Abiotic resource depletion potentials (ADPs) for elements revisited—updating ultimate reserve estimates and introducing time series for production data

Guinée and Heijungs (1995) defined the ADP for resource i as follows:

Here, P_i is the world annual production (kg/year) of resource i, and R_i its ultimate reserve (kg). The ADP of resource i is the ratio for resource i divided by that of a reference resource, indicated as ref. Usually, this is antimony (Sb), but the precise choice is irrelevant in the final results, comparable with the choice between kilogram and pound.

Adopting the ADP as characterization factor, the corresponding category indicator result, here referred to as AD, is calculated as follows:

where m_i is a product’s use of resource (element) i (kg) and the summation is understood to run over all elements covered.

When a normalization step is part of the LCIA phase, a category total can be found for the impact category abiotic depletion by calculating the category indicator result for a specified reference region and time period. The elementary flows (i.e., resource extractions from nature) for a specified reference region and time period are indicated by M_i and are usually measured in per-time units, such as kg/year. The category total for abiotic resource depletion (TAD) is then found by:and the normalized category indicator result for abiotic resource depletion (NAD) as:

When AD is in the unit kg Sb-equivalent and TAD is in the unit kg Sb-equivalent/year, NAD is in the unit year; the mathematics demonstrate that it is independent of the choice of the reference substance (see Appendix A - ESM). A special situation is when the resource’s global production (P) that is used in constructing the ADP and the elementary flows (M) that are used for the category total coincides, that is, when they are for the same region (world) and year. In that case, M_i = P_i andand

Despite this possible identification of M_i with P_i, we will keep the symbols separate, as it might be useful to choose different regions or time periods for the production in calculating the ADP and the category total. As will be shown later in this article, the ADP might be based on the world resource production (P) in a specific year t, or an average or cumulative production over a period, while the world resource use that is used to calculate the category total for abiotic resource depletion might be based on a specific recent year y.

Because annual production differs per year, ADPs also differ, and so do the category indicator results. When relevant, we will add an extra subscript t to indicate the year, creatingP_{i, t}, ADP_{i, t}, AD_t, etc. Table 1 summarizes the terminology and symbols.

For updating production data, the annual reporting by USGS (US Geological Survey 2018) and BGS (British Geological Survey 2018) was utilized. Production data were downloaded for the time series 1900–2015 from the USGS website and for the period 1970–2016 from the BGS website. For more recent years, world production data reported by USGS refer to world mine production. Some historic data refer to smelter production and may thus include secondary resources. BGS provides production data for different stages of production and distinguishes between “metal, mine”, “metal, refined”, and “metal, smelter” for most metals. As the ADP refers to primary production, we used the production reported under “metal, mine” for our updated work. This should imply that we included only primary production data excluding secondary (recycled) material.²

USGS and BGS data for the same resource may differ for many reasons (like differences in units (e.g., tonnes ore versus tonnes element content) or institutions reporting to the geological surveys or completeness of the minerals for which an inventory is made). A thorough analysis of the differences in reported element extractions was beyond the scope of this article. In this article, USGS production data have been adopted as baseline data for the calculation of ADPs. The main reason for this is that USGS data are provided mostly in tons of element (in contrast to BGS data), which is the entry needed for the calculation of ADPs. For resources where USGS lacks entries (like for aggregates and clays), BGS data were used. Data on platinum group of metals (PGM) and titanium minerals were also retrieved from BGS because these data enabled breakdown into different PGM elements and titanium ores. Breaking down aggregate data on rare-earth minerals is based on Adibi et al. (2018). For some elements data are not reported by either USGS or BGS, such as cerium, hafnium, ruthenium, and scandium. Production data for these elements were taken from an EU report (Deloitte Sustainability et al. 2017). This report provides average data for these elements over the period 2010–2014. It has been assumed that these elements have also been used in the same amount in previous periods (1970–2009).

USGS and BGS also report production data for some minerals that are extracted to be used as such. This concerns minerals like bentonite, diatomite, and gypsum. In the actual markets, these minerals are considered to be irrelevant for the total primary extraction of metals and elements as such (dedicated extractions). However, by extracting these minerals, the elements are also extracted (non-dedicated extractions) from the earth crust. In this way, the extractions of these minerals also contribute to the total production of element i (P_i) in the ADP equation. The production data for these minerals were therefore also taken into account in the estimation of the P_i. An average composition of these minerals in terms of elements was used for converting them into elementary compositions (see ESM1: sheet “minerals to elements matrix”).

For updating the ultimate reserves, we need basic earth data and data on crustal content. Basic earth data on the volume and mass of the earth, ocean, and atmosphere were taken from Harte (1988). For the calculation of the ultimate reserves, the crustal content concentrations as compiled by Rudnick and Gao (2014) were used. Their report is considered the standard in geological sciences and is accepted by industry as the reference for average crustal concentrations (Drielsma et al. 2016). Lacking data were supplemented by estimations from Lide (1990). The data on the thickness of the different layers were taken from Rudnick and Gao (2014). Based on this, the average depth of upper earth crust was assumed to be 12 km. At present, the deepest mines are at approximately 4-km depth.

Because production data show large fluctuations over time, the ADPs will fluctuate over time. We developed three different strategies for dealing with this time issue when calculating ADPs:

Below, we present the detailed mathematical procedures for each.

A time series of ADPs on the basis of that year’s production is found through

A moving average uses the production of that year and a number of previous years. If the number of years used in the averaging is indicated by m, we use

For the first few years, we take a shorter time span because the time series does not provide the necessary data for the full calculation. With this, we use

The idea behind a cumulative ADP is that the more of a resource has been taken out, the more serious the problem becomes. The cumulative production up to a year is given by:

The physical unit of \( \overset{\sim }{P} \) is not kg/year, but kg. The corresponding cumulative ADP is found as follows:

For purposes of normalization, category totals were developed in a similar way:

Note that theoretically, the t in ADP and P can reflect different years and should therefore have different symbols. However, for ease of notation, the same t has been adopted above. Whenever relevant, we will indicate in the captions of tables and graphs which t has been taken for the ADP and which one has been adopted for the P.

Finally, for purposes of comparison with the existing practice, the 1999 ADPs as defined in Guinée et al. (2002) and compiled in the impact assessment spreadsheet CML-IA 2002 (CML-IE, 2018) were used. This set of ADPs from the CML-IA spreadsheet is based on yearly extractions of elements as compiled by USGS (US Geological Survey 2018). The extraction rates of elements refer to the year 1999. The calculation of the crustal content is based on Clarke and Washington (1924), Guinée and Heijungs (1995), and van Oers et al. (2002).

The first strategy is to calculate a time series of ADPs with that year’s production data. Full results can be found in ESM1, sheet a&g (ADPs); columns E–AX. It appears that the ADPs of individual elements differ in many orders of magnitude, for instance for aluminum, it is around 10⁻⁸ kg antimony-equivalent/kg, for chromium, it is around 10⁻³, and for gold, it is around 10³. To give an idea of the pattern over time, Fig. 1 shows the time series of a few high-value ADPs (ADPs for year t based on that year’s production data). We clearly see that there is quite some variation of the ADP values over time, but also that there is quite some consistency, e.g., gold is always higher than tellurium and palladium, and almost always higher than platinum. We also see that there are missing values (shown in the graph as zero), due to either missing production or reserve data, for the first few years for rhenium and palladium. The yearly variations of most ADPs and calculating ADPs with the most recent production data imply that ADPs should basically be updated every year, which may not be very practical and attractive for many applications, including policy setting or frameworks.

The second strategy is to calculate a time series of ADPs on the basis of a moving average of the production. Full results can be found in ESM1, sheet “a&g (ADPs) recommended ADP & norm data”; columns BH–DH. Figure 2 shows the effect of averaging over 5 years for the ADP of copper compared with yearly updated ADPs (first strategy). With a moving average over 5 years, we clearly see that the peaks are dampened and that a longer time window leads to smoother and more stable lines. From a practical and policy point of view, stable values are convenient. But of course, they may miss certain information that might be relevant. For instance, in Fig. 2, we see a highly fluctuating annual ADP between 1995 and 2005, which in the 5 years moving average is largely invisible. We could also adopt longer averaging periods, e.g., 20 or 40 years, in which case, the line will almost be flat. That would solve the issue of fluctuating ADPs, but it would also not show any trends (see section 4).

The third option is to calculate ADPs on the basis of the cumulative production up to the year of interest. In the definition of the ADP, we see \( \frac{P}{R^2} \), where the P is quite arbitrarily defined on a per annum basis. In the second strategy, we adopted a 5 years average for the production but we might also have taken a longer or shorter accounting period. While 1 year (first strategy) is the extreme for a shorter period, the extreme for a longer period is by looking at the cumulative production. With this, cumulative ADPs can be calculated. Full results can be found in ESM1, sheet “a&g (ADPs) recommended ADP & norm data”; columns GS–IL. We could even correct the reserve R for the cumulative production, deducting every year a part of the remaining reserve (see columns IN–KG). If we do so, we find for most elements no or negligible differences³ (see ESM1, tab sheet “Table 4’), and therefore, the corrected R is excluded from hereon. Table 3 illustrates the differences between the annual ADP, the moving average ADP, and the cumulative ADP for all resources. Figure 2 highlights the case of copper. For an interpretation of the results and recommendations on which ADP type to use, refer to “Discussion”.

Figure 3 presents the results of the category total calculated according to the three alternative strategies.

Similar to the ADPs, we see a strongly fluctuating pattern when the category total is calculated per year adopting the ADP per year (first-strategy ADP). The fluctuations are again flattened out a bit when the category total is calculated per year adopting the ADP based on a moving average (second-strategy ADP), and results in an even smoother line with an overall upward trend when the category total is calculated per year adopting the ADP based on the cumulative production up to the year of interest (third-strategy ADP). The strong fluctuating pattern raises the question whether it is the result of one or a few resources, or if it is due to an overall underlying cause, such as economic development.

To analyze this, we first study the contribution by different resources in the category total. In doing this, first observe that we use a global reference area and a calendar year as a reference time period. So, for the global annual use M_i, we take the global extraction P_i for 2015. Because the category total is the result of a sum over all resources,we can study the quantity TAD_i, defined byas the share of resource i in the category total. There appears to be a very uneven distribution here. For barium, for instance, the contribution is less than 0.1%, while it is around 70% for gold (it varies between 61 and 80%, depending on the year). In Table 2, the top 20 contributions of elements to the category total (TAD₂₀₁₅) are presented (adopting first-strategy ADPs and reference year 2015). The total list can be found in ESM1 (sheet “recommended ADP & norm data” a&g (norm. data, year 2015)’; columns E–AX).

Just a handful of elements contribute more than 1%: besides gold, these are antimony, cadmium, chromium (in some years), copper, lead, mercury (in some years), molybdenum, palladium (in some years), platinum, silver, tellurium, and tin (in some years). Figure 4 shows their share pattern over time (adopting first-strategy ADPs).

Note that this dominance of the category total by just a few elements is a phenomenon also seen for other impact categories in normalization (Wegener Sleeswijk et al. 2008; Van Oers and Huppes 2001). For example, the category total for global warming is 65% due to CO₂; the ozone depletion category total is 44% due to CFC-12; the human toxicity category total is 27% due to chromium; the freshwater ecotoxicity category total is 67% due to aldicarb etc. (see the “impact assessment” spreadsheet CML-IA (CML-IE). The dominance of the abiotic resource depletion category total by gold is thus not a unique phenomenon within the current state-of-the-art of normalization. Limitations of characterization models and data do of course influence these results, but given the problem definition and ADP-model used, the dominance of the category total result by a few elements is a mathematical fact. Considering the price, demand and reserves of gold, this dominance of gold is also an understandable result given the ADP model.

Above, we studied the temporal variation of ADPs and their contribution to the category total. In this section, we consider the effect of two choices we made, i.e., the choice of reference substance and the treatment of the variability of the production data as uncertainty.

For the calculation of the ADPs, data are needed and assumptions are made which might lead to different characterization factors and category total for abiotic resource depletion based on the ADP.

Basic earth data are used to calculate the volume and mass of the continental earth crust, like depth, surface area, and density. Together with the concentration of the elements in the continental crust, these data are used to estimate the ultimate reserves. Since—in the end—ADP is a relative expression of the potential contribution of elements to depletion, the relative availability of elements as expressed in R is of importance, and the absolute values of the reserves are of less importance. Consequently, the adopted earth crust’s depth, area, and density will not influence the contribution of the elements and thus the value of the ADP as long as within the volume of the compartment (area × depth), the concentration is considered to be homogeneous. This is most likely not entirely reflecting reality but data on concentrations provided by Rudnick and Gao (2014) are given as average data for the upper earth crust and as such considered homogeneous.

In the present calculation of ADPs, only the continental crust is considered. Mining of the oceanic crust is not foreseen in the near future. Again, expanding the volume of the crust by taking into account the oceanic crust will not affect the ADPs considerably as long as the composition of the oceanic crust is more or less the same as the composition of the continental crust. However, concentrations of elements in the oceanic crust are believed to be very different compared with the continental crust (White and Klein 2014), so if mining of the crust beneath the oceans is considered feasible, oceanic crust could become important.

For the ultimate reserve, crustal content data from Rudnick and Gao (2014) were adopted. Significant changes in crustal content data are not expected, but if they occur, these should lead to further updates of ADPs and category total in future.

Production data are compiled by many different geological surveys, like USGS and BGS. Different geological surveys report sometimes different production data. There are many different reasons why reported production data may differ, such as different principles of the classification of minerals, differently assumed concentrations of elements in ores and minerals, different units (tons ore, tons element, tons compound), and different responses the surveys get from geological surveys, statistical bureaus, and industry. USGS and BGS meet yearly to mutually align their data, at least for some of the most important elements and metals. A yearly mutual tuning of production data for all the elements and minerals by all stakeholders is recommended—not just for the purposes of ADP calculations, but for improvement of all future efforts to model raw material supply.

A key challenge addressed in this work has been how to deal with the fact that production data annually fluctuate, resulting in potentially widely fluctuating ADP values and category totals over different years. In addition to just adopting the most recent production value for calculating the ADP and category total, we have analyzed two other options for calculating the ADP:

The differences between the 2 alternative options appear to be small (see Table 3), and they go in different directions for different resources. Table 3 shows individual ADPs for individual elements. The differences between the 3 different calculation procedures for the ADP (i.e., ADP_2015, ADP_2015 (moving average), and ADP_2015 (cumulative)) of a specific element have a maximum variation between 0.2 and 1.8, and only between 0.7 (rhenium) and 1.5 (tellurium) for the eight highest ADPs (Fig. 1) (see ESM1, Table 3 sorted). The key point in making a choice between the annual ADP, a moving average ADP (1), or a cumulative ADP (2) and between the different time horizons for averaging basically comes down to addressing two questions:

Which option is taken is a normative choice and open for debate which is inextricably linked to the aim of the impact assessment method. If the method should reflect trends, the most suitable option would be to adopt the average production over 5 years (fourth column in Table 3) as longer averaging periods will increasingly flatten out the trends (see Fig. 2). In this case, however, the use of crustal content (ultimate reserves) might no longer be appropriate. If the ADP should reflect how problematic society’s use of primary natural resources cumulatively is, the cumulative ADP over time for which we have production data (last column of Table 3) is the most appropriate option to adopt, with or without correction of the R for the cumulative production (in practice, there is no difference between the two). Since the latter is in line with the original aim of the ADP method, the cumulative option is recommended here.

Independent of the choice for a specific calculation procedure, agreement on an update procedure of ADPs and category totals is required when the ADP is used for public policies such as the EPD and PEF (Wardenaar et al. 2012). Assuming that USGS and BGS will continue to publish production data in the upcoming years, a recurring period will need to be defined after which ADPs and category totals will be updated to the latest production data. An obvious option for this is to calculate new ADPs and category totals every 5 years, as this would follow update cycle for other impact categories such as the update of GWPs by the Intergovernmental (IPCC, 2018).

For a consistent calculation of normalized indicator results for ADP-based abiotic resource depletion, it is crucial that the ADPs and category total are based on the same characterization model, use the same input data for P and R, and represent the same spatial scale (i.e., spatial scales of characterization model and normalization reference area), and that the category total is based on the same complete set of elements as for which ADPs have been derived.⁵

Considering the reference period, the question is what to adopt as “world annual extraction” for calculating the category total. Should the same choice be made for this “world annual extraction” (M_i) as made for the ADP (cumulative production over a number of years, P_i)? Our answer would be “no”. We argue that in line with global warming and other impact categories, the production data for the most recent and completely documented reference year should be adopted, for all impact categories. The aim of normalization is to compare characterization results with the same reference information, e.g., characterization results for the world for a given year. Preferably, the same reference year (and region) should be adopted for all “world annual emission/extraction” data sets for all impact categories, which is also how these data have been collected so far (van Oers and Huppes 2001; Wegener Sleeswijk et al. 2008; Sala et al. 2015). Depending on the reference year adopted for other impact categories, the proper reference year for the “world annual extraction” for the ADP should be selected from the accompanying spreadsheet ESM1, e.g., 2010. Of course, we could also take the cumulative production over a number of years as reference but then we should take a similar reference for other impact categories (e.g., cumulative emissions over the same period). That would require cumulative emission data, which may not always be available.

Considering the reference area, we want to briefly point at an inconsistency in the use of the reference area as currently adopted in the PEF. In the current PEF, ADPs based on the USGS reserve base are combined with European (“domestic”) annual extractions for 2010 published by Sala et al. (2015). Compared with the category totals developed in this article, Sala et al. (2015) adopted another reference area for normalization (EU 27 instead of the world), use the USGS reserve base instead of crustal content for the R, and moved uranium from minerals to he fossil energy category while changing the name of this category to “energy carriers”. This EU 27 normalization is problematic in several ways: the P in the ADP reflects world production and is compared with an M in normalization that reflects EU 27 production. This results in an inconsistent calculation of normalized indicator results since not all elements, for which there are ADPs, are mined within the EU 27 but rather imported (either as resource or as processed resource in materials and products). Altogether, this means that inventory data on metal, mineral, and fossil resources appear to be available for only 34 out of 74 elementary flows for which there is an ADP (Sala et al. 2015). Similar problems will also affect normalization results for other impact categories. Using crustal content data for R, considering uranium as mineral and not as energy carriers, and adopting 2010 as reference year for normalization, Fig. 5 shows the influence of adopting the EU 27 as the normalization reference area (covering 34 elements) instead of the world (covering 74 elements) for the cases of 1 kg copper and 1 kWh of electricity Dutch mix, adopting the “ILCD 1.0.8 2016 midpoint” methods as implemented in ecoinvent v3 except for the ADP, which is taken from this study for the year 2015 (ADP₂₀₁₅). We see some remarkable differences between the global and EU 27 results, for the mineral resource, toxicity-related, and the freshwater eutrophication categories. All of these differences are related to the difference in the reference area for normalization.⁶

In general, the following reasoning can be useful for analyzing the validity and reliability of normalized indicator results. Assuming representation by perfect data and full coverage of all possible elementary flows, in general, one expects that a typical functional unit should show normalized indicator results for the different impact categories in the same order of magnitude. In appendix A (ESM), a mathematical proof is given for this statement. However, as perfect data and coverage and the “typical functional unit” do not exist, practice may turn out differently. Specific products may show (normalized) indicator results for abiotic resource depletion (or any other impact category) that deviates from the order of magnitude from the other impact categories. For example, the normalized indicator results for ADP-based abiotic resource depletion for 1 kg of copper are up to a factor or 3 higher in comparison with other impact categories. This is largely explained by the fact that resources play a dominant role in the inventory result of 1 kg of copper production and by copper’s relatively high ADP, which is due to its relatively low crustal content combined with relatively high production. Note that the normalized characterized scores only reflect the relative ranking of resources within the impact category, without any indication of the seriousness of the problem compared with other impact categories (which is weighting). Although copper’s ADP is relatively high due to how the factor is calculated, the USGS estimates copper reserves amount to 790 million tonnes, and copper resources are currently estimated over 5000 million tonnes (USGS 2018). Thus, the result should be interpreted carefully by the study user.

However, if normalized indicator results for a certain impact category are by 1 or more orders of magnitude higher than for other categories, this potentially points at consistency and/or coverage problems as explained above and as well-known for the toxicity-related impact categories (Heijungs et al. 2007; Prado et al. 2016).