Digital twins (DT) for water resource recovery facilities (WRRF) are different from regular process models. They require 1) a physical plant twin; 2) automatic data exchange with the real plant; 3) possibility to dynamically update models when or if required. Their use has the potential to improve understanding of plant behaviour and unmeasured variables; move towards proactive decision making at the plants when including influent forecasts; improve data quality control when comparing simulation results to measured values; and be used for predictive maintenance. The model used in a DT can be mechanistic (i.e., describing underlying mechanisms/physics), data driven (empirical, based on observed relationships between variables) or a combination of both (hybrid model). Most of the commercially available (mechanistic) wastewater process simulators include the option to use them in (near) real time as digital twins. Fault detection is important for DTs to avoid use of faulty input data. Methods range from dimensional reduction techniques to process models and statistical control charts. Automated methods for gap filling and corrections of sensor values based on laboratory measurements can be used to correct faulty data. Forecasts of influent flow rate and concentration of pollutants can be useful for optimization and “what if”-scenarios. Forecast models can be data driven (e.g., many examples with time series models and artificial neural networks available in the literature) or detailed mechanistic models. Common for most examples is that weather forecasts (temperature and precipitation) are used, and the model accuracy of course depend on the quality of the forecast. Automatic calibration can be used for both data driven/hybrid models (i.e., re-training) and mechanistic models. For mechanistic models, examples in the literature include simple changing of measured influent fractions or settler solids separation efficiency to global optimization of multiple variables over a plant-wide model. Automatic calibration can be done at fixed intervals or based on performance evaluation. Model predictive control (MPC) has been widely studied in simulated settings, with few real examples for WRRFs. For digital twins, the possibility to combine a mechanistic model with influent forecasts and numerical optimisation for, e.g., setpoints over a future time interval to achieve a certain goal is promising. The faster control applications can then be handled using regular PID-controllers. Few examples of implemented digital twins for WRRFs have so far been published in the literature. Here, one example of a digital twin is presented. It includes automatic data transfer, automatic calibration, and forecasts, but is (at the time of writing based on the available literature) only used as an advisory tool and not for direct control. Digital twins of water resource recovery facilities are complex with many different parts and models that work together. They can be used for fault detection, predictions, and optimization/control. This report summarizes some of the components that can be used to build digital twins, which ones to include of course depends on the scope and goals of the specific project. In all cases, the flow of data from collection to use must be well designed to avoid unnecessary interruptions in operation.