Unraveling the complexities of urban fluvial flood hydraulics through AI

Abstract

As urbanization increases across the globe, urban flooding is an ever-pressing concern. Urban fluvial systems are highly complex, depending on a myriad of interacting variables. Numerous hydraulic models are available for analyzing urban flooding; however, meeting the demand of high spatial extension and finer discretization and solving the physics-based numerical equations are computationally expensive. Computational efforts increase drastically with an increase in model dimension and resolution, preventing current solutions from fully realizing the data revolution. In this research, we demonstrate the effectiveness of artificial intelligence (AI), in particular, machine learning (ML) methods including the emerging deep learning (DL) to quantify urban flooding considering the lower part of Darby Creek, PA, USA. Training datasets comprise multiple geographic and urban hydraulic features (e.g., coordinates, elevation, water depth, flooded locations, discharge, average slope, and the impervious area within the contributing region, downstream distance from stormwater outfalls and dams). ML Classifiers such as logistic regression (LR), decision tree (DT), support vector machine (SVM), and K-nearest neighbors (KNN) are used to identify the flooded locations. A Deep neural network (DNN)-based regression model is used to quantify the water depth. The values of the evaluation matrices indicate satisfactory performance both for the classifiers and DNN model (F-1 scores- 0.975, 0.991, 0.892, and 0.855 for binary classifiers; root mean squared error- 0.027 for DNN regression). In addition, the blocked K-folds Cross Validation (CV) of ML classifiers in detecting flooded locations showed satisfactory performance with the average accuracy of 0.899, which validates the models to generalize to the unseen area. This approach is a significant step towards resolving the complexities of urban fluvial flooding with a large multi-dimensional dataset in a highly computationally efficient manner.

Spatiotemporal Variability of Channel Roughness and its Substantial Impacts on Flood Modeling Errors

“…A set of ML‐based regressors was chosen for the predictive analysis (Figure S5 in Supporting Information S1), including Random Forest (RF), Multi‐layer Perceptron (MLP), Extreme Gradient Boosting (XGB), and K‐Nearest Neighbors (KNN). RF is an ensemble learning method for regression operated by constructing a collection of multiple decision trees when training the model (Mehedi et al., 2022). An MLP is a fully connected type of feed‐forward neural network (Gaudart et al., 2004).…”

“…The relative feature importance of the predictors was studied by analyzing the Permutation Feature Importance (PFI) technique in the computational domain (Mehedi et al., 2022; Mi et al., 2021). In PFI, the impact of shuffling the values of a feature (e.g., NDVI) over the target variable (e.g., roughness coefficient) is quantified to observe the response in the output variables due to the change in the input variables.…”

“…In meandering channels, the irregularity of channel geometry and sinuosity (Figure S1 in Supporting Information S1) can also play a role in determining n (James, 1994; Naghavi et al., 2023). Nevertheless, past studies have mostly focused on a certain type of stream or a limited number of sites (Bhusal et al., 2022; Mehedi et al., 2022; Mohanta et al., 2018; Roushangar & Shahnazi, 2021; Yang et al., 2021), and a large dataset is lacking to enable large‐scale Machine Learning (ML) studies of n that would provide holistic and systematic perspectives. Since all these channel properties covary, we hypothesize that catchment area or discharge as integrative signals, along with local environmental predictors, could present a pathway toward better determination of n at large scales.…”

Predicting Suicidal Ideation, Planning, and Attempts among the Adolescent Population of the United States

“…The models selected were Decision Tree (DT), Random Forest (RF), Support Vector Machines (SVM), Logistic Regression (LR), and eXtreme Gradient Boosting (XGB). The rationale behind choosing these specific models lies in their demonstrated efficacy in predicting binary targets [34][35][36].…”