Recycling sludge in agriculture? Assessing sustainability of nutrient recovery in Italy

Abstract

Using a hybrid multi-regional input–output approach, we traced sustainability footprints of a nutrient recovery strategy from sewage sludge applied in Italy. We then compared the results with the most common landfilling practice. Overall, accounting for indirect global upstream effects, using sewage sludge for organic fertiliser production generates more jobs and reduces more greenhouse gas emissions than landfilling. By contrast, landfilling stimulates the whole economy more, generating higher indirect turnover and reduces energy carrier use more. Finally, we accounted for uncertainties in these results using an error propagation method based on Monte Carlo simulations.

1. Introduction

The link between a prosperous society, a competitive economy and a healthy planet places sustainable agro-food systems at the heart of the European Green Deal (EC, 2020a). Agro-food systems play a crucial role in ensuring food security and achieving key targets by 2030, such as reducing greenhouse gas (GHG) emissions by 55 per cent, pesticide use by 50 per cent and fertiliser use by 20 per cent compared with 1990 figures (EC, 2023a; Montanarella and Panagos, 2021). However, supply chain shocks and shortages, such as those arising from the pandemic of coronavirus disease 19 pandemic, climate-induced extreme events¹ and the Ukraine–Russia conflict, have brought to the fore the systemic structural weaknesses in agricultural systems, thus weakening the ability to achieve these sustainability targets (Behnassi and Haiba, 2022; Sperling et al., 2022).

Disruptions in the trade of raw materials and primary products induce greater volatility in agri-food commodities and fertiliser prices, which have skyrocketed (Behnassi and El Haiba, 2022). In the fertiliser market, prices already doubled between the summer of 2020 and the end of 2021 (Smith, 2022) because of (i) the rising price of natural gas (which accounts for 80 per cent of the operating costs in nitrogen fertiliser production), (ii) market distortions because of the pandemic (e.g. China suspended fertiliser exports until the end of June 2022 to ensure domestic availability) and (iii) the increase in agricultural commodity prices, which incentivised higher fertiliser use. The war in Ukraine has exacerbated this trend, causing fertiliser prices to increase by 3–43 per cent (depending on the fertiliser type) in March 2022 compared with February 2022 prices (EC, 2022a) and impacting the European Union (EU),² in particular. Although the current increase in fertiliser prices has been mitigated to some extent by the current agronomic season,³ there are major concerns about the risk of fertiliser unavailability in the coming period. Changes in fertiliser prices are also more volatile than changes in the prices of other agricultural inputs because of demand rigidity, which has been enforced because fertilisers are essential for production and have few substitutes (Beckman and Riche, 2015).

While there are handy substitutes for some agricultural production inputs (e.g. capital for labour), chemical fertilisers tend to have few alternatives. One potential solution, along with technological developments and smart farming (Moysiadis et al., 2021), is the application of organic fertilisers, such as manure and treated sewage sludge⁴ (Kumar Bhatt, Labanya and Joshi, 2019). The market share of organic fertilisers (without manure) is estimated to be 6 per cent of that of inorganic products (Chojnacka et al., 2019). Due to the drastic increase in the prices of mineral fertilisers and natural gas in the years 2007 and 2009, during this period, the agricultural use of sludge-based fertilisers increased by 51 per cent in France and more than 100 per cent in the USA (Lupton, 2017). The same phenomenon is expected in the coming years; to address the economic effects of the Ukraine crisis, for example, Italy enacted an urgent directive (Decree-Law No. 21/2022) to allow for the replacement of chemical fertilisers with sludge-based organic fertilisers. Assessing the sustainability of sludge management has then become an even more crucial task for scientists (Kumar, Chopra and Kumar, 2017).

Sustainability is a condition of balance, resilience and interconnectedness that enables society to maintain and sustain its needs and those of future generations (Morelli, 2011). Sustainable development is therefore based on three pillars: economic, social and environmental (Purvis, Mao and Robinson, 2019). In accordance with this view, our paper is a preliminary sustainability assessment based on empirical evidence gathered from a case study in Italy integrated with a multi-regional input–output (MRIO) framework. The use of this framework allows us to begin investigating the extent to which a waste-based nutrient recovery strategy can be responsible for contributing to economic, social and environmental targets. This pressing question is challenged by the way in which impacts can be monitored, particularly with regard to changes in demand in global upstream supply chains.

Sewage sludge is the by-product of municipal or industrial wastewater treatment plants (WWTPs). Because of population growth, urban planning and industrial development, the volume of sewage sludge produced has rapidly increased (Kumar, Chopra and Kumar, 2017). In addition, the implementation of Directive 91/271/EC to improve wastewater collection and treatment is also causing a significant increase in annual sewage sludge production in the EU (Kelessidis and Stasinakis, 2012).

In Italy, wastewater and sludge management is an expensive and long-standing problem. Since 2004, Italy has been violating EU regulations governing the collection and treatment of urban wastewater (Directive 91/271) by failing to ensure that 620 agglomerations (i.e. cities, towns and settlements) have adequate WWTPs; as a result, Italy faces fines and penalties of more than 60 million euros per year (Court of Justice of the European Union, CJEU, 2018; EC, 2022b). Furthermore, in line with the waste hierarchy in the EU, landfilling should be the very last resort (Directive 2008/98/EC). However, Italy recycles only 44.1 per cent of over three million tons of sludge produced per year (ISPRA, 2022), whereas out of the 1.6 million tons of sludge disposed of, 15 per cent is landfilled, 7.3 per cent incinerated (ISPRA, 2022) and the remaining ends up in a limbo of pre-treatment classifications for later landfilling or incineration, but whose ultimate end is unknown.⁵ In April 2022, in violation of Directive 1999/31/EC restricting the landfilling of waste, the CJEU sentenced Italy for failing to permanently close or restore all 44 landfills identified in March 2019 in violation with this Directive.

Along with Italy, other Member States are failing to properly manage wastewater and recycle sewage sludge. To date, other 16 Member States⁶ have active violations of Directive 91/271 (for a total of 2,381 agglomerations without adequate WWTPs) and/or Directive 1999/31/EC (for a total of 262 landfills to be closed or rehabilitated) (EC, 2023c), a relevant issue especially for some (new) EU Member States (e.g. Malta, Croatia and Romania), where 67–100 per cent of sewage sludge is landfilled (Hudcová, Vymazal and Rozkošný, 2019; Kelessidis and Stasinakis, 2012). From a bioeconomy perspective, however, rather than a problem, wastewater and sewage sludge management offers significant opportunities for strategic innovation in nutrient recovery for agricultural purposes.

Rich in nutrients and organic matter, stabilised sewage sludge can improve soil fertility, texture and chemical–physical properties. Several field trials were conducted to test the response to the application of sewage sludge fertilisers in comparison with inorganic fertilisers: Debiase et al. (2016) observed a 16 per cent increase in winter wheat grain yield in Italy, Koutroubas et al. (2020) observed a 14 per cent increase in sunflower achene yield in Greece and Dubis, Szatkowski and Jankowski (2022) obtained equal or slightly higher yields for an energy crop in Poland. In longer experiment settings (more than 10 years), Kätterer, Börjesson and Kirchmann (2014) reported a 6 per cent increase in grain yield of oats and spring barley and Gómez-Muñoz, Magid and Jensen (2017) reported up to a 43 per cent increase in the N concentration and, consequently, in the protein content and quality of oat grain and straw. There is also growing interest in investing in new technologies to recover phosphorus from sewage sludge (as a phosphorus-rich waste) (Tonini, Saveyn and Huygens, 2019; Di Capua et al., 2022) to address the emerging global challenge of phosphorus scarcity (Cordell et al., 2011); it is estimated that sewage sludge can cover up to 15–30 per cent of current mineral phosphate fertiliser demand in Europe (Battista et al., 2020; Smol, 2019). All cited studies have stressed the importance of using properly processed sludge to avoid critical levels of soil contamination (e.g. heavy metals, organic compounds and pathogens) (Fijalkowski et al., 2017).

In the EU, the safe reuse of sewage sludge in agriculture is regulated by Directive 86/278/EEC. The more-than-35-year-old directive allows Member States to set different precautionary standards and limit values for sludge use on farmland; for instance, application doses can range from a minimum value of 1.66 tons of sludge per hectare per year, as in Germany and the Czech Republic, to maximum annual loads of 100 tons of sludge (dry matter) in Lithuania for the purpose of rehabilitating damaged sites or cultivating energy crops (European Commission, 2018). After an evaluation of the directive launched on 16 June 2020, the EC has emphasised that the directive no longer meets actual needs and current expectations (EC, 2020b). Thus, the commission is considering revising the wastewater treatment and sewage sludge directives as part of the New Circular Economy Action Plan adopted on 11 March 2020. In addition, given the promising technical progress made, the new EU Fertiliser Regulations will establish new conditions to facilitate the access of organic and waste-based fertiliser to the EU Single Market (EC, 2022c).

Our study contributes to this strand of literature in several ways. Although several studies have been conducted to assess the sustainability of sludge management strategies, there is a lack of a systematic approach (Yoshida, Christensen and Scheutz, 2013). In particular, the decision on the system boundary is at the expert’s discretion, leading to truncation errors that limit the robustness of the applied analysis (Crawford et al., 2018; Ward et al., 2018). In the present study, we address the sustainability assessment by applying a demand-driven MRIO model capable of covering an infinite order of contributions from upstream production processes (Wiedmann, Lenzen and Barrett, 2009). We quantify the direct and indirect impacts on cumulative intermediate expenditure, job creation, GHG emissions and energy consumption that a sludge-based nutrient recycling strategy adopted in Italy can have within the global supply chain. The same is done with landfilling, the results of which are then used as a comparative reference for the recycling strategy. Researchers, policymakers and practitioners can benefit from our research to improve the sustainability of sludge management by gathering important insights from management and policy perspectives on the economic structure of a nutrient recovery strategy and associated issues, such as sustainability impacts embodied indirectly in production, consumption and global trade flows.

Besides the contemporary relevance of the topic, this article contributes to the literature on production economics and ecological footprinting by applying a formal uncertainty analysis to the input–output framework. The use of input-output (IO) tables to assess systems-wide impacts is not new to the agri-food supply chains literature, but their popularity increases with the increase in the level of details with which these databases reflect the industrial interdependencies within an economy (Hughes, 2003; Okuyama and Santos, 2014). Agricultural economists have used IO tables to examine macroeconomic spillover effects on output and employment from a change in economic activity, such as a change in the mix of forest plantings (Eiser and Roberts, 2002), the establishment of a large-scale ethanol plant (Thomassin and Baker, 2000) or a new rural development policy (Hyytiä, 2014; Cruz et al., 2017). More recently, Wahdat and Lusk (2023) used IO tables to assess the vulnerability of food industries to upstream industries from the perspective of intermediate inputs and labour. However, the influence of uncertainty remains unclear in most studies. The present study complements this aspect by extending the MRIO frameworks into stochastic forms through a Monte Carlo simulation, thus providing uncertainty intervals for the estimated total impact multipliers.

2. Methodological framework

In any impact assessment, the determination of an appropriate system boundary affects the reliability of the results. Identifying which activities to include in the analysis is often left to the judgement of the practitioner (Rajagopal Vanderghem and MacLean, 2017), whose arbitrary selection of a finite boundary can lead to truncation errors (Crawford et al., 2018). To avoid cutting out relevant interactions with the wider economy, we applied a hybrid MRIO analysis, which can handle infinite supply chain systems in the sustainability assessment (Wiedmann, Lenzen and Barrett, 2009) by combining detailed process-based data with environmental-extended MRIO database. Several researchers have successfully applied the same techniques to estimate the carbon footprint of a selected sector (e.g. Malik et al., 2014; Lenzen et al., 2018; Wei et al., 2021).

2.1. MRIO analysis

MRIO databases are spatially explicit representations that describe the economic interdependencies among sectors and global agents (e.g. households, governments, the capital sector and stocks) within and between countries (Figure 1).

In this study, we relied on the latest version (v.8.3) of Exiobase for the year 2020 (Stadler et al., 2018). Exiobase follows a standard supply-use structure featuring 49 regions |$\left( r \right)$| (27 EU Member States, 17 major economies and five regions in the rest of the world), 163 industries |$\left( i \right)$|⁠, 200 commodities |$\left( c \right)$| and 6 final demand compartments |$\left( f \right)$|⁠. As a tool for impact assessments, Exiobase includes economic, social and environmental indicators |$\left( k \right)$| as satellite data accounts |$Q$| (or the so-called stressor matrix) (Malik et al., 2019; Södersten and Lenzen, 2020), describing the amount of stressors needed (e.g. working hours and GHG emissions) to produce the total output of industry |$i$| in region |$r$|⁠.

Thanks to this extension, MRIO datasets are becoming a popular tool for conducting consumption-based impact assessments. Based on the well-known IO framework developed by Leontief (1936), it is possible to now estimate direct and indirect (i.e. upstream) effects associated with a particular final demand shock and derive total impact multipliers (e.g. carbon footprint expressed in kilograms of CO₂-eq. emitted per unit of final demand produced). Following the conceptual framework of Lenzen and Rueda-Cantuche (2012) that formalises the application of IO-type calculations on supply-use tables, the calculation of impact multipliers in this study is conducted as follows.

Let |$T$|(Figure 1) be an |$N \times N$|monetary MRIO transaction matrix represented by

with |$U$| being the commodity-by-industry use matrix of size |$P \times S$|(with |$P = c \times r$| and |$S = i \times r$|⁠), showing the domestic intermediate (diagonal submatrices in Figure 1) and bilateral trade (off-diagonal submatrices) production structures, and |$V$| the industry-by-commodity |$\left( {S \times P} \right)$| make matrix, showing the outputs by industries in region r. |$T$| satisfies the national accounting identity:

where |${e_P}$| and |${e_S}$| are summation sub-vectors of lengths |$P$| and |$S$|⁠, respectively, |${y_c}$| the vector of final demand of products of size |$P \times 1$| and |$q$| and |$g$| the vectors of total product and industry outputs, respectively (which form the gross output vector |$x$| of size |$N \times 1$|⁠). The vector |$y$| is given by |$y = Y{e_M}$|⁠, with |$Y$| being the |$P \times M$| matrix of final demand compartments satisfied by domestic production (diagonal submatrices in Figure 1) and direct imports (off-diagonal submatrices) and |${e_M}$| the row summation vector of sizes |$M \times 1$|⁠. Equation 2 can be subjected to Leontief’s quantity demand–driven formalism (Lenzen and Rueda-Cantuche, 2012) and transformed into

where |$I$| is an |$N \times N$|identity matrix, |$D = V{\hat q^{ - 1}}$| is the market share matrix of size |$S \times P$| and |$B = U{\hat g^{ - 1}}$| is the use coefficient matrix (input structures) of size |$P \times S$|⁠, where the hat symbol |$\left( ^{\wedge} \right)$| identifies the diagonalisation of the total economic outputs.

Lenzen and Rueda-Cantuche (2012) provided a clear explanation of how to derive (equation (4)) the industry-by-industry |$\left( {S \times S} \right)$| input–output Leontief inverse matrix constructed from equation (3) on the basis of the fixed product sales structure,

with |$I$| being the |$S \times S$| identity matrix and |$DB$| (i.e. the |$A$| matrix in conventional IO analysis) the industry-by-industry |$\left( {S \times S} \right)$| technical coefficient matrix .

The Leontief inverse matrix represents the cumulative intermediate consumption of inputs generated by a unit of output. It is used in conjunction with satellite data accounts of environmental-extended MRIO database to compute direct and indirect (i.e. upstream) impacts associated with a particular final demand |$y^{\prime}$|⁠. Let |$Q$| (Figure 1) be the satellite block of size |$k \times S$|⁠, whose elements describe the amount of stressor |$k$| (e.g. GHG emissions and employment) required to produce the output of industry |$i$| in region |$r$| (Södersten and Lenzen, 2020). Denoting the direct impact multiplier matrix |$Q^{\prime}= Q{\hat g^{ - 1}}$| (with |${\hat g^{ - 1}}$|⁠, the |$S \times S$| matrix resulting from diagonalizing the inverse of |$g$|⁠), whose elements describe the amount of direct (i.e. on-site) stressor required to produce one unit of output of industry |$i$| (Södersten and Lenzen, 2020) in region |$r$|⁠, the total impact multiplier matrix |$\Lambda $| of a certain final demand |$y^{\prime}$| is obtained by

where |$y^{\prime}$| is a |$S \times 1$| vector with a non-zero element limited to the sector for which a final demand shock is simulated. For instance, if |$Q{^{\prime}_k}$| is the amount of GHG emission per unit of production in recycling sewage sludge, the element |${\Lambda _k}$| describes the total amount of GHG emitted (i.e. carbon footprint) by all industries as a result of delivering a unit of recycled sewage sludge to final consumption.

The estimated total multiplier matrix |$\Lambda $| can be further decomposed via production layer decomposition (PLD) (Lenzen et al., 2018). Noting that the series expansion of the Leontief matrix can be written as |${L_{I,ii}} = \left( {I + DB + \left( {DB} \right)\left( {DB} \right) + \cdots } \right)$| (Lenzen and Rueda-Cantuche, 2012), the basic IO equation becomes

where # denotes element-wise multiplication. The term |$D{B^n}$|captures the contributions from supply chains of nth order, and the sum of all these contributions is called the nth production layer (Lenzen et al., 2018). Thus, the term |$Q^{\prime}{\rm{\# }}y^{\prime}$| represents the direct impacts on production for a final demand shock |$y^{\prime}$|⁠, the first-order term |${\rm{Q^{\prime}\# }}\left( {{\rm{DB}}} \right)y^{\prime}$| refers to the impacts on direct suppliers and |${\rm{Q^{\prime}\# }}{\left( {{\rm{DB}}} \right)^2}y^{\prime}$| indicates the impacts on the suppliers of direct suppliers and so on.

2.2. Hybridisation and empirical application

MRIO sectors are generally aggregated and not sufficiently detailed (Lenzen, 2000; Suh et al., 2004) for the purpose of our analysis. To address this limitation, we augmented the MRIO table with new columns and rows containing details on the inputs and outputs of the production studied (Malik et al., 2015), whose data were collected from an Italian empirical case study. This hybridisation procedure provides both specificity and completeness to the sustainability impact analysis; process-based data provide a detailed representation of the production (i.e. the nutrient recovery strategy from sewage sludge), while MRIO analysis captures the total impacts and eliminates truncation errors (Crawford et al., 2018; Malik et al., 2019).

Process-based data (Table 1) were collected through a case study conducted in Pavia province, Northern Italy, in 2020. Pavia is the leading Italian and European producer of rice, with about 80,000 hectares cultivated. The area has intensive farming systems and a low livestock farming density. Over the years, this has created a great demand for organic fertilisers to restore soil fertility and correct pH variations as a result of mineral fertilisation and submersion for rice cultivation. Site selection was based on purposive and convenience criteria on (i) the representativeness of a complete, qualified and actual bio-based ecosystem; (ii) accessibility of processing plant data and (iii) willingness of the companies to participate in the study. Data collection was conducted at both the sectoral and product levels. Several sources of information, including semi-structured interviews, questionnaires and legal reports, were triangulated to gather evidence.

The biorefinery collects about 165,000 tons of sewage sludge from 45 WWTPs annually, serving 226 municipalities. Each WWTP holds a tender, and the company that offers the best techno-economic conditions (i.e. lowest price) is awarded the contract for sludge collection. The large availability of biomass allows many players to enter the sector. Organic fertilisers produced from sewage sludge, on the other hand, currently have no economic value, which is hampered by consumer scepticism about the safety of their use for human health and the environment. Therefore, the company charges money only for the waste collection service (on average, 100 euros per ton of sewage sludge), which is counted in this framework as final consumption.

By-product withdrawal is handled by land transportation. Before being collected, sludge is analysed for its chemical profile. The biomass is treated as a resource rather than waste and is conditioned by sulphuric acid, lime and gypsum to produce 160,000 tons of organic fertilisers and biosolids. On average, the dry matter content is 28.5 per cent. The chemical characteristics per dry matter include 1.6 per cent total nitrogen, 1.5 per cent total phosphorus, 0.4 per cent potassium and 24 per cent organic C content. Only after an evaluation of the chemical profile conformability are the organic fertilisers distributed free of charge to local farmers. The biorefinery serves a total of 10,000 hectares, mainly for rice or corn production. In addition to the distribution of organic fertilisers, consulting services are provided for ploughing and fertilisation. Next, a third chemical control is carried out on the soil. The collected data were reported in the Italian region of Exiobase with proper codes, identified by correspondence tables using the Harmonised System and the Statistical Classification of Products by Activity.

For comparative purposes, we compared the sustainability impacts of the nutrient recovery strategy with landfilling. To this regard, we used the ‘Landfill of waste: Food’ sector of Exiobase of the Italian region, with related satellite data accounts. Although generic and aggregated, the sector is also representative of the landfilling of sewage sludge because in Italy, food waste and sewage sludge are often stored and landfilled together. Finally, as the two waste management strategies are perfect substitutes, we considered the collection and landfill service offered at the same price (100 euros per ton of sewage sludge) as the service supplied by the biorefinery. However, we recognise that there are several sources of uncertainty in the analysis that needs to be addressed (Section 2.4).

2.3. Economic, social and environmental indicators

As sustainability assessment, we adopt a Triple Bottom Line (TBL) (profits, people and planet) approach. Developed in the field of accounting and corporate responsibility, the TBL considers the multiple dimensions of sustainability and integrates the assessment of business operations, environmental concerns and social responsibilities of suppliers (Ellis et al., 2022; Sarkis and Dhavale, 2015; Norman and MacDonald, 2004).

Based on the TBL framework and most common approaches in the literature, we select four indicators from the available satellite data accounts in Exiobase. We use the sum of intermediate uses of goods and services by all industries in the economy (i.e. cumulative intermediate expenditure) (Wiedmann, Lenzen and Barrett, 2009) as an indicator of the industry impact on economic growth (also referred to as the economic stimulus indicator); it accounts for the total purchase of inputs by an industry to carry out its operations, which creates production opportunities upstream in the supply chain and therefore stimulates indirect turnover. In assessing social impacts, we select total employment, expressed as Full Time Equivalent (FTE), that is, the number of full-time workers employed for 40 h a week. Finally, we assess the GHG emissions and energy consumption as environmental indicators related to climate changes and the contemporary energy crisis. GHG emissions are expressed in tons of carbon dioxide equivalent (t CO₂-eq) calculated through the widely used 100-year global warming potential (GWP100), which accounts for the different GWP of CO₂, methane (CH₄) and nitrous oxide (N₂O) (Pachauri and Reisinger, 2007). Energy consumption is expressed in tera joule (TJ) of total energy carrier use, thus including electricity and heat as well as solid, liquid and gaseous fuels.

In our case study, the nutrient recovery strategy requires a direct expenditure of 10.42 million euros per year, 84 employees (FTE) and 5.48 TJ of energy carriers. No information was available regarding GHG emissions. Therefore, the GHG emissions reported by Murray, Horvath and Nelson (2008) for lime stabilisation (550 kg of CO₂ per dry ton of sludge at 20 per cent dry matter content) were used as a reference.

2.4. Uncertainty analysis

To address part of the uncertainty surrounding our assumptions, we determine the stochastic variation of the total impacts using a Monte Carlo simulation (Lenzen, Wood and Wiedmann, 2010). Uncertainty is propagated using standard deviations |${\sigma _Q},{\sigma _T},$|and |${\sigma _Y}$| for perturbating the basic data items |$Q,T,$| and |$Y$| (Figure 1), with the perturbated footprints |$\Lambda $| calculated from 10,000 simulations (Lenzen et al., 2018). The dispersion (or uncertainty) of estimated footprint measures is then derived from the statistical distribution of the perturbations.

More specifically, we approximated the logarithmic absolute error of |$\log x$| as

where |${r_x}$| is the relative standard deviation (RSD) of |$x$|⁠. The log-normality assumption ensures that the Monte Carlo perturbations do not extend towards negative values that pose problems for IO analysis. Hence, the perturbed entries of the MRIO coefficient can be computed as |${Q^P} = {10^{{{\log }_{10}}Q + \nu {\sigma _{{{\log }_{10}}Q}}}}$|⁠,|${T^P} = {10^{{{\log }_{10}}T + \nu {\sigma _{{{\log }_{10}}T}}}}$| and |${Y^P} = {10^{{{\log }_{10}}Y + \nu {\sigma _{{{\log }_{10}}Y}}}}$|⁠, where |$\nu $| denotes a vector of normally distributed random numbers |$\nu \in N\left( {0\left| 1 \right.} \right)$|⁠. The perturbations are different for each element in the matrices. The logarithmic perturbations can be computed as per equation (7). The perturbated |${x^P}$|is obtained by summing |${T^P}$|and |${Y^P}$|to maintain the balance in the IO table (Wei et al., 2021). In this exercise, we assumed that the satellite accounts (matrix |$Q$|⁠) exhibit an RSD of 30 per cent, whereas the final demand (⁠|$Y$|⁠) and transaction matrices (⁠|$T$|⁠) exhibit a relatively low RSD of 10 per cent to avoid over-perturbations of gross output (⁠|$x$|⁠), which is known with a relatively high degree of confidence (Lenzen, Wood and Wiedmann, 2010).

3. Results

3.1. Direct and indirect impacts

Total impacts include impacts directly related to the Italian industrial production and indirect externalities triggered along the global upstream supply chains. Table 2 shows the results, comparing the sustainability of the nutrient recovery strategy with landfilling for a million-euro (MEUR) demand shock |$\left( {y^{\prime}} \right)$|⁠, which corresponds to 10,000 tons of processed (recycled or landfilled) sewage sludge (recalling that both waste collection services cost 100 euros per ton of sewage sludge).

The direct impacts |$\left( {Q^{\prime}} \right)$| of nutrient recovery express the intensity of the satellite accounts collected from the case study divided by the total industrial output (e.g. 84 employees per 16.5 million euros of industrial output). Direct landfill impacts, on the other hand, are those reported by the MRIO (Exiobase) database for the waste landfill sector of Italy. The total impacts multipliers |$\left( \Lambda \right)$| characterise the direct and indirect impacts embodied in a unit of final demand and are derived with the IO analysis (equation 5). Overall, accounting for the upstream spillover effects, using sewage sludge for organic fertiliser production generates more jobs and reduces more GHG emissions than landfilling. By contrast, landfilling stimulates the economy more and reduces energy carrier use more. Comparing |$Q^{\prime}$| with |$\Lambda $|⁠, we observe that energy consumption is mainly indirect (on average, 96.6 per cent |$\left[ {1 - \frac{1}{2}\left( {\frac{{0.33}}{{10.64}} + \frac{{0.24}}{{6.72}}} \right)} \right]$| of the total impact), as are the social (70.5 per cent) and economic (52.7 per cent) impacts. By contrast, GHG emissions are mainly direct (87.6 per cent).

3.2. PLD

Several upstream suppliers are required to provide inputs that are ultimately necessary for the waste management service that WWTPs purchase. The PLD analysis unravels the contributions to the footprints of different sectors disaggregating the total impacts by upstream production layers |$\left( \Lambda \right)$|⁠. Each production layer signifies the sum of all contributions from supply chains of nth order (Section 2.1). We provided two sets of PLDs (Figures 2 and 3) for both nutrient recovery and landfilling strategies (for the same demand shock |$y^{\prime}$| of one million), illustrating the different sectors involved in the cascading effects and footprints associated with the two sewage sludge management strategies.

3.2.1. Social and economic footprints

Figure 2 illustrates the socioeconomic requirements of the two sewage sludge management alternatives. Layer 1 illustrates the on-site impacts |$\left( {Q^{\prime}} \right)$| occurring at the production plants reported in Table 2 (e.g. 0.63 million of cumulative intermediate expenditure and 5.09 FTEs promoted by one million services purchased for recycling 10,000 tons of sewage sludge). Layer 2 includes all contributions to footprints from direct suppliers (e.g. producers of lime, sulphuric acid and gypsum (Table 1) in the nutrient recovery strategy). Layer 3 includes the suppliers of suppliers (e.g. suppliers of the energy needed in the production of lime purchased by the waste management industry). After the fifth production layer, the graphs tend to converge to the total impact |$\left( \Lambda \right)$| reported in Table 2, as the contributions from additional suppliers of suppliers become marginal. Six production layers are sufficient to account for more than 96 per cent of the socioeconomic impacts. Excluding footprint contributions after the second production layer would cause a truncation error of 26 per cent in economic terms and 32 per cent in social terms.

The nutrient recovery strategy needs to source many inputs from the transportation, research and development (R&D), manufacturing and chemical sectors in order to support capital and operating costs. Further upstream, direct suppliers need inputs, with business services (including wholesale and retail trade), electricity, fuels and mining products as the major commodities. The substantial differences between the nutrient recovery and landfilling strategies mainly concern R&D and mining, which are socioeconomically relevant in the nutrient recovery strategy but not in landfilling, in which other business activities (including commission trade and renting of machinery and equipment) play the most important role.

Regarding the geographical distribution of footprints, most inputs are purchased domestically, and the main trading partners are the same for both strategies. The nutrient recovery strategy accounts for 94 per cent of the cumulative expenditure and 87 per cent of jobs at the national level, +4 per cent and +13 per cent, respectively, compared with landfilling. Germany is the most economically involved country, given its high trade of industrial machinery, equipment and motor vehicles with Italy, followed by China and the USA for chemical imports. At the social level, jobs abroad are mainly related to mining (in particular, the extraction of crude petroleum) in the Asia and Pacific area.

3.2.2. Environmental and energy footprints

Figure 3 illustrates the indirect GHG emissions and energy consumption footprints. As much as 95 per cent of the environmental impacts are domestic in origin (Table 2). For illustration purposes, we represented only indirect emissions (Figure 3), so production layer 1 starts from the origin. For completeness, 1,100 and 3,678.12 t CO₂-eq. (⁠|$Q^{\prime}$| in Table 2) should be recalled in Figures 2a and 3a, respectively. Including direct impacts, six production layers account for 98 per cent of the total GHG emissions. Excluding impacts beyond direct suppliers (i.e. production layers beyond the second), the underestimation of footprints is 10 per cent for the nutrient recovery strategy and 4 per cent for landfilling, which are lower than those for other footprints because GHG emissions occur primarily on-site.

The main sectors responsible for indirect GHG emissions in waste management strategies are mining, energy, transportation and other services. Mining includes the externalities associated with natural gas and crude oil extraction in Russia, Asia and the Pacific. In addition to these, landfilling generates the intermediate demand for several emission-intensive services for the disposal of other wastes (e.g. paper, wood and textiles), some of which are imported (e.g. landfilling from Portugal).

In terms of energy carriers (including electricity, heat and solid, liquid and gaseous fuels), the nutrient recovery strategy involves greater energy demand than landfilling. During sewage sludge recycling for nutrient recovery, energy is required to treat and stabilise the biomass with chemicals (sulphuric acid and lime) and minerals (gypsum). In our case study, the biorefinery relies mainly on renewable energy (photovoltaics), which accounts for 56 per cent of the total energy consumption. On the contrary, the total energy consumption of landfilling comes mainly (30 per cent) from petroleum refineries. The rest come precisely from the chemical and mining industries, whose energy costs usually account for most of the gross production costs. In this case, if the impact assessment had stopped at the direct supplier level, it would have underestimated the total energy consumption by 30 per cent for the nutrient recovery strategy and 59 per cent for landfilling. Finally, in line with the profile of responsibility for GHG emissions, part of the total energy consumption comes from abroad (16 per cent for nutrient recovery and 37.6 per cent for landfilling), particularly natural gas imported from Russia and Middle Eastern countries and nuclear electricity from France.

3.3. Uncertainty analysis

The uncertainty of total impacts was determined by uncertainties in basic data, which refer to uncertainties of the data sources and to the assumptions of the IO analysis (e.g. fixed proportions or constant prices). To estimate the errors associated with multipliers |$\left( \Lambda \right)$|⁠, we run 10,000 Monte Carlo simulations, including parametrical uncertainty of the entire MRIO database, process-based data and satellite data accounts assuming log-normally distributed errors in the basic statistical matrices |$Q,T,$| and |$Y$|⁠.

Figure 4 shows the frequency distributions of the perturbated values of the results of the hybrid MRIO analysis using Monte Carlo simulations. We found that the multipliers of nutrient recovery are certain at the 95.5 per cent level of confidence: (i) between 1.00 and 1.73 million euros of cumulative expenditure, (ii) between 16.84 and 28.92 FTEs, (iii) between 885.55 and 1,952.63 t CO₂-eq. and (iv) between 8.16 and 14.84 TJ.

Table 3 compares the frequency distributions (normalised by the baseline results) of the waste management strategies using the interquartile range (IQR) and coefficient of quartile variation (CQV). Except for GHG emissions, the same perturbations of the data sources generated more uncertain multipliers |$\left( \Lambda \right)$| for the nutrient recovery strategy than for landfilling. This can be observed by comparing the IQR or CQV values between waste management strategies (Table 3). These results can be explained as follows. Considering that the multiplier computation involves numerous additions of elements in |$T$|⁠, the errors in the source data cancel out because of their stochastic nature. Emissions were less affected by this condition, being mainly on-site (Table 2). The same applies to the nutrient recovery strategy, which has a production function (Table 1) involving fewer sectors than the general landfilling sector of Exiobase.

Based on their variance and median, the distributions of an indicator may overlap with one another and contradict the baseline comparison made in Table 2. Despite the higher uncertainty, the distributions of GHG emissions overlap less than the other indicators do, and only in 0.05 per cent of cases does the nutrient recovery strategy generate more GHG emissions than landfilling does. Similarly, the probability that this type of sewage sludge recycling consumes less energy in total than landfilling is also low (1.19 per cent). For socioeconomic indicators, on the other hand, overlaps are likelier, with a 24.05 per cent probability that replacing landfilling with sewage sludge recycling for nutrient recovery would possibly stimulate the economy more and with 37.62 per cent generating fewer jobs than before.

4. Discussion

The literature shows that experts’ discretion regarding which activities and products to include in the total impact assessment often leads to truncation errors (Crawford et al., 2018). Applying a hybrid MRIO analysis, instead, we have handled the global upstream supply chains involved in the sustainability footprints of an Italian nutrient recovery strategy from sewage sludge. The results indicate that the sum of the impact contributions from supply chains of the sixth order (i.e. six production layers) accounts for more than 95 per cent of the total footprints. In addition, studying the indirect impacts of the waste management strategies turned out to be crucial to assessing their sustainability.

Considering indirect impacts can overturn conclusions drawn from a superficial first glance. For example, considering only the on-site (direct) social impacts, recycling sewage sludge for agricultural purposes appears to be less labour intensive than landfilling, potentially reducing jobs by 190 for one million ton of sewage sludge recycled rather than landfilled. However, when the analysis also considers the indirect social impacts associated with the final demand for the sludge management service (i.e. the FTEs promoted upstream to meet the demand for inputs needed in order to recycle sewage sludge for agricultural purposes), the opposite is true, and for every ton of sludge not landfilled but recycled, 112 jobs are generated, especially for the R&D sector. Indeed, R&D plays a crucial role along the entire supply chain, promoting production safety and plant efficiency. Technological innovation, patents and investments influence the competitiveness of waste management strategies. In this study, R&D investments also represent a cost-saving opportunity for agricultural producers, as technological progress is directed towards the provision of no-cost services related to precision agriculture (e.g. detailed information maps of soil conditions).

Overall, sewage sludge management can have relevant environmental and economic impacts. In terms of GHG emissions, sewage sludge recycling shows the best performance compared to landfilling, supporting evidence from previous observations (Liu et al., 2013). The poorer performance of landfilling depends mainly on the large amount of methane leakage, which Liu et al. (2013) estimated to be 76 per cent of the total emissions in sewage sludge landfilling, while 42 per cent in the fertiliser production scenario. This also implies that GHG emissions from landfilling are unlikely to be reduced in the future, for example, through renewable energy. In 2020, the amount of sewage sludge disposed of in landfills in Italy was more than 1.6 million tons (53.3 per cent of the total amount). According to Table 2, in the extreme scenario in which the entire amount disposed of in landfills is directed to the nutrient recovery strategy (i.e. 160 million of demand shock), there is a potential reduction in total GHG emissions of 0.44 million tons CO₂-eq. (0.277 t CO₂-eq. per ton of sewage sludge recycled instead of landfilled). However, this emission reduction comes at a cost. Recycling sewage sludge otherwise landfilled and producing organic fertiliser can reduce the (intermediate) use of goods and services by all industries in the economy (i.e. cumulative intermediate expenditure) by 35.2 million (22 euros per ton of sludge). Hence, these findings suggest that the nutrient recovery as a strategy to reduce GHG emissions would cost the society (in terms of cumulative expenditure) 79.4 (22/0.277) euros per ton of CO₂-eq. Moreover, if all sewage sludges destined for landfills were converted to organic fertiliser, Italy would experience an increase in energy consumption of 627.2 TJ (0.392 GJ per ton of sewage sludge). The reduction in GHG emissions takes this increase into account. However, special attention must be given to the energy resources and sources used by the country so as not to compromise the resulting environmental benefits. Nonetheless, as the uncertainty analysis shows, GHG emission reduction was the most certain (at a 99.95 per cent confidence level) among the estimated impacts.

This study suggests how the circular management of sewage sludge for agricultural purposes can be a valuable option in contributing significantly to the goals of the European Green Deal. Given the reduced GHG emissions compared to landfilling, the nutrient recovery strategy can contribute positively to the EU’s 2030 GHG emission reduction target of at least 55 per cent below 1990 levels. In addition to reducing emissions, efficient nutrient recovery from sewage sludge has the potential to reduce mineral fertilisers consumptions (including imports) as aimed by the Farm to Fork Strategy and strengthen agricultural production processes, making them more efficient and resilient, with high yields and high profit margins (Dimitriou and Rosenqvist, 2011). From a social point of view, together with the reorientation of employment towards cleaner production, the creation of more job opportunities in the nutrient recovery strategy than in landfilling can also generate further job growth. To ensure that nutrient recycling is sustainable, renewable energy sources play an essential role; energy efficiency must be prioritised to suggest possible abatement actions. Finally, in economic terms, landfilling performs better by stimulating the whole economy with more indirect turnover; however, a more detailed analysis is needed (Section 4.1) to identify other indirect impacts not captured by the estimated total multipliers.

Policymakers need to ensure more consistent and up-to-date regulatory frameworks. In particular, the European Directive 278/1986 should consider the technological and scientific developments achieved in recent years to prevent harmful effects on soils, vegetation, animals and humans when regulating the use of sewage sludge in agriculture. In addition, while various regulations consistently commit companies to numerous stringent safety standards, they are not adequately brought to the public’s attention. Achieving an efficient and circular transition requires consumer awareness (Qi and Roe, 2017) of the responsibilities and protections associated with sludge management, which ensure food safety and the absence of environmental damage if properly respected. Several countries, such as the UK, the USA, Australia and New Zealand, have developed specific assurance schemes for fertilisers made from sewage sludge. This scheme reduces barriers to use by formalising the products, establishing standardised production procedures, limiting contaminant content and testing protocols (Moya, Parker and Sakrabani, 2019).

4.1. Limitations and further research

The generalisability of our results is subject to certain limitations. One source of weakness is the focus on a single case study and country. In fact, the results correspond to a specific nutrient valorisation strategy involving sewage sludge in Italy, where the biorefinery sources the primary inputs. If modelling was done in a different country, the direct impacts might be similar, but the total impacts would change according to the specific sourcing of inputs of the country’s sectors. Furthermore, the MRIO analysis requires several assumptions, such as a fixed production structure, constant returns to scale and fixed commodity prices that limit the ability to cope with modern economic systems. Therefore, our analysis has mainly emphasised short-term effects. Further research is necessary to evaluate the long-term effects of recycling and processing sewage sludge for agricultural purposes. This research should include model simulations to determine the optimal rates of sludge recycling over time (Chen, Ruijs and Wesseler, 2005) and employing dynamic approaches that consider factors such as the increasing opportunity cost of landfill space, whose capacity is limited, and nutrient availability changes. These aspects were not addressed in the current study. Additionally, it is relevant for a long-run perspective to consider the potential effects of substituting synthetic fertilisers. We currently lack empirical evidence regarding the extent to which chemical fertilisers have been replaced by the organic alternative under investigation. Nonetheless, we estimated the potential overall impacts of using chemical fertiliser for a specific crop (i.e. paddy rice) through secondary sources. The results indicate that the inclusion of these substitution effects does not affect the findings presented and discussed, given their low relative impact on the estimated sustainability footprints (for a detailed report, see Supplementary Information). However, more attention needs to be paid to these substitution effects when studying indicators where mineral fertiliser application has a significant contribution, such as for human toxicity, particulate matter, acidification, eutrophication and freshwater ecotoxicity (Fusi et al., 2017). Next, MRIO databases treat capital expenditures and depreciation of existing capital assets as exogenous components, given the lifespan of fixed assets exceeding the length of the 1-year accounting period used in the national accounts (Södersten and Lenzen, 2020). This implies that upstream impacts associated with the production of capital goods are not included in the estimated impact multipliers; Södersten and Lenzen (2020) have recently suggested a supply-use approach to capital endogenisation in MRIO analysis, which should be referred for future research. Finally, although we attempted to capture stochastic variations of data and calculation procedures, our uncertainty analysis did not address systematic error sources (such as changes in import structures or choice of currency conversion factors).

5. Conclusion

Replacing sewage sludge landfilling with other waste management solutions can result in significant impacts on the socioeconomic balance of the existing economic system, as well as on the environment. Several studies have demonstrated the agronomic benefits of applying treated sewage sludge on soil properties and crop yields, but to the best of our knowledge, none has consistently focused on the overall economic, social and environmental implications in a single framework. Accordingly, the present study was designed to consistently determine the footprints for all sustainability dimensions of agricultural sewage sludge treatment and to compare them with those of landfilling.

Using a hybrid MRIO analysis, we evaluated direct and indirect footprints. Overall, the results show that recycling sewage sludge for agricultural purposes promotes job creation and reduces GHG emissions more than landfilling does; the latter, by contrast, stimulates greater economic growth and lower energy consumption. Indirect impacts play a key role in these sustainability performances, which are characterised by different levels of uncertainty. We accounted for uncertainties in the results using an error propagation method based on Monte Carlo simulations. The GHG emissions per final demand of sludge treatment were the most uncertain of the estimated footprints. However, even when the uncertainty of the estimate is considered, the production of organic fertilisers from sewage sludge otherwise destined for landfills has a high probability of reducing GHG emissions in the country where production takes place. There is also likely to be a significant increase in energy consumption, which we should pay special attention to, so as not to compromise the resulting environmental benefits.

Acknowledgement

The authors would like to thank the reviewers for their feedback on earlier drafts of the paper.

Funding

This work was supported by the H2020 BioMonitor project [grant agreement No. 773297].

Supplementary data

Supplementary data are available at ERAE online.

Conflicts of interest

The authors have no conflicts of interest to declare.

Footnotes

References