Prediction of Peak Water Levels during Tropical Cyclones with Deep Learning

Storm-driven flooding is a severe hazard for coastal communities and regions. Computational models can predict the combined effects of tides, winds, and flooding due to tropical cyclones, including in real-time, but requirements for the models’ runtime make it challenging to consider simulations of the full range of storm uncertainty. To address this problem, researchers have developed neural networks, trained on libraries of storm surge simulations, to predict ensembles of coastal flooding in seconds. However, existing neural networks do not consider the interaction between storm surge and astronomical tides nor storms of any duration. Moreover, they are trained on datasets tailored to represent only extreme conditions. We aim to develop a neural network to predict the peak total water levels for any storm at specific locations along the North Carolina coast.

In this research, we implemented a neural network to predict peak values for total water level (tides and storm surge) at multiple stations, considering astronomical tides and storm tracks of any duration as inputs. To create the training library, we simulated 1,813 synthetic tropical cyclones based on historical data in the North Atlantic Ocean, with a specific focus on storms that affect North Carolina. These simulations used a full-physics hydrodynamic model with variable spatial resolution of about 50 m near the coast. The outputs were downscaled to grayscale images with a higher and constant resolution of 15 m, enhancing the flood predictions by considering small-scale topographic features, and then used as training data for the neural network. The many-to-one deep learning model predicts a single peak total water level in time at multiple locations in space using time series of the offshore astronomical tide and track parameters as inputs. We used the model to make probabilistic predictions of peak total water levels for observed and perturbed tracks of several historical storms that affected North Carolina.

We showed that the neural network performed well (with errors ranging from 8 to 43 cm) in predicting peak total water levels at nine locations in North Carolina. We applied the neural network to make probabilistic predictions of peak total water levels for observed and perturbed tracks of historical storms. For each storm, the neural network predicted at nine stations for 101 storm scenarios (the true/historical storm and 100 perturbations) in less than 10 seconds. The performance for the observed historical storms was similar to those obtained in process-based simulations, but with a significant gain in computational runtime.

TA Cuevas López (2024). “Prediction of Peak Water Levels during Tropical Cyclones with Deep Learning,” North Carolina State University.

North Carolina Center for Coastal Algae, People, and Environment

The NC C-CAPE: North Carolina Center for Coastal Algae, People, and Environment will investigate the health effects of various microcystin (MC) mixtures, and it will elucidate links between environmental and climatic drivers and harmful algal bloom (HAB) dynamics, MC congener composition, and toxin contamination in oysters and blue crabs. We will determine and the health effects of MC-mixtures on hepatic toxicity, NAFLD and hepatocellular carcinoma in model systems and humans. The Center’s Community Engagement Core will use the principles of data justice to address HAB exposure and prevention, where community members are experts, rather than objects of research, and have the capacity to conduct critical and systemic inquiry into their own lived experiences. The Administrative Core will provide efficient and effective fiscal and scientific leadership and promote interactions and collaborations across all Center components and beyond. Project 1 will advance our understanding of HAB dynamics and MC contamination in seafood, combining state-of-the-art in situ observing technologies and targeted field surveys. In addition, experimental work will elucidate trophic transfer of toxins in oysters and blue crabs. Project 2 will define how MC mixtures influence mechanisms of liver toxicity and resulting risk of adverse health outcomes in regulatory-relevant mammalian models as well as at-risk human populations. Project 3 will integrate highly diverse data sets and coastal circulation modeling within a probabilistic (Bayesian) modeling framework to elucidate environmental controls on MC distribution in water and seafood and assess MC exposure risk in a changing climate. NC C-CAPE will provide significant insight to guide efforts to implement effective monitoring approaches, inform guideline values for safe consumption of water and seafood, deliver predictive tools to assess emergent and future toxin exposure risk, and will leverage community engagement initiatives to fill data gaps and improve oceans and human health.

A Schnetzer, SM Belcher, BB Cutts, DR Obenour, T Ben-Horin, JC Dietrich, C Hoyo, NG Nelson, R Paerl. “North Carolina Center for Coastal Algae, People, and Environment (NC C-CAPE).National Institutes of Health, National Institute of Environmental Health Sciences, Centers for Oceans and Human Health 4: Impacts of Climate Change on Oceans and Great Lakes, 2024/02/01 to 2029/01/31, $6,913,382 (Dietrich: $467,482).

Conferences: Fall 2023

Poster: Fall 2023 Conferences

Efficiency Gains for Spectral Wave Models in Coupled Frameworks

We propose to modernize a spectral wave model to allow for more flexibility and efficiency within a coupled modeling framework. It is now commonplace for spectral wave models to run alongside other models for circulation and related coastal processes. These models can be coupled within sophisticated frameworks or at the source-code level. However, the widespread use of coupled models has also led to the identification of inefficiencies. Spectral wave models tend to be computationally expensive, and this cost can be amplified when they are coupled with other models. There are known methods for reducing the cost of spectral wave models, such as the nesting of nearshore and regional domains with offshore forcing from other sources, but these methods may have challenges in a coupled framework, such as the need to interpolate between nested domains. The coupling overhead can be (and has been) minimized, but there may be additional methods to further reduce costs without sacrificing predictive accuracy.

Thus, there are remaining research questions related to how to improve the performance of a spectral wave model in a coupled modeling framework. What are the tradeoffs when a spectral wave model is nested nearshore and receives boundary conditions from other sources? Over what period should the spectral wave model simulate as a storm approaches a coast? Can this research lead to guidance or best practices for coupled modeling applications? This project will focus on the Simulating WAves Nearshore (SWAN) model and SWAN+ADCIRC framework, but the project findings will be transferable to other spectral wave models and frameworks. We aim to improve the ability to nest spectral wave models in both space and time, via modernization of boundary conditions and a coupled model controller, and thus improve computational efficiency.

JC Dietrich. “Efficiency gains for spectral wave models in coupled frameworks.” Department of Defense, Broad Agency Announcement, Engineer Research and Development Center, Coastal Hydraulics Laboratory, 2023/09/22 to 2025/09/21, $191,353 (Dietrich: $191,353).