Paradigm for Substance Flow Analyses

2. Principles and Procedures

2.1 The Principle of Substance Balance
2.2 The SFA System
2.3 Procedures
2.3.1 Dealing with Uncertainty
2.3.2 Cross-check
2.3.3 Assessment of Consumption
2.3.4 Modelling
2.3.5 Sun-down chemicals

2.1 The Principle of Substance Balance

Substance flow analysis (SFA) is an analytical tool, that is used for achieving the understanding of the flow of substances within a given system defined in space and time.

Any substance flow analysis is based on the principle of substance balance:

(Eq. 1) Input + formation = output + degradation

As concerns elements, the mass is always constant, and there is no possibility of formation and degradation. For elements the substance balance is thus a simple mass balance: input = output. In the case of chemical compounds that are composed of more than one element, 'formation' and 'degradation' may have a significant impact as chemical, biological or thermal processes may lead to both formation and degradation of such substances.

Expression (1) may be formulated in a number of ways, depending on the system it aims to describe. If the purpose as here is to assess the consumption of a product in Denmark, the expression will be:

(Eq. 2) Imports + Danish production = exports + consumption

Û

(Eq. 3) Consumption = Danish production + imports - exports

In all substance flow analyses, it is important to determine the boundaries of the analysis in time. The substance flow analyses covering the Danish society as a whole have most often run for a period of one year. However, many products will have a useful lifetime of more than one year, and, at the same time, the market is constantly changing. For this reason it is often relevant to include changes in the stock building of the substance.

Within the boundaries in time, the substance balance can consequently be expressed:

(Eq. 4) Input + formation = output + degradation + accumulation

In this expression, 'accumulation' refers to the changes in stock building within the period covered. The equation is discussed further in section 2.2.1.

Expression (1)-(3) all describe simple substance balances, which can be made more or less complex depending on the desired level of detail. In fact, SFAs consist of the systematic use of such substance balances to describe a larger system. The substance balances can be used to describe a static situation by simple bookkeeping or be build into more complex modelling of chances in the balances, which can be used to describe the dynamic of the system and forecast future situations.

2.2 The SFA System

As mentioned before the SFA system is defined by the boundaries in space and time.

Boundaries in space

The focus of the 'core SFA' of this paradigm is the Danish society and all the processes taking place from the extraction/import of raw materials to the final disposal of residual products. The Society may as well be designated the 'technosphere' or the 'physical economy'. The society is surrounded by 'The Environment' divided on the compartments: 'soil', 'water' and 'air'. The 'core SFA' does not cover the flows between these compartments.

A simplified model of this system is shown in Figure 1.

Most of the arrows shown in the figure may indicate several different types of transport. For instance, 'recycling' covers:

	The chemical substance in isolation or contained in materials from which the substance can be recovered by industrial processing (e.g. lead plates in batteries for cars)
	The chemical substance as a trace element and contaminant in other materials which are recycled (e.g. lead as a trace element in paper, metals, glass and plastic)
	The chemical substance contained in residual products used for the manufacturing of other industrial products (e.g. lead contained in fly ash from coal-fired power plants where the ash is used for manufacturing of cement and concrete)

In the figure it has been chosen to indicate 'landfilling' as a separate level in the system between 'treatment of waste/residuals ' and the environment (represented by the boxes 'soil', 'water' and 'air'). In reality, for many organic substances 'landfilling' will be a form of waste treatment, as degradation/transformation will take place in a landfill. In the case of persistent chemical substances, landfills should be regarded as a type of storage from where these substances can slowly disperse into the environment by leaching or evaporation.

Figure 1
Diagram showing the principles of the transport and turnover of chemical substances in Denmark

Defining the boundary of the analysis at the Danish borders means that emission of the substance by extraction and processing abroad of imported raw materials are not covered by the analysis. Similarly emissions by processing of exported waste is excluded. It should be noted that other SFA/MFA approaches focus more on these life stages as decribed in Bringezu 1997.

Concentration and accumulation of the substance in the environmental compartments are not covered by the 'core SFA ' but may be included in an extension to the analysis as described in section 3.7.2.

Boundaries in time

In the 'core SFA' the system in time is defined as one year. It means that the analysis aim at estimating the actual flows between the compartments the year of reference.

In fact it is often necessary to include information from a broader time period as the emissions to the environment and loss to waste the year of reference can only be estimated on the basis of information on historical consumption figures. For 'sun-down' chemicals, the emission to the environment and loss to waste may for instance be much larger than the actual consumption.

For the discussion of the environmental implications of the present consumption of the substance it may be relevant to expand the boundaries in time and try to predict future emissions as result of nowadays consumption. Such predictions are included in the optional extension described in 3.7.1.

2.3 Procedures

Whereas the principles of SFAs are very simple, the practical procedures will seldom be. The reality, which is described by means of an SFA, is often complex and it may be difficult to obtain the necessary data. In addition the data, which can be obtained, is seldom 100% accurate, but usually subject to varying degrees of uncertainty.

This section is not intended to thoroughly go through all procedures, but to focus on some procedures that by experience often cause difficulties.

The procedures used in the individual case will depend on the pattern of application and flow of the chemical substance to be analysed as well as the purpose of the analysis and the level of detail and reliability considered appropriate. According to experience and internationally proposed framework, the following phases of the SFA study is proposed:

	Phase 1: Goal and system definition
	Phase 2: System analysis
	Phase 3: Inventory, evaluation of data and modelling
	Phase 4: Interpretation of the results

Compared to the steps introduced in the Introduction a step with 'system analysis' is added. The main reason is that the SFAs covered by this paradigm in general are projects where goal definition and interpretation of the results are carries out at two levels. The goal definition and interpretation of the results by the initiator of the analysis (here the Danish EPA) in a way encompass the goal definition and interpretation of the results by the researcher.

Phase 1. Goal and system definition

The initial goal and system definition is often carried out by the Danish EPA or other clients. The present paradigm is pretended to provide a reference point both in the preparation of call for tenders and preparation of tenders. After initiation of the project the goal and system definition may be changed in cooperation between the researcher and the client on basis of the system analysis.

Phase 2: System analysis

The system analysis is considered a separate phase because this is often the first phase involving the researcher and the outcome of this phase may effect both the goal and system definition and the inventory. In this phase, the whole system is considered thoroughly and all types of transport and processes, which have an impact on the flow of the substance/material through society, are identified. This phase forms the basis of the subsequent inventory, as it is during the system analysis phase that the requirements are determined as to what data and degree of accuracy are needed. Such a system analysis will be based on knowledge about industrial applications (intended applications and application as a trace element and contaminant) of the chemical substances in question. In addition, detailed knowledge about treatment of waste and disposal of residual products in Denmark is required (cf. Section 4).

It is important to keep in mind that the flow system established in this phase would seldom be perfect. Presumably, there will be aspects which have been overlooked and types of transport/processes considered to be essential will turn out to be unimportant. In practice, it is an iterative process through which the system is modified continuously as more profound knowledge is obtained.

Phase 3: Inventory and modelling

This phase starts out with the collection of easily accessible data. In this context, easily accessible data comprises statistical data about the production, import and export of relevant raw materials, semi-manufactured goods and finished products. In addition, there will be monitoring data concerning finished products and waste streams, etc.

On this basis, a strategy is worked out for the continued collection of data. Should data be collected from enterprises and if so, what information from what enterprises etc.? Are individuals at Danish research institutions likely to possess information etc.? If a questionnaire is used or interviews are made, how can the questions be phrased so as to obtain the needed answers? Are certain persons likely to be reluctant to supply information and perhaps give wrong information, and how can this situation be prevented and monitored? The relevant sources of information will be plentiful and depend on the types of information required (cf. Section 4).

The above questions are primarily meant as a checklist for the considerations to be kept in mind in the planning of the data collection. General rules cannot be laid down. The right solution in the individual situation must be based on experience and the personal preference of the researcher. If a wrong solution is chosen at first, the analysis will have to start all over again, until sufficient data have been collected for a realistic estimate to be made.

An example of estimation of the total consumption of a substance for a specific application based on information from producers is shown in box 1.

During the inventory the pieces of the puzzle (the individual data) are put together. Often, the figures will not correspond precisely and there will be a need for a more detailed evaluation of the data to find out why. Can the discrepancies be explained by inaccuracies of estimates and monitoring results or has something been overlooked or based on wrong estimates? Again, this is an iterative process as there may be a need to improve the collection of data until a satisfactory result is achieved (the jigsaw puzzle is complete).

Some of the flows can due to the lack of monitoring data only be determined indirectly using models. This will be further discussed in section 2.3.4.

Phase 4: Interpretation of the results

Interpretation and discussion of the results can be done at more levels. The SFAs carried out according to the old paradigm (Hansen & Boisen 1993) has been criticised not to go into more detail with the interpretation of the results and keep the analysis and interpretation at a level of bookkeeping. This criticism is right as much of the interpretation is left for the authorities and other readers of the reports.

The basic interpretation, which should be part of the reporting, includes a discussion of the obtained flow chart based on the cross-checks which are discussed further in section 2.3.2. The main sources of emission to the environment and losses to waste are pointed out and emissions of potential importance where more data are needed are flagged.

A further discussion of the emissions of the substance in terms of contribution to environmental impact or sustainability indicators may be relevant, but such an impact assessment is not included in the SFAs covered by this paradigm.

The interpretation of the results in terms of regulatory actions is kept out the analysis and reporting, but left for the authorities initiating the analysis. By the interpretation of the results the researcher may for example point out the main sources of discharges of the substance to the aquatic environment, but does not propose specific regulatory actions to minimise these discharges. The reasons for this is, as mentioned in the Introduction, that one of the goals of the analysis is to provide a common understanding for all stakeholders of the risk minimisation process.

2.3.1 Dealing with Uncertainty

Almost all data used for the SFAs will be subject to a certain degree of uncertainty. The sources of uncertainty will not be discussed in detail here. A systematic way of describing the sources of uncertainty can be found in /Hoffmann et al. 1997/. In this chapter will be focused more on how to represent and manage the uncertainty.

The uncertainty should be specified for all figures. It is recommended to represent uncertain figures as probability distributions and indicate the uncertainty by intervals; e.g. 200-250 tonnes.

For most figures there will be a 'true value'; e.g. the actual consumption of the substance within the reference year, but the researcher conducting the SFA does not know this value. When a number of monitoring data exist, it may be possible to use standard statistics to estimate a confidence interval assuming the data are normal distributed, but for most figures in the analysis the uncertainty has to be determined by 'expert judgements'.

As the figure is uncertain it will not be possible to represent the figures by an interval within which the 'true value' at 100% certainty can be found.

In order for the intervals not to be unreasonably wide, it is recommended to use intervals representing a 90% (or 80%) certainty level. In other words: the figures are represented by the interval within which the author estimates the 'true value' with 90% certainty can be found. In addition it can be assumed that the probability is normal distributed with the mean value as the most probable and a symmetric distribution around the mean.

This means that the width of the interval directly indicates the uncertainty on the results. It makes no sense to state: 'The consumption is roughly estimated at 100-110 tonnes'. A rough estimate will inherently be quite uncertain and if the estimate is rough it will be more correct to state: 'The consumption is roughly estimated at 20-200 tonnes'.

By experience most people have a tendency to underestimate the uncertainty. To get an idea of how wide the intervals should be it is recommended to practice with standard statistics e.g. in spread-sheet applications. If e.g. three independent information sources estimate the consumption to be 200, 300 and 400 tonnes, respectively, and we assume the information sources to be equally trustworthy, the 80% confidence interval calculated with standard methods will be 191-408 tonnes.

If all figures are represented as probability distributions a consistent way of adding up the figures is to use the rules for e.g. addition of normal distributions. Principles and software for carrying out life cycle assessments using figures represented as probability distributions is described in /Hoffmann et al. 1997/.

Managing the figures by these rules of probability distribution arithmetic, however, makes the analysis less transparent and it is recommended simply to add up figures by adding up the lowest and highest values within the intervals, respectively.

 

Consumption with product A	100-300
Consumption with product B	300-600
Total	400-900

The consequence of using this simple addition principle is that the intervals representing the total actually represent a higher certainty level than the two input-data.

The interpretation of the intervals used in the analysis should be described in the 'Preface' or 'Introduction' of the report. For example, it should be mentioned whether the intervals for input data are considered to represent 80% or 90% certainty levels?

In a few instances when a number of analysis data are available it will be possible to use standard statistical methods e.g. to calculate a confidence interval on the mean value. As an example data on the concentration of the substance in sewage sludge may be available from many sewage treatment plants and a 90% confidence interval can be calculated presuming that the analyses represent random samples.

Most often the available date, however, cannot be presumed to be random samples from a common distribution, and it may be expedient to divide the data into more distributions, treating each distribution independently. By this method the uncertainty on the total result will often be lower.

For instance it is relevant to treat analyses of flue gas from solid waste incinerators as three distributions depending on the present flue gas cleaning technology: dry, semi-dry or wet cleaning technology.

Beside the natural variation among the samples each analysis are subject to an uncertainty dependent on the applied analysis method. As far as the uncertainty is unbiased (symmetric distributed around the mean), this uncertainty will simply make a part of the measured variation between the samples and need no specific treatment. It is, however, essential to be sure that the analysis method not systematically overestimate or underestimate the actual content of the samples. Especially when using older analysis results it is thus necessary to assess whether new analysis methods has revealed that the applied method overestimate or underestimate the actual content.

2.3.2 Cross-checks

The most important tool to handle and evaluate the uncertainty is 'cross-checking'. Whenever there is a possibility to cross-check, it should be done in order to verify the coherence of the SFA. One of several important cross-checks consists in checking whether the consumption by the manufacturers of raw materials corresponds to the consumption of raw materials and semi-manufactured goods registered in the statistics. If these data correspond, the estimated consumption for different purposes is likely to be true. If the data do not correspond, one ore more estimate(s) may be wrong or stock building etc. may be taking place.

In reality, many of the data collected for SFAs are so uncertain that they do not really have any value until they have been cross-checked. SFAs for chemical substances at the national level in many ways resemble jigsaw puzzles. The best way to prove that the jigsaw puzzle has been completed correctly is to make certain, that it is not possible to point out pieces, which do not fit together.

The following types of cross-checks based on the application of mass-balances is considered to be relevant:

	Does the consumption of raw materials/semi-manufactured goods registered in the statistics correspond to the consumption of raw materials of the manufacturers?
	Does the consumption of raw materials of the manufacturers correspond to their production and estimated losses to wastewater, soil, air and waste as well as the volume recycled?
	Does the estimated total loss to solid waste from enterprises and consumers correspond to the volumes measured in solid waste?
	Does the estimated total loss to wastewater from enterprises and consumers correspond to the volumes measured in wastewater?
	Does the estimated total loss to chemical waste correspond to the volumes received by Kommunekemi (the Danish central treatment facility for chemical waste) or other recipients?
	Do the volumes presumed to be recycled correspond to the volumes received by the recycling companies?

Box 1
Evaluation of information received from manufacturers

This box describes an example of the collection and evaluation of data from manufacturers. The example is hypothetical, but must be considered representative of the type of assessments to be carried out as part of SFAs for chemical substances at the national level. Based on statistical information, the annual consumption of the substance X for manufacturing of products in Denmark is known to be approx. 1,500 tonnes. It is also known, that this consumption is distributed on the fields of application A, B, C, D and E. The task is to determine this distribution as precisely as possible. For the field of application A, the Danish Technological Institute informs that 3 large-scale manufacturers exist in Denmark each with a market share of 20-30% as well as 5-6 small-scale manufacturers. It is decided to collect information from the three large manufacturers as well as two of the small ones. Due to severe competition, most manufacturers do not wish to indicate their own consumption. However, they are willing to provide an estimate of the total consumption for the field of application A in Denmark. The following information is obtained: Manufacturer 1: Total consumption estimated at 200-250 tonnes Manufacturer 2: Total consumption estimated at 200-300 tonnes Manufacturer 3: Total consumption estimated at approx. 250 tonnes Manufacturer 4: Own consumption approx.: 2 tonnes. Market share estimated at 5-10%. Manufacturer 5: Total consumption estimated at 100-200 tonnes As manufacturer (1)-(3) are the 'large' manufacturers, they are presumed to have a better impression of the market than manufacturer (4)-(5). The preliminary figure for the consumption within the field of application A is therefore estimated at 250 tonnes with an inaccuracy margin of ± 50 tonnes. The final figure can only be determined when information has been collected for all the fields of application A-E. As it turns out, the total consumption estimated on the basis of the information collected from the manufacturers amounts to 1,500-1,800 tonnes. As the statistical information is considered to be reasonably reliable (at any rate, more reliable than much of the information obtained from the manufacturers), the total consumption for the fields of application A-E is estimated at approx. 1,500 tonnes. Consequently, the information for each field of application has to be re-evaluated in order to find out where there is an estimate, which may be too high. As a result of this re-evaluation, it is decided to change the estimate for field of application A to approx. 200 tonnes with an inaccuracy margin of +50 tonnes.

2.3.3 Assessment of Consumption

By experience there is often some confusion regarding the interpretation of the terms 'consumption' and 'supply' used in the analyses. For this reason the consumption assessment will be discussed in more detail here although the consumption assessment is only one part among others of the analysis.

The substance is most often both consumed with raw material and semi-manufactures for production of finished products in Denmark and consumed with finished products sold at the market.

In this paradigm the term 'Consumption in Denmark' represents the consumption of the substance with finished products. The consumption of the substance with finished products during the year of reference is equal to the total content of the substance in finished products sold in Denmark during the year. This is independent of whether the products are actually used up by the consumer during the year or they last for many years.

For example, the flow of the brominated flame retardant TBBPA with printed circuit boards is illustrated in figure 3.1.

The consumption of TBBPA in Denmark is represented by the consumption of the substance with the finished electronic products. Beside this it is relevant to assess the consumption of TBBPA for production of compounds, the consumption of laminates for production of printed circuit boards, and the consumption of printed circuit boards for production of electronic products. The main emphasis of the assessment of production of semi-manufactures is the losses by the processes, and consequently the focus should be on processes where losses occur.

Consumption for production processes is in the SFA specifically termed 'Consumption for production processes in Denmark'.

Figure 3.1
Flow of TBBPA with printed circuit boards (modified from Lassen et al. 1999 B)

By assessing the consumption of the substance with raw materials and semi-manufactures for production in Denmark e.g. from the import/export statistics it is important to avoid "double account" where the same quantity is accounted two times. This is further discussed in section 3.4.1 of appendix 1.

Supply

The terms 'supply' and 'consumption' describe in fact the same reality from two different points of view. In this paradigm, however, the term 'supply' is used in a pure trade statistical context where:

Supply = Danish production + imports - exports

The equation is similar to Equation 3, but consumption is replaced by supply. The term is used when data from the statistics is presented and the supply is calculated as a single value from the available statistical data. The supplied commodities may both be consumed for production processes and as finished products, and in addition the statistical data may be vitiated by errors. It is an essential part of the SFA to determine to what extent this registered supply actually is equal to the consumption.

Accumulation

The consumption as defined above does actually not represent the total input to the Danish society; as also emissions and waste from production processes must be considered inputs (and at the same time outputs).

For the calculation of the present accumulation of the substance in the Danish society equation 4 can be modified in the following way:

(Eq. 5) Accumulation = consumption(finished products) + loss(production processes) - total emissions - total loss to waste

'Loss (production processes)' represents emissions to the environment and loss to waste from production processes. The loss can be ideally be calculated as:

Loss(production processes) = import (raw materials) + extraction(raw materials) - export(raw materials) - production(finished products)

The picture is, however, in reality often very complicated due to intensive import and export of semimanufactures. The losses from the production processes are therefore most often calculated from the figures on production and typical emission factors. If the number of Danish producers is small the loss from production processes may be estimated from information obtained from all producers.

These losses are also included in 'total emissions' and 'total loss to waste'. The reason for this emphasis on the loss from production processes is that by experience these losses are often omitted from the assessment of inputs and only included as emissions.

It should be noted that the accumulation in the society does not say anything specific about the accumulation of the products sold the year of reference. For substances where the consumption is declining the accumulation may be negative meaning that the amount of the substance accumulated in the society is declining.

2.3.4 Modelling

In principle all potential flows identified during the system analysis should be estimated. Allthoug the 'core SFA' is aimed at being a simple 'bookkeeping' of the substance flows it is often necessary to use simple or more complex models to determine some of the flows. In addition some of the extensions include more complex modelling.

It is essential to keep in mind that the SFA is not only a systematic way of representing the available information on the flows of the substance. The SFA is a systematic way of representing all flows; whether or not data are available. It is not uncommon that less trained researchers confuse the available information on flows with the "true" flows.

Less investigated substances

When carrying out SFAs for less investigated substances there may not be any Danish monitoring data of the substance in wastewater, residual products, flue gasses, etc. For most substances data on emissions from production processes are also absent.

In this situation it has to be decided whether the SFA should exclude this part of the analysis or the most probable values should be estimated on the basis of analyses from foreign countries or theoretical models.

If models and foreign results are used in the SFA, it should be clearly indicated all through the report. For instance such estimates has been used in SFA for brominated flame retardants /Lassen et al. 1999 B/.

Foreign analysis results

When using foreign analysis results the use pattern of the substance in the foreign country should be comparable to the use pattern in Denmark. For substances with same use pattern in Denmark and Sweden, by way of example, analysis from Swedish wastewater treatment plants may be used in the absence of Danish result, when all assumptions are clearly indicated and it is indicated that the estimates are based on Swedish results. Such results can be used to indicate whether Danish analysis programmes should be initiated.

Models based on physical/chemical properties

In the absence of directs measurements of emissions from production processes or products in use the emissions may be estimated from models e.g. based on physical/chemical properties of the substances and the materiel they make part of. Such estimates will usually be very uncertain and only the order of magnitude can be estimated.

Process specific emission factors has been developed in EU for use in Risk Assessments and emission factors for different industrial sectors can be found in the "Use Category Documents". Similar documents are under development under the auspices of the OECD. An example of using the "Use Category Documents" can be found in SFA for brominated flame retardants /Lassen et al. 1999 B/.

Models based on historical consumption figures

Often the loss of the substance to e.g. solid waste can only be determined from information on historical consumption figures as the products discarded today may have been introduced into the society many years ago.

Today loss of the substance with a specific product will depend on two parameters:

	An assessment of the life-time distribution of the product
	An assessment of the annual consumption of the substance with the product within a historical period covering the whole range of the life-time distribution

A detailed example can be found in SFA for cadmium, where the amount of NiCd disposed of in 1998 is estimated based on the basis of such a model (Drivsholm et al. 2000).

These models in many ways resemble the scenarios for the future loss/disposal of the substance discussed in Appendix 1 section 3.7.1.

2.3.5 Sun-down chemicals

For sun-down chemicals, chemicals on the way to be phased out, there is some specific procedures that will be mentioned here.

For substances, which are commonly used for the application in question in Denmark, the total consumption in Denmark may be extrapolated from information from the main producers and suppliers. This procedure will, however, not apply for 'sun-down chemicals', because the manufactures will usually not know whether other manufactures still use the chemical or use a substitute. In addition it will often be so that the largest producers have substituted the substance, but it is still used by small producers. In this instance the total consumption cannot be extrapolated from information from large suppliers representing e.g. 80% of the market of the product, as the remaining 20% may represent the total consumption in Denmark. It is thus necessary that the assessment cover practically all suppliers. An example of such an analysis can be found in SFA for chlorinated hydrocarbons (Maag 1999).

This means that it is necessary to have a nearly 100% coverage which can only be obtained by questionnaire surveys or telephone interviews. By experience the reply percentage of questionnaire surveys is quite low unless the inquiries are followed up by direct telephone calls.