Small utilities often lack the required amount of data to train machine learning-based models to predict pipe failures, and hence are unable to harness the possibilities and predictive power of machine learning. This study evaluates the generalizability and transferability of a machine learning model to see if small utilities can benefit from the data and models of other utilities. Using nine Norwegian utilities’ datasets, we trained nine global models (by merging multiple datasets) and nine local models (by utilizing each utility's dataset) using random survival forest. Several pre-processing techniques including addressing left-truncated break data and break data scarcity are also presented. The global models and three of the local models were tested to predict the pipe failure of the utilities which were not included in their training datasets. The results indicate that the global models can predict other utilities with sufficient accuracy while local models have some limitations. However, if a representative utility with a sufficiently large (and information rich) dataset is selected, its model can predict the other utility's pipe breaks as accurate as the global models. Furthermore, survival curves for defined cohorts as proxies for uncertainty, and variable importance show that pipes with and without previous breaks behave extremely different. With the understanding of models’ generalizability and transferability, small utilities can benefit from the data and models of other utilities.

@en

This paper aims to stimulate discussion based on the experiences derived from the QUICS project (Quantifying Uncertainty in Integrated Catchment Studies). First it briefly discusses the current state of knowledge on uncertainties in sub-models of integrated catchment models and the existing frameworks for analysing uncertainty. Furthermore, it compares the relative approaches of both building and calibrating fully integrated models or linking separate sub-models. It also discusses the implications of model linkage on overall uncertainty and how to define an acceptable level of model complexity. This discussion includes, whether we should shift our attention from uncertainties due to linkage, when using linked models, to uncertainties in model structure by necessary simplification or by using more parameters. This discussion attempts to address the question as to whether there is an increase in uncertainty by linking these models or if a compensation effect could take place and that overall uncertainty in key water quality parameters actually decreases. Finally, challenges in the application of uncertainty analysis in integrated catchment water quality modelling, as encountered in this project, are discussed and recommendations for future research areas are highlighted.@en

Receiving water quality simulation in highly urbanised areas requires the integration of several processes occurring at different space-time scales. These integrated catchment models deliver results with a significant uncertainty level associated. Still, uncertainty analysis is seldom applied in practice and the relative contribution of the individual model elements is poorly understood. Often the available methods are applied to relatively small systems or individual sub-systems, due to limitations in organisational and computational resources. Consequently this work presents an uncertainty propagation and decomposition scheme of an integrated water quality modelling study for the evaluation of dissolved oxygen dynamics in a large-scale urbanised river catchment in the Netherlands. Forward propagation of the measured and elicited uncertainty input-parametric distributions was proposed and contrasted with monitoring data series. Prior ranges for river water quality-quantity parameters lead to high uncertainty in dissolved oxygen predictions, thus the need for formal calibration to adapt to the local dynamics is highlighted. After inferring the river process parameters with system measurements of flow and dissolved oxygen, combined sewer overflow pollution loads became the dominant uncertainty source along with rainfall variability. As a result, insights gained in this paper can help in planning and directing further monitoring and modelling efforts in the system. When comparing these modelling results to existing national guidelines it is shown that the commonly used concentration-duration-frequency tables should not be the only metric used to select mitigation alternatives and may need to be adapted in order to cope with uncertainties.@en

The calibration of models for urban drainage systems has become more and more important as especially the usage of detailed models has increased considerably over the last years as the basis for planning and design. Still the effects originating from the choice of data used for model calibration are little known and advice on planning measurement campaigns for model calibration is limited, especially for small and medium-sized municipalities. The choice of measurement sites (number and location) within a sewer system is affecting the robustness of the calibration and in consequence the assessment of the modelled system behaviour. This paper discusses the calibration of a hydrologic-hydrodynamic model using the representative example of a small municipality. Different calibration scenarios were created using a model-based approach, focusing on varying availability of in-sewer measurement data. To assess the performance of different scenarios and validate the respective models, different model outputs were compared. The different calibration scenarios resulted in high variations in the model performances. The number and location of used calibration points influence model performance significantly. Predicted CSO volumes deviate from a set of given reference values in ranges between 1% and 253% for one, −21% to −5% for two and 1% to 237% for five used calibration points, depending on the rainfall data input. Consequently, the design of measurement campaigns for calibration data is a very sensitive decision in the modelling process. The model performance further influences design and decision-making processes, which are then perceptible in economic and functional aspects.

@en

Pipe failure modelling is an important tool for strategic rehabilitation planning of urban water distribution infrastructure. Rehabilitation predictions are mostly based on existing network data and historical failure records, both of varying quality. This paper presents a framework for the extraction and processing of such data to use it for training of decision tree-based machine learning methods. The performance of trained models for predicting pipe failures is evaluated for simple as well as more advanced, ensemble-based, decision tree methods. Bootstrap aggregation and boosting techniques are used to improve the accuracy of the models. The models are trained on 50% of the available data and their performance is evaluated using confusion matrices and receiver operating characteristic curves. While all models show very good performance, the boosted decision tree approach using random undersampling turns out to have the best performance and thus is applied to a real world case study. The applicability of decision tree methods for practical rehabilitation planning is demonstrated for the pipe network of a medium sized city.

@en

Although calibration of a hydrodynamic model depends on the availability of measurement data representing the system behavior, advice for the planning of necessary measurement campaigns for model calibration is scarce. This work tries to address this question of efficient measurement site selection on a network scale for the objective of calibrating a hydrodynamic model case study in Austria. For this, a model-based approach is chosen, as the method should be able to be used before measurement data is available. An existing model is assumed to represent the real system behavior. Based on this extended availability of “measurement data” in every point of the system, different approaches are established to heuristically assess the suitability of one or more pipes in combination as calibration point(s). These approaches intend to find suitable answers to the question of measurement site selection for this specific case study within a relatively short time and with a reasonable computational effort. As a result, the relevance of the spatial distribution of calibration points is highlighted. Furthermore, particular efficient calibration points are identified and further measurement sites in the underlying network are recommended.@en

The hydraulic verification of combined sewer systems as well as the assessment of combined sewer overflows (CSOs) can be conducted using a hydrodynamic model. Unfortunately, long-term simulations with hydrodynamic models for the assessment of CSOs can cause unacceptably long computation times. Using only a series of storm events instead of a precipitation continuum can reduce this time and enables parallel simulation of single storm events. We introduce a method to select this series of storm events. The method’s parameters have been optimized to replicate the overflow volume of the continuous simulation and to minimize the overall computation time. This optimization revealed a generally applicable parameter set that results in series of storm events that are shorter than the precipitation continuum by 86.2–95.2% for the investigated cases. Additionally, the deviation of overflow volume between continuous simulation and series simulation ranges between only 0.1% and 4.1%.@en