In the following sections, we analyze how reference situations are used in current life cycle impact assessment models addressing biodiversity. First, it is discussed how reference situations fit in the established LCA framework. In the second part, the practical challenges associated with applying these reference situations are discussed.
3.1 Reference situations in the LCA framework, a theoretical background
The first studies discussing the application of a reference situation for biodiversity in LCA of products state that the reference for land use should describe the situation in the absence of the studied product system (Weidema and Lindeijer
2001; Milà i Canals et al.
2006,
2007; Michelsen
2008), i.e., a baseline reference situation that describes a human-free situation against which the impacts of the studied land use are assessed. The rationale behind this lies in the fundamental goal of LCA, which is the compilation and evaluation of the inputs, outputs, and the associated potential environmental impacts of a product system (covering both goods and services) across the whole life cycle. Its results are used to aid in decision-making, choice of environmental performance indicators, and support market claims (ISO
2006), often aiming at improvement in the system (Ekvall and Tillman
1997; Baumann and Tillman
2004). In order to achieve these fundamental goals, LCA divides the world into a technosphere and an ecosphere in which changes to the ecosphere (as a result of activities in the technosphere) can be considered unintentional “man-made” consequences (Hauschild et al.
2017). However, the location of the boundary of these two spheres is quite abstract and therefore often debated with regard to agricultural systems for example (Hauschild et al.
2017). Yet, as this is the way LCA is constructed, several authors point out that including parts of the technosphere in the reference situation should preferably be avoided in attributional LCA (Soimakallio et al.
2015; Brander
2016). Herewith, they exclude reference situations that are not completely natural, like current land use mix as proposed by Koellner and Scholz (
2008).
The reasoning above explains the preference for baseline or other natural reference situations in current LCA practices, i.e., a historical “natural” situation, a natural counterfactual situation, or a re-naturalization situation.
When using historical baseline references, usually a point in time is selected to compare against current conditions to. Without a sufficiently distant past, however, these references may fail to fully inform about past human impacts. Choosing a timeframe is therefore not straightforward; how far back in time do we have to go? Obviously, going back to times in which the wooly mammoth was still around, some 10,000 years ago, would not appear be relevant for current assessments of current land use impacts, but identifying a closer “pristine nature” time period might be challenging, as human land use history varies strongly from region to region (e.g., Hilding Rydevik et al.
2018). To be able to compare impacts of land use in different landscapes, time frame and level of degradation of the reference should be equal, if not to be compromised (Nielsen et al.
2007).
The change of ecosystems without human interventions is normally neglected when natural baselines are used. In principle, the effect of land interventions on biodiversity can only be explained if the expected trends of biodiversity in the absence of these interventions are known, thus if a natural counterfactual reference situation is used. In landscape-level ecological research, these types of expected patterns on landscape scale have been used and termed
neutral landscape models (NLM) in the tradition of neutral or null models in ecology (Ricotta et al.
2002).
Potential natural vegetation (PNV), which has been used in LCIA models, is such a neutral model, developed to express the biotic potential of a region with regard to all site factors relevant for vegetation development (Ricotta et al.
2002), thus the hypothetical natural status of vegetation that could be outlined for the present time or for a certain earlier period, imagining that the area would never have been subjected to human interventions (Tüxen
1956). PNV is a much-debated concept with many different interpretations (Curran et al.
2016). It has been interpreted both as a future hypothetical state after all human interventions had stopped (a re-naturalization reference), and as a pre-anthropogenic disturbance state, i.e., some kind of “climax” natural state (a historical reference) (Chiarucci et al.
2010). In the LCA literature, the latest recommendations point out that the appropriate definition should be “hypothetical biotic potential” of a region based on patterns in existing remnants of maximally undisturbed vegetation in which case existing remnant vegetation must act as a source for colonizing species following abandonment and succession (Curran et al.
2016). This makes the PNV a re-naturalization reference, rather than a natural counterfactual reference situation.
Re-naturalization has been identified as the most suitable reference situation for use in attributional LCA by several authors (Milà i Canals et al.
2007; Soimakallio et al.
2015; Curran et al.
2016; Koponen et al.
2018), as land occupation postpones natural regeneration of the land and the fact that the resulting final ecosystem could be of a different “quality” than the original. These impacts are permanent and need to be accounted for (Weidema and Lindeijer
2001) (Fig.
1). However, by solely using a re-naturalization reference situation, we would only catch how much impact additional land use would cause. Earlier land use impacts would be impossible to calculate without comparison to the situation as it was before the land use started (historical) or the counterfactual reference situation. Mathematically, in an assessment using only a re-naturalization reference situation to quantify biodiversity impacts, further degradation of the land due to changed land use management might actually result in a lower estimated impact if the degradation in question limits the capability of the land to regenerate (Soimakallio et al.
2015). In the same way, use of land with a high potential to regenerate would cause a higher impact than the same usage of land with a lower regeneration potential (although in some cases, this effect might be compensated for by a higher production ratio). To prevent this potential “gap” in impact results, it would be of importance that an assessment that quantifies land occupation impacts related to a re-naturalization reference also includes the quantification of permanent impacts (related to a baseline reference as natural situation or natural counterfactual situation). In other words, complying to the UNEP-SETAC framework for land use impacts in LCA would mean that two reference situations would have to be established: both a natural situation or natural counterfactual situation as well as the re-naturalization reference. The use of a double reference situation to calculate land use impacts is, however, rarely seen in impact assessment models for biodiversity (or any other land use impacts), as it is far from practical considering the difficulties and uncertainties involved in applying reference situations, as described in the section below. We thus discourage the distinction between occupational and permanent biodiversity impacts unless a double reference situation can be established. As a re-naturalization reference situation by principle only catches occupational impacts and neglects permanent impacts, a re-naturalization reference is unsuitable in assessments using a single reference situation. There are, however, examples where the natural or baseline reference situation is implicitly present. For example, in the model by Curran et al. (
2016) based on conservation status of species (the IUCN red list of threatened species), a baseline reference situation is applied due to the use of contemporary threat/rarity data. In this case, the application of a re-naturalization reference situation allows for assessing the potential permanent impacts expressed as the extinction risk of species.
A fourth group of reference situations used in LCA, although considerably less often so than the baseline references discussed so far, are the target reference situations. Several authors point to the reference situation as a possible benchmark of a desired direction for ecosystem management, improvement, or restoration (Curran et al.
2016; Teixeira et al.
2016; Verones et al.
2017). Some authors argue that regardless of which reference system is chosen, the reference situation reflects per definition a desired situation (Cao et al.
2017), others, like us, regard a target situation as one of multiple types of reference situations, besides, e.g., a (hypothetical) situation representing conditions in the absence of human intervention or a current situation (Frischknecht et al.
2016b). This view is motivated by the fact that a specific type of natural reference situation (i.e., a human free reference or re-naturalization reference) will not represent a desired situation in all cases. For example, the re-naturalization reference does not address active land restoration or other forms of human land management in habitats depending on this (Souza et al.
2015). Lindner et al. (
2014) proposed a biodiversity impact assessment model applying the desired (target) state of biodiversity as defined in national strategy documents as the reference situation. The model has been applied by, e.g., Lindqvist et al. (
2015), and further developed by Winter et al. (
2018). In the case of Lindqvist et al. (
2015), it should be noted that the advocated desired state ended up being interpreted as a maximum natural value, which illustrates the earlier stated fact that target references can take different shapes. We elaborate on the differences between baseline reference situations and target reference situations in Sect.
4.
3.2 Challenges in applying baseline reference situations
Going beyond theory, successfully applying a reference situation into LCA practice, means that the reference situation has to be translated into a set of functional reference conditions, a set of attribute values or quantifiable characteristics of the reference situation (Miller et al.
2012). These types of attributes will depend on the indicators chosen to assess biodiversity and can be physical, chemical, or biological parameters of organisms, ecosystem functions, or structures, and could be represented by single values or a distribution (Miller et al.
2012).
In LCIA, there is a wide variety of indicators applied, see, e.g., Curran et al. (
2011), Michelsen and Lindner (
2015), Souza et al. (
2015), and Winter et al. (
2017) for reviews. In brief, the most frequently used biodiversity indicator in LCIA models is α-diversity (number of species in an area) calculated by use of a species-area relationship (SAR) model. Lately, the diversification of indicators for capturing biodiversity has increased, as has the tendency to recommend multiple indicators. The list of suggested indicators in LCIA now includes ecological scarcity (ES), ecological vulnerability (EV) and structural indicators (Michelsen
2008), functional diversity (Souza et al.
2013), and indicators based on expert knowledge (e.g., Jeanneret et al.
2014; Lindner et al.
2014; Lindqvist et al.
2015). Indicators capturing the genetic level of biodiversity remain few (Winter et al.
2017).
The reference situation is decisive for what indicators can be applied, as data must necessarily be available for the reference situation. Irrespective of LCIA model and indicator selected, finding reliable data for baseline situations is often complicated, and more so when that baseline is not integrated in the indicators used but rather relying on the assessment of actual sites representing the baseline reference situation. Such sites are difficult to find in many parts of the world, as few or no areas remain that have not been under human influence of some kind. Therefore, the sites chosen are often those considered to be “the best of what’s left.” As a result, if we aim for natural situations, we often end up with a reference that departs from the natural state by some unknown amount. The struggle to find a baseline reference situation that is measurable shows clearly in current LCA studies, where reference situations as potential natural vegetation, pre-anthropogenic disturbance state, and relaxation potential are all approximated with data of semi-natural land use in the region (Weidema and Lindeijer
2001; De Schryver et al.
2010; Souza et al.
2013). Due to the varying departure of semi-natural from the natural, semi-natural conditions cannot deliberately be used as a proxy for the natural (FAO
2002).
In many cases where we desire naturalness, we have to accept that in some places, the baseline has shifted and that we have lost the chance to conserve (or even identify the loss of) the most vulnerable habitat or species. There is then no other option than to estimate the reference situation by quantifying the biological condition at a set of sites that are either minimally or least disturbed by human activity (Stoddard et al.
2006). However, the characterization of even sub-optimal sites as reference situations can still provide a valuable guide. This kind of logic introduces the notion of “best-available condition” or “least-disturbed condition” as the most appropriate, and indeed only, baseline reference for many areas of the world (Gardner
2010; Stoddard et al.
2006).
Available monitoring data provide a starting point for historical baseline references. However, most biodiversity monitoring schemes were initiated late in the twentieth century, whereas land use, just like most of the anthropogenic pressures that are currently impacting biodiversity, has been operating for over centuries or even millennia (Mihoub et al.
2017). This mismatch between the long history of land use and the limited temporal frame of the biodiversity monitoring schemes, together with geographical and taxonomic biases of the data, limits the assessment of full impacts on biodiversity (Mihoub et al.
2017). Other problems arise when historical data is not linked to a bigger monitoring plan, as time frame and level of degradation of the reference should be equal to prevent misleading results, i.e., due to climate variation (Nielsen et al.
2007; Borja et al.
2012). In conclusion, due to the absence of historical records in many parts of the world, a baseline reference situation is often inapplicable.