Towards the development of an online platform for an industry metabolic pathway

Water being at the centre of a literal nexus of services, research is understandably moving towards methodologies exploring more holistic approaches such integrated water resources management (Savenije & Van der Zaag 2008), or the concept of a dynamic metabolism modelling of urban water services (Behzadian et al. 2014; Venkatesh et al. 2014) or circular economy (CE). Industrial symbiosis (IS) can be defined as engaging ‘traditionally separate industries in a collective approach to competitive advantage involving physical exchange of materials, energy, water, and/or by-products’ (Chertow 2000). A clear emphasis is placed on benefits from collaboration and opportunities for convenient exchanges of useable by-products resulting from spatial proximity. CE relies on ‘decoupling growth from the consumption of infinite resources, reducing waste and pollution, reusing products and materials, and the regeneration of natural systems’ (Ellen Macarthur Foundation 2010). Yazan et al. (2022) reviewed decision support tools for smart transition towards CE and identified eight mainstream techniques used for analysing IS. These are: (1) agent-based modelling (Batten 2009) that simulates interactions between independent decision making entities; (2) material passports (Hansen et al. 2018) that tracks over the life cycle of an object of its circular value, and the uses for its products and components; (3) machine learning and rule-based algorithms (Van Capelleveen 2020) that utilise training data to predict outcomes; (4) environmental assessment and accounting methods from the field of industrial ecology (Daddi et al. 2017) that investigate material and energy flows; (5) game theory (Jato-Espino & Ruiz-Puente 2021) that considers the interaction between players making rational decisions based on their individual goals and interests; (6) geographical information systems (GIS)-based exploration and scoring methods (Van Capelleveen et al. 2018) that focus on region identification tools; (7) material selection methods (Ramalhete et al. 2010) that help designers select the most appropriate material candidate based on its technical properties; and (8) network and infrastructure optimisation techniques (Kastner et al. 2015) that look for ways to connect plants to facilitate and optimise the transportation of by-product exchange.

The IS support framework (Yazan et al. 2022) emphasises the need for modelling, in particular, agent-based models, to investigate the factors that impact on the willingness of individual businesses to co-operate, thereby improving the efficacy and sustainability of the IS solutions. Thus, a modelling solution, whether it is system dynamics, discrete-event simulation or agent-based simulation, is important for decision-makers to explore the factors that contribute to the emergence of IS, those that are important for the potential adopters, etc. However, a challenge is the use of a simulation-based decision support system by the non-experts, which in our example, are the individual business owners. Arguably, the use of an online platform that separates the detailed simulation model (developed by programmers and simulation experts) from its intended end users, but is yet sufficiently versatile to enable novel forms of experimentation by the end users, would allow further enable the adoption of CE concepts among stakeholders. The novel forms of visualisations proposed in this paper go beyond the development of interfaces which only allow the users to change the model parameters. Indeed, and based on the domain of interest, the user interface should facilitate the development of new pathways among different businesses, whereby the waste generated by one entity becomes an input (raw material) for a subsequent entity, thereby forming a chain of input–output processes among various businesses. The input–output linkages may evolve through time. Thus, a platform targeted at end users of IS solutions should allow reconfiguration of the core model logic related to input–outputs; additionally, the platform should enable such model-level reconfiguration to be performed by non-experts through novel visualisations. Towards this, the paper extends our previous work on designing online platforms with highly visual and interactive interfaces for the NEXTGEN serious game (Evans et al. 2023; Khoury et al. 2023) that supported system dynamic models on CE of water (Evans et al. 2023). In this paper, we extend the state-of-the-art in modelling and visualisation by articulating the need for an online decision support tool that allows the stakeholders to make changes to the core model logic using novel visualisations; a tool which allows the non-experts to engage with the experimental modelling and simulation approach through novel interfaces, helping them to explore opportunities to move towards the CE of water through IS and to discover the resultant benefits.

The technical novelty of this work lies in the flexibility and modularity of the hybrid framework that combines a system dynamic modelling (SDM) simulation engine with an interactive online visualisation system. Improving on the design principles and lessons learned from the NEXTGEN project (Evans et al. 2023; Khoury et al. 2023), the ULTIMATE platform presents enhanced adaptive visualisation capabilities with flexible and modular modelling capacities. Four different interactive systems are showcased: a greenhouse located in the Netherlands, a collective olive mills wastewater transport and treatment system from Israel, a juice factory from Greece, and a hybrid abstract model combining all three systems (referred to as the integrative model). The platform is fully functional and available for all to use online (https://ultimate-www.cafloodpro.com/).

This work presents a hybrid modelling framework aiming at exploring a metabolic network of industries (metabolic network as in biology where metabolic reactions produce different compounds reused by other reactions) built around technological examples of CE for water and IS. The hybrid framework integrates methods and techniques from multiple disciplines, namely, systems modelling and computer simulation, and applied computer science. While SDM is used to model the underlying system of interest (which, in our case, is IS), a plethora of techniques from applied computer science have been deployed to develop the online platform and novel forms of visualisations.

The resulting hybrid modelling framework combines multiple instances of SDMs into one seamless simulation engine and then connects it to a novel user interface enabling the flow control, activation, and chaining of these models representing industrial activities with different input/outputs.

The SDMs, used as modelling modular building blocks in our framework have four generic characteristics (Naugle et al. 2024):

• Models are based on causal feedback structure where feedback loops are either positive or negative loops.

• Accumulations and delays are foundational, leading to a differentiation between ‘stocks’ or variables in which something accumulates, and ‘flow variables’ that determine changes to those stocks.

• Models are equation-based and are therefore simulated by calculating the value of each variable at a starting time step, and then by updating the values for all variables at the next time step until the end of simulation.

• The concept of time is theoretically continuous. In practice, it is implemented by segmenting the conceptually continuous time horizon into common discrete time steps (i.e. hourly time steps to simulate daily rainfall events over 20 years).

One of the initial concepts within the ULTIMATE project was demonstrating novel technologies within the water cycle to showcase wastewater is not only reusable but also a resource-rich medium. To present the potential benefits of adopting a CE approach for wastewater management, three case studies were chosen: a juice factory from Greece, a greenhouse located in the Netherlands, and collective wastewater transport from olive mills and treatment system in Israel. For each of the case studies, a real-time simulation engine was developed that was based on surrogate models of technologies and processes relating to water, energy, and material reuse.

The Netherlands case study investigates the potential of electrodialysis (ED), aimed at reclaiming water and assessing the feasibility of nutrient recovery from wastewater discharged by a collective of greenhouses. ED is an innovative electrochemical membrane technology that can be a low-energy-cost alternative for treating wastewater for this application. To simulate this, a demand-driven mass-balance model was developed (Figure 3). To provide a baseline comparison between energy use for specified water recovery rates the electrochemical membrane technology is compared against a reverse osmosis approach for water treatment/recovery.

This model is demand driven whereby water and nutrient demands are determined by the greenhouse characteristics such as crop types, irrigation methods, area, etc. To meet the water demands the original model had access to a combination of water sources: harvested rainfall, surface water, reclaimed treated water and other water. From the nutrient aspect the model only considers nutrients via fertigation whereby these originate from an external additive sources within the mixing tank and via any reclaimed nutrients present within the treated drainage water. While the water reclamation approaches allow for the recovery and reuse of water, some losses still occur within the system relating to evapotranspiration, plant take up, and discharged water, where the quality of discharged water cannot be reused for irrigation within the system without additional treatment.

To simulate the demand within the system, the inputs relating to water and nutrient demand estimates were derived from selected crop types. Ten types of nutrients (nitrates, ammonium, phosphate, potassium, calcium, magnesium, sulphur, chlorides, sodium, and sodium tolerance limit) are tracked through the system. The mmol/L requirements for fertigation water are estimated from literature along with percentage uptake rate estimates that can be customised by the user of the model to infer values of nutrients present in the collected discharge water that is to be treated. Utilising laboratory data, percentage-based recovery parameters for different nutrients are estimated for different water recovery rates and target conductivity/water quality values (Table 2).

The model allows the user to select the desired quality (in terms of conductivity levels) of the treated water for reuse, which subsequently determines the quantity of nutrients being recovered, and energy required to recover this volume of water at this quality.

While the parameters affect the quality of recovered treated water, they also determine the quality of the discharged water that cannot be reused directly within the Netherlands case study. From the individual model perspective this discharged wastewater would need to undergo additional treatment prior to being discharged to a municipal wastewater treatment plant (WWTP).

Within the singular model setup OMWW is mixed with domestic wastewater at a ratio of 0.5 m³ of OMWW to every 95.5 m³ of domestic wastewater treating around 120 m³ per day. This mix is anaerobically treated at the decentralised AAT to reduce its COD value to acceptable levels prior to being discharged into the sewer system to continue to the municipal WWTP. In addition to the COD reduction, the anaerobic treatment of the OMWW and domestic WW produces biogas at a volume ranging from 8 to 15 m³/day.

The three case studies have demonstrated the modelling of different CE applications. These applications could be further linked as a CE system that the outputs of one model could serve as inputs or complementary inputs to the other. An integrative model, that combines the three case studies can be applied to analyse the overarching CE system.

An online decision support tool for a collective (such as a group of greenhouse farmers) requires a responsive interface that must be able to send back results to the user quasi-instantly. This requires an SDM to run in the background and compute the solution of a given problem on demand with minimum latency. Furthermore, some portion of a given model would have to be reused as a component in a different model later, so there is a need for a modular reusable structure. The simulation engine running the SDMs was therefore implemented in the Julia programming language (Bezanson et al. 2012), a recent scripting language with sufficient speed and flexibility to satisfy both requirements. As mentioned in Section 2, out hybrid modelling framework that integrates techniques from SDM and applied computer science (online platform, visualisation) is based on the Model-View-Controller architecture.

Experts in each case study have asked for the ability to visualise the impact of changes directly during a presentation, may it be online or via some in-person demonstration on a common screen. They also underlined the usefulness of being able to access the system from different devices ranging from a laptop to a smartphone so that participants might be able to concurrently explore different sets of operational parameters, compare possible outcomes and discuss pros and cons.

Note that there is also a system/options box that allows the user to mute audio voice comments and switch between the present 2D projection and an isometric projection of the interactive board. There is also a ‘style’ switch that allows to change from a ‘corporate’ visual style suitable for a business-oriented audience to a ‘cartoon’ visual style more relatable to the general public.

Additional results are shown as animated charts to users:

• The average yearly use of rainwater,

• Treated drain water, and other sources (such as tap water),

• The nutrient levels in pre-treatment water as well as water quality before reuse and after discharge,

• Sankey diagrams that illustrate the flows of consumed and reused water and nutrients,

• The resulting quantity of fertilisers used and saved.

OMWW is highly toxic, only a limited amount can be spread on fields per hectare per year (users can set the amount spreadable on fields), and if discarded in water bodies without pretreatment, it can cause severe problems for the aquatic environment and can prevent normal wastewater treatment plants to function properly. Users are therefore being allowed to change parameters related to how concentrated the produced wastewater is through seasonal scheduling, and how it is transported and treated. There is also the possibility of creating a new abstract olive mill with new characteristics such as the distance to the Karmiel AAT plant, the distance to the new decentralised AAT plant, the daily wastewater production, and the olive fields area in order to evaluate changes in results such as transportation costs.

In the results section of the interface, a table visible in Figure 12 shows a cost-benefit analysis related to wastewater transportation, treatment and disposal, as well as associated biogas generation and possible VACs recovery as it can use the same filtration system as the juice factory.

To help with decision support in IS involving models from the three case studies seen above, a simulation engine running a hybrid model is created and is connected to a bespoke user interface, leading to explore the space of solutions via experimentation.

To create an integrative model (i.e., a combined model that captures the processes related to the three case studies) for capturing a metabolic network of industries, the outputs and inputs of respective models and a chain of operation should be considered. For example, with two hypothetical processes in a CE, process A and process B, treated water output from process A could be at a sufficient quality standard for use as input to process B; however, treated water from process B may not be at a sufficient quality standard for use as an input to process A.

Having built a simulation engine running the SDMs, the next step is then to design a bespoke user interface that can support symbiosis decision making where whole businesses can join or leave. We refer to this combined interface as the metabolic network of industries interface, and which is discussed next.

The decision support tool, which is based on the hybrid modelling approach (Section 2), is comprised of a frontend web page (or client) that interacts with a remote server running the SDMs. The simulation engine consists of three models running concurrently in a Docker container, meaning that all the code, libraries, and dependencies are packaged in a self-contained virtualised unit that is easy to run on different systems without needing tailored installation.

The web programming technology is based on JavaScript open source visualisation libraries (such as D3js and MaterialUI) integrated within the React framework, an industry standard for creating reusable components for user interfaces. The modular aspect of the code allows us to easily create new variations of decision systems for future work.

Having learned lessons from the deployment experience of the past NEXTGEN project (https://nextgenwater.eu/), we have decided to avoid running the server through cloud services due to the prohibitive costs for long-term hosting (beyond 2 years), the complexity of the installation process, and the added maintenance work due to compulsory and unwanted ‘upgrades’ to the system. Instead, we have chosen to revert to hosting the platform on an Ubuntu Linux server, in the end, it proves to be an overall much simpler and more cost-effective solution.

Exploration via the user interface shows that the benefits obtained when considering each industry in isolation are somewhat different from the benefits obtained by enabling IS.

When looking at each industry separately:

• The greenhouse case study shows (see Figure 10) substantial benefits in terms of water and energy savings can be shown (saving 70% of the water and 90% of the energy footprint linked to water treatment) while using the decision platform due to the use of ED as opposed to reverse osmosis.

• The juice factory case study shows that VAC recovery technology holds promise when extracting polyphenols from orange wastewater (with an estimated added value of 15 million euros/year for one factory as seen in Figure 8).

• The olive mills case study shows that the cost associated with the transportation and treatment of OMWW can also be significantly reduced when optimising the placement of a decentralised AAT plant (Figure 12).

When connecting all three previous case studies to explore the viability of a nascent ‘metabolic pathway’ of industries, what looks like initially modest benefits points towards potential gains of a greater magnitude. From the point of view of water availability, the decision support tool shows that enabling IS between the juice factory and the greenhouse increases the amount of reused wastewater, and therefore lowers the overall consumption of surface and tap water. IS also shows promise to lower the toxicity of the wastewater discharged to domestic WWTPs (by using partially treated wastewater from another source to dilute olive mills discharges) in the case of olive mills. There are probably many possible similar ways to use partially treated wastewater to lower the toxicity of discharges in other industries and hence avoid potential fines or penalties.

If each industry is a node that can be linked by IS to other nodes or industries (by reusing locally available by-products), greater benefits start to appear when increasing the number of connected nodes.

If industries (or connected nodes) within a catchment are all reusing wastewater discharged by each other, the sum of the volumes of wastewater reused by each node will rise dramatically and easily surpass the total volume of freshwater abstracted from the river. With a growing number of connected nodes, IS would greatly increase efficiency not only in term of water reuse, but also for other by-products such as nutrients, heat, and materials. Metals, for example, have a high potential as reusable by-products due to their high concentration in domestic wastewater and continuously increasing thermodynamic rarity (the energy required for mining a mineral from the earth core, as well as smelting and refining it). Concentrations of metal in domestic wastewater shown in Appendix Table 4 built from literature (Sewage sludge management in Germany 2013; Westerhoff et al. 2015) can potentially rise from 1 to 4 orders of magnitude when mixing with industrial wastewater. Reusing metals found in runoff water in a metabolic network of industries could become for example a potent way to lower carbon emissions.

The web-based decision co-creation platform shows substantial advantages from a water, energy, and nutrient reuse perspective to making use of ED in the case of a greenhouse. Similarly, the use of VAC recovery technology shows a reduction in overall water footprint due to the capability to reuse some of the partially treated water discharged by the juice factory, and presents a significant opportunity to extract expensive VACs at a reasonable cost. This technology can also be used to greatly reduce the high toxicity of OMWW using AAT in a cost-effective way.

These three separate examples of industrial activities, when linked together in the decision support tool, interconnected by the physical exchange of materials, water, and other by-products – help users uncover the usefulness of what might be the basis for a metabolic network of industries (as in biology where metabolic reactions produce different compounds reused by other reactions).

The potential for IS is not just limited to the exchange of water or wastewater, but also heat, nutrients, metals and all sorts of local by-products. The resource efficiency of such a metabolic network of industries is also very likely to grow significantly with the number of industrial nodes connected.

As future work, we intend to go beyond the scope of the ULTIMATE project and present the co-creation decision platform to major actors in the pharmaceutical and food industries in Europe, as well as farmers, and see if they can be engaged into starting a nascent metabolic network of industries based on water reuse and nutrients, metals, and VAC recovery.

The work outlined in this paper was supported by the ULTIMATE project (indUstry water-utiLiTy symbIosis for a sMarter wATer society), funded by the European Union's Horizon 2020 research and innovation programme (GA 869318) as part of the circular economy call CE-SC5-04-2019.

Grant agreement No. 869318. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.