M}{\displaystyle \sum _{m^{\prime }=1}^M}{\displaystyle \frac{{xSS}_{m{m}^{\prime }}}{D_t}}{dSS}_{m{m}^{\prime }}\right),\hfill \end{array}$$

(1)

Subject to:

The constraints were formulated according to an unbalanced transshipment problem as the total daily production exceeds the total daily consumption. Equation (2) limits the supplying quantity up to the production quantity of that corresponding production region.

$$\sum _{j=1}^J{xPE}_{ij}\le s_i \forall i \in F,$$

(2)

As the economic centers are assumed to hold no inventory, (3) equates the receiving amount to the redistributed amount at each economic center. For a particular economic center, vegetables are received from farmers, other economic centers, the nearest outbound railway station and redistributed either to the nearest inbound railway station or to other retailers.

$$\sum _{i=1}^I{xPE}_{ij}+{xSE}_m+\sum _{j^{\prime }\ne j}^J{xEE}_{j^{\prime }j}={xES}_m+\sum _{j^{\prime }\ne j}^J{xEE}_{j{j}^{\prime }}+\sum _{k=1}^K{xER}_{jk}\forall j \in E and m \in St,$$

(3)

Equation (4) limits the receiving amount to a particular railway station to the same amount redistributed to other stations and retailer locations and also to that transported back to the corresponding economic center.

$${xES}_m+\sum _{m^{\prime }}^M{xSS}_{m^{\prime }m}=\sum _{m^{\prime }=1}^M{xSS}_{m{m}^{\prime }}+\sum _{k=1}^K{xSR}_{mk} +{xSE}_m \forall m \in St,$$

(4)

Equation (5) limits the receiving amount to a retailer DS division greater than the corresponding demand.

$$\sum _m^M{xSR}_{mk}+\sum _{j=1}^J{xER}_{jk}\ge d_k \forall k \in R,$$

(5)

Equations (6)–(9) make sure the vegetable flow between production AI regions, economic centers, and retailer locations is positive.

$${xPE}_{ij}\ge 0 \forall i\in F and for all j\in E,$$

(6)

$${xES}_m\ge 0 \forall for all m\in St,$$

(7)

$${xSS}_{m{m}^{\prime }}\ge 0 \forall m, m^{\prime }\in St,$$

(8)

$${xSR}_{m^{\prime }k}\ge 0 \forall k\in R and for all m^{\prime }\in St,$$

(9)

3.2. Case Study

Considering the limited adoption of multi-modalism in domestic vegetable supply chains in Asian countries, the vegetable supply chain in Sri Lanka was used as a suitable case study. The supply chain structure with a product flow from the farmer, through economic centers to retailers was modeled, as it was the most common supply chain structure within the country. The Agricultural Inspection (AI) region (1027 regions) was set as the smallest production unit and the District Secretariat (DS) division (256 regions) was set as the smallest consumption unit. Although the supply chain is quite extensive due to the large number of AI regions and DS divisions captured, the supply (production) and demand (consumption) were found to be not evenly distributed across the supply chain. Due to the uneven distribution of production and consumption across the supply chain, all the production AI regions and consumption DS divisions were not equally contributing to the functioning and performance of the supply chain. Therefore, multi-stage sampling was used, accompanied by the Pareto principle, in each stage. In the first stage, the districts that account for the top eighty percent of the production or consumption were included in the broader sample. Then, in the second stage, the producing AI regions and consuming DS divisions contributing to the top eighty percent of the district production and consumption were drawn from the broader sample to be included in the final sample. Accordingly, 187 production AI regions and 157 consumption DS divisions were included in the sample. However, all economic centers (i.e., 13) were considered as processing units. The monthly production quantity for each AI region was extracted from the records maintained by the Department of Agriculture, Sri Lanka to estimate the daily production quantity. The consumption amounts were estimated by multiplying the daily per capita consumption rates of the rural (165.89 g per day), estate (108.38 g per day), and urban (180.55 g per day) regions by the corresponding population in the rural, estate, and urban regions of each DS division, respectively.

The distances (i) from the production AI regions to the economic centers; (ii) from the economic centers to the nearest railway station; and (iii) from the railway station to the retailer DS divisions were calculated using the “Haversine” formula (Equations (10)–(12)).

$$\mathrm{a}={sin}^2\left({\displaystyle \frac{\mathsf{\Delta }\mathrm{l}\mathrm{a}\mathrm{t}}{2}}\right)+\mathrm{c}\mathrm{o}\mathrm{s}\left(\mathrm{l}\mathrm{a}\mathrm{t}1\right)\times cos\left(\mathrm{l}\mathrm{a}\mathrm{t}2\right)\times {sin}^2\left({\displaystyle \frac{\mathsf{\Delta }\mathrm{l}\mathrm{o}\mathrm{n}}{2}}\right)$$

(10)

$$\mathrm{c}=2\times atan2(\surd \mathrm{a},\surd (1-\mathrm{a}\left)\right)$$

(11)

$$\mathrm{D}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{e}=\mathrm{R}\times \mathrm{c}$$

(12)

Here, the latitudes and longitudes were taken in radians, and R, the earth radius (6371 km), was taken in kilometers. Currently, transportation occurs from the farmers to economic centers, between economic centers, and from the economic centers to retailers via trucks with a load-carrying capacity of 10 tons. It was noted that all the economic centers were located within a 15 km radius of the nearest railway station. Therefore, the downstream segment of the supply chain can be reconfigured by adopting rail transportation as indicated in Figure 1b. Accordingly, it is feasible to dispatch the vegetables collected at the economic centers to the downstream retailers via another economic center supported by rail transportation or existing truck routes, considering the unit transportation cost. Once the MILP problem is extended to accommodate the structural changes, as above, the volume and direction of the flow between each pair of supply chain entities were determined to minimize the transportation costs, food miles, GHG emissions, and empty truck haul distances at the supply chain level.

4. Results

As per the existing supply chain configuration, the delivery of vegetables between the economic centers is carried out using trucks only. As such, the MILP model was solved for the existing supply chain configuration to determine the food miles, transportation cost, CO₂eq emissions, and empty truck haul distance based on the current structure. When adopting bi-modality, rail transportation from Nanuoya to Colombo Fort (zSS2,6), Bandarawela to Beliatta (zSS3,4), Kandana (zSS3,5) and Narahenpita (zSS3,10), and Thambuththegama to Ratmalana (zSS11,7) railway stations is proposed, based on the results obtained by solving the MILP model, in order to minimize the total transportation cost. From the total vegetable volume, 19.2% utilized the bi-modal transportation option. As reported in

Table 1. Comparison of performance measures.

Performance Measure		Existing Configuration	Bi-Modal Configuration	Saving
Food Miles (tkm)	Farmer to Economic Center	5405	2716	32%
	Between Economic Center	172	-
	Economic Center to Nearest Station	-	673
	Economic Center to Retailer	9739	1687
	Nearest station to Economic Center	-	673
	Between Stations	-	1053
	Stations to Retailer	-	4328
	Total	15,316	10,477
Transportation Cost		1,391,569	888,340	36%
CO2eq Emission		1299	841	35%
Empty Truck Return Distance		15,316	9424	38.5%

, with the existing uni-modal transportation system, the total food miles traveled to fulfill the daily vegetable demand is 15,316 tkm. Food miles are calculated in ton-kilometers by multiplying the distance (km) travelled by the volume of vegetables transported (ton). Adoption of the bi-modal structure has reduced the food-miles down to 10,477 tkm. Thus, the adoption of multi-modality has the potential to reduce the total food miles by 31.6% with the use of railway transportation for transporting vegetables between the economic centers only. As rail transportation is capable of transporting larger loads for long distances, the transportation frequency reduces and, as a result, the total number of food miles traveled reduces, as well.

The transportation cost is calculated by multiplying the per-kilometer cost by the distance traveled and the number of trips. In terms of costs, the existing supply chain configuration with uni-modal transportation resulted in a total cost of LKA 1,391,569 to deliver the daily vegetable requirement. By comparison, the proposed supply chain configuration with bi-model transportation (truck and railway combination) has the potential to reduce the total supply chain transportation cost down to LKA 888,340, indicating a 36.2% reduction. In the proposed configuration with bi-modal transportation, 19.2% of the vegetable volume flows through trains, which is economically beneficial compared to using truck transportation. Here, for truck transportation, the cost of both forward and return (empty haul) trips were taken into account; however, for rail transportation, only the forward transportation was taken into account. The emission level was calculated by multiplying the emission rate for each transportation mode by the corresponding ton–kilometer value. In terms of environmental impact, the existing uni-modal transportation configuration resulted in a total of 1298.8 t CO₂eq emissions and the proposed bi-modal transportation configuration resulted in a total of 841.45 t CO₂eq emissions indicating the potential for reducing CO₂ emissions by 35.2% when adopting bi-modal transportation.

Other than comparing the economic and environmental performance, facility (capacity) utilization was also assessed for the existing uni-modal and proposed bi-modal transportation configurations. The empty truck haul in the uni-modal (truck) structure is taken as half of the total distance traveled, while omitting the return trip distance in the bi-modal structure. According to the results, the total amount of empty truck hauls in the existing structure is 15,316 km, which can be reduced down to 9423.6 km by 38.5% with the adoption of rail transportation between economic centers or the nearest railway stations. This computation included the transportation between farmers to economic centers, between economic centers, and between railway stations, as well as from economic centers/railway stations to the retailer DS divisions.

5. Discussion

By adopting rail transportation (instead of truck transportation) between Nanuoya to Colombo Fort (zSS2,6), Bandarawela to Beliatta (zSS3,4), Kandana (zSS3,5) and Narahenpita (zSS3,10), and Thambuththegama to Ratmalana (zSS11,7) railway stations, the existing uni-modal transportation structure was converted into a bi-modal transportation structure. According to the results obtained by solving the MILP model, for both the existing uni-modal and the proposed bi-modal configurations, the supply chain performance was assessed using the transportation cost, food miles, CO₂eq emissions, and the empty truck hauls involved. With the adoption of multi-modality, significant savings were delivered against all four performance metrics, i.e., food miles, transportation costs, CO₂eq emissions, and empty truck hauls.

Unnecessary empty truck hauls were quite significant within the existing uni-modal vegetable supply chain in Sri Lanka. This has direct implications for price, as the transportation cost accounts for loaded vegetable transportation, as well as empty truck hauls. Railway transportation is flexible in many ways as wagons can be added and removed to cater to the required space. Furthermore, the same train could be used for transporting both passengers and vegetables, thereby further leveraging scale economies. There is also the added flexibility to allow the train capacity to be used for passenger transportation during peak hours. As most of the bulk transportation is happening between economic centers, replacing truck transportation with rail transportation is appropriate in terms of reducing empty truck return frequency.

Given that consumers are increasingly aware of and concerned about the freshness of vegetables they consume and the environmental impact of transportation, reducing food miles could positively contribute to addressing both of these aspects while achieving cost savings. Because of the higher load-carrying capacity of trains, the use of rail transportation reduces the frequency of rides, thereby reducing the total amount of distance traveled and the amount of food miles traveled. By solving the MILP model, the total quantity of vegetables moved between economic centers or railway stations remains the same while the total distance traveled is reduced. Overall, with the adoption of bi-modality in the vegetable supply chain, the amount of food miles could be reduced by 31.6%.

As the per-ton–kilometer emission rate of a truck (0.0848 CO₂eq/tkm) is higher than the rate of a train (0.0402 CO₂eq/tkm), by replacing vegetable transportation between economic centers by trains, the amount of CO₂eq emissions can be substantially reduced. Additionally, a reduction in distance traveled directly contributes to a reduction in CO₂eq emissions. Overall, the adoption of multi-modality could potentially reduce CO₂eq emissions by 35.2%. The limited studies undertaken into the perishable food context have not evaluated the applicability of bi-modalism in terms of emissions. Furthermore, the adoption of multi-modal transportation eliminates unnecessary empty truck hauls and food miles traveled, further contributing to reducing the total transportation cost by 33.8%. Multi-modality has not been assessed in terms of empty truck hauls so far. The application of multi-modal transportation in the perishable food supply chain context is frequently assessed in terms of cost. A study undertaken on the vegetable supply chain in Spain has reported cost savings of 15% [37], and the poultry supply chain in Mississippi has reported cost savings of 30% by implementing multi-modality [33].

6. Conclusions

This study was conducted to test the applicability of bi-modal transportation in a constrained environment. As multi-modality has seen limited applications in the context of agri-food supply chains, Sri Lanka’s mainstream vegetable supply chain was chosen as a suitable case study. A combination of truck and rail transportation was considered against the existing uni-modal vegetable supply chain that uses trucks only. The vegetable transportation between economic centers was replaced by railway station to railway station vegetable transportation using trains. An MILP problem was formulated to determine the optimal flow quantity and the direction of transportation so that the total transportation cost is minimized.

The results obtained by solving the MLP model suggested that rail transportation from Nanuoya to Colombo Fort (zSS2,6) Bandarawela to Beliatta (zSS3,4), Kandana (zSS3,5) and Narahenpita (zSS3,10), and Thambuththegama to Ratmalana (zSS11,7) railway stations could potentially reduce the total food miles (32%), CO₂eq emissions (35%) and empty truck hauls (38.5%). The reduction in food miles and empty truck hauls reinforced the reduction in total transportation costs (36%). The reduction in food miles could indirectly contribute to reducing post-harvest losses while also addressing the customers’ concerns regarding environmental impacts. The reduction in transportation costs is also significant, particularly in light of the affordability burden imposed by increasing fuel prices in the context of Sri Lanka. These findings highlight the potential of multi-modal transportation to enhance the sustainability of vegetable supply chains.

Multi-modality is applicable even for constrained environments in achieving sustainable development goals. Specifically, for agri-food systems, multi-modality is applicable under good handling practices during transit. Under Sri Lankan conditions, even with its limited railway network, multi-modal transportation is highly applicable to replace the transportation between economic centers, as bulk deliveries occur in that segment of the supply chain.

The findings of this study can be useful in terms of their contribution to practices; for example, informing decision-makers of the merits of establishing a vegetable transportation railway schedule in the country, which aligns with the economic center operating in time windows. Prior to the establishment of this railway schedule, it can be tested on a simulation platform to see the impact in terms of multiple sustainability dimensions. The study could be extended to account for the post-harvest losses corresponding to each transportation mode and between handling operations subject to the availability of data.

This study has considered all vegetable types as one commodity, but practically, different vegetable types need different conditions during transportation. Therefore, in future studies, the model can be amended to account for the different categories of vegetables. Additionally, the mathematical model used in this study has assumed all the stations have adequate storage facilities and that employees are capable of handling the vegetable stocks received. However, in real-world situations, the stations might need infrastructure upgrades for handling the vegetable stock. Moreover, it has been assumed that in each and every economic center, the amounts received will be sent to retailers on the same day. Sometimes, the oversupply is kept in cold storage and sold for a lower price the next day. These variations can be explored in future studies by adding constraints on the shelf-life time windows of the vegetables to the model. Moreover, further investigations can be undertaken regarding integrating railway schedules to determine the lead times and their impact on retaining food quality and shelf life. Additional supply chain design decisions such as facility location and capacity addition, subject to various uncertainties, can also be accounted for to investigate the benefits of multi-modal transportation.