* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download alexander b. polonsky
Scientific opinion on climate change wikipedia , lookup
Iron fertilization wikipedia , lookup
Climatic Research Unit documents wikipedia , lookup
Climate change in Tuvalu wikipedia , lookup
Public opinion on global warming wikipedia , lookup
Effects of global warming on humans wikipedia , lookup
Hotspot Ecosystem Research and Man's Impact On European Seas wikipedia , lookup
Solar radiation management wikipedia , lookup
Global warming wikipedia , lookup
Surveys of scientists' views on climate change wikipedia , lookup
Climate change and poverty wikipedia , lookup
Climate change, industry and society wikipedia , lookup
Climate change feedback wikipedia , lookup
Future sea level wikipedia , lookup
Numerical weather prediction wikipedia , lookup
Climate sensitivity wikipedia , lookup
Years of Living Dangerously wikipedia , lookup
Ocean acidification wikipedia , lookup
Effects of global warming on Australia wikipedia , lookup
IPCC Fourth Assessment Report wikipedia , lookup
Attribution of recent climate change wikipedia , lookup
Physical impacts of climate change wikipedia , lookup
Atmospheric model wikipedia , lookup
Effects of global warming on oceans wikipedia , lookup
Global warming hiatus wikipedia , lookup
PROBLEMS OF ADEQUACY OF OCEANOGRAPHIC AND METEOROLOGICAL DATA AND THEIR ASSIMILATION FOR LONG-TERM CLIMATE VARIABILITY STUDY ALEXANDER B. POLONSKY Marine Hydrophysical Institute, Sevastopol, the CRIMEA (E-mail: [email protected]) 1. PRELIMINARY REMARKS Ocean variability is a crucial factor in regulating the decadal-to-century changes of the coupled ocean-atmosphere system and hence of the climate (Bjerknes, 1964; Kushnir, 1994; Polonsky et al., 1995). Certainly in principle, atmosphere can vary in the broad range of periods (including decadal-scale, see e.g., [Power et al., 1995]) without active sea-air interaction. It seems however, this statement concerns mostly the planets without Oceans. The Earth on the contrary is more close to the aquaplanet. Oceans are more slow and inertial component of the coupled system. Low-frequency changes of the meridional ocean circulation and of the surface heat fluxes due to the sea surface temperature (SST) variability are the principal causes of the long-term variability of the coupled system resulting in the climate changes. So, the reliable detection of the low-frequency ocean variability is crucial factor to study the long-term climate changes. Because of the limitations in the quantity and quality of oceanographic observations, the joint analysis of different datasets and discussion of the problem of climatic signal detection are necessary in order to study the long-term variability. This is the aim of the present paper with the strong focus to the deep Atlantic Ocean. It continues the discussion of this topic initiated at the 10th Conference on Applied Climatology (Polonsky, 1997). 2. RESULTS AND DISCUSSION 2.1 Deep-sea observations Deep-sea observations have been performed for more than a century. Before the 1970's, Nansen profiles were the main source of deep-sea data. Then the XBT and CTD soundings essentially replaced the Nansen measurements. However, the routine oceanographic observations are too sparse and noisy for the reliable detection of lowfrequency (long-term) changes of the oceanic fields in the deep ocean (below about upper 800 m layer), except perhaps near a few standard oceanographic sections and some specific regions with strong oceanographic activity (Dzhiganshin and Polonsky, 1993; Parilla et al., 1994). Moreover even in these regions, activity was concentrated in recent decades. This means that even there, only decade-to-decade (not longerterm) variations may be estimated with reasonable accuracy. Regular XBT profiles have been collected during the last 25 years and the maximum depth of XBT sounding is usually 800 m. Therefore the XBT data cannot eliminate the lack of long-term deep-sea observations. At the same time, from the point of view of a long-term variability study, the systematic discrepancy between the Nansen, CTD and XBT data sets may also be an important problem (Gaysky et al., 1992). While the systematic errors may be corrected in some cases (as done e.g., by Wright (1986) for SST), the random high-frequency noise cannot be eliminated in principle. This may be only reduced using optimal filtering or/and interpolation, if one has at least a few observations near a considered grid point. However, high-amplitude noise and the small amount of deep sea observations do not permit to obtain an accurate depiction of "ocean climate" even for the most recent decades. Moreover, the unknown statistical structure of regional deep sea oceanographic fields leads to errors of interpolation which are of the same order as the typical magnitude of low-frequency variability (Gaysky et al.,1992). As a result of all kinds of errors the joint combined climatic datasets consisting of all types of observations, such as Levitus' (1982) datasets, demonstrate strong artificial vertical inversions due to different space-time distribution of various measurements on different standard levels (Bulgakov et al., 1990). Recently White (1995) gave the assessment of statistical structure of the temperature fields in the upper 400m layer in all Oceans to the North of about 30˚S. For the tropical Atlantic his results confirm the assessment of Gaysky et al. (1992) He concluded that the sampling rate is adequate for detecting the space-time evolution of biennial signal over the interior global ocean, but it is inadequate for detecting year-to-year changes in upper ocean heat storage averaged over large portions of the global ocean. It is clear the situation with the decade-to-decade changes is much worse. In fact, rather optimistic White’s result is due to quite numerous XBT observations during 80’s. During most decades of XX century the amount of observations is poor and even the biennial fluctuations in the upper 400 to500m layer cannot be detected reliably. Below 500600m, all the time of oceanographic observations is not suitable to detect the long-term ocean variability over the globe. The precise long-term observations are necessary to improve such case. The regular observations at the Ocean Weather Station (OWS) and the World Ocean Circulation Experiment (WOCE) field program are good examples of such strategy.It is unacceptable (from the point of view of the long-term variability study) to interrupt these space-time series. Numerous attempts to estimate the long-term trends in deep-sea fields of the North Atlantic Ocean, where generally the most intense observational activity in XX century occurs, gave few significant results. I’ll put attention to the some of them. First, the positive temperature and salinity trends in the deep North Atlantic Ocean during recent decades were found and documented by Antonov and Grojsman (1989), Levitus (1989) and Parrilla et al. (1994). Secondly, the warming of the subsurface equatorial Atlantic during the summer and autumn months of the 1980's in comparison with previous decades was documented by Polonsky et al. (1995). Thirdly, cooling and freshening of high-latitude North Atlantic waters in the 1980's were found by Read and Gould (1992). Besides, the coupled nature of the decadal-scale variability of the subtropical gyre was demonstrated recently by Polonsky and Kuzmin (1999) using historical data sets from 1969 to 1992. They showed that the term W T (where: W is the decadal z fluctuations of vertical Ekman velocity due to the decadal change of horizontally-inhomogeneous wind field over the subtropical Atlantic and T z is the average vertical temperature gradient in the subtropical thermocline) is very important in the decadal heat balance of the Atlantic subtropical thermocline. Moreover, the phase characteristics of the decadal fluctuations of vertical Ekman velocity over the eastern subtropical Atlantic and of vertical temperature gradient in the subtropical thermocline are similar one another. The westward propagation of the decadal temperature signal due to the linear Rossby waves (as predicted by the standard theory) is absent in the subtropical gyre because the Rossby waves are modified by the eastward average currents. It should be emphasized that all these results were obtained using simple statistic processing without assimilation procedure. It is not sure that assimilation of recent global and/or regional datasets in ocean or coupled models (or/and other recent techniques, e.g., ensemble averaging) may specify the above results or improve essentially their statistical significance. In fact, uncertainty of the surface boundary conditions and model drift may obscure the real long-term ocean variability. Let us analyze briefly these problems. 2.2 Surface fields Both large-scale climatic-mean surface wind and surface heat fluxes are estimated with a typical error of 30 to 50% at best. For instance, Harrison (1987) argued that wind stress of Hellerman and Rosenstein (1983) is overestimated by about 40%.Large discrepancy between winds of Hellerman and Rosenstein (1983) and of the European Center for Medium Range Weather Forecast (ECMWF) was found by Trenberth et all (1989). There are two sources of discrepancy in different wind climatic sets. The first of these is due to different techniques of wind observation and of wind stress calculation (space averaging, drag coefficient, etc.). The second is due to the difference in the periods of averaged observations. For example, Hellerman and Rosenstein (1983) used the ship observations collected during about century, while the ECMWF wind stress was calculated using fourdimensional assimilation of the available global information for about a decade (seven years). As a result, the simulated ocean responses to two of these wind fields differ essentially without any changes of other parameters of the model (Chervin and Semtner, 1993). It is not clear whether this change of response is due to real differences in the wind climate or not. Trenberth et all (1990) analyzed the differences of zonally integrated Sverdrup transports produced by winds of ECMWF and of Hellerman and Rosenstein (1983), while Landsteiner et all (1990) analyzed the same differences produced by three wind stresses over the tropical Pacific in 1979 to 1981, namely marine climatology, fourdimensional assimilation of all available information into the GCM and satellite data. Reference is further made to Trenberth et all (1990). They argued "that in spite of the expected reduction in wind stress due to the use of a drag coefficient that is about 20% smaller, the added variance captured by the ECMWF analyses has apparently been compensated to some extent... the 1980-86 period (period of the ECMWF wind averaging - AP) may have been unusual." Landsteiner et all (1990) intended but could not check the existence of the Sverdrup balance in the equatorial Pacific because the differences between three their assessments were too large. Reynolds et all (1995) recently compared the simulation results concerning the sea level variability in the equatorial Pacific using NMC (U.S.National Meteorological Center), FSU (Florida State University) and satellite (ERS-1) wind forcing and TOGA-TAO data. They found large differences between these results especially in the Eastern equatorial Pacific. FSU wind agreed with TOGA-TAO data best of all. Systematic errors of definition of monthly wind can be caused by the transition in the 1960's to instrumental measurements instead of visual estimation using the Beaufort scale. It results in artificial positive trends (see for example, [Isemer and Hasse, 1991]). Actually, wind speed data, obtained using the Beaufort scale, underestimate the wind speed. The instrumental measurements began at the end of the 1940's, their fraction in total number of observations has increased in the 1950's. Approximately in the middle of 1960's they already prevailed in the magnitudes of monthly mean wind speed. This may lead to an artificial trend of about 0.1 m/sec per 10 years. However, the artificial linear trend was not found in the North Atlantic by Voskresenskaya and Polonsky (1994) using comparison of measured wind and index of circulation for 1957 to 1990. The positive wind speed trend in the tropical and subtropical Atlantic after the Second World War was determined also in the independent estimates by Bigg (1993) and Wolter and Hastenrath (1989) using ship observations and measurements at the coastal stations. Voskresenskaya and Polonsky (1994) showed that the trend of wind speed in the North Atlantic is even more statistically significant than the trends of thermal characteristics and the humidity deficiency. On the other hand they showed that low-frequency change in wind field is better approximated by a nonlinear trend and that this change is actually a quasi-periodical interdecadal variability which strongly correlates with high-latitute temperature change and hence with overturning intensity (see also, [Polonsky and Voskresenskaya, 1996 a, b]). However, information is insufficient to estimate the multidecadal and century variability of the wind field over entire ocean basins. Moreover, high-frequency changes of the wind field should also be taken into account for simulation of the climate processes (Polonsky et al., 1992). This means we should have at least daily global wind fields for a long (about century) period if we want to study the multidecadal changes in the coupled ocean-atmosphere system or in the Oceans. Typical absolute errors of calculation of heat gain using climatic monthly hydrometeorological fields and bulk formula are more than 15 and more than 50 Watt per square metre in the midlatitudes and in the tropical regions, respectively (Gleckler, 1993; cf. also results by Budyko, 1963, and Bunker, 1976). Lack of precise longterm cloudiness observation is one of the main causes. Therefore the current heat flux assimilation may not improve essentially the results of simulation of long-term oceanic climatic change. However, some significant results concerning a trend-like variability of the sea surface heat fluxes were obtained using the climatic monthly fields after the Second World War and bulk formula. For instance, Voskresenskaya and Polonsky (1994) found that significant surface cooling of the North equatorial Atlantic during the 1960's and 1970's was accompanied by reduced interface humidity contrast and air-sea temperature difference and intensification of the trade winds. Then (in 1980’s), the sign of linear trend of the surface heat flux has been changed. As a result, Polonsky and Voskresenskaya (1996a) extracted the significant parabolic trend of the total heat flux over the most North Atlantic in 1957 to 1990. It is due mostly to the parabolic trend of the interface humidity contrast which is, in fact, the manifestation of the 50-60yr variability in the coupled system. Advective ocean processes regulate this variability. Kushnir (1994) analyzed the differences of SST, sea surface pressure and wind between anomalously warm periods in the North Atlantic (1950 to 1964) and anomalously cold one (1970 to 1984). He showed that North (extratropical) Atlantic warming is accompanied by cyclonic circulation anomalies in mid-latitude atmosphere. It seems that the recent techniques (such as assimilation procedure) cannot specify the above results. Certainly, the assimilation of meteorological data is very effective for the short-term weather forecast. The best atmospheric models, such as ECMWF, are using the assimilation during about 20 years. However, the model's improvement, change of parameterization, etc., lead to artificial changes in the results of simulation if one deals with the long-term atmosphere variability. Therefore, the atmosphere fields produced by such models without correction are useless for the detection of interdecadal variability. Results of the reanalysis from 1979 those are available now are suitable to study the interannual-to-decadal (not longerterm) climatic fluctuations. It seems, the use of the coupled models permits to avoid the problems of the surface field inadequacy. However, on the other hand, the other serious problems arise if one uses the interactive models. 2.3 Model inadequacy Inadequacy of models include a lack of complete parameterization of relatively small-scale and high-frequency processes in the coupled oceanatmosphere system. Recent coupled oceanatmosphere models consist of a relatively highresolution atmospheric part and a relatively coarse oceanic component (Gates et al., 1993; Phillips, 1994). This difference is due to the various Rossby scales in the atmosphere and oceans. Therefore the typical space step of a few degrees in the recent coupled models is eddyresolving for the atmosphere but not the oceanic part. Certainly, the resolution of coupled models may be improved essentially in the near future. However, it is not clear whether a more detailed resolution itself may improve essentially the results of climate modeling. For example, a comparison of the land surface climatology produced by two versions (R15 and R42) of the NCAR model showed only some regional differences (Bonan, 1994). This is not the case for the parameterization of cloudiness, smallscale sea-air interaction and especially mixing processes in the oceans. Small-scale turbulence, double diffusion and internal waves in the oceans are the principal mechanisms of the highfrequency thermodynamics of the coupled system which are responsible for the mixing and dissipation. They cannot be simulated directly in large-scale coupled models because of extremely small space-time scales. However, they are the crucial factors regulating the long-term thermodynamics and properties of the coupled ocean-atmosphere system. Accordingly, a complete parameterization of mixing processes in the oceans is extremely important. Unfortunately this problem is not solved. Different parameterizations of the upper mixed layer processes and deep-sea mixing give very different results even in the ocean-only models (Polonsky, 1989; Hirst and Cai, 1994; Marotzke, 1997). For instance, Marotzke (1997) showed that the meridional heat transport is proportional to Kv½ (Kv is the coefficient of vertical mixing) if Kv varies in the range permitting the steady solution. At the same time, Weber (1998) clear demonstrated that the strong sensitivity of the interactive ocean-atmosphere model to the reasonable change of the coefficient of the vertical diffusion in the ocean occurs. Moreover, ocean temperature and salinity in the coupled model are more sensitive to this coefficient than in the ocean-only model. Another problem of the coupled system is the large time scale of the ocean along with a poor knowledge of its initial state. At least partly the model drift is due to the above problems. In fact, the "climate drift in coupled models is associated primarily with systematic errors in the component models which manifest themselves as incompatibilities in the surface fluxes required by the ocean to maintain the observed climatology, and those surface fluxes which are predicted by the atmospheric component of the model" (Moore and Gordon, 1994). Flux correction is not a principal solution of this because "although such flux corrections are independent of any anomalies in either the control or perturbation runs, and have no interaction with the simulated response, they may influence a model's ability to portray longer-term climate change. It is hoped that future improvements in coupled models will make the use of such flux corrections unnecessary" (Gates et. al., 1994). Recently Dijkstra and Neelin (1999) clear showed that multiple equilibria can strongly depend on the flux correction procedure. It seems that the assimilation of current oceanographic data into the coupled models may help to reduce the model drift. However on the other hand, the initialization of large-scale oceanic fields is necessary because otherwise the artificial model drift may increase. Numerous attempts of initialization of the oceanic fields using high-resolution ocean models were not quite effective especially in the tropical ocean (see discussion of this problem in [Bulgakov et al.,1990]). The changes in the oceanic fields strongly depend on the model parameters and time of initialization. Uncertainty of climate boundary/initial conditions and model deficiencies may lead to a wrong final state of the coupled system after initialization. We should also remember that the change of climate in the coupled ocean-atmosphere models looks like a nonlinear chaos. The state of such system strongly depends on parameterization of smallscale processes (see above and [Fleming,1993; Tziperman et al., 1995]). So, we do not know reliably even the present climatic state of the coupled system. Accordingly, forecasts of climatic tendency using coupled models, such as (Davies and Haat, 1994; Rumstorf, 1995; Stouffer et al., 1994), may be considered as very rough assessments. They underestimate usually the natural variability in the coupled system and overestimate the humaninduced changes. However, the models' improvement may lead to more reliable results in the future. Assimilation of precise oceanographic data in such models may be a useful tool for the study of the long-term variability of the coupled system over a few years/decades. At the same time, preliminary analysis of different data sets is essential as a first step in the study of the longterm variability of the coupled system. 3. CONCLUSION AND PROPOSITIONS What should we do first of all during the next few decades for the development the study of the long-term climatic variability of the coupled ocean-atmosphere system on the globe, taking into account an uncertainty of natural climate ocean state/boundary conditions, etc.? It is clearly necessary to perform precise global (including deep sea) observations and to develop coupled models. However, a fast progress in the study of the long-term climatic variability even using the assimilation procedure, ensemble averaging and improved models is not likely. Duration of observations should be of the same order as the typical time scale of processes to be studied. This means we should continue to support the global observational network at least during the next few decades and to put attention to analysis of different kinds of observations. From the oceanic point of view the most important long-term observations are as follows: -the volunteer observing system including the XBT program; -the precise deep oceanographic observational programs, such asOWS, WOCE and TOGA-TAO; -long-term satellite observations. In the best case all of these should be continued and intensified especially in the region with poor network as was recommended, for instance, by the CLIVAR-GOALS Workshop (Melbourne, Australia, 10-12April 1995) or by Godfrey et al. (1995). Recent PIRATA program spreading the TOGA-TAO-type observations at the tropical Atlantic is a good example of such strategy. In another case it is necessary to choose more informative and less expensive program. First of all this is VOS. However, the absence of deepsea observations in the VOS program does not permit to restrict our ocean observations only to the VOS. Precise long-term deep-sea observations are absolutely necessary also for the study of low-frequency climatic change. At least, the present repeated oceanographic sections should be performed during a few decades. If we change the sections' location we cannot detect reliably the trend-like deep ocean variability (even if we perform accurate observations). The same problems are well-known in the analysis of meteorological observations (Robeson,1995). Another recommendation is to renew the observations at the OWS or at least at the some of them as soon as possible. 4. REFERENCES Antonov, D.I., and Grojsman,P.A.,1988: Temperature change below active layer in the North Atlantic Ocean (in Russian). Meteorology and Hydrology, No.3, 57-63. Bigg, G.R., 1993: Comparison of coastal wind and pressure trends over the Tropical Atlantic: 1946-1987. Int. J.Climatol., 13, 411-421. Birman, B.A., and Posdnjakova, T.G., 1992: Climatic monitoring of the virtual heat fluxes at the ocean surface (in Russian). Meteorology and Hydrology, No.4, 102-104. Bjerknes, J., 1964: Atlantic air-sea interaction. Adv. Geophys., 10, No.1, 1-82. Bonan, G.B. 1994: Comparison of the land surface climatology of the National Center for Atmospheric Research community climate model 2 at R15 and R42 resolutions. J. Geophys. Res., 99, 5D,10357-10364. Budyko, M.I., 1963: Atlas of heat balance of the World (in Russian). GGO. Leningrad. Bunker, A.F., 1976: Computations of surface energy flux and annual air-sea interaction cycle of the North Atlantic Ocean. Mon. Weath. Rev., 104, 8, 1122-1140.. Bulgakov,N.P. et all 1990: Structure and seasonal variability of the hydrophysical fields of tropical and sub-tropical Atlantic as a result of adaptation (in Russian). MHI press, Sevastopol, 36p. Chervin, R.M., and Semtner, A.J., 1993: Sensitivity of a global ocean model to alternative wind stress prescription. pp.13-14 in Preprints of 4th International Conference of Southern Hemisphere Meteorology and Oceanography.(29 Mar.-2 Apr. 1993, Hobart, Australia). AMS, Boston, MA. Davies, H., and Haat, B., 1994: The problem of detecting climatic change in the presence of climatic variability. J. Meteor. Soc. Japan, 72, 5, 765-771. Dijkstra, H.A., and Neelin, J.D. 1999: Imperfections of the thermohaline circulation: multiple equilibria and flux correction. J.Climate, 12, 5, 1382-1392. Dzhiganshin, G.F., and Polonsky, A.B., 1993: Interannual variability of the oceanographic fields in the tropical Atlantic (in Russian). Oceanology, 33, 1, 654-660. Fleming, R. J., 1993: The dynamics of uncertainty: application to parameterization constants in climate models. Climate Dynamics, 8, 3, 135150. Gates, W. L.,et all 1993: An intercomparison of selected features of the control climates simulated by coupled ocean-atmosphere general circulationmodels.WCRP82,WMO/TD-No.574,46 p. Gaysky, V.A. et all, 1992: Statistical structure of the temperature field in the tropical Atlantic during GATE (in Russian). Collected papers: Automatized control of measurements in the Oceans. Marine Hydrophy-sical Institute, Sevastopol, 69-82. Gleckler, P.J., 1993: Uncertaintly estimates in global ocean surface heat fluxes. pp.93-98 in Proceedings of 17th Annual Climate Diagnostics Workshop. Norman, Oklahoma, Oct. 18-23 1992. Ghil, M., and Melotti-Rizotti, P., 1991: Data assimilation in meteorology and oceanography. Advances in Geophysics, 33, 141-266. Godfrey, J.S. et all, 1995: The role of the Indian Ocean in the global climate system: recommendations regarding the Global Ocean Observing System (OOSDP Background Report No. 6): Ocean Observing System Development Panel, February 1995, 89 p. Harrison, D.E., 1987: Equatorial sea surface temperature sensitivity to wind forcing. p. 413419 in Further progress in equatorial oceanography., Nova University Press. Hellerman, S., and Rosenstein, M., 1983: Normal monthly wind stress over the World Ocean with error estimates. J.Phys. Oceanogr., 13, 4, 1093-1104. Hirst, A.C., and Cai, W., 1994: Sensitivity of a World Ocean GCM to changes in subsurface mixing parameterization. J. Phys. Oceanogr., 24, 1256-1279. Kushnir, Y., 1994: Interdecadal variations in the North Atlantic sea-surface temperature and associated atmospheric conditions. J. Climate, 7, 141-157. Isemer, H-J., and Hasse, L., 1991: The scientific Beaufort equivalent scale: effects on wind statistics and climatological air-sea flux estimates in the North Atlantic Ocean. J.Climate, 4, 819-836. Landsteiner, M.C.et all, 1990: On the sensitivity of Sverdrup transport estimates to the specification of wind stress forcing in the tropical Pacific. J. Geophys. Res. 95, C2,16811691. Levitus, S., 1982: Climatological atlas of the World Ocean. NOAA Prof. Paper-13, U.S. Government Printing Office., Washington, 174 p. Levitus, S., 1989: Interpentadal variability of temperature and salinity in the deep North Atlantic, 1970-74 versus1955-59. J. Geophys. Res., 94, 16125-16131. Levitus, S., and Gelfeld, R.D., 1992: National Oceanographic Data Center inventory of physical oceanographic profiles. U.S. Department of Commerce, NOAA, USA, 242 pp. Marotzke, J., 1997: Boundary mixing and the dynamics of three-dimensional thermohaline circulation. J. Phys. Oceanog., 27, 8. 17131728. Moore, A.P., Gordon, H.B., 1994: An investigation of climate drift in a coupled atmosphere-oceansea ice model. Climate Dynamics, 10, 81-96. Parilla, G. et all, 1994: Rising temperature in the subtropical North Atlantic Ocean over the past 35 years. Nature, 369, 48-51. Phillips, T.J.,1994: A summary documentation of the AMIP models. PCMDI Report No.18, Livermore, 343 pp. Polonsky, A.B., 1989: Upper mixed layer in horizontally inhomogeneous ocean and its modelling (in Russian). VNIIGMI-MCD, Obninsk-Sevastopol, 234 p. Polonsky, A.B. 1997: The problems in study of the low-frequency variability in the coupled oceanatmosphere system due to the inadequacy of historical datasets. Preprint volume of the 10 th Conference on Applied Climatology, (20-23 Oct.1997, Reno). AMS, Boston, MA, 240-243. Polonsky, A.B., Baev, S.A., Diakite, A., 1992: On the response of the ocean upper layer to synoptic variability of the atmosphere. Dyn. Atmos. Oceans, 16, 225-248. Polonsky, A.B. et all, 1995: Global manifestation of ENSO events and long-term coupled ocean-atmosphere variability. pp. 94-95 in Abstracts of International Scientific Conference on TROPICAL OCEANS GLOBAL ATMOSPHERE (TOGA 95), 2-7 April 1995, Melbourne, Australia. Polonsky, A., and Kuzmin A. (1999). What regulates the decadal-scale variability in the North Atlantic. Ocean. Meteorology and Hydrology, (submitted). Polonsky, A.B., and Voskresenskaya, E.N.,1996a: Trend-like variability in the coupled ocean-atmosphere system in the North Atlantic in 1957 to1990. GAOS, (submitted). Polonsky, A.B., and Voskresenskaya, E.N., 1996b: Low-frequency changes of meridional Ekman transport in the North Atlantic, (in Russian). Meteorology and Hydrology, No.7, 89-99. Power, S., Tseitkin, F., and Dix, M. 1995: Stochastic variability at the air-sea interface on decadal time scales. Gephys. Res. Lett., , 22, 19, 2593-2596. Read, J.-F., and Gould, W.J., 1992: Cooling and freshening of the subpolar North Atlantic since the 1960's. Nature, 360, 55-57. Reynolds, R.M., et al., 1995: Comparison of the effects of different wind stresses on tropical Pacific model sea level. WCRP/WMO No.717, 896-900. Robeson, S. M., 1995: Resampling of networkinduced variability in estimates of terrestrial air temperature change. Climate Change, 29, 213229. Rumstorf, S. 1995: Bifurcations of the Atlantic thermohaline circulation in respons to changes in the hydrological cycle. Nature, .376, 145149. Stouffer R.J., Manabe S., and Vinnikov R.Ya., 1994:Model assessment of the role of natural variability in recent global warming. Nature, 367, 634-636. Trenberth,K. E., Large, W. G., and Olson, J. G.,1989: A global ocean wind stress climatology based on ECMWF analyses. NCAR Tech. Note NCAR/TN-338+STR, 93 p. Trenberth, K.E., Large, W.G., and Olson, J.G., 1990: The mean annual cycle inglobal ocean wind stress. J. Phys. Oceanogr., 20, 1742-1760. Tziperman, E., Cane, M.A., and Zebiak, S.E., 1995:Irregularity and locking to the seasonal cycle in an ENSO prediction model as explained by the quasi-periodicity route to chaos. J. Atmos. Sci.,52, 293-306. Voskresenskaya, E.N., and Polonsky, A.B., 1994: Low-frequency variability of sensible and latent heat fluxes in the North Atlantic in 1957 to 1990 (in Russian). Meteorology and Hydrology, No.9, 58-69. Weber, S.L., 1998: Parameter sensitivity of a coupled atmosphere-ocean model. Clim. Dyn., 14, 3, 201-212. White W.B., 1995: Design of a global Observation System for Gyre-Scale Upper Ocean Temperature Variability. Progress in Oceanography, 36, 3, 169-218 Wolter, K., and Hastenrath, S., 1989: Annual cycle and long-term trends of circulation and climate variability over the tropical oceans. J. Climate, 2, 1329-1351. Wright, P.B., 1986: Problem in the use of ship observations for the study of interdecadal climate changes. Mon.Weather Rev., 114, 1028-1034.