Scientific Computing Seminar

Date:

Friday, June 24, 2005

Time:

1:00pm-2:00pm

Location:

50A-5132

Seminar Speaker:

Larry Breaker
Moss Landing Marine Laboratories
http://www.mlml.calstate.edu/?id=114

Title:

Linear and Nonlinear Analysis of an 85-Year Time Series of Sea Surface Temperature in Monterey Bay

Abstract:

Daily sea surface temperatures (SSTs) have been acquired in southern Monterey Bay at the Hopkins Marine Station since 1919. This time series is one of the longest oceanographic records along the west coast of North America. The record in its original form was nonhomogeneous, contained outliers and numerous gaps, and the observations were often collected at different times of day. The record has been reconstructed and some of the steps in the reconstruction process are initially described. The reconstructed record has subsequently been examined for nonstationarity, nonlinearity, spectral content, long-term trends, and change points. A new test for stationarity indicates that the record is stationary for frequencies up to ~ 8 cycles per year, but nonstationary at higher frequencies. Using moving windows together with higher statistical moments, it appears that certain events such as warm water intrusions and El Nino episodes contribute to nonlinearity. The annual cycle together with a strong semiannual component in the power spectrum suggest a harmonic relationship between them. Bispectra, singular spectra, wavelet spectra, and Hilbert spectra confirm this relationship. These results also imply the dangers in applying linear analysis techniques to data that are essentially nonlinear. There is a long-term warming trend of ~+0.01C/year, based on a linear, least- squares fit to the record, a result that implies a gradual and continuous warming process. However, a CUSUM of the Hopkins record indicates rather clearly that warming in Monterey Bay has occurred in steps, as indicated by change points in the CUSUM plot. The first step occurred in 1929, and a second in 1976, and there may be others. Finally, we conclude that the value of the Hopkins record is primarily due to two factors, first, its length and second, its location.

Sponsor of Seminar:

Yu-Heng Tseng

Scientific Computing

Contact Esmond G. Ng EGNg@lbl.gov