SCIENCE REPORTS

Other TIP pages

TIP homepage

Other TIP 3 pages

TIP 3 Table of contents

Summary

Opening of the meeting

Summary of current conditions in the Tropical Pacific

National reports

Program status reports

Science reports

Recommendations and acknowledgements

ENSO Teleconnections and Impacts in Korea
Won-Tae Kwon, Meteorological Research Institute, Korea
Introduction. Unlike over North America and tropical regions, it seems that Korea does not have a strong signal in climate variation associated with El Nino (Ropelewski and Halpert 1987). Furthermore, the simultaneous and lagged correlation maps of climate elements with the Pacific sea surface temperature (SST) fields lack regions of high correlation coefficients in the eastern tropical Pacific. Research efforts in long-range forecasting at the Meteorological Research Institute, Korea Meteorological Administration (KMA) will be summarized here.

Two research projects for the development of the long-range forecasting models using statistical approaches were initiated in 1993 to provide better objective methods for the operational forecasting. KMA issues a monthly forecast at the beginning of every month, and seasonal forecast at the beginning of each season, mainly using statistical methods. Objectives of the research are to develop multiple regression models for the area averaged monthly temperature and precipitation using predictors from the monthly 500 hPa heights and Pacific sea surface temperature (SST), and analog and anti-analog methods for the same predictands using the predictor fields.

Data Used in the Long Range Forecast Research. As a first approximation, predictands of monthly mean temperature and precipitation are averaged over fifteen stations with 30 years of data. The 500 hPa heights of NMC (U.S. National Meteorological Center) octagonal grid and JMA (Japan Meteorological Agency) monthly mean SST are used as predictors. The 500 hPa heights are averaged for each month and interpolated for 468 grid points from the original 1977 grid points; and the monthly heights are normalized by subtracting means for twelve calendar months and dividing by standard deviations. The SST fields are then interpolated into a 4 deg latitude-longitude grid from the original 2 deg latitude-longitude grid over the Pacific domain, from 100E to 70W and 40S to 60N and a similar normalization procedure has been applied.

Development of the multi-regression models. Simultaneous and lagged (up to 12-month lag) correlation coefficients of temperature and precipitation with the Northern Hemisphere 500 hPa heights and Pacific SST fields are computed for the 30-year period. Areas showing correlation coefficients greater than 0.5 (statistically significant using a two-tailed test) are chosen for further analysis. For the next step, we divide the 30 years into three ten-year periods and find correlation coefficients for each period; points showing fluctuating coefficients are discarded. Finally, predictors for the monthly temperature and precipitation are selected from the 500 hPa heights and the Pacific SST fields for the construction of multiple regression models. We experimented in making forecasts using these multiple regression models with minor success.

Development of the analog and anti-analog method. EOF analyses have been carried out for the 500 hPa and 100 hPa heights and the Pacific SST fields. EOF patterns for the monthly and 3-month periods are analyzed and the first ten EOF patterns are selected for the computation of analog and anti-analog years for each month. First, we experimented with this analog and anti-analog method for the 500 hPa heights. Since some EOF temporal-coefficient time series have better correlation with the predictands, the simultaneous and lagged correlation coefficients of temperature and precipitation with the EOF temporal coefficients are being analyzed in order to define the climate state vector.

ENSO and its impacts on the Korean climate system. For 19 ENSO events since 1910, the three years of temperature and precipitation data from the year preceding the ENSO event to the year following the ENSO event have been analyzed using a composite analysis. The results showed several statistically significant points, especially in precipitation. One of the possibilities is associated with the Changma (the Korean long rain period associated with the East Asian Monsoon): June precipitation of the ENSO year is later than normal but July and August precipitation are slightly higher than the mean, suggesting the Changma period starts later than usual and the front stays longer over the Korean Peninsula during the whole season. There is a significant negative anomaly for September precipitation of the ENSO year and a positive anomaly for November of the same year.

Future Research and Development Plans. The multiple regression models and analog and anti-analog method should be tested for a year and improved before being put into operation. We also plan to develop a long-range forecasting system including data collection, model forecasting, and graphic presentation before the end of 1995. Another project will start at the end of 1994 to collect data for the global climate diagnostics on a quasi-real-time basis and to develop a seasonal to annual climate forecasting model.

Reference

Ropelewski, C.F. and M.S. Halpert, 1987: Global and regional scale precipitation patterns associated with El Nino/Southern Oscillation. Mon. Wea. Rev., 115, 1606 - 1626.

These comparisons show that assimilation of observed subsurface temperature data into oceanic initial conditions results in significantly improved forecasts. Further, a significant part of the additional skills due to assimilation of subsurface temperature data may come from assimilation of subsurface temperature data collected by the TOGA-TAO buoys, especially for forecasts with lead times longer than two to three seasons.

Our results are also suggestive that with data assimilation, use of higher quality winds which further improves accuracy of oceanic initial conditions, may further improve forecast skills. However, this result is far from conclusive. It is possible that errors in the Coupled Global Circulation Model (CGCM) and the imbalance between the equilibrium thermal states of the oceanic initial conditions and the CGCM may have overshadowed the impact of improved oceanic initial conditions to ocean predictions, and therefore further examination of this issue with improved models and initializations is necessary.

The analysed winds at ECMWF and the United Kingdom Meteorological Office (UKMO) were compared. The differences appear at first sight to be relatively small (0.005 N/m2 on the equator and 0.015 N/m2 in the trades, but these wind differences can lead to differences of 1øC when used to force an ocean model, indicating that the differences are significant. The comparison was over the period 1992 - 93. There was a marked sustained difference in the wind stress and SST fields during 1993, corresponding to the ECMWF winds being stronger (more easterly) than the UKMO.

The drag coefficient appears to vary by a factor of two for the same wind speed. One explanation is that it may depend on sea state. To test the impact of different wave drag formulation, the stresses from a control integration and an active wave simulation were used to force an ocean model. The active wave simulation consisted of a wave model coupled to the ECMWF model running in climate mode. The control was equivalent but used a drag coefficient which did not depend on sea state. The relationship between sea state and stress was as given by Janssen (JPO, 1989). The differences in SST were as large as 1C, comparable to the size of interannual anomalies, and therefore potentially significant. In much of the equatorial belt the non-local influence of swell was significant. Further work is required to assess the significance of these results in a wider context.

Instability waves in the eastern Pacific are a prominent feature of ocean infrared satellite images, and of many ocean models. They are detectable to a limited degree in the TAO array. Explanations are that the waves arise as shear instabilities between the North Equatorial Countercurrent (NECC), the Equatorial Undercurrent and the South Equatorial Current. However, by using ECMWF winds to force an ocean model, it appears that the phase of the waves particularly at the start of their season and possibly at the end may not be random, but rather be triggered by Kelvin waves and their reflected Rossby waves generated by westerly wind bursts occurring in the western Pacific.

Reference

Janssen, Peter A.E.M., 1989: Wave-induced stress and the drag of air flow over sea waves. J. Phys. Oceanogr., 19, 745 - 754.

Finally, heat transfer due to mixing is estimated by using the observed turbulent energy, and it is found that mixing accounts for about 70% of warming during the measurement period.

Empirical Orthogonol Function (EOF) and isopycnal analyses suggest wave motion and intrusion of off-equatorial water are the main processes for the variation of temperature and salinity in the upper ocean, along with flux through the sea surface and the mixing process in the mixed layer. The spectra of the first two EOF's exhibit peaks at one-half day and two days.

Zonal currents are principally eastward except in the uppermost layer above 100 m in which changes of the current direction are found frequently. Mean eastward current corresponding to the EUC has two maxima at 70 - 90 m and 210 m. The depth of the EUC core tends to shift in the vertical direction with time scales from several days to a few months. The layered structure of the zonal currents are represented by those of EOF modes. For example, the first EOF mode at the western site, contributing to about 60% of the total variance, has no nodes in the depth range and a maximum at 90 m, meaning that the entire layer changes with the same phase. In the uppermost layer above 100 m, strong eastward currents reaching about 80 cm/s are found in January 1992. This may be caused by the occurrence of a westerly burst in the western equatorial Pacific. The meridional current has a time average of about zero throughout almost the entire layer, and its variations are about half the magnitude of zonal current variations. Spectral analyses indicate that dominant changes are found at periods of about 10 days, 15 - 20 days, and 30 - 60 days.

Time series during the Intensive Observing Period (IOP) are compared with those from two sites (2N,156E and 2S,156E) in which moored ADCP measurements were made by Dr. A. Kaneko in Hiroshima University in Japan. Fluctuations with a period of about 2 days are found in many time series of the meridional current, and is dominant from the middle of November to the middle of December 1992.

The objective of the present study is to investigate interannual variations of subsurface temperature in the equatorial Pacific. Thus, high frequency variations with periods less than 60 days are removed by applying a digital filter to the original data. A difficulty in using TAO data for investigations of low-frequency variations arises because of data gaps. In the present study, data gaps appearing in the smoothed filtered data are filled by using the correlation method of Johnson and McPhaden (1992). Most of the data gaps are filled with this technique, but some gaps can not be filled when all vertical levels are missing for several months. MSSA is then applied to fill the remaining gaps. The filling is done by fitting the first 10 eigenvectors to the data before and after the data gap period, using the fits to interpolate across the gaps.

It is also shown by using MSSA that in all locations over the equatorial Pacific, first and second eigenmodes are characterized by the El Nino and quasi-biennial periods, respectively. Largest amplitudes of the modes appeared at the thermocline depth which appears at 150 m depth in the western Pacific and 100 - 50 m depth in the eastern Pacific. The present results also indicate that a strong El Nino signal appears first in the deep western Pacific thermocline, preceding the eastern Pacific El Nino signal by about one and one-half years. The propagation characteristics of the El Nino signal are consistent with what Latif and Graham (1992) showed with their tropical ocean model.

References

Johnson, E.S. and M.J. McPhaden, 1992: On the structure of intraseasonal Kelvin waves in the equatorial Pacific Ocean. J. Phys. Oceanogr., 22, 951 - 962.
Latif, M. and N.E. Graham, 1992: How much predictive skill is contained in the thermal structure of an oceanic GCM?. J. Phys. Oceanogr., 23, 608 - 625.

The surface relative to bottom dynamic height fluctuations observed in situ during the 6 - 7 month experiment had a standard deviation of 5 cm with excursion on the order of ñ15 cm. Energetic steric sea level variability was found to exist on short time scales on the order of hours to a few days, with in particular, the quasi permanence of strong semi-diurnal internal tides. Such internal tides were noted to induce changes in surface dynamic height with a standard deviation of 2 dyn cm. At the shallower of the two sites, 2S,156E, a possible nonlinear rectification of the internal tide was observed occasionally to change the dynamic height by as much as 30 cm over less than an hour. On time scales longer than the 10 day repeat of the TOPEX/Poseidon satellite, the low-frequency fluctuations of dynamic height were related to interannual variations corresponding to the 1991 - 93 ENSO, the seasonal cycle and intraseasonal variations associated with the 40 - 60 day oscillations of the equatorial zonal wind field.

Instantaneous comparisons between the TOPEX/Poseidon altimeter retrievals and 5 minute dynamic height were performed with the altimeter data corrected using several tide models, the barotropic tide measured in situ, ECMWF surface air pressure, and the surface air pressure measured in situ. Depending on the choice of the environmental corrections applied to the altimeter data, the rms differences between the TOPEX/Poseidon and the in situ measurements of sea level were as low as 3.3 cm at 2S,156E and 3.7 cm at 2S,164E. When the high frequency energy was filtered out of both time series by applying a 30-day Hanning Filter, the in situ and satellite data were found to be highly correlated, r=0.96 and 0.93, with rms differences = 1.6 and 1.8 cm, respectively at 2S,156E and 2S,164E.

An ocean model data assimilation system has been extensively tested over the period 1980 to the present. The assimilation method uses sequential optimal interpolation with 10-day or one-month data windows. The model has been tested with Florida State University (FSU) winds and with Bureau of Meteorology surface wind analyses, and uses observed SST and climatological sea surface salinity for the surface boundary conditions. The data assimilation model has demonstrable skill measured against the routine analysis system, particularly through extreme conditions (warm and cold events).

The improved statistical forecast analysis system was used to analyze the relative information content of the TAO and VOS XBT sampling networks. Three experiments were conducted: one using all available subsurface in situ data, one with only mooring (TAO) data, and one with only VOS data. In the near equatorial region, TAO clearly is the dominant source of information, as expected, and nearly all the important variability is captured with just TAO alone. The VOS system misses some of the low frequency variability and much of the high frequency variability though, perhaps surprisingly, it does occasionally capture important variability associated with westerly wind bursts. Assessed over a broader domain (20S - 20N), the complimentary nature of the two systems is evident and suggests the combined information content is superior to either individual contribution. This is reinforced by examination of the effective observation count (i.e., an estimate of the number of independent observations actually presented for analysis) which shows that over the tropical domain the individual systems are contributing about the same information content (at least as counted by this system) and that redundancy is relatively small.

A preliminary examination of these same observing networks was also undertaken through coupled model hindcast experiments using 400 m depth-averaged temperature analyses to initialize the model. The coupled model has been developed by Richard Kleeman at BMRC and its adjoint has been developed by Andrew Moore. Depth-averaged temperature anomalies were found to be a suitable proxy for model thermocline deviations. Data assimilation has been shown to improve this model's skill over the 1982 - 1991 period. It was found that this skill (hindcast anomaly correlation for SST in the NINO3 region) was not degraded by omitting either the TAO or the VOS data, though the rms error for the TAO-alone experiment was somewhat higher. For this period, and using the intermediate coupled model and its adjoint, it would seem that improvements can be obtained with relatively small amounts of subsurface information. Further experiments are in progress.

The TAO time series have also been used to estimate integral time scales in several frequency partitions by filtering the data and examining the autocorrelation functions. A principal finding is that for high-passed data (half-power point at 150 days), the integral time scale of both SST and 20C depth is about 5 days. This suggests that one degree of freedom is gained roughly each five days of unaliased observation of upper equatorial Pacific temperature.

Zonal and meridional space scales are estimated by finding cross-correlations for all station pairs along a given meridian or latitude circle. These correlations were weighted by their significance and fit with analytic functions, either a Gaussian if the correlations were all positive, or a combined cosine-exponential function if a wavelike process was evident. For SST, the equatorial zonal scales were found to be about 75 deg longitude for unfiltered data (dominated by the annual cycle), but greater than 100 deg for interannual variability. Corresponding scales for 20C depth were much shorter, reflecting the prevalence of Kelvin waves. High-passed scales of 20C depth were about 10 deg longitude, consistent with 50 - 60 day period intraseasonal Kelvin waves. The zonal spacing of the array is comparable to the wavelength of tropical instability waves, making the array unsuitable to examine the zonal scales of SST for high-frequency measurements. Low-frequency meridional scales of SST were larger than the extent of the array, particularly in the east where the annual and interannual variations of the cold tongue encompass a wide region. In the west these scales were somewhat shorter. Meridional scales of 20C depth were shorter than those for SST, due to the meridional scale of the Kelvin wave variations. Scales of high-passed variability were of the order of 1 - 2 deg latitude, suggesting that the TAO buoy spacing is just barely close enough to resolve these scales.

TAO Project Office
NOAA | Pacific Marine Environmental Laboratory
7600 Sand Point Way NE
Seattle, WA 98115
atlasrt@noaa.gov
Credits | Disclaimer | Privacy Policy