Preliminary work on the grey water footprint
In the context of this study, grey water footprint refers to the idea of expressing water pollution in water volumes by converting substance loads into the dilution volume that would be required to dilute the pollution below substance-specific limits. Conversely, a substance-specific dilution factor can be derived from this. This idea is basically not new because already in 1974 it has been pointed out that an average dilution factor of 10 for wastewater flows is at least required to “dilute[d] [polluted water] in order that concentrations of pollutants be reduced to an “acceptable” level”, although the approach has not yet been called grey water footprint32. Almost 20 years later, annual freshwater runoffs have been studied to find out what proportion of Earth’s freshwater is actually accessible to humans and a then common dilution factor of 28 litres per second per 1000 people33 has been used to take into account the amount of dilution required for waste water treatment. While these studies still used average dilution factors to describe human-induced water pollution in principle, a more recent study suggested to use substance-specific dilution factors34, e.g. 100 for N (related to cubic metres) considering a permissible limit of 0.01 kg m−3. Shortly after, the expression grey water footprint has been introduced12, which has since become common, by defining it as the substance load divided by a specific threshold value that is valid for the receiving water body. Such thresholds can originate from generally applicable national or international water quality standards. In a further study, the authores changed their approach slightly by carrying out the dilution with regionally actually available water, so to speak, and not with demineralised water13. Mathematically, this means that the substance load is divided by the difference between threshold and natural background concentration. This can also be understood as the amount of water needed to assimilate pollution through substances14. However, the authors have emphasized that dilution is not a “free pass” for water pollution, but a method to quantify it volumetrically in order to be able to reduce it. Since then, the grey water footprint has been widely applied in many different contexts creating a large pool of literature, not all of which we can discuss here. In the field of agriculture it was for example used to express the impact of agricultural emissions of N and P on water quality on the global level15,16 and it has been further used to reveal the impact of N fertilizer emissions on ecosystems.
As part of a generally critical examination of existing water footprint methods, a concept of the water scarcity footprint for LCA was introduced in order to set a different focus in the area of water footprinting: the water compartments of a catchment are no longer considered separately (e.g. surface water, groundwater, rainwater), but as a hydrological unit in a catchment. This is to take into account the fact that the use of water from any compartment can contribute to regional water scarcity. In the case of the grey water footprint, which is referred to as the qualitative water scarcity footprint and described by the virtual dilution volume within the approach, regional aspects are taken into account by requiring to orient the dilution on the naturally prevailing water quality. For this purpose, the load is divided by the natural, or geogenic background concentration, if not greater than general water quality standards, referred to as target concentration. Subsequently, in the context of global supply chains, dilution is done fairly with demineralised water for all applications. The argument behind this is that a consumer or product should not take advantage of a lower water quality footprint just because regional water is naturally cleaner. On the other hand, naturally already more pre-loaded water can have this effect, too, but here the target concentration is used to mitigate the effect in the case of excessively pre-loaded water. And in the line of reasoning of the approach, a consumer or product cannot be held responsible for the natural state of a water body. Target concentrations, also applied for artificial substances with no geogenic background concentration, are taken from the WHO drinking water standard35 for the reason that water pollution can contribute to scarcity if user’s requirements are no longer meet. To consider all users we see drinking water standards to be the most appropriate, as this ensures that there is no danger to humans and nature. The approach uses a mathematical case differentiation and has the advantage that no negative values are possible. It is modified in the following to be applicable beyond the LCA context for which it has been originally presented.
Water quality footprint of agriculture
Water pollution from agriculture follows a certain pattern that must be taken into account when calculating water quality footprints: in the beginning, there is the application of fertilisers, manure and pesticides to the field. Part of it is taken up by plants and extracted by harvesting, released into the atmosphere or is washed away with the surface runoff, the latter referred to as emission to surface water here. What remains is absorbed by the soil or as solution in the soil and is basically available for interaction with groundwater. In fact, however, only a small part of it, the emission to groundwater, actually reaches groundwater depending on substance-specific impact pathways that contain effects such as attachment to soil particles or degradation. Impact pathways are highly substance-specific, very complex and dependent on spatial and temporal conditions (e.g. Siebert et al.36, Borggaard & Gimsing37, Batjes et al.38, Sattari et al.39, Wick et al.40, Papadopoulos et al.41). The water quality footprint of agriculture summarises the emission to surface water and to groundwater. Existing approaches usually use leaching-run-off-rates to calculate the share of the application that ends up in water bodies16, referred to as loads. It is also common practice42 to translate loads into water volumes by dividing the loads by a certain concentration, e.g. substance-related geogenic background. In this way, water quality footprints can also be expressed in terms of water volume, i.e. the volume necessary to dilute the pollution down to the reference concentration, and can be used in the same way and in addition to quantitative water footprints of water withdrawals or consumption.
Substance-specific features
Within this study, water pollution from agricultural application of N, P, and G is considered. N and P are the most important plant macronutrients in terms of quantity (e.g. FAOSTAT Database), that are discharged to a large extent into inland and coastal water bodies19. The third macronutrient, potassium, often accounts for only a fraction of the total nutrients applied (e.g. FAOSTAT Database) and is therefore neglected in this work, as are micronutrients. The second source of agricultural water pollution are pesticides which comprise chemicals or microorganisms used to destroy or inhibit organisms or viruses that are harmful to crop growth. Due to the large number of pesticides, not all can be considered here. Based on its global importance20, G is chosen to demonstrate the approach and present first results. In the following, relevant characteristics of N, P and G are described and evaluated for the calculation approach.
N contributes to many different pools in the atmosphere, biosphere and hydrosphere and in soils in different chemical forms. Atmospheric N is in the form of inert, elemental N2 gas. Next to it there are different reactive forms of N, such as ammonia [NH3] and ammonium [NH4+], nitric oxide [NO], nitrogen dioxide [NO2], nitrous oxide [N2O], nitrate [NO3−] and nitrite [NO2−]43, which have strongly increased due to biological fixation of N2 through leguminous crops, combustion of fossil fuel and, in particular, production of synthetic fertilizer43. For plant nutrition, ammonium [NH4+] or nitrate [NO3−] are required. Also, organic N-nitroso compounds (R-N = O) are present in the environment. The following points are important to balance agricultural N applications: next to mineral fertilizer input, manure as well as biological fixation and atmospheric deposition contribute to the N input, whereas extraction by plants and denitrification, which is the anaerobic conversion of nitrate to N2, or N2O by soil bacteria and subsequent release into the atmosphere, are relevant for the output36. Additionally, emission to surface water by runoff and soil loss are considered in a balance44. The remaining quantity is either adding to the stock of N in the soils or is emitted to groundwater. Stock addition can be neglected due to high water solubility and low adsorption of N compounds. Hence, the remaining quantity can be considered as emission to groundwater. However, it is still important to note that during the residence time of the groundwater, some of the N is denitrified, which is taken into account in the calculation. In principle, all listed N compounds are water soluble and can contribute to the total N content of water bodies. Throughout this study, N in water is always reported as total dissolved inorganic N. The water quality footprint consists of the sum of emission to surface water and groundwater minus denitrification.
As regards P, monophosphine, PH3, is the only gaseous form of P, which can be neglected with regard to the whole P cycle. In the absence of an atmospheric cycle, plants can obtain P only from the soil38. There P can occur inorganically in primary P minerals, dissolved in the form of H2PO4− or HPO42− or organically in various forms. Plants must make bound P available as H2PO4− in order to be able to absorb it. Harvesting removes P from the soil indirectly, while surface runoff (emission to surface water) and soil loss remove it directly, which together would result in a negative P balance without external supply. Addition of P through manure or mineral fertiliser would avoid that the P is mined from the soil. Depending on their specific soil characteristics, P is easily available for plants or not and whether the soil storage of P is accumulating or not38. These effects are important from an agronomic perspective and have a noteworthy influence on the yield and the amount of P storage in the soil mainly determines the emission to water bodies. Previous work on P leaching to groundwater has concluded that leaching is negligible compared to surface runoff: next to discharge of untreated waste waters, the surface runoff from crop fields and pastures is the globally most important source of excess P in surface water bodies45. P leaching may occur in saturated soils, that have faced long-term over-fertilisation46, however, application has already been declining since the 1980s in countries with historical over-fertilisation, especially in Europe39. Other studies indicate that P leaching can be important in flat landscapes in the presence of certain subsoil properties47,48. But these are mainly findings from laboratory experiments or specific sites, so that a general indicator for all soil types has not been established yet. Hence, P leaching is neglected in this study and the water quality footprint considers only run-off emission to surface water.
As regards G, it is an artificial substance which means that there are no primarily natural cycles. The mechanisms and effects of G dissipation on certain compartments have variously been studied spatially and temporarily. However, we still know too little about the potential establishment of anthropogenic cycles and interactions (e.g. G as P source49). Consequently, inputs from sources other than direct application are neglected in this study. While highly complex and site specific, there are hints that G in general shows a similar behaviour than P with two exceptions: cultivated soils can be saturated with P, which is not likely for G, and G is also degraded by microbes37. To calculate the emission to surface water and groundwater, (bio)degradation, sorption and leaching to groundwater have to be considered as impact pathway.
To account for the complexity and high spatial dependence of the impact pathways, the emission to surface water and groundwater for N, P and G is determined from highly developed models with spatial resolution. These data are also associated with high uncertainties, but represent a suitable state-of-the-art basis that takes substance-specific characteristics into account and is continuously being developed. Our focus here is not the provision of data, but the description of a suitable methodology and representation of water quality footprints with available data.
Application of water scarcity footprint methodology
Agricultural water pollution through N, P and G is expressed in volumes of virtual water to dilute the emission to surface and groundwater. The VDV is calculated with demineralized water VDVdem by dividing the substance-specific load s in a catchment area i by the geogenic background concentration cgeo,s in case of naturally occurring substances or the target concentration ctarg,s in case of anthropogenic substances18. For the purpose of this study, the calculation is adapted: load s is renamed to loads,em to make it clear that it is the emission to surface and groundwater. VDVs are calculated at country level for the producer countries supplying the German bioeconomy, so that the numerator i stands for the individual countries here and the functional unit (FU) is m3 per country i. For N and P, the calculation follows Eq. (1), if cgeo,s ≤ ctarg,s, and Eq. (2), if cgeo,s > ctarg,s. For pesticides, the calculation follows Eq. (2):
$${{VDV}}_{{dem},s,i}[{m}^{3}]=\frac{{{load}}_{s,{em}}\left[{kg}\right]}{{c}_{{geo},s}\left[kg\,{m}^{-3}\right]}$$
(1)
$${{VDV}}_{{dem},s,i}\left[{m}^{3}\right]=\frac{{{load}}_{s,{em}}\left[{kg}\right]}{{c}_{{targ},s}\left[kg\,{m}^{-3}\right]}$$
(2)
The N loadN,em is calculated from a N cropland mass balance with the general equation mt=x+1 = mt=x + dm dt−1, where m represents a stock, t = x a certain point in time, here the beginning of an agricultural period, and t = x + 1 a later point in time, here the following agricultural period. The total N storage Nstor,t=x+1 is the sum of the storage Nstor,t=x resulting from the previous agricultural period, the fertilizer input Nfert,t=x+1, the input from livestock excretions Nex,t=x+1, atmospheric deposition Ndep,t=x+1 and biological N fixation Nfix,t=x+1 minus N extraction by harvested plant parts Nharv,t=x+1, ammonia volatilization Nvol,t=x+1, soil loss Nloss,t=x+1, denitrification Ndenit,t=x+1 and the loadN,res,t=x+1. Emission to surface water from runoff and emission to groundwater through leaching are both included in loadN,res,t=x+1. Converting the formula gives Eq. (3):
$${{load}}_{N,{em},t=x+1}= {N}_{{stor},t=x}-{N}_{{stor},t=x+1}+{N}_{{fert},t=x+1}+ {N}_{{ex},t=x+1}+{N}_{{dep},t=x+1}\\ +{N}_{{fix},t=x+1} -{N}_{{harv},t=x+1}-{N}_{{vol},t=x+1}\!-\!{N}_{{loss},t=x+1}-{N}_{{denit},t=x+1}$$
(3)
The global model IMAGE-GNM19 provides all relevant parameters for the years 1995 and 2020 on grid cell level (0.5 by 0.5 degree, Supplementary Table 1) except for denitrification which cannot be obtained for a single year directly from the current publication of IMAGE-GNM. It is calculated following van Drecht et al.50 under the assumption of steady state for every year from 1990 to 2010 (Supplementary Data 5). The average of the years 1990 to 2020 is taken as factor fdenit (Supplementary Table 1) to be multiplied with leaching. Country values are aggregated with the help of GIS-based zonal statistics.
The P loadP,em corresponds to the emission to surface water here, as emission to groundwater is neglected. It is taken from the global model IMAGE-GNM19 for the years 1995 and 2020. It is assumed that the surface runoff reaches surface water bodies in the year of application.
The G loadG,em is calculated from the gridded application rates according PEST-CHEMGRIDS20 for the year 2020 by multiplication with a mass fraction which is obtained from the USEtox® model21,22. The mass fraction is the proportion of G that is emitted into freshwater (surface water and groundwater) after transport, sorption and (bio)degradation in the soil. USEtox® provides mass balances for several thousand organic and inorganic substances on different scales considering interactions between indoor compartments, air, agricultural and natural soil, freshwater and coastal marine water by modelling fate and exposure. Actually, these results are further used to derive characterisation factors for the assessment of the effect of substances on humans and ecosystems in life cycle impact assessment modelling. Here, we use it to describe the impact pathway of G, because we consider it an advanced model for modelling the fate of chemicals that is the consensus method for determining human and ecotoxicity in life cycle impact assessment. As regards spatial resolution, USEtox® considers wind speed, precipitation, groundwater level and runoff on continental and sub-continental level and distinguishes between agricultural and natural soils51. Input data from IMAGE-GNM and USEtox® have not been validated in the course of this study. IMAGE-GNM has a good validation status on long time series for N and P concentrations in different river basins around the world, but it is not possible to validate the fluxes of surface runoff and groundwater flow19,46.
Country-level geogenic background concentrations cgeo,s for N and P are compiled using a literature research (Supplementary Data 6 and 7). Where country-level data are not available, continent-level data are taken, and if also not available, the global median is used. The target concentration ctarg,s is taken from the World Health Organisation drinking water standard35 where data from different countries were analysed to determine median values. In case of artificial substances, geogenic background concentrations do not play a role and no regional differences need to be considered. What is important instead is the toxicity to organisms reflected in the global threshold values from the drinking water standard. The drinking water standard is used for surface water and groundwater, because the different water bodies of a catchment are not treated separately from each other, but as a unit18. The target concentration is also used, if the geogenic background concentration is zero or greater than the target concentration.
The calculated dilution volume is presented in m3 per inhabitant of Germany and categorised as low (<4.6 m3 per German), medium (4.6 to 460) and high (>460) by comparing it with the GDW, the German direct drinking water withdrawal per German inhabitant (127 L d−1 in 2020 according to Destatis25, which equals 46 m3 a−1). This classification considers that usually more than 90% of the product water use in general are to be found in the upstream supply52. A virtual dilution volume of 460 m3 per German or smaller, compared to which the GDW of 46 m3 is one tenth, is consequently still in an acceptable range and marks the border to the category high here. It is important to note, that we compare purely calculated virtual volumes, which are not consumed in reality, to actually used volumes of water, such as the GDW. Thereby, we want to show comparatively how much water would be needed if water pollution were eliminated by dilution to make the extent of water pollution tangible. For comparison purposes, the dilution volumes of the three substances are discussed side by side, but in the overall view for the German bioeconomy, only the largest volume is listed, the so-called critical volume. The largest volume contains the smaller dilution volumes and dilutes all examined substances.
Although recommended, we refrain from using AWARE water stress factors, as the underlying dataset of water availability and water use refers to the year 2010. Instead, the indicator withdrawal-to-availability ratio (WTA), initially presented by Alcamo et al.23, is used. We use an own calculation of WTA with data from AQUASTAT24 to cover the years both 1995 and 2020 and to receive country level values. They are calculated as the total freshwater withdrawal divided by the total renewable freshwater resources per country (Supplementary Figs. 2, 5). Since the water footprint method presented here is based on an LCA approach, but no LCA is conducted in accordance with the relevant guidelines, we refrain from weighting by multiplication as usually recommended in LCA. Also, routinely performed weighting can be misleading depending on the scope of a study and is hence not recommended for all applications of the water footprint53. Hence, we conduct the water stress analysis by using a colour scheme8 to designate low, middle and high water stress levels besides the calculated virtual dilution volumes, where no water stress corresponds to 0 < WTA ≤ 0.1, low water stress to 0.1 < WTA ≤ 0.2, medium water stress to 0.2 < WTA ≤ 0.4 and high water stress to WTA > 0.4. We have merged the no stress and low stress category to present three categories in total.
Share of the German bioeconomy on agricultural water pollution
Country-level VDVs are multiplied by the share of the German bioeconomy on the agricultural production in a country. Agricultural production includes the eight primary crop categories paddy rice; wheat; cereal grains nec; vegetables, fruits and nuts; oil seeds; sugar cane, sugar beet; plant based fibres and crops nec. Shares are based on the share of raw material input into the German bioeconomy (RMI, including German domestic consumption and export) in a country’s agricultural production EXIOBASE for the years 1995 and 20208, and have been updated meanwhile according to a newer version of EXIOBASE. Shares are predominantly only available at the regional level with five large rest-of-world regions covering a large number of countries. Moreover, the shares refer to production quantities, while the total substance loads are related to area. With the help of agricultural production data according to quantities and areas and the total agricultural area of a country (FAOSTAT), shares of rest-of-world regions are scaled down to country level and all shares are recalculated to area-related shares. To do so, FAOSTAT agricultural data were assigned to the eight primary crop categories of EXIOBASE. As this study supplements an existing monitoring framework for the German bioeconomy with an indicator for water quality, it is calculated with an identical data basis, even though MRIOs with higher resolution are now available. Germany’s share of agricultural production in a country is related to the N, P or G emissions of the entire agricultural sector, thus neglecting the actual composition of imports by crop and the effects of crop-specific emission intensities. Disaggregation by crop can make the results bioeconomy-specific, but is not possible with the currently available database, as EXIOBASE only breaks down by eight crop classes and IMAGE-GNM by three, all of which are not bioeconomy-specific (such as biofuels). In the continuation of the monitoring programme, the data basis will be successively refined as planned and results with less uncertainty and higher resolution will be presented.
