Using Machine Learning Methods for Urban Flood Modeling in New York City

Flooding in urban areas has become a growing concern as cities expand and the effects of climate change intensify. This issue is particularly pronounced in densely populated cities like New York City (NYC), where large expanses of impervious surfaces, such as concrete and asphalt, impede water absorption, leading to rapid runoff and heightened flood risks. NYC’s complex urban environment, with its intricate infrastructure and high population density, is especially vulnerable to these flood events, demanding sophisticated methods for prediction and management.

In response to this challenge, our study delves into the key factors driving urban flooding in NYC, applying machine learning techniques to develop a predictive model that accurately simulates flood events. By leveraging a wide range of data, including precipitation records, Digital Elevation Models (DEM), 311 complaints, and real-time flood measurements from FloodNet sensors, we aim to provide a comprehensive view of the factors contributing to NYC's flood risks. Machine learning models such as Long Short-Term Memory (LSTM), Support Vector Machine (SVM), and Random Forest (RF) were applied to enhance the accuracy of flood simulations and improve our understanding of flood dynamics within the city.

The integration of 311 complaint data provided valuable insight into the specific times and locations of flood incidents, establishing a spatial and temporal map of urban flooding. Additionally, FloodNet sensors offered precise measurements of flood depths, enabling a more accurate calibration of our models. Through our analysis, we identified street slopes and clogged catch basins as significant contributors to flooding, beyond the expected influence of precipitation levels. The LSTM model, in particular, proved highly effective at predicting flood depths by capturing the temporal relationships between rainfall patterns and sensor data.

These machine learning models offer a powerful tool for forecasting flood scenarios under various conditions, providing urban planners and emergency response teams with the ability to anticipate and mitigate flood risks. By employing these models, cities like NYC can take proactive measures to minimize the impact of flooding, improving overall resilience to extreme weather events. Beyond immediate applications in flood management, this research contributes valuable knowledge to the broader domains of urban hydrology and disaster preparedness, offering novel insights into the drivers of urban flooding.

This study underscores the significant potential of machine learning in advancing the field of urban flood risk management. As climate change exacerbates the frequency and severity of urban flood events, the integration of AI into urban planning and water management systems will be essential for enhancing resilience. The framework established in this study not only provides a valuable tool for flood prediction in New York City but also serves as a transferable model that can be adapted to other urban environments facing similar hydrological challenges globally. This research contributes to the ongoing efforts to build more sustainable and adaptive cities in the face of growing environmental uncertainties.