Paige A. Hovenga | Geology Reports

USGS2022

Combining process-based and data-driven approaches to forecast beach and dune change

Producing accurate hindcasts and forecasts with coupled models is challenging due to complex parameterizations that are difficult to ground in observational data. We present a calibration workflow that utilizes a series of machine learning algorithms paired with Windsurf, a coupled beach-dune model (Aeolis, the Coastal Dune Model, and XBeach), to produce hindcasts and forecasts of morphologic change along Bogue Banks, North Carolina. Neural networks paired with genetic algorithms allow us to fine tune calibration parameters for the hindcast, and then a long short-term memory neural network, trained on the hindcast, produces a 4-year forecast. We compare our hindcasts to observations from 2016 to 2017 and find they successfully reproduce observed modes of dune and beach change except for seaward growth of the dune face. We compare our forecasts to observations from 2016 to 2020 and find that they produce reasonably accurate predictions of dune change except when there are significant instances of erosion during the forecast period.

North Carolina↗

USGS2021

Labeling poststorm coastal imagery for machine learning: Measurement of interrater agreement

Classifying images using supervised machine learning (ML) relies on labeled training data—classes or text descriptions, for example, associated with each image. Data-driven models are only as good as the data used for training, and this points to the importance of high-quality labeled data for developing a ML model that has predictive skill. Labeling data is typically a time-consuming, manual process. Here, we investigate the process of labeling data, with a specific focus on coastal aerial imagery captured in the wake of hurricanes that affected the Atlantic and Gulf Coasts of the United States. The imagery data set is a rich observational record of storm impacts and coastal change, but the imagery requires labeling to render that information accessible. We created an online interface that served labelers a stream of images and a fixed set of questions. A total of 1,600 images were labeled by at least two or as many as seven coastal scientists. We used the resulting data set to investigate interrater agreement: the extent to which labelers labeled each image similarly. Interrater agreement scores, assessed with percent agreement and Krippendorff's alpha, are higher when the questions posed to labelers are relatively simple, when the labelers are provided with a user manual, and when images are smaller. Experiments in interrater agreement point toward the benefit of multiple labelers for understanding the uncertainty in labeling data for machine learning research.

Earth and Space Science↗