Environmental Systems Analysis of Meals

Transcription

1 SIK-rapport Nr Environmental Systems Analysis of Meals Model Description and Data Used for Two Different Meals Ulf Sonesson Jennifer Davis 1

2 SIK-rapport Nr Environmental Systems Analysis of Meals Model Description and Data Used for Two Different Meals Ulf Sonesson Jennifer Davis SR 735 ISBN

3 Summary This report is part of FOOD 21, a research program working within the area of sustainable food production. One part of the FOOD 21 program is environmental systems analysis, in which both the agricultural as well as the post-farm system are studied. The aim with the present study was to investigate the environmental impact and resource use for the entire food chain from farm to fork for integrated food chains, i.e. not single food products. In the first study the objectives were to compare three different ways of producing a single meal: home made, semi-finished and a ready-to-eat meal. A meatball meal was chosen for this purpose, partly because it is a common meal in Swedish diets and also because the meal is available as a semi-finished and ready-to-eat meal. Results from this first study, and discussions with stakeholders in the food chain determined the main direction in the second study. Here, two different meals based on chicken were chosen for analysis: one home made and one semifinished meal. The objectives were to explore how different improvement measures in the meals supply chain would affect the environmental impact of each meal. Overall, the objective of the project was also to develop a simulation model suitable for analyses of this type. This report gives detailed information on the model and data used in the two studies. The results from the studies and a more thorough background, including references to other literature, as well as analyses and discussions are presented in Sonesson et al. (2005a), and an impending article (Davis & Sonesson, 2005). Results from both case studies show that raw material utilisation in the post-farm system is very important. Since the largest environmental impact occurs early in the meal s life cycle, i.e. in agriculture, any food wasted in industry and households means that significant environmental burden has taken place to no use. By reducing wastage after the farm, the impact from the farm is also reduced, as less material from the farm is then needed to produce the meal. The chicken study shows that small measures in terms of reducing wastage do have an impact, so the potential for improvements is large. The holistic approach used in the simulation model is important as it allows for analysis of the impact of changes to the system. When researching possible ways to sustainable production and consumption of food, it is crucial to have a system s perspective to ensure that improvements in one part of the system do not lead to negative impacts in other parts. Furthermore, the approach also enables you to pinpoint where in the system improvement efforts are best focused. 3

4 Table of Contents Summary... 3 Introduction... 5 Aim and Objectives... 5 Structure of the Presentation of the Model and Input Data... 6 Model description... 6 General description... 6 General modelling approach... 9 Description of the models in the foreground system Description of the models in the background system Result Management General data used in the models Scenarios studied Scenarios studied in the meatball meal study Scenarios studied in the chicken meal study Data used in the meatball meal study Data used in the chicken meal study Results Results for the meatball meal case study Results for the chicken meal case study Discussion and conclusions Discussion Main conclusions Recommendations for future work References

5 Introduction This report is part of FOOD 21, a research program working within the area of sustainable food production. The overall long term goal of the FOOD 21 program is to define optimal conditions for sustainable food production that generate high quality food products (FOOD 21, 2004). One part of the FOOD 21 program is environmental systems analysis, in which both the agricultural as well as the post-farm system are studied. The overarching aim with the systems analysis projects is to analyse the environmental impact of the food chain with a holistic approach, using research results from other parts of the FOOD 21 program. The focus of the present study is how different processing in the post-farm system affects the overall environmental impact of the food chain (including impacts from the farm). The working process has been to first analyse three different ways to produce a single meal (a meatball meal). The results from this study were then disseminated through meetings with stakeholders in the supply chain: food industries (Findus, Procordia/Orcla and Swedish Meats), a representative for the retail sector (COOP), and a consumer organisation (Konsumentföreningen Stockholm). This was with the aim to distribute the results, but also to discuss the next steps in the study: which areas did they pinpoint as relevant to analyse? With these discussions as a starting point, the aims and scenarios of the second study were determined. This procedure was beneficial as we received better understanding of the needs of stakeholders in the food chain, but will also be valuable in future research projects (the discussions with the stakeholders resulted in many interesting research questions, more than could be covered within this project). This report gives detailed information on the model and data used in the two studies. The results from the studies and a more thorough background, including references to other literature, as well as analyses and discussions are presented in Sonesson et al. (2005a), and an impending article (Davis & Sonesson, 2005). We would like to express our thanks to Anna Flysjö, SIK, for contributing to the data collection, and other colleagues at SIK for discussions on food processing technology. The contact persons from companies and organisations are also greatly acknowledged for their help and information. Aim and Objectives The aim with the present study was to investigate the environmental impact and resource use for the entire food chain from farm to fork for integrated food chains, i.e. not single food products. In the first study the objectives were to compare three different ways of producing a single meal: home made, semi-finished and a ready-to-eat meal. A meatball meal was chosen for this purpose, partly because it is a common meal in Swedish diets and also because the meal is available as a semi-finished and ready-to-eat meal. Results from this study, and discussions with stakeholders in the food chain determined the main direction in the second study. Here, two different meals based on chicken were chosen for analysis: one home made and one semi-finished meal. The motivation for exploring chicken based meals was partly to give the study another dimension (otherwise we could have continued to explore the meatball meals further), and partly because consumption of chicken has increased significantly in the past ten years in Sweden. The objectives were to explore how different improvement measures in the meals supply chain would affect the environmental impact of each meal. The scenarios of each study are described further in the Scenario chapter (page 51). 5

6 Overall, the objective of the project was also to develop a simulation model suitable for analyses of this type. The aim of this report is to present all data used and the model employed for performing the analysis. This is very important since it is crucial to supply all information used for large systems analyses like this, in order to make the study credible. Structure of the Presentation of the Model and Input Data The amount of data used in the presented study is large; hence the data presentation is structured according to how data is used and where in the study and model. 1. The model is described so the reader can understand how it uses the input data to calculate the emissions and resource use. 2. Input data that are general for all scenarios in both studies are presented for each part of the model, as energy transports, cooking etc. 3. The three scenarios for the meatball meal study are described in terms of how the food chain is designed and what assumptions have been made. 4. Input data that are specific to each meatball meal scenario is described 5. The four scenarios for the chicken meal study are described in terms of how the food chain is designed and what assumptions have been made. 6. Input data that are specific to each chicken meal scenario is described 7. Results for the two studies are presented Model description General description The model developed is basically a material and substance flow model, where energy flows are also accounted for. The simulation model SAFT (Systems Analysis of Food processing and Transport) is constructed to facilitate simulation experiments for different food supply chains from farm gate to consumption. The results are consumption of resources and emissions from the system. Methods from Life Cycle Assessment (LCA) are used regarding characterisation of emissions and resource use (for more information about LCA see e.g. Lindfors et al., 1995 and CEN, 1997). The systems perspective is crucial when defining systems boundaries, also in this respect the modelling resembles LCA. Using LCA vocabulary, the function of the system is to transport and process food raw material from farms to the point of consumption. In order to take into account all effects of the food flow both upstream systems (e.g. agriculture, energy production, packaging production) and downstream systems (e.g. waste management) are included. In Figure 1 the principal system boundaries are described. Within the core system the models are more detailed and facilitate simulations if different technologies and organisations. For the systems in the background system, static LCA input data is used to calculate energy use and emissions per kg of input needed or outflow treated. 6

7 Background system Core system T T T Agriculture Industry Retail Consumption Fertiliser production Energy System Packaging production Residue and waste treatment Water production Sewage treatment Figure 1. Principal system boundaries for the SAFT model, T denotes transports. Since one important function of food (however not the only) is to supply the human population with nutrients, it is important to be able to track the main nutrients through the system. Different system will be more or less efficient in this respect. Moreover, unwanted substances in foods, e.g. cadmium, can also be of interest to trace. This makes a material/substance flow modelling (MFA/SFA) approach appropriate. Conclusively, the SAFT model is an MFA/SFA model that uses LCA methodology to evaluate both the sustainability issues relating to resource use and emissions as well as mapping the substance flows through the system. The SAFT model has a modular approach, each sub system is modelled individually and connected to the rest of the model by its in- and outflow of energy and materials. The energyand material flows are described by vectors (Table 3, Table 4 and Table 5). This means that new processes can be included in the existing model if new systems are studied In Figure 2 the top level of the SAFT model is depicted, arrows indicate material- and energy flows. Each box is further divided in sub-systems, often on several levels of detail, where the connections and flows are described graphically. 7

8 Figure 2. Top level of the SAFT model. Arrows indicate material- or energy flows, note that the direction of energy flow arrows are opposite the actual flow, since it is information of amount and type of energy needed in other parts of the models that is transferred. 8

9 General modelling approach The model is intended to be modular, i.e. consist of independent sub-models that may be combined in many ways. The model is constructed in the software MATLAB/Simulink. All individual sub-models are connected with a data file (m-file in MATLAB) where all data for that process is inserted. This approach facilitates easy documentation of data used in a study. This means that the SIMULINK model describes the structure of the models and the principal causal connections between in- and outflows, and the m-files contain all data used. All submodels are listed in Table 1 along with the corresponding init-file. These individual sub-models can be used solitarily, but can also be used for building Clustered models. Such clustered models are used to organise product flow, and individual sub-models are parts where energy consumption, emissions etc are calculated. These clustered models are constructed for each study performed, depending on what food that are studied and how the flows are organised in the system. The clustered models used in this set-up of SAFT are shown in Table 2. 9

10 Table 1. List of all individual sub-models in SAFT and their corresponding init-file Sub-model Models used in both studies Truck and trailer, heavy truck, light truck, Pickup, Car Retail Private households Cooking Residue treatment Energy Production Packaging production Air emissions Water emissions Drinking milk dairy Mill Bakery Models used in the Meatball meal study Agricultural production (milk, cattle, pigs, wheat, potato, carrot) Abattoir Carrot Packer Potato Packer Meat ball production Mashed potato production Ready-to-eat meal production Models used in the chicken meal scenarios Agricultural production (milk, chicken, wheat, potato, onion, lettuce, carrot, apple) Chicken abattoir Hash manufacture Carrot Packer Potato Packer Lettuce packer Apple packer Corresponding init-file TransportInit RetailInit HhInit CoInint ResidueInit EnergyInit PackInit Weighingfactors Weighingfactors DairyInit MillInit MillInit AbbatiorInit CarrotInit PotatoInit AbbatiorInit PotatoInit ReadyInit FarmInit AbattoirInit ReadyInit CarrotInit PotatoInit LettuceInit AppleInit 10

11 Table 2. Clustered sub-models used in the study. Clustered model Distribution Retail and home transport Households Individual sub-model included Heavy Truck Truck and Trailer Retail Car Pick-up Cooking Storing Init file for data in the clustered model (except the ones used for the included individual sub-model) HhInit, RetailInit - CoInit HhInit SAFT is basically a material flow model, i.e. a certain amount of food raw material enters the model at one end, the product flows through the system and all use of energy and emissions that occur as a result of the flow is calculated. Finally, consumed products leave the system at the other end. The time scale is yearly averages due to the purposes of the model, to study future systems overall environmental performance. Throughout the SAFT model vectors are used to describe all physical flows between sub-models, as material and energy. There are three different vectors; 1. The material flow vector, which describes the flow of material and has 60 single positions (Table 3) 2. The second one describes the use of net energy and has 45 positions (Table 4) 3. Primary energy carriers, this vector describes the consumption of primary energy carriers (e.g. amount of oil and coal extracted from the earth s crust and amount of biofuel used), and has 15 positions (Table 5). Obviously, all positions in the vectors are not relevant for all flows, often there are only a few positions used for a certain flow. The fact that both single substances and materials are included means that both chemical composition and type of material for e.g. packaging can be described by one vector, which simplifies the modelling work. The rationale to use such extensive vectors is that it offers flexibility regarding what questions that can be analysed with the use of the SAFT model. 11

12 Table 3. The vector that describes all physical flows between sub-models in the SAFT model Position Substance etc. Position Substance etc. 1 C-tot 31 Pb 2 C- slow (lignin, humus) 32 Cd 3 C-medium (cellulose, hemicellulose) 33 Hg 4 C-Fast (sugar, starch) 34 Cu 5 C-Fat 35 Cr 6 C-Protein 36 Ni 7 COD 37 Zn 8 VS (DM-Ash) 38 Particles/susp. 9 DM 39 Volume, packaging included 10 CO 2 -fossil origin 40 Plastic, HDPE (Primary package) 11 CO 2 -biological origin 41 Plastic, LDPE (Primary package) 12 CO 42 Plastic, PP (Primary package) 13 CH 4 43 Plastic PS (Primary package) 14 VOC (volatile hydrocarbons) 44 Plastic PET (Primary package) 15 PAH 45 Cardboard (Primary package) 16 Phenols 46 Laminate PE/Cardb. (Primary package) 17 PCB s 47 Laminate PE/Cardb./al (Primary package) 18 Dioxins 48 Aluminium (Primary package) 19 H 2 O 49 Glass (Primary package) 20 N-tot 50 Plastic HDPE (Secondary package) 21 + N-NH 3 /NH 4 51 Plastic LDPE (Secondary package) 22 N-NO X 52 Plastic PP (Secondary package) 23 - N-NO 3 53 Corr. cardboard (Secondary package) 24 N-N 2 O 54 Wood (Secondary package) 25 S-tot 55 Steel (Secondary package) 26 S-SO X 56 Aluminium (Secondary package) 27 P-tot 57 Sewage 28 Cl-tot 58 Solid org. Waste 29 K-tot 59 Burnable solid waste 30 Ca-tot 60 Other waste 12

13 Table 4. The vector that describes the energy flows in SAFT Position Content Comment 1 District heating, mix 1 Use of heat 2 District heating, mix 2 d:o 3 District heating,mix 3 d:o 4 Fossil oil d:o 5 Coal d:o 6 Fossil gas d:o 7 Solid biofuel d:o 8 Electricity d:o 9 Biogas d:o 10 Heat pump d:o 11 Extra d:o 12 Extra d:o 13 Extra d:o 14 Extra d:o 15 Extra d:o 16 Base load, grid 1 (Swedish average) Electricity 17 Base margin load, grid 1 d:o 18 Top load grid 1 d:o 19 Base load, grid 2 (EU-average) d:o 20 Base margin load, grid 2 d:o 21 Top load grid 2 d:o 22 Base load, grid 3 (OECD average) d:o 23 Base margin load, grid 3 d:o 24 Top load grid 3 d:o 25 Extra d:o 26 Extra d:o 27 Extra d:o 28 Extra d:o 29 Extra d:o 30 Extra d:o 31 Diesel oil Fuel used locally 32 Petrol d:o 33 Fossil gas d:o 34 Biogas d:o 35 RME (Rape methyl esther) d:o 36 Methanol d:o 37 Ethanol d:o 38 Base load, grid 1 (Swedish average) d:o 39 Base margin load, grid 1 d:o 40 Top load grid 1 d:o 41 Extra d:o 42 Extra d:o 43 Extra d:o 44 Extra d:o 45 Extra d:o 13

14 Table 5. The vector that describes the use of primary energy carriers, it is only used within the sub-model Energy Production Position Energy carrier Comment 1 Oil 2 Coal 3 Fossil gas 4 Biofuel 5 Biogas 6 geotermic energy 7 Uranium 8 Hydropower In this study amount of MJ hydropower is as used primary energy carrier 9 Extra 10 Extra 11 Extra 12 Extra 13 Extra 14 Extra 15 Extra Description of the models in the foreground system Transports All transport models have the same basic structure, first the fuel consumption is calculated as a function of amount of goods, load size and distances, thereafter the emissions are calculated as a function of amount and type of fuel. It is possible to adjust the energy consumption for certain traffic situations or transport modes, as freeze transports. All transport models are connected to the m-file transportinit.m where data are set. Truck and Trailer In Figure 3 the truck and trailer sub-model is shown. Below a description of the model follows, step by step, mainly from left to right. Shaded blocks in Figure 3 are input data set in transportinit.m : At product the product to be transported enters the model as the 60-position vector. Max load, m 3 gives maximum load volume on the vehicle, this together with data on total volume of the load (calculated in Volume, m 3 ) gives number of loads if volume is limiting. Max load, tonnes gives maximum load weight on the vehicle, which together with total weight of the load (calculated in weight ton ) gives number of loads if weight is limiting. At N:o loads the number of loads are compared and the highest is chosen for further calculations. Number of loads ( n:o loads ) and distance per load ( Distance driven per load ) gives the total distance driven per year for the goods. At Average load, ton/load the average load in kg is calculated. This figure is used to create the quotient between maximum load and real average load, this is done in average load/max load. 14

15 Data on fuel consumption per km at maximum load ( Fuel cons. full load ) and empty vehicle ( Fuel cons. empty ) are set in transportinit. At working points in between, the fuel consumption is assumed to change linearly. The average load as part of maximum load together with this interpolated fuel consumption results in fuel consumption per km for the actual transport. The total distance driven ( Total distance km/year ) and the average fuel consumption ( Aver. Fuel cons. ) together with two correction factors ( Allocation factor, Traffic factor ) gives the total fuel consumption per year ( MJ fuel used ). The Allocation factor is a dimensionless unit that is used if the transported goods is co-transported with some other goods and only a part of the transport should be allocated to the studied goods (if this is not the case, the Allocation factor is one). Traffic factor is a figure that facilitates compensation for extraordinary traffic situations, as distribution transports with a lot of stops or driving mainly in city areas where traffic congestion occur frequently. If the transport is a refrigerated or freeze transport this can also be accounted for using this parameter. It is possible to use seven different fuels and combinations thereof (diesel, petrol, fossil gas, methanol, Rape methyl ester (RME), ethanol, biogas). In transportinit a figure 0-1 (part of fuel mix) for each fuel type is set and the sum for all fuels must equal 1. It is not an actual fuel mix that is given, but parts of the truck fleet that are fuelled with the different fuels. The amount of fuel (in MJ) calculated in MJ fuel used is multiplied with this vector describing the fuel mix and the resulting amount of each fuel calculated in MJ of each fuel type is delivered to the surrounding system via the outport Energy. The vector of amount of fuels used is also used to calculate the emissions. In Air emission1 the amount of diesel used (from the vector described above) is multiplied with the emission vector emissions for diesel trucks. The same operation is performed for the other fuels, and the sum of all emissions is delivered to the surrounding system via the outport Air emissions. The product that is transported is not affected by the transport and leaves the transport model via outport Product out Data on where the transports are performed ( Part of transport in urban areas and Part of transport in rural areas ) together with Total distance km/year gives distance driven in urban and rural areas ( vehicle*km in urban areas and vehicle*km in rural areas ). This information may be used for a simple assessment of the impacts of transport that is not covered by energy consumption and emissions. Heavy Truck, Light Truck These models have identical structures as Truck and Trailer, it is only data on fuel consumption, maximum load volume and maximum load weight that differs. 15

16 TransAirEmTruckTrailer(:,1) em i ssi o ns for u(31) diesel trucks Air emissions1 MJ diesel used TransAirEmTruckTrailer(:,2) em i ssi on s for u(32) petrol trucks MJ petrol Air emissions2 used u(33) MJ fossil gas TransDistPerLoadTruckTrailer1 used Air emissions3 T ransairem T ruckt railer(:,3) Distance driven per load TransAllocationFactorTruckTrailer1 emissions for Allocation factor fossil gas trucks u(34) TransMaxVolumeTruckTrailer1 MJ biogas T ranst rafficfactt ruckt railer1 used Air emissions4 Max load, m3 T ransairem T ruckt railer(:,4) Traffic factor emissions for u(35) u(44)/1000 biogas trucks MJ RME Product1 Volume, m3 used Air emissions5 TransAirEmTruckTrailer(:,5) 1 MATLAB em i ssi on s for Mux Function RME trucks u(36) Product Total distance MJ fuel used Mux N:o loads km/year MJ methanol used Air emissions6 f(u) TransAirEmTruckTrailer(:,6) em i ssi on s for weight ton u(37) Product2 methanol trucks MJ ethanol used Air emissions7 TransAirEmTruckTrailer(:,7) TransMaxLoadTruckTrailer em i ssi o ns for ethanol trucks Max load, tonnes TransLoadFactorTruckTrailer1 MJ of each Load factor (0-1) fuel type Average load, Type of TransFuelUsedTruckTrailer1 ton/load fuel used average load/ max load TranskmUrbanTruckTrailer1 T ran sm JPe rkm Ful l T ruckt rai l er vehicle*km in urban areas Fuel cons. full load TranskmRuralTruckTrailer1 Aver. fuel cons. TransMJPerkmEmptyTruckTrailer vehicle*km in rural areas Fuel cons. empty TransPartUrbanTruckTrailer1 Part of transport in urban areas TransPartRuralTruckTrailer1 Part of transport in rural areas Figure 3. The Truck and trailer sub-model. Shaded blocks are indata set in transportinit.m 16 Sum1 1 Product out 2 Air emissions 3 Energy

17 Industries Drinking milk dairy In this model energy consumption, local air emissions, use of water, wastage and waste generation is calculated as a static function of the flow of milk to the dairy. By changing input data in the init-file, different fuels for local heat production can be changed. The inflow to the Drinking milk dairy is used to calculate consumption of water, energy use and direct emissions as a function of amount of milk processed. The amount of milk wasted and the proportion of the wasted milk going to sewage and feed respectively, is also calculated using percentages. Amount of package used, both primary and secondary, is calculated as a function of amount of milk delivered from the dairy. The amount of different package material is sent to the packaging production model where use of energy and emissions are calculated. The package material is also added to the vector that describes the product flow. Abattoir The abattoir model consists of three parallel sub-models; cattle, pig and poultry slaughter. They have the same structure but the input data differs. The inflow of animals, described with the 60 position vector, is used to calculate the use of water and energy and also direct emissions for the process. The total inflow is then partitioned by percentages to residual products, waste and meat. The residual products ( slaughterhouse waste ) are sent to the model Residues treatment, the waste is sent to the Waste management sub-model. The amount of package used is calculated as a function of the total amount of meat leaving the abattoir. Mill The mill model uses the incoming vector describing the grain, to calculated wastage, energyand water using static parameters, based on performance per unit grain milled. The percentage of incoming grain that is sold as flour and bran is also calculated. The amount of packaging material used and waste generated is calculated as a function of the amount of flour produced. Potato Packer The model for Potato packer is constructed the same way as the mill model, the difference is the input data used in the data files. Carrot Packer The model for Carrot packer is constructed the same way as the mill model, the difference is the input data used in the data files. Bakery The model for Carrot packer is constructed the same way as the mill model, the difference is the input data used in the data files. Meat ball production The model for Carrot packer is constructed the same way as the mill model, the difference is the input data used in the data files. 17

18 Mashed potato production The model for Carrot packer is constructed the same way as the mill model, the difference is the input data used in the data files. Ready-to-eat meal production The model for Carrot packer is constructed the same way as the mill model, the difference is the input data used in the data files. Chicken hash production The model for chicken hash production is constructed the same way as the mill model, the difference is the input data used in the data files. Distribution The sub model for distribution is a combination of some other sub-models that partly use the same data, hence together they make up an individual sub model. The reason for modelling distribution in this way is that the mode of transport is coupled to where the food is to be delivered. The structure of this part of the food chain is also different depending on what food that is studied. In this setup of the SAFT- model there are different distribution systems for the three different receivers of food, External retailers, Neighbourhood retailers and e- shopping (se the respective heading for a detailed explanation of the individual sub-models). This leads to the structure described in Figure 4 and Figure 5, where the former describes the top level of the distribution model, and the latter distribution of each product. In Figure 5 the distribution model for dairy is shown, the corresponding sub models for other products are identical in structure. The inflow is divided according to how large part of the product that is sold in the different retail types (external, neighbourhood and e-shopping). For each transport the fuel consumption and emissions are calculated as described above under heading Transports. The product flow is not affected by the transport but is passed on to the retail sub model. In the examples in the figure, the distribution to neighbourhood retail is done with a heavy truck and to the other two retail types with truck and trailer. This can be changed, and also combinations of different transport modes can be modelled by inserting other transport models or combination of transport models. 18

19 Figure 4. The clustered model Distribution 19

20 Figure 5. The model for distribution of dairy products to retailers Retail and Home transport Background The retail- and home transport systems are so closely connected so they are put together to form a clustered model. It seems likely that one type of retailing structure is connected with certain types of home transport. For example the part of the customers that use car are lower in small, local shops than in large, more externally situated (Sonesson et al., 2005b). Moreover, e-shopping is naturally followed by some kind of delivery, either to homes or some spreading points. However, the retail and home transport parts are mainly modelled in a structure that separates them as much as possible. The clustered model consists of two main parts, Retail and Home transport. The connection between them is that the flow of food that enters Retail 1 automatically continues to Car 1. This means that everything bought in Retail 1 is transported home by Car 1 (except the part that is transported home by walking etc., which is omitted from the study, see below). This structure is chosen since both transport distances and shopping frequency are assumed to be depending on type of shop, as described above. Retail There are three types of retailers included in the model, the dividend between them are where they are situated. 1. External retail, this type is a large shop situated away from dwelling areas and relies on the customers having cars, only a small part of the customers use public transports. 20

21 2. Neighbourhood retail, this type is situated close to dwelling areas, the home transport may be performed either by car or walking, biking and public transport 3. E-shopping, this retailing type is characterised by the fact that the food is not collected by the customer at all, it is delivered by the retailer. How this delivery is performed may differ, that is decided in the transport models following the retail model. This model structure facilitates differences in energy use etc. between different stores. In reality External retailers are probably more energy efficient than Neighbourhood retail since the former are larger. Since we have not found any reliable data on the differences, both these models of retail use the same data. The differences are probably small considering the total impact from a products entire life cycle (Carlsson & Sonesson, 2000). The inflow of each food under study is divided in parts that are sold in each type of retail. This is done with parameters set in RetailInit and states between 0-1 for each type, where the sum must equal 1. This partitioning is done in the Distribution block Before the flows of foods are entered to the three retail models, they are sorted in types, in block called Sorting of products in types. This is a block that must be tailored for each product flow, since there is no information in the vector that can be used for automatic sorting. The foods are put in one of six type foods (types): 1. Frozen, long storage 2. Frozen, short storage 3. Cold, long storage 4. Cold, short storage 5. Normal, long storage 6. Normal, short storage The reasons for this are presented below. In the next step the food enters the retail sub-models, the model for External Retail is presented in Figure 6. In this example it is shown that there are six flows of food through the model. The energy consumption is mainly depending on whether the food is frozen, cold or kept in room temperature and the retention time in the shelf. To manage this, the incoming flow of food must be connected to the right inport ( frozen, cold or room temp ). In Figure 6 the model for External retail2 is presented. The flow from each inport is then further divided in two flows, long and short retention time, the partitioning is done by 0-1 parameters set in retailinit. Thereafter the energy consumption and amount of bags are calculated for each of the six alternatives ( frozen/long, frozen/short, cold/long, cold/short. Room temp/long and Room temp/short ). Finally all energy consumption are summed at Total energy and delivered to the surrounding system as are the amount of bags used and the food flow, now completed with bags. In this way all foods studied can be handled with the same model and still take differences between types of food into account. The model is a simplification, but according to Carlsson & Sonesson (2000) the main differences are included using this structure. The alternative would be to set average retention time for each food, but that is data that can not be attached to the flow of food, since it differs between retail types. Then one has to have a specific model for each type of food, but that bring about more work that can be justified from the importance of retailing shown in Carlsson & Sonesson (2000). 21

22 Product out 1 Frozen products long shelf time 2 Frozen products short shelf time 3 Cold products long shelf time Energy turnov er out Frozen products in To waste management Added packages Waste to other use Frozen products, long shelf time Product out Energy turnov er out Frozen products in To waste management Added packages Waste to other use Frozen products, short shelf time Product out Energy turnov er out Frozen products in To waste management Added packages Waste to other use Cold products, long shelf time Mux Mux 2 Energy turnover 1 Food out, 6 vectors 4 To waste man. 4 Cold products short shelf time Product out Energy turnov er out Frozen products in To waste management Added packages Waste to other use Cold products, short shelf time 3 Added packages (to pack.prod.) Product out 5 Other products long shelf time 6 Other products short shelf time Energy turnov er out Frozen products in To waste management Added packages Waste to other use Other products, long shelf time Product out Energy turnov er out Frozen products in To waste management Added packages Waste to other use Other products, short shelf time 5 Wastege returned to supplier Figure 6. The model for External retail The next level of the retail models is shown in Figure 7. The food enters the sub-model at Product in. At MJ used the energy consumption is calculated as a function of kg food passing the model), as well as amount of bags (paper and/or plastic). The bag material is both delivered to the Packaging production model for calculations of manufacturing the material ( Packaging material ) and added to the product ( Add bags ). The latter is to reflect what actual happens; bags are brought home. The food together with bags leaves the model at Product out. 22

23 Retail1ReturnColdShortTime Return kg 5 Waste to other use 1 Frozen products in Sum 1 Product out u(9)+u(19) Fcn Retail1EnergyColdShortTime MJ 2 Energy turnover out Retail1PackageAdded Package added 4 Added packages Mux Demu Ground2 Mux1 Demux Ground1 Terminator Retail1WastageColdShortTime Waste, kg u(9)+u(19) waste kg Ground3 Ground Mux Mux 3 To waste management Figure 7. The second level of the External retail model Home transport, car The model for transport by car is somewhat different in structure than truck transports. The difference is that the fuel consumption is not at all correlated to the amount of goods transported. The rationale for this is that the effect of kg goods on fuel consumption is negligible, transporting 10 or 40 kg in a vehicle weighing 1000 kg uses practically the same volume of fuel. After the fuel consumption is calculated the model is similar to Truck and Trailer The average distance, shopping frequency and the number of cars used in the area under study is used to calculate the total distance driven. Thereafter the fuel consumption is calculated by data on MJ/km. In Figure 8 the structure of the Car model is presented. No of Cars is calculated as a function of Total number of Households, Part of Customers in retail1 that use car and Part of households that shop in Retail 1. No of Cars is the number of households that use car for the specific transport. The total distance driven is calculated as a function of No of Cars, Shopping frequency Distance per Trip and Weeks per year. The resulting amount of kilometres is multiplied with Allocation factor and Traffic factor (se heading Truck and Trailer) and Fuel Cons.. This results in MJ fuel used. After that point the model for Car is similar to the Truck and Trailer model. At the inport product in the goods is entering the model. The food is not affected by the transport and leaves the model by the outport Product out. The flow of food through the 23

24 model is used to detect whether the model is used at all. In weight kg the weight is calculated. This figure is then compared to a zero in Zerocheck1 and Zerocheck2. If the flow is zero the results from the model will be vectors of zeros both for Energy and Air emissions. This check is necessary since the Car sub-model is included in the model block Retail and home transport and depending on the choices for structure of retail one of the two Car models may not be used. Since the energy consumption is calculated independently of the product flow the model will generate energy consumption and emissions even if no products are transported. Home transport, delivery with pick-up This model is used for e-shopping followed by home delivery of groceries. The model is almost similar to the Car model; the only difference is that the total distance driven is calculated in a somewhat different way (Figure 9). Distance per Trip is the number of kilometres driven each delivery round, Shopping frequency is number of deliveries per week and Number of Hh per round. These parameters are used to calculate the total driving distance. 24

25 1 Product in u(8)+u(22) weight kg TransAirEmCar(:,1) emissions for u(31) diesel cars MJ diesel Air emissions6 used TransAirEmCar(:,2) emissions for u(32) petrol cars MJ petrol Air emissions13 used u(33) TransAllocationFactorCar1 TransAirEmCar(:,3) MJ fossil gas Allocation factor used Air emissions7 emissions for fossil gas cars TransTrafficFactorCar1 Traffic factor u(34) MJ biogas used Air emissions10 52 TransAirEmCar(:,4) Sum1 Weeks per year emissions for u(35) biogas cars MJ RME TransDistPerLoadCar1 used Air emissions11 TransAirEmCar(:,5) Distance per trip emissions for RME cars u(36) TransTripsPerWeekCar1 MJ methanol Shopping frequency Total distance TransAirEmCar(:,6) used Air emissions9 km/year and car MJ fuel used emissions for No of methanol cars u(37) cars MJ ethanol Retail1PartCar used Air emissions8 TransAirEmCar(:,7) Part of Customers in retail1 that use car emissions for ethanol cars RetailPartLargeMilk Part of households that shop in retail 1 TransNumberOfHouseholds Total number of households MJ of each TransMJPerkmCar fuel type Type of Fuel cons. TransFuelUsedCar1 fuel used TransPartUrbanCar1 Part of transport in vehicle*km urban areas in urban areas 0 Constant2 TransPartRuralCar1 Part of transport in vehicle*km rural areas in rural areas Figure 8. The Car modell. Shaded blocks are indata set in transportinit.m 25 Switch2 Switch3 1 Product out Switch zeros(si ze(1:60)) Switch1 zeros(si ze(1:45)) TranskmUrbanCar1 TranskmRuralCar1 3 Air emissions 2 Energy

26 1 Product in u(8)+u(22) TransAirEmPickup(:,1) weight kg em issions for u(31) diesel pick-ups MJ diesel Air emissions6 used TransAirEmPickup(:,2) emissions for u(32) petrol pick-ups MJ petrol Air emissions13 used TransAllocationFactorPickup1 u(33) TransAirEmPickup(:,3) Allocation factor MJ fossil gas emissions for used Air emissions7 fossil gas pick-ups TransTrafficFactorPickup1 Traffic factor u(34) TransAirEmPickup(:,4) 0 MJ biogas em issions for used Air emissions10 52 Display1 biogas pick-ups Sum1 Weeks per year u(35) MJ RME TransDistPerLoadPickup1 TransAirEmPickup(:,5) used Air emissions11 Distance per trip em issions for RME pick-ups u(36) TransDeliveryFreqPickup1 MJ methanol Shopping frequency TransAirEmPickup(:,6) used Air emissions9 TransHhPerTripPickup1 Total distance MJ fuel used em issions for km /yea r methanol pick-ups Number of Hh per round u(37) MJ ethanol used Air emissions8 No of households TransAirEmPickup(:,7) RetailPartHomeDelivMilk em issions for ethanol pick-ups Part of product that is sold by home delivery TransNumberOfHouseholds MJ of each number of households fuel type TransFuelUsedPickup1 Type of fuel used TransMJPerkmPickup Fuel cons. pickup TransPartUrbanPickup1 vehicle*km Part of transport in in urban areas urban areas 0 TransPartRuralPickup1 vehicle*km Constant2 Part of transport in in rural areas rural areas Figure 9. The Pickup modell. Shaded blocks are indata set in transportinit.m 26 Switch2 Switch3 1 Product out Switch ze ro s(si ze(1 :60 )) Switch1 ze ro s(si ze(1 :45 )) T ra n skm Urb a n P i cku p 1 T ra n skm Rura l P i cku p 1 3 Air emissions 2 Energy

27 Households The model for household consists of three parts: storing, sorting/preparing and cooking. They are connected in that sequence, and the model following requires input from the former. Storing The storing model receives input from the retail model, sorted according to storage condition; freezer, refrigerator or room temperature. For freezers and refrigerators the energy use is calculated as a function of storage time and volume of the stored item, and there are different models for different equipment, both types and sizes. For a detailed description of the storage models see Sonesson et al. (2003) The wastage during storing is also set, specific for each product stored. In the storing model the packaging included, both primary food packaging and bags from the grocer s in the flow is identified and sent to the waste management model Sorting/preparing before cooking This model facilitates the calculation of inputs and changes in the raw material before cooking. In the study presented it was mixing water and flour to produce dough, peeling of potatoes and carrots, mixing water and mashed potato powder and mixing minced meat to make meatballs. The amount of food wasted in these processes, as when peeling, is calculated and delivered to the waste management model. Cooking, including dishwashing The cooking sub model consists of four technologies for domestic cooking: Boiling in water, frying in pan, baking in oven and microwave oven. These models are thoroughly described in Sonesson et al (2003). In short the input data includes preparation time, physical data on foods, batch size, type of equipment, amount of water evaporated and specific energy use for the different modes of cooking. Description of the models in the background system Residues treatment The model for residue treatment uses the inflow from the different industries to calculate the direct emissions and energy- and water used to process the residual products. In the case that the residual product can be used, for example whey as feed, it also calculates the amount of alternative products saved by the use of the residual product. The environmental impact and resource use for that alternative product is subtracted from the direct emissions and resource use resulting from the processing. LCA data for alternative products are used as input data for this latter part. This sub model is specific for each type of residual product and alternative product; it is not a general model that can be used in any system. Energy system Background In the SAFT model all sub-models (e.g. transports, dairy) calculates the amount of energy needed for the flow of food through the model. Three forms of energy is used, Heat, Electricity and Fuel. Heat means processes etc. that use heat but does not burn fuel locally to produce it, an example is district heating used. Electricity is of course use of electricity bought from the grid. With Fuel we mean energy that is produced locally, i.e. fuel for trucks or gas used for drying in an industry. The amount of energy used in each submodel is delivered to the Energy Production sub-model. In the model we have tried to mirror the reality regarding where emissions are calculated, so emissions from production of 27

28 Heat is calculated in Energy Production as are emissions from production of Electricity. The amount of energy used in each sub-model is delivered to the Energy Production sub-model. Emissions from Fuel on the other hand, are calculated where it is combusted, i.e. in the sub-model itself. The amount of fuel is however also delivered to Energy System where calculations of emissions for producing and distributing the fuel are calculated ( pre-combustion ). In Figure 12 a principal description of how energy is managed in the SAFT model is shown. Model description This sub-model calculates the consumption of primary energy carriers (PEC) and emissions from the energy system. The inflow to the model is information on amount of energy needed in all other sub-models, in vector form (see Table 4). In cases where some part of the system under study delivers a negative energy consumption (e.g. Waste Management if packaging is incinerated) it calculates saved PEC and emissions. In Figure 10 the top level of the model is presented. All vectors entering the model are added to a vector describing the whole systems net energy turnover ( sum ). This vector is divided ( Demux ) in three 15-vectors, one for each energy type (see Table 4). These 15-vectors are indata for one sub-model each, where emissions and use of PEC are calculated. In Figure 11 an example of this level is presented, it is the sub-model for Heat production that is shown. The inflow, the 15-vector, is divided in scalars, each representing the amount of one type of energy used. The first position in the vector is district heating, and since district heat is produced with a mix of fuels that might change, an extra step is needed. In the function District heating the amount of MJ used is multiplied with a 15-vector describing the mix, thus producing a 15-vector with the same content as the one that entered the model (except for the first position, district heating, which is empty). Thereafter the amount of heat produced from different fuels directly in the system can be added to the amount of fuels used in the district heating system and delivered to the third level. These scalars enter the third level of the model ( calc. for waste, calc. for oil etc.) Within these sub-model the emissions and use of PEC is calculated as a static function of the amount needed in the system. In these calculations both direct emissions from combustion as well as indirect emissions ( precombustion ) is included. The amount of PEC for example, include both the energy content in the fuel itself and the energy needed to supply the fuel to the system, and the same for the emissions. From the sub-models Heat and Fuel, information on amount of electricity used within these sub-models are sent to the sub-model Electricity. This is done in order to take into account electricity used in heat- and fuel production. 28

29 Primary EC Air emissions In Water emissions Electricity used (15 v ec.) 6 Heat Demux 10 Demux 11 electricity used (15 v ec) Primary EC In Air emissions Water emissions Fuel Sum 3 Mux Mux1 Figure 10. The top level of the Energy Production sub-model 6 Water emissions heat (use+prod.) 3 Air emissions heat (use+prod.) From heat prod. Primary EC From sub-models Air emissions From f uel prod. Water emissions Electricity 7 Water emissions electr. (use+prod.) 4 Air emissions electr. (use+prod.) 8 Water emissions fuel (only prod.) 5 Air emissions fuel (only prod.) 2 Net energy consumption (One 45 vector/sub-model) 29 Mux Mux 1 Primary EC

30 Primary EC In1 Air emissions Water emissions Electricity used calc. for waste Primary EC In1Out1 District heating Demux Sum4 In1 Air emissions Water emissions Electricity used calc. for Fossil oil 1 Primary EC Sum Primary EC Demux1 electricity2 Sum5 Sum3 In1 In1 Air emissions Water emissions Electricity used calc. for coal Primary EC Air emissions Water emissions Electricity used calc. for Fossil gas Sum1 2 Air emissions Primary EC Sum6 In1 Air emissions Water emissions Electricity used calc. for Biofuel Primary EC 1 In Demux M u x Sum7 In1 Air emissions Water emissions Electricity used calc. for Biogas 3 Water emissions Sum8 Mux Sum2 Demux electricity electricity1 M u x Mux1 4 Electricity used (15 vec.) Sum9 Figure 11. The second level of Energy production, the example is Heat (from the previous figure). The other two ( Electricity and Fuel ) are structurally identical 30