Environmental Product Declarations of Structural Wood: A Review of Impacts and Potential Pitfalls for Practice

: The use of wood and timber products in the construction of buildings is repeatedly pointed towards as a mean for lowering the environmental footprint. With several countries preparing regulation for life cycle assessment of buildings, practitioners from industry will presumably look to the pool of data on wood products found in environmental product declarations (EPDs). However, the EPDs may vary broadly in terms of reporting and results. This study provides a comprehensive review of 81 third-party veriﬁed EN 15804 EPDs of cross laminated timber (CLT), glulam, laminated veneer lumber (LVL) and timber. The 81 EPDs represent 86 different products and 152 different product scenarios. The EPDs mainly represent European production, but also North America and Australia/New Zealand productions are represented. Reported global warming potential (GWP) from the EPDs vary within each of the investigated product categories, due to density of the products and the end-of-life scenarios applied. Median results per kg of product, excluding the biogenic CO 2 , are found at 0.26, 0.24, and 0.17 kg CO 2 e for CLT, glulam, and timber, respectively. Results further showed that the correlation between GWP and other impact categories is limited. Analysis of the inherent data uncertainty showed to add up to ± 41% to reported impacts when assessed with an uncertainty method from the literature. However, in some of the average EPDs, even larger uncertainties of up to 90% for GWP are reported. Life cycle assessment practitioners can use the median values from this study as generic data in their assessments of buildings. To make the EPDs easier to use for practitioners, a more detailed coordination between EPD programs and their product category rules is recommended, as well as digitalization of EPD data. C.E.A. and A.W.; methodology, F.N.R. and C.E.A.; investigation, C.E.A., F.N.R., A.W. and R.N.H.; data curation, C.E.A. and F.N.R.; writing—original draft preparation, F.N.R. and C.E.A.; writing—review and editing, A.W., R.N.H. and H.B.; fund-ing acquisition, H.B. and F.N.R. authors manuscript.


Introduction
Activities of the building and construction sector contribute notably to environmental impacts. For instance, as much as 39% of global CO 2 emissions are associated with the construction and use of buildings [1]. Almost one third of these emissions arise from the production of building materials [2]. Additionally, the sector uses vast amounts of energy, mineral, and metal resources for construction as well as operation of existing buildings [3]. As a response to the negative impacts, the building industry and regulation is moving towards measures to reduce emissions as well as energy and material uses.
For decades, building regulation has mainly focused on energy efficiency in the operational phase. However, recent developments acknowledge the growing importance of embodied impacts of the building and construction sector, i.e., emissions and resource uses embedded in production, use, and end-of-life of materials [4][5][6]. The sector itself is attentive to the sustainability challenge, and in several countries, regulation is being prepared to tackle the negative environmental side effects of the material production and use. Countries, such as Finland, France, Denmark, Sweden, and New Zealand, are in the process of

•
What is the availability of data from third-party verified EN 15804 EPDs on structural wood? • Which environmental impacts do the EPDs report, and what affects the comparability of the EPDs? • How may the EPDs be used for informing the building design process, and which potential pitfalls should users be aware of?

Data Collection Process
The compilation of data was limited to third-party verified EPDs that follow the European Standards EN15804:2012 + A1:2013 and EN15804:2012 + A2:2019 for construction works [32,41]. A study by Jane Anderson, 2021, [42] identifies a comprehensive list of international EPD programs that include EPDs, which apply to the two European Standards for EPDs on construction works, respectively. This study uses the list of EPD programs identified by Anderson as a basis for the systematic compilation of EPDs. Searches were conducted in each of the EPD programs using the search words 'wood' and 'timber' as well as equivalent words in French, Spanish, and German, depending on the geographical scope of the EPD program in question. The systematic search of the EPD programs resulted in 72 EPDs. An additional google search was conducted with the search phrase 'EN 15804 wood timber EPD'. This resulted in an additional 9 EPDs.
Besides only focusing on EPDs concerning construction works, this study considers one type of product exclusively: structural wood elements. We chose this limitation because structural elements most often constitute the greater part of the environmental impact from a building, thereby representing the biggest opportunity to reduce environmental impacts from a building. As a final focus of the data collection process, we only included EPDs that were valid at the time of the data collection, that is spring 2021.

Data Extraction Criteria
Next, after the data collection process, we extracted data from all EPDs according to a defined set of data extraction criteria. The data extraction criteria were generated to create a solid foundation for the data analysis and consists of a list of criteria that could prove to be relevant in the further analysis. Table 1 presents a comprehensive list of all data extraction criteria. Table 1. Data extraction criteria used for the EPD review.

EPD id
The unique id for every EPD.

EPD owner
The owner of the EPD.
Product identification A short description of the product including wood type (for instance spruce, pine, or fir) and moisture content.
Product category Structural wood within one of the four categories: sawn timber, glulam, cross laminated timber, and laminated veneer lumber.

Density
Average density for the product as stated in the EPD.
Validation from/to date Dates from which the EPD is valid from and to.
Geographical representativeness Geographical scope that the product in the EPD is valid for Temporal representativeness Temporal scope that the EPD represents.

Data Extraction Criteria Description
European Standard The European Standard that the EPD is based on, either EN15804:2012+A1:2013 or EN15804:2012 + A2:2019 [32,41] Product Category Rules Specific product category rules that the EPD is based on.
Functional Unit Declared unit defined in the EPD, which serves as a basis for the calculations in the EPD.
LCA database LCA database used for the calculation of the environmental impacts in the EPD.
LCA software LCA software used for the calculation of the environmental impacts in the EPD.

End of life scenarios
End-of-life scenario examined in the EPD. In cases where the EPD entails several end-of-life scenarios (for instance, the landfill scenario, incineration scenario, recycling scenario), all end-of-life scenarios are registered.

Environmental impacts
Environmental impacts per functional unit as stated in the EPD.

Exclusion Criteria
In the data extraction process, we defined a set of exclusion criteria used to scope the study. One important criterion was the sole focus on structural wood elements' EPDs within one of the four categories, as described in Ramage et al., 2017 [17] • Sawn timber: Timber cut from logs in different sizes, shapes, and different types of wood. • Glued laminated timber (glulam): Glued laminated timber comprised of multiple layers of timber bonded together with an adhesive to form structural beams. • Cross Laminated Timber (CLT): Cross laminated timber comprised of multiple layers of wood panel bonded together perpendicular to one another with an adhesive to form a uniform wood panel with structural properties. • Laminated Veneer Lumber (LVL): Laminated veneer lumber comprised of multiple layers of thin wood bonded together with an adhesive to form structural elements, such as beams.
In addition, to limit the data extraction process, we chose to only include system boundaries related to the production phase, the construction process phase, the end-of-life phase, and the benefits and loads beyond the system boundary. Figure 1 illustrates the system boundaries included in this study with a blue highlight. The system boundary follows the European Standard for EPDs of construction products [32,41]. The newest version of the European Standard for EPDs sets a minimum requirement that the life cycle stages A1-3, C3, C4, and D should be declared in future EPDs [5].
A final exclusion criterion defined in the study is the focus on a selected range of environmental indicators. A study by Dias et al. (2020) suggests that the environmental indicators acidification potential of land and water (AP), eutrophication potential (EP), and global warming potential (GWP) are the most relevant impact categories for softwoods, hardwoods, and glulam [43]. To further investigate the resource uses of construction wood, we also included the resource use categories abiotic depletion potential (ADPe), total use of non-renewable primary energy resources (PENRT), and total use of renewable primary energy resources (PERT). Table 2 presents a comprehensive list of all environmental indicators investigated in the study. Since the updated EN 15804: A2 defines other environmental indicators to be declared in EPDs than EN 15805: A1, we also included indicators from both versions of the standard in the study.

Data Normalization
To enable specific levels of comparison between the environmental impact potential results extracted from the EPDs, we normalized all impact potentials to the same unit, namely the unit of environmental indicator per kilogram of product. The normalization was based on the functional unit and the density of the product declared in the EPD and using Equation (1): where Inorm,i is the normalized impact potential for the environmental indicator i, Ii is the impact potential for the environmental indicator provided in the unit m 3 in the EPD, and ρ i is the density of the given product in the unit kg/m 3 .
Other comparative analyses in this study require a normalization of results to exclude the biogenic CO2 embedded in the product. The majority of EPDs declare the amount of biogenic CO2 in the product. In some other cases, the biogenic carbon is declared, which is then converted into biogenic CO2 by multiplying with a factor 44/12, which is the factor describing the difference in molar mass between CO2 and C according to EN 16449. In the few cases where neither biogenic CO2 nor biogenic carbon were declared, the biogenic CO2 was calculated based on the calculation rules from standard EN 16449, presented in Equation (2):

Data Normalization
To enable specific levels of comparison between the environmental impact potential results extracted from the EPDs, we normalized all impact potentials to the same unit, namely the unit of environmental indicator per kilogram of product. The normalization was based on the functional unit and the density of the product declared in the EPD and using Equation (1): where I norm,i is the normalized impact potential for the environmental indicator i, I i is the impact potential for the environmental indicator provided in the unit m 3 in the EPD, and ρ i is the density of the given product in the unit kg/m 3 .
Other comparative analyses in this study require a normalization of results to exclude the biogenic CO 2 embedded in the product. The majority of EPDs declare the amount of biogenic CO 2 in the product. In some other cases, the biogenic carbon is declared, which is then converted into biogenic CO 2 by multiplying with a factor 44/12, which is the factor describing the difference in molar mass between CO 2 and C according to EN 16449. In the few cases where neither biogenic CO 2 nor biogenic carbon were declared, the biogenic CO 2 was calculated based on the calculation rules from standard EN 16449, presented in Equation (2): where 44 and 12 represent the molar mass of CO 2 and C, respectively; 0.5 is the default share of carbon content in wood; and M d is the dry mass of the wood.

Uncertainty Assessment
To examine the inherent uncertainty of the EPDs, we performed an uncertainty assessment of all EPDs included in the study. The uncertainty was assessed by applying the method presented in Waldman et al. (2020). Waldman et al. (2020) presents a method that enables a quantitative assessment of the variations in the underlying data in EPDs. The method considers five parameters that can potentially affect the data quality and specificity of EPDs, namely whether the EPDs are: 1.
Time specific.
The first four parameters represent whether the EPD captures an industry or one manufacturer (manufacturer specific), multiple facilities or one specific facility (facility specific), a range of products or a single product (product specific), and a specific point in time or a time period (time specific). Finally, the fifth parameter, supply chain specific, captures the degree to which the EPD is based on generic upstream data and on upstream data that are representative of the actual processes.
Each of the parameters are given a data quality factor (DQ f ), a Z-value, that quantifies the specificity. If the EPD is specific to the manufacturer, facility, or product, a factor of 2% is applied to the EPD impact potential results for each parameter. If the EPD is not specific to the three parameters, a factor of 20% is applied for each parameter. Parameter 4 about time specificity is set to 20% for all cases, because all EPDs rely on inventory data from past production years. The percentages may be set at different values, but in this study, the percentages for parameters 1-4 follow the defaults described in Waldman et al. [39]. Parameter 5 about supply chain specificity is set to 10% based on the assumption that data from production sites and key suppliers are specific, and that this data represents 90% of the impact, the remaining 10% being non-specific.
The data quality factors for each parameter are combined into one uncertainty score using Equation (3) describing the uncertainty of the impact potential results provided in the EPDs: where Z M is the DQ f for manufacturer specific, Z F is the DQ f for facility specific, Z P is the DQ f for product specific, Z T is the DQ f for time specific, and Z S is the DQ f for supply chain specific.

Characteristics of EPDs on Structural Wood
The review paper analyzed 81 EPDs found from the search. The EPDs belong to 11 different EPD program holders. Four of the EPDs contain documentation and results for more than one product. Therefore, a total of 86 products are documented in the EPDs, distributed by the EPD programs as shown in Figure 2. Furthermore, a total of 17 of the EPDs from Environdec and EPD Denmark contain more than one end-of-life (EoL) scenario for the declared product. The total amount of additional product scenarios reported in the EPDs amount to 64 EoL scenarios. The Environdec EPD program holds most of the available EPDs and is also the program where the EPDs contain the higher number of scenarios. This reflects Environdec's broader geographical coverage that spans a range of countries, whereas most of the other programs hold EPDs representing products primarily within the regional or national context of the program.
In total, 16 of the EPD documents, i.e., 20%, were available in a format other than the PDF file format, either as a CSV or XLS format.
Details about the EPDs included in this review can be found in the Supplementary Materials.
scenarios. This reflects Environdec's broader geographical coverage that spans a range of countries, whereas most of the other programs hold EPDs representing products primarily within the regional or national context of the program.
In total, 16 of the EPD documents, i.e., 20%, were available in a format other than the PDF file format, either as a CSV or XLS format.
Details about the EPDs included in this review can be found in the Supplementary Materials.

Validity of EPDs
The available EPDs were approved in the period 2015-2021. Especially, the period 2018-2020 saw an increase in the number of published EPDs of structural wood products, as seen in Figure 3. In relation to wood products, the reporting of biogenic carbon content and sequestered CO2 is an important reporting parameter. The EN 15804: A2 version released in 2019 contains an updated list of environmental indicators to report, for instance, concerning the GWP, which will be reported in the details of GWPfossil, GWPbiogenic, and GWPLULUC. The implementation of the EN 15804: A2 is shown in the EPDs published after publication in 2019.
The validation period of an EN 15804 EPD is 5 years, although several of the EPDs in the sample have had their validation period extended by a revision of the EPD. As seen in Figure 3, a high number of EPDs were published in 2018-2020 based on the EN 15804: A1, and these EPDs will be valid up until 2025. Hence, for several years to come, the available EPDs on structural wood will contain different reporting formats on GWP and other environmental indicators, such as AP and EP.
Apart from the use of EN 15804: A1 and EN 15804: A2 as core PCRs, the majority of the EPDs state that they align with the specific PCRs of their respective program holder. The PCRs may prescribe additional documentation and formatting requirements. An example of additional reporting requirements for biogenic CO2 is found in the EPD-Norway program, where the PCR in use, even before the EN 15804: A2, required specific documentation of the biogenic CO2 at different oxidation rates. Number of EPDs CLT [20] Glulam [27] LVL [5] Timber [34]

Validity of EPDs
The available EPDs were approved in the period 2015-2021. Especially, the period 2018-2020 saw an increase in the number of published EPDs of structural wood products, as seen in Figure 3. In relation to wood products, the reporting of biogenic carbon content and sequestered CO 2 is an important reporting parameter. The EN 15804: A2 version released in 2019 contains an updated list of environmental indicators to report, for instance, concerning the GWP, which will be reported in the details of GWP fossil , GWP biogenic , and GWP LULUC . The implementation of the EN 15804: A2 is shown in the EPDs published after publication in 2019.
The validation period of an EN 15804 EPD is 5 years, although several of the EPDs in the sample have had their validation period extended by a revision of the EPD. As seen in Figure 3, a high number of EPDs were published in 2018-2020 based on the EN 15804: A1, and these EPDs will be valid up until 2025. Hence, for several years to come, the available EPDs on structural wood will contain different reporting formats on GWP and other environmental indicators, such as AP and EP.
Apart from the use of EN 15804: A1 and EN 15804: A2 as core PCRs, the majority of the EPDs state that they align with the specific PCRs of their respective program holder. The PCRs may prescribe additional documentation and formatting requirements. An example of additional reporting requirements for biogenic CO 2 is found in the EPD-Norway program, where the PCR in use, even before the EN 15804: A2, required specific documentation of the biogenic CO 2 at different oxidation rates.

Scope of EPDs
The scope of life cycle stages varies across the sample of EPDs. A total of 14 differently defined scopes of life cycle modules are found in the sample of EPDs. As shown in Figure 4, 18 of the EPDs include a scope where only life cycle modules A1-A3 and potentially A4 are included. The remaining 68 EPDs include some combination of EoL modules C1-C4 as part of the scope.
The 18 EPDs that include only production modules A1-A3 (and potentially A4) often report only the biogenic CO2 uptake in the structural wood and not the release, leading to unbalanced mass flows. When these unbalanced numbers are applied by EPD users, this can lead to misrepresentations of the actual impact of the product, because the product reports a negative GWP, i.e., it seems like a climate mitigation measure to use the product in question. However, according to the updated standard EN 15804: A2, EoL modules shall be reported. This leads to some degree of mass balancing of the biogenic CO2, although a chosen EoL scenario of landfilling may still affect the total balancing. This is further explained in Section 3.2. A total of 71 EPDs report the loads and benefits of the next product system, module D, as part of the scope. The module D scenarios represent a broad range of substitution scenarios covering energy conversion from different sources as well as recycling and reuse scenarios. Module D impacts do not belong to the product system under study but rather the next product system. Because module D, according to the standards, is to be reported separately from the life cycle impacts of a product (to avoid double counting in relation

Scope of EPDs
The scope of life cycle stages varies across the sample of EPDs. A total of 14 differently defined scopes of life cycle modules are found in the sample of EPDs. As shown in Figure 4, 18 of the EPDs include a scope where only life cycle modules A1-A3 and potentially A4 are included. The remaining 68 EPDs include some combination of EoL modules C1-C4 as part of the scope.
The 18 EPDs that include only production modules A1-A3 (and potentially A4) often report only the biogenic CO2 uptake in the structural wood and not the release, leading to unbalanced mass flows. When these unbalanced numbers are applied by EPD users, this can lead to misrepresentations of the actual impact of the product, because the product reports a negative GWP, i.e., it seems like a climate mitigation measure to use the product in question. However, according to the updated standard EN 15804: A2, EoL modules shall be reported. This leads to some degree of mass balancing of the biogenic CO2, although a chosen EoL scenario of landfilling may still affect the total balancing. This is further explained in Section 3.2. A total of 71 EPDs report the loads and benefits of the next product system, module D, as part of the scope. The module D scenarios represent a broad range of substitution scenarios covering energy conversion from different sources as well as recycling and reuse scenarios. Module D impacts do not belong to the product system under study but rather the next product system. Because module D, according to the standards, is to be reported separately from the life cycle impacts of a product (to avoid double counting in relation The 18 EPDs that include only production modules A1-A3 (and potentially A4) often report only the biogenic CO 2 uptake in the structural wood and not the release, leading to unbalanced mass flows. When these unbalanced numbers are applied by EPD users, this can lead to misrepresentations of the actual impact of the product, because the product reports a negative GWP, i.e., it seems like a climate mitigation measure to use the product in question. However, according to the updated standard EN 15804: A2, EoL modules shall be reported. This leads to some degree of mass balancing of the biogenic CO 2 , although a chosen EoL scenario of landfilling may still affect the total balancing. This is further explained in Section 3.2.
A total of 71 EPDs report the loads and benefits of the next product system, module D, as part of the scope. The module D scenarios represent a broad range of substitution scenarios covering energy conversion from different sources as well as recycling and reuse scenarios. Module D impacts do not belong to the product system under study but rather the next product system. Because module D, according to the standards, is to be reported separately from the life cycle impacts of a product (to avoid double counting in relation to the next system), the reported numbers from module D were not further investigated in this study. Figure 5 shows a heat map of the country of origin of the 81 structural wood products represented in available EPDs. There is a notable share of European countries with five or more EPDs on structural wood products. Specifically, the countries in which regulation is prepared at the moment (France, Nordic countries) display higher numbers of available EPDs.

Geographical Coverage of EPDs
Note that language barriers may have prevented collection of EPDs from countries in Asia, Africa, and South and Central America. However, it is likely that there are none, or very few, EN 15804-compatible EPDs from these regions, as the EN 15804 is Eurocentric in its origin from The European Committee for Standardization. Still, the existence of EPDs from Canada, US, Australia, and New Zealand serve to show that the structure and content of the verified EN 15804 EPDs are seen as useful in their own regional contexts.
Buildings 2021, 11, x FOR PEER REVIEW 9 of 18 to the next system), the reported numbers from module D were not further investigated in this study. Figure 5 shows a heat map of the country of origin of the 81 structural wood products represented in available EPDs. There is a notable share of European countries with five or more EPDs on structural wood products. Specifically, the countries in which regulation is prepared at the moment (France, Nordic countries) display higher numbers of available EPDs.

Geographical Coverage of EPDs
Note that language barriers may have prevented collection of EPDs from countries in Asia, Africa, and South and Central America. However, it is likely that there are none, or very few, EN 15804-compatible EPDs from these regions, as the EN 15804 is Eurocentric in its origin from The European Committee for Standardization. Still, the existence of EPDs from Canada, US, Australia, and New Zealand serve to show that the structure and content of the verified EN 15804 EPDs are seen as useful in their own regional contexts.

Reported Environmental Impacts-GWP
In Figure 6, the reported GWP from production modules (A1-A3), EoL modules (C3-C4), and the total of these are presented in box plots. The upper line of the box is the 75th percentile, and the lower line of the box is the 25th percentile. The median of the data set is denoted with a line in the box. The average is represented by the x. The procedure for selecting and normalizing data for the graph is as follows: • For A1-A3: All unique products are included, i.e., unique production impacts reported by the EPD sample. • For C3-C4: All individual scenarios for EoL are included. One unique product may have up to five EoL scenarios • For A1-A3-C3-C4: All individual scenarios are included, confer C3-C4 above. In cases where only cradle-to-gate impacts are reported, including biogenic CO2, the biogenic CO2 is counterbalanced in a manual step based on the reported (or estimated) biogenic CO2. Details can be found in the methods section about data normalization.

Reported Environmental Impacts-GWP
In Figure 6, the reported GWP from production modules (A1-A3), EoL modules (C3-C4), and the total of these are presented in box plots. The upper line of the box is the 75th percentile, and the lower line of the box is the 25th percentile. The median of the data set is denoted with a line in the box. The average is represented by the x. The procedure for selecting and normalizing data for the graph is as follows: • For A1-A3: All unique products are included, i.e., unique production impacts reported by the EPD sample. • For C3-C4: All individual scenarios for EoL are included. One unique product may have up to five EoL scenarios • For A1-A3-C3-C4: All individual scenarios are included, confer C3-C4 above. In cases where only cradle-to-gate impacts are reported, including biogenic CO 2 , the biogenic CO 2 is counterbalanced in a manual step based on the reported (or estimated) biogenic CO 2 . Details can be found in the methods section about data normalization.
The trend across product categories is that production modules report negative values of CO 2 e emissions due to the sequestration of biogenic CO 2 in the tree growth (module A1). As reflected in the C3-C4 numbers, biogenic CO 2 is released at the EoL modules of products. The numbers for total GWP (A1-A3 + C3-C4) amount to median values of approximately 110 kg CO 2 e/m 3 for CLT and glulam, and 58 kg CO 2 e/m 3 for timber (including both hardwood and softwood). A difference between the totals of the engineered products CLT and glulam and the total of timber is expected, due to the higher density as well as the additional processes and materials associated with the production of CLT and glulam. The LVL sample consists of only four different products spanning 10 scenarios, which is considered too small a sample for deriving statistically based values. The span between the LVL data points is notable, however.

Variations Due to Density and EoL
The variation of the total GWP for glulam and for timber is also notable, especially, where several data points between the minimum and the median fall at values below 0 kg CO 2 e/m 3 . Hence, even though the median value is at 58 kg CO 2 e/m 3 for timber in general, several products present a negative total GWP. These negative GWP numbers are rooted in the EoL scenario definitions applied for the products in the EPD. According to the EN 15804:A2, the biogenic CO 2 sequestered in the wood at growth shall be counterbalanced at the EoL modules. The total counterbalancing of sequestered CO 2 is not always followed, in the sense that released CO 2 from EoL is less than the biogenic CO 2 uptake in the production modules of the life cycle. Several of the EPDs are made according to program-specific PCRs that allow for inclusion of additional EoL scenarios. Specifically, landfill scenarios, where permanent storage of the majority of biogenic CO 2 is assumed, generate negative GWP values. The variation between reported impacts from EoL scenarios in the EPDs is shown in Figure 7a, normalized per kg of structural wood product. The trend across product categories is that production modules report negative values of CO2e emissions due to the sequestration of biogenic CO2 in the tree growth (module A1). As reflected in the C3-C4 numbers, biogenic CO2 is released at the EoL modules of products. The numbers for total GWP (A1-A3 + C3-C4) amount to median values of approximately 110 kg CO2e/m 3 for CLT and glulam, and 58 kg CO2e/m 3 for timber (including both hardwood and softwood). A difference between the totals of the engineered products CLT and glulam and the total of timber is expected, due to the higher density as well as the additional processes and materials associated with the production of CLT and glulam. The LVL sample consists of only four different products spanning 10 scenarios, which is considered too small a sample for deriving statistically based values. The span between the LVL data points is notable, however.

Variations Due to Density and EoL
The variation of the total GWP for glulam and for timber is also notable, especially, where several data points between the minimum and the median fall at values below 0 kg CO2e/m 3 . Hence, even though the median value is at 58 kg CO2e/m 3 for timber in general, several products present a negative total GWP. These negative GWP numbers are rooted in the EoL scenario definitions applied for the products in the EPD. According to the EN 15804:A2, the biogenic CO2 sequestered in the wood at growth shall be counterbalanced at the EoL modules. The total counterbalancing of sequestered CO2 is not always followed, in the sense that released CO2 from EoL is less than the biogenic CO2 uptake in the production modules of the life cycle. Several of the EPDs are made according to programspecific PCRs that allow for inclusion of additional EoL scenarios. Specifically, landfill scenarios, where permanent storage of the majority of biogenic CO2 is assumed, generate negative GWP values. The variation between reported impacts from EoL scenarios in the EPDs is shown in Figure 7a, normalized per kg of structural wood product.
CLT A1-A3 [19] CLT C3-C4 [27] CLT A1-A3 + C4-C4 [30] Glulam A1-A3 [27] Glulam C3-C4 [31] Glulam A1-A3 + C4-C4 [39] LVL A1-A3  Figure 7a furthermore illustrates how 14 of the product scenarios use a mixed scenario approach for the EoL modelling. For instance, a declared EoL impact represents a scenario in which 33% is recycled, 16% is incinerated with energy recovery, and 17% is landfilled. These kind of specific mix scenarios seem to be applied to represent national averages of wood waste treatment. However, it makes it difficult to use the EoL of the EPDs in local settings, where another EoL scenario mix may be more representative. Figure 7b illustrates the reported GWP from A1-A3 of the EPDs, normalized by density. Figure 6 indicates almost a factor 2 difference between GWP timber and CLT/glulam for A1-A3 + C3-C4. However, the difference is reduced drastically when looking at A1-A3 normalized by weight. For CLT, the median is at −1.36 kg CO2e/kg; for glulam, at −1.31 kg CO2e/kg; and for timber, at −1.38 kg CO2e/kg. When further normalized to exclude biogenic CO2, the median values are at 0.26, 0.24, and 0.17 kg CO2e for CLT, glulam, and timber, respectively, as shown by the black marks in Figure 7b. Figures 6 and 7a,b thus show that the median values of reported GWP may vary up to a factor 2 between categories of structural wood product, but this difference is largely explained by the density of the products as well as the EoL scenarios applied in the EPDs.

Reported Environmental Impacts-Other than GWP
GWP is the more reported impact category in LCA studies and practice. However, in line with the EN 15804 standard, the EPDs also declare a range of other indicators for potential impacts and resource uses. Figure 8a-e show the correlation between GWP from A1-A3 and a selection of additional impacts and resource uses, all normalized to one kilogram of wood. The R2 values for the correlation between GWP and acidification potential AP (a), eutrophication potential (EP) (b), and primary energy, non-renewable, total PENRT (d) indicate a fit to the trend line between 9% and 18%, which cannot be characterized as a strong fit to the correlation model. The bad fittings to models imply that the use of GWP as a single point of reference is not appropriate to capture equivalent potentials in these other impact and resource use categories. In the case of abiotic depletion potential for elements ADPE (e), the sample is characterized by large variations within the Figure 7a furthermore illustrates how 14 of the product scenarios use a mixed scenario approach for the EoL modelling. For instance, a declared EoL impact represents a scenario in which 33% is recycled, 16% is incinerated with energy recovery, and 17% is landfilled. These kind of specific mix scenarios seem to be applied to represent national averages of wood waste treatment. However, it makes it difficult to use the EoL of the EPDs in local settings, where another EoL scenario mix may be more representative. Figure 7b illustrates the reported GWP from A1-A3 of the EPDs, normalized by density. Figure 6 indicates almost a factor 2 difference between GWP timber and CLT/glulam for A1-A3 + C3-C4. However, the difference is reduced drastically when looking at A1-A3 normalized by weight. For CLT, the median is at −1.36 kg CO 2 e/kg; for glulam, at −1.31 kg CO 2 e/kg; and for timber, at −1.38 kg CO 2 e/kg. When further normalized to exclude biogenic CO 2 , the median values are at 0.26, 0.24, and 0.17 kg CO 2 e for CLT, glulam, and timber, respectively, as shown by the black marks in Figure 7b. Figure 6andFigure 7a,b thus show that the median values of reported GWP may vary up to a factor 2 between categories of structural wood product, but this difference is largely explained by the density of the products as well as the EoL scenarios applied in the EPDs.

Reported Environmental Impacts-Other Than GWP
GWP is the more reported impact category in LCA studies and practice. However, in line with the EN 15804 standard, the EPDs also declare a range of other indicators for potential impacts and resource uses. Figure 8a-e show the correlation between GWP from A1-A3 and a selection of additional impacts and resource uses, all normalized to one kilogram of wood. The R2 values for the correlation between GWP and acidification potential AP (a), eutrophication potential (EP) (b), and primary energy, non-renewable, total PENRT (d) indicate a fit to the trend line between 9% and 18%, which cannot be characterized as a strong fit to the correlation model. The bad fittings to models imply that the use of GWP as a single point of reference is not appropriate to capture equivalent potentials in these other impact and resource use categories. In the case of abiotic depletion potential for elements ADPE (e), the sample is characterized by large variations within the reported results, spanning from a minimum of 4.9 × 10 −9 kg Sb-e/kg to a maximum of 2.7 × 10 −5 kg Sb-e/kg.
The use of primary energy, renewable, and total PERT (c) covers not only the renewable energy used for processes along the life cycle modules, e.g., wood chips for kiln drying, but also integrates the renewable feedstock energy embodied in the product. This feedstock energy is balanced in the life cycle of the product, like the biogenic CO 2 , albeit with a reverse sign. Hence, it is typically accounted for as a resource use in the A1 module and counterbalanced in the C3 or C4 module. The balancing of the feedstock energy is found around the lower heating value of dry wood, which is around 19 MJ/kg wood.
reported results, spanning from a minimum of 4.9 × 10 −9 kg Sb-e/kg to a maximum of 2.7 × 10 −5 kg Sb-e/kg. The use of primary energy, renewable, and total PERT (c) covers not only the renewable energy used for processes along the life cycle modules, e.g., wood chips for kiln drying, but also integrates the renewable feedstock energy embodied in the product. This feedstock energy is balanced in the life cycle of the product, like the biogenic CO2, albeit with a reverse sign. Hence, it is typically accounted for as a resource use in the A1 module and counterbalanced in the C3 or C4 module. The balancing of the feedstock energy is found around the lower heating value of dry wood, which is around 19 MJ/kg wood.

Uncertainty Assessment of EPD Sample
The sample of EPDs represent varying degrees of specificity in terms of manufacturer, facility, co-production, time of data representation, and supply chain. The uncertainty of the EPD sample is assessed with the method by Waldman et al. 2020, as described in the method section. Figure 9 shows the distribution between the different categories that the EPDs fall in, depending on the specificity. As seen in the figure, the majority of

Uncertainty Assessment of EPD Sample
The sample of EPDs represent varying degrees of specificity in terms of manufacturer, facility, co-production, time of data representation, and supply chain. The uncertainty of the EPD sample is assessed with the method by Waldman et al., 2020, as described in the method section. Figure 9 shows the distribution between the different categories that the EPDs fall in, depending on the specificity. As seen in the figure, the majority of CLT EPDs represent a single manufacturer, at a single production site, producing only one product. In total, 45% of all EPDs in the sample are of this degree of specificity. At the other end of the scale, 46% of all EPDs in the sample represent various manufacturers, at various production sites. This lower degree of specificity dominates the collection of timber EPDs.
CLT EPDs represent a single manufacturer, at a single production site, producing only one product. In total, 45% of all EPDs in the sample are of this degree of specificity. At the other end of the scale, 46% of all EPDs in the sample represent various manufacturers, at various production sites. This lower degree of specificity dominates the collection of timber EPDs. Figure 9. Specificity of EPDs in the areas of manufacturer, facility, and co-production, presented as category 1-4. Calculating the uncertainty of the EPDs (the ±ZEPD), values to represent the specificity degrees are found as ±21%, ±30%, ±36%, and ±41% for category 1, 2, 3, and 4, respectively. According to Waldman et al., these ZEPD factors are added to the reported results to "…form the bounds of reasonable confidence, meaning that it is reasonably likely that any given specific instance of a product (e.g., a specific kg of rebar) in the population (i.e., all the products represented by the EPD) will have a value within that range." [39]. Using timber as an example, the factors for individual EPDs were added to the reported GWP and are shown in Figure  10. Note that the GWP numbers are adjusted for biogenic CO2 content, since this follows the mass and hence should not be affected by uncertainty in the same way as other processes contributing to GWP. The nature of the ±ZEPD calculation means that higher-end results are more affected in an absolute sense. For instance, the latter value in Figure 10 has a calculated uncertainty span of ±41% and thus spans from 287-691 kg CO2e/m 3 . This may come across as a suspiciously large span, leading to questioning of the method behind the ±ZEPD. However, the specific EPD in question, an EPD of Australian hardwood, is transparent in its own reporting of variations between the 33 manufacturers behind the EPD. Hence, variations Calculating the uncertainty of the EPDs (the ±Z EPD ) , values to represent the specificity degrees are found as ±21%, ±30%, ±36%, and ±41% for category 1, 2, 3, and 4, respectively. According to Waldman et al., these Z EPD factors are added to the reported results to " . . . form the bounds of reasonable confidence, meaning that it is reasonably likely that any given specific instance of a product (e.g., a specific kg of rebar) in the population (i.e., all the products represented by the EPD) will have a value within that range." [39]. Using timber as an example, the factors for individual EPDs were added to the reported GWP and are shown in Figure 10. Note that the GWP numbers are adjusted for biogenic CO 2 content, since this follows the mass and hence should not be affected by uncertainty in the same way as other processes contributing to GWP.
CLT EPDs represent a single manufacturer, at a single production site, producing only one product. In total, 45% of all EPDs in the sample are of this degree of specificity. At the other end of the scale, 46% of all EPDs in the sample represent various manufacturers, at various production sites. This lower degree of specificity dominates the collection of timber EPDs. Figure 9. Specificity of EPDs in the areas of manufacturer, facility, and co-production, presented as category 1-4. Calculating the uncertainty of the EPDs (the ±ZEPD), values to represent the specificity degrees are found as ±21%, ±30%, ±36%, and ±41% for category 1, 2, 3, and 4, respectively. According to Waldman et al., these ZEPD factors are added to the reported results to "…form the bounds of reasonable confidence, meaning that it is reasonably likely that any given specific instance of a product (e.g., a specific kg of rebar) in the population (i.e., all the products represented by the EPD) will have a value within that range." [39]. Using timber as an example, the factors for individual EPDs were added to the reported GWP and are shown in Figure  10. Note that the GWP numbers are adjusted for biogenic CO2 content, since this follows the mass and hence should not be affected by uncertainty in the same way as other processes contributing to GWP. The nature of the ±ZEPD calculation means that higher-end results are more affected in an absolute sense. For instance, the latter value in Figure 10 has a calculated uncertainty span of ±41% and thus spans from 287-691 kg CO2e/m 3 . This may come across as a suspiciously large span, leading to questioning of the method behind the ±ZEPD. However, the specific EPD in question, an EPD of Australian hardwood, is transparent in its own reporting of variations between the 33 manufacturers behind the EPD. Hence, variations The nature of the ±Z EPD calculation means that higher-end results are more affected in an absolute sense. For instance, the latter value in Figure 10 has a calculated uncertainty span of ±41% and thus spans from 287-691 kg CO 2 e/m 3 . This may come across as a suspiciously large span, leading to questioning of the method behind the ±Z EPD . However, the specific EPD in question, an EPD of Australian hardwood, is transparent in its own reporting of variations between the 33 manufacturers behind the EPD. Hence, variations between −42% and +90% are reported for the kiln-dried hardwood. This shows that the uncertainty of EPDs is not to be neglected when numbers from EPDs are used further, for instance in building design evaluations.

Implications for Use and Development of EPDs
EPDs are intended for use in the decision-making concerning building design and material choice. This study mapped the availability and characteristics of EPD data, and points to some areas of attention for the uptake, use, and development of EPD data in the building industry.
The mapping shows a high concentration of EPDs in the European countries, especially covering the countries where regulation is on the political agenda. This indicates how industry in those countries is preparing for an increased demand for product data to support decision-making in the building design process.
Recent research and practice highlight the necessity of digitalization in the design process [44,45], and it is thus imperative that EPD data becomes available in digital formats that support a digitalized work processes. At the current state, only 20% of the EPDs on structural wood are available in digital formats other than pdf.
The use of representative data is a key concern in building-LCA, and the vast amount of available EPD data thus require that EPD users critically assess the data they apply, increasing the workload dramatically. Depending on the building design stage, it may be appropriate to use generic data or it may be appropriate to use product-specific data. However, in their current forms, it can be difficult to assess the representativeness of data because documentation varies hugely, not only between EPD program holders, but sometimes also between EPDs in the same program. For the sake of the EPD users, EPD programs should consider harmonizing documentation requirements to ensure transparency at the highest level possible.

Comparison with Database Values
Building-LCA practitioners in search for data have the option of using one of the publicly available databases for construction materials provided by companies, research institutions, or governmental bodies. Examples of these available databases include ICE from the UK [46], KBOB from Switzerland [47], and Ökobaudat from Germany [48]. In connection with the ongoing regulatory development in Sweden, a national database, the Klimatdatabas [49], is also in its testing phase. These databases have in common that they offer building-LCA practitioners data covering a large range of construction materials. The databases provide generic and/or average data, and Ökobau also contains digitalized EPDs for specific products.
A comparison of the median values from the current study with generic GWP values found in the databases provide an overview as shown in Figure 11. The median values derived from the EPDs are displayed for modules A1-A3 and normalized to kg of product without biogenic CO 2 . As seen in Figure 11, the EPD medians are, for all product categories, in the lower end of the scale. This is expected, since the publicly available databases for construction products, in general, choose to display conservative values for the generic data. A similar pattern was found by Hill and Dibdiakova in a comparison of wood product EPDs with values from the ICE database [50].
The median GWP values found in this study are based on a broad sample of EPDs for structural wood products. Hence, median values may be used by building-LCA practitioners in cases where no regional generic data exist, or the values may be used for LCA screenings early in the building design, where specific products are not yet decided upon. The large spans within the categories (see Figure 6) and the uncertainties of the EPDs (see Figure 10) should, however, be kept in mind. Figure 11. A1-A3 GWP values per kg of wood product (without biogenic CO2) from the current study and as reported in publicly available databases for construction products.

Potential Pitfalls in the Use of EPDs
The comparison of GWP and other environmental impacts reported from EPDs showed notable variations between the different product categories at a comparative basis of 1m 3 , but less variation at a comparative basis of 1 kg. The density and the EoL scenarios were shown to be key factors in determining the total GWP. EPD users thus need to be specifically aware of these factors when determining whether an EPD may be seen as representative.
Although GWP is receiving special attention by the building industry, results from this study also show that EPD users should be cautious in seeing GWP as an indicator for the products' overall environmental performance. The weak correlation between GWP and several other indicators is a specific point of attention for EPD users working under multiple LCA criteria, for instance via certification systems for sustainable buildings.
The embedded uncertainty in reported EPD impacts is an additional point of attention for the EPD users. One level of uncertainty pertains to the specificity of data, as investigated in this study by use of the ZEPD method, i.e., how representative is the data in terms of time and place to the actual product considered for use. An additional level of uncertainty pertains to the methodological approaches applied by the EPD producers, in the cases where the EN 15804 standard is not specific, leaving room for interpretation. This is, for instance, the case of allocation of impacts between the main product and byproducts [51] and for choosing scenarios for modules C and D [52]. Both levels of uncertainty should be recognized, and preferably integrated in further use of the numbers.

Limitations of Study
This study is a review of EPD data, which is a rapidly changing field of industry activity. The validity of an EPD is five years, so the EPD data form a a dynamic pool of material data. Hence, this study may be seen as a snapshot of a field of data in development. In a future with digitized EPDs, the review exercise of this study may be conducted less laboriously, and the results and values be subject to regular updates.
The main limitation of the study is found in the sometimes diverging approaches to calculating and reporting results in the EPDs. Even though all included EPDs are in line with the EN 15804 standard, the varying program-specific PCRs result in numbers that may be difficult to compare. For instance, the PCR from UL Environment guides calculations based on the Traci characterization method [53]. For GWP, this means a difference to the CML characterization method because they build on different assessment reports

Potential Pitfalls in the Use of EPDs
The comparison of GWP and other environmental impacts reported from EPDs showed notable variations between the different product categories at a comparative basis of 1 m 3 , but less variation at a comparative basis of 1 kg. The density and the EoL scenarios were shown to be key factors in determining the total GWP. EPD users thus need to be specifically aware of these factors when determining whether an EPD may be seen as representative.
Although GWP is receiving special attention by the building industry, results from this study also show that EPD users should be cautious in seeing GWP as an indicator for the products' overall environmental performance. The weak correlation between GWP and several other indicators is a specific point of attention for EPD users working under multiple LCA criteria, for instance via certification systems for sustainable buildings.
The embedded uncertainty in reported EPD impacts is an additional point of attention for the EPD users. One level of uncertainty pertains to the specificity of data, as investigated in this study by use of the Z EPD method, i.e., how representative is the data in terms of time and place to the actual product considered for use. An additional level of uncertainty pertains to the methodological approaches applied by the EPD producers, in the cases where the EN 15804 standard is not specific, leaving room for interpretation. This is, for instance, the case of allocation of impacts between the main product and by-products [51] and for choosing scenarios for modules C and D [52]. Both levels of uncertainty should be recognized, and preferably integrated in further use of the numbers.

Limitations of Study
This study is a review of EPD data, which is a rapidly changing field of industry activity. The validity of an EPD is five years, so the EPD data form a a dynamic pool of material data. Hence, this study may be seen as a snapshot of a field of data in development. In a future with digitized EPDs, the review exercise of this study may be conducted less laboriously, and the results and values be subject to regular updates.
The main limitation of the study is found in the sometimes diverging approaches to calculating and reporting results in the EPDs. Even though all included EPDs are in line with the EN 15804 standard, the varying program-specific PCRs result in numbers that may be difficult to compare. For instance, the PCR from UL Environment guides calculations based on the Traci characterization method [53]. For GWP, this means a difference to the CML characterization method because they build on different assessment reports from the International Panel of Climate Change IPCC. The differences in GWP from the two methods are considered negligible [54] but may nevertheless be considered a bias to results. Likewise, the different ways of reporting GWP from the EN 15804:A1 and the GWP total in EN 15804:A2 are considered to have a minor effect on the comparability [54].

Conclusions
This study provides a comprehensive review of 81 third-party verified EPDs of structural timber, identified as complying with EN 15804. The 81 EPDs represent 86 different products and 152 different product scenarios. The EPDs mainly represent European production, but North America and Australia/New Zealand productions are also represented. More than 80% of EPDs include a more complete life cycle with EoL options, and this number will grow in the transition from versions EN 15804: A1 to EN 15804:A2.
Reported GWP from the EPDs vary broadly within each of the investigated categories of CLT, glulam, LVL, and timber. The variations were shown to be closely related to the density of the products and the EoL scenarios applied. When normalizing the results per kg of product, and further excluding the biogenic CO 2 , median values of 0.26, 0.24, and 0.17 kg CO 2 e were found for CLT, glulam, and timber, respectively.
The results further showed that the correlation between GWP and other impact categories is limited, and that the inherent data uncertainty may add up to ±41% to reported impacts when assessed with an uncertainty method from the literature. In the EPDs, uncertainties of up to 90% for GWP are reported as an example.
Building-LCA practitioners can use the median values from this study as generic data in their assessments of buildings. To make the EPDs easier to use for practitioners, a more detailed coordination between EPD program holders is recommended, as well as the development of digitalized EPD data.

Conflicts of Interest:
The authors declare no conflict of interest.