|TIP 4 Science reports|
Variability in the Tropical Atlantic
A second mode of variability has many similarities to ENSO, featuring anomaly patterns that are focused on the equator, with weakened (strengthened) easterly trades associated with equatorial warm (cold) events. The structure of the anomaly patterns is strongly suggestive that this coupled mode is dynamically akin to ENSO, and local to the Atlantic. The suggestion is reinforced by modeling experiments using an intermediate coupled model analogous to that developed by Cane and Zebiak for the tropical Pacific. The model reproduces several of the key features of the observed equatorial variability, including some features that are distinct from their Pacific counterparts. Differences in the parameters controlling local coupling can account for the fact that Atlantic and Pacific variability exhibit similar time scales, despite the large differences in basin size. The mechanisms accounting for the coupled Atlantic mode are discussed, as are the competing effects of local versus remote forcing, and the implications for prediction of Atlantic interannual variability.
Surface and Subsurface Climatic Variability in the Tropical Atlantic Ocean
J. Servain, ORSTOM, France
A living Example. An exceptional warming of the sea surface temperature (up to 2°C) initiated in the beginning of 1995 over the northern tropical Atlantic basin. At the time this report was written (September 1995) the warming episode is still in progress, continuing to spread southward. This event appears to be linked to abnormal atmospheric conditions which have been responsible during the summer 1995 for large flood disasters in Morocco (more than 200 victims), and for intensive hurricanes which recently devastated regions in the West Indies and Gulf of Mexico. Meanwhile, another warming event (possibly linked to the first one), emerged with the same intensity in the eastern equatorial Atlantic. Associated with the second warming event, a rising up to 6 centimeters of the sea surface in the same region was observed, by both the TOPEX/POSEIDON altimeter, and by a tide-gauge located at Sao Tome Island. Apparently, this second warm event was responsible for an unsuccessful tuna fishery (tuna disappeared from its usual locations in the open ocean), and for weak sardinella catches along the West African coast. This large and complex abnormal 1995 event in the tropical Atlantic is the most recent example of what may occur in this region in terms of interannual climatic variability. From results extracted from the French oceanic operational model (this operational version of the LODYC's OGCM run at Meteo-France with an operational AGCM), it is established that the present equatorial warm event could have had strong abnormal signatures (Kelvin and Rossby wave propagations) below the surface. Unfortunately, due to the lack of continued subsurface monitoring in the equatorial Atlantic (as it is presently available in the equatorial Pacific, thanks to the TAO array implemented during the TOGA program), it is not easy to validate subsurface variability in simulations of the Atlantic operational model.
More Generally. Tropical Atlantic climatic variability acts according to two main modes. When strictly observed along the equatorial region, the abnormal events mostly occur during a relatively short period (a few months), and they are primarily consistent with the equatorial wave propagation dynamics. The seasonality of this mode is strong. Basically, wind perturbations occurring in the western equatorial region during the northern spring are followed during the next summer by abnormal oceanic conditions in the Gulf of Guinea. It was the case, for instance, for both the 1968 and 1984 warm events (Servain, 1984; Delecluse et al., 1994). Due to the typical equatorial dynamics, these events are highly predictable (Servain and Arnault, 1995).
When the entire tropical basin is considered, a very low frequency decadal signal reflects a meridional dipole pattern (Servain, 1991). This dipole pattern is temporally related to an abnormal meridional displacement of the ITCZ. Because the ITCZ is mostly located north of the equator (close to 5°N on average), the equatorial region (including the whole Gulf of Guinea), is part of the southern pole. When the dipole pattern is well established, it can be efficiently used to predict abnormal occurrences in the rainy season (February May) of northeast Brazil. For instance, during the periods November 1973 April 1974, and November 1984 April 1985, the ITCZ was located abnormally south, and it was associated with an abnormally cold northern basin and an abnormally warm southern basin. As a consequence, the 1974 and the 1985 rainy seasons over northeast Brazil were exceptionally wet.
However, on the surface, the range of the interannual signal remains relatively modest (about a third) compared to the range of the seasonal signal. Until recently, very little was known about the observed interannual variability in the subsurface. Using the last 15-years of vertical temperature profiles derived from the XBT measurements in the eastern equatorial Atlantic, it is found that the interannual variability in the subsurface temperature is considerably larger than that in the surface. In fact, below the surface, both seasonal and interannual temperature variations are of the same magnitude. This result seems to support the idea that the importance of subsurface dynamics in the interannual component of the eastern equatorial Atlantic is very similar to that which occurs in the eastern equatorial Pacific.
If the modes of climatic variability in the tropical Atlantic can be relatively well distinguished (an equatorial mode and a meridional mode), then there are multiple dynamical origins of this variability. Local atmosphere-ocean feedback could combine with remote teleconnections from higher latitudes (Deque and Servain, 1989), and from other equatorial regions. Thus, a recent study (Delecluse et al., 1994) indicated that oceanic and atmospheric abnormal conditions in the equatorial Pacific, associated to the 1982-1983 ENSO event, could have initiated the exceptional warm conditions which occurred during 1984 in the tropical Atlantic. According to this assumption, it is shown here that about 50% of the variance of the zonal wind stress in the western equatorial Atlantic during the Northern spring can be explained only by the Southern Oscillation index (SOI) integrated during the preceding November March period. Furthermore, a long-term trend noted in the SOI from the beginning of the 1970's (more warm events than cold events in the Pacific) seems well related with a similar trend in the strengthening of the western equatorial region.
Servain, J., 1984: Reponse oceanique a des actions eloignees du vent dans le Golfe de Guinee en 1967 - 1968. Oceanol. Acta, 7, 297-307.
Deque, M. and J. Servain, 1989: Teleconnections between tropical Atlantic sea surface temperatures and midlatitude 50 KPa heights during 1964-1986.J. Climatol., 2, 929-944.
Servain, J., 1991: Simple climatic indices for the tropical Atlantic Ocean and some applications. J. Geophys. Res., 96, 15,137 15,146.
Delecluse, P., J. Servain, C. Levy, L. Bengtsson, and K. Arpe, 1994: On the connection between the 1984 Atlantic warm event and the 1982-1983 ENSO.Tellus, 46A, 448-464.
Servain, J. And S. Arnault, 1995: On forecasting abnormal climatic events in the tropical Atlantic Ocean. Annales. Geophys., in press.
Mixed-Layer Model for the Tropical Atlantic Circulation
With the purpose of showing the present status of the modeling effort, the results of low-resolution (9/10 of a degree) runs executed for assessing the model's performance with imposed zonal boundary conditions were presented. Basically two experiments were discussed. The first one used the same forcing fields based on COADS data set used by Bleck (pers. comm.) for studying the North Atlantic circulation. In the second experiment, precipitation data compiled at NCAR by Spencer were used instead of that from COADS. The results show that the model's mixed layer salinity distribution is considerably sensitive to this thermodynamic forcing.
Assimilation of Satellite Altimeter Data in Ocean Models
T. Rosati, NOAA/GFDL, Princeton University, NJ, U.S.A.
The Derber and Rosati (1988) Optimal Interpolation scheme has been used to assimilate both TOPEX altimetry data and XBT temperature profiles. The method of assimilating the TOPEX data is an extension of the Mellor and Ezer (1991) statistical extrapolation procedure applied to the global ocean and with temporal variability in the correlation functions. The impact of each data set is examined and in the tropical Pacific the results are compared to the TAO mooring sites. In regions where the variability is high the TOPEX data better resolves those features than with XBT data only. However, in situ data is needed to define the mean state.
Derber, J. and A. Rosati, 1989: A global oceanic data assimilation system. J. Phys. Oceanogr., 19, 1333-1347.
Mellor, J.L. and T. Ezer, 1991: A Gulf Stream model and altimetry assimilation scheme. J. Geophys. Res., 96, 8779-8795.
The Role of in situ Data in the NMC's Climate Forecast System (M. Ji, NOAA/NMC, U.S.A.)The observed climatic state since 1990 has shown noticeably different behavior from that during the early 1980s. This recent period was characterized by a more persistent nature of warmer than normal SSTs and weaker than normal trade winds in the tropical Pacific. Prediction of El Ni¤o episodes during the 1990s has been more difficult than prediction of those episodes during the 1980s. Four sets of parallel forecast experiments for the 1992 1995 period were carried using oceanic initial conditions produced with various wind forcing but without subsurface data assimilation. This is in order to explore the potential predictability that may come from improved wind stress forcing especially from including in situ observations of surface winds from the TAO buoys. The forcing fields used are a) from the NMC operational analyses; b) from the FSU analyses based on ship/buoy observations; c) from the European Remote Sensing Satellite (ERS-1) winds; and d) from the NMC operational winds but improved by combining observed surface winds from the TAO buoys. These are denoted as NMC, FSU, ERS-1, and TAO, respectively. Of these 4 wind fields, the TAO and FSU winds utilized nearly the full potential of the TAO observations while the NMC and ERS-1 have little or none of the TAO data in them. The forecast skill from all four sets of experiments are low, as expected from the NMC coupled model without using data assimilation. However, skills from the FSU and TAO experiments are maintained nearly constant with forecast lead time up to 12 months while those from the NMC and ERS-1 forecasts dropped significantly after about 2 seasons.
A number of recent studies have shown that with assimilation of observed in situ subsurface temperature data into ocean initial conditions, forecast skill for coupled models can improve significantly. However, model errors may play a significant role, i.e., these conclusions may be dependent on coupled models used in the studies. To further quantify potential improvement in coupled model predictability by using data assimilation to obtain better defined oceanic initial conditions, pseudo forecasts, i.e., perfect coupled model experiments where the winds from the atmosphere were taken from the FSU analyses, were performed. The ocean initial conditions used in these experiments were obtained from analyses, i.e., assimilation of subsurface temperature data into ocean model, and from simulation, i.e., no data assimilation was done. These pseudo forecasts result in higher skill when oceanic initial conditions were obtained from analyses than those from simulation. This indicates that presently, errors in the winds and in the ocean model are sufficiently large to require data assimilation to compensate for them. Additionally, skills from both sets of pseudo forecasts are significantly higher than forecast skill achieved using the NMC coupled model, indicating that presently model errors are the main limiting factor for higher forecast skill.
New techniques for assimilation of sea level data from TOPEX/POSEIDON are currently being developed and implemented at a number of institutions including at the NMC. Satellite data can provide basin wide coverage and an observing system based on satellites is much easier to implement than those of in situ instruments based (albeit more expensive). However, present sea level data from satellite can provide a good estimate of anomalies but not the mean state. Assimilation of sea level anomalies require good estimates of the mean subsurface temperature structure which are provided by in situ data from TAO and XBTs. The sea level assimilation system developed at NMC is based on using both in situ and remotely sensed data. The addition of satellite sea level data improved the quality of ocean analyses.
Prediction of the ENSO Phenomenon Using a Coupled Model
A.J. Busalacchi, NASA/Goddard Space Flight Center, U.S.A.
Short-term climate prediction, i.e., seasonal to interannual, has begun to generate a demand for operational ocean observations. The TOGA observing system was designed to respond to this need to monitor and predict El Ni¤o and the Southern Oscillation. As the TOGA observing system evolved, so too did the utilization of these observations in modeling studies. Going into the GOALS/CLIVAR program, data assimilation methodologies will provide the link between coupled ocean-atmosphere models and the observations. ENSO prediction skill will be a valuable metric for assessing observational, modeling, and assimilation strategies. This paper will address the merits of assimilating data into a coupled forecast model. Over the past several years ENSO forecasts have benefited from the assimilation of observations into ocean or atmosphere models in a stand-alone mode as part of initialization procedures performed in a serial manner. Assimilation approaches have also been used in support of parameter estimation for coupled models. Here we will present a coupled initialization procedure that relies on the assimilation of wind observations as a means of dealing with the issues of filtering and balance of initial conditions to ENSO forecasts. The methodology is to initialize the forecast system in a coupled manner, using a simple data assimilation procedure in which the coupled model wind stress anomalies are nudged toward observations. The new data assimilation procedure for the intermediate coupled model developed by Cane and Zebiak yields substantially improved ENSO forecasts for the 1980s and comparable skill for the 1970s versus previous forecasting procedures. As in the earlier forecasts with the same model, only wind information is assimilated. The improvement is attributed to this use of a truly coupled ocean-atmosphere initialization. The results suggest that ENSO is more predictable than previously estimated. This new method of data assimilation effectively filters out small-scale, high frequency variations that rapidly contaminate forecast skill. The initialization procedure effectively lessens the mismatch between initial conditions and the model's intrinsic variability, while retaining the essential large-scale, low-frequency information, resulting in improved forecast performance and enhanced stability out to 24 months. The coupled approach to initialization also eliminates the well-known "spring barrier" to ENSO prediction, implying that this characteristic may not be intrinsic to the real climate system.
A Modest Modal
Model of Kelvin Waves During the Onset of the 1991 92 El Nino
the Tropical Indian Ocean Circulation
Levitus, S., 1982: Climatological ATLAS of the World Ocean, NOAA Professional Paper 13, U.S. Gov. Printing Office, Washington, D.C., 173 pp.
Zonal Displacement of the Western Pacific Warm Pool Associated with a Series of ENSO Events
J. Picaut, ORSTOM, Noumea
In the equatorial Pacific, the zonal displacements of the eastern edge of the warm pool, subject to insignificant seasonal variations, are dominated by strong interannual variations in phase with the Southern Oscillation Index. Gill suggested in 1983 that these displacements were caused solely by horizontal advection associated with zonal current anomalies during the 1971-73 El Nino-La Nina period. The purpose of the present study, based on observational findings and modeling analyses, is to assess Gill's suggestion over several El Nino and La Nina events using a combination of data sets and numerical models.
The observational approach is based on SST data and several types of derived and directly measured surface current data. The data sources include Reynolds' SST, GEOSAT (November 1986 February 1989), TOPEX/POSEIDON (October 1992 January 1995), moored mechanical current meters and ADCPs at five equatorial sites (1986 - 1994), and surface drifters (1987 - 1993). The modeling approach is based on two types of model: a 10-vertical mode linear model forced by the 1961 - 94 FSU wind stress, and a high resolution OGCM (LODYC/OPA) forced by the wind stress and heat flux over the 1986 - 89 period.
From the basin-wide GEOSAT-derived surface current anomalies, and inferred water parcel displacements, it is demonstrated that the eastward (westward) displacement of the eastern edge of the warm pool was primarily due to horizontal advection during the 1986 - 87 El Nino (1988-89 La Nina). As a corollary, El Nino (La Nina) warm (cold) SST anomalies in the central-western equatorial Pacific were mainly the result of anomalous zonal advection. These low frequency displacements and associated temperature anomalies appear to be the result of a succession of higher frequency forcing, i.e., a succession of local wind forcing and its remote Kelvin and m-1 Rossby wave response.
The extension of this observational study to the 1989-95 period, and the ensuing long-lasting 1991-94 El Nino, supports the idea that anomalous zonal advection is an important mechanism for the displacement of the eastern edge of the warm pool. However, due to incomplete or inaccurate in situ data between the GEOSAT and TOPEX/POSEIDON missions or due to other important physical processes, it is difficult to firmly establish the dominance of zonal advection in the slow recovery of the warm pool from its far western position during the 1988 89 La Nina to its mean position prior to the 1991-94 El Nino.
For the 1961-94 period, horizontal displacements of the eastern edge of the warm pool are well simulated by the trajectory of the water parcels simulated using the linear model. The role of local wind forcing, remote equatorial wave response and vertical modes is investigated, and in particular it appears that the first vertical mode is dominant in this multi-year simulation. Coincident with the 1986-89 GEOSAT period, the OGCM reproduces reasonably well the main sea-surface temperature, current and equatorial wave features associated with the 1986-89 El Nino-La Nina. In this simulation, horizontal anomalous advection is notable for the El Nino-La Nina displacements of the eastern edge of the warm pool, and contributions of heat flux and other advective terms are contrasted.OSSE on the Buoy Network of JAMSTEC
A. Sumi, University of Tokyo, Japan
An Observation System Simulation Experiment (OSSE) has been conducted to evaluate the performance of the buoy network proposed by JAMSTEC.
Simulated data were used in the GFDL high-resolution Pacific Ocean model (1/3° in latitude, 1/3° - 1° in longitude and 20 levels) at the University of Tokyo (Masumoto and Yamagata, 1993). These simulated data have been assimilated through ODAS (Ocean Data Assimilation System) of JMA (Yoshikawa et al., 1995), where the Brian-Cox model was used (1/2° - 2° in latitude, 2.5° in longitude and 20 levels).
Six experiments are conducted:
Masumoto, Y. and T. Yamagata, 1993: Simulated seasonal circulation in the Indonesian Seas. J. Geophys. Res., 98, 12,501 12,509.
Yoshikawa, I., M. Kimoto, and M. Ishii, 1995: Ocean data assimilation system for climate monitoring at JMA. In:Proceeding of the Second International Symposium on Assimilation of Observations in Meteorology and Oceanography. WMO/TD, NO. 651.
|Dicussions and recommendations >>|
|Home | Project overview | Data display | Data delivery | El Niño & La Niña | Site map|