ALIGNED D1.2 Description of scientific methods (Task 1.4 Framework for interpreting uncertainty)

Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 1 Work Package 1 – Shared modelling framework and learnings Task 1.4 – Framework for Interpreting uncertainty D1.2 – Description of scientific methods Guide on the appraisal of uncertainty in the LCA of biobased products Lead Contractor: Aalborg University (AAU) Author(s): M. Pizzol, Marcos D. B. Watanabe, M. Tschulkow, U. Javourez This document is a part of the ALIGNED project (grant no. 101059430) deliverable D1.2. It contains a guide on the appraisal of uncertainty in the LCA of bio-based products. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 2 PROJECTS DETAILS Project title Aligning Life Cycle Assessment methods and bio-based sectors for improved environmental performance. Project acronym ALIGNED Start / Duration 01/10/2022 – 36 months Type of Action RIA Website www.alignedproject.eu DELIVERABLE DETAILS Dissemination level Public Nature Report Due date (M) M18 (March/2024) Submission date 31/03/2024 DELIVERABLE CONTRIBUTORS Name Organisation Job title Deliverable leader Massimo Pizzol AAU Professor Contributing Author(s) Marcos Djun Barbosa Watanabe Maxim Tschulkow, Ugo Javourez NTNU ANTW INSAT Postdoc Postdoc Postdoc Reviewer(s) Jurjen Spekreijse BTG Researcher Nariê Rinke Dias de Souza NTNU Postdoc Ugo Javourez INSAT Postdoc Final review and quality approval Massimo Pizzol Dalia Stakenaite AAU AAU Professor Project manager DOCUMENT HISTORY Date Version Name Changes 2023/08/20 v.0.1 Guide on the appraisal of uncertainty in the LCA of bio-based products Improved tutorials, revised according to comments of internal review 2024/01/30 v.0.2 Guide on the appraisal of uncertainty in the LCA of bio-based products Added tutorials, revised according to comments of review round 2024/03/18 v.1.0 Guide on the appraisal of uncertainty in the LCA of bio-based products Final 2025/09/30 v.1.1 Guide on the appraisal of uncertainty in the LCA of bio-based products Revised version after testing in the case studies during the course of the project. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 3 TABLE OF CONTENTS Acronyms and abbreviations ......................................................................................................... 5 Executive summary ........................................................................................................................ 6 Introduction ................................................................................................................................... 7 1. Types of uncertainty: classification used in this document ................................................... 8 Data uncertainty .................................................................................................................... 8 Model uncertainty ................................................................................................................. 8 2. General approaches to analysis of uncertainty and sensitivity ........................................... 11 3. Uncertainty analysis – focus on data ................................................................................... 12 3.1. Defining uncertainty for input data, tiered approach ................................................ 12 Rough estimate. ................................................................................................................... 12 Pedigree matrix. ................................................................................................................... 12 Empirical estimates. ............................................................................................................. 12 3.2. Propagating uncertainty from input data to output, tiered approach ...................... 13 Analytical method. ............................................................................................................... 13 Stochastic method. .............................................................................................................. 14 4. Sensitivity analysis – focus on data ...................................................................................... 16 5. Uncertainty analysis – focus on model ................................................................................ 17 6. Sensitivity analysis – focus on model ................................................................................... 18 7. Guidelines for appraisal of uncertainty in the ALIGNED model framework ........................ 20 7.1. Uncertainty in background modelling (T1.1).............................................................. 21 7.2. Uncertainty in foreground modelling (T1.2) .............................................................. 22 7.2.1. Dynamic carbon flux model ............................................................................... 22 7.2.2. Constraints to biomass availability .................................................................... 23 7.3. Uncertainty in Life Cycle Impact Assessment (T1.3) .................................................. 24 7.4. Uncertainty in socio-economic assessment (T1.5) ..................................................... 25 8. Concluding remarks ............................................................................................................. 26 9. References............................................................................................................................ 27 Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 4 List of tables Table 1 Example of calculating uncertainty estimates for LCA model inputs from repeated measurements. ............................................................................................................................ 13 Table 2 Example of calculating uncertainty estimates for LCA model outputs with the analytical method. ........................................................................................................................................ 14 Table 3 Uncertainty and sensitivity of data in modelling of prospective scenarios using IAMs within the ALIGNED framework. .................................................................................................. 21 Table 4 Uncertainty and sensitivity in modelling dynamic carbon fluxes within the ALIGNED framework. ................................................................................................................................... 22 Table 5 Uncertainty and sensitivity in modelling constraints to biomass availability within the ALIGNED framework. ................................................................................................................... 23 Table 6 Uncertainty and sensitivity in life cycle impact assessment within the ALIGNED framework. ................................................................................................................................... 24 Table 7 Uncertainty and sensitivity in socio-economic assessment within the ALIGNED framework. ................................................................................................................................... 25 List of figures Figure 1 Uncertainty and sensitivity analysis for a generic model, for example a LCA model. ... 10 Figure 2 Practical toolbox for handling uncertainty and sensitivity of data and models in the ALIGNED framework (OAT: One At Time, GSA: Global Sensitivity Analysis). .............................. 11 Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 5 Acronyms and abbreviations ABBREVIATIONS Description EPD Environmental Product Declaration GSA Global Sensitivity Analysis LCA Life cycle assessment LCI Life cycle inventory LCIA Life Cycle Impact Assessment OAT One at the Time WP Work Package Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 6 Executive summary This guide presents practical approaches to handle uncertainty in the Life Cycle Assessment (LCA) of bio-based products within the ALIGNED project. The primary aim is to improve decision making in the bio-based industries and sectors – because such transition is heavily informed by and dependent on comparative assessment studies. As in the case of all ALIGNED WP1 outputs on methodological framework, all guidance is here provided using a tiered approach, i.e. different options are provided to perform a specific task or apply a specific method, in order of increasing accuracy but also increasing complexity. This is reported in a specific action paragraph in each section. The document is accompanied but other tools such as tutorials and calculators in excel and python that are made available in the T1.4 repository. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 7 Introduction Nowadays the number of environmental assessments and related modelling approaches and tools is booming. Besides LCA studies under the ISO framework (ISO, 2020), the Environmental Product Declaration system (EPD International, 2023) and now the Product Environmental Footprint (DG Environment, 2021) guidelines from the European Commission, but also studies under the framework of Greenhouse Gas Emission protocol (WRI & WBCSD, 2011) and the Science Based Targets and many others are all essentially based on quantitative models to calculate and return a numerical estimate of environmental footprint. All the results of these models are affected by uncertainty. Uncertainty can be defined as unknowns about the reality. Since a model is a simplified representation of reality and our understanding of reality is always incomplete, then the lack of knowledge about how the model should be designed is transferred to the results of such model (Lo Piano & Benini, 2022). In even simpler words the uncertainty of the result of a product footprint model can be intended as a range: every produced number should be intended as one in a distribution of possible outcomes. Uncertainty is related to sensitivity. While uncertainty analysis is about the qualification and quantification of the uncertainty in the inputs and outputs of a model, sensitivity analysis focuses on understanding how the changes in the inputs of a model influence the model outputs. Sensitivity analysis helps shaping uncertainty analysis and vice-versa. Recently several studies have argued for an increased focus on uncertainty and sensitivity analysis (Lo Piano & Benini, 2022; Saltelli, Bammer, et al., 2020) and on the limit of LCA models and in general of the limits of quantification in decision making for sustainability (Saltelli, Benini, et al., 2020). Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 8 1. Types of uncertainty: classification used in this document There are different ways of defining uncertainty and for the LCA domain. Igos et al. (2019) suggest to classify uncertainty either according to its intrinsic nature - epistemic or aleatory, or according to its location in a LCA model. In the latter case one can distinguish between uncertainty regarding the structure of the model, the quantities used in the model, or the context in which the model is used. From Igos et al. (2019): “Regarding the nature of uncertainty, epistemic uncertainty is due to a lack of knowledge or representativeness and can be reduced with more research and efforts (e.g., more collected data, higher model complexity). Aleatory or ontic uncertainty is due to the inherent variability and the lack of determinability of the system (inherent randomness of nature, observer effect) and cannot be reduced. Both natures of uncertainty can be present for quantity uncertainty, model structure uncertainty, and context uncertainty” (Igos et al., 2019). In their report on the prospective assessment of bio-based technologies the JRC and European Commission (2022) also classify uncertainties depending on either their nature (epistemic and ontic) their location (data, model, context), and their scale (from moderate to deep uncertainty). We refer the reader to the original report (European Commission et al., 2022) for a more extensive description of each. Not all these classifications are however equally useful for practical purposes. Since this is a guide addressing LCA practitioners in the bio-based industries, and users of the models provided by the project, the pragmatic choice done in the ALIGNED project was to simplify the classification to two types of uncertainty: data and model uncertainty – as this also allows to define appropriate handling strategies in terms of uncertainty and sensitivity analysis respectively. Data uncertainty refers to the choice of the numerical values to be used in the LCA model. Examples: lack of knowledge regarding…. …the carbon content of a specific tree species. …the amount of feedstock used to produce a biobased product in a year. …the amount of carbon emissions generated by composting a biobased product. Model uncertainty refers to all the possible way data are combined into a LCA model and the operations done with these data. Examples: lack of knowledge regarding the choice of…. …characterisation factors for the global warming potential (GWP20 or GWP100). …type of decay of biomass in the ground (linear versus nonlinear). …temporal range to calculate the yearly increment in biomass supply (5 years versus 20 years) Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 9 …type of probability distribution selected for each parameter/input (uniform, triangular, etc.) In Figure 1, a LCA model can be intended as a black box where input data points are transformed into output data points. When the specific value of the input is not precisely known (“cloud” of values rather than single value), we have uncertainty. The same applies when there are unknowns regarding how to structure the model (“shifting” shape of the box rather than single shape). Sensitivity is instead the relation between input and output (arrows). While some techniques like stochastic error propagation allow to quantify the uncertainty associated with the output of a LCA model starting from the uncertainty in the inputs, techniques like local and global sensitivity analysis allow explain to what extent changes in model input or model structure, that are due to uncertainty, lead to variations in model results. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 16 4. Sensitivity analysis – focus on data The most practical approach is to use OAT (One At Time) sensitivity analysis. Once a set of specific data or parameters’ values is chosen, these are varied one by one and the difference in output is then compared to the difference in input using sensitivity coefficients that can be calculated in different way (Bisinella et al., 2016). Imbeault-Tétreault et al. (2013) defines a sensitivity index Sx,h as the relative variation in output h caused by a relative variation in input x. While this index should be calculated using the partial derivative of impact score h according to x, in practice the LCA practitioner usually calculated it based on discrete data, as the ration between the relative difference in output and input. This is defined as “sensitivity ration” by Bisinella et al. (2016) and approximated to the value obtained using the derivatives. See an example of calculation in a separate Excel file (ALIGNED-T1.4Sensitivity-Ration-example-AAU.xlsx) and in a separate notebook tutorial (ALIGNED-T1.4-OATtutorial-AAU.ipynb). The result is a ranking of the most sensitive parameters based on their index value, where higher value indicates higher sensitivity of the results to changes in the value of the parameter. The limitation is that a data point or parameter might have different sensitivity on the results under different modelling assumptions. For example, the sensitivity to energy use can depend on the carbon intensity of the energy mix assumed. The sensitivity might also change depending on the values taken by of the rest of the input parameters, this is the case when there are interactions between parameters and non-linear models. Action: for the product system under analysis, select a list of parameters for testing sensitivity. This selection can be informed by previous experience and familiarity with the model, results from a contribution analysis, special importance for the decision maker of the study, or because they are highly uncertain (lack of knowledge) or very variable (multitude of values can be expected). Test the sensitivity of the parameters by changing their value (e.g. by 10% as in the notebook), calculating new results, and then calculating sensitivity ratios. With more time and resources available, increase the number of parameters under analysis. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 17 5. Uncertainty analysis – focus on model To understand the uncertainty in the results that is due to modelling choices, the suggested approach is to use scenarios. Scenarios are intended as plausible assumptions regarding a specific model structure, for example the assumption about a constrained flow or about what activity is substituted by a co-product, or model is used to calculate characterisation factors or emissions at the end of life, or the geographical location of a supplier, or the composition of the energy mix used. The modelling choice might affect multiple activities simultaneously in a complex way, for example the choice of allocation key (economic vs mass) or the choice of modelling approach (attributional versus consequential) or even of background system model (cut-off versus APOS versus consequential for the same ecoinvent version) are all modelling choices that affect simultaneously several activities in the product system. In practice the analysis is performed by changing the assumption and thus the model structure according to different scenarios, and then calculate new model outputs. This can be iterated for several options, see a simplified example in the notebook on changing electricity mix (ALIGNEDT1.4-Model-Uncertainty-tutorial-AAU.ipynb). Note that the use of different scenarios here is practically similar but conceptually different from doing an OAT sensitivity analysis on the modelling choice. It is practically similar because in the way it is implemented because it consists in a calculating a series of results based on different model configurations. However, since each modelling choice is a discrete choice - because either one model structure or another structure can be considered each time - the “variable” that is modified is a categorical type of variable and not a continuous one as in the case of OAT. In other words, it is the type of exchanges that is modified whereas in OAT the numerical value of an exchange is modified. The presence of a categorical variable prevents to calculate sensitivity rations on the modelling choice as in the case of OAT (previous section). The scenario analysis approach is then conceptually different from sensitivity analysis because its objective is quantification of uncertainty. The objective of the uncertainty analysis is to obtain a range or distribution of output values due to different modelling choices, while the objective of OAT and GSA is to obtain a measure of sensitivity of results due to the change in model structure. Action: for the product system under analysis identify a set of discrete modelling choices that are expected to influence results. (cf. also tables 3-7). For these, perform calculations using different scenarios, i.e. model structures obtained from different assumptions, and present results together making clear what is the quantitative difference that is due to the assumption made. A common approach is to test the sensitivity to the choice of energy mix. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 18 6. Sensitivity analysis – focus on model To overcome the limitation of OAT, the only approach that can allow to investigate sensitivity to model assumptions is Global Sensitivity Analysis (GSA) (Cucurachi et al., 2016; Saltelli, 2005). GSA is a systematic perturbation of the model, in the sense that all parameter values are changed simultaneously using a predefined sampling strategy (e.g. Morris, Hypercube latin, etc.) and the influence of each parameter on the results under varying conditions is then determined using appropriate indices of sensitivity (Delta, Sobol). This approach allows to mitigate the limits of OAT previously mentioned. For example, calculating the average and standard deviations of the sensitivity ratios for one parameter while all the others are being modified allows to derive sensitivity indices for the parameter that take into account possible changes in model structure (Morris method). GSA is however more complex than OAT to implement in practice in LCA and it is not feasible with commercial LCA software and had larger computational requirements the higher the number of parameters are included in the analysis, because this results in a larger number of combinations between these parameters with the need to perform a larger number of simulations. In some situations GSA might also be not strictly necessary: Kim et al. (2022) show that since several LCA models are linear or close to linear (the output is directly proportional to a change in input over the entire input space, i.e. across all possible values of all inputs) a simpler approach can be applied, i.e. using a correlation analysis. The tiered recommendation is to first investigate qualitatively if any model assumption might change the results of OAT, and if this is found to be the case perform a simplified version doing multiple OAT on a limited set of selected critical modelling assumptions and investigate the differences in results using sensitivity ratios. A second step can be to perform a larger simulation considering several key parameters and calculating correlation coefficients for each of those (Kim et al., 2022). Guidance for carrying out this type of analysis is provided in a separate notebook tutorial (ALIGNED-T1.4-GSA-tutorial-corrAAU.ipynb). A third and more complex but also more informative approach is to apply GSA using a predefined sampling strategy and then analysing results using corresponding sensitivity indexes. Guidance is provided in separate notebook tutorial (ALIGNED-T1.4-GSA-tutorial-FAST-AAU.ipynb). It should be noted that the GSA approach described in previous sections is in fact an approach to estimate the sensitivity to the choice of parameters values when also considering the structure of the model. However, the approach does not strictly assess sensitivity to the use of different model structures and consequently to different modelling choices, and neither does assess the or consider the combination of choices made on parameter values and modelling structure. GSA and scenario analysis can in principle be coupled to address this challenge (Blanco et al., 2020) but the theoretical approach is not yet practical operational in existing software tools to consider a large numbers of modelling choices and scenarios as in common LCA practice. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 19 Action: for the product system under analysis identify first qualitatively whether relations of interdependence between parameters require the need for a GSA. If this is the case, identify a set of parameters that must be included in the GSA. Define sampling strategies for these parameters and perform a simulation on the different models obtained from the combination of different values for different parameters. For simpler approach use uniform distributions to sample the parameters and use correlation indices to measure the sensitivity. With more resource and skills available use specific sampling strategies and specific sensitivity indexes. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 20 7. Guidelines for appraisal of uncertainty in the ALIGNED model framework In the following, recommendations are provided for understanding the uncertainties in the models used in the ALIGNED framework for assessment of bio-based products. These are organized according to the tasks in the Work Package 1 of the project, that loosely follow the ISO phases of LCA. For each model in the framework, indications are provided regarding uncertainties of data and model type. Additional guidance is provided that illustrates practical tools to be used in the uncertainty and sensitivity analysis of the LCA of bio-based products. Action: when using the approaches, methods, and tools within the aligned modelling framework, read the indications provided in the tables below before performing uncertainty and sensitivity analysis. These can e.g. guide in the choice and selection of the parameters for a sensitivity analysis as well as in the understanding of the major sources of uncertainty and consequent strategies for reduction of uncertainty where possible (e.g., via additional collection of data and checking the soundness of assumptions with specific stakeholders) or for management of uncertainty where reduction is not possible (e.g. nuancing the presentation of results by reporting on the uncertainties). Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 21 7.1. Uncertainty in background modelling (T1.1) Table 3 Uncertainty and sensitivity of data in modelling of prospective scenarios using IAMs within the ALIGNED framework. Uncertainty Sensitivity Data As prospective background databases are based on premise tool (Sacchi et al. 2022), uncertainties are related mostly to the modelling structure from output data of Integrated Assessment Models (IAMs). The compilation of future background databases also depends on the choices made by the developers such as when considering new inventories representing novel technologies being gradually implemented in key industries such as power generation, fuels steel, cement, and transport at global level. Introduction of alternative inventories for novel technologies is possible through userdefined scenarios. Although dataand time-consuming, this feature can be useful to incorporate projections for a sector, product, or technology that may not be adequately addressed by standard IAM scenarios. Compared to the standard approach (e.g. fixed ecoinvent database), the adoption of prospective background databases coupled with outputs from IAMs already represents an important advance towards a better sensitivity analysis of the effects from diverse assumptions of the background database used in LCAs. The possibility of making projections of future databases for different years (from 2005 to 2100, with time steps which vary from five to ten years) using diverse combinations of modelling assumptions can be understood as an extra layer of complexity when expressing possible future realities affecting background databases. Model Uncertainties are related mostly to the modelling structure from Integrated Assessment Models (IAMs), Shared socioeconomic pathways (SSPs), and climate policy implementation assumptions which, in turn, are largely affected by the potential trajectory selected for atmospheric radiative forcing associated to the Representative Concentration Pathways (RCPs). Market structure can vary depending on the selected LCA approach (e.g. consequential or attributional). Considering that the world supply chain is aggregated into a few regions, there is an intrinsic uncertainty associated to LCAs studies adoption a smaller geographic scope (e.g., countries, states, counties, etc.). Sensitivity is indirectly covered by the model’s scenario analysis. For a selected year of the future database, it is possible to select between two IAMs of high reliability (REMIND and IMAGE). In terms of narratives or storylines for the future, the current version of premise covers three SSP options in REMIND (SSP1Taking the green road, SSP2Middle of the Road, and SSP5-Fossilfueled development: Taking the highway) and one SSP option in IMAGE (SSP2). Besides, different climate policy assumptions here ranked by order of stringency, can be selected: no policy implementation (Base), National Policies Implemented (NPi), Nationally Determined Contributions (NDC), and the achievement of different CO2 emission peak scenarios by 2100 (PkBudg1150 and PkBudg500) according to Paris Agreement Objectives. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 22 7.2. Uncertainty in foreground modelling (T1.2) 7.2.1. Dynamic carbon flux model Table 4 Uncertainty and sensitivity in modelling dynamic carbon fluxes within the ALIGNED framework. Uncertainty Sensitivity Data Not all species-specific data are available and some are assumed. Example: “share of above ground biomass harvested” for “Kiggelaria africana” as the data were not available. There is high variability across tree species in the growth and biological parameters to it is recommended to calculate results for more than one to account for this. Monte Carlo simulation is not implemented in the model but the model is parametrized and can thus support this type of simulation (with appropriate Excel add-ins). Results are very sensitive to the chosen rotation period value as well as carbon content in the biomass, it is recommended to choose these as accurately as possible. The model is parametrized so allows to easily calculate the sensitivity to changes in parameter values. Note that the model is nonlinear. Notes from sensitivity/correlation test performed in excel, based on ~30 scenarios: • Life cycle CO2 uptake (C balance): Strong correlation with S1 carbon factor (0.9+) and S1 basic wood density (0.5+), moderate with S1 rotation time (0.3+) • Life cycle CO2 uptake (GTP): strong correlation with S1 rotation time (0.8+), moderate with S1 wood density (0.5+), S1 carbon content (0.6+) • Life cycle CO2 uptake (GWP): strong correlation with rotation time (0.9+), moderate with carbon factor (0.4+) • Life cycle CO emissions (C-balance): strong correlation with S1 carbon content (0.8+), moderate with S1 wood density (0.6+) Model The choice of substituted activity regarding the biomass from thinning is associated with uncertainty. The uncertainty concerns the assumption that biomass from thinning is burned and thus substitute short-rotation wood. What is uncertain is the location of the substituted activity e.g. short rotation wood from south American plantation rather than European ones. This needs to be modelled ad hoc in the model. The choice of carbon pool and the choice of indicator for the climate impact are very impactful on the results. The model is parameterized and can in principle support the use of global sensitivity analysis. The model is nonlinear. Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 23 7.2.2. Constraints to biomass availability Table 5 Uncertainty and sensitivity in modelling constraints to biomass availability within the ALIGNED framework. Uncertainty Sensitivity Data Data from FAOSTAT are assumed to have high reliability, however some of those are estimated by FAOSTAT and not the result of direct measurement / reporting. It is possible to calculate uncertainty estimates of the model coefficients (annual increments) using the standard error of the regression (SER estimate). This gives a quantitative indication of the size of uncertainty in the estimated historical increment. The model is based on historical data from FAOSTAT. These are representative of past conditions and thus have high uncertainty in making predictions to the future. It is recommended to substitute historical data with scenario ones when available, e.g. FAO agricultural outlook (https://www.oecd.org/publications/oecdfao-agricultural-outlook-19991142.htm). Data are per country level, there is variability across countries and it is recommended to perform the analysis for multiple countries. The model is substantially sensitive to the choice of the timeframe for the analysis (start year and end year chosen to calculate the historical increment). For the same number of years under analysis, it is recommended to calculate results using different start and end years (e.g. 20142019 and 2015-2020) to nuance the conclusions. Model The model is based on linear regression, this model might not reflect fully the reality of the phenomenon as growth. The use of other nonlinear models is possible to obtain higher accuracy. Use of R2 and AIC statistics is recommended to quantify the accuracy of the prediction. The model is substantially sensitive to the choice of the timeframe for the analysis (start year and end year chosen to calculate the historical increment). For the same data sources, it is recommended to calculate results using different periods (e.g.5y range and 10y range) to nuance the conclusions. The model to calculate the composition of the country mix is substantially sensitive to the choice of countries to be included in the analysis. It is recommended to include a minimum of ten countries in the analysis. A non -exhaustive list of strategies for choosing the countries to include are on following, to be chosen based on what is most sound in the specific case (for regional markets choosing the neighbouring trading area might be sufficient but for good traded globally a larger number of producing countries should be considered). Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 24 7.3. Uncertainty in Life Cycle Impact Assessment (T1.3) Table 6 Uncertainty and sensitivity in life cycle impact assessment within the ALIGNED framework. Uncertainty Sensitivity Data Characterization factors (CFs) are sometimes provided with uncertainty ranges (e.g. for Iordan et al., (2023) biodiversity losses due to GHG emissions). Temperature impulse response functions (to compute e.g. GTP midpoint metrics) have uncertain variable calibrations, as these depend on the selected background climate model (Olivié et Peters, 2013) CF proportionally affects the LCA scores. For CF with uncertain ranges documented, an OaT can be performed with the brightway framework. Model Which climate metric to choose (see recommendations of task 1.3), as the further in the impact pathway (the cause-effect chain of ecosystem modelling), the higher the uncertainty: selecting different metrics to reduce the uncertainty. Conclusions are sensitive to the time horizon selected when computing relative metrics such as GWP and GTP. Duration of the life cycle duration is also relevant. Both can be assessed as perturbation/scenario analysis (what if..), to see if conclusions change. For endpoints damages indicators (e.g. biodiversity loss), assess at least two different LCIA method (e.g. LC-impact and Impact world +). Horizon Europe grant agreement N° 101059430. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 25 7.4. Uncertainty in socio-economic assessment (T1.5) Table 7 Uncertainty and sensitivity in socio-economic assessment within the ALIGNED framework. Uncertainty Sensitivity Data Depending on the TRL of the technology/process/product data can be subject to uncertainty. In general, uncertainty tends to be the highest under low TRL and gradually decreases with increasing TRL (maturity). Potential mitigation strategy: scenario analysis. All indicators are primarily focused on a microalga case study. However, the goal was also to focus on the biobased sector. Hence, the indicators can be used as a starting point for the Aligned project and need to be slightly adjusted in a further stage. Potential mitigation strategy: Project internal expert consultation. Social indicators are usually qualitatively assessed and are subjective. It is hard to quantity them. For now, the social indicators are based on a microalga case study. Potential mitigation strategy: further literature review, scenario analysis. Specifically for social indicators, the value of the country-specific indicator is based on the average of the entire country which might deviate from the correct value that is determined on a firm level. Potential mitigation strategy: further literature review, scenario analysis. Data can be sensitive to the choice of country, especially for environmental indicators and social indicators. Certain indicators such as electricity mix (environmental) or wages (social) can deviate depending on the country/region. Potential mitigation strategy: It is recommended to include a pre-defined number of countries that is used among all parts of the novel Aligned framework. Model The prospective aspect is still lacking. The model is constructed in such a way that it represents the present. Specifically, the social indicator unit does not have a prospective nature. For now, the model is decoupled from the consequential model that is currently under development in the Aligned project. The model does not include learning effects/curves. Potential mitigation strategy: Aligned and harmonized learning model across the entire Aligned model. The model is built in an integrated way. All technological data points are connected to the economic, environmental, and societal part of assessments. Hence, a deviating technological parameter will cause a deviation in results in all three domains.