How useful are data-driven methods for building-level greywater treatment from a process engineering perspective?

• Evaluating data-driven methods' usefulness from goals, motivations, and added value. • Data-driven methods show high adaptability in treatment optimization and monitoring. • Boundary conditions and interpolation intervals are crucial for models' usefulness. • Online sensor adoption over grab sampling boosts data resolution & model reliability. • Inter/intra-building transferability is key for their practical applications. Data-driven methods are increasingly explored in various fields, demonstrating their potential. Compared with reviewing case-specific applications, this article critically assesses the usefulness of data-driven methods application from a process engineering perspective. Specifically, we focus on building-level greywater treatment, a highly decentralized treatment option. The application of data-driven methods in this field presents both operational constraints (e.g., lack of professional staff for maintenance) and technical challenges (e.g., highly variable influent loadings). Therefore, we review data-driven methods applied to greywater treatment processes suitable for building-level applications to explore their effectiveness. This review includes both regression and classification-based methods applied in lab- and pilot-scale studies across the following treatment processes: (i) filtration, (ii) electrocoagulation, (iii) nature-based solutions, (iv) membrane bioreactors, and (v) adsorption. We further evaluate the practical usefulness of these data-driven methods from a process engineering aspect based on their ability to meet study objectives, their motivations, and the added value they provide to process engineers. For instance, these methods aid in identifying key operational factors for treatment optimization and improving water safety by developing early-warning systems based on data-driven monitoring methods, reducing the need for chemical additives and labor-intensive laboratory analyses. However, practical applications could be hindered by ill-defined model boundaries, insufficient sampling resolution, and poor input selection. Additionally, compared with applying data-driven methods in centralized wastewater treatment plants, emphasizing model transferability (including both intra- and inter-building transferability) is necessary to enhance the scalability and practical applicability of data-driven models at the building level. This review, grounded in process engineering-based evaluation, bridges the gap between research advancements and practical needs in building-level greywater treatment, contributing to the broader application of data-driven methods in water management.