Health benefits, ecological threats of low-carbon electricity

This work builds on an integrated hybrid LCA model, THEMIS (technology hybridized environmental-economic model with integrated scenarios), which has been developed to evaluate life cycle impacts of global electricity system scenarios. THEMIS represents the global economy in nine regions, incorporating regional adaptations of the ecoinvent LCA database for materials and selected manufacturing processes and the EXIOBASE multiregional input-output model for inputs of services. The 2 °C and baseline scenarios of the Energy Technology Perspectives of the International Energy Agency provided parameters describing region-specific technology performance and electricity generation mixes in 2010, 2030 and 2050. Prospective LCAs of material production from the New Energy Externalities Development for Sustainability (NEEDS) project were used to represent important technology improvements in the mitigation scenario. The electricity generation is modeled in the foreground based on original inventories collected by a team of experts under the auspices of the International Resource Panel [17,18]. A key feature of an increasingly clean power system is that more of the environmental impacts occur in the construction of the power system, and if needed, the provision of the fuels. Changes in the power system reduce the pollution caused in particular in manufacturing processes, creating a virtuous circle. We capture the positive feedback of clean electricity on the construction of the power system through integrating the foreground life cycle inventories of electricity production into the background by replacing electricity in ecoinvent and EXIOBASE. Thus, the model captures important improvements both in material production and in the electricity used in manufacturing. The THEMIS model is documented in a separate method paper [16] and has been applied in a number of assessments [19–26].

Life cycle inventories for solar technologies (specifically photovoltaics and concentrating solar power), hydropower, wind power, natural gas, and coal power, were collected by a team of experts under the auspices of the International Resource Panel. These detailed bottom-up life cycle inventories of electricity generation technologies tally up emissions and resource use caused by the manufacturing of power plants and associated infrastructure, the production of fuels, and the operation and dismantling of power plants. The 450 page report of the Resource Panel offers a description of the technologies, the inventories, and further assessments, e.g. of the scientific literature on ecological impacts. We supplemented the original technology portfolio with data on nuclear power [26] and biopower based on forest residues and short-rotation coppice [27,28]. Nuclear power inventories were adapted from ecoinvent 2.2 [29]. Biomass feedstocks can be classified in four main economic categories, from lower to higher costs: wastes (e.g. organic waste, manure), processing residues (e.g. timber residues, black liquor), locally collected feedstocks (e.g. agricultural and forestry residues, energy crops), and internationally traded feedstocks (e.g. roundwood or biomethane). These feedstocks may undergo pretreatment to improve transportation and conversion processes, such as drying, pelletisation, briquetting, torrefaction, pyrolysis, or hydrothermal upgrading [30]. Due to limited life cycle data available on each of these options, we principally modeled two types of biomass feedstocks used for electricity generation: forest residues and lignocellulosic biomass from short rotation wood crops. Across all regions and years, we assumed a fifty-fifty split between these two biomass feedstocks. In energy scenario literature, dedicated lignocellulosic, woody or grass-type energy crops is generally expected to be the most important type of biomass feedstock in the future. Agriculture or forestry residues, are often important, but their use varies across models. First generation energy crops, including sugar cane and palm oil crops, play only a small role in long-term scenarios [31,32]. We assumed that short rotation wood crops is overall representative for lignocellulosic energy crops in general, a simplifying assumption that is also made in some energy scenario models (table S1 available at stacks.iop.org/ERL/12/034023/mmedia in electronic supplement of Rose et al [31]). For forest residue biomass, life cycle inventories were adapted from Singh et al [27]. For crop-based biomass, we used data on the amount of diesel, nitrogen, phosphorus and potassium fertilizer, chemicals and irrigation for existing bioenergy crops [28]. We also included inputs of diesel, fertilizer and chemicals to the production of cuttings [33]. For emissions of nitrogen compounds from crops, we assumed the following factors: 0.016 kg N₂O to air [34], 0.05 kg ammonia (NH₃) to air [35], 0.003 kg NOx to air [35], and 0.3 kg nitrate (${\rm NO}_3^{\minus}$) to water (derived from [34,36]) per kg of N fertilizer added. Assumed emission factors for phosphorus compounds were: 0.5 kg phosphate and 0.2 kg particulate phosphorus to water per hectare per year, based on values reported in [35,36]. We treated the use of herbicides and pesticides as emissions to agricultural soil. Fuel and auxiliary input requirements and emissions associated with the operation of biomass power plants to produce electricity was modelled based on ref [27]. For biopower, average global yields for the modelled energy crops were taken to be 190 GJ ha⁻¹ yr⁻¹ in 2010, 400 GJ ha⁻¹ yr⁻¹ in 2030, and 500 GJ ha⁻¹ yr⁻¹ in 2050, based on the most optimistic assumptions found in the literature (see figure 3 in [37]). Sensitivity analysis on feedstock mix is provided in the supplementary information (figure S3).

The call for indicators aggregating different environmental mechanisms reflects the limits of considering many factors in decision making [38,39]. Further, environmental impacts are only one aspect of technologies, along with cost, ease of operation, reliability, etc. Yet, scientists have been slow to embrace a comprehensive indicator, citing imperfection in knowledge that requires value judgments to resolve, and the incommensurability of risks or damages that affect different individuals, groups, species or ecosystems [40]. Typically, economists monetize ecosystem damages through estimates of external costs caused by pollution. In contrast, environmental scientists model the strengths of different environmental impacts and quantify their contribution to a common indicator, such as human health and ecosystem quality [41,42]. While both approaches have been used in life cycle assessment (LCA), we prefer the latter approach, which is more accepted in LCA and subject to major research efforts. In this paper we applied ReCiPe [14], a widely used method for life cycle impact assessment, with many person-years of dedicated development effort. Many energy LCAs [16] rely on ReCiPe midpoint indicators, which express the contribution of product systems to a large set of environmental mechanisms (also called impact categories). Human health impacts of energy LCAs were previously analyzed by ref [43]. This letter reports the first assessment of potential damage to both human health and ecosystem quality caused at the endpoint level of all major electricity generation technologies, as well as global power system scenarios.

The ReCiPe indicator for damage to human health incorporates the aggregate effects of the following environmental mechanisms: air pollution, human toxicity (via carcinogenic and non-carcinogenic damage), ionizing radiation (carcinogenic and hereditary effects), and climate change (table 1). The term 'air pollution' represents the effects of particulate matter formation (inhalation exposure to particulate matter in the air), photochemical oxidant formation (inhalation exposure to ozone and other oxidants), and ozone depletion (exposure to increased UV radiation). Human health damage was measured in disability-adjusted life years (DALYs), which combine years lost due to premature mortality, and years lived with a disability, or in poor health [14]. The ecosystem quality indicator was calculated from aggregated species diversity effects of the following environmental mechanisms: terrestrial, freshwater and marine ecotoxicity (increased concentration of toxic chemicals), terrestrial acidification (change in base saturation), freshwater eutrophication (algae growth, hypoxia of aquatic milieus), and climate change (temperature increase and loss of species). The unit for damage to ecosystem quality is species.yr, derived from the potentially disappearing fraction (PDF) approach [44]. PDF is a measure quantifying the fraction of today's present species that will potentially become extinct in a specific geographical location due to an emission or anthropogenic intervention. In that, it is a measure for loss of species richness, i.e. potential species extinction. PDF includes losses that happen right after the intervention, and also time-integrated damages. For example, a pulse of CO₂ emissions, still leads to species loss after 100 years. In ReCiPe [14], PDF was developed to species.yr, to have damages in absolute terms. The indicator species.yr is based on the species richness of different environmental compartments (freshwater, marine, terrestrial), which allows the damages in these compartments to be combined. ReCiPe is based on global species densities for these different environmental compartments, which are multiplied by the damage in PDF gives a weighted damage over all species in all compartments (assuming equal weight for all species). For the sake of legibility in presenting the final unit results, midpoint indicators with an endpoint characterization factor were aggregated into six distinct groups, as shown in table 1.

Scenario assessment results were based on vintage capital modelling, as in [16,45]. The electricity system life cycle inventories were broken down by life cycle stages: construction, operation and maintenance, and decommissioning. For every year from 2010 to 2050, the total environmental impact from the electricity sector was calculated as the sum of the environmental impacts of capacity increase (construction), the operation of existing plants, and the repowering of retired plants. The capacity figures were derived from the IEA's Baseline and BLUE Map scenarios' data on power plant installed capacities (table 2) [13]. Combining these capacity values with the lifetime of the various technologies, we were able to derive the capacity increase, operation, decommissioning and repowering rates for each technology and region. Finally, the indicators were all scaled to 100 in 2008 to show their relative variation until 2050.

The high ecosystem impact of biopower, with land use largely offsetting the benefits of CO₂ emission mitigation, is a novel finding. We hence would like to discuss bioenergy in more detail.

We do not account for potential land use change related emissions or the difference in timing between the emission of CO₂ and its uptake [52]. Life cycle GHG emissions of our biopower systems come mostly from fuel production and harvesting. Disregarding the impact of climate change, we find that the combustion-related emissions from fossil power and biopower cause larger human health impacts than technologies that do not rely on combustion for primary power generation.

Biopower shows a relatively high contribution to the ecosystem damage indicator from terrestrial ecotoxicity. The metolachlor used as herbicide during the agricultural phase is responsible for most of this contribution (more than 90%), due to a substantially larger ecotoxicity characterization factor. If this herbicide use is avoided, as it is for most of the biomass sources involved in the survey underlying our data [28], the ecotoxicity levels of biopower would fall below those of its fossil counterparts. Although classified as a potential carcinogen by the U.S. Environmental Protection Agency [53], metolachlor is widespread in the United States. The actual toxicity of metolachlor on humans and ecosystems is still being analyzed [53–56], and actual effects are difficult to assess because acute toxicity mostly occurs in combination with other substances [57].

Regarding land occupation, biomass plantations require about two orders of magnitude more surface area than any other technology, although the impact of this land use is less per unit area than that of coal mines. The high pressure of agricultural systems on ecosystems through land occupation is reflected in the endpoint indicators: the potential damage to ecosystems from biopower is comparable to the impact of coal power through climate change. One should note that land transformation (land use change) and land occupation (land use) are different impacts, with different consequences. Occupation impacts are impacts caused by ongoing land use, thus maintaining a changed ecosystem quality in comparison to a reference state, which effectively accounts for the delay in potential recovery. Transformation impacts relate to the one-time event of land use change, which in ReCiPe is modeled as the change in species richness during a recovery period [58].

We found that the choice of biomass feedstock supply considerably alters land occupation impacts of bio-sourced electricity generation. Depending on the feedstock, and its assumed energy yield, these impacts may be as large as 0.46 m²a kWh⁻¹ for crop-based biomass with CCS if implemented today, to as little as 0.06 for forest residues and 2050 efficiencies. At a global level, the feedstock mix influences significantly the stress on land occupation (figure S3). For forest residues as feedstock, we did not account for the land occupation associated with the forest area from which the residues are sourced because timber is the main output of forestry, and the choice to harvest forest residues (in addition to timber) does not increase the forest area needed. Implementing methods for assessing the ecological impact of removing this extra biomass from the forest would require more information on the operations on the ground [59], but could potentially nuance our results. Adopting an endpoint perspective on the deployment of low-carbon electricity generation—that is, focusing on the ultimate damage to ecosystems and human health—shows that land occupation may become a major contributor to the threat to ecosystems. This finding is not surprising, because land use is already one of the main drivers of global biodiversity loss [60].

In general, methods for quantifying impacts on ecosystem quality in LCA are fast developing, including the introduction of additional stressors, finer spatial detail and inclusion of taxa-specific characteristics or vulnerabilities. As is also the case for this study, land use very often turns out to be a dominant driver of impacts on biodiversity and therefore deserves special attention. In the last decade numerous methods for quantifying impacts from land use have been proposed in an LCA context, ranging from methods specific to individual countries or taxonomic groups to methods that are applicable on a global level and for multiple land use types, spatial levels and taxonomic groups. Curran et al [61] provide an overview of 20 land use models developed for LCA assessments. The dominant metric for current land use methods is 'species richness' (16 of the models in Curran et al), which restricts the assessments to changing species numbers, but does not include other information, such as abundance or vulnerability. Most of the currently used and proposed land use methods are related to endpoint indicators, very frequently the potentially disappeared fraction of species (PDF). However, discussions whether absolute species losses (instead of relative ones) would not be more meaningful are ongoing [58], especially because losses at local and global levels can be wrongly assumed to be the same [42]. Currently, midpoint indicators used for land use assessments are usually restricted to the quantification of the amounts of land used or transformed, which the inventory parameters [58]. While this is easy to quantify, it does not reflect impacts very well.

The UNEP-SETAC life cycle initiative has recently made a first recommendation for a land use assessment method, which is quantifying impacts for 6 different taxonomic groups for 6 different land use types in all 825 terrestrial ecoregions of the world and includes the vulnerability of species. However, the method is also only developed on an endpoint level and, due to its novelty and therefore limited experience in test cases, recommended for hotspot analyses only [62].

Also, while spatial aspects have gained attention in recent years, this is much less the case for temporal aspects, such as dependencies on the timing of harvesting of agricultural crops [58].

The methods we employed to quantify environmental impacts are state-of-the-art, however, the characterization factors used to convert life cycle inventory values to environmental impact scores would be more accurate if they were spatially explicit. The characterization factors represent an average of all global ecosystems and habitats, but the impacts depend on local circumstances. In emerging spatially-explicit impact assessment methods, such as LC-Impact [63], spatial differentiation includes ecoregion, watershed or even pixel level detail, and leads to widely varying characterization factors due to differences in ecosystem sensitivity or species richness [59]. For example, the ecosystem impacts of land use can vary by up to four orders of magnitude among the various ecoregions and land use types [64]. This wide range leads to a larger variation in results. However, the poorest ecoregions are not very productive, so the actual impact for realistic biomass supply scenarios will not vary as much. These emerging spatially explicit assessment methods could not be applied to the present inventory results because the energy scenarios that are available do not specify the location where biomass is harvested. Employing spatially explicit impact assessment would require systematic spatial detail in energy scenarios and life cycle inventory data. Given the importance of land use for the ecosystem impacts of future energy systems, it would be pertinent to develop such scenarios and inventory data in order to explore the potential of growing biomass in areas of lower ecosystem diversity, which again can be used to derive policies that ensure that feedstock will in fact be sourced from such low-impact regions. Such scenarios would consider a wider variety of biomass sources and conversion technologies.

Our LCA results are robust with respect to the substantial co-benefits of replacing coal and gas with solar, wind, hydro and nuclear power for both human health and ecosystem quality. These co-benefits are a result of the significant air and water pollution caused by extraction, transport and combustion of coal and gas, as well as the substantial land use associated with coal mining. These patterns were already suggested by analyses of individual environmental mechanisms [16] or individual pollutants [8–10] rather than integrated endpoints. High non-CO₂ pollution impacts for CCS have been consistently found in the LCA literature [65–67], although other works suggests that impacts may be comparable [11]. Our work confirms these findings but also indicates that the avoided climate change impacts are larger than the additional non-climate pollution impacts.

There are significant trade-offs for ecosystem quality associated with biopower and smaller trade-offs associated with pollution from fossil power with CCS. From a co-benefit perspective, the reliance on a large-scale utilization of biopower, not least to achieve negative emissions after 2050, appears to be a weak point of present mitigation scenarios [68]. To understand the potential ecosystem damage of increased land occupation and changed emissions better, future research should develop scenarios for the location of biomass plantations and power plants, allowing the application of methods for site-specific impact assessment [64]. A larger variety of biomass supplies may be explored, with value-chain specific ecological impacts. Our present analysis considers a combination of forest residues and fast rotation croppies, which have relatively small land use impacts compared to conventional forestry or agriculture. To limit ecosystem quality impacts, policy makers should seek technology and management options for ensuring a biomass supply with lower than average land use impacts. Our findings also provide a rationale for investigating alternative carbon-negative technologies that lead to lower land use impacts.

The findings of our work should put to rest residual myths about adverse health and ecosystem impacts associated with the high energy use and material requirements of producing and installing solar and wind power plants and put in perspective the health impacts associated with ionizing radiation from nuclear power. Adopting the right mix of low-carbon technologies for electricity generation brings multiple benefits to human and ecosystem health while having the potential to stabilize global temperature.