1 Introduction

Life cycle assessment (LCA) is a promising method to account for the life cycle environmental performance of a product from raw material extraction, through manufacturing and transportation, to demolition and disposal. As required by ISO 14040, four phases are normally involved in a LCA study and they include: (i) goal and scope definition, (ii) inventory analysis, (iii) impact assessment, and (iv) interpretation (ISO 2006b). The first phase is to define the study objectives, product system, target audience, functional unit, etc. In the life cycle inventory (LCI) phase, data are collected to provide the inventory results. In the third phase, life cycle impact assessment (LCIA), which is a complex process, converts the LCI results into various category indicators. The LCI and LCIA results are then interpreted in the final phase in which sensitivity analysis or uncertainty analysis can be conducted.

When performing LCIA, assessors may adopt the midpoint and endpoint approaches. As defined in (ILCD 2011), the midpoint and endpoint approaches are characterization models which provide indicators at different levels. The endpoint approach evaluates the environmental impact at the areas of protection (AoP) level, such as human health, ecosystem, and resource. In contrast, the midpoint approach assesses the environmental impact at a level in cause-effect chain from the release of substance or consumption of resource to the endpoint level.

Using the midpoint LCIA methods, such as CML (Guinée 2001) and TRACI (EPA 2003), the environmental interventions are represented as a set of indicators including carbon dioxide (CO2) for climate change, chlorofluorocarbon (CFC) for ozone depletion, nitrogen oxides (NO x ) for eutrophication, sulfur dioxide (SO2) for acidification, etc. Since the midpoint approach concludes a set of indicators for dozens of impact categories, it is difficult to interpret the midpoint results (Yi et al. 2011). In contrast, the endpoint LCIA methods like Eco-indicator 99 (Goedkoop et al. 1999) convert indicators of the impact categories into three or four damage categories (i.e., AoP) only, i.e., human health, ecosystem, and resource, resulting in a relatively easier interpretation. The damage caused by environmental interventions is estimated in damage modeling through which damage factors are derived to facilitate the endpoint evaluation. Since LCA comprises LCI data over time and space, it is challenging to provide a characterization model at the endpoint level. The downside of the endpoint approach is that, due to the involvement of additional steps like the fate modeling and damage modeling, such a complex characterization model could introduce some uncertainties to the endpoint results (Finnveden et al. 2009). Bare et al. (2000) advocated that a consistent framework is inevitable to provide the LCIA results at both the midpoint and endpoint levels. Such a request led to the development of ReCiPe (Goedkoop et al. 2009; Heijungs et al. 2003). Although ReCiPe strives to provide the midpoint and endpoint assessments in a consistent way, the implementations of the two approaches may result in different interpretations.

While many literature reviews, statistical analyses, and case studies have been conducted to compare the existing LCIA methods (Notarnicola et al. 1998; Schulze et al. 2001; Pennington et al. 2004; Brent and Hietkamp 2003; Dreyer et al. 2003; Renou et al. 2008; Landis and Theis 2008; Amani and Schiefer 2011; Berger and Finkbeiner 2011; Pizzol et al. 2011; Cavalett et al. 2013; Monteiro and Freire 2012), there has been no consensus on when to use an endpoint-based method and how to help the end users interpret the endpoint results. In some LCA applications, such as Environmental Product Declaration (EPD) (ISO 2010) and its superstructure Product Category Rules (PCR) (CEN 2012), a midpoint LCIA method of CML is selected. Others (EMSD 2006) may prefer an endpoint LCIA method like Eco-indicator 99.

Until now, very few studies have been conducted to compare the midpoint and endpoint approaches. A research by Bare and Gloria (2006) indicated that the midpoint results are more comprehensive to cover possible environmental interventions, while the endpoint approach may neglect some aspects in deriving the damage indicators. Bare (2010) concluded that the midpoint results can provide analyses in a more scientific way, while the endpoint approach should provide additional support. A recent study by Yi et al. (2011) compared the midpoint and endpoint approaches of LIME (Life Cycle Impact Assessment based on Endpoint Modeling) through a set of solid waste management scenarios and concluded that the results of the midpoint and endpoint are in general consistent except when the primary contributor is not included in the estimated categories. However, it is still unclear if the two approaches can lead to different interpretations.

The construction industry consumes a large volume of materials. It has been reported that buildings account for 40 % of the world’s material and energy flows (Horvath 2004). A study (Zabalza Bribián et al. 2011) reported that the civil and construction works are responsible for 60 % of the raw material consumption. On average, the annual production of concrete in the world is around 1 ton per human being (Van den Heede and De Belie 2012; Flower and Sanjayan 2007). The demand for construction materials will continue to grow in developing countries like China, where the building floor area will increase by 30 billion square meters from 2005 to 2020 (Huang et al. 2013). Throughout the life cycle of a building, the construction material stage represents 10–20 % of the energy consumption (Ramesh et al. 2010). Cement as a commonly used construction material accounts for up to 10 % of the weight of concrete and is responsible for 5–7 % of the anthropological carbon emissions (Meyer 2009).

In recent studies, LCA has been conducted to calculate the environmental impact of construction materials (Huntzinger and Eatmon 2009; Huberman and Pearlmutter 2008; Rajagopalan et al. 2010; Landis et al. 2009; Ibbotson and Kara 2013; Collins 2010; Bovea et al. 2009; Hammond and Jones 2008; Bolin and Smith 2011; Menzies 2013; Irfan 2011; Zygomalas et al. 2012; Puettmann et al. 2012), building components (Ortiz et al. 2010; Taborianski and Prado 2012), and the entire building (Zhang et al. 2006; Lee et al. 2009). Nonetheless, the discussion on the midpoint and endpoint approaches is rarely found when a LCIA method is adopted in these studies. Ignoring the difference between the midpoint and the endpoint approaches may result in an inappropriate interpretation on the LCA results.

The present work aims at addressing the differences between the midpoint and endpoint approaches of ReCiPe and using them to analyze the environmental impact of commercial buildings in Hong Kong. In this study, 23 construction materials are examined, covering over 99 % total environmental burden of the materials used in commercial buildings in Hong Kong. Comparison between the midpoint and endpoint based on the normalization results of the selected materials is conducted. The environmental impact of a commercial building in Hong Kong is then analyzed by both the midpoint and endpoint approaches. The difference between the midpoint and endpoint results is investigated. Based on these results, suggestions on when to adopt an endpoint approach and how to interpret the different results from the two approaches are provided. Finally, the limitations and recommendations of the midpoint and endpoint approaches are discussed.

2 Materials and methods

2.1 Impact assessment

The development of ReCiPe was largely due to the need of harmonizing the existing midpoint and endpoint methods to eliminate the barriers of LCIA method selection in LCA design (Goedkoop et al. 2009). ReCiPe is a newly developed LCIA package based on the concepts of CML 2001 and Eco-indicator 99. The name ReCiPe carries twofold meanings: (i) it provides a recipe for the calculation of LCIA; and (ii) it is an acronym of the main contributors which include RIVM and Radboud University, CML and PRé Consultants (Goedkoop et al. 2009). In this study, ReCiPe is selected since the results in both the midpoint and endpoint levels can be provided.

At the midpoint level, 18 impact categories are defined to cover a large variety of impacts. The indicators of impact categories represent the interventions of common concern. Normalization is included in the midpoint by applying a set of midpoint normalization factors obtained from the statistics for the reference year of 2000 (Sleeswijk et al. 2008). Two tiers of normalization factors are available for the European and the World scales, so that the midpoint ReCiPe is capable of calculating the impact in different regional scales. Although weighting factors can be assigned in the midpoint, ReCiPe does not provide a weighting analysis at the midpoint.

At the endpoint level, analysis is available on the impact categories as well as the damage categories. The damage categories describe three AoP which include human health, ecosystems, and resources. Impact categories contribute differently to the three damage categories. In the category of human health, ReCiPe implements the disability-adjusted life years (DALY) which considers the year of life lost and the year of life disabled due to environmental interventions. Damage to ecosystems is described by species lost in a predefined period as a result of emissions to terrestrial, freshwater, and marine systems. Damage to resources is calculated as the economic loss caused by the marginal increase in costs due to the extraction of a resource (Goedkoop et al. 2009). The essence of ReCiPe lies in its ability to link the midpoint and endpoint by means of the endpoint characterization model. The midpoint characterization results are converted into the endpoint characterization results through a set of endpoint characterization factors. The development of connection between the midpoint and the endpoint is the most important outcome of ReCiPe, so that the two approaches are linked as both of them are based on the midpoint characterization results. Normalization and weighting factors are provided in the European and World scales for the three damage categories. A single score can be derived by aggregating the weighting results. Figure 1 shows the connection factors between the elements in impact assessment.

Fig. 1
figure 1

Connection factors between elements in impact assessment

Full size image

ReCiPe employs a cultural theory (Hofstetter 1998) with three archetypes being used to describe the three groups of considerations and assumptions (Goedkoop et al. 2009). Individualist (I) considers the short-term impact due to the most relevant chemicals. On the other hand, egalitarian (E) is based on the precautionary principle that considers the long-term perspective and involves more risk. Hierarchism (H) is a balanced perspective based on the common policy principles. In addition to I, H, and E, ReCiPe provides another set of weighting factors (A) by averaging the weighting factors of the three perspectives. In this study, the balanced term (H) which is recommended as a default choice is adopted. Since there is no special preference on the weighting factors, the averaged value will be adopted. The ReCiPe Midpoint (H)-World and ReCiPe Endpoint (H)-World H/A are, therefore, used to demonstrate the difference between the midpoint and the endpoint approaches.

2.2 Selection of construction materials

EMSD (2006) studied the environmental impact of construction materials based on the bills of quantities of 28 commercial buildings in Hong Kong. The 20 most influential construction materials according to the single score of Eco-indicator 99 contribute over 99 % to the total environmental impact of all building materials (Table 1). Apart from the top 20 materials, three materials namely cast iron, copper, and paint are included for comparison as they are commonly used in commercial buildings. The results based on the selected materials should, therefore, represent most of the environmental impact of materials in a commercial building in Hong Kong.

Table 1 The selected construction materials and the corresponding processes in LCI database. Contributions to environmental pollution are extracted from EMSD (2006)
Full size table

A LCI database of Ecoinvent (Frischknecht et al. 2007) is used to simulate majority of the selected materials. The selection of the processes in Ecoinvent becomes a problem when several options are available in the database. Only the common process is chosen to represent the environmental performance of a material. For instance, five concrete types are provided in Ecoinvent while the most commonly used one with the strength of B45/35 is adopted. Aluminum is modeled as primary aluminum and secondary aluminum in Ecoinvent, while the production mix of 68 % primary and 32 % secondary aluminum is used to reflect the worldwide production. Three materials viz. galvanized steel, access floor, and precast concrete component are not available in Ecoinvent. Galvanized steel can be modeled by US LCI (NREL 2004). As for access floor, it is modeled by referring to galvanized steel following the assumptions of EMSD (2006). The inventory of precast concrete components was solicited through a field survey (Dong et al. 2013). The LCIA results of the selected materials are generated by SimaPro 7.

2.3 Comparison of structure

This study compares the midpoint and the endpoint LCIA results for individual construction materials as well as the materials used for a commercial building. The comparison begins with a very representative construction material, concrete, due to its large environmental impact (Table 1). All the selected materials are used for the rest of the comparison. The comparison is primarily based on the normalization results as the characterization results cannot be directly compared while the midpoint of ReCiPe provides no analysis on weighting.

To comprehensively interpret the difference caused by the two approaches, the material inventory of a commercial building is analyzed. The results of the building case study can offer an overview of the impact on a building project in real practice due to the switching between the midpoint and endpoint approaches. It can also help improve our understanding beyond the analysis of an individual construction material.

3 Results and discussion

3.1 Interpretation on the LCIA results of concrete

By referring to Table 2, the midpoint approach provides results for the impact categories in two tiers: characterization and normalization. The midpoint characterization results in different units can be difficult to interpret if a LCA participator has insufficient knowledge on the impact categories. For example, the greenhouse gas emissions to produce 1 kg concrete are 0.13 kg CO2 eq. (carbon dioxide equivalent), though it may not be clear on the consequences of the emissions. The normalization results at the midpoint level can be further analyzed by involving the normalization factors so that the results are normalized to a reference system and hence facilitate the comparison. Climate change has a midpoint normalization factor of 1.45E−4 persons∙year/kg (point), derived from the total anthropogenic carbon emissions of 4.18E13 kg CO2 eq. in 2000 when the world’s population was 6,115 million. The value of 1.8E−5 indicates that the production of 1 kg concrete is equivalent to the emission contributed per capita in 9 min. The normalization results sharing the same unit are hence comparable.

Table 2 LCIA results of 1 kg concrete using the midpoint and endpoint approaches of ReCiPe with the unit of normalization and weighting results being point (pt)
Full size table

The endpoint approach carries out the analysis for both the impact and the damage categories. The characterization results can be compared between the impact categories within the same damage group from the endpoint approach. Damage to human health and ecosystems caused by concrete is mostly attributed to climate change, while fossil depletion is responsible for the damage to resources (Table 2). Although the endpoint approach provides results of various impact categories, the analysis is only conducted for the concerned damage areas, while other aspects are eliminated owing to a lack of knowledge and the unavailability of emission factors (Goedkoop et al. 2009). The impact categories under the human health damage group, e.g., photochemical oxidant formation, particulate matter formation, ionizing radiation, etc. can cause damage to not only the human health but also the ecosystems. On the other hand, acidification, eutrophication, ecotoxicity, and land transformation may also influence human health. However, since these effects are not clearly understood, they are neglected in the current endpoint LCIA methods. Another deficiency of the endpoint analysis is its incomprehensive coverage of the damage categories which might lead to biases and errors of the final interpretation. The endpoint normalization is based upon the fraction of damage per capita which is the aggregation of several impact categories. Therefore, the endpoint normalization results are apparently smaller than that of the midpoint (Table 2). In the damage assessment level, the endpoint results can be interpreted as an integration of the midpoint results. The endpoint weighting factors of ReCiPe are defined through a panel survey and usually politically concerned. Both the midpoint and endpoint approaches are based on the midpoint characterization results, and a difference may exist in the following steps of normalization, damage assessment and weighting.

Comparison of the normalized scores at the midpoint and endpoint levels (Table 2) reveals that climate change, human toxicity, freshwater eutrophication, and marine ecotoxicity share the same order of magnitude at the midpoint level, while the normalized scores of these impact categories at the endpoint level differ by for example 8 orders of magnitude between climate change (human health) and marine ecotoxicity. The inconsistent normalized scores are attributed to the different normalization factors under the two approaches. The midpoint normalization factor is 1.45E−04 of climate change (human health) and 4.14E−04 of marine ecotoxicity, whereas the combined normalization factor (product of endpoint characterization and endpoint normalization) is 1.04E−04 of climate change (human health) and 4.89E−11 of marine ecotoxicity. This suggests that the midpoint and endpoint approaches can lead to inconsistent results of materials which are dominated by different impact categories.

It can be realized that the calculation process in LCIA is in fact a way to extract information from abundant data so that the results can be easily interpreted in either a single score or by means of damage indicators. The emission inventory is aggregated to the indicators of impact categories which are further combined to derive the single score by applying the characterization factors, normalization factors, and weighting factors. A good LCIA method should be able to provide those factors in a scientific way, e.g., through the fate modeling, numerical modeling, data mining, questionnaire survey, Delphi method, etc., so that any artificial influence can be minimized. A schematic diagram is given in Fig. 2 to illustrate the data extraction process in LCIA. The pyramid shape represents the data loss from LCI to a single score. The inventory of environmental interventions includes all the recognized emissions caused by a product or process. By performing LCIA, the results become less comprehensive but more concise. The endpoint approach hence provides a more concise way to interpret the LCA results.

Fig. 2
figure 2

Schematic illustration of LCIA procedures

Full size image

3.2 Results of selected materials

The results of the selected construction materials are studied in this section. Contributions from the impact categories to the endpoint single score are presented in Fig. 3, where four impact categories are found to be the most influential to the endpoint single score, viz. climate change, human toxicity, particulate matter formation, and fossil depletion. In addition, agricultural land occupation caused by plywood has a large impact on the single score as well.

Fig. 3
figure 3

Contributions from impact categories toward endpoint single score

Full size image

Figure 4 gives the ranking of impact categories based on the ratio between the combined normalization factors and the midpoint normalization factors (as defined in Fig. 1). It is found that for most of the cases, the midpoint normalization factors are larger than the combined normalization factors, suggesting that the midpoint normalization results are greater than the endpoint normalization results (Table 2). An impact category with a smaller ratio indicates that the midpoint normalization result receives a greater emphasis than the endpoint normalization result.

Fig. 4
figure 4

Ranking of impact categories based on the ratios of combined normalization factor/midpoint normalization factor

Full size image

In Fig. 5, the normalization results of impact categories are aggregated to normalized scores that are compared between the selected materials. In general, the midpoint agrees with the endpoint, though disagreement can still be found. For example, the normalized score of paint is larger than that of structural steel in the endpoint while the midpoint results of the two materials are in a reverse order. This is because the midpoint normalized score of structural steel is dominated by marine ecotoxicity, freshwater ecotoxicity, and metal depletion. However, the combined normalization factors of the three impact categories are comparatively small which are ranked 13th, 14th, and 10th among the 18 impact categories. As a result, the difference caused by the three impact categories becomes negligible in the endpoint. On the other hand, the great combined normalization factor of fossil depletion with the 2nd rank can magnify the difference at endpoint, resulting in a larger endpoint normalized score of paint than that of structure steel. The magnifying effect of fossil depletion on the endpoint normalized score can also be observed when comparing other materials, e.g., “plastic, rubber, polymer” and “tiles”. As shown in Fig. 4, the ratio of fossil depletion is the greatest among the impact categories, indicating the high normalized scores of fossil depletion at endpoint. Climate change, fossil depletion, and particulate matter formation of which the ratios are comparatively large can contribute significantly to the endpoint single score. This suggests that the materials being dominated by the three categories may demonstrate a greater environmental influence according to the endpoint single score as compared with those materials which are dominated by other impact categories.

Fig. 5
figure 5

Comparison of the total environmental performance of the selected materials based on normalization results (functional unit—1 kg)

Full size image

The ranking of materials as given in Fig. 5 is based on the averaged impact scores. However, when taking uncertainties into account, the ranking may be rearranged. In order to examine the influence of uncertainty, a Monte Carlo analysis is carried out for structural steel and paint. The model is run for 500 times. It is found that the midpoint normalized score of paint has a standard deviation of 4.97E−3 and a mean of 3.20E−2. The average midpoint normalized score of a nearby point rebar is 3.01E−2 (Fig. 5). In that case 3.20E−2 to 4.97E−3<3.01E−2, indicating paint has more than 16 % possibility to be ranked lower than rebar in terms of the midpoint normalized score. Similarly, the orders of structural steel and plastic laminate may be reversed at the midpoint.

In Table 3, the rankings of impact categories are compared for the selected materials. The ranking of impact categories presents completely different features between the midpoint and the endpoint approaches. In the midpoint, human toxicity, marine ecotoxicity, and freshwater ecotoxicity are the most contributive impact categories for most of the selected materials. Nonetheless, the endpoint results show that fossil depletion, climate change (human health), and particulate matter formation are more influential. This can be explained by Fig. 4, in which the ratios of the three categories are larger than the others. The different rankings of scenarios and impact categories as shown in Fig. 5 and Table 3 imply that the implementation of the midpoint and endpoint approaches can lead to different interpretations in and after the normalization step when more than one impact categories are considered.

Table 3 Ranking of impact categories for construction materials based on normalization results per kilogram construction material. Contributions of impact categories are shown (normalization results are equally weighted)
Full size table

3.3 Results of a commercial building

The studied building is a 30-storey commercial building in Hong Kong. The building provides 43,210 m2 of construction floor area (CFA) (Zhang et al. 2013). The results of the building can provide further insights when dealing with real case study. The key construction materials in the commercial building are given in Table 4. It should be noted that only the materials listed in Table 1 are analyzed, while the other materials are ignored due to their relatively insignificant impact. The LCIA results of rebar are used to represent the environmental impact of steel in the studied building.

Table 4 Major materials of a commercial building in Hong Kong with the data calculated from Zhang et al. (2013)
Full size table

The comparison is conducted at the normalization level. As shown in Fig. 6, there is no obvious relationship between the midpoint and the endpoint results in terms of the ranking of impact categories. This is consistent with the observation in the previous sections for individual materials. For instance, freshwater ecotoxicity contributes significantly in the midpoint approach, while it receives the lowest rank in the endpoint approach. Similar observations can be found for the categories of fossil depletion, freshwater eutrophication, particulate matter, etc. The underlying rationale behind this disagreement is the effects of the midpoint normalization factors and the combined normalization factors, which can be explained by the different ratios in Fig. 4.

Fig. 6
figure 6

Comparison between normalization results of the case building using midpoint and endpoint approaches based on the resource use for the commercial building

Full size image

A further step is to study the contributions from individual materials to the total environmental impact using the midpoint and endpoint approaches (Fig. 7). The results show that the contributive percentage of concrete is larger in the endpoint than in the midpoint. In contrast, steel accounts more environmental impact in the midpoint than in the endpoint. This is attributed to the different values of normalization factors of the two approaches. Climate change (human health) ranks high in the endpoint while low in the midpoint, corresponding to the larger contribution of concrete in the endpoint. It should be noted that in Table 1, concrete has been estimated as the most contributive material to the environmental impact of a building, while in Fig. 7, steel is responsible for more impact. This is because Table 1 divides the steel types into rebar, galvanized steel, structural steel, and stainless steel, while Fig. 7 considers the aggregated effects of all steel types. The results in Figs. 6 and 7 indicate that precaution should be paid before conducting a LCIA analysis since an interpretation based on the midpoint and endpoint results can be very different.

Fig. 7
figure 7

The contribution analysis of the environmental impact of construction materials of the case building using midpoint and endpoint approaches based on the resource use for the commercial building (normalization results are equally weighted)

Full size image

ISO 14040 points out that no scientific basis is available for weighting. The methodology to define the weighting factors remains controversial as different LCIA methods may adopt distinctive ways to quantify the weighting factors. Eco-indicator 99 (the predecessor of ReCiPe endpoint method) constructs the weighting factors through a panel survey with 365 interested persons (Goedkoop et al. 1999). The life cycle impact assessment method (LIME) in Japan develops the weighting factors by conjoint analysis based on 1,000 responses for its second version LIME2 (Itsubo et al. 2012). In ReCiPe, weighting factors are assigned by a similar way to Eco-indicator 99. The weighting is not provided in the level of impact categories but is only applicable to the endpoint damage categories. The endpoint approach defines the weighting factors for the three time perspectives as well as their average. The emphasis on damage categories shifts from human health for short-term perspective (I) to ecosystems for long-term perspective (E). The human health and ecosystems share the same weighting factor in the averaged set (A). Among the three damage categories, weighting factor of resources is the smallest. The choice among the weighting factors should hence depend on which damage category is mostly concerned in the LCA study.

The ReCiPe Endpoint (H)-World H/A is used in the above analysis. Here, the effects of three weighting methods are compared by calculating the contributions from the damage categories to a single score for the studied building. As observed in Fig. 8, the single score being calculated by the equal weighting factors is apparently higher than the single scores from other weighting methods. Therefore, the weighting factors would not only determine the contributions of the damage categories toward the total impact but they could also influence the value of the single score.

Fig. 8
figure 8

The contribution from damage categories to single score using different weighting factors based on the resource use for the commercial building (H: balanced-term; A: averaged perspective; Equal: equally weighted)

Full size image

3.4 Difference between midpoint and endpoint

This study reveals that if the ranking of construction materials is calculated for an individual category, the midpoint and the endpoint are consistent in ReCiPe. Consequently, the two approaches can lead to the same ranking if the LCA study is performed for individual impact categories.

If all the impact categories are considered, the ranking of construction materials is in general consistent. However, differences can still be found when the materials are dominated by distinct impact categories. By examining the normalization results of impact categories, the endpoint results are smaller than the midpoint results, except for the category of fossil depletion. In the endpoint, fossil depletion, climate change, and particulate matter formation are emphasized by applying a great value of combined normalization factors, while in the midpoint, the categories of toxicity become more important. This can lead to a risk of steering toward the desired results. As if when one option is favored, the approach with the best outcomes for this option could become the choice. For example, when selecting materials for a building construction project, the contractor may prefer the endpoint approach if the selected materials are dominated by toxicity or ecotoxicity. To prevent any subjective adoption of the midpoint and endpoint approaches, it is suggested that the estimation should be given in the individual impact categories in addition to an impact score, following the requirements given in ISO 14044.

The ranking of impact categories can be very different when using the midpoint and endpoint approaches of ReCiPe. This is valid for the selected construction materials and the commercial building. When comparing the importance of impact categories, the midpoint approach is suggested if there is no special request on the damage assessment, as the midpoint approach should be able to generate more reliable results. If the endpoint characterization results are used, the midpoint characterization results should also be provided for reference. If the endpoint single score is used, it should according to ISO (2006a) be complemented with the underlying midpoint results without using weighting.

The midpoint and endpoint may lead to different interpretations if the contributions from the subprocesses to the total environmental performance of commercial buildings are analyzed. This can cause a problem when LCA is used to detect the hot spots. For instance, the contribution of concrete is more significant in the endpoint than in the midpoint. To reduce the environmental impact of a commercial building, the contractor may wish to use more cementitious substitute (e.g., fly ash, blast furnace slag, silica fume, etc.) for concrete if the endpoint score is referred. Nevertheless, more effort may be paid to adopt recycled steel when the midpoint approach is used. To resolve this problem, participators should define the important impact categories clearly in the goal and scope definition stage. In other words, it should be determined in what impact categories the hot spots will be detected. For commercial buildings in Hong Kong, if the concerned categories are climate change and energy consumption in general, the contributions of materials toward climate change and energy consumption can hence be used to identify the hot spots.

As pointed out in ISO 14040 / 44, the weighting is based on value choices with no scientific basis. The implementation of a single score is hereby limited. The results in this study indicate that the weighting factors can change the single score of the studied building. The single score is larger if the damage categories are equally weighted. The choice among the weighting factors should depend on which damage category is mostly concerned in the LCA study.

Under most PCRs, the evaluation stops at the midpoint. This is also true for the PCR of construction works (EN 15804:2012). As discussed above, the midpoint is more reliable while the endpoint could lead to larger uncertainties. The implementation of the midpoint seems to be a wise choice for most cases. However, the endpoint approach has advantages as it can provide easy-to-understand results by considering the damage. The implementation of the endpoint approach should be encouraged when environmental LCA is integrated with life cycle costing (LCC) (Hunkeler et al. 2008) and social LCA (SLCA) (UNEP 2009) in a life cycle sustainability assessment (LCSA) (UNEP 2012). The indicator of human health (DALY) can be directly used to assess the social impact. Moreover, the unit of resources is monetary-based, so that the result of resources can be substituted to LCC as externalities.

One should note that the above discussion is based on ReCiPe. For the LCA studies using other methods such as CML, Eco-indicator 99, TRACI, EPS (Steen 1999), or IMPACT 2002+ (Jolliet et al. 2003), the situation can be rather different. For example, if CML is applied as the midpoint method while Eco-indicator 99 is used as the endpoint method, the results will vary at the characterization level rendering them incomparable. If both the midpoint and endpoint results are desired in a LCA study, the LCIA methods with both the midpoint and endpoint approaches like ReCiPe are strongly recommended.

3.5 Limitation and recommendations

LCIA conducts analyses to extract essential information from an inventory of environmental interventions. In this process, part of information is lost while the LCIA results are largely determined by a set of factors (characterization factors, normalization factors, etc.). The way to evaluate the environmental performance of a product is schematized at levels of midpoint and endpoint in LCIA. The endpoint approach provides results with higher degree of interpretation but larger uncertainty. On the other hand, the midpoint approach is more reliable but does not give information of damages. The midpoint and endpoint are only two representative sections along the cause-effect chain. There could be more tiers and one example is the concentration of pollutants which can be derived by fate modeling. The midpoint approach provides comprehensive results covering many impact categories, while the endpoint gives concise information of damages in three AoP.

In LCIA, the factors are defined based on the current knowledge of the impact categories and may not reflect the real situation. For example, the eutrophication process is calculated in the LCIA with indicators of nutrients in the midpoint and potentially disappeared fraction of species in the endpoint. The damage factor of freshwater eutrophication is obtained by a regression analysis between species loss and phosphorus concentration. However, eutrophication in aquatic system is actually determined by many complex hydrodynamic (e.g., flow velocity, turbulence, stability) and biochemical (e.g., intake by algae and other aquatic lives) processes, which have not been fully understood. Thus, the eutrophication process cannot be precisely modeled by the damage factor as defined in LCIA.

LCA is a comparison tool that examines the environmental impacts of products by considering the potential exposure rather than the actual exposure. There are other assessment tools that focus on the actual exposure, such as risk assessment and environmental impact assessment. These assessment techniques need extra effort (e.g., field monitoring, computational fluid dynamics modeling, data mining, etc.). The advantage of LCA lies in its ability to describe the studied system with a holistic and easy-to-use LCIA method. LCA deals with the whole life cycle chain of products, which is usually not covered by other assessment tools. The environmental impacts in LCA are calculated from a set of well-defined factors, which help simplify the studied problem and accelerate the analysis.

LCA can be integrated with other techniques to provide more accurate analysis. One example is to integrate it with geographic information system (GIS) to provide detailed spatial analysis (Geyer et al. 2010). This, however, requires more effort to collect site-specific data. In the construction industry, the integration of LCA with building information modeling (BIM) has been attempted by Kulahcioglu et al. (2012).

4 Conclusions

The midpoint and the endpoint approaches can provide assessment at different levels resulting in distinctive interpretations. This study has examined the implementation of the midpoint and endpoint approaches of ReCiPe by analyzing 23 construction materials as well as a commercial building. The results in this study indicate that the difference in the two approaches can be detected when several impact categories are taken into consideration. The rankings of impact categories demonstrate completely different features due to the effects of midpoint normalization factors and combined normalization factors. The ranking of materials is slightly different under the two approaches when the materials are dominated by different impact categories. Uncertainty can influence the rankings of materials with similar impact scores. In the case study of a commercial building in Hong Kong, the contribution from concrete is larger in the endpoint approach due to the greater factor value of climate change (human health) in the endpoint approach. In terms of weighting, the single score and contributions from damage categories may vary when different weighting factors are used.

The midpoint approach is able to provide analysis for a set of impact categories despite the results are difficult to interpret. The endpoint approach includes the damage assessment and introduces more uncertainties to the results. The midpoint approach is in general preferred since it can provide reliable assessment, while the endpoint approach gives additional information of damage with a higher degree of interpretation. In the LCIA calculation, the original information is whittled down through the calculation in LCIA from LCI to a single score and the essential information of high concern is extracted. Compared with midpoint results, endpoint results are less comprehensive but more concise.

Suggestions are provided to help designers interpret results from the two approaches, in particular, for the assessment of commercial buildings in Hong Kong. Since the midpoint and endpoint approaches can lead to different outcomes, any subjective adoption to steer toward the desired results should be prevented. The critical impact categories, in which the environmental performance is compared and analyzed, should be identified in the goal and scope definition. Characterization results per impact category should be given when the endpoint approach is used. If a single score is reported, it is mandatory to supply this result with the underlying midpoint LCIA results. The endpoint weighting factors should be properly selected to reflect the importance of AoPs. These suggestions are proposed based on ReCiPe and may not be applicable to methods with entirely different concepts such as the midpoint by CML and the endpoint by Eco-indicator 99. When an endpoint result is asked for, the LCIA methods with both midpoint and endpoint analyses like ReCiPe are strongly recommended.