<link rel="stylesheet" href="styles.cf390c6bfbe7555f.css">

A SPECTRAL DATA ASSIMILATION SYSTEM FOR WAVEWATCH III USING THE FOUR-DIMENSIONAL LOCAL ENSEMBLE TRANSFORM KALMAN FILTER

Loading...
Thumbnail Image

Files

Publication or External Link

External Link to Data Files

Date

Advisor

Ide, Kayo

Citation

Abstract

Ocean waves are created from energy passing through water with severe weather, such ashurricanes, creating hazardous waves. These hazardous waves pose risks to offshore operations, maritime operations, and coastal communities. Accurate wave forecasting provides critical information to these fields. Conventional wave data assimilation (DA) systems typically update mean wave parameters such as significant wave height, which limits their ability to represent the full wave spectrum and enforce physical consistency. This dissertation develops and evaluates a spectral wave data assimilation system for WAVEWATCH III based on the Local Ensemble Transform Kalman Filter. The LETKF analysis directly updates weights of the ensemble members, which makes it a versatile analysis for updating different analysis control variables. This proposed system introduces a logarithmically transformed spectral analysis control variable to weakly constrain the non-negative model state from WAVEWATCH III. Then, the 3D and 4DLETKF are compared to determine the impact of incorporating the model trajectory into the analysis. Lastly, shortest path and shortest distance localization are compared to determine the impact of observations across landmasses. Together, these factors result in a unique wave DA system that focuses on improving the physical characteristics of the wave spectrum. Observing system simulation experiments (OSSEs) were performed using significant wave height observations from buoys and satellite altimeters. Results show that spectrally-averaged root mean square error (RMSE) is similar across the different tested analysis control variables at the time of analysis. However, the analysis in significant wave height space reveals more meaningful corrections in RMSE and bias over a 24 hr forecast period. In significant wave height forecasts, changing the analysis control variable leads to significantly different forecasts. Comparisons between 3D and 4D-LETKF configurations demonstrate that 4D DA achieves persistent improvements to significant wave height forecast skill across each of the analysis control variables. The comparison between shortest path and shortest distance localization shows minimal aggregate differences because of the limited regions affected by the change in localization method and because the model, WAVEWATCH III, is wind-driven. These findings demonstrate the importance of the analysis control variable and wave evolution for wave DA systems.

Notes

Rights

Endorsement

Review

Supplemented By

Referenced By