Multiple methods of spectral analysis with applications to the Florida Current

Kevin Patrick McKone


Analysis is presented on observed and simulated volume transport time series in the Intra-Americas Sea (IAS). The simulated transport time series were produced by NLOM, the Naval Research Laboratory (NRL) Layered Ocean Model. An observed transport in the Florida Straits is compared with a NLOM simulated transport in the Florida Straits. A good comparison using spectral analysis between the two transports lead to further investigations of simulated transport time series in the IAS. Seasonal and intra-annual signals were seen in the simulated volume transports using multiple methods of spectral analysis. A barotropic Rossby wave model was used to describe the phase difference seen between the seasonal signal of different simulated transports in the Caribbean. One-point correlation maps suggest regional winds might play a role in forcing the seasonal signal in the Florida Current. In oceanology, spectra of discrete time series are difficult to estimate because of their shortness. This problem of relatively short, discrete time series is common in all of geophysics. Estimates of these spectra using different methods of spectral analysis can give surprisingly different results. Therefore, it is often difficult to interpret whether a signal is real or a manifestation of the spectral method being used. The use of multiple methods of spectral analysis can help with this difficulty. The importance of using multiple methods of spectral analysis is that a signal that is real should be impervious to the method being used. Four methods of spectral analysis are applied to volume transport time series from NLOM, along with an observed time series of volume transport in the Florida Current at 27° N . These methods are the periodogram, Welch's overlapping method, multitaper method of spectral estimation, MTM and autoregressive spectral analysis. This analysis is performed on observed transport data to produce a metric spectrum that is compared to the spectrum from the NLOM output. With good model-data comparison in the Florida Current, further analysis using univariate and bivariate spectral analysis of volume transports from NLOM in the IAS and Florida Current was done. From this analysis, a seasonal signal which is seen in the Florida Current, is also seen in some of the volume transports in the IAS. One interesting feature seen in the NLOM output in the IAS, was that the seasonal signal was not present in all transports. A strong nine month signal was seen in some of the IAS transports. Loop Current eddy shedding in the Gulf of Mexico is one hypothesis for this strong nine month signal. Bandpass filtering is used to isolate the seasonal signal in the simulated volume transports. Cross spectral analysis techniques are then applied to transport pairs. An estimate of the velocity of signal propagation between the transport pairs for high coherence signals was found using a lead/lag model. The results of which were interpreted in terms of barotropic Rossby waves. One-point correlation maps were produced to determine if seasonal wind stress was a possible source of forcing for the seasonal signal seen in the Florida Current and some IAS transports. High correlation of wind stress near the Florida Current suggest that some of the Florida Current's seasonal signal is forced by regional winds.