Gémeos digitais de rios para o conhecimento de processos fluviais, facilitar o planeamento e a gestão do risco e promover a participação cidadã.

Project DT4Rivers is geared towards building Digital Twins of rivers (RDTs) and to make them available for stakeholders through a Digital Twin Environment. Two case studies are envisaged – rivers Sorraia and Jamor – testing the articulation and convergence of multiple sources of data and results of different simulation tools. The objectives of DT4Rivers are O1. to install the physical components (sensing networks, citizen science observatories) and physical-virtual communication protocols of the RDTs; O2. to develop its modelling components, featuring a deterministic numerical simulation layer and data-driven simulation tools, based on Machine Learning tools, trained with weather Radar and locally sensed data; O3. to develop the Digital Twinning Environment i.e., a multi-organization and multi-river digital platform to host specific RDTs and establish communication and interaction with diverse stakeholders. River Digital Twins (RDTs) may be used to improve the knowledge of river's physical processes but also to predict the hydro-morphological outcome of river interventions. They are powerful tools to engage riverine communities, ensuring their wellness and safety. As key challenge, we seek agile flood nowcasting, by using data-driven and data assimilation techniques, drawing from weather Radar images. To meet these general objectives, the specific goals of DT4Rivers are as follows. G1 To install new meteorological and hydrometric networks of Sorraia and Jamor: Entails to physically install two new meteorological stations, at Jamor and Sorraia catchments, and water level probes and water quality sensors (proposed locations in Annex T2). Also to establish the Edge Computing strategy for local data processing and the wireless Internet of Things (IoT) protocol to relay the data. (Pursued in Task 2). G2 To establish the citizen river communities Citizen river communities will be built through workshops and training, involving project team members and agents from the municipalities, abiding the principles of Citizen Science of the European Citizen Science Association and General Data Protection Regulations. (Task 3). G3 To build data-driven models for precipitation and inundation nowcasting It involves testing Generative Adversarial Neural Networks to build data-driven models to nowcast precipitation, trained with existing Coruche Radar data. Radar data is to be transposed to precipitation on-the-ground and water levels in critical river sections trough Convolutional Neural Networks (CNN). Ground-truth from the udometer networks and from hydrometric stations will be used for training. (Tasks 4 and 5). G4 To build deterministic simulation tools A shallow water model will be used for river planning and to improve the accuracy of flood forecasting. We construct a 4D VAR data assimilation algorithm for hydrometric data. The strategy is to use a forward-mode differentiation and iteratively run the hydrodynamic model to calculate the cost function gradient. The strategy will be tested on a small-scale laboratory environment with panoptic data acquired in controlled conditions. (Tasks 5, 6 and 7). G5 To construct River Digital Twin Environments The concept of River Digital Twin Environment (RDTE), a web-based platform to host the RDTs, requires new features for handling real-time precipitation and hydrometric data, to handle citizen contributions, in the form of photos and videos, and a new rendering layer. (Task 8).