This study focuses on exploring the potential of using Long Short-Term Memory networks (LSTMs) for low-flow forecasting for the Rhine River at Lobith on a daily scale with lead times up to 46 days ahead. A novel LSTM-based model architecture is designed to leverage both historical observation and forecasted meteorological data to carry out multi-step discharge time series forecasting. The feature and target selection for this deep learning (DL) model involves evaluating the use of different spatial resolutions for meteorological forcing (basin-averaged or subbasin-averaged), the impact of incorporating past discharge observations, and the use of different target variables (discharge Q or time-differenced discharge dQ). Then, the model is trained using the ERA5 dataset as meteorological forcing, and employed for operational forecast with ECMWF seasonal forecast (SEAS5) data. The forecast results are compared to a benchmark process-based model, wflow_sbm. This study also explores the flexibility of the DL model by fine-tuning the pretrained model with limited SEAS5 dataset. Key findings from feature and target selection include: (1) opting for subbasin-averaged meteorological variables significantly improves model performance compared to a basin-averaged approach. (2) Utilizing dQ as the target variable greatly boosts short-term forecast accuracy compared to using Q, with a mean absolute error (MAE) of 25 m³ s⁻¹ and mean absolute percentage error (MAPE) of 0.02 for the first lead time, ensuring reliability and accuracy at the onset of the forecast horizon. (3) While incorporating historical discharge improves the forecasting of Q, its impact on predicting dQ is less pronounced for short lead times. In the operational forecast with SEAS5, compared to the wflow_sbm model, the DL model exhibits skill in forecasting low flows as evidenced by Continuous Ranked Probability Skill Score (CRPSS) median values of all lead times above zero, and better accuracy in forecasting drought events within short lead times. The wflow_sbm model shows higher accuracy for longer lead times. In the exploration of fine-tuning approach, the fine-tuned model generates marginal short-term enhancements in forecasting low-flow events over a non-fine-tuned model. Overall, this study contributes to advancing the field of low-flow forecasting using deep learning approach.

Hydrodynamic models are used to analyse water networks (water distribution, drainage, surface water, district heating, etc.). The non-linear nature of water flows necessitates the use of iterative solution methods in hydraulic modelling. This requires a relatively large computational effort. To reduce this effort, networks, network forcing and/or the flow in networks are often simplified and analysed using the Graph Theory. The simplification options depend on the network characteristics. There are many topological features to describe Graph-based networks. In this paper, these characteristics are summarised, applied on 7 urban drainage networks and discussed. As the topological features do not describe the networks in a uniform manner, a new type of topological characterisation of looped drainage networks (Network Linearisation Parameter, NLP) is proposed based on linearized hydraulics and bottlenecks identified in paths to outfalls.

Drinking water distribution networks (WDNs) are a crucial infrastructure for life in cities. Deterioration of this ageing, and partly hidden from view, infrastructure can result in losses due to leakage and an increased contamination risk. To counteract this, maintenance strategies are required to maintain the service level. Information on the most critical elements of a WDN, with respect to the functioning of the system as a whole, is essential for prioritising maintenance or rehabilitation activities. In this study a Graph theory based method is developed and applied for efficiently identifying the most critical elements. The main advantage of this method is that it avoids the need to perform elaborate hydrodynamic model calculations. Instead, the structure of the network is the main starting point. The results show that the structure of the network is more decisive than the hydraulics with respect to the criticality of the system’s performance as a whole. Results depict that the suggested approach is applicable not only to the main (primary) network, but also to the capillaries which are normally beyond the scope of the traditional methods applied so-far because of the complexity of the networks and the required calculation time.

The description of intertwined ecological processes in surface waters requires a holistic approach that accounts for spatially distributed hydrological/water quality processes. This study describes a new approach to model dissolved oxygen (DO) based on linked hydrodynamic and closed nutrient cycle ecological models. Long term datasets from the River Dommel (Netherlands) are used to determine: 1) if this methodology is suitable for modelling DO concentrations, 2) the model sensitivity to various levels of nutrients input, and 3) the DO production and consumption processes and their response to nutrient input changes. Results show that seasonal dynamics of DO are well quantified at long timescales; the sensitivity of DO to different pollutant sources exhibits significant seasonal variation and the largest influences on DO are aeration and mineralization of organic material. The approach demonstrates an ability to consider the impacts of nutrient input and long term vegetation maintenance on ecological quality.