Variable seasonal and subseasonal oscillations in sea level anomaly data and their impact on sea level prediction accuracy W. Kosek 1,2, T. Niedzielski 3,2, W. Popiński 4, M. Zbylut 1,A. Wnęk 1 1) Environmental Engineering and Land Surveying Department, Agriculture University of Krakow, Poland 2) Space Research Centre, Polish Academy of Sciences, Warsaw, Poland 3) Institute of Geography and Regional Development, University of Wrocław, Poland. 4) Central Statistical Office of Poland, Warsaw, Poland VIII Hotine-Marussi Symposium, VI 2013, Rome, Italy
DATA Gridded Sea Level Anomaly (SLA) data 1 o ×1 o for φ, Λ obtained from the Archiving, Validation and Interpretation of Satellite Oceanographic (AVISO) data in with one week sampling interval. Data are produced from observations carried out by the satellites TOPEX/Poseidon, ERS, Jason-1, Jason-2 and Envisat.
Time-frequency analysis of sea level anomaly data where: - SLA time series - broadband oscillation with central frequency ω -geographic latitude and longitude - half of the bandwidth - parabolic transmittance function - sampling interval of the SLA data T – mean period of broadband oscillation
Mean amplitude spectrum where: n – number of SLA data, k – the number of points to be drop at the beginning and at the and of filtered time series due to filter errors, T=Δt/ω - the mean period of broadband oscillation. Time variable amplitude spectrum
Mean amplitude spectrum of the whole ocean, northern and southern hemispheres ω 4ω4ω 3ω3ω 2ω2ω
Kuroshio CurrentGulf Current Antarctic Circumpolar Current
Time variable amplitude spectrum
PROGNOCEAN - Near real time system and service for sea level prediction Version This is Prognocean 1.0.beta - solely experimental solution. Testing phase of Prognocean 1.0.beta is in progress. The system may be switched off without publishing any notice. See Disclaimer below for terms of use. Objective The near real time system and service for sea level prediction, known also as Prognocean is designed, implemented and based at the University of Wroclaw (Poland). The initiative is supported by the Foundation for Polish Science through the European Regional Development Fund and the Innovative Economy Programme and aims to compute predictions of Sea Level Anomaly (SLA) maps in near real time. Initially, daily data are predicted and lead time does not exceed two weeks. Solutions for longer lead times will be available later, when time series sampled every 7 days are modelled. Along with predictions, Root Mean Squared Error (RMSE) of predictions is computed in near real time so that the users are able to evaluate the performance of the system and service.
The mean prediction error of sea level anomaly data for 2 weeks in the future computed using the combination of the polynomial harmonic extrapolation model and 1) autoregressive prediction, 2) threshold autoregressive prediction. Polynomial harmonic + autoregressive Polynomial harmonic + threshold autoregressive
The mean prediction error of the SLA data for 2 weeks in the future and the mean amplitude of the annual oscillation
Computation of phase variations in real-valued time series Combination of complex demodulation and the Fourier Transform Low Pass Filter - CD+FTLPF Combination of the FTBPF and Hilbert transform - FTBPF+HT
Combination of complex demodulation and the Fourier transform low pass filter (CD+FTLPF) 1. Multiplication of the time series by complex-valued harmonic with frequency : 2. Filtration of the transformed signal using FTLPF of complex-valued time series: 3. Computation of instantaneous phases:. - transmittance function, λ - window halfwidth
Combination of the FTBPF and Hilbert transform (FTBPF+HT) 1. Computation of the oscillation with central frequency by the FTBPF: λ - window halfwidth 2. Forming of the complex-valued series using the Hilbert transform of the filtered oscillation: or 3. Computation of instantaneous phases:. - transmittance function
Conclusions The FTBPF analysis of sea level anomaly data reveals that the annual oscillation has a broadband character. This creates oscillations with frequencies being an integer multiplicity of the annual frequency. The amplitude maxima of all these shorter period oscillations are located almost in geographic regions of the annual oscillation amplitude maxima. The mean prediction errors for 2 weeks in the future of sea level anomaly data are usually big in geographic regions of annual oscillation amplitude maxima. The increase of the prediction errors of sea level anomaly data is mostly caused by variable phases and amplitudes of the broadband annual oscillation.