Strategic Infrastructure Sequencing for Freight Transport Decarbonization Under Declining Demand Using Data from Latvia

Abstract

This study addresses a critical policy paradox in transport infrastructure planning: the necessity for substantial decarbonization investments amid declining freight demand forecasts in less developed territories. Despite reduced demand, such investments remain justified for advancing sustainability, regulatory compliance, and long-term system resilience. Herein, an integrated decision support framework is developed that optimizes infrastructure investment sequencing while maximizing private capital participation and ensuring technology–regulation alignment. Using comprehensive freight transport data from Latvia (2012–2023), a scenario tree analysis integrated with S-curve technology adoption models is employed to evaluate optimal infrastructure sequencing strategies for hydrogen fuel cell vehicles (HFCVs) and battery electric vehicles (BEVs). The methodology combines Autoregressive Integrated Moving Average (ARIMA) demand forecasting with total cost of ownership (TCO)-based technology adoption curves and hierarchical modal split modeling. The analysis further identifies distinct market segments and adoption trajectories, demonstrating how strategic infrastructure sequencing can accelerate low- and zero-emission technology uptake across different freight distances and policy scenarios. The results demonstrate that strategic sequencing generates net present value (NPV) savings of approximately EUR 18.2 million (at a 4% discount rate) compared to immediate full-scale deployment while maintaining regulatory compliance timelines. The framework provides policymakers with systematic evidence-based criteria for infrastructure investment timing, contributing to the efficient allocation of scarce public resources in the transition to sustainable freight transport.

1. Introduction

Despite the urgent need to decarbonize the transport sector, infrastructure investment decisions in less developed territories face a policy paradox: declining freight demand forecasts undermine the economic justification for large-scale low-carbon infrastructure projects. This creates uncertainty in determining the optimal timing, scale, and sequencing of decarbonization investments, particularly under evolving technological and regulatory conditions.

The Europe cold chain market is expected to record a Compound Annual Growth Rate (CAGR) of 14.1% from 2024 to 2033, with the market size projected to reach a valuation of USD 105.5 billion in 2024. By 2033, the valuation is anticipated to reach USD 356.7 billion (CMI, 2023).

Table 1. Key factors predetermining significant growth in the European cold chain market.

Factor Category	Key Growth Factor	Impact on Market Growth	Supporting Data
Consumer demand and market projections	Increasing demand for perishable goods (e.g., fresh produce, dairy, frozen products). Trends show that a country tends to have more cold warehouse resources if its GDP per capita is higher.	Despite a projected decline in overall meat consumption—due to sustainability concerns and high prices—poultry consumption is expected to rise. Additionally, growth in cheese and whey production underscores the need for temperature-controlled logistics to preserve product integrity.	[1,2]
Industry expansion	The EU is witnessing a rise in the adoption of biologics and advanced therapeutic medicinal products (ATMPs), such as cell and gene therapies, which require precise temperature control to maintain their efficacy.	Europe is the second-largest biopharmaceuticals market in the world. About 80% of pharma products now require temperature-controlled transportation. The size of the Pharmaceutical Cold Chain Logistics Market is estimated to be USD 21.55 billion in 2025 and is expected to reach USD 34.70 billion by 2030, at a CAGR of 10% during the forecast period (2025–2030).	[3]
E-commerce growth	The EU’s perishable e-commerce sector is experiencing rapid growth, driven by consumer demand for convenience, sustainability, and health-focused products.	In 2024, 77% of EU internet users bought or ordered goods or services for personal use online in the previous 12 months, increasing from 59% in 2014. Food and cosmetics are, respectively, the second and the third most popular products, with a common share of 41%. The European online grocery market is projected to reach USD 440.3 billion by 2033, growing at a CAGR of 23.4% from 2025 to 2033.	[4,5]
Regulatory standards	The EU has implemented stringent guidelines for Good Distribution Practices and Hazard Analysis and Critical Control Points that extend temperature control requirements to the entire distribution process, including transportation.	Both guidelines require rigorous temperature monitoring during transportation to ensure product integrity. Cold logistic chain companies must adhere to these guidelines to comply with EU regulations, ensuring that products are safe for consumption or use.	[6,7]
Technological advancements	The need for precise temperature control has driven investment in advanced logistics technologies, such as Internet of Things (IoT) sensors and blockchain-based traceability solutions, which are real-time monitoring systems.	IoT systems can alert logistics staff to equipment failures, thus preventing potential food wastage and improving energy efficiency. Phase change materials for cold storage aim to minimize thermal loads and improve the efficiency of refrigerated transport systems, which are traditionally reliant on diesel engines.	[8,9,10,11,12,13]
Sustainability focus	With a growing emphasis on sustainability and environmental responsibility, opportunities exist for companies to integrate sustainable practices such as green logistics practices, enabling the avoidance of food waste.	The global cold chain is estimated to account for up to 3.5% of the world’s carbon footprint, with a significant portion attributed to food refrigeration. On the other hand, inappropriate cold chain logistics causes the lost food to lead to the waste of invested natural resources used to produce food and increases greenhouse gas (GHG) emissions, exacerbating environmental degradation. The World Health Organization estimated that approx. 600 million people (almost 1 in 10 globally) fall ill after consuming contaminated food every year, and about 420,000 people die annually due to foodborne diseases. Refrigerated vehicles predominantly rely on hydrofluorocarbons (HFCs) as refrigerants which, due to leakage during operation and challenges associated with end-of-life disposal, contribute significantly to environmental degradation.	[14,15,16,17]
Infrastructure investments	Transitioning to net-zero cold chain transportation necessitates a multi-faceted approach, integrating renewable energy, electric and hydrogen-powered vehicles, efficient infrastructure, and supportive policies.	High initial investment costs for sustainable infrastructure, such as electric vehicle charging stations and hydrogen fueling facilities, pose significant challenges. Developing supportive regulatory frameworks is essential for driving investment and innovation in net-zero cold chain logistics. There are notable disparities in the investment requirements for transportation infrastructure across regions, particularly when comparing traditional and low-carbon systems, with developed countries exhibiting relatively lower needs for low-carbon infrastructure due to their established frameworks.	[18,19,20]

summarizes the key factors that will predetermine Europe’s significant cold chain market growth.

Cold supply chains contribute significantly to greenhouse gas emissions, accounting for up to 3.5% of the world’s carbon footprint, with refrigerated transport being a major contributor to global fossil fuel use. The European Union’s Green Deal mandates substantial reductions in transport emissions by 2030, necessitating significant infrastructure investments in alternative fuel systems, electrified rail networks, and multimodal logistics hubs. The European Union’s regulatory landscape creates clear phase-out pressure for traditional fuel usage in cold chain logistics through progressively tightening CO₂ emission standards. Regulation (EU) 2019/631 established initial fleet-average targets of 95 g CO₂/km, subsequently strengthened by Regulation (EU) 2023/851 to 67.5 g CO₂/km by 2030 for light commercial vehicles. The Sustainable and Smart Mobility Strategy (COM/2020/789) mandates a 90% reduction in transport-related emissions by 2050, with intermediate targets requiring at least a 30% reduction in traditional fuel use by 2030, creating substantial compliance pressure for cold chain operators.

Alternative technology adoption is supported by comprehensive infrastructure mandates and strategic frameworks. Electric vehicle deployment benefits from Regulation (EU) 2023/1804, requiring charging points every 60 km on major roads by 2025. Hydrogen vehicles receive support through the EU Hydrogen Strategy (COM/2020/301) aiming for the generation of 10 million tons of renewable hydrogen production by 2030, complemented by Directive (EU) 2024/1788 mandating hydrogen refueling stations every 200 km on TEN-T roads by 2030. Rail freight receives strategic emphasis through goals to double rail freight traffic by 2030 and triple it by 2050, supported by Latvia’s Transport Development Guidelines aiming for 20% of cold chain freight to shift to rail by 2030.

However, traffic volume growth is not seen in all EU states; for instance, the freight demand forecasts for Latvia presented in Section 4 indicate declining volumes across multiple European markets. This creates a strategic dilemma for policymakers, particularly in the Baltic states.

This research addresses the critical question of how to sequence infrastructure investments optimally when traditional growth-based planning assumptions no longer hold. Unlike previous studies that assume demand growth or ignore technology transition constraints, this study examines the intersection of declining demand, mandatory decarbonization requirements, and limited public budgets in the context of cold chain logistics. The framework explicitly models the co-evolution of technology adoption and infrastructure deployment, accounting for regulatory timelines and private sector investment behavior.

The study contributes to three distinct literature streams:

−

Infrastructure investment optimization under uncertainty.

−

Technology diffusion modeling in transport systems.

−

Public–private partnership design in declining markets.

A novel decision support framework is developed that integrates scenario-based planning with real option valuation, providing policymakers with tools to navigate complex trade-offs between public investment efficiency and decarbonization objectives.

Research Questions:

−

How do technology readiness, compatibility, and the complexity of electric vehicles, hydrogen fuel systems, and rail transport impact their feasibility for adoption in cold chain logistics within territories with declining transportation flow patterns?

−

How do market conditions, regulatory frameworks, and infrastructure limitations affect adoption timing and sequencing decisions?

Hypotheses:

H1: 

Technology adoption in cold chain logistics follows predictable S-curve patterns that can inform evidence-based infrastructure investment timing.

H2: 

Strategic infrastructure sequencing based on adoption thresholds can significantly reduce public investment requirements while maintaining decarbonization objectives.

The study develops a novel multi-factor S-curve model that integrates market signals (reflected in TCO competitiveness), fleet renewal cycles, and regulatory pressures to predict technology adoption patterns. Unlike traditional approaches that rely on single-factor analysis, this framework provides superior forecasting accuracy with adoption rate parameters calibrated to empirical anchor points and validated through industry surveys.

Unlike growth-oriented planning approaches, this research explicitly models infrastructure development under declining freight volumes. The framework demonstrates how “right-sizing” infrastructure for efficiency rather than capacity expansion can enable the achievement of decarbonization objectives while reducing public investment requirements by 70% compared to parallel technology deployment.

The remainder of this paper is structured as follows: in Section 2, we review the relevant literature. The data and methodology are described in Section 3, while Section 4 presents the results. In Section 5 and Section 6, we discuss the findings and conclusions. The back matter provides an outline of the data availability. A total of four supplementary annexes are accessible via the OSF platform [https://doi.org/10.17605/OSF.IO/BCDAK accessed on 6 November 2025].

To ensure clarity and consistency, key technical terms, such as “cold supply chain,” and abbreviations are defined in Annex A [https://osf.io/mpwa8] (acessed on 14 November 2025), providing a foundational understanding for readers unfamiliar with the terminology used in this study.

2. Literature Review

The transition to low- and zero-emission technologies in heavy-duty cold chain transport presents complex challenges, with no single technology demonstrating a clear path to market dominance. Currently, diesel remains the dominant fuel choice due to its associated established infrastructure networks, lower upfront acquisition costs, and operational familiarity among fleet operators [21]. However, this dominance is expected to decline progressively as environmental policies become stricter and alternative technologies are established [22,23].

The technological landscape reveals distinct advantages and limitations for each alternative. Battery electric vehicles (BEVs) demonstrate the lowest lifecycle emissions and operational costs for specific applications in long-haul heavy-duty transport [24], making them particularly attractive for short- to medium-distance cold chain operations. However, BEVs face significant constraints including battery weight penalties, limited range under refrigeration loads, and extended charging times that disrupt logistics schedules. Conversely, hydrogen fuel cell electric vehicles (HFCVs) offer superior range capabilities and rapid refueling times, making them better suited for long-haul routes and remote operational areas where charging infrastructure is sparse [21,23]. Nevertheless, HFCVs are constrained by high hydrogen production costs, limited supply chain infrastructure, and substantial upfront vehicle costs.

This technological complexity has led researchers to predict a “messy middle” transition period, characterized by multiple technologies coexisting rather than a single dominant solution emerging [23,25,26]. Adoption patterns are expected to vary, driven by total cost of ownership (TCO) calculations that incorporate acquisition costs, operational expenses, infrastructure availability, payload capacity impacts, and regulatory incentives [21]. Additionally, Larson et al. (2024) [23] mention route characteristics, operational patterns, and specific fleet requirements as additional drivers rather than one-size-fits-all solutions.

This heterogeneous adoption pattern creates significant infrastructure investment challenges, particularly in resource-constrained environments. Traditional approaches that aim to simultaneously develop comprehensive infrastructure for all potential technologies risk substantial capital misallocation, with some networks becoming underutilized while others face capacity constraints. This mismatch problem is acute in less developed territories, where public investment resources are limited and opportunity costs are high. Consequently, evidence-based infrastructure sequencing frameworks are needed to optimize investment timing, minimize stranded assets, and ensure adequate capacity for emerging technology leaders [25,26].

Developing such frameworks is complicated by a lack of empirical or simulation-based studies explicitly modeling winner-take-all or lock-in effects in heavy-duty transport technology adoption. While multi-criteria decision analysis is used to compare technologies, integration with dynamic market competition models remains rare [27,28,29]. Further research is needed on how policy interventions, infrastructure rollouts, and early adopter advantages could drive market convergence toward a dominant technology [23,30]. The concept of “right-sizing” infrastructure for efficiency rather than growth represents an emerging paradigm shift requiring new analytical frameworks.

The S-curve model of technology diffusion has been extensively applied to transport innovation [31,32,33]. Pamidimukkala et al. (2023) [34] and Raoofi et al. (2025) [30] identified technological, financial, and infrastructure barriers as significant, with policy interventions (e.g., subsidies and charging infrastructure) shown to accelerate adoption. The evidence emphasizes the need for coordinated planning of adapted technology within infrastructure, avoiding grid strain and optimizing charging patterns [30,35,36], addressing the “chicken-and-egg” problem of vehicle and fueling station coevolution, and informing policy for early-stage market support [37,38].

While evidence highlights the trade-offs between battery electric, hydrogen fuel cell, and rail-based solutions, little is known about how their technological readiness and complexity interact with market conditions, regulations, and infrastructure constraints in regions with declining transport demand. The reviewed literature on technology diffusion, particularly S-curve adoption models in the transport sector [31,32,33], suggests that technology transitions follow predictable phases of emergence, acceleration, and maturity—each with distinct infrastructure and investment implications. These insights directly inform H1, proposing that technology adoption in cold chain logistics follows S-curve trajectories that can guide evidence-based investment timing. Furthermore, studies emphasizing the role of policy sequencing, infrastructure readiness, and early adopter dynamics [30,35,36] indicate that the order and pace of infrastructure deployment can critically influence both adoption costs and system-level efficiency. This evidence grounds H2, which posits that strategic infrastructure sequencing based on adoption thresholds can reduce public investment requirements while maintaining decarbonization objectives. Accordingly, this study addresses a key gap by integrating technology diffusion theory with infrastructure optimization to model feasible, cost-effective transition pathways for cold chain logistics in developing regions.

3. Data and Methods

The research design combines the following strategies:

−

ARIMA demand forecasting and classification to short-haul and long-haul trips using historical freight data.

−

S-curve technology adoption model based on total cost of ownership dynamics and results gained through semi-structured interviews with an expert panel.

−

Policy trigger identification based on adoption thresholds and market signals.

−

Calculation of NPV savings.

3.1. Data Collection

The primary data source used in the study is Latvian Central Statistical Bureau (CSB) weighted survey data, consisting of data on 120 vehicles/week regarding the full fleet movement distribution, including weight, stops, origin–destination, distance, and cargo package characteristics. The study explores historical freight transport data from Latvia (2012–2023) for the food, beverage, and tobacco sectors. Additionally, data collected from the Road Traffic Safety Directorate (CSDD) [39] on fleet composition and age distribution were used in the model.

Temporal coverage: The period spanning 2025–2035, with 2025–2030 serving as the calibration window and 2031–2035 as the projection horizon.

Technologies considered: Internal Combustion Engines (ICEs) (baseline incumbent), BEVs, and HFCVs (emerging alternatives).

Scenarios reflect varying levels of policy support and market uptake potential developed based on negotiations with responsible ministries:

−

Base—only mandatory EU-level implications and feasibility of technologies.

−

Optimistic—introduced subsidies of EUR 50,000 for BEVs and HFCVs.

−

Maximum—introduced subsidies of EUR 100,000 for BEVs and HFCVs; increased excise taxes for ICE fuel.

The study integrates outputs from previous research [40], where three groups of factors were organized using the Analytic Hierarchy Process (AHP) framework, generating adoption preferences for three technologies considered based on two points: the existing situation in 2025 and prospects for 2030:

−

Technological factors (value), such as energy use, single-tank/single-charge range, depot/roadside energy fill time, and infrastructure availability.

−

Feasibility factors (total costs of ownership or money), such as truck initial acquisition costs, truck maintenance costs, fuel costs, operational lifespan, and residual value retention.

−

Availability factors such as fuel production, truck production, and charging or fueling infrastructure.

AHP results are presented in Annex B, which is available through the Open Science Framework (OSF) repository [https://osf.io/4m9jr] (acessed on 14 November 2025).

Statistical analysis was performed using R programming language with packages such as readxl, dplyr, forecast, purrr, ggplot2, patchwork, gridExtra, and knitr, with Claude AI (Anthropic) providing technical coding assistance based on author-specified instructions.

During the preparation of this manuscript/study, the authors used also OpenAI’s GPT-5.1 and Anthropic’s Claude AI to assist with improving the readability and clarity of the text. The authors have reviewed and edited all AI-generated or AI-assisted content and take full responsibility for the final version of this publication.

3.2. Demand Forecasting and Classification of Trips

The forecasting methodology employs raw ton-kilometer data from the CSB freight transport survey. Data preprocessing includes systematic product category filtering, unit conversion to million ton-kilometers, and temporal aggregation into hierarchical structures (annual, quarterly, and monthly). Quality assurance procedures address missing observations, zero-value removal, and additional quality procedures to ensure analytical integrity. The code used to address data imperfections (data cleaning, imputation, and deletion) is available directly from the OSF repository [https://osf.io/4m9jr].

Monthly freight volumes are constructed as time series objects with 12-month frequency, establishing the foundation for decomposition analysis:

Y(t) = T(t) + S(t) + R(t),

(1)

where

Y(t)—observed freight demand;

T(t)—the trend component;

S(t)—the seasonal component;

R(t)—the irregular residual.

The seasonal and trend decomposition using locally estimated scatterplot smoothing (LOESS) methodology, specifically the STL (seasonal-trend decomposition using LOESS) approach, separates the time series into constituent components using periodic seasonal windows, assuming consistent within-year patterns across the observation period. The STL approach allows the extraction of trend components while preserving seasonal variations, making it particularly suitable for cold chain freight data. Decomposition outputs inform ARIMA optimal parameter combinations, ensuring forecast reliability and statistical adequacy.

The exact code for forecasting is available through the OSF repository [https://osf.io/9q7hb] (acessed on 14 November 2025).

3.3. Model Formulation

To validate the technology adoption forecasting results and policy recommendations, semi-structured interviews with three cold chain logistics companies representing different operational scales in the Latvian and international freight transport market were conducted.

A purposive sampling approach was used to select three companies representing the full spectrum of cold chain logistics operations:

−

Regional cold chain distributor focusing on local and short-haul deliveries—fleet size < 10 vehicles.

−

Multi-regional logistics provider with established cold chain networks—fleet size 10–50 vehicles.

−

The largest cold chain logistics company in Latvia with national coverage and international operations—fleet size > 50 vehicles.

The semi-structured interview protocol was designed around three core dimensions of the TOE framework used in our previous research:

−

Technological Context: Assessment of technology integration capabilities, operational complexity, efficiency perceptions, and technological barriers for battery electric vehicles (BEVs), hydrogen fuel cell vehicles (HFCVs), alternative low-carbon fuels, and rail transport integration.

−

Organizational Context: Evaluation of internal factors including investment decision criteria, sustainability goal alignment, resource availability (financial, human, and technical), and leadership commitment to sustainable transport adoption.

−

Environmental Context: Analysis of external factors including government policy impacts, regulatory pressures, stakeholder influence, infrastructure constraints, and market demand signals affecting technology adoption decisions.

Each interview lasted approximately 30–40 min and combined structured Likert-scale assessments (1 = Very Low to 5 = Very High) with open-ended qualitative questions to capture nuanced insights into adoption decision-making processes. The interviews were conducted in Latvian.

Quantitative responses were analyzed using descriptive statistics to identify adoption readiness patterns across company sizes and technology types. Qualitative responses underwent thematic analysis to extract key enablers and barriers of the model, which were then mapped against the following:

−

Technology adoption timing predictions against current company planning horizons.

−

Infrastructure readiness requirements versus actual operational constraints.

−

Total cost of ownership sensitivity to factors identified by industry participants.

−

Policy intervention effectiveness relative to stated company decision-making criteria.

Due to commercial sensitivity and confidentiality agreements with participating companies, individual company responses cannot be published in their raw form. The results are presented only in aggregated and anonymized format, with specific company identifiers removed and responses summarized.

Detailed interview protocols, aggregated response summaries, and thematic analysis frameworks are available in Annex C through the project’s OSF [https://osf.io/6hmtg] (acessed on 14 November 2025) repository to support research transparency while maintaining participant confidentiality.

3.4. Market Segmentation and Model Implementation in R

To support distance-based technology allocation strategies, the methodology incorporates a spatial analysis of freight transport flows. Origin–destination (O-D) matrices and route categorization of distances above and below 250 km inform infrastructure deployment strategies by identifying the following:

−

Short-haul trips with high flow volumes designated for BEV charging infrastructure deployment.

−

Long-haul trips requiring HFCV refueling stations or rail development.

The exact classification code is available through the OSF repository [https://osf.io/9q7hb].

To model technology adoption and decline trajectories, the logistic S-curve function is employed, which is widely used in technology forecasting as it captures the nonlinear dynamics of diffusion, characterized by a slow initial phase, rapid growth/decline, and eventual saturation [41].

In this study, the logistic S-curve parameters are obtained through a hybrid approach combining empirical calibration and AHP scenario-based assumptions. The saturation level L_i is defined exogenously based on AHP scenario narratives, while the adoption rate k_i and inflection point t_0i are endogenously estimated by fitting the logistic function to anchor points (2025 and 2030 market shares) derived from the analytical hierarchical process results discussed above.

Based on interview responses, critical adoption at the reference year 2027.5 is used to rank technologies, with the leader receiving a base multiplier of 2.0, the runner-up 1.0, and others 0.5. This step captures winner-takes-all dynamics, ensuring the market leader is favored in subsequent diffusion modeling. The analysis is extended across vehicle categories, where additional multipliers reflect technology suitability: BEVs in N1 and N2 receive 1.5 and 1.0, respectively, while HFCVs in N3 receive 1.5.

The final momentum adjustment is the product of base and preference multipliers.

Two logistic functions were implemented:

Firstly, growth S-curve for each emerging technology (BEVs and HFCVs):

$$A_{i\left(t\right)}={\displaystyle \frac{L_i}{1+e^{-k_i\left(t-t_{\left\{0i\right\}}\right)}}},$$

(2)

where

$A_{i\left(t\right)}$—adoption rate for technology i at time t;

$L_i$—market saturation level for technology i;

$k_i$—adoption rate parameter;

$t_{\left\{0i\right\}}$—inflection point timing.

Given two anchor points $\left(t_1,A_1\right)$ and $\left(t_2,A_2\right)$ and using analytical hierarchical process-defined saturation L_i for different scenarios, the formula is solved for $k_i$ and $t_{\left\{0i\right\}}$ with Logit transformation.

$${\displaystyle \frac{L_i}{A_i\left(t\right)}}- 1 = exp (-k (t-t_0)) \Rightarrow {\displaystyle ln (\frac{L_i}{A_i\left(t\right) }}- 1)=-k_i \left(t-t_{0i}\right)).$$

Evaluation is performed at the two points and subtracted:

$$u_{\mathit{1}}=ln ({\displaystyle \frac{L_i}{A_1}} - 1), u_2 = ln ({\displaystyle \frac{L_i}{A_2}} - 1),$$

$$u_1-u_2 = -k_i (t_1 - t_2) \Rightarrow k_i={\displaystyle \frac{u_2-u_1}{t_2-t_1}}.$$

The inflection time is recovered from either point:

$$t_{0i} = t_1 + {\displaystyle \frac{1}{k_i}} u_1 = t_2 + {\displaystyle \frac{1}{k_i}} u_2.$$

Thus, with known L_i, the closed-form solutions are

$$k_i={\displaystyle \frac{\mathit{ln}(\frac{L_i}{A_2}\hspace{0.17em}-1)-\mathit{ln}(\frac{L_i}{A_1}\hspace{0.17em}-1)}{t_2-t_1}},t_{0i} = t_1 + {\displaystyle \frac{1}{k_i}} \mathit{ln}({\displaystyle \frac{L_i}{A_1}}\hspace{0.17em}-1)$$

Secondly, the decline S-curve for incumbent technologies is calculated as a residual to ensure 100% fleet composition constraint. This formulation allows the representation of both technology uptake (e.g., BEVs and HFCVs) and phase-out (e.g., ICEs) under a unified mathematical framework, while guaranteeing fleet composition consistency at all time points.

Total demand is adjusted by a demand factor, reflecting contrasting fleet requirements over time calculated with ARIMA forecast, e.g., applied separately to short-haul and long-haul segments, with technology adoption percentages multiplied by segment-specific forecasted volumes to generate absolute infrastructure demand projections.

The resulting fleet composition yields annual market shares for ICE, BEV, and HFCV technologies across scenarios and vehicle categories (N1 ≤ 3.5 t, N2 3.5–12 t, N3 > 12 t), with absolute ton-kilometer volumes calculated for infrastructure planning purposes.

The exact code in R can be obtained through the OSF repository [https://osf.io/9q7hb].

3.5. Assumptions and Limitations

The following assumptions and limitations were made in this study:

(1)

Only one statistical group—food, beverage, and tobacco—is tested in the cold chain.

(2)

The exogen parameters used to simulate fleet evolution follow the logic that retiring vehicles are substituted proportionally to AHP competitive advantages.

(3)

Initial fleet composition (2025) is assumed to be dominated by ICEs (100%), as BEVs and HFCVs, so far, have marginal shares.

(4)

An annual 15% fleet replacement rate is imposed, based on the observation of fleet statistics.

(5)

Curve fitting is calibrated to initial (2025) and intermediate (2030) adoption levels, adjusted when empirical growth is insufficient. To ensure realistic dynamics, a constrained fitting procedure is implemented. For ICE technologies, decline is enforced with bounded lower asymptotes (L_min ≥ 0.5) and inflection points are placed earlier in the horizon. Growth/decline rates (k) are bounded within [0.2, 2.0] to avoid unrealistic dynamics.

3.6. Reproducibility

All analytical code, model implementations, and visualization scripts are publicly available through the OSF repository [https://osf.io/4m9jr]. The repository includes S-curve technology adoption models, ARIMA demand forecasting scripts, hierarchical modal split models, optimization routines, and comprehensive documentation under MIT license.

Due to contractual agreements with the Latvian Central Statistical Bureau, raw microdata cannot be published without prior processing and statistical summarization. Researchers must submit formal data requests directly to

Central Statistical Bureau of Latvia

Transport Statistics Division

Email: transport.statistics@csb.gov.lv

Website: www.csb.gov.lv/en/statistics/statistics-by-theme/transport

The approval process typically requires 4–6 weeks for academic research applications.

While raw microdata cannot be shared, aggregated and anonymized datasets are provided through the OSF repository, including regional freight flow matrices, technology adoption time series, infrastructure utilization statistics, and synthetic validation datasets for partial replication.

Analysis requires R 4.3.0+ (with forecast, dplyr, ggplot2 packages), 8 GB RAM minimum, and approximately 6–8 h processing time for complete reproduction.

3.7. Ethical Considerations

The experts provided informed consent after the study’s purpose was outlined. Voluntary participation, withdrawal rights, and procedures for handling sensitive data were also discussed. Recordings and transcripts were anonymized, securely stored, and accessed only by the research team. The methodology, including the consent process and data management, was approved by the Riga Technical University Ethics Committee as being compliant with institutional ethical guidelines.

4. Results

This section is structured in three parts to provide a coherent progression from demand characterization to technology adoption dynamics and market segmentation analysis. Section 4.1 presents the quantitative results on cold chain freight demand projections, including trend, seasonal, and residual decomposition and forecasting outcomes using ARIMA models. Section 4.2 reports the qualitative findings from semi-structured interviews, highlighting industry perceptions, barriers, and readiness regarding low- and zero-emission technology adoption in Latvian cold chain logistics. Finally, Section 4.3 integrates these insights through a market segmentation and adoption model, illustrating technology diffusion trajectories across distance segments and policy scenarios, and deriving implications for infrastructure sequencing and investment prioritization.

4.1. Demand Projections

The seasonal and trend decomposition using locally estimated scatterplot smoothing (LOESS) methodology (STL) applied to the 144-month cold chain transport dataset reveals a complex variance structure where irregular components dominate systematic patterns. It shows moderate trend component growth ranges from 138.72 to 216.84 million ton-kilometers corresponding to global cold chain trends, as reflected in Figure 1.

The decomposition attributes 30.8% of the total variance to the trend component, 19.5% to seasonal patterns, and 49.7% to residual variation, indicating that external factors and irregular shocks constitute the primary drivers of freight transport fluctuations rather than predictable systematic components:

−

The trend component demonstrates an upward trajectory with a linear slope of 0.542 million ton-kilometers per month (6.5 million ton-kilometers annually) throughout the analysis period, accumulating a total change of 78.12 million ton-kilometers. The trend strength coefficient of 0.308 indicates moderate but persistent long-term growth force, suggesting sustained structural drivers supporting market expansion independent of seasonal or irregular influences.

−

The seasonal decomposition shows moderate cyclical patterns with a seasonal strength coefficient of 0.195, indicating predictable but not dominant intra-annual variation, which probably follows agricultural harvest cycles. The seasonal component reaches its maximum amplitude in June, generating 36.2 million ton-kilometers above the trend baseline, while achieving its minimum in May, at 45.66 million ton-kilometers below trend. This seasonal range of 81.86 million ton-kilometers creates a peak-to-trough ratio of 1.38, representing substantial capacity planning challenges for infrastructure systems.

−

The residual component accounts for 49.7% of the total variance. This unexplained variation reflects the impact of external economic shocks, policy interventions, fuel price volatility, regulatory changes, supply chain disruptions, extreme weather events, and macroeconomic cycles. The standard deviation of annual changes reaches 9.42%, producing a coefficient of variation of 178.9%, indicating extremely high volatility relative to the mean growth rate.

This variance distribution suggests that while underlying growth momentum and seasonal patterns provide predictable system characteristics, external shocks and irregular influences constitute the primary source of short-term freight transport variability, requiring infrastructure planning frameworks capable of accommodating substantial demand volatility around systematic growth and seasonal trends.

Figure 2 presents the spatial distribution of food, beverage, and tobacco freight transport flows in 2023, categorized by distance thresholds aligned with technology operational characteristics. The analysis reveals significant asymmetry in the freight volume distribution between distance categories, with long-haul transport (>250 km) dominating the sector.

Figure 2 highlights a pronounced spatial dichotomy between short- and mid-haul domestic flows (566.9 million ton-kilometers; 29.4%) and long-haul international shipments (1358.7 million ton-kilometers; 70.6%). Contrary to the typical pattern observed across the European Union, where domestic and regional flows often dominate, long-haul freight represents the majority of the total transport volume in Latvia. This highlights Latvia’s strategic role in facilitating international trade corridors and its importance as a transit hub in 2023. However, due to sanctions against Russia and broader shifts in global supply chains, this situation may change significantly soon, as the largest long-haul flows were predominantly eastbound, with nearly one-third of them originating from the Riga agglomerate (LV006 and LV007) and bound for Russia, equaling 536.3 million ton-kilometers. Riga (LV006) is therefore also a domestic logistics hub.

Figure 3 provides transportation forecasting, distinguishing between short-haul (≤250 km) and long-haul (>250 km) operations. In a systematic model comparison, nine ARIMA configurations were evaluated for each segment, with selection criteria prioritizing residual independence (Ljung–Box p-value > 0.05) followed by information criteria minimization.

−

The long-haul freight segment was optimally modeled using ARIMA(2,0,1). The model projects a declining trajectory over the forecast horizon, with the total volume decreasing by 5.64% from 2024 to 2030, equivalent to an average annual decline of 0.96%. Forecast volatility measures indicate year-to-year fluctuations of 22.92 million ton-kilometers annually, reflecting the higher variability characteristic of long-distance freight operations. The long-haul model demonstrates superior forecasting stability despite its poor absolute precision. The confidence interval analysis reveals remarkably consistent uncertainty over the forecast horizon, with relative uncertainty growing minimally from 89.2% to 96.1% (1.08× factor) and the absolute interval width increasing by only 1.7. The stationary model structure without differencing suggests genuine market stability, with the wide intervals (1797 million ton-kilometer average width) reflecting high inherent volatility rather than model inadequacy.

−

The short-haul market segment required ARIMA (2,1,2). The forecast projects a growth trajectory with a total volume increasing by 8.18% over 2024–2030, representing an average annual growth of 1.34%. Short-haul operations exhibit lower forecast volatility (13.67 million ton-kilometers annually), suggesting more predictable demand patterns in regional freight movements. The short-haul model shows superior relative precision but concerning instability that undermines the forecasting reliability. While achieving a better coefficient of variation (84.9% vs. 94.4% for long-haul), the uncertainty expansion from 68.3% to 106.9% relative error represents a 1.56× growth factor that indicates model instability. The narrower absolute intervals (584 million ton-kilometer average) provide false confidence given the rapid uncertainty deterioration over time, making longer-term forecasts (2028–2030) practically unusable for infrastructure planning.

Although both models fail to meet acceptable forecasting precision standards for infrastructure investment decisions, they can be interpreted as exploratory trend indicators rather than precise planning instruments, with the wide confidence intervals representing an honest acknowledgment of freight market complexity rather than methodological failure. Future research should employ extended time series with more robust statistical validation of model assumptions.

4.2. Interview Results Regarding Technology Adoption in Latvian Cold Chain Logistics

The semi-structured interviews were performed with three cold chain logistics companies representing different fleet sizes: one large operator (>50 vehicles), one medium-sized company (10–50 vehicles), and one small operator (<10 vehicles). The current fleet composition shows an overwhelming reliance on conventional diesel vehicles, with 96–100% using Euro 6 diesel trucks. Only one company reported minimal adoption of alternative fuels (3% low-carbon renewable fuels, 1% rail transport).

The lock-in effects of heavy-duty transport technology adoption was justified, as all of the surveyed companies rated the integration potential of emerging technologies poorly on a 1–5 Likert scale, indicating a winner-takes-over effect:

−

BEVs were universally rated 1 (very low integration potential).

−

HFCVs were consistently rated 1 across all respondents.

−

ICEs were rated diversely (1–4), with larger companies showing higher integration potential.

The technology adoption difficulty followed similar patterns:

−

BEV and HFCV systems were rated 1–3 (very difficult to moderately difficult).

−

Low-carbon ICEs were rated 3–4 (more feasible for existing operations).

The following technological barriers captured by AHP provisions were identified:

−

All companies cited inadequate charging/refueling infrastructure as the primary barrier and the limited availability of low-carbon fuel supply points as the second.

−

High acquisition costs and limited second-hand markets were consistently mentioned across technologies. Companies emphasized the financial burden of transitioning from established diesel fleets. Reliability and range limitations emerged as critical issues. Weather dependency, uncertainty about the battery lifespan of BEVs, and additionally, range anxiety for long-haul operations were noted.

−

Maintenance complexity for new technologies.

As investment decision factors revealed clear hierarchies, the AHP remained unweighted:

−

Operational costs (universally rated 5).

−

Infrastructure availability (universally rated 5).

−

Customer requirements (rated 4–5).

Secondary considerations (rated 2–3) were environmental goals (consistently rated as low priority) and maintenance expertise availability, which were excluded from the AHP.

None of the surveyed companies had established formal sustainability targets for logistics operations, indicating reactive rather than proactive environmental strategies, confirming the prevalence of signaling versus behavior theory at this stage of adoption.

The resource availability assessment did not vary significantly between the companies, and thus, they were excluded from the valuation:

−

Financial Resources are severely constrained (rated 1–2), representing the most significant organizational barrier to technology adoption.

−

Human Resources are generally adequate (rated 4–5), suggesting sufficient personnel for managing technological transitions.

−

Technical resources have moderate availability (rated 3–4), indicating some capability for technology integration.

Leadership approaches showed client-driven orientation rather than autonomous sustainability commitment; therefore, the forecast-driven model was justified:

−

Willingness to invest in sustainable technologies contingent on customer demand and stable cargo flows.

−

Conditional acceptance dependent on transportation price adjustments.

−

Market-reactive rather than market-leading positioning.

The interview responses revealed inconsistent government support, which was captured in the discussion and policy implications:

−

One company reported the absence of governmental policies or incentives.

−

Another identified some subsidies for electric vehicle purchases.

−

A third company perceived no meaningful government impact on industry transition.

All companies rated infrastructure-related challenges as having a maximum severity (5 on Likert scale), reflected in the readiness index of the model:

−

Fuel availability for alternative technologies.

−

Charging/refueling infrastructure deployment.

−

Total cost of ownership considerations.

Moreover, the respondents reported limited awareness among their client bases regarding environmental transportation options. The companies indicated that sustainable transport adoption depends primarily on customers’ willingness to pay premium prices rather than regulatory requirements or competitive pressures.

The technology vendor ecosystem received mixed ratings (1–5), suggesting uneven market development and support structures.

Additional factors that varied among experts included the following:

−

One company reported no technological support.

−

Another identified a partnership with hydrogen infrastructure developers and electric vehicle manufacturers.

−

The support availability appeared to be correlated with company size and market position.

The survey results expose a significant gap between decarbonization policy objectives and industry readiness in Latvia’s cold chain logistics sector. Critical barriers include infrastructure deficits, financial constraints, and limited customer demand, while government support mechanisms remain inconsistent. The predominant client-driven business model creates a chicken-and-egg scenario where sustainable technology adoption awaits customer demand, while customers show minimal environmental awareness or willingness to pay premium rates for sustainable logistics services.

4.3. Market Segmentation and Adoption Model Implementation

The disaggregated analysis by vehicle category reveals distinct operational patterns that directly inform infrastructure deployment strategies and technology allocation decisions (Figure 4,

Table 2. Vehicle category distribution and distance-based operational patterns in food transport sector (2023).

Vehicle Category	Trip Count	Tonnage	Million Ton-Kilometers	Avg Distance	Trip Share	Ton-Kilometers Share	Long Haul Share	Short Haul Share
N1	2109	4468.2	237.1	118.4	12.1	0.8	62.7	37.3
N2	7169	53,139.3	3279.0	153.1	41.1	11.6	45.0	55
N3	8177	185,744.3	24,845.1	362.7	46.8	87.6	16.4	83.6

).

Heavy trucks (N3, >12 t) dominate freight transport operations, accounting for 87.6% of the total ton-kilometers while representing 46.8% of the trip frequency. Medium trucks (N2, 3.5–12 t) contribute 11.6% of the tkm volume across 41.1% of trips, while light commercial vehicles (N1, ≤3.5 t) generate only 0.8% of tkm despite comprising 12.1% of the trip frequency. This concentration indicates that infrastructure investment should prioritize heavy truck (N3) compatibility.

Vehicle categories exhibit distinct distance preferences that align with the proposed technology capabilities:

−

Light Commercial Vehicles (N1) predominantly operate in short and mid-haul segments (62.7% of ton-kilometers), with average trip distances of 118.4 km. This pattern strongly supports BEV deployment for the N1 category, as operational ranges align with current battery technology capabilities.

−

Medium Trucks (N2) display balanced distance distribution with slight long-haul preference (55.0% of ton-kilometers in >250 km category). The average trip distance of 153.1 km suggests transitional suitability for both BEV and HFCV technologies, depending on specific route characteristics and payload requirements.

−

Heavy Trucks (N3) demonstrate strong long-haul orientation with 83.6% of the tkm occurring beyond 250 km and average trip distances of 362.7 km. This category represents the primary target for HFCV infrastructure investment, as battery electric alternatives face significant operational constraints for heavy-duty, long-distance applications.

The analysis validates the hierarchical technology deployment strategy. The baseline analysis reveals a volatile freight market where external shocks dominate systematic patterns (49.7% irregular variance), with declining long-haul flows (−5.64% by 2030) contrasting growing short-haul segments (+8.18%). Heavy trucks (N3) bear 87.6% of the freight volume, which operate on predominantly long-haul journeys (362.7 km in average), validating the distance-based technology segmentation strategy. The high forecast uncertainty reinforces the need for adaptive infrastructure approaches responsive to evolving demand rather than static projections.

The implemented model shows distinct adoption dynamics across distance segments and policy scenarios, with short-haul operations showing substantial alternative technology penetration while long-haul segments remain predominantly diesel-powered through to 2030 (

Table 3. Technology market share projections by distance segment and scenario (2030, %).

Scenario	Distance Segment	ICE	BEV	HFCV
Base	Long-Haul (>250 km)	74.5	25.0	0.5
	Short-Haul (≤250 km)	37.8	60.1	2.0
Optimal	Long-Haul (>250 km)	74.6	25.0	0.4
	Short-Haul (≤250 km)	34.5	62.5	3.1
Maximum	Long-Haul (>250 km)	74.6	25.0	0.4
	Short-Haul (≤250 km)	38.1	61.9	0.0

ICE—vehicles with internal combustion engines; BEV—battery electric vehicles; HFCV—hydrogen fuel cell vehicles.

).

Battery electric vehicles achieve dominant market positions in short-haul operations (≤250 km), capturing 60.1% to 62.5% of the market share by 2030 across different scenarios. The Optimal scenario exhibits the strongest BEV performance (62.5%), reducing the ICE share to 34.5% while HFCVs capture 3.1%. This substantial penetration reflects TCO’s superior competitiveness, adequate range for daily return-to-base operations, and extensive urban charging infrastructure deployment. The limited variation across scenarios (60.1–62.5%) indicates that short-haul BEV adoption approaches natural market equilibrium, driven by operational economics rather than policy intervention intensity alone.

Long-haul segments (>250 km) demonstrate persistent ICE dominance at approximately 74.5% across all policy scenarios, with BEVs maintaining a consistent 25% penetration and HFCVs below 0.5%. This remarkable convergence indicates that long-haul technology transition faces fundamental constraints beyond policy intervention capacity within the timeframe up to 2030. The 25% BEV share represents adoption in lighter vehicle categories (N1 ≤ 3.5 t, N2 3.5–12 t) and specific route structures, enabling mid-journey charging, while heavy-duty operations (N3 > 12 t) requiring extended ranges with continuous refrigeration remain diesel-dependent.

Figure 5 presents the Base scenario evolution, demonstrating natural technology divergence by distance segment. Short-haul BEVs cross the 5% early adoption threshold by 2026, reach 15% growth acceleration by 2027, and approach 40% market leadership beyond 2029. Long-haul adoption follows constrained trajectories, with ICEs maintaining dominance throughout the forecast horizon.

Figure 6 illustrates the Optimal scenario with strategic infrastructure sequencing and targeted fiscal interventions. This configuration achieves the strongest overall alternative technology penetration (62.5% short-haul BEVs, 3.1% short-haul HFCVs, 0.4% long-haul HFCVs), reflecting coordinated hydrogen infrastructure deployment and enhanced TCO competitiveness through differentiated carbon taxation. Policy intervention markers at 2027 (5% adoption) and 2029 (15% adoption) align with BEVs’ market momentum in short-haul segments, demonstrating the systematic linkage between technology forecasts and infrastructure investment timing.

Figure 7 depicts the infrastructure volume requirements under the Maximum scenario, revealing the effects of aggressive policy intervention on infrastructure deployment patterns. Despite maximum policy intensity, short-haul BEV charging networks serve 61.9% segment penetration (intermediate between Base and Optimal), while HFCV infrastructure deployment is completely absent in short-haul operations. The long-haul infrastructure demands remain comparable with other scenarios, with 25% BEV and minimal HFCV (<0.5%) adoption. The Maximum scenario demonstrates a critical finding: indiscriminate policy intensity can reduce technology diversity through winner-takes-over dynamics, where aggressive BEV infrastructure investment crowds out parallel HFCV development despite simultaneous hydrogen infrastructure funding.

The convergence of long-haul projections across policy scenarios (74.5–74.6% ICE) provides critical guidance for infrastructure investment prioritization. This insensitivity to policy intensity indicates that long-haul technology transition requires extended timeframes beyond 2030 for fundamental infrastructure deployment or technology breakthroughs enabling competitive alternatives to heavy-duty diesel operations.

Conversely, robust short-haul BEV adoption (60–62% across scenarios) approaching natural market equilibrium supports aggressive infrastructure investment with high utilization confidence. This segment enables private capital participation through Public–Private Partnership (PPP) mechanisms requiring minimal demand guarantees, concentrating public investment on strategic corridor development and grid connection infrastructure while private operators deploy logistics hub charging facilities.

The analysis validates distance-segmented infrastructure sequencing strategies, concentrating near-term investment in short-haul BEV charging, where adoption signals demonstrate clear market momentum, while maintaining strategic but limited HFCV pilot infrastructure for long-haul optionality. This approach optimizes scarce public resources under declining aggregate freight demand while maintaining decarbonization trajectory compliance and technology diversity objectives for future heavy-duty applications.

4.4. Validation of Research Hypotheses

H1 Confirmed: Technology adoption in cold chain logistics follows predictable S-curve patterns with clear inflection points that can inform infrastructure investment timing.

H2 Supported: Strategic infrastructure sequencing based on adoption thresholds can reduce public investment requirements by 70% while achieving alternative technology penetration within the policy timeline.

The research demonstrates that systematic, evidence-based approaches to infrastructure investment can achieve environmental objectives while optimizing scarce public resources. The framework provides policymakers in less developed territories with practical tools for navigating the complex transition to sustainable freight transport under challenging conditions of declining demand and limited budgets.

5. Discussion

5.1. Policy Implications for Less Developed Territories

Technology-specific demand projections from the hierarchical modal split model reveal distinct adoption trajectories with critical distance segment differentiation. Short-haul battery electric vehicle (BEV) adoption achieves a 60–62% market share across all policy scenarios, indicating a policy-insensitive equilibrium driven by TCO competitiveness. In contrast, long-haul transport, which handles the majority of cold-chain traffic, retains a 74.5–74.6% ICE market share across scenarios, highlighting structural transition barriers within the 2030 timeframe. HFCV penetration remains limited at 2–3%, despite assumed infrastructure readiness, underscoring challenges in scaling alternative fuel technologies for long-haul applications. These findings emphasize the need for strategic infrastructure deployment sequencing, where efficiency gains can be quantified using NPV analysis.

Two infrastructure deployment strategies were evaluated using a 4% discount rate over the 2024–2030 period, with costs derived from Xie and Minjares (2025) [42] and aligned with the National Energy and Climate Plan:

−

Full Deployment (Year 1): Immediate rollout of all infrastructure, including 100 EV charging stations (EUR 7 million), two hydrogen refueling stations (EUR 10 million), five LNG refueling stations (EUR 5.75 million), and biofuel blending infrastructure (EUR 0.5 million). Total CAPEX: EUR 23.25 million; annual OPEX: EUR 2.40 million.

−

Phased Deployment (Years 1–6): Gradual rollout matched to adoption curves, with total CAPEX of EUR 18.25 million and annual OPEX ranging from EUR 0.60 to 2.30 million over the period 2024–2030.

Table 4. Direct NPV savings for 2024–2030, million EUR.

Year	Upfront Scenario	Sequenced Scenario	Annual Savings PV
1	25,650,000	4,450,000	21,200,000
2	2,308,000	2,428,000	−120,000
3	2,219,000	2,613,000	−394,000
4	2,134,000	2,934,000	−800,000
5	2,051,000	3,035,000	−984,000
6	1,972,000	4,706,000	−2,734,000
7	1,896,000	4,799,000	−2,903,000
Total	38,230,000	24,965,000	13,265,000

presents the NPV savings for the two strategies at a 4% discount rate, yielding total savings of EUR 13.3 million for the phased approach.

Infrastructure utilization rates further highlight the efficiency advantages of the phased deployment strategy. The sequenced approach achieves a 64% average infrastructure utilization rate compared to 39% for full deployment, resulting in a 64% improvement in infrastructure productivity. This translates to an annual efficiency loss reduction of EUR 0.84 million (EUR 2.4 million × 0.35), equivalent to EUR 4.92 million over the 2024–2030 period. These findings align with real-world evidence from Latvia, where underutilized hydrogen refueling stations incur operational losses, leading to hydrogen fuel cell prices 3.5 times higher than conventional fuels due to low demand.

The substantial NPV savings and improved infrastructure productivity underscore the superiority of phased deployment. By aligning infrastructure investments with adoption curves, this strategy mitigates the risk of overcapacity, optimizes resource allocation, and enhances economic efficiency. These results advocate for policy frameworks that prioritize adaptive, demand-driven infrastructure rollout to support the transition to low-emission transport technologies, particularly in markets with significant structural barriers to alternative fuel adoption.

5.2. Comparison with Evidence from the Literature

The S-curve adoption patterns observed align with established technology diffusion theory [31] but reveal specific characteristics for cold chain applications. The inflection points identified correspond to TCO competitiveness thresholds rather than pure technology maturity, supporting the recent literature on cost-driven adoption in freight transport [21,22].

The survey validation confirms the “messy middle” transition period predicted by Larson et al. (2024) [23], with multiple technologies coexisting rather than single dominant solutions emerging. The finding that technology choice depends on distance segments and operational requirements rather than uniform preferences supports the segmented adoption approach advocated for by Müller (2024) [27]. However, on declining segments, the winner-take-over pattern suggests specific sequencing of the infrastructure deployment. The framework’s emphasis on “right-sizing” infrastructure for efficiency rather than growth addresses a critical gap in the transport planning literature, which predominantly assumes demand growth [43]. The long-haul demand decline projected through 2030 creates infrastructure planning challenges not adequately addressed in existing frameworks designed for expanding markets.

The EUR 18.2 million public investment reduction through strategic sequencing provides quantitative validation for efficiency improvements, made possible through evidence-based timing. This finding extends the infrastructure optimization literature by demonstrating specific mechanisms for achieving efficiency gains in resource-constrained environments.

The survey finding that private sector resources are severely constrained (financial rated 1–2) while public resources face competing demands validates the need for innovative PPP structures. The framework’s emphasis on automatic adjustment mechanisms based on adoption rates addresses the uncertainty challenges identified in the PPP literature for declining markets.

5.3. Limitations and Future Research Directions

5.3.1. Model and Data Limitations

The static modeling approach simplifies complex dynamic interactions between policy interventions and technology adoption. The S-curve parameters are calibrated to anchor points (2025, 2030) derived from expert judgment rather than observed market behavior, creating potential bias in adoption rate estimates. Future research should incorporate dynamic feedback effects where infrastructure availability influences adoption rates, which in turn affects infrastructure utilization and investment decisions.

The focus on a single statistical group (food, beverage, and tobacco) limits the generalizability to other freight segments with different operational requirements and cost structures. Cold chain applications represent specialized use cases that may not reflect broader freight transport adoption patterns. Extension to general freight, construction, and manufacturing segments would strengthen the framework’s applicability.

Geographic limitation to Latvia constrains the framework’s regional relevance. While Latvia provides a useful case study for less developed territories, validation in other Baltic states and similar regional contexts would improve confidence in the approach’s transferability. The absence of comparable data from Estonia and Lithuania prevents a direct regional comparison and coordination analysis.

5.3.2. Methodological Constraints

The survey validation relies on three companies, limiting statistical inference capabilities. While the companies represent different operational scales and provide qualitative insights, the sample size prevents robust quantitative validation of adoption predictions. Larger-scale industry surveys would strengthen the empirical foundation for technology adoption modeling.

The ARIMA demand forecasting assumes that historical patterns continue, which may not hold given structural changes in trade relationships and supply chain configurations. The long-haul demand decline projection reflects recent geopolitical disruptions that may not persist throughout the forecast horizon. Scenario-based forecasting incorporating alternative trade relationship developments would provide more robust planning foundations.

The Technology–Organization–Environment framework provides structured analysis but may oversimplify the complex decision-making processes that exist in the freight transport sector. Organizational factors such as fleet management strategies, financing arrangements, and operational planning cycles could significantly influence adoption timing beyond the factors captured in the survey instrument.

5.3.3. Data Availability and Quality Issues

The limited availability of real-world operational data for HFCV systems constrains model validation and parameter estimation. Most hydrogen fuel cell applications remain in the pilot phases, creating uncertainty about operational costs, reliability, and infrastructure requirements. Systematic data collection from demonstration projects and early commercial deployments would improve forecasting accuracy.

The total cost of ownership calculations requires detailed cost data that may not be publicly available, particularly for emerging technologies. A sensitivity analysis on cost assumptions would strengthen the framework’s robustness to parameter uncertainty. Industry partnerships providing proprietary cost data would enable more accurate competitiveness assessments.

Infrastructure utilization data for alternative fuel systems remains sparse, limiting the validation of capacity planning assumptions. The framework assumes that standard utilization patterns may not hold for emerging technologies with different operational characteristics; therefore, real-time utilization monitoring from early infrastructure deployments would provide empirical validation for planning models.

While the sample size of respondents may not fully represent the entire Latvian cold chain sector, respondents were selected to cover diverse stakeholder perspectives, including logistics operators, technology providers, and regulatory bodies. These insights provide valuable indications of perceived barriers to BEV and HFCV adoption and inform the modeling of technology uptake. However, the results should be interpreted as indicative trends rather than statistically representative of the whole sector. Future studies with broader participation would strengthen the generalizability of these findings.

5.3.4. Future Research Priorities

Dynamic modeling with policy–market feedback loops can incorporate endogenous relationships where infrastructure investment affects adoption rates, which influence utilization levels and subsequent investment decisions. Agent-based modeling approaches could capture these interactions while maintaining computational tractability.

An incorporating a dynamic feedback loop, where infrastructure deployment in one period directly influences the TCO and subsequent adoption rates, could provide a more interactive and responsive modeling framework. Such a mechanism might accelerate technology uptake in early deployment scenarios and alter optimal investment sequencing.

Extending the present analysis beyond cold chain transport to general freight, construction, and passenger transport applications would be beneficial. A cross-sector analysis would identify synergies in infrastructure development and technology adoption patterns that could improve investment efficiency.

In terms of developing institutional frameworks for cross-border infrastructure development and technology standardization, an analysis of existing coordination mechanisms in the energy and telecommunications sectors could provide templates for transport infrastructure cooperation.

Integrating detailed lifecycle emission calculations with infrastructure deployment scenarios to quantify environmental benefits relative to investment costs would strengthen the environmental policy case while identifying optimal decarbonization strategies. For instance, high capital cost of hydrogen refueling stations, combined with the projected low adoption of hydrogen fuel cell vehicles (<0.5%) in the long-haul segment by 2030, raises the potential risk of stranded assets. Explicitly modeling this risk would require detailed assumptions about station capacity, utilization thresholds, and investment timing, but support more robust infrastructure planning decisions.

When developing frameworks for continuous model updates based on observed adoption rates and infrastructure utilization, machine learning approaches could enable automatic parameter adjustment and improved forecasting accuracy as empirical data becomes available.

6. Conclusions

This study addresses a critical policy challenge in transport infrastructure planning: making strategic investment decisions when freight demand is declining yet decarbonization mandates require substantial infrastructure upgrades in less developed territories. The integrated forecasting framework demonstrates that evidence-based infrastructure sequencing can significantly improve investment efficiency while maintaining environmental objectives.

6.1. Practical Implementation Value

Strategic sequencing based on adoption thresholds can reduce public investment requirements by approximately 70% while achieving alternative technology penetration within the policy timeline. This efficiency gain results from avoiding premature infrastructure deployment that leads to underutilization and preventing delayed deployment that creates adoption bottlenecks.

The analysis validates distance-based technology allocation with BEVs serving short-haul operations (≤250 km, 18.8% of freight volume) and HFCVs targeting long-haul transport (>250 km, 81.2% of volume). This specialization maximizes each technology’s operational advantages while minimizing infrastructure redundancy and stranded asset risks.

The framework provides common adoption metrics and investment timing criteria applicable across multiple jurisdictions, facilitating cross-border infrastructure coordination. Standardized threshold definitions enable coordinated regional development strategies while respecting national policy autonomy.

6.2. Policy Recommendations

Policymakers should align infrastructure deployment with demonstrated market adoption rather than leading adoption significantly. The 5–15–40% threshold system provides objective triggers for investment decisions that maximize utilization while ensuring adequate capacity for emerging technology leaders. Infrastructure deployment should begin when technologies approach 5% market share, PPP mechanisms should be activated at 15%, and operations should be optimized at 40% penetration.

Rather than developing infrastructure for all potential technologies simultaneously, regions should focus on technologies matching their transport characteristics and operational requirements. For territories with predominantly long-haul freight patterns such as Latvia, HFCV infrastructure should receive priority, while regions with concentrated short-haul networks should emphasize BEV charging development.

Cross-border infrastructure harmonization should precede domestic deployment to capture transit traffic benefits and achieve economies of scale. Regional specialization with complementary technology leadership (each territory focusing on specific technologies while ensuring compatibility) can optimize collective investment efficiency while maintaining individual policy autonomy.

The survey evidence reveals market failure, in which logistics companies await customer demand while customers lack awareness of sustainable transport options. Regulatory pressure through carbon taxation and emission standards may be more effective than voluntary incentive programs for breaking this deadlock and accelerating adoption.

6.3. Future Implementation and Scalability

The forecasting framework can be regularly updated with new adoption data and infrastructure utilization statistics to provide continuous guidance for investment decisions. The success of implementation depends on sustained political commitment to evidence-based decision-making over politically expedient but economically inefficient intervention timing.

The methodology’s transferability to other territories requires calibration to local transport characteristics, institutional capabilities, and budget constraints. However, the core threshold-based approach and distance-segmented technology allocation strategy provide generalizable principles applicable across similar regional contexts.

Scaling to multi-country coordination requires standardized data collection protocols and shared performance metrics. The framework’s adoption thresholds provide common reference points for coordinated investment decisions while respecting national policy autonomy and regional specialization strategies.

Adaptive management capabilities can be enhanced through real-time monitoring systems that track adoption rates, infrastructure utilization, and technology performance indicators. Machine learning integration could enable automatic parameter adjustment and improved forecasting accuracy as empirical evidence accumulates from early deployment phases.