Enhancing interpretability and generalizability of deep learning-based emulator in three-dimensional lake hydrodynamics using Koopman operator and transfer learning: Demonstrated on the example of lake Zurich.

Three-dimensional lake hydrodynamic model is a powerful tool widely used to assess hydrological condition changes of lake. However, its computational cost becomes problematic when forecasting the state of large lakes or using high-resolution simulation in small-to-medium size lakes. One possible solution is to employ a data-driven emulator, such as a deep learning (DL) based emulator, to replace the original model for fast computing. However, existing DL-based emulators are often black-box and data-dependent models, causing poor interpretability and generalizability in practical applications. In this study, a data-driven emulator is established using deep neural network (DNN) to replace the original model for fast computing of three-dimensional lake hydrodynamics. Then, the Koopman operator and transfer learning (TL) are employed to enhance the interpretability and generalizability of the emulator. Finally, the generalizability of DL-based emulators is comprehensively analyzed through linear regression and correlation analysis. These methods are tested against an existing hydrodynamic model of Lake Zurich (Switzerland) whose data was provided by an open-source web-based platform called Meteolakes/Alplakes. According to the results, (1) The DLEDMD offers better interpretability than DNN because its Koopman operator reveals the linear structure behind the hydrodynamics; (2) The generalization of the DL-based emulators in three-dimensional lake hydrodynamics are influenced by the similarity between the training and testing data; (3) TL effectively improves the generalizability of the DL-based emulators.

An integrated assessment of drainage system reconstruction based on a drainage network model.

In order to mitigate urban flooding and combined sewer overflows, an integrated assessment method was proposed to identify the optimum reconstruction scheme of a drainage system by considering environment, economy, and society. The integrated assessment framework consisted of the drainage system model establishment, analytic hierarchy process theory, and regret value method. Five drainage system reconstruction schemes for Chaohu city were proposed in this study, and they were evaluated according to nine assessment factors by the integrated assessment method at the initial and future stages. The integrated assessment results show that setting up interceptive equipment for a combined drainage network is the optimal reconstruction scheme at both the initial and future stages of the life cycle. This means that an interceptive combined drainage network is better than a separate drainage network or setting up storage tanks in particular situations from a comprehensive perspective.

Modelling the effects of water diversion and combined sewer overflow on urban inland river quality.

In order to assist and optimize the operation of a clean water diversion project for the medium-sized inland rivers in Chaohu, China, an integrated hydrodynamic and water quality model was used in this study. Sixteen diversion scenarios and five sewage interception scenarios were defined to assess the improvement of water quality parameters including ammonia nitrogen (NH3-N), total phosphorus (TP) and chemical oxygen demand (COD) under different diverted water flows, diverting times, diverting points, diverting routines and sewage interception proportions. An index of pollutant removal rate per unit diverted water flow (PRUWF) was proposed to evaluate the effect of the clean water diversion. Results show that operating conditions played important roles in water quality improvement of medium-sized inland rivers. The optimal clean water diversion was operated under the conditions of a flow rate of 5 m3/s for 48 h with an additional constructed bridge sluice. A global sensitivity analysis using the Latin Hypercube One-Factor-at-a-Time (LH-OAT) method was conducted to distinguish the contributions of various driving forces to inland river water restoration. Results show that sewage interception was more important than diverted water flow and diverting time with respect to water quality improvement, especially for COD.

Anthropogenic Influences of Paved Runoff and Sanitary Sewage on the Dissolved Organic Matter Quality of Wet Weather Overflows: An Excitation-Emission Matrix Parallel Factor Analysis Assessment.

The quality of dissolved organic matter (DOM) in a wet weather overflow (WWF) can be broadly influenced by anthropogenic factors, such as nonpoint sources of paved runoff and point sources of sanitary sewage within the drainage networks. This study focused on the anthropogenic influences of the paved runoff and sanitary sewage on the DOM quality of WWF using excitation-emission matrix parallel factor analysis (EEM-PARAFAC). Results show that (1) EEM-PARAFAC fitted terrestrial humic-like, anthropogenic humic-like, tryptophan-like, and tyrosine-like components can be regarded as indicators to identify the types of sewage overflows and the illicit connection status of drainage systems. (2) A short emission wavelength (em: 302-313 nm) peak of the tyrosine-like component occurred in the reserved sanitary sewage, while a type of longer emission wavelength (em: 321-325 nm) peak came from the sump deposit. These tyrosine-like components were gradually evacuated in the initial phase of the overflow process with the fading of their EEM signals. Fluorescence signal transformations of all the components confirmed the potential ability of EEM-PARAFAC to monitor the dynamic changes of the primary pollutant sources. (3) The input of the newly increased sanitary sewage had a dominant influence on the quality and yield of the WWF DOM.