Safe Aggregation in Modular Sustainability Assessment: A Method for Interface-Aware Pareto Screening

In the world of industrial ecology, scientists and engineers try to understand the full environmental cost of the things we use, from the electricity that powers our homes to the materials that build our cities. They do this by breaking complex systems down into smaller, manageable pieces called modules. Imagine a vast network of factories, farms, and recycling centers, each performing a specific task like refining oil or generating power. To study the whole system, researchers build digital models of these individual tasks and then connect them together, tracing how materials and energy flow from one step to the next. This approach, known as life cycle assessment, allows them to calculate the total environmental burden, such as carbon emissions or water use, associated with a product. However, a persistent challenge arises when these models become too large or too complex. To make them manageable, modelers often simplify the details of intermediate steps, grouping different types of materials or conditions into a single, generic category. For instance, they might treat all electricity as the same, regardless of whether it was generated at noon or midnight, or they might ignore the specific pressure of a gas stream, assuming that only the total amount of gas matters. The critical question is whether these simplifications are safe. If a modeler discards a detail that turns out to be important for the next step in the chain, the final calculation of the system's total impact could be completely wrong, potentially leading to the selection of a technology that is actually worse for the environment than the one they rejected.

Juan J. Segura, a researcher at Universidad Andres Bello, has developed a precise method to answer this question, determining exactly when it is safe to simplify a model and when it is dangerous to do so. His work focuses on the interface between different parts of a system, which is the point where one process hands off its output to the next. He proposes a rule that checks whether the downstream process, the one receiving the material, actually cares about the specific details being removed. If the receiving process is insensitive to a particular detail, such as the temperature of a gas or the specific location where it was produced, then that detail can be safely aggregated or ignored without changing the final environmental conclusion. However, if the receiving process is sensitive to that detail, the information must be kept. Segura demonstrates that ignoring this rule can lead to a specific and serious error: a model might incorrectly identify a technology as the best option and discard a superior alternative simply because the model failed to see a crucial difference between them.

To prove his point, Segura constructed a specific test case involving five different technologies that all produce the same basic material. In this scenario, the technologies produced materials with different combinations of purity, location, and temperature. The downstream process depended critically on the purity and location of the material, but by construction, it was insensitive to the temperature. When the modelers used a traditional approach, they looked only at the total amount of material produced and the energy used to make it, ignoring the differences in purity and location. Under this simplified view, one of the technologies appeared to be the most efficient and was selected as the winner, while a superior alternative was discarded. However, when the full system was assembled and the specific requirements for purity and location were taken into account, the situation reversed. The technology that had been discarded turned out to be the only option that could satisfy the system's needs without incurring massive extra burdens downstream. The technology that looked good in isolation actually required so much extra effort downstream to meet the specific purity and location requirements that it became the worst option overall. This example showed that a premature simplification had eliminated the true best solution.

Segura then took this concept further by running a massive computer simulation to see how often this kind of mistake happens in the real world. He generated over 160,000 different modular systems, varying the number of technologies, the complexity of the environmental impacts, and how sensitive the downstream processes were to specific details. The results were stark. As the sensitivity of the downstream process increased, the risk of making a wrong decision by simplifying the model grew dramatically. In cases where the downstream process was highly sensitive to details, the traditional method of simplifying the model failed to find the best solution in nearly 74 percent of the simulations. It consistently removed the very technologies that would have been the best choice for the complete system. In contrast, the method Segura proposed, which he calls interface-aware screening, never made this mistake. By checking whether the downstream process could actually distinguish between the different options before making a decision, this method preserved the correct solution every time, even while still allowing for the removal of truly irrelevant details.

The core finding of this research is that the level of detail required in a model is not a fixed standard that applies to every situation. Instead, it depends entirely on what the next step in the chain needs to know. A model does not need to be detailed everywhere, but it must be detailed enough where it matters. If a downstream process cannot tell the difference between two states, then those states can be treated as the same. But if the downstream process reacts differently to them, the model must keep them separate. This insight provides a clear, auditable rule for scientists and engineers building these complex sustainability models. It allows them to simplify their work without fear, provided they can prove that the simplification does not hide a difference that the rest of the system cares about.

This approach changes how we think about building models for the future. It suggests that we can create libraries of reusable process models that are certified to be safe for specific types of connections. Instead of guessing whether a model is detailed enough, modelers can now calculate the exact level of detail required based on the intended uses of that model. This is particularly important for prospective assessments, where scientists try to predict the environmental impact of technologies that do not exist yet or that will operate in very different future contexts. By ensuring that the right details are preserved, this method helps prevent the selection of technologies that look good on paper but fail in practice. It offers a way to navigate the complexity of industrial systems with confidence, ensuring that the path we choose for a sustainable future is based on accurate information rather than convenient shortcuts. The work does not claim to solve every problem in environmental modeling, but it provides a fundamental tool for knowing when a model is ready to be used and when it needs more detail, turning a matter of judgment into a matter of calculation.