Open AccessPrediction skill of nearshore profile evolution models
Author(s) -
Plant Nathaniel G.,
Holland K. Todd,
Puleo Jack A.,
Gallagher Edith L.
Publication year - 2004
Publication title -
journal of geophysical research: oceans
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2003jc001995
Subject(s) - bathymetry , forecast skill , hindcast , sediment transport , parameterized complexity , sampling (signal processing) , statistics , significant wave height , parametrization (atmospheric modeling) , predictive modelling , geology , environmental science , mathematics , sediment , computer science , climatology , physics , wind wave , geomorphology , oceanography , filter (signal processing) , combinatorics , radiative transfer , quantum mechanics , computer vision
The hindcast prediction skill of a beach profile evolution model has been evaluated using bathymetric observations obtained at Duck, North Carolina. The model included coupling and feedback between evolving bathymetry, wave‐averaged hydrodynamics, and parameterized cross‐shore sediment transport. Statistically optimum predictions were obtained by tuning free model parameters using rigorous inverse methods. When compared to persistence predictions (i.e., substitution of the initial, observed profile at all prediction times), significant prediction skill was found for prediction periods longer than 3 days and shorter than 17 days. The average skill (defined as 1 minus the ratio of prediction to persistence error variances) for a typical 5‐day prediction was 0.4, and the maximum skill was 0.8. In contrast to several previously published comparisons with field data, the present approach yielded significant predictive skill during conditions dominated by onshore sediment transport. To make significant predictions, it was necessary to vary model parameters values, which showed a dependence on wave conditions at the seaward boundary of the model domain. Finally, interpretation of the estimated parameter values and prediction skills suggests that (1) errors in existing hydrodynamic models contributed significantly to profile prediction error; (2) a transport model using only second‐order statistics of hydrodynamics was as accurate as (and sometimes more accurate than) a typically used energetics model; (3) existing sediment transport models contained significant errors in the formulations describing downslope and bed load transport; and (4) when driving models with observed flow fields, sampling limitations contribute to profile prediction error.