New Approaches to Reducing Environmental Harm from Combustion Engine Vehicles, the Case for the Application of Systems Thinking

1. Introduction

Since its conception, the internal combustion engine (ICE) has been subject to continuous research and development in an attempt to improve reliability, optimise performance and fuel economy, and latterly, to reduce gaseous tailpipe emissions and particulates. Throughout the early to mid-2000’s climate mitigating initiatives were directed towards the reduction of greenhouse gas (GHG) emissions, specifically carbon dioxide (CO₂), and manufacturers were encouraged to develop small capacity high speed diesel engines to power the light duty vehicle (LDV) fleets. With the shift in consumer attitudes following the ‘dieselgate’ [1] scandal and the resulting move away from diesel powered LDVs, in part driven by government policies aimed at reducing nitrogen oxides (NO_x), attention has turned to the longer term solution of phasing-out ICE powered LDVs.

In response to the requirement to reduce gaseous tailpipe emissions of the LDV fleet, there has been a move to zero tailpipe emission battery electric vehicles (BEV). However, whilst this transition to BEV and the removal of GHG vehicle tailpipe emissions represents a clear strategic vision, the issues of cost and infrastructure have limited this transition predominately to the markets of the Global North and China. Hence, large parts of the global vehicle market, those outside of the Global North and China, will remain aligned to the combustion engine vehicle for the foreseeable future. Further, the expected increase in the number of LDVs, to a figure of approximately 1.7 to 2 billion cars globally in 2040 [2,3], will be primarily realised in those markets outside of the Global North i.e. the developing countries of the Global South.

The requirement is that in those markets that are unable to transition to BEV and who remain wedded to ICE, that all available options to reduce gaseous tailpipe emissions are pursued. However, the consequence of a change in focus of the industry to meeting the high development costs associated with the transition to BEV, together with the longer term reduction in market for ICE in the Global North, has led to a question over investment in further development of cleaner ICE technologies. This requires that alternative approaches are considered to reduce tailpipe GHG emissions under the conditions of lower investment.

Optimising the environmental performance of the ICE through identifying and exploiting the critical relationships between existing emissions reduction strategies is one approach. Previous research into individual vehicle tailpipe gaseous emissions reduction strategies has revealed that the impact of individual actions can be leveraged when deployed in combination e.g. deploying vehicle lightweighting with a concurrent change to bioethanol [4,5,6,7,8]. Therefore, research in this paper looked to apply the approach of systems thinking to the challenge of reducing ICE emissions. With systems thinking and by employing qualitative systems mapping using causal loop diagrams (CLDs) it is possible to portray the dynamics, inter-relationships and feedback characteristics present within complex systems. Through identifying and exploiting the critical relationships between existing emissions reduction strategies, further improvements in the environmental performance of ICE vehicles could be realised. Specifically in this case the interconnections and interdependencies between engine downsizing, vehicle lightweighting, engine turbocharging and bioethanol fuels, previous investments in intellectual property (IP) and manufacturing could be exploited for the benefit of further emission reduction.

Section 2 provides a short introduction to systems thinking, a description of system dynamics (SD) methodology and CLDs. Section 3 describes the methodology; section 4 presents the complex systems model and the associated sub-system diagrams accompanied by results and interpretations of their meaning and their effect on the GHG emissions reference mode. Discussions are presented in section 5 followed by conclusions and suggestions for future research directions in section 6. The research focusing exclusively on SI engine powered LDVs.

2. Systems Thinking, Systems Dynamics Modelling & Causal Loop Diagrams

Systems thinking is a method that can be used to represent causal relationships between the variables that define a complex system. The definition of systems thinking varies according to the discipline [9] and has been defined and redefined numerous times since the term was first coined by Barry Richmond [10]. Currently, no uniformly accepted definition exists as demonstrated by the works of numerous researchers, [9,10,11,12], who have compared the various definitions proposed by leading experts in the field. The authors of this paper have adopted the definition put forward by Amissah, et al, [11] who state, ‘systems thinking is aimed at understanding relationships between components and their overall impact on system outcomes (i.e., intended and unintended) and how a system of interest similarly fits in the broader context of its environment’ (p.1) as the frame of reference for their work.

System dynamics modelling is a method used to analyse system related problems in which time is an important factor [13]. System dynamics uses simulation modelling tools to interpret the structure of the system [14], and one of these tools is causal loop diagrams. CLDs form the initial stage in the development of a system dynamics model [15] and are a visual diagramming technique [16] that qualitatively maps the elements/variables of the system and shows the causal relationships that are formed among them [14]. They communicate sources and implications of interactions and feedback within the system. In practice, CLDs display variables as text and the causal relationships between them are represented by arrows. The arrows indicate the direction of causality, the type of relationship (i.e., proportional or inverse), and whether there is any delay in an expected effects’ occurrence. Causal links must have either a positive (+) or negative (-) polarity and a collection of links that provides positive feedback is called a reinforcing loop. A collection of links that provides negative feedback is called a balancing loop [17]. Negative loops are self-correcting and counteract change [18]. Loop polarity may be established by counting the number of negative links, if this is an odd number then the loop is a balancing one. However, this method of determining loop polarity only works properly if the link polarities are correctly assigned. An alternative way to establish loop polarity is to trace the effect of a small change in one of the variables as it circulates around the loop [17].

CLDs do not predict what is going to happen to a system, instead they simply represent the structure of it and how it could behave in time under certain circumstances [14]. The method is a qualitative tool for modelling and relies on the researcher’s intuitions [19], however, when multiple loops interact it is often impossible to use intuition to determine what the dynamic effects will be and therefore it is necessary to develop a quantitative model using stocks and flows, applying mathematical equations to CLD models to allow for simulation [16,20]. To the best of the authors knowledge no work exploring the variables in the system of interest had previously been published and given that systems thinking and the associated dynamic modelling has predominately been applied to the solution of problems in the business and public policy domains it was unclear if this technique would be appropriate or what its true value would be. However, according to Sterman [17] ‘the usefulness of models lies in the fact that they simplify reality, creating a representation of it we can comprehend’ (p.89). CLDs are flexible and can be used to portray the feedback structure of a system from any domain. It was recognised therefore that adopting a systems thinking approach and the construction of CLDs would provide a useful visual representation of this complex system and its behaviour, and that this virtual realisation may inform experiments in the real-world, leading to the desired improvements.

3. Methodology

The research adopted the approach outlined in Rahman [21], which was a development of the framework developed by Sterman [17]. The approach is comprised of five structured stages, these being:

i.

Problem articulation (boundary selection).

The problem definition was identified as being the achievement of reduced tailpipe GHG emissions by leveraging existing technology and IP. Boundary selection is listed as part of problem articulation when considering textbook approaches to system dynamics modelling – the distinction between endogenous, exogenous and excluded variables. A taxonomy approach was adopted, considering primary physical systems (turbocharging; engine-downsizing; bioethanol; and lightweighting). A review of the literature was undertaken to identify and classify variables.

ii.

Formulation of a dynamic hypothesis.

The dynamic hypothesis is a view point of the modeller about what structures and interactions exist that can generate the expected behaviour. For each primary physical system the endogenous and exogenous variables identified from stage 1 are mapped to show the relationships – these include positive and negative relationship as well as complex and conditional.

iii.

Simulation model formulation.

In developing a system model, the final stage is formulation. Model formulation concerns the aggregation of the individual dynamic hypotheses from stage 2 – leading to an understanding how the actual system behaves.

iv.

Model testing.

Model testing is asking questions of the model to explore its functioning, and evaluation is related to confirmation of the dynamic hypotheses. The methodology was to ask questions based on observations extracted from literature and to ensure that the functioning of the model was apt based on feedback loops identified (present or absent), the classification of feedback loops (reinforcing or balancing), etc.

v.

Policy design and evaluation.

Whilst policy design and evaluation is the final stage of the approach proposed by Rahman, this research does not propose or evaluate, but elaborates upon the role of current policy design in the system, how that intervention could be developed and where there may be opportunity for additional intervention.

The following sections detail the application of the above methodology, with section 4.0 providing the key results and resulting system model and section 5.0 a discussion with regards the implications of the insights arising from the same.

4. Results

4.1. Problem Articulation (Boundary Selection).

A comprehensive review of literature revealed that there was the potential to achieve environmental gains by combining different mitigating strategies, such as the work by Lewis, et al, [4] who concluded that the combination of mass reduction and use of advanced engines in the same platform reduces life cycle GHG emissions. Similarly, work by Leach, et al, [2] concluded that ‘the fuel consumption and hence the GHG impact of spark ignition engines can be significantly reduced using existing technology’ (p.14). Based upon a review of the literature, the mitigating strategies of interest selected were: engine downsizing; vehicle lightweighting; engine turbocharging; and bioethanol fuels. Based upon these strategies the endogenous, exogenous and excluded variables were identified based on their prevalence and significance within the literature (Table 1).

4.2. Dynamic Hypothesis

The dynamic hypotheses were created based around the emission reduction strategies of interest, these being; turbocharging, engine downsizing, bioethanol and vehicle lightweighting. It should be noted that these strategies become ‘variables’ within the causal loop diagram presented in Figure 5, and are therefore referred to interchangeably throughout the following text and in Table 2, Table 3 and Table 4.

4.2.1. Turbocharging

The literature review revealed the key variables related to the turbocharging sub-model to be as listed, with references, in Table 1 and shown diagrammatically in Figure 1. It can be seen that there are two inputs and four main outputs. From the literature, Yang, et al [23] states that a turbocharger reclaims energy from the exhaust gas to boost the intake air thus improving the power density of the engine, which is one of the key enablers to achieve engine downsizing. The reduction of fuel consumption is today a significant factor in engine design and development [22] as it is a means of reducing harmful gaseous tailpipe emissions, and is primarily achieved through engine downsizing, increasing the engine compression ratio and the fitment of a turbocharger. However, due to the higher in-cylinder compression pressures and higher combustion temperatures, the likelihood of ‘engine knock’ is increased leading to reduced engine efficiency as a result of increased heat transfer through the cylinder walls, plus the likelihood of catastrophic mechanical damage to the piston, cylinder, valves and other engine components. One solution to mitigate this is the use of higher octane fuels [26]. Turbocharging can also make a significant contribution to enhancing engine performance [25] and in conjunction with advanced engine mapping, more efficient compression ratios and advanced automotive technologies, the problem of turbo lag can be almost entirely eliminated and engine response improved.

From Table 1, it can be seen that some variables assigned to the turbocharging sub-model were considered to be exogenous and others were excluded from the model. Whilst costs and user behaviour are important factors they were excluded as they were considered to be outside the scope of this work. From Figure 1 it can be seen that connections between variables are indicated by arrows, the direction of the arrow indicating the causal flow. On the input side the use of waste energy from the engine exhaust gas is used to drive the turbocharger and high octane fuels are used to eliminate combustion knock. Plus (+) signs have been added to the outputs to indicate that a reinforcing causal relationship exists such that if one variable increases or decreases the same behaviour would be seen in the other, e.g., the effect of turbocharging will positively improve engine performance and fuel economy.

4.2.2. Engine Downsizing

The key variables for the engine downsizing sub-model are shown in Figure 2, these being chosen as a direct result of the findings of the literature review, a full list of references are provided in Table 1. Research confirms that engine downsizing is an important strategy in the drive to reduce fuel consumption [8] and GHG emissions and is primarily achieved by a reduction in engine displacement volume and the fitment of a turbocharger, which provides high density air to the combustion chambers, thus increasing torque, power density [23] and engine efficiency [26]. Jo, et al [8] state ‘Downsizing along with turbocharging also increases efficiency by changing the engines’ operating regions to where pumping, friction and heat transfer losses are relatively low’ (p.1). The reduction in the mass of the engine also contributes to vehicle lightweighting and the enhancement of fuel economy [29]. The emissions regulations variable is also included as such regulations are the main driver for the introduction of emission reduction strategies. Cost factors were excluded from the model as firstly they were, as with the turbocharging sub-model, considered to be outside the scope of the work. Additionally, research by Ward, et al, [27] concludes that cost savings are relatively small for ICE powered vehicles. Distance travelled is a key factor in engine life, however, this was also excluded from the sub-model as the increase in environmental impact due to engine wear is unlikely to be large. Downsizing and pressure boosting tends to stress the engine beyond the limits seen in classical design [28]. The importance of coatings for engine parts and in design pairing of low viscosity motor oils with the engine characteristics to reduce engine component wear and provide improvements in fuel economy as discussed by Lee, et al, [28] were initially considered but ultimately excluded due to the relatively small fuel consumption gains arising. Connections between the variables are, as with the turbocharging sub-model, represented by arrows showing the direction of flow. Plus (+) signs have been added to the outputs to indicate the reinforcing causal relationships.

4.2.3. Bioethanol

The key variables for the bioethanol sub-model were established from the extensive literature review and are shown in Figure 3, with a full list of references provided in Table 1. Three inputs and four outputs were identified. Alimin, et al, [36] observe that ‘policies to reduce emissions and conserve energy are the primary motivators for engine modifications and fuel science’ (p.1). In the case of this paper, the fuel science being confined to bioethanol derived from feedstocks such as corn. Biofuels can have a positive effect on climate change as the feedstock removes CO₂ from the atmosphere whilst the crop is growing. The type of feedstock used to produce the biofuel influences its effect on climate change due to the impact of nitrous oxide (N₂O) emissions from the fertilizer used to grow the crop [34]. N₂O has a global warming potential (GWP) factor of 298 compared to 1 for CO₂. Puricelli, et al, [35] comment that ‘biofuels can be a promising alternative to fossil fuels, provided they are produced from advanced feedstocks. In particular, feedstocks that lead to land use change should be avoided’ (p.15). Most research concludes that the use of bioethanol can lead to reductions in tailpipe GHG emissions, for example, research by Milovanoff, et al [30] states that ‘replacing conventional gasoline with mid-level ethanol blends (15–30% ethanol by volume) can reduce fossil fuel use and GHG emissions’ (p.1). Work by Kroyan, et al, [32] showed that the use of an E22 blend (22% of ethanol by volume) reduced the CO₂ emissions in their specific model by 1% despite an increase in fuel consumption of 7.76%. Ethanol has a higher research octane number (RON) and (latent) heat of vaporisation than petrol, the latter giving rise to a cooling effect in the engine, thus, enhancing knock resistance [26,31]. The potential of bioethanol to control engine knock means that it has been widely adopted in downsized, high compression, turbocharged engines. Work by Obeid, et al, studied engine brake power vs. relative air-fuel ratio and concluded that it was possible, using E20 fuel, to operate the turbocharged engine tested in their research at the full load using a lambda value of 0.98 without any engine knock. To obtain the same power from E0 (petrol only) required a lambda value of 0.91. The knock resistance properties of bioethanol being attributed to the fuels superior combustion properties.

A number of factors were excluded from the bioethanol model as listed in Table 1. Whilst all of these factors are important they were considered to be outside the scope of the present work. From Figure 3 it can be seen that renewable fuel policy drives the amount of feedstock and resultant conversion, which in turn has a reinforcing causal relationship with renewable fuel. Simply put, if policy dictates an increase in renewable fuel, then there will be an increase in feedstock and conversion leading to an increased quantity of renewable fuel. There are four outputs each of which has a reinforcing causal relationship with bioethanol.

4.2.4. Vehicle Lightweighting

The key variables for the vehicle lightweighting sub-model were derived from the literature review and comprise of two inputs and three outputs as shown in Figure 4. Vehicle lightweighting is a key strategy for improving fuel consumption and reducing GHG emissions [2,27,37,39,40]. According to Joost, as cited by Ward, et al [27] for an ICE powered vehicle a 7% increase in fuel economy is achievable for a 10% reduction in weight. Additionally, life cycle assessment (LCA) results from work by Raugei, et al, [38] suggest that reductions in environmental impact of approximately 7% are possible in most lifecycle impact categories, hence, further strengthening the case for vehicle lightweighting. Primary lightweighting is concerned with reducing weight from the body in white/chassis, whereas, secondary lightweighting is achieved through engine downsizing and the use of lighter components in the powertrain, suspension and other vehicle sub-assemblies. Engine downsizing is important as it is an effective strategy for improving fuel economy and also contributes to the total secondary mass reductions. Lewis, et al [5] conducted evaluations of life cycle energy and GHG emissions of baseline and lightweight internal combustion engine vehicles (ICEV), hybrid electric vehicles (HEV) and plug in hybrid electric vehicles (PHEV) with aluminium and aluminium/high strength steel, using a range of scenarios.

The results showed that, with the inclusion of secondary mass reductions, lightweight vehicles are 16% lighter than baseline vehicles, and that 35 – 41% of the reduction was due to secondary mass reduction. Similarly, work by Lewis, et al, [4] reveals that by combining vehicle lightweighting techniques with a high efficiency petrol engine (i.e. downsized and boosted) in the same platform, more reductions in GHG emissions and life cycle energy can be achieved. In their work they showed that a naturally aspirated ICE combined with a body in white subject to a 35% mass reduction provided a 7% reduction in life cycle impact and that a high efficiency petrol engine combined with a body in white that had no mass reduction provided a 22% reduction in life cycle impact. By contrast, a high efficiency petrol engine combined with a body in white subject to a 35% mass reduction provided a 27% reduction in life cycle impact. They attributed the decrease in life cycle impacts entirely to a reduction in fuel consumption.

Fuel consumption is powerfully influenced by vehicle weight [6] and according to Koffler, et al, cited by Del Pero, et al, [6] about one third of an ICE vehicles’ total fuel consumption directly depends on its weight. Vehicle dynamic performance has been included in the model as this is enhanced by lightweighting, especially by secondary reductions to unsprung components, which have a positive effect on the dynamics, manoeuvrability and stability [40]. As with the other sub-models, some factors were excluded as they were considered to be outside the scope of the study. These being, the extraction of raw materials, manufacturing, end of life recycling, technological change, materials selection and cost. As can be seen from Figure 4, arrows show connections and indicate the direction of flow. Plus (+) signs have been added to the outputs to show their reinforcing causal relationship with vehicle lightweighting. The full list of references relating to this sub-model are provided in Table 1.

4.3. Simulation Model Formulation

The CLD in Figure 5 illustrates the causal loops and causality chains that exist within the whole system. The loops are assigned identification letters and numbers and the reinforcing characteristics of the loops are indicated by the letter ‘R’. No ‘balancing’ loops were identified. To aid clarity the respective causality chains are presented in Table 2. Each sub-system and variable was also assigned to a category as shown in Table 4. The categories being colour coded as shown in the key inside Figure 5. The ‘blends’ variable was assigned to the technology category as creation of fuel blends is the result of the application of scientific knowledge. Blends are engineered and may become a resource due to technology, hence, in the current work they were assigned using this reasoning. Similarly, this thinking was also applied to high octane rating (RON) fuels. The ‘renewable fuel policy’ variable was assigned to the regulation category as many countries have policies that require fuel suppliers to supply a given percentage of renewable fuel to the market as in the United Kingdom, or which requires petrol to contain a minimum volume of renewable fuel as in America. In Europe, the Renewable Energy Directive [41] places a binding target on the share of renewable energy in the total EU energy consumption by 2030. Section 4.4.3 explores the distribution of the different categories within the individual causal loops within the whole system model.

Figure 5. Complete causal diagram for the whole system.

Figure 5. Complete causal diagram for the whole system.

4.4. Model Testing

The model testing phase required asking questions of the model. For the loops identified the behaviour is outlined in Table 2. We need to understand how the loops interact and whether the requirement is to drive positive feedback loops to reinforce them or whether combinations are necessary. For example, the reinforcement of L1 would be negated if the L11 was not also concurrently reinforced as a downsized engine in an existing vehicle frame would potentially increase emissions. The following sections therefore explore the functioning/driving of each loop, the dependencies and the strength of the relationship with technology, policy, resources and behaviour. The ambition is to develop a clear understanding of the system with the purpose of identifying the parts of the system where intervention would be productive.

4.4.1. the Causal Loops

As stated in section 2, when interpreting the causal loops it is important to appreciate how the system of interest fits in the broader context of its environment and how the virtual model informs the design and implementation of solutions in the real world. In reality, the virtual model informs the real world and drives physical changes, which themselves can then be fed back into the virtual model to stimulate further improvements. In the present work the descriptions of each loop, as given below, should be considered in the context the United Nations sustainable development goals (SDG), which are external to the system but are the overarching driver for the reduction of harmful tailpipe GHG emissions and which have a direct impact on the behaviour of the system through the emissions regulations and renewable fuel policy. In order to provide further context and to illustrate how the loops relate to society and the current environment, examples are included in some of the explanations thus demonstrating the potential behaviour of the particular loop.

• Causal loop L1 illustrates how, driven by the need to comply with stricter emissions regulations, engine downsizing acts with vehicle lightweighting to produce an increased reduction in tailpipe GHG emissions. The loop is considered to be reinforcing as the combination of strategies has a positive effect on reducing tailpipe GHG emissions. Similarly, if there were a reduction in the requirements of the emissions regulations or less emphasis on engine downsizing, vehicle lightweighting or a combination of any of these, there would be a negative effect on the reduction of tailpipe GHG emissions. Simply stated if the cause increases the effect increases and vice versa.
• Causal loop L2 shows the reinforcing causal relationships between engine downsizing, fuel economy, tailpipe GHG emissions and emissions regulations. Basically, engine downsizing is a technique for improving fuel economy, and, by association, reduces tailpipe GHG emissions, thus helping to meet emission regulations. Within this loop the engine downsizing and emission regulations variables are significant, if there is a change in either, there would be a commensurate change in the other variables. For example, when the Euro 7 emission standard is implemented in July 2025 for new ICE powered vehicles sold in Europe, there will be a direct impact on the other variables in this loop.
• Causal loop L3 builds on the relationships seen in loop L1, through the inclusion of improved fuel economy, as this is a further causal effect of engine downsizing and vehicle lightweighting. The loop is considered, as with loop L1, to be reinforcing. Improved fuel economy is particularly important in some global markets as, for example, this is a significant strategy to meet American Corporate Average Fuel Economy (CAFE) standards, which are designed to increase fuel economy. The latest requirement being 2% per year for model years 2027-2031 [42].
• Causal loop L4 builds on loop L2 by introducing the ’improved efficiency’ variable, where efficiency is defined as an overarching term covering engine and overall vehicle efficiency. The causal relationships in this loop are considered to be reinforcing.
• Causal loop L5 shows the reinforcing relationship between engine downsizing, reduced tailpipe GHG emissions and the emission regulations.
• Causal loop L6 illustrates the reinforcing relationship between turbocharging and engine downsizing, demonstrating the positive relationship turbocharging has on engine downsizing as a key enabler. In this loop it may be concluded that engine downsizing drives the increase or decrease of the turbocharging variable through the need to meet the UN SDGs. As demand for ICE vehicles increases, predominately in the global south, and authorities in these developing countries adopt more stringent emissions standards the demand for downsized engines will increase with a commensurate rise in demand for turbochargers. This demand extending to HEVs. Whilst forecasts vary most sources predict a rise in the demand for turbochargers, with a peak being reached in, or around, 2027.
• Causal loop L7 shows that in the context of emissions regulations, engine downsizing in combination with turbocharging can lead to improved fuel economy and therefore reduced tailpipe GHG emissions.
• Causal loop L8 illustrates the direct relationship between bioethanol and renewable fuel policy, where the causal relationship is reinforcing because the effect of policy will either result in an increase or decrease in bioethanol quantity. In this loop it can be seen that there is a quite straightforward causal relationship between the two variables, however, the quantity of bioethanol is not only dependent on changes to the renewable fuel policy but is also directly exposed to influences outside of the system. An example of this is the current conflict in Ukraine, which, as one of the main global corn producers has had a significant detrimental impact on global corn supplies.
• Causal loop L9 shows, in response to emission regulations, that engine downsizing in combination with turbocharging and the use of bioethanol can lead to increased engine performance/power density and reduced tailpipe GHG emissions.
• Causal loop L10 illustrates how the causal relationship between bioethanol, blends, bioethanol’s high octane (RON) qualities and suitability for use with turbocharging can lead to increased engine performance/power density.
• Causal loop L11 illustrates the positive causal relationship between the emissions regulations variable, vehicle lightweighting and reduced tailpipe GHG emissions. This is a reinforcing loop as a change in one variable will cause an increase, or decrease, in the other variables.
• Causal loop L12 shows the simple relationship between the feedstock/ conversion and renewable fuel. If there is an increase in feedstock and conversion there will be an increased quantity of renewable fuel and vice versa. As with causal loop L8, the relationship between the variables appears quite straightforward but they are arguably more vulnerable to outside influences than some other variables in the system. Feedstock supplies can be disrupted by conflict or global supply problems and conversion can be affected by land use issues and the competition between food production and liquid renewable fuel, leading to a significant impact on the system.
• Causal loop L13 expands on causal loop L12 by introducing the variables ‘renewable fuel policy’ and ‘bioethanol’, showing how a change in renewable fuel policy will have a reinforcing effect on the other variables in the causal loop.

Table 2. List of loops and their causality chains in the whole system CLD.

Loop no	Balancing/reinforcing	Causality chain
L1	Reinforcing	Engine downsizing→ vehicle lightweighting→ reduced tailpipe GHG emissions→ emissions regulations→
L2	Reinforcing	Engine downsizing→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L3	Reinforcing	Engine downsizing→ vehicle lightweighting→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L4	Reinforcing	Engine downsizing→ improved efficiency→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L5	Reinforcing	Engine downsizing→ reduced tailpipe GHG emissions→ emissions regulations→
L6	Reinforcing	Turbocharging→ enables engine downsizing→ engine downsizing→
L7	Reinforcing	Turbocharging→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→ engine downsizing→
L8	Reinforcing	Bioethanol→ renewable fuel policy
L9	Reinforcing	Bioethanol→ reduced tailpipe GHG emissions→ emissions regulations→ engine downsizing→ turbocharging→ increased engine performance/power density
L10	Reinforcing	Bioethanol→ blends→ high octane rating (RON) → use with turbocharging→ turbocharging→ increased engine performance/power density→
L11	Reinforcing	Emissions regulations→ Vehicle lightweighting→ reduced tailpipe GHG emissions→
L12	Reinforcing	Feedstock/ conversion→ renewable fuel→
L13	Reinforcing	Renewable fuel policy→ feedstock/ conversion→ renewable fuel→ bioethanol→

Table 2. List of loops and their causality chains in the whole system CLD.

Loop no	Balancing/reinforcing	Causality chain
L1	Reinforcing	Engine downsizing→ vehicle lightweighting→ reduced tailpipe GHG emissions→ emissions regulations→
L2	Reinforcing	Engine downsizing→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L3	Reinforcing	Engine downsizing→ vehicle lightweighting→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L4	Reinforcing	Engine downsizing→ improved efficiency→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→
L5	Reinforcing	Engine downsizing→ reduced tailpipe GHG emissions→ emissions regulations→
L6	Reinforcing	Turbocharging→ enables engine downsizing→ engine downsizing→
L7	Reinforcing	Turbocharging→ improved fuel economy→ reduced tailpipe GHG emissions→ emissions regulations→ engine downsizing→
L8	Reinforcing	Bioethanol→ renewable fuel policy
L9	Reinforcing	Bioethanol→ reduced tailpipe GHG emissions→ emissions regulations→ engine downsizing→ turbocharging→ increased engine performance/power density
L10	Reinforcing	Bioethanol→ blends→ high octane rating (RON) → use with turbocharging→ turbocharging→ increased engine performance/power density→
L11	Reinforcing	Emissions regulations→ Vehicle lightweighting→ reduced tailpipe GHG emissions→
L12	Reinforcing	Feedstock/ conversion→ renewable fuel→
L13	Reinforcing	Renewable fuel policy→ feedstock/ conversion→ renewable fuel→ bioethanol→

4.4.2. Relative Loop Importance

The relative importance of the strategies was considered by adopting a process similar to the degree centrality method used to determine network centrality [43]. In this method the number of connections to the turbocharging, engine downsizing, bioethanol and vehicle lightweighting reduction strategies/variables were counted to reveal which of these strategies had the most connections and therefore the most importance according to this method. From Table 3 it can be seen that turbocharging and engine downsizing were equally important and were slightly more important than vehicle lightweighting. Regarding the relative importance of the individual causal loops, none was identified as being of more importance than any of the others, rather, they all play a significant role in describing the system and demonstrating the complex relationships between turbocharging, engine downsizing, bioethanol fuel and vehicle lightweighting.

Table 3. Number of connections to each main sub-model variable.

	Turbocharging	Engine downsizing	Bioethanol	Vehicle lightweighting
Number of connections	7	7	6	5

Table 3. Number of connections to each main sub-model variable.

	Turbocharging	Engine downsizing	Bioethanol	Vehicle lightweighting
Number of connections	7	7	6	5

4.4.3. Assignment of Categories

To facilitate a deeper understanding each strategy/variable was assigned to one of four categories as shown in Table 4. Using the causality chain information from Table 2 for each causal loop it can be seen that variables from the technology category are present in each loop and that variables from the regulations category feature in all loops except loop L8. In loops L1, L6 and L10 the technology category is dominant whereas in loop L2 variables from the behaviour category dominate. In loops L3, L7 and L9 the technology and behaviour categories are of equal importance, however, in loop L4 the behaviour category is dominant. In loops L5 and L11 there is no dominant category. Loop L8 shows that the resource and regulation categories are of equal importance. Loop L12 shows that the technology and resources categories are of equal importance, whereas in loop L13 the resources category dominates.

Table 4. Assignment of sub-model and variables to categories.

Regulation	Technology	Behaviour	Resources
Emissions regulations	Turbocharging	Improved engine response	Use of waste energy
Renewable fuel policy	Vehicle lightweighting	Increased engine performance/power density	High octane fuel
	Feedstock/ conversion	Improved fuel economy	Renewable fuel
	Blends	Improved efficiency	Bioethanol
	Engine downsizing	Reduced tailpipe GHG emissions
	High octane rating (RON)	Improved vehicle dynamics
		Use with turbocharging
		Enables engine downsizing

Table 4. Assignment of sub-model and variables to categories.

Regulation	Technology	Behaviour	Resources
Emissions regulations	Turbocharging	Improved engine response	Use of waste energy
Renewable fuel policy	Vehicle lightweighting	Increased engine performance/power density	High octane fuel
	Feedstock/ conversion	Improved fuel economy	Renewable fuel
	Blends	Improved efficiency	Bioethanol
	Engine downsizing	Reduced tailpipe GHG emissions
	High octane rating (RON)	Improved vehicle dynamics
		Use with turbocharging
		Enables engine downsizing

4.4.4. Comparative Frequency

Comparative frequencies (occurrences) were derived for each of the categories listed in table 4, based on the number of times variables belonging to each category occurred in the thirteen loops shown in figure 5, the results of which are provided in table 5. It was noted that the technology category had the largest number of occurrences followed by the behaviour, regulations and resources categories. This order of the categories is expected as changes in technology would mostly result in a positive behavioural change, and, similarly, the impact of regulatory changes would result in a similar outcome. For example in causal loop L4 a change in the emissions regulations may trigger a change in engine downsizing activity that in turn improves efficiency and fuel economy leading to a reduction in tailpipe gaseous emissions. From the technology category the ‘engine downsizing’ variable was seen to be most significant as it appeared in eight of the loops. Similarly, the ‘emissions regulations’ variable from the regulations category and the ‘reduced tailpipe GHG emissions’ variable also featured in 8 loops. The behaviour category contained the largest number of variables although some of these were not included in any of the loops.

Table 5. Number of occurrences in each category.

	Number of occurrences
Technology	19
Behaviour	17
Regulation	10
Resources	6

Table 5. Number of occurrences in each category.

	Number of occurrences
Technology	19
Behaviour	17
Regulation	10
Resources	6

5. Discussion

In this work we applied a systems thinking approach and created a full CLD (Figure 5) to visualise the system of interest. To our knowledge this is a novel approach that has not previously been undertaken and therefore this work is a first step in the exploration of this complex system and brings a new dimension to aid understanding of the interactions between the chosen variables as well as offering a methodology that may be replicated with other similar systems. Through the full system CLD we are able to understand the important elements of the system and how tailpipe GHG emissions can be affected by them and how changes to variables will impact the system in general, thus, informing how the different relationships might be positively exploited. For example, the full system causal diagram shown in Figure 5, supported by the literature, reveals how engine downsizing and vehicle lightweighting can deliver improved fuel economy and reduced tailpipe GHG emissions in response to the requirements of the emissions regulations (loop L3). Unsurprisingly, in response to emissions regulations if turbocharging is coupled with engine downsizing and the use of bioethanol (loop L9) we see that an increase in engine performance and power density plus a reduction in tailpipe GHG emissions is also possible. It follows therefore that it could be advantageous to combine all or some of these variables to deliver further tailpipe GHG emissions reductions, thus bringing the benefits of reduced tailpipe GHG emissions, improved fuel economy and enhanced engine performance together.

The advantage of being able to see a visual representation of the system is that links that may not have been obvious can be identified, and links that were thought to exist and didn’t can be exposed. For example, in the CLD shown in Figure 5, it can be seen that bioethanol has a direct link to the reduction of tailpipe GHG emissions, which may not have been obvious. Indeed, from the CLD it is possible to see that bioethanol is linked to other variables and as an octane enhancer has the ability to control engine knock in turbocharged engines. Time delays have deliberately not been included in the CLD as it could be argued that there will always be some delay in cases of cause and effect. An example being the delay that may occur from the change in emissions regulations before the effect on the related variables is visible. In reality, manufacturers will make changes well in advance of dates that are mandated by regulations and the linked legislation, as is the case in the transition to HEVs where many manufacturers have plans to transition their vehicle ranges to HEV only, in advance of the date set for the cessation of the sale of new ICE powered vehicles.

In the previous sections of this paper the factual interpretations pertaining to the sub and whole system models and the individual causal loops have been presented. In addition, the relative importance of each of the main strategies/variables was considered and comparative frequencies were calculated. This being an alternative approach to the adoption of adjacency matrices as proposed by Beck, et al [44]. These sections provide valuable information and offer insight into the areas of the system where significant leverage points exist. According to Meadows [45] , ‘leverage points are places in a complex system where a small shift in one thing can produce big changes in everything’ (p.1), however, determining leverage points is not an intuitive process and rigorous analysis of the system is required.

The research presented in this paper established that the engine downsizing and regulations variables are of particular significance. This suggests that optimisation of downsized engines is a prime leverage point for intervention and exploitation, perhaps through the adoption of technologies such as mild hybrid battery assist systems, high efficiency turbochargers, improved turbocharger matching or Atkinson cycle operation, or a combination of some or all. It is clear that the emissions regulations are also a leverage point where significant change in the system can be asserted. However, whilst this might be an obvious point for intervention, care must be taken to ensure that the system changes are driven in the correct direction to achieve the desired effect. Meadows [45] states that positive (reinforcing) feedback loops ‘are sources of growth, explosion, erosion, and collapse in systems’ (p.11) and suggests that control must involve slowing down the positive feedbacks. An alternative approach arising from this research, is to seek gains through the use of combinations and consideration of the interrelationships between the loops. For example, if a regulation or policy change causes a change in one loop but this is not considered in neighbouring loops or elsewhere in the system the maximum benefit of the intervention may not be reaped. In the example given in section 4.4 the reinforcement of loop L1 would be negated if the loop L11 was not also concurrently reinforced as a downsized engine in an existing vehicle platform may potentially increase tailpipe GHG emissions. Similarly, it can be seen in loop L9 that in response to emissions regulations the tailpipe GHG emissions can be reduced by the use of bioethanol in conjunction with engine downsizing and turbocharging whilst providing increased engine performance/power density. However, the full benefit of any system change, in this case due to regulations, may not be realised if the changes are not applied to the other loops in the system. In this example if a change was produced in the vehicle lightweighting loop L3, then it is likely that the overall gains would be greater. So, in order to realise maximum benefit, intervention at one point is insufficient.

There are many constraints that influence the adoption of technologies and which may contribute to the reason some desirable changes are not made. Using the two preceding examples it could be that a vehicle lightweighting is not adopted because of considerations around manufacturing cost and the target retail price of the product. This illustrates that some of the causal loops in the system are working relatively well, e.g., engine downsizing, which has assisted the development of larger vehicles with smaller capacity ICEs, combined with lower fuel consumption and tailpipe emissions when compared to older vehicles of equivalent size, especially as vehicle mass has been increasing. However, it could be argued that this has created a rebound effect where many consumers are now choosing larger, heavier vehicles, especially sport utility models, and, as a result, the composition of the vehicle fleet is now made up of larger vehicles, rather than smaller ones, thus arguably driving the system the wrong way. There are also loops within the system that do not appear to be working as well as they might. Research by the Global Fuel Economy Initiative [46] identifies that specific fuel consumption can vary significantly depending on the technology applied. This implies that there is scope to gain improvements in the loops that have connections to the ‘improved fuel economy’ variable in the system.

This research has allowed us to create a comprehensive CLD, which has revealed the breadth and complexity of the system and which also allows the identification of system vulnerability. An example of this vulnerability can be seen in loop L12 and the relationship between feedstock/ conversion and renewable fuel. If there was a reduction in the feedstock/ conversion possibly as a result from an influence outside the system, a commensurate reduction in renewable fuel would result causing significant disruption to the system. As previously stated, this work represents a first step in understanding the system at the centre of this research and whilst the work does not offer particular solutions, important factors have been identified that could form the basis for further investigation.

6. Conclusions and Future Research

The purpose of this research was to identify the relationships that could be leveraged to provide reductions in tailpipe GHG emissions whilst considering the challenges of cost. To achieve this objective a system thinking approach was employed and a CLD was created to represent the whole system revealing the causality chains and relationships in the context of the GHG emissions reference mode. The relationships between turbocharging, engine downsizing, bioethanol fuel and vehicle lightweighting were exposed and the paths of causal interaction with policy were also explored. Whilst accepting that qualitative models have limitations due to the difficulty of being able to fully investigate the behaviour that emerges without numerical data, the research has successfully validated the hypothesis. The research has revealed the reinforcing behaviour of the individual causal loops and the pivotal influence of the engine downsizing and emissions regulations variables, whilst providing an ordered visual representation of the system. Moreover, the findings concur with those of Lewis, et al [4] who concluded that advanced downsized/boosted engines (with gasoline or ethanol) and lightweight materials provide complimentary benefits. The work has shown that some causal loops are functioning well and are strong whereas others are working less well and are relatively weak. The effect of this is to restrict achievement of the goal to accelerate reductions in tailpipe GHG emissions in the global fleet. In order to achieve this goal the weak loops will have to be strengthened.

This work has shown that a systems thinking approach can be successfully applied to represent this type of complex system and is a viable methodology that can be adopted by other researchers wishing to explore similar systems. Also, that CLDs are a powerful tool for mapping the feedback structures of such systems [17] and qualitive data has an important role to play in the pursuit of finding solutions to reduce vehicle gaseous tailpipe emissions.

Notwithstanding Sterman’s statement that causal diagrams can never be comprehensive or final, rather, they are always provisional, [17] one recommendation is made for the future direction of this research. There could be value in adopting the use of Participatory Systems Mapping (PSM) to garner the collective input from a variety of stakeholders in order to deepen understanding of the system and possibly identify additional causality chains of importance.