From toxic releases to damages on human health:a method for life cycle impact assessment, with a case study on domestic rainwater use

Life Cycle Assessment (LCA) is a tool developed to evaluate the environmental impact of a product or a system. After a decade of research in the LCA field, significant progress has been achieved but methodologies for the assessment of toxicological impacts on human health are still in the development phase. This dissertation contributes to the research required in this field. More specifically, its main objective is to develop a Life Cycle Impact Assessment (LCIA) procedure for human health respecting the guidance developed under the umbrella of the Society of Environmental Toxicology and Chemistry (SETAC). This means that we aim to implement an original procedure to quantify the potential carcinogenic and noncarcinogenic effects of toxic releases on human health (chapters 2 and 3), and to develop a new method describing the fate of atmospheric releases and the resulting exposure on humans (chapter 4). A framework summarized in figure 5.1 is also proposed to combine the effect assessment with the fate and exposure assessment, in order to derive a so-called human damage factor (chapter 5). A set of heavy metals (cadmium, chromium(VI), chromium(III), copper, methylmercury, beryllium, lead and inorganic arsenic) and of criteria air pollutants (CO, SO2, NOx and fine particles) is chosen for a full application of the procedure developed in this dissertation. The use of this procedure to the Cycleaupe case study is also part of the objectives of this research. This study aims to determine whether systems using rainwater or reducing water consumption are "friendlier" from an environmental perspective than conventional toilet flushing (chapter 6). Figure 5.1. Overview of the framework proposed in this thesis for assessing the damage induced on human health by a toxic released into air. In chapters 2 and 3, a new paradigm based on the effect dose ED10h is derived from the Risk Assessment concept of benchmark dose. It is proposed and explored for the first time in LCIA. The ED10h is defined as the best estimate of the dose which induces a 10% added risk over background for humans. Carcinogenic and noncarcinogenic risks towards humans are characterized by drawing a straight line from the ED10h down to the origin of the dose-response function. The slope of this straight line is called the slope factor and is denoted βED10. The linear dose-response function without threshold, which is assumed in this ED10-approach, is discussed. The ED10h is calculated for chemicals with bioassay data available in the Integrated Risk Information Service (IRIS) database provided by the US Environmental Protection Agency (US EPA). New correlations between the ED10h and the more widely available tumor dose TD50a (for carcinogenic effects) and the No Observable Adverse Effect Level NOAEL (for noncarcinogenic effects) are determined. They are applied to quantify the slope factor of more than 900 chemicals. A weighting of the different health outcomes associated with chemicals is proposed, based upon the Disability Adjusted Life Years per affected person (DALYp) concept. For carcinogenic endpoints, the DALYp is calculated for different types of tumors, using data reported in the literature. This shows that all cancers have more or less the same severity and an average DALYp of 11.1 years of life lost per affected person is derived. For noncarcinogenic effects, a simplified classification of the adverse effects into three categories is chosen and a DALYp of 11.1, 1.1 and 0.11 years of life lost per affected person is respectively assigned to each of the three categories. Finally, the slope factor βED10 and the DALYp for each substance are combined together in an original way to derive its effect factor. This effect factor is expressed in years of life lost per absorbed mass. Appendix 1.1 summarizes the effect factors calculated for more than 900 toxic releases. Effect factors for carcinogenic outcomes range from 1.3·10-9 for cinnamyl anthranilate up to 3.4·10-1 [yr lost / mg absorbed] for 2,3,7,8-tetrachlorodibenzo-p-dioxin. Effect factors for noncarcinogenic endpoints range from 4.2·10-12 for 1-Chloro-1,1-difluoroethane to 1.4·10-3 [yr lost / mg absorbed] for beryllium. In chapter 4, a semi-empirical approach is developed to evaluate the fate and exposure for atmospheric releases of metals, carbon monoxide (CO), sulfur dioxide (SO2), nitrogen oxides (NOx) and fine particles. For that purpose, we apply for the first time in LCA the concept of exposure efficiency, which is defined as the ratio between the dose absorbed by the population and the emission inducing that absorption. Three types of exposure efficiency are defined for a world release into air of a given compound. A specific exposure efficiency is directly based on the rural and urban concentrations inhaled by humans. A continental exposure efficiency is defined by considering an uniform world continental concentration over urban and rural inhabited regions (marine and desert regions are excluded). A global exposure efficiency issimilarly defined from the global world concentration of a substance. Exposure efficiencies are calculated for fine particles, CO, NOx and SO2. The specific exposure efficiency ranges from 3.9·10-6 to 2.4·10-5 [mg absorbed / mg emitted], demonstrating that only a very small fraction of an air release is inhaled by humans. The exposure efficiency for metals after inhalation is assumed to be equal to the exposure efficiency for fine particles, since airborne metals are attached to particulate matter. If atmospheric deposition on an agricultural soil occurs, humans can be exposed through a transfer into food products. A first evaluation of this transfer indicates that it can increase the exposure efficiency of metals released into air by a factor 5 up to 70. Specific exposure efficiencies are selected in this thesis to describe the fate and exposure of atmospheric releases. We show for the first time that specific exposure efficiencies are higher by a factor 3 than continental exposure efficiencies, indicating that the use of one-box continental models tend to underestimate the exposure efficiency that can be expected in the real world. This is due to the fact that higher emissions occur in highly populated regions. As a first approximation, the factor 3 could be used as a corrective factor to derive the specific exposure efficiency from the exposure efficiency predicted by one-box continental models. In chapter 5, exposure efficiencies presented in chapter 4 and effect factors presented in chapters 2 and 3 are multiplied to derive the so-called Human Damage Factors (HDF). The damage factors are expressed in years of life lost per emitted mass. Using that factor, the emission of a substance can be converted into its potential damage induced on humans. The damage factors are calculated for NOx, SO2, CO and fine particles, as well as for the selected set of metals released into air or into agricultural soils (see appendix 1.2 for the summarized results). When the transfer into food products is not accounted for, the damage factors for the studied metals range from 1.7·10-11 for chromium(VI) up to 1.3·10-8 [yr lost / mg emitted] for beryllium. Lead has the highest damage factor (1.9·10-8 [yr lost / mg emitted]) if transfer into food products is considered. Damage factors ranging from 2.7·10-10 to 6.6·10-10 [yr lost / mg emitted] are found for NOx, SO2 and fine particles, while carbon monoxide is characterized by a damage factor 103-folds lower. Per emitted mass, metals inhaled by humans induce damages of the same order of magnitude than NOx, SO2 and fine particles; when atmospheric deposition on agricultural soils and its subsequent transfer into food are accounted for, metals present higher damage factors. An indirect validation of the damage factors is presented for SO2, NOx, CO, fine particles and some metals, by applying their damage factors to their total emissions over Switzerland and Europe. The evaluated damages are plausible and in accordance with results reported in other studies. In chapter 6, a Life Cycle Analysis is performed to compare five scenarios for toilets flushing. This LCA is the first one carried out on the whole water cycle, including both thesimilarly defined from the global world concentration of a substance. Exposure efficiencies are calculated for fine particles, CO, NOx and SO2. The specific exposure efficiency ranges from 3.9·10-6 to 2.4·10-5 [mg absorbed / mg emitted], demonstrating that only a very small fraction of an air release is inhaled by humans. The exposure efficiency for metals after inhalation is assumed to be equal to the exposure efficiency for fine particles, since airborne metals are attached to particulate matter. If atmospheric deposition on an agricultural soil occurs, humans can be exposed through a transfer into food products. A first evaluation of this transfer indicates that it can increase the exposure efficiency of metals released into air by a factor 5 up to 70. Specific exposure efficiencies are selected in this thesis to describe the fate and exposure of atmospheric releases. We show for the first time that specific exposure efficiencies are higher by a factor 3 than continental exposure efficiencies, indicating that the use of one-box continental models tend to underestimate the exposure efficiency that can be expected in the real world. This is due to the fact that higher emissions occur in highly populated regions. As a first approximation, the factor 3 could be used as a corrective factor to derive the specific exposure efficiency from the exposure efficiency predicted by one-box continental models. In chapter 5, exposure efficiencies presented in chapter 4 and effect factors presented in chapters 2 and 3 are multiplied to derive the so-called Human Damage Factors (HDF). The damage factors are expressed in years of life lost per emitted mass. Using that factor, the emission of a substance can be converted into its potential damage induced on humans. The damage factors are calculated for NOx, SO2, CO and fine particles, as well as for the selected set of metals released into air or into agricultural soils (see appendix 1.2 for the summarized results). When the transfer into food products is not accounted for, the damage factors for the studied metals range from 1.7·10-11 for chromium(VI) up to 1.3·10-8 [yr lost / mg emitted] for beryllium. Lead has the highest damage factor (1.9·10-8 [yr lost / mg emitted]) if transfer into food products is considered. Damage factors ranging from 2.7·10-10 to 6.6·10-10 [yr lost / mg emitted] are found for NOx, SO2 and fine particles, while carbon monoxide is characterized by a damage factor 103-folds lower. Per emitted mass, metals inhaled by humans induce damages of the same order of magnitude than NOx, SO2 and fine particles; when atmospheric deposition on agricultural soils and its subsequent transfer into food are accounted for, metals present higher damage factors. An indirect validation of the damage factors is presented for SO2, NOx, CO, fine particles and some metals, by applying their damage factors to their total emissions over Switzerland and Europe. The evaluated damages are plausible and in accordance with results reported in other studies. In chapter 6, a Life Cycle Analysis is performed to compare five scenarios for toilets flushing. This LCA is the first one carried out on the whole water cycle, including both the water supply and the wastewater treatment. The drinking water supply system, the rainwater recuperation system and the wastewater treatment system are included in the system boundaries. Results demonstrate that economic toilets (3.5 [l/flushing]) lead to a significant reduction of the energy requirements compared to conventional toilets (9 [l/flushing]). A conventional water supply and a rainwater recuperation with a storage tank of 10 m3 are characterized by similar energy consumption. A rainwater storage tank of 20 m3, designed to be completely independent of the conventional water supply system, is energetically disadvantageous. Calorific losses, linked to the temperature increase of flushing water within the house, have a significant contribution to the energy requirement. The advantage of economic toilets is confirmed when looking at the inventory emissions. An initial LCIA was performed using the critical surface-time CST95 method of Jolliet and Crettaz [1997]. It showed that the conventional scenario using economic toilets (CONVeco) is the most advantageous for all impact classes. When applying the human damage factors developed in this thesis (see chapter 5), the conventional scenario (CONVeco) is still characterized by lower impacts on humans than the recuperation scenario (REC10eco). However, the substances having the major effect on human health differ from those found with the CST95 method; reasons for that change are discussed