Hessel C. Winsemius

Large-scale optical non-contact surface velocimetry provides an inexpensive, flexible, and scalable approach for remote monitoring of river flows by tracking the displacement of tracers on the river surface. However, its accuracy can be compromised by surface waves, such as wind waves and standing waves, which may propagate at speeds different from the underlying flow. Recently developed physics-based wave imaging approaches address this limitation by exploiting the spectral signature of surface-wave dynamics to estimate flow velocity. Despite their potential, these approaches have not yet been directly benchmarked against more established surface velocimetry techniques such as Large-Scale Particle Image Velocimetry (LSPIV) and Space-Time Image Velocimetry (STIV), making the selection of the appropriate method for a specific application difficult.Here, we present the first systematic benchmark of a novel physics-based velocimetry software, Image Wave Velocimetry Estimation (IWaVE). We tested the new approach against LSPIV and STIV across a range of challenging measurement conditions: a wind-swept tidal channel, a large river, and a morphodynamically heterogeneous mountain stream. In all tested conditions, IWaVE provided reliable velocity measurements, whereas LSPIV and STIV failed under strong winds and in the absence of suitable surface tracers. These results demonstrate the robustness of physics-based wave imaging for conditions that challenge conventional image-based velocimetry. The enhanced measurement capability of this new method extends non-contact flow monitoring applications to a broader range of environments beyond the current capabilities of non-contact surface velocimetry approaches.

Autonomous camera-based systems combined with image velocimetry analyses, enables operational river flow measurements in rapidly responding rivers. The open-source OpenRiverCam (ORC) software stack supports edge and cloud video processing, time-series generation, and rating-curve development, enabling fully operational, scalable non-contact water level and discharge estimation with relatively affordable camera systems.Despite these advances, image calibration remains a major bottleneck for broad uptake, as it typically requires high-precision surveying of non-collinear ground control points to constrain the camera's pose. This process is often complex and relies on instruments that are not readily available to many users.We investigate a photogrammetry-based alternative workflow for camera pose estimation for possible integration in ORC: during camera installation, users collect a set of smartphone photographs from multiple viewpoints near the camera location. A photogrammetric reconstruction using these photos together with a sample video from the installed camera, jointly estimates the camera pose and lens parameters. The resulting camera pose is then used to orthorectify videos in operational data collection. Using controlled experiments and field experiments in New Zealand, Zambia and The Netherlands, we assess here (i) the accuracy of reconstructed 3D coordinates compared to traditional calibration, (ii) methods to robustly constrain the horizontal plane, (iii) the number of photographs required, and (iv) the influence of GPS accuracy on the solution.This approach aims to significantly simplify calibration workflows and lower the barrier to deploying camera-based river monitoring systems.

Deployment and upscale of traditional monitoring systems have different challenges, specially in developing countries, because of high investment costs, difficult operation and maintenance. In the last years there have been several low-cost, open-source products developed as well as initiatives with the objective of tackling down those issues.Herein we present the setup and first leanings of the project L-DaaS “Local people for Discharge monitoring as a Service” where we proposed a scheme of using an open-source software for flow monitoring and a low-cost hardware which use standard components with a business model centred around a local enterprise in charge of the operation and maintenance. The project was executed in Zambia with a locally-driven environmental monitoring company based in the same country. Key stakeholders where the Water Resource Management Authority of Zambia and hydropower operators.Open-source and low-cost system are not by themselves a solution, it is also needed to create a sustainable and scalable business model which fosters affordable, efficient, and locally supported water management solutions.

TEMBO Africa is a Horizon Europe project that seeks to improve in situ sensing of weather and water in sub-Saharan Africa. To ensure beyond-the-project sustainability, we are using innovative sensors to measure variables such as rainfall, bathymetry, river flow, and large-scale soil moisture. TEMBO also develops services for hydropower, agriculture, and disaster management. These services will produce societal and economical value, for which governments and companies are willing to pay. These payments, in turn, serve to maintain the observation networks. One guiding principle is that the new data gathering method should cost less than 10% of existing methods in term of total costs of ownership. This principle implicitly pays special attention to the local availability of human resources. Many monitoring projects in Africa consist of installation by experts from the Global North, followed by a short training of local technicians. This works nicely until something breaks down. In TEMBO, African universities and spin-off companies are co-developing the technologies such that any operational problems can be solved without flying in expensive foreign experts. In this presentation, we will go through the sensor innovations and how these feed into different products and services.

Measuring river surface velocity enables river discharge estimation, a fundamental task for hydrologists, environmental scientists, and water resource managers. While traditional image-based velocimetry methods are often effective, they struggle to produce complete velocity fields under complex environmental conditions. Poor lighting, reflective glare, lack of visible surface features, or excessive turbulence can all result in regions where feature tracking fails, leading to gaps in the resolved velocity field. Addressing these gaps through the reconstruction of missing velocity measurements is an important research challenge. Recently, researchers have employed deep learning to address various hydrology problems, demonstrating promising improvements. In this work, we propose a neural operator-based model to address the challenge of missing velocities in river surface velocimetry. Specifically, our model is based on the Fourier neural operator with a graph-enhanced lifting layer. It is trained on the river surface velocimetry reconstruction task using a self-supervised paradigm. Once trained, it can be used to infer missing velocities in unseen samples. Experiments conducted on a dataset collected from a river in the Netherlands demonstrate our approach’s ability to accurately infill missing surface velocities, even when faced with large amounts of missing data. We attribute this robustness to the neural operator’s ability to learn continuous functions, which enhances our model’s capacity for high-level feature representation and extraction. Our findings suggest that the reconstructed velocity fields produced by our model can act as reliable ground truth data for deep learning-based methods. In the future, we aim to improve our model’s performance and generalization by incorporating additional data collected from a wider range of rivers and under varying environmental conditions.

Herein we present an analysis of the performance of the Image Wave Velocimetry Estimation (IWaVE), a python library for image-based river discharge calculations. IWaVE simultaneously performs a 2D velocimetry analysis and calculates the stream depth through 2D Fourier transform, exploiting the sensitivity of water wave dynamics to flow conditions. Unlike existing velocimetry approaches such as Particle Image Velocimetry (PIV), Particle Tracking Velocimetry (PTV) or Space-Time Image Velocimetry (STIV), the uniqueness of this approach lies in: 1) velocities that are advective of nature can be distinguished from other wave forms such as wind waves. This makes the approach particularly useful in estuaries or river stretches affected strongly by wind, or in shallow streams in the presence of standing waves. 2) The velocity is estimated based on the physical behavior of the water surface, accounting for the speed of propagation of waves and ripples relative to the main flow. This makes the approach more robust than traditional methods when there are no visible tracers. 3) If the depth is not known, it can be estimated along with the optimization of x and y-directional velocity. Depth estimations are reliable only in fast and shallow flows, where wave dynamics are significantly affected by the finite depth.We analyzed 2 videos recorded from a drone on a site in the Netherlands over a tidal channel in Zeeland at Waterdunen - Breskens. One of the videos has strong winds, which creates waves moving upstream. ADCP measurements for both videos are available. The videos were taken at different moments during different tidal conditions, they were processed using IWaVE, a LSPIV and a STIV methods. The results show that the LSPIV, STIV and IWaVE are in good agreement with the ADCP measurements for the case where there is no wind. However when there is wind the LSPIV and STIV methods fail to obtain the correct surface velocity, while the velocity calculated with IWaVE is in good accordance with the ADCP.

Flood risk assessments are increasingly guiding urban developments to safeguard against flooding. These assessments, consisting mainly of hazard and risk maps, make use of interconnected models consisting of a chain of climate, hydrological, hydraulic, and impact models, which are increasingly run interactively to support scenario modelling and decision-making in digital twins. To maintain interoperability, transparency, and reusability of this chain and the assessments themselves, using a workflow manager to manage the inter-model dependencies is a natural fit. However, composing and maintaining workflows is a non-trivial, time-consuming task, and they often have to be refactored for new workflow engines, or when changing compute environments, even if the workflow conceptually remains unchanged. These issues are particularly relevant in the development of digital twins for climate adaptation, where flood risk assessments serve as input to indicate high-risk areas. The complex model chain underpinning such digital twins can benefit greatly from transparent workflows that can be easily reused across different contexts.To address these challenges, we developed the HydroFlows Python framework for composing and maintaining flood risk assessment workflows by leveraging common patterns identified across different workflows. The framework allows users to use one of the many steps available in the library or define workflow steps themselves and combine these into complete workflows which are validated on the fly. Available workflow steps include building, running, and postprocessing of models. Execution of the workflow is handled by one of the workflow managers to which our workflow description can be exported, such as Snakemake or tools with CWL support. This flexibility allows users to easily scale their workflows to different compute environments whenever the computational requirements demand so.We demonstrate the flexibility of the HydroFlows framework by highlighting how it can be used to create complex workflows needed for digital twins supporting climate adaptation. HydroFlows not only enhances the flexibility and portability of the digital twin modelling workflows but also facilitates the integration of digital twin tooling and advanced computing and processing solutions to support interactive flood risk assessments in federated compute and data environments.

Understanding extreme discharge behavior is of importance for flood design and risk management. For example, estimates of large extreme discharge return periods such as the 100-year return period or higher are often needed as a basis for flood hazard maps or dike design. Yet, frequency analysis based on decade-long discharge records show a large uncertainty for these frequencies, among others due to the statistical uncertainty from the distribution parameters. This is not the case for the shape parameter, a key parameter that describes the upward or downward curvature of the tail of the distribution and thus an indicator of extreme discharge behavior. This study provides robust estimates of the shape parameter by using the 1,040 years of synthetic daily discharge generated for the Meuse catchment as part of the EMfloodResilience project from the Interreg Euregio Meuse-Rhine program. The spatially-distributed hydrological model wflow_sbm, calibrated and validated for the Meuse catchment, is forced with 16 synthetic climate ensembles of 65 years representative for the current climate from the physically-based KNMI regional climate model RACMO climate model at a daily and hourly time step. The annual maxima (AM) from hydrological years (Oct-Sep) are retrieved from these continuous time series, and a GEV distribution is fit to the AM. We observe a clear spatial pattern of the shape parameter across the Meuse catchment. Using this large dataset of shape parameters, we also review the possible reasons for the different tail behavior obtained with respect to rainfall statistics, catchment characteristics and river systems following the In doing so, we aim to bridge the extreme value statistical modelling with our current understanding of the extreme hydrological signatures present in the catchment.

Modelling of compound flood events, the assessment of their impact, and assessing mitigation and adaptation measures is in increasing demand for local authorities and stakeholders to support their decision making. Additionally, the severity of extreme events driving compound flooding, including storms and heavy rainfall, is projected to increase under climate change. To support local communities in flood risk management, complex modelling systems involving multiple cross-disciplinary models need to be orchestrated in order to effectively and efficiently run a wide range of what-if scenarios or historical events to understand the drivers and impacts of compound floods. The large volume and variety of data needed to configure the necessary models and simulate events strain the reproducibility of modelling frameworks, while the number of events and scenarios demand increasingly powerful computing resources. Here we present a solution to these challenges using automated workflows, leveraging the Common Workflow Language standard. The presented workflows update a base model configuration for a user-specified event or scenario, and automatically reruns multiple defined scenarios. The models are executed in containers and dispatched using the StreamFlow workflow manager designed for hybrid computing infrastructures. This solution offers a single, uniform interface for configuring all models involved in the model train, while also offering a single interface for running the model chain locally or on high performance computing infrastructures. The allows researchers to leverage data and computing resources more efficiently and provide them with a larger and more accurate range of compound flood events to support local authorities and stakeholders in their decision making.