Partitioning product footprint changes into yield and environmental improvement effects: toward life cycle assessment of agricultural system transitions

Recent agricultural policies require the complementary use of area-based and product-based environmental indicators to assess agricultural system transitions because both alleviating productivity-environment trade-offs and increasing food value chain sustainability are crucial in successful policy decisions. However, there is a dilemma due to the exclusiveness between representing trade-offs in the transition paths and connecting the calculated results to the assessment of downstream food products. Here, this paper proposes a procedure to resolve the dilemma. The dilemma-resolving procedure is to partition product footprint changes into yield and environmental improvement effects in the same unit as the product footprint. After specifying a typology of transition toward agricultural sustainability, the partitioning procedure was developed on the basis of mapping from a two-dimensional space (area-based indicators) to a one-dimensional space (product-based indicators). To demonstrate the effectiveness and performance of the partitioning procedure, this paper applied it to the three categories of agricultural system transitions, including those to organic agriculture, integrated production, and production systems involving new agricultural inputs such as microbial inoculants and biostimulants, using the cases of the impact category of global warming (carbon footprint). The transition dataset (matched pairs) was constructed using a bibliographical survey. The results indicate that the partitioning procedure is effective and has good performance: (1) it was able to apply to all the cases in the dataset and to classify all the cases into six specified transition types; (2) it was differentiable between the three transition categories; and (3) it was able to explain the features of each transition category. These results imply that the perspective on agricultural system transitions can be integrated with the perspective on food value chains. Therefore, productivity-environment trade-offs during the research and development phase of agricultural technologies can be linked to environmental mitigation practices along the food value chains. The results imply that, by applying the partitioning procedure, the recent agricultural policy framework contained two exclusive perspectives can be explained consistently. Every stakeholder related to agricultural policy making should be explicit about product-environment trade-offs during agricultural system transitions, as well as about food value chain sustainability. The partitioning procedure facilitates such stakeholder practices.


Introduction
biodiversity on farmlands (Kleijn et al. 2011), and to close yield gaps (Mueller et al. 2012). These practices include negative emission technologies (Northrup et al. 2021) and microbiome management strategies (French et al. 2021). Policy decisions about the implementation of sustainable agricultural practices should rely on multiple indicators (footprints) to measure the performance of new practices switched from conventional ones, wherein the indicators should summarize the cradle-to-grave information across entire food value chains. The former perspective on the performance measurement is the key to resolve the trade-offs between productivity and environmental burden (U.S. Department of Agriculture 2020) and the latter life-cycle perspective is the basis for the Farm to Fork Strategy (European Commission 2020a).
This duality of perspectives corresponds to two types of environmental indicators (environmental impacts per functional unit) used in life cycle assessment (LCA) of agriculture (see, e.g., Nemecek et al. 2011;Hayashi 2013). They include area-based indicators, in which impacts are measured per unit of land area (and time), and product-based indicators, in which impacts are measured per unit of product. The former is linked to the narratives on ecosystem services provided by agriculture (Kremen and Miles 2012); it comprises a background for the use of the land area as the functional unit for organic agriculture from the early stage of methodological development (Haas et al. 2001). Productivity-environment trade-offs-improved productivity entails decreased environmental performance per area and vice versa-constitute a primary concern in agricultural system transitions toward sustainability. The latter corresponds to the view that agricultural production is a decisive factor in global commodity supply chains (Vidergar et al. 2021); therefore, the functional unit of product is common in the practice of product environmental footprint (PEF) (European Commission 2018). Under these circumstances, the complementary use of both indicators (i.e., functional units) has been recommended and practiced in comparative LCA of agricultural systems (see, e.g., Nemecek et al. 2011;Hayashi 2013;Hokazono and Hayashi 2015); in fact, both GHG emissions per product (product carbon footprint) and those per area have been used in many earlier LCA studies (see Tables S1 and S2). Furthermore, a visualization framework to integrate both indicators has already been proposed within LCA (Hayashi et al. 2016).
However, there is a dilemma. If we select area-based indicators, we can express productivity-environment tradeoffs to envision the paths of agricultural system transitions, although the results are not applicable to the assessment of downstream products such as processed foods and dishes because of inconsistency in the unit. On the contrary, if we select product-based indicators, we can use the results in the assessment of downstream products, although the trade-off information disappears. To resolve this dilemma due to exclusiveness, establishing coexistence of both the representability of the transition paths and connectivity between upstream and downstream processes is required.
Here, this paper proposes a procedure to resolve the dilemma by partitioning the changes in product footprints into yield and environmental improvement effects, both of which are defined in the unit of product, by integrating land-and product-based expressions. By the term product footprints, this paper indicates the results of product LCA, which include calculated numerical values of PEFs and carbon footprint. The yield improvement effect means the share of positive changes in product footprints caused by crop productivity improvement (positive changes in crop yield), and the environmental improvement effect indicates the share of those caused by land-based environmental improvement (positive changes in environmental impacts per area). The partitioning procedure developed is the first to show how many degrees of product footprint changes are caused by productivity improvement (the yield improvement effect) and by land-based environmental improvement (the environmental improvement effect) on the two-dimensional space that has the same unit (e.g., kg CO 2 equivalent per product) as the measure of product footprints. In other words, so far, it was not possible to illustrate the degree of contributions made by yield and environmental improvements because there were no procedures to partition the changes in the product footprint into the two effects. The two partitioned effects can be used in LCA of foods and drinks at the consumption stage because they are measured in the same unit as the product footprint.
After showing the procedure to distinguish both effects and providing a typology of transitions, the usefulness of partitioning in understanding transitions to sustainability is demonstrated by applying the procedure to earlier publications on comparative LCA of agricultural production systems. Three categories of transitions were investigated, including those to organic agriculture, integrated production, and production systems using new agricultural inputs such as microbial inoculants (Ahmad et al. 2018) andbiostimulants (Jardin 2015;Yakhin et al. 2016). Although this paper analyzes the cases of the impact category of global warming (carbon footprint), the partitioning procedure this paper proposes can cover the other impact categories because the same two-dimensional spaces can be defined between, for example, crop yield and eutrophication measured by, e.g., kg phosphate equivalent. Through the development of the partitioning procedure, this study aims to support the understanding and practice of an agricultural policy framework compatible with productivity-environment trade-offs during agricultural system transitions and with the sustainability of the food value chain.

Visualization of the transition to agricultural sustainability
The foundation used to partition product footprints into yield and environmental improvement effects is visualized as transitions illustrated in land-oriented and product-oriented expressions (Hayashi 2013). In the land-oriented expression based on area-based indicators, a two-dimensional space is defined by crop yield as the horizontal axis and the environmental impact per area as the vertical axis, and the transition paths for agricultural systems are illustrated using arrows, in which sources correspond to conventional systems (references for comparisons) and destinations correspond to alternative systems (Fig. 1a). In the latter expression based on product-based indicators, environmental impacts per product (product footprints) for each source and destination of the arrows are visualized (Fig. 1b). Although both expressions can be linked because environmental impact per product is defined by environmental impact per area (a value on the vertical axis) divided by crop yield (a value on the horizontal axis), as illustrated in Fig. 1, an inconvenience in applying LCA to analyze the transition of the agricultural systems appears. The productivity-environment trade-offs shown in land-oriented expression, which are useful agronomic information for realizing sustainability, disappear in product-oriented expression, which is a default application mode of product LCA used in calculating PEFs.

Typology of transitions
Before illustrating the partitioning procedure to resolve the analytical inconvenience, we will elaborate on the typology This classification is not consistent with the environmental impacts per product (product footprints) in the sense that A, B, and F are indifferent and C, D, and E are also indifferent ( Fig. 1b) because these are based on ratios and are related to the inconvenience caused by the disappearance of trade-off information. The method to be adopted to resolve this problem is proposed in the next subsection and is shown in Fig. 2b. In contrast to Fig. 2a, which uses different units for horizontal and vertical axes, Fig. 2b is based on the same unit for both axes, and the sum of both axes measures the environmental impacts per product (product footprints), which is shown as an isoquant line.

Partitioning procedure
The partitioning procedure proposed in this study is based on the mapping from the two-dimensional space (area-based indicators) to the one-dimensional space (product-based indicators). The procedure preserves the productivityenvironment trade-offs contained in the two-dimensional space. The implications of the procedure are dependent on the transition types as follows (type D is excluded from the following explanation): • Win-win transition (A): The partition provides the share of contributions between the improvement in crop yield and the environmental impact. • Trade-off transition 1 (B and C): This transition illustrates the number of degrees of environmental impact that are canceled by a decrease in yield, which is important for assessing organic conversion. • Trade-off transition 2 (E and F): This transition illustrates the number of degrees of environmental impact saved by an increase in yield, which is important for agricultural technologies that aim to improve productivity by applying agricultural inputs such as microbial materials, although environmental impact per area is not necessarily improved.
The partitioning procedure is illustrated for each transition type (Fig. 3) and the following is an example for type A: 1. The first step is the decomposition of the vectors. ����� ⃗ RA = ����� ⃗ RO + ����� ⃗ OA for transition type A (Fig. 3a). 2. The second step is a coordinate transformation from the two-dimensional space between the crop yield (y) and the environmental impact per area (e) to a one-dimensional space defined as the environmental impact per product (e/y). For example, R = y R , e R is changed into θ R = e R ∕y R . 3. The third step is the calculation of the yield improvement effect, which is defined as R − O , and the environ- The same procedure is applicable to the other transition types, and the cases for types B and F are illustrated in Fig. 3b and c.

Path dependency and negative emissions
Path dependency must be considered in the partitioning procedure. There are two decomposition paths for the transition vectors on which the results of the partitioning depend: the yield-environment path (R-O 1 -B in Fig. 4) and the environment-yield path (R-O 2 -B). Path dependency can be identified by comparing the following ratio between the two paths: the environmental impact difference divided by the yield difference. The ratio in the two-dimensional space is transformed into the one-dimensional space. Transition type B is used for explanation as follows ( Fig. 4a): For the vector decomposition ����� ⃗ Fig. 3, the ratio can be transformed into where A implies the coordinate transformation for A, where the coordinate x is divided by the coordinate y. If there is no difference in yield between the two agricultural systems, the yield improvement effect is defined as 0.
In contrast, for the decomposition ����� ⃗ can be transformed into Therefore, there exists a difference in the ratio between the two paths; the ratio of the environmental impact difference to the yield difference is multiplied by y R ∕e R for the yield-environment path and by y B ∕e B for the environmentyield path.
Although the above result implies that the calculation using the geometric mean of both multipliers  Fig. 3 Partitioning procedure of product environmental footprint into the yield and environmental improvement effects for each transition type. a Procedure for type A, b procedure for type B (almost the same for type C), and c procedure for type F (almost the same for type E) would be possible, there is an additional issue in assessing the transition of the agricultural systems. Negative GHG emissions can occur in the case of, for example, organic agriculture, and this causes a problem related to the coordinate transformation of negative emission cases; in the region where y is positive and e is negative, the sign of the transformation of O 2 − B is the opposite of the expectation (Fig. 4b).
The same logic is applicable to transition type A (transition types A and B are the cases where negative GHG emissions occur). Therefore, this study proposes to use the partition procedure based on the yield-environment path.

Data on agricultural system transitions
To demonstrate the utility of the partitioning procedure, three categories of agricultural system transitions were analyzed in this study. First, the transition to organic agriculture, which plays an important role in agri-environmental policies in the EU and Japan. The European Commission set a target to increase agricultural land under organic farming to at least 25% by 2030 (European Commission 2020a, b), and the Ministry of Agriculture, Forestry and Fisheries of Japan has set an objective of making organic farmlands at least 25% of farmlands by 2050 (The Ministry of Agriculture, Forestry and Fisheries of Japan 2021). Second, transition to integrated production. Although integrated production can be classified as a type of conventional agriculture, this study distinguishes integrated production from conventional production. Third, transition to agricultural systems using new materials such as microbial inoculants and biostimulants. This study tried to include additional integrated production practices such as integrated pest IPM) and biological and ecological control, which are based on a wide range of ecosystem services. However, no cases were present in the dataset created by scrutinizing extracted publications. The transition dataset (matched pairs) was constructed using a bibliographical survey using the Web of Science Core Collection on May 20, 2022. Earlier publications on LCA studies under each transition type were defined as the intersection ("AND") of publications on LCA and those on each transition type (Table 1). Publications until 2021 were included in the dataset.

Queries
Life cycle assessment Articles related to LCA were defined as the union ("OR") of the query "ALL = "life cycle assessment$"" and "ALL = "life cycle analys?s"". The number of articles in the former category was 29,855 and that in the latter category was 3,878; the union was 32,394.
Transition to organic agriculture The articles on this transition were defined as the union of queries using multiple word combinations related to organic agriculture (Table 1). The number was 19,880 and the intersection with the above LCA articles was 265.
Transition to integrated production The query "integrated production$" was used for this transition and the union was 31.

Transition to agricultural systems using inoculation (microbial inoculants and biostimulants)
The query words "inoculant$" and "biostimulant$" were used for this transition and the union was 4 and 11, respectively.

Construction of the dataset
After extraction, the dataset was created by checking each publication (Tables S1 and S2). Articles showing results (numerical values) on crop yield and life cycle GHG emissions (per area and/or per product) of both alternative and conventional production systems were included. This study only analyzed crop production systems, and animal production systems were excluded. If results in the articles were shown only in graphical formats, emails were sent to the authors requesting them to provide numerical results, although no responses were available, and the articles were not inserted. Furthermore, the following adjustment was applied: • Minimal tillage was not included in conventional production, although integrated production was considered a type of conventional production in the case of the transition to organic agriculture. • Several publications on the transition to organic agriculture and integrated production that were not shown in the literature search results were also inserted into the dataset. This adjustment was made because the search results were incomplete. For example, although the query result on organic agriculture included the papers on integrated production, the search result on integrated production did not include the same papers. Therefore, the data on the transition to integrated production were expanded using the data accessed through the results of the query on the transition to organic agriculture.

Testing the partitioning procedure
This paper analyzes the three categories of agricultural system transitions using the cases of the impact category of global warming (carbon footprint) to demonstrate the following three research questions. The first question is whether the partitioning procedure can be feasible and applicable to case studies in the real world. Although the partitioning procedure was developed to deal with realworld cases, the effectiveness of the partitioning procedure should be verified by using actual case studies. The second is how differentiable the partitioning procedure is for the three categories of agricultural system transitions. The third is how the partitioning procedure can explain the features of each agricultural system transition. Differentiability and explainability measure the performance of the partitioning procedure and are important in designing sustainable agricultural systems.
To investigate the differentiability of the three types of agricultural systems transitions, statistical analyses were performed using R version 4.1.2 (R Core Team 2021). For the Steel-Dwass test (calculating the p-value for the observed Dwass, Steel, Critchlow, Fligner W statistic), the function "pSDCFlig" in the version 1.16 of the package "NSM3" (Schneider et al. 2021) was applied; the Monte Carlo method with 100,000 permutations was used. As the gathered samples included a wide variety of crops, varieties, and cropping types, the values for the multiple comparison tests (improvement in carbon footprint and Organic "organic agriculture" OR "organic conversion$" OR "organic cultivation" OR "organic farm$" OR "organic farming" OR "organic field$" OR "organic practice$" OR "organic production$" OR "organic management" 19,880 265 55 137 Integrated production "integrated production$" 2,182 31 8 12 Inoculation "inoculant$" 6,783 4 2 3 "biostimulant$" 2,228 11 1 1 Others "integrated pest management" OR "IPM" 28,996 18 0 0 "biological control$" OR "biocontrol$" OR "biological pest control$" 74,035 11 0 0 "ecological control$" OR "ecological pest control$" 501 0 0 0 yield and environmental improvement effects) were normalized using the midpoint of the carbon footprint value for the conventional system and that for the revised systems (organic agriculture, integrated production, and the systems using inoculation). Additionally, to visualize the relationship between two indices to feature the transition to organic agriculture (one is an index calculated from the result of the partitioning procedure and the other is a conventional index used in meta-analysis), the locally weighted scatter plot smoothing line and 95% confidence intervals were used. Packages such as "ggplot2" version 3.3.5 (Wickham 2016) and "ggExtra" version 0.10.0 (Attali and Baker 2022) were used to draw histograms and density plots.

Effectiveness of the partitioning procedure
The partitioning procedure developed in this study was able to apply to all the cases in the dataset as expected (Tables S1 and S2); negative emission cases were correctly calculated. As a result, we were able to classify all the cases into each transition type (A to F) and category (an alternative agricultural system as a destination of the system transition) as shown in Table 2. In the classification, the transition cases in which the yield difference was presumed to be zero were classified as type B rather than type A, and as type E rather than type D. Eighty-nine percent of the cases were related to the transition to organic agriculture.

Differentiability of transition categories
The results of partitioning of carbon footprint improvement into yield and environmental improvement effects can be illustrated visually using means and standard errors of yield and environmental improvement effects for each transition category (Fig. 5). They revealed the differences among transition categories, which were not identified only by carbon footprint improvement without partitions. The differences were evaluated by statistical tests (Fig. 6). First, whether the samples for the values (improvement in carbon footprint and yield and environmental improvement effects) come from populations with the same mean was checked using the Kruskal-Wallis rank sum test. The results indicated that although there are no significant differences in the improvement in carbon footprint (p = 0.79), significant differences were observed in the yield improvement effect (p < 0.01) and the environmental improvement effect (p < 0.01). Second, to differentiate the three transition types, the Steel-Dwass test was applied. Significant differences were observed in yield and environmental improvement effects between organic agriculture and the other two systems (Fig. 6). However, the difference was not significant for the improvement in carbon footprint.

Explainability of transition categories
Another utility of the partitioning procedure is the enhancement in explainability of product footprints, the performance of featuring agricultural system transitions using partitioned product footprints. Tendencies in each transition category were specified using the number of transition types (A-F).

Transition to organic agriculture
All transition types were found in earlier case studies on the transition to organic agriculture (Table 2); in terms of numbers, the transitions were in the following order: #B > #C > #D > #A = #E > #F; the most preferable transition is A. In general, when the environmental performance of agricultural products (carbon footprint) is improved in conversion to organic management, the agronomic performance (crop yield) tends to decrease (#B > #A > #F). In contrast, the deterioration in the environmental performance of agricultural products (carbon footprint) during the conversion to organic agriculture is mainly due to  Inevitably, when the environmental improvement effect increases, the yield improvement effect decreases ( Fig. 5; Figs. S1, S2, S3, and S4), which is related to the result mentioned above that differentiates between organic agriculture and the other two systems in terms of yield and environmental improvement effects.
To further clarify the point, the relationships between the cancelation ratio (the yield improvement effect divided by the environmental improvement effect, which illustrates the number of degrees of the positive environmental improvement effect that were canceled by the negative yield improvement effect) and the response ratio (environmental impact per product for organic agriculture divided by that for conventional agriculture, which is a common indicator in meta-analysis) in types B and C were analyzed (Fig. 7). For the cancelation ratio, the boundary between types B and C is -1 (the larger, the better), while for the response ratio, the boundary is 1 (the smaller, the better). Figure 7 illustrates that the distribution of the cancelation ratio is left-skewed compared to that of the response ratio, types B and C are not distinguished clearly in both ratios, and the cancelation ratio captures different aspects from the response ratio as shown in the smoothing line.

Transition to integrated production
The types of transition to integrated production were A, B, E, and F and the number of each transition type was 4, 6, 1, and 1, respectively. In most cases, the environmental improvement effect was positive, while the yield improvement effect was small; that is, the transition to integrated production did not involve a decrease in yield, although the environmental improvement effect was smaller than that of the transition to organic agriculture.

Transition to agricultural systems using new materials such as microbial inoculants and biostimulants
The types of transition to these agricultural systems were A and F. In the case of the transition to agricultural systems using microbial inoculants, all the transition types observed were A; for that using biostimulants, the type observed was F, although the decrease in the environmental improvement effect was small. These results are consistent with those that were theoretically expected. Carbon footprints will be improved due to the introduction of the new materials when both the yield and environmental improvement effects are positive (type A) and even when the environmental improvement effect is slightly negative (type F).

Revealing productivity-environment trade-offs in product footprint changes
The results clarified the effectiveness and performance (differentiability and explainability) of the partitioning procedure by applying it to the assessment of agricultural system transitions. These results imply that the partitioning procedure can resolve the dilemma involved in the use of area-based and product-based indicators; that is, we can now reveal productivity-environment trade-offs using productbased indicators (product footprints). This means that perspectives on agricultural system transitions have been integrated with perspectives on food value chains and that the recent agricultural policy framework (European Commission 2020a; U.S. Department of Agriculture 2020; The Ministry of Agriculture, Forestry and Fisheries of Japan 2021) can be explained consistently.

Reference for the research and development of agricultural technologies
The revealed trade-offs between yield and environmental effects can be a reference system for evaluating the research and development of agricultural technologies aiming at the transition to sustainability. Although increasing crop yield is one of the primary objectives in agricultural research and development, which are closely related to increase in food production and reduction of hunger, more attention will be required in the formulation of yield responses. For example, Kløverpris et al. (2020) define the yield effect due to changes in agricultural practices through a comparison of an alternative (microbial inoculation) system with a reference system (i.e., the agricultural system transition in the terminology of this study), as the environmental impacts from changes in crop production elsewhere by applying system expansion. This perspective based on modeling of indirect land use change (Schmidt et al. 2015) is important in the LCA of agricultural technologies, because if we include the effect into the LCA of organic conversion, its tendency of performance decline caused by yield decrease expands further. However, a review study reveals that the yield gap between organic agriculture with diversification practices including multi-cropping and crop rotations and conventional agriculture without such practices is smaller (Ponisio et al. 2015) and that the estimation of overseas land conversion is not straightforward (Smith et al. 2019). The explicit distinction between the yield improvement effect presented in this study and the yield effect defined by them is the first step for 1 3 further methodological development; however, the formulation of the yield effect needs careful attention because it implies an induced effect rather than the yield increase. The full assessment of induced effects should not be limited to crop yield and should include material inputs such as fertilizers and pesticides.

Partitioning procedure to link two perspectives on agricultural production
These considerations indicate the necessity of the explicit formulation of product-and area-based indicators and of the establishment of a link between them. Product footprints are product-based indicators, in which the assessment of environmental impacts along global food value chains is the main issue. In contrast, area-based indicators include ecosystem (provisioning, regulating, and supporting) services provided by agricultural land (not by agricultural goods). Earlier studies on yield comparisons between organic and conventional agriculture (de Ponti et al. 2012;Seufert et al. 2012) can be classified under the category of the provisioning service.
What this study has developed is the linkage based on the explicit distinction of the two types of indicators. The importance of this achievement can be stressed in two ways. First, previous literature surveys and meta-analyses on the comparison of life cycle environmental impacts between organic and conventional agriculture (Tuomisto et al. 2012;Meier et al. 2015;Clark and Tilman 2017) are not explicit about the link. If the partitioning procedure developed in this study is applied, the productivity-environment tradeoffs can be shown explicitly using the same unit as product footprints. Second, yield and environmental improvement effects can be distinguished with respect to the functional units used in the Product Environmental Footprint Category Rules (European Commission 2018). It implies the utility of the developed partitioning procedure in environmental labeling policies because the partitioned two effects are shown in the same unit as the product footprints. Although this study demonstrated the utility of the partitioning procedure using an impact category of global warming (carbon footprint), the developed procedure is expected to improve informed decision-making processes using LCA. Organic is better. Conventional is better.

Conventional is better.
Organic is better.

Conclusions
There was a dilemma involved in the use of both area-based and product-based indicators in the assessment of agricultural system transitions; on one hand, area-based indicators can exhibit trade-offs in the transition paths to agricultural sustainability, although they cannot be connected to the assessment of downstream products; and on the other hand, product-based indicators cannot exhibit trade-offs, although they can be connected to the assessment of downstream processes. To resolve the dilemma due to the exclusiveness, this article proposed a partitioning procedure of the product footprint changes during agricultural system transitions into yield and environmental improvement effects.
To demonstrate the effectiveness and performance of the partitioning procedure, this paper applied it to the three categories of agricultural system transitions, including those to organic agriculture, integrated production, and production systems involving new agricultural inputs such as microbial inoculants and biostimulants, using the cases of the impact category of global warming (carbon footprint). The results indicate that the partitioning procedure is effective and has good performance: (1) it was able to apply to all the cases in the dataset and to classify all the cases into each transition type (A to F); (2) it was differentiable between the three transition categories; and (3) it was able to explain the features of each transition category. These results imply that the perspective on agricultural system transitions can be integrated with the perspective on food value chains; productivity-environment trade-offs during the research and development phase of agricultural technologies can be linked to environmental mitigation practices along the food value chains. Furthermore, they mean that the recent agricultural policy framework can be explained consistently; the partitioning procedure can facilitate every stakeholder related to agricultural policy making to be explicit about product-environment tradeoffs in product footprint changes during agricultural system transitions, as well as about value chain sustainability.
Funding This work was supported in part by the Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (KAKENHI) Grant Number 18K11745.

Data availability
The data used for this research are provided in the electronic Supplementary Information (two files).

Conflict of interest The author declares no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.