Insufficient public and municipal investment represent a major challenge for the long term management of urban drainage systems. Utilities are challenged to develop efficient rehabilitation strategies in order to maintain the level of service. Closed-circuit television (CCTV) inspection is used since the 1980’s as industry standard for sewer investigation system and structural performance evaluation. Due to budget restrictions, inspection rates are generally low and municipalities tend to inspect only a small part of their network (e.g. in France, less than 5% according to Ahmadi et al., 2014c). Since the definition of rehabilitation strategies is limited by the lack of information about sewer condition and remaining life, deterioration models have been developed to forecast the evolution of the system according to its current and past condition. One of the main factors hampering the uptake of deterioration modelling by utilities is the lack of real scale evidence of the tangible benefits provided. In particular, most utilities are concerned by the minimum amount of CCTV data required and the relevance of using such models on their networks with limited data availability. Finally, most utilities acknowledge the uncertainties in the procedure of sewer condition assessment, mainly due to the subjectivity of the coding operator. There is a strong need to quantify precisely the uncertainty of the sewer condition assessment procedure and its influence on the outcomes of deterioration modelling. The thesis aims at addressing these gaps by assessing the performance of sewer deterioration modelling using a case study with high CCTV data availability and by identifying the influence of CCTV data quality and availability on modelling performance. The study has been performed with a statistical (GompitZ) and a machine learning (Random Forest) deterioration models using the extensive CCTV database of the cities of Braunschweig and Berlin in Germany. Our results show, that at network level, both machine learning and statistical models can simulate with sufficient accuracy the condition distribution of the network, even in case of low data availability. At the pipe level, the machine learning model outperforms the statistical model. Regarding CCTV data uncertainty, our results highlight that the probability to inspect correctly a pipe in poor condition is close to 80-85% and thus the probability to overestimate the (good) condition of the pipe is close to 15-20% (False Negative). The impact of the uncertainties on the prediction of a deterioration model is not negligible. The analysis shows that the required replacement rate to maintain a constant proportion of segments in poor condition is underestimated if the uncertainties are not included in the analysis.

Kanalalterungsmodelle, mit denen sich der Zustand von Abwasserkanälen simulieren lässt, können wertvolle Werkzeuge für die Sanierungsplanung sein. Dennoch werden sie in Deutschland bisher nur von wenigen Kanalnetzbetreibern eingesetzt. Im Rahmen des Forschungsvorhabens SEMA-Berlin wurden verschiedene Modellansätze getestet und hinsichtlich ihrer Prognosequalität bewertet. Für den Modellaufbau wurden die Ergebnisse von mehr als 100 000 TV-Inspektionen sowie Daten zu den individuellen Kanaleigenschaften und Umgebungsfaktoren der Stadt Berlin verwendet. Die Untersuchungen zeigen, dass das statistische Modell GompitZ die Zustandsverteilung des Kanalnetzes mit einer Genauigkeit von 99 % wiedergeben kann. Mit Random Forest, einem Modell des maschinellen Lernens, kann mit einer Trefferquote von 67 % vorhergesagt werden, welcher Kanal sich im schlechten Zustand befindet. Die Ergebnisse können dafür genutzt werden, prioritäre Haltungen für Kanalinspektionen zu identifizieren und Investitionen so zu steuern, dass der Zustand der Kanalisation langfristig erhalten oder sogar verbessert wird.

For ensuring microbial safety, the current European bathing water directive (BWD) (76/160/EEC 2006) demands the implementation of reliable early warning systems for bathing waters, which are known to be subject to short-term pollution. However, the BWD does not provide clearly defined threshold levels above which an early warning system should start warning or informing the population. Statistical regression modelling is a commonly used method for predicting concentrations of fecal indicator bacteria. The present study proposes a methodology for implementing early warning systems based on multivariate regression modelling, which takes into account the probabilistic character of European bathing water legislation for both alert levels and model validation criteria. Our study derives the methodology, demonstrates its implementation based on information and data collected at a river bathing site in Berlin, Germany, and evaluates health impacts as well as methodological aspects in comparison to the current way of long-term classification as outlined in the BWD.

Deterioration models can be successfully deployed only if decision-makers trust the modelling outcomes and are aware of model uncertainties. Our study aims to address this issue by developing a set of clearly understandable metrics to assess the performance of sewer deterioration models from an end-user perspective. The developed metrics are used to benchmark the performance of a statistical model, namely, GompitZ based on survival analysis and Markov-chains, and a machine learning model, namely, Random Forest, an ensemble learning method based on decision trees. The models have been trained with the extensive CCTV dataset of the sewer network of Berlin, Germany (115,258 inspections). At network level, both models give satisfactory outcomes with deviations between predicted and inspected condition distributions below 5%. At pipe level, the statistical model does not perform better than a simple random model, which attributes randomly a condition class to each inspected pipe, whereas the machine learning model provides satisfying performance. 66.7% of the pipes inspected in bad condition have been predicted correctly. The machine learning approach shows a strong potential for supporting operators in the identification of pipes in critical condition for inspection programs whereas the statistical approach is more adapted to support strategic rehabilitation planning.

Closed Circuit Television Inspection is used since decades as industry standard for sewer system inspection and structural performance evaluation. In current practice, inspection data are helpful to support asset management decisions. However, the quality and uncertainty of sewer condition assessment is rarely questioned. This article presents a methodology to determine the probability to underestimate, overestimate or accurately estimate the real condition of a pipe using visual inspection. The approach is based on the analysis of double inspections of the same sewer pipes and has been tested using the extensive data-set of the city of Braunschweig in Germany. Results indicate that the probability to inspect correctly a pipe in poor condition is close to 80%. The probability to overestimate the condition of a pipe in bad condition (false negative) is 20% whereas the probability to underestimate the condition of a pipe in good condition (false positive) is 15%. Finally, sewer condition evaluation can be used to assess the general condition of the network with an excellent accuracy probably because the respective effects of false positive and false negative are buffered. © 2017 Informa UK Limited, trading as Taylor & Francis Group.