D5.1 Initial preparatory and testing report in case studies

D5.1 Initial preparatory and testing report in case studies February 2025 Data driven implementation of hybrid nature-based solutions for preventing and managing diffuse pollution from urban water runoff Ref. Ares(2025)4439475 - 03/06/2025 2 D5.1 Initial preparatory and testing report in case studies D5.1 Initial preparatory and testing report in case studies Work Package WP5 Deliverable lead VCS Author(s) Laura Roed Bonde (VCS) Contributors: All D4RUNOFF partners Contact [email protected] Grant Agreement number 101060638 Start date of the project / Duration 1 September 2022 / 42 months Type of deliverable (R, DEM, DEC, other) R Dissemination level (PU, SEN) PU Project website www.d4runoff.eu R=Document, report; DEM=Demonstrator, pilot, prototype; DEC=website, patent fillings, videos, etc.; OTHER=other PU=Public, SEN=Sensitive, limited under the conditions of the GA Document history Version Date Authors (organisation) 0.1 30.01.2025 Laura Roed Bonde (VCS) 0.2 20.02.2025 Jesús Fernández Águila (ITG), Nicolás Morales (AQU), Begoña Espiña (INL), Anders Risbjerg Johnsen (GEUS), Thomas Karlsson (UCPH), Jorge Rodríguez (UC), Federica Guerrini (MITIGA), Simone Lippi (ACQUE) 0.3 23.02.2025 Laura Roed Bonde (VCS) 1.0 24.02.2025 Anders Risbjerg Johnsen (GEUS), 2.0 30.05.2025 ITG, AQU, INL, UCPH, UC, MITIGA, and VCS have addressed the PO’s comments regarding editing and have provided a more detailed description of the next steps and strategy in Section 2. Furthermore, additional information has been included in Section 4 concerning next steps, testing, cooperation, and stakeholder interaction. 3 D5.1 Initial preparatory and testing report in case studies ACKNOWLEDGEMENTS This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101060638. COPYRIGHT STATEMENT The work described in this document has been conducted within the D4RUNOFF project. This document reflects only the D4RUNOFF Consortium views, and the European Union is not responsible for any use that may be made of the information it contains. This document and its content are the property of the D4RUNOFF Consortium. All rights relevant to this document are determined by the applicable laws. Access to this document does not grant any right or license on the document or its contents. This document or its contents are not to be used or treated in any manner inconsistent with the rights or interests of the D4RUNOFF Consortium or the Partners detriment and are not to be disclosed externally without prior written consent from the D4RUNOFF Partners. Each D4RUNOFF Partner may use this document in conformity with the D4RUNOFF Consortium Grant Agreement provisions. 4 D5.1 Initial preparatory and testing report in case studies Executive Summary This deliverable, D5.1, is produced as the result of the first set of actions implemented within Tasks T5.1 to T5.5. It consolidates the information obtained from preparatory and ongoing testing activities across several key areas: the detection methods developed in WP1, the risk assessment and mapping methodology from WP4, the online sensor system from WP2, the AI-assisted platform also from WP4, and the MCDA framework from WP3. The report details the testing and validation of innovative analytical techniques for characterizing urban runoff pollutants, the refinement of risk assessment models through the integration of new field data, the evaluation of an online monitoring system under real-world conditions, and the deployment of an AI-driven decision support platform. Early findings indicate that the advanced detection methods have significantly improved our understanding of the pollutants present in urban runoff, while initial sensor tests reveal that further refinements in communication and system adjustments are necessary to address the complexities of natural samples. In addition, the risk assessment activities have provided valuable insights for model calibration, and the integration of GIS and real-time data underscores the need for standardized metadata and robust sensor connectivity. Feedback from the various work packages highlights that while WP1’s analytical methods have advanced pollutant characterization, WP2’s sensor system requires enhanced communication protocols, WP3’s MCDA approach must be adapted to better reflect local cost and operational practices, and WP4’s AI-assisted platform would benefit from further data harmonization and stronger cross-WP collaboration. Looking ahead, the next steps include completing the sampling and analysis, iteratively refining the risk assessment models, finalizing the performance of the online monitoring system, transitioning the AI-assisted platform to full deployment across demo sites, and adapting the NBS library and MCDA methodology to produce feasible Nature Based Solution proposals. In parallel, local working groups will be established at each case study to provide continuous feedback, ensuring that all testing and implementation activities align with practical urban runoff management needs. Overall, this report documents key learnings from the initial testing phase and lays the foundation for further improvements and targeted actions in the subsequent project phases. 5 D5.1 Initial preparatory and testing report in case studies Table of Contents 1 INTRODUCTION ............................................................................................................ 8 1.1 Purpose of the document ....................................................................................... 8 1.1.1 Scope of the document ...................................................................................... 8 2 Summary of Data and Testing Activities .................................................................... 9 2.1 Task 5.1: Implementation of Novel Detection Methods........................................... 9 2.1.1 Detection methods ............................................................................................. 9 2.1.2 Sampling ...........................................................................................................10 2.1.3 Status ................................................................................................................10 2.2 Task 5.2: Risk Assessment and Risk Mapping ......................................................11 2.2.1 Summary of the Risk Assessment Methodology (WP4).....................................11 2.2.2 Implementation of the Risk Assessment Module ...............................................12 2.2.3 Next steps .........................................................................................................14 2.3 Task 5.3: Implementation of Online Sensors .........................................................14 2.4 Task 5.4: AI-Assisted Platform Implementation .....................................................16 2.4.1 AI-Assisted platform operation ..........................................................................16 2.4.2 Preparatory actions for platform implementation in case studies .......................18 2.4.3 Testing activities ...............................................................................................35 2.4.4 Conclusions and next steps ..............................................................................38 2.5 Task 5.5: Multi-Criteria Decision Analysis (MCDA) ................................................39 3 Feedback to WP1-WP4 ................................................................................................41 4 Action Plan for Further Testing ..................................................................................43 4.1 Local working groups ............................................................................................46 5 Conclusion ...................................................................................................................46 6 Acronyms.....................................................................................................................47 6 D5.1 Initial preparatory and testing report in case studies List of Figures Figure 1 Comparison between hourly rainfall data from one station (Sensor 5414, courtesy of VCS), rainfall data provided by the Danish Meteorological Institute (DMI), and the total precipitation from the reanalysis model ERA5Land for November 2023; a more thoroug .....13 Figure 2 Land cover for the larger Odense area on the island of Funen, Denmark. Each colour corresponds to a different land surface type. ........................................................................13 Figure 3 Images of the multi-analyte monitoring platform used for metals, microplastics and CECs detection (from deliverable 2.5). .................................................................................14 Figure 4 Scheme of the final multi-analyte monitoring platform, including sample collection, metals detection and detection of plastics and CECs. ..........................................................15 Figure 5 Diagram of the main components integrating the AI-Assisted platform and its operation in the project's demo sites, covering from the input data used to the outputs produced. .............................................................................................................................17 Figure 6 Location of the Danish demo site of the D4RUNOFF project, highlighting the municipal boundary of Odense, which defines the spatial domain in the current phase of the project for its implementation in the AI-Assisted platform. .....................................................................19 Figure 7 Land use within the spatial domain of the Danish demo site defined in the AI-Assisted platform. The water collected by the sanitation infrastructure in the study area mainly originates from residential areas (red zones) and, to a lesser extent, from activities in the tertiary sector (grey zones). ........................................................................................................................20 Figure 8 Spatial distribution of the sewer network in the Danish demo site, highlighting the main components that constitute it and will be integrated into the AI-Assisted platform. ................21 Figure 9 Location of the rainwater basins in the municipality of Odense, along with the proposed locations of the three main NBS selected for implementation and validation within the AI-Assisted platform: (1) a rain garden located in Træøen (Bolbro) and two retention ponds in the Bellinge Fælled area (2) and near the intersection of Niels Bohrs Alle/Ørbækvej (3). .22 Figure 10 Example of external GIS data sources available for integration into the AI-Assisted platform, providing contextual information for the Danish demo site: (a) Terrain elevation (Digital Elevation Model), (b) Hydrogeology, (c) Hydrology, and (d) Water hardness levels. .23 Figure 11 Example of 1-hour interval precipitation data recorded in the vicinity of Odense between January 2024 and February 2025. Rainfall datasets will be integrated into the AIAssisted platform and used for numerical and AI-based model development within the Danish demo site..............................................................................................................................24 Figure 12 Location of the Spanish demo site of the D4RUNOFF project, highlighting the municipal boundary of Cantabria (red dashed line) along with the spatial domain defined in the current phase of the project for its implementation in the AI-Assisted platform. ....................25 Figure 13 Land use within the spatial domain of the Spanish demo site defined in the AIAssisted platform. The water collected by the sanitation infrastructure in the study area comes mainly from urban areas (gray zones). .................................................................................26 Figure 14 Spatial distribution of the combined sewer network in the Spanish demo site, highlighting the main components that constitute it and will be integrated into the AI-Assisted platform. ...............................................................................................................................27 Figure 15 NBS considered in the spatial domain of the Spanish demo site to be defined in the AI-Assisted platform together with the sampling points and location for monitoring CECs. Due to the different scales, the permeable parking lot is represented by a point ..........................28 Figure 16 Example of external GIS data sources available for integration into the AI-Assisted platform, providing contextual information for the Spanish demo site: (a) Hydrography, (b) 7 D5.1 Initial preparatory and testing report in case studies Edaphology, (c) Terrain elevation (Digital Elevation Model), and (d) Soil permeability types. .............................................................................................................................................29 Figure 17 Location of the Automatic Weather Stations in the domain of Santander. Main parameters observed: temperature, wind velocity, rainfall or humidity. .................................30 Figure 18 Example of 1-hour interval precipitation data recorded at Santander weather stations. ................................................................................................................................30 Figure 19 Location of the Italian demo site of the D4RUNOFF project, highlighting the municipal boundary of Pontedera (white dashed line) along with the spatial domain defined in the current phase of the project for its implementation in the AI-Assisted platform. ................................31 Figure 20 Land use within the spatial domain of the Italian demo site defined in the AI-Assisted platform. The water collected by the sanitation infrastructure in the study area comes mainly from urban areas (yellow zones) and industrial areas (blue zones). .....................................32 Figure 21 Spatial distribution of the combined sewer network in the Italian demo site, highlighting the main components that constitute it and will be integrated into the AI-Assisted platform. ...............................................................................................................................33 Figure 22 Example of external GIS data sources available for integration into the AI-Assisted platform, providing contextual information for the Italian demo site: (a) Terrain elevation (Digital Elevation Model), (b) Geology, (c) Hydrology, and (d) Soil impermeability levels. .................34 Figure 23 Example of 15-minute interval precipitation data recorded in Pontedera between 2020 and 2024. This dataset will be integrated into the AI-Assisted platform and used for numerical and AI-based model development within the Italian demo site. ............................35 Figure 24 Screenshot of the tested connection between devices in WP2 and AI-Assisted Platform to acquire observed data ........................................................................................37 Figure 25 Screenshot of some MQTT actions tested during the initial activities to validate the communication between the AI-Assisted Platform and the WP2 sensors .............................38 8 D5.1 Initial preparatory and testing report in case studies 1 INTRODUCTION 1.1 Purpose of the document This report is a key deliverable from the initial actions implemented within Tasks T5.1 to T5.5 of the D4RUNOFF project. Its aim is to consolidate and present the preliminary findings and data gathered from preparatory and ongoing testing activities, thereby establishing a solid foundation for subsequent development and implementation phases. 1.1.1 Scope of the document The document covers the initial preparatory and testing activities conducted under Tasks T5.1 through T5.5. It includes the evaluation of advanced detection methods, risk assessment and mapping approaches, online sensor integration, AI-assisted platform operations, and multicriteria decision analysis. The report is intended to capture both the successes and challenges encountered, along with feedback to refine project methodologies. Structure of the document The report is organized into the following sections: • Introduction: Outlines the purpose, scope, and relevance of the report within the context of WP5 and its integration with WP1–WP4. • Summary of Data and Testing Activities: Details the specific testing activities, methodologies, and data collection efforts across the tasks. • Feedback to WP1–WP4: Compiles insights, recommendations, and suggestions for improvement derived from the initial testing phase. • Action Plan for Further Testing: Describes the planned next steps, timeline, and responsibilities for the upcoming testing and validation phases. • Conclusion: Summarizes key findings and outlines implications for future project phases. Next step of the document Following this report, the next steps involve refining the testing methodologies based on the feedback collected, enhancing data integration across pilot sites, and progressing with further testing and validation activities. The insights gained will guide the development of improved versions of the project’s technical modules and support more targeted actions in the subsequent phases of the D4RUNOFF project. 9 D5.1 Initial preparatory and testing report in case studies 2 Summary of Data and Testing Activities The upcoming testing and implementation phase will focus on deploying the developed tools and methodologies across the three pilot sites - Odense, Santander, and Pontedera. A coordinated cooperation strategy has been established to support this effort, ensuring alignment between technical partners and local stakeholders through dedicated workspaces, structured data integration, and feedback loops. This cross-site collaboration will enable sitespecific adaptation of the AI-Assisted platform and other components, ensuring that local urban runoff challenges are effectively addressed. The outcomes of this cooperation will be documented in upcoming deliverables D5.3 and D5.4, and further validated through stakeholder engagement activities, including the final D4RUNOFF Open Day in Odense. A more detailed description of the implementation strategy and site-level preparations is provided in Section 2.4. 2.1 Task 5.1: Implementation of Novel Detection Methods This section details the testing and validation of the innovative detection methods developed in WP1 for characterizing urban runoff pollutants. Task 5.1 focuses on applying advanced analytical techniques — including liquid chromatography-high resolution mass spectrometry (LC-HRMS) for both target and non-target screening, hydrophilic interaction liquid chromatography-HRMS (HILIC-HRMS) for suspect screening of mobile compounds, and digital droplet PCR for quantifying antibiotic resistance markers — to real-world water samples. Composite samples are collected from case study sites in Santander, Odense, and Pontedera during urban runoff events, enabling the project team to evaluate these methods under operational conditions. The resulting data not only provides essential feedback to refine these detection approaches but also lays the groundwork for subsequent tasks (T5.2–T5.6) within the project. 2.1.1 Detection methods LC-HRMS. A novel LC-HRMS detection method for urban runoff target, suspect, and nontarget screening (NTS) was developed and validated in WP1. The method is described in deliverables 1.1, 1.2, and 1.6. It will be implemented in T5.1 for quantification of known chemicals of emerging concern (CECs) as well as identification and quantitative estimation of unknown runoff pollutants. HILIC-HRMS. In WP1, a HILIC-HRMS method was developed for suspect-screening ofmobile and very mobile compounds, which are otherwise poorly retained in reversed-phase liquid chromatography. As in WP1, the method will cover a large list of suspect compounds, but with special emphasis on the fate of the biocides that were detected in the WP1 inventory samples. Digital droplet PCR. Digital droplet PCR (ddPCR) will be used to quantify absolute levels of selected antibiotic resistance genes, covering the four classes developed in WP1. In addition, ddPCR will measure the 16S rRNA gene to estimate the total bacterial community and calculate the relative incidence of antibiotic resistance. These analyses will be complemented by traditional plate counts of thermo-tolerant E. coli as a marker for faecal contamination. 16 D5.1 Initial preparatory and testing report in case studies 2.4 Task 5.4: AI-Assisted Platform Implementation The objective of task T5.4 of the D4RUNOFF project is the implementation of the AI-Assisted platform, which is being developed within WP4, at the three demo sites of the project: Odense (Denmark), Santander (Spain), and Pontedera (Italy). This implementation aims to test and validate the functionality and operation of the AI-Assisted platform in real-world environments. The initial phase of T5.4 has focused on preparatory activities, primarily analysing the available information and data at each pilot site which will later feed the AI-Assisted platform. Additional activities have been developed to verify the connection between sensors and the platform, as well as ensuring the proper storage of data and information generated within D4RUNOFF. Once the development of the AI-Assisted platform is completed within WP4 (expected between M30-M32), task T5.4 will move into a new phase focused on the actual deployment of the platform at the three demo sites. The following sections of Deliverable D5.1 will describe the preparatory activities carried out so far in task T5.4 for the implementation of the AI-Assisted platform, as well as an overview of preliminary testing and validation activities. 2.4.1 AI-Assisted platform operation The AI-Assisted Urban Runoff Management platform is being developed within WP4 of the D4RUNOFF project as a web-based GIS tool designed to support urban runoff and stormwater management. The platform will integrate data from online sensors, AI-based models, results from numerical models, external databases, and GIS-based tools, enabling advanced analysis and decision-making processes for different types of users: 1) policymakers, 2) technical operators, 3) the scientific community, and 4) civil society and citizens. One of the platform’s key strengths is its ability to unify and process data from various sources, serving as a digital environment that centralizes information and modelling results generated throughout the project. Specifically, it integrates: • The characterization of urban runoff pollutants developed within WP1. • Data collected by novel sensors developed within WP2, including measurements of CECs and other pollutants. • Tools and methodologies from WP3, such as the NBS Library, preliminary design calculations, and GIS-based MCDA (Multi-Criteria Decision Analysis) • Information from the three pilot sites, used for testing and validation within WP5. • Tools for raising citizen awareness about urban runoff issues through Serious Games and increasing their engagement with the project. These tools will be used within WP6 for project dissemination. Figure 5 presents a diagram summarizing the main components and the planned operation of the AI-Assisted platform in the three demo sites of the D4RUNOFF project. Its operation is based on the information and data available at the pilot sites, which serve as inputs to the platform. The platform can receive real-time data from online sensors or external sources, as well as storing the information in a structured and organized manner for further processing and deployment. 17 D5.1 Initial preparatory and testing report in case studies Figure 5 Diagram of the main components integrating the AI-Assisted platform and its operation in the project's demo sites, covering from the input data used to the outputs produced. The operation of the AI-Assisted platform is based on three core components: 1) Data Gathering, Merging and Data Life Cycle Support. This part of the AI-Assisted platform is responsible for gathering and managing data from IoT and non-IoT sensors, external web services, and other relevant sources. Data processing includes cleaning, merging, and structuring the information to ensure its usability. 2) Calculation and Modelling Engine. This component of the platform supports calculations, analyses, and modelling activities within the project based on the available data and information. 3) Decision Making. This part of the AI-Assisted platform integrates different components/modules to facilitate decision making in developing effective urban runoff and stormwater management plans and implementing NBS. The decision-making block was conceptualized in four dimensions: Smart operation functional module, Dynamic risk mapping functional module, Policy and regulation, and Citizen engagement. The AI-Assisted platform has been designed to operate through dedicated workspaces, one for each demo site, where information, available data, and generated results will be stored, calculated and displayed. After the collection, storage, and processing of data and information from each pilot site through the Data Gathering and Calculation Engine modules, the platform will produce results to support decision-making from different perspectives. Below is a list of some of the outputs that will be generated and deployed in each of the three demo sites once the AI-Assisted platform is fully implemented: • Deployment of information on CECs and conventional pollutants (WP1 and WP2) in the demo sites. 18 D5.1 Initial preparatory and testing report in case studies • Deployment of general information on Nature Based Solutions (NBS) and Engineering Drainage Systems (EDS) and recommendations for solutions to implement. • Deployment of information and monitoring data on the performance of NBS implemented in the demo sites. • Recommendations on the feasibility of implementing NBS in the demo sites from a technical perspective. • Information on the impact of policy changes related to the adoption of NBS for urban runoff CEC mitigation. • Evaluation of the effectiveness of NBS implementation in the demo sites through the comparison of simplified hydraulic model results. • Results from AI-based models. • Information on pollution risk areas through risk mapping. • Tools for data visualization and analysis from different sources. • Serious Games inspired by the demo sites for awareness and citizen engagement. • Insights on the impact of climate change on runoff water and its contamination. 2.4.2 Preparatory actions for platform implementation in case studies The implementation of the AI-Assisted platform in the demo sites of the project and its proper functioning depend on the availability of information and data at each site. So far, the work within task T5.4 has focused on preparatory activities, analysing the three demo sites and assessing the available data. The following sections describe the preparatory activities carried out at each demo site. 2.4.2 Odense Case Study (Denmark) 2.4.2.1.1 Spatial domain definition The operation of the AI-Assisted platform in a specific location requires the initial definition of a spatial domain that encompasses the study area, where the various elements of the case study will be positioned (e.g., NBS and drainage systems, sewer network, wastewater treatment plants (WWTP), sensors, sample locations, etc.). As an initial step toward the future implementation of the AI-Assisted platform at the Danish demo site, a spatial domain has been defined, covering the entire municipality of Odense, with an area of approximately 304 km2. Figure 6 shows the location of the Danish demo site, highlighting the municipal boundary of Odense, which serves as the defined spatial domain for its implementation in the AI-Assisted platform. The platform was developed within WP4 to ensure that the spatial domain of study areas can be uploaded via a specific file format (GeoJSON), which includes both geometry and additional information about the site. 19 D5.1 Initial preparatory and testing report in case studies Figure 6 Location of the Danish demo site of the D4RUNOFF project, highlighting the municipal boundary of Odense, which defines the spatial domain in the current phase of the project for its implementation in the AI-Assisted platform. 2.4.2.1.2 Combined sewer network evaluation In the context of urban runoff pollution and water management, which is the focus of the D4RUNOFF project, sewer networks play a key role. For this reason, the AI-Assisted platform has been designed to store geospatial information and the characteristics of the main elements that make up the sanitation infrastructure in the different case studies. As part of the preparatory tasks for task T5.4, detailed analysis has been conducted on the information provided by the partner responsible for the site (VCS) regarding the sanitation infrastructure of the Danish demo site. Odense is Denmark’s third-largest city, with a sanitation infrastructure capable of transporting and treating wastewater generated by a population of approximately 200,000 inhabitants. The sewer network includes both combined sewer systems and separate sewer systems in different areas of the city. Figure 7 presents a land use map of the municipality of Odense. The water collected by the sanitation infrastructure in the Danish pilot site primarily originates from residential areas and, to a lesser extent, from activities in the tertiary sector. 20 D5.1 Initial preparatory and testing report in case studies Figure 7 Land use within the spatial domain of the Danish demo site defined in the AI-Assisted platform. The water collected by the sanitation infrastructure in the study area mainly originates from residential areas (red zones) and, to a lesser extent, from activities in the tertiary sector (grey zones). For the implementation and evaluation of the performance of the sanitation infrastructure within the AI-Assisted platform, five types of elements constitute the main components of the sewer network of Odense: manholes, pipes, pumping stations, spill points, and WWTP. A specific GIS database has been prepared during the preparatory activities of task 5.4 as a starting point for its subsequent upload to the platform. As the sanitary infrastructure of Odense is of considerable size, future phases of task T5.4 will determine whether to upload information for the entire system into the platform or to focus on specific areas where project-specific data is generated within D4RUNOFF, and urban runoff will be evaluated. Figure 8 presents the spatial distribution of the different elements that make up the sewer network in the Danish demo site, as visualized in the prepared GIS database. The sewer system of Odense has been examined to facilitate its seamless incorporation into the AI-Assisted platform upon completion of its development, as outlined in Section 4.3.3 of Deliverable D4.2 within WP4 of the project. Data on the infrastructure can be integrated into the platform either through individual uploads, element by element, or via bulk imports using standardized GeoJSON files that ensure compatibility with the system. This analysis aims to confirm that the platform’s architecture supports the accurate storage and management of the diverse components comprising the sewer network in the Danish demo site. The available data on the sewer network of Odense includes more than 1,500 km of installed pipelines, with most of the pipes (over 80%) having a circular cross-section. The majority of the pipes have a diameter of less than 1 m, although larger-diameter pipes (>1.2 m) are also 21 D5.1 Initial preparatory and testing report in case studies present in smaller quantities. The water is transported to three WWTPs for treatment, requiring 158 pumping stations to lift the water and reach its destination. The three main wastewater treatment plants in Odense are Ejby Mølle Renseanlæg, Nordvest Renseanlæg, and Nordøst Renseanlæg. The first of these has a capacity of 385,000 PE and serves as a model of efficiency and sustainability in wastewater treatment, combining technology, optimization, and resource recovery to generate clean energy and reduce its environmental impact. Figure 8 Spatial distribution of the sewer network in the Danish demo site, highlighting the main components that constitute it and will be integrated into the AI-Assisted platform. 2.4.2.1.3 NBS evaluation The AI-Assisted platform has been designed to store information about NBS implemented at the pilot sites, contributing to urban runoff management alongside conventional infrastructure, and to serve as a monitoring tool for the performance of specific NBS equipped with sensors. Preparatory activities carried out within task T5.4 regarding the installed NBS at the three demo sites indicated that the Danish pilot has the highest number of operational NBS. For example, more than 20 stormwater basins and over 50 rain gardens have been implemented in Odense. However, despite their large number, the vast majority of NBS at the Danish demo site either lack monitoring systems or have only limited sensors for operational control. Within the D4RUNOFF project and the AI-Assisted platform, implementation and validation efforts will focus on proposed NBS options to be considered for the Danish demo site in Odense: a sustainable drainage system in Helsingborggade and two water basins in 22 D5.1 Initial preparatory and testing report in case studies Trykstokken and Risingsvej. Figure 9 shows the locations of the three proposed NBS that will be evaluated within the AI-Assisted platform at the Danish demo site. In future phases of task T5.4, detailed information about the chosen NBS will be integrated into the platform, including their characteristics, operational data, and available measured data for further analysis and evaluation. Figure 9 Location of the rainwater basins in the municipality of Odense, along with the proposed locations of the three main NBS selected for implementation and validation within the AI-Assisted platform: (1) a sustainable drainage system in Helsingborggade and two water basins in Trykstokken (2) and Risingsvej (3). 2.4.2.1.4 GIS-based tools identification The D4RUNOFF platform is being developed as a web-based GIS tool designed to integrate external data sources, enriching the specific project-generated information with broader contextual insights. As part of the preparatory work for task T5.4, multiple data providers relevant to the Danish demo site have been identified, with the potential to be incorporated into the platform through WMS (Web Map Service) or similar integration methods. These external sources offer valuable datasets, including digital terrain models, geological and hydrogeological information, slope analysis, orthophotos, or hydrographic data. The AIAssisted platform is built to support a variety of base maps, allowing customization based on user preferences while displaying spatial distributions of key project data. This approach enhances geospatial analysis and overall understanding of the site. Figure 10 presents examples of maps from external sources available for incorporation into the AI-Assisted platform, offering contextual information on terrain elevation, hydrogeology, hydrology, and water hardness levels. 23 D5.1 Initial preparatory and testing report in case studies Figure 10 Example of external GIS data sources available for integration into the AI-Assisted platform, providing contextual information for the Danish demo site: (a) Terrain elevation (Digital Elevation Model), (b) Hydrogeology, (c) Hydrology, and (d) Water hardness levels. 2.4.2.1.5 Preliminary analysis of available data The operation of the AI-Assisted platform being developed within the D4RUNOFF project relies on data collected from sensors installed at the pilot sites or obtained from external sources. These data play a crucial role in its function as a Decision Support System (DSS) being designed to receive and process data in real-time from both sensors and external providers. As part of the preparatory activities for task T5.4, external monitoring sensors operating at the Danish demo site have been identified as potential data sources for integration into the AIAssisted platform. One of the key external data providers is the Danish Meteorological Institute (DMI), which supplies weather forecasts, climate monitoring, and meteorological, oceanographic, and atmospheric studies. For the Odense area, the DMI provides both historical meteorological data and forecast models relevant to the D4RUNOFF project, including precipitation, temperature, humidity, and other parameters. A particularly critical component for the project and the efficient operation of the AI-Assisted platform is the availability of high-frequency precipitation data, as these measurements are essential for assessing runoff generation caused by rainfall events of varying intensities and their correlation with key pollution indicators. High-quality, highfrequency precipitation data are already available in Odense through the partner responsible for the Danish pilot site. However, additional precipitation and meteorological data can be integrated into the platform via the services provided by the DMI. a) b) d) c) 24 D5.1 Initial preparatory and testing report in case studies In this context, the DMI offers precipitation data recorded at different time intervals in Odense. Figure 11 presents an example of precipitation data measured every hour between January 2024 and February 2025, which can be incorporated into the platform and utilized for developing numerical and AI-based models for the Danish demo site. Figure 11 Example of 1-hour interval precipitation data recorded in the vicinity of Odense between January 2024 and February 2025. Rainfall datasets will be integrated into the AIAssisted platform and used for numerical and AI-based model development within the Danish demo site. 2.4.2 Santander Case Study (Spain) 2.4.2.2.1 Spatial domain definition Defining the spatial domain constituted the first activity within the preparatory activities to kick off the Santander case study, as it establishes the area of interest for the rest of activities to be performed in the platform such as data storage, analysis and result generation. In the case of Santander, the whole municipality, which covers 36 km² is considered. To use these boundaries in the platform, GeoJSON files were elaborated and uploading to the platform incorporating other relevant data such as the population in this area. Figure 12 depicts the location of the Spanish demo site, being highlighted in red the municipal boundary of Santander and its location in the north of the Iberian Peninsula. 0 2 4 6 8 10 12 14 16 18 20 Precipitation (mm) 25 D5.1 Initial preparatory and testing report in case studies Figure 12 Location of the Spanish demo site of the D4RUNOFF project, highlighting the municipal boundary of Cantabria (red dashed line) along with the spatial domain defined in the current phase of the project for its implementation in the AI-Assisted platform. 2.4.2.2.2 Combined sewer network evaluation To implementing the AI-Assisted platform at the Spanish demo site, the information on the sewer system within the Santander area is of paramount relevance to better understand the current infrastructure which is being used to manage urban runoff. The information on which the subsequent analysis of the wastewater and drainage infrastructure of Santander is based has been supplied by the lead partner for the case study (AQUALIA). The initial understanding of this information has been the main target for the work undertaken during the initial phase of T5.4. As detailed below, Santander global water cycle counts with a hybrid system which comprises two types of NBS, a wetland in “Las Llamas” and a permeable pavement parking lot close to it. Additionally, a traditional system consisting of several pumping stations that convey water towards the wastewater treatment plant, which discharges the treated effluent into the Cantabrian Sea is also present. As depicted by Figure 13 , the collected water comes mainly from urban areas and commercial and industrial units. The current challenges faced by the system are mainly connected with the risk of its collapse due to water runoff and stormwater overflows incidents when combined sewer overflows may occur. 32 D5.1 Initial preparatory and testing report in case studies analysis of the drainage and wastewater infrastructure in Pontedera, provided by the project’s lead partner for the case study (ACQUE SPA), has been the focus of the preparatory activities carried out in the initial phase of task T5.4. Pontedera relies predominantly on a CSN that conveys both stormwater and WWTP located to the east of the city, which discharges the treated water into the aquatic environment near the Arno River. As shown in Figure 20, which illustrates land use in the Italian demo site, the water collected by the drainage and sanitation infrastructure mainly comes from urban and industrial areas. However, during periods of intense rainfall, there is potential for discharges of polluted water into the aquatic environment due to the insufficient capacity of the CSN to collect stormwater and wastewater. Figure 20 Land use within the spatial domain of the Italian demo site defined in the AI-Assisted platform. The water collected by the sanitation infrastructure in the study area comes mainly from urban areas (yellow zones) and industrial areas (blue zones). For the implementation and evaluation of the performance of the sanitation infrastructure within the AI-Assisted platform, six types of elements constitute the main components of the CSN of Pontedera: manholes, pipes, pumping stations, discharge points or outlets, overflows or spill points, and the WWTP. In addition, 1086 subcatchments are identified in the urban area contributing runoff water to specific manholes within the urban area of Pontedera. A specific GIS database has been prepared during the preparatory activities of task 5.4 as a starting point for its subsequent upload to the platform. Figure 21 presents the spatial distribution of the different elements that make up the CSN in the Italian demo site, as visualized in the prepared GIS database. The sewerage infrastructure of Pontedera has been analysed to ensure its successful integration into the AI-Assisted platform once its development is complete, as described in Section 4.3.3 of deliverable D4.2 within WP4 of the project. Infrastructure data can be uploaded to the platform either individually, element by element, or in bulk using preformatted GeoJSON 33 D5.1 Initial preparatory and testing report in case studies files that the platform can interpret and manage efficiently. The primary goal of this data analysis is to guarantee that the platform’s design and development allow for the correct storage and management of the various types of elements present in the CSN of the Italian demo site. The available data on sewer network pf Pontedera includes more than 60 km of installed pipelines. Over 60% of these have a circular cross-section, while more than 20% have a rectangular profile, and the remaining ones are irregularly shaped. Concrete and masonry account for 87% of the pipeline materials, while plastic materials (PVC) represent only 7%. These material distributions reflect the network’s age: most pipelines (59%) were installed around 1970, although significant renovations and new installations took place between 2016 and 2018, accounting for 32.5% of the total network. Most of the network consists of mediumsized diameters (200–800 mm), with large-diameter pipelines (>1 m) being relatively scarce. The pipelines transport both wastewater and stormwater to the WWTP, where the main spillways are also located. Figure 21 Spatial distribution of the combined sewer network in the Italian demo site, highlighting the main components that constitute it and will be integrated into the AI-Assisted platform. 2.4.2.3.3 NBS evaluation For the Italian demo site, the specific NBS to be implemented in the AI-Assisted Platform for assessing their impact on urban runoff management will be determined in the upcoming phases of the D4RUNOFF project. 34 D5.1 Initial preparatory and testing report in case studies 2.4.2.3.4 GIS-based tools identification The platform being developed within D4RUNOFF is a web-based GIS tool capable of connecting to external sources to incorporate relevant information that contextualizes and enhances the specific data and results generated within the project. During the preparatory activities for task T5.4, various external sources have been identified as potential data providers for the Italian demo site, which could be integrated into the platform via WMS (Web Map Service) or similar services. Specifically, several external sources have been identified for potential integration with the platform, providing contextual information on terrain elevation (digital terrain models), geology, slopes, orthophotos, hydrography, impermeability levels, and more. The AI-Assisted platform has been designed to support various base maps (customizable according to user preferences) from external sources, alongside the spatial distribution of project-specific data and results. This integration enhances the comprehension and geospatial analysis of the information. Figure 22 presents examples of maps from external sources available for incorporation into the AI-Assisted platform, offering contextual information on terrain elevation, geology, hydrology, and soil impermeability levels. Figure 22 Example of external GIS data sources available for integration into the AI-Assisted platform, providing contextual information for the Italian demo site: (a) Terrain elevation (Digital Elevation Model), (b) Geology, (c) Hydrology, and (d) Soil impermeability levels. a) b) d) c) 35 D5.1 Initial preparatory and testing report in case studies 2.4.2.3.5 Preliminary analysis of available data External monitoring sensors operating at the Italian demo site have been identified as potential data sources for integration into the AI-Assisted platform within the D4RUNOFF project. One of the key external data providers is the Regional Hydrological Service (SIR) of the Tuscany Region, which manages a network of monitoring stations collecting real-time hydrological and meteorological data. For the Pontedera area, the SIR provides hydrometeorological data from its monitoring stations, including precipitation levels, river water levels, temperatures, and other relevant parameters. Particularly crucial for the D4RUNOFF project and the effective operation of the AI-Assisted platform is the availability of high-frequency precipitation data, as these measurements enable the assessment of runoff generation caused by rainfall events of varying intensities and their correlation with key pollution indicators. In this context, the SIR provides precipitation data recorded at 15-minute intervals in Pontedera. Figure 6 presents an example of precipitation data measured every 15 minutes between 2020 and 2024, which will be incorporated into the platform and utilized in the development of numerical and AI-based models for the Italian demo site. Figure 23 Example of 15-minute interval precipitation data recorded in Pontedera between 2020 and 2024. This dataset will be integrated into the AI-Assisted platform and used for numerical and AI-based model development within the Italian demo site. 2.4.3 Testing activities The core of the testing activities for the AI-Assisted platform at the three demo sites will take place starting from month 30 of the project (M30). During this phase, information and data about each site will be integrated into the platform as a foundation for model generation and the production of results that will support platform users in making informed decisions regarding urban runoff. However, during the initial phase of task T5.4, in addition to the previously mentioned preparatory actions, some preliminary activities related to the validation of data upload and reception on the AI-Assisted platform have also been carried out. 0 2 4 6 8 10 12 Precipitation (mm) 36 D5.1 Initial preparatory and testing report in case studies The platform will operate using results and data generated within different work packages of the D4RUNOFF project. On one hand, it will utilize results from laboratory analyses of runoff water samples taken at the pilot sites using innovative contaminant characterization methods developed within WP1. On the other hand, it will incorporate data collected by the novel sensors developed within the framework of WP2 for the identification and monitoring of CECs and new pollutants. The initial testing activities have focused on validating the proper storage of results from WP1 and data from WP2, as well as improving the platform to enable real-time communication with the novel sensors being developed within the project. First, actions have been taken to validate the proper storage of the results generated within WP1 of the project in the AI-Assisted platform, using the data model designed for this purpose within WP4 (see Deliverable D4.2, Section 4). This model aims to ensure the unambiguous storage, organization, and structuring of the information from the collected samples at the pilot sites by identifying their location based on interrelated elements. The testing activities carried out during the initial phase of task T5.4 focused on verifying the proper storage of preliminary information and results from WP1 on the platform. These activities have helped refine and improve the storage system for these results through a data model structured into the following elements, from the top level to the bottom: 1) Site. Defines the general location where samples are collected. The sites correspond to the three pilot locations of the project, along with external collaboration sites. 2) Station. Specific sampling location within a site, covering different scales: 1-2 largescale/composite sources and 3-5 small-scale/local sources. Local stations (e.g., rain beds, storm drain runoff) must have a sufficiently large catchment area to ensure representative sampling. 3) Sampling Point. The exact location within a station where a sample is collected. Samples can be taken from different elements of the NBS or sewer network (e.g., inlet, outlet, bypass pipes). They are categorized as "lab" (samples analyzed in a laboratory) or "device" (samples collected by sensors for in situ analysis). 4) Sample. A portion of a substance (water or soil) collected at a sampling point. A sample is mainly identified by its sampling point, collection date, and additional parameters. 5) Analysis. The process of determining the measured result of a parameter within a sample. Each analysis is mainly defined by the sample (sampling point + date), the analysed compound/parameter, the measurement result, and its units. Second, a set of preliminary tasks has been also undertaken during this phase to initiate the validation of the data communication with sensors developed in WP2 for the identification of new pollutants and CECs. To streamline the acquisition to the AI-Assisted platform from these devices, a significant enhancement has been made thanks to the implementation of MQTT protocol that allows to interact with the sensors for some of the following purposes: 1) Data acquisition. Automatic data acquisition form observed parameters measured by the abovementioned sensors have been tested as shown in Figure 24. 2) Actions. Several examples of MQTT actions were examined during this initial stage to validate that the platform can interact with the devices. For instance, one action could 37 D5.1 Initial preparatory and testing report in case studies be used to initiate a sampling process (Figure 25). Moreover, over this period, the possibility of triggering these actions based on other observed parameters (e.g. rainfall or precipitation probability) or a defined schedule has begun to be considered. In the coming months, additional tests are expected to be conducted to fine-tune these sensors operation scenarios Figure 24 Screenshot of the tested connection between devices in WP2 and AI-Assisted Platform to acquire observed data 38 D5.1 Initial preparatory and testing report in case studies Figure 25 Screenshot of some MQTT actions tested during the initial activities to validate the communication between the AI-Assisted Platform and the WP2 sensors 2.4.4 Conclusions and next steps During the initial phase of task T5.4 for the implementation of the AI-assisted platform at the demo sites, efforts have focused on preparatory activities, analysing the characteristics and specific features of the three pilot sites and evaluating the available information and data at each location. Additionally, the availability of external information sources, including GIS tools and real-time data sources, has been identified as potential integration pathways for the platform. Preliminary testing activities were also carried out during the first months of task T5.4. These tests verified, on one hand, the proper storage of project-generated data and results in the AIassisted platform and, on the other hand, the communication between the platform and the novel sensors being developed in D4RUNOFF. Particularly significant were the improvements made to the platform for using the MQTT protocol to interact with the sensors, as well as the tests conducted to validate platform-sensor communication. In the next phases, testing and validation of the AI-assisted platform at each demo site will be further developed, integrating the available information and data from each location. These data sources will serve as the foundation for the development and execution of AI-based models within the platform at the three pilot sites, as well as for generating numerical results to assess the effectiveness of NBS implementation. The progress and achievements of the AIassisted platform implementation in future phases of task T5.4 will be detailed in Deliverables D5.3 (intermediate testing) and D5.4 (final testing and conclusions). 39 D5.1 Initial preparatory and testing report in case studies 2.5 Task 5.5: Multi-Criteria Decision Analysis (MCDA) Case study leaders (VCS, ACQUE and AQUALIA) participated directly during the development of the MCDA methodology in the WP3. Thanks to this, it was possible to identify some potential limitations for its application from the beginning, as for example the scarce power of decision of the utilities now or the lack of consideration of some new environmental criteria. Now, in the WP5, the limitations are addressed with three main objectives: • Perform the application of the WP3 results to the case studies, in coordination with the WP4 and the T5.4, using the related functionalities of the D4RUNOFF platform. • Put into value the construction of NBS in the different locations and learn from it to update the parametric design library and the ranking obtained from the MCDA. • Propose the construction of new NBS soon using the MCDA and the GIS analysis of the selected areas in each city. The preliminary actions developed during these last months with the support of the case study leaders (VCS, ACQUE and AQUALIA) have been: 1. Review the deliverables D3.1, D3.2 (both public and available in https://d4runoff.eu/results/), D3.3 (sensible and available only for the partner organizations) and D3.4 (public summary). 2. Look for NBS under construction now and soon (e.g., the SSF wetland in Pontedera) to compare the designs with the ones that can be obtained from the parametric library. 3. Share the available information about the cost of construction and maintenance of urban drainage techniques in each country (i.e. national database of prices, including not only pipes but also NBS) to particularize some figures used in the MCDA. 4. Validate the ranking of NBS obtained from the MCDA and evaluate if it needs to be adapted to each specific location to generate a particular ranking. 5. Confirm the selection of the specific study area in each city (e.g., Las Llamas valley in Santander) to focus there the analysis of the potential NBSs to execute in the future solving existing problems related with urban drainage (this means update the results of the D3.3 if needed). From the results of the WP3 and the responses of the case study leaders, the effectiveness of the real application of the MCDA has been discussed during the preliminary actions of this task 5.5. VCS states that most of the NBS techniques considered in the ranking performed with the MCDA methodology are not commonly used in Denmark for rainwater management. The most used ones are the rain beds (bioretention areas) with a variable cost that goes from 16.000 to 31.500 €for around 20 square meters, which means 2 to 3 times the average cost considered in WP3. Consequently, the NBS ranking from the MCDA needs an important adaptation to the local practice in Denmark to ensure a proper application. Mainly particularizing the list of techniques and the vocabulary use together with the real costs of construction. ACQUE is now working in constructing a SSF wetland in Pontedera, taking advantage of this to check the preliminary results obtained from the parametric library developed in the WP3. The comparison shows that the library is useful for the initial design of different alternatives but not yet for real projects where all the details must be described. Moreover, as in Denmark, in Italy is also needed a selection of alternatives and a vocabulary adaptation to apply the NBS ranking obtained from the MCDA performed in the WP3. 40 D5.1 Initial preparatory and testing report in case studies Finally, AQUALIA highlights the importance of one of the initially identified limitations: utilities do not decide what NBS must be applied where. Consequently, even though the company in charge of the water management agrees with the NBS ranking coming from the MCDA, it is the corresponding water authority the one deciding, depending on the feasible solutions available in the market in that specific moment. With all, in Santander, the importance of the wetland constructed in Las Llamas Park in growing thanks to the D4RUNOFF project, and similar NBS are being considered in the plans development of the city. In the three case studies, the selected study areas have been confirmed. In the next months of the WP5, the MCDA methodology will be adapted to each location, collecting the recommendations from the three case studies and implementing the improvements needed for the tailored application in each of the selected areas in Odense, Pontedera and Santander. The consideration of local updated costs in each location (Santander, Odense and Pontedera) could affect the approach for the selection of NBS, modifying the initial proposed ranking and consequently the potential locations. The local authorities will have to decide if the initial weighting is suitable or they prefer to change it. As reference, with the objective weighting methodology used, the Entropy Weight Method (EWM), the construction costs have a weight of 13,44% and the maintenance costs a 15,75%, being the two more important indicators to consider. 41 D5.1 Initial preparatory and testing report in case studies 3 Feedback to WP1-WP4 Although WP1–WP4 activities is almost completed by M30, feedback remains crucial as WP5 is currently testing and applying the developed methods and technologies in practice. In this deliverable, the feedback is presented not as a basis for adjusting WP1–WP4, but as a reflection on the achieved results and an evaluation of their implementation. This documentation of key learnings ensures that we capture important insights and identify any future needs for adjustments in the application of these methods. Moreover, the feedback serves to address questions regarding our approach and helps improve guidelines for similar projects in the future. WP1 – Contaminant Characterization and Analytical Methods • In WP1, it became clear that thorough planning of the sampling process is essential. In particular, several sampling locations (especially in Odense) had to be changed or even cancelled because the limitations of the Nature Based Solutions were not fully understood beforehand. • Additionally, the safe transportation of sampling equipment must be ensured to prevent damage during transit. • Moreover, a high degree of coordination is required to ensure that physical sampling is effectively synchronized with sensor data collection, considering both weather conditions and the practical challenges of collecting samples concurrently with sensor operations. WP2 – Online Monitoring System • Initial tests with the online monitoring system are scheduled to start in month 31. • Close collaboration is essential to resolve technical issues and ensure smooth integration of the sensors with the AI-assisted platform. WP3 – Multi-Criteria Decision Analysis (MCDA) and NBS Evaluation • The academic approach initially adopted for the MCDA has highlighted discrepancies when compared to professional practice, particularly regarding cost considerations and terminology. • It is recommended that WP3 leverage the outcomes of tasks T5.5 and T5.6 to adapt the MCDA methodology to the local contexts of Denmark, Italy, and Spain. • These adaptations are crucial for producing a more realistic ranking of proposed NBS options for urban water management. • Adapt the MCDA methodology to better reflect local conditions by updating cost data and aligning vocabulary, ensuring the decision framework is practical for professional water management applications. WP4 – AI-Assisted Platform and Data Integration • The primary focus during this phase has been the evaluation of available data at each pilot site, including two main groups: o GIS Data Sources: Assessment of metadata from sewer networks, NBS identification layers, and base maps (e.g., aerial images, land use). o Real-Time Data: Verification of data connections from internal sensors and external providers (e.g., weather forecast services).