AI/ML for Hydrologic Modeling

Forecasting Multi-Step-Ahead Street-Scale Nuisance Flooding using a seq2seq LSTM Surrogate Model for Real-Time Application in a Coastal-Urban City

Binata Roy, Jonathan L. Goodal, Diana McSpadden, Steven Goldenberg, Malachi Schram 

Journal of Hydrology

August 2025

https://doi.org/10.1016/j.jhydrol.2025.132697

In coastal-urban cities facing an elevated risk of nuisance flooding (by rain and tide) due to increased heavy rainfall, sea level rise, urbanization, and aging drainage systems, real-time flood forecasting at the street-scale can provide useful information to transportation decision-makers. Physics-Based Models (PBMs) that offer high accuracy come with high computational runtimes and costs that limit their application for real-time flood forecasting. To address this challenge, Machine Learning (ML) surrogate models trained from PBMs have been proposed to provide street-scale flood forecasts. Previous related studies have focused on using Long Short-Term Memory (LSTM) architectures to model hourly flood depth on streets. While LSTM models can capture input sequences effectively, they fall short in accurately preserving output sequences, limiting their suitability for multi-step-ahead forecasts. The seq2seq LSTM architecture offers a key advantage here by capturing the full sequence of input–output, making it potentially more suitable for multi-step-ahead flood forecasts compared to traditional LSTM models. However, seq2seq LSTM has not been tested for street-scale flood forecasting, particularly for rapidly fluctuating nuisance flooding events which require special attention to its temporal sequences. Hence, in this study, we applied the seq2seq LSTM model to explore multi-step-ahead street-scale nuisance flooding and compared its results to the traditional LSTM model as a benchmark model. LSTM and seq2seq LSTM surrogate models were applied to 22 flood-prone streets in Norfolk, Virginia, as a case study with a 4-hr (short-term) and 8-hr (long-term) lead time. The models were trained with environmental (rainfall and tide) and topographic (elevation, Topographic Wetness Index, and Depth-To-Water) features along with PBM-derived water depths for different storm events. The results demonstrated satisfactory performance of both LSTM and seq2seq LSTM surrogate models throughout the forecast period compared to the PBM. However, the seq2seq LSTM showed lower Mean Absolute Error (MAE)/ Root Mean Square Error (RMSE) and higher Nash–Sutcliffe Efficiency (NSE)/ correlation than the LSTM across most lead times, particularly for long-term forecasting due to its supremacy in handling both input–output sequences together, which is missing in the traditional LSTM. For example, in the long-term, the average RMSE ranges were 0.0268–0.0373 m for LSTM and 0.0226–0.0319 m for seq2seq LSTM, while in the short-term, they were 0.0263–0.0293 m and 0.0261–0.0283 m, respectively. Additionally, while both models exhibited similar performance in distinguishing flooded and non-flooded streets for flood depth ≥ 0.1 m, the seq2seq LSTM model demonstrated superior performance for higher flood depths (such as ≥ 0.2 m and ≥ 0.3 m). Once trained, inference took only 0.09 to 0.11 s (short-term) and 0.30 to 0.35 s (long-term) per storm event for the 22 streets, making the application highly suitable for real-time decision-making during nuisance flood events.

FloodForecaster: A domain-adaptive geometry-informed neural operator framework for rapid flood forecasting

Mehdi Taghizadeh, Zanko Zandsalimi, Mohammad Amin Nabian, Jonathan L Goodall, Negin Alemazkoor

Journal of Hydrology

January 2026

https://doi.org/10.1016/j.jhydrol.2025.134512 

Abstract
High-resolution flood forecasting over large domains requires models that are both computationally efficient and generalizable. While traditional physical models are too slow, existing machine learning methods often struggle with complex geometries or adapting to new regions. This paper introduces FloodForecaster, a novel deep learning framework built upon a time-dependent geometry-informed neural operator (GINO). This core predictive model synergizes graph neural operators to process irregular terrain with Fourier neural operators to efficiently capture global flood dynamics, demonstrating superior accuracy and stability over state-of-the-art GNN baselines. To ensure transferability, the framework integrates a domain adaptation technique using a gradient reversal layer, which encourages the model to learn domain-invariant physical features. Evaluated on a multi-segment case study, this adaptation technique proves more effective than conventionally fine-tuning the base GINO model. While standard fine-tuning leads to catastrophic forgetting of source-domain knowledge, our approach successfully preserves the model’s original expertise while also effectively learning the hydraulics of new river segments. This successful adaptation is also highly data-efficient, achieving strong performance with as few as 10 training simulations from a new domain and reducing prediction error by approximately 75% compared to fine-tuning. FloodForecaster thus offers a significant advancement toward developing transferable, real-time flood forecasting systems.

Toward reproducible and interoperable environmental modeling: Integration of HydroShare with server-side methods for exposing large-extent spatial datasets to models

Young-Don Choi, Iman Maghami, Jonathan L Goodall, Lawrence Band, Ayman Nassar, Laurence Lin, Linnea Saby, Zhiyu Li, Shaowen Wang, Chris Calloway, Hong Yi, Martin Seul, Daniel P Ames, David G Tarboton

Environmental Modelling & Software

January 2025

https://doi.org/10.1016/j.envsoft.2024.106239 

Abtract

The principal objective of this project is to advance integrated modeling as an approach for building next generation watershed models to support environmental management. Integrated modeling is a systems-based approach for constructing models through the use of predefined modeling components. Each component represents some spatially-explicit system process, and the modeler defines how these components are coupled to simulate system response. Because components can be developed and maintained by different groups, yet can still be coupled within a modeling system, integrated modeling offers a transformative approach for constructing next generation, community-supported environmental models. This project will address a fundamental question in integrated modeling, that is the transfer of boundary conditions between spatially and temporally misaligned components, and produce results that will be used in future efforts in integrated modeling. This project will focus on the following research tasks to advance the science and adoption of integrated modeling: (1) to investigate scaling issues in integrated modeling, in particular algorithms for transferring values between coupled process-level components that operate on different spatial or temporal scales; (2) to prototype a set of process-level components for an integrated modeling system and apply these components to improve understanding of the performance and accuracy of integrated modeling approaches; and (3) to educate a new generation of environmental modelers in integrated modeling techniques to foster a community of integrated watershed modelers. If successful, this project will provide important guidance to integrated watershed modeling activities in the U.S. and abroad.