Agnieszka Indiana Olbert

Coastal cities located in estuaries often face significant risks from both riverine and tidal flooding due to their low-lying locations. Accurately predicting flood water levels in a complex urban environment is challenging because multiple factors interact – upstream river flow, heavy rainfall, tides and storm surges all play a role. This research explores a new approach to compound coastal-fluvial flood forecasting using deep learning: a combination of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. The goal is to forecast water levels in an estuary as well as flood water levels over an urban floodplain with a lead time of 1 to 33 hours. The model is specifically designed to effectively combine various hydrological, coastal and meteorological data sources.Coastal city of Cork, the second-largest city in Ireland and one that is frequently affected by compound coastal-fluvial flooding is used as a case study. In this research, we use high-resolution precipitation forecasts from the NWP model operated by Met Eireann, river flow data from the River Lee catchment, and tidal/surge information from the MSN_Flood hydrodynamic model of the Cork Harbour.Our proposed CNN-LSTM architecture combines the strengths of these deep learning methods. The CNN component efficiently identifies important spatial patterns from the rainfall forecasts and model outputs that suggest potential flooding. The LSTM component then captures how water levels change over time, enabling the model to learn the evolution of flood conditions.Historical flood events in Cork City form the basis for training our deep learning model. This historical data, combined with real-time data streams from NWP, river gauge records and hydrodynamic model, allows the CNN-LSTM network to learn the intricate relationships between upstream riverine conditions and downstream sea water levels. This system has the potential to significantly improve flood preparedness and response in Cork City, enabling earlier warnings and proactive measures to protect communities from flood damage. Additional data, such as soil moisture and land cover information are also used to enhance the model’s accuracy and robustness.

Climate change is anticipated to alter the frequency and intensity of climate extremes, with profound implications for vulnerable regions such as the Caspian Sea. This study investigates future changes in extreme precipitation events over the Caspian Sea, a region particularly sensitive to climatic shifts due to its geographical position and isolation from major oceans. Using bias-corrected precipitation data from the General Circulation Models- Coupled Model Intercomparison Project Phase 6 and historical reference data from ERA5, this research applies advanced extreme value analysis to examine trends under two climate scenarios. The results reveal an increase in average precipitation from the historical period to the future period (1980-2100), highlighting the region’s vulnerability to climate change. Furthermore, the analysis projects a rise in the frequency and magnitude of extreme precipitation events, with more intense droughts and floods expected to emerge. Extreme value distributions fitted to the precipitation data confirm these findings, showing higher tail probabilities for extreme events under future scenarios. These changes underscore the need for climate adaptation strategies to mitigate potential socio-economic and ecological impacts. By providing a comprehensive evaluation of precipitation extremes and their future trends, this study contributes valuable insights into the expected hydrological changes in the Caspian Sea region, which can inform water resource management, disaster risk reduction, and climate policy.

Floods are one of the most pervasive and costly natural disasters that result in notable socio-economic impacts. Flood forecasting systems are crucial for effective flood risk management, hazard assessment, and decision makings. However, accurate modelling and forecasting of flood events is difficult due to uncertainties in forecasting of meteorological and hydrological input variables. Forecasting is additionally complicated when floods are compound. Conventional flood forecasts typically consider only one driver at a time, whether it is ocean, fluvial or pluvial without considering the often compound nature of flood events. In this study, a novel two-step approach to forecasting of compound coastal-fluvial floods was developed. The two-model hydrodynamic-machine learning forecasting system combines and links the hydrodynamic model with machine learning (ML) model, where the hydrodynamic outputs representing various probability flood events are used to train the ML algorithm in order to predict the inundation patterns resulting from a combination of coastal and fluvial flood drivers occurring simultaneously. In step one, the MSN_Flood model was used to simulate the compound coastal-fluvial flooding over coastal city of Cork, Ireland. In step two, the hydrodynamic model outputs such as time variable water depths across urban floodsplains were used to train the ML model. In total 7 ML models (Support Vector Regression (SVR), Support Vector Machine (SVM), Radial Basis Function (RBF), Linear Regression (LR), Gaussian Process Regression (GPR), Decision Tree (DT), and Artificial Neural Network (ANN)) were applied in this study. Results show that the ML models, when trained on accurate hydrodynamic model outputs, can provide reliable estimates of flood inundation and associated water depths. Three ML models: RBF, ANN and DT show particularly strong performance and are found suitable for flood forecasting. In overall, the novel coupled hydrodynamic-ML system can be successfully used for flood forecasting.

Flood early warning systems are vital for preventing flood damages and for reducing disaster risks. Such systems are particularly important for forecasting compound events where multiple, often dependent flood drivers co-occur and interact. In this research an early warning system for prediction of coastal-fluvial floods is developed to provide a robust, cost-effective and time-efficient framework for management of flood risks and impacts. This three-step method combines a cascade of three linked models: (1) statistical model that determines probabilities of multiple-driver flood events, (2) hydrodynamic model forced by outputs from the statistical model, and finally (3) machine learning (ML) model that uses hydrodynamic outputs from various probability flood events to train the ML algorithm in order to predict the spatially and temporarily variable inundation patterns resulting from a combination of coastal and fluvial flood drivers occurring simultaneously.The method has been utilized for the case of Cork City, located in the south-west of Ireland, which has a long history of fluvial-coastal flooding. The Lee River channelling through the city centre may generate a substantial flood when the downstream river flow draining to the estuary coincides with the sea water propagating upstream on a flood tide. For this hydrological domain the statistical model employs the univariate extreme values analysis and copula functions to calculate joint probabilities of river discharges and sea water levels (astronomical tides and surge residuals) occurring simultaneously. The return levels for these two components along a return level curve produced by the copula function are used to generate synthetic timeseries, which serve as water level boundary conditions for a hydrodynamic flood model. The multi-scale nested flood model (MSN_Flood) was configured for Cork City at 2m resolution to simulate an unsteady, non-uniform flow in the Lee River and a flood wave propagation over urban floodplains. The ensemble hydrodynamic model outputs are ultimately used to train and test a range machine learning models for prediction of flood extents and water depths. In total, 23 machine learning algorithms including: Artificial Neural Network, Decision Tree, Gaussian Process Regression, Linear Regression, Radial Basis Function, Support Vector Machine, and Support Vector Regression were employed to confirm that the ML algorithm can be used successfully to predict the flood inundation depths over urban floodplains for a given set of compound flood drivers. Here, the limited flood conditioning factors taken into account to analyse floods are the upstream flood hydrographs and downstream sea water level timeseries. To evaluate model performance, different statistical skill scores were computed. Results indicated that in most pixels, the Gaussian Process Regression model performs better than the other models.The main contribution of this research is to demonstrate the ML models can be used in early warning systems for flood prediction and to give insight into the most suitable models in terms of robustness, accuracy, effectiveness, and speed. The findings demonstrate that ML models do help in flood water propagation mapping and assessment of flood risk under various compound flood scenarios.