Dominant Modes Of Variability Research Articles

El Niño in the 1990s

The evolution of the unusual warming events of the first half of the 1990s is discussed based on the National Center for Environmental Prediction ocean re‐analysis data, covering 1980–1995. The warmings occurring in 1991–1992, 1993, and 1994 are viewed as individual episodes, rather than as one long El Niño. It is shown that the warm episodes of the 1990s differ from previous El Niños in two important ways. First, subsurface anomalies are more strongly coupled to the mixed layer across the equatorial Pacific in the 1990s than in earlier years. Second, a persistent warm SST anomaly in the central Pacific has been instrumental in initiating warm events. Both of these factors interfere with the dominant mode of variability in the tropical Pacific air‐sea system, namely the delayed oscillator mechanism. We postulate that the low skill of dynamical atmosphere‐ocean models in predicting the recent warmings is related to a weakened delayed oscillator mechanism.

Stochasticity and Spatial Resonance in Interdecadal Climate Fluctuations

Ocean–atmosphere interaction plays a key role in climate fluctuations on interdecadal timescales. In this study, different aspects of this interaction are investigated using an idealized ocean–atmosphere model, and a hierarchy of uncoupled and stochastic models derived from it. The atmospheric component is an eddy-resolving two-level global primitive equation model with simplified physical parameterizations. The oceanic component is a zonally averaged sector model of the thermohaline circulation. The coupled model exhibits spontaneous oscillations of the thermohaline circulation on interdecadal timescales. The interdecadal oscillation has qualitatively realistic features, such as dipolar sea surface temperature anomalies in the extratropics. Atmospheric forcing of the ocean plays a dominant role in exciting this oscillation. Although the coupled model is in itself deterministic, it is convenient to conceptualize the atmospheric forcing arising from weather excitation as having stochastic time dependence. Spatial correlations inherent in the atmospheric low-frequency variability play a crucial role in determining the oceanic interdecadal variability, through a form of spatial resonance. Local feedback from the ocean affects the amplitude of the interdecadal variability. The spatial patterns of correlations between the atmospheric flow and the oceanic variability fall into two categories: (i) upstream forcing patterns, and (ii) downstream response patterns. Both categories of patterns are expressible as linear combinations of the dominant modes of variability associated with the uncoupled atmosphere.

DECADAL VARIATIONS OF THE SOUTHERN HEMISPHERE CIRCULATION

A data set of monthly mean surface and sea-level pressure observations from a number of stations in the Southern Hemisphere (SH) for the period since 1901 has been used to investigate interannual and interdecadal variations of the SH circulation. A fairly uniform network of 62 stations was selected with data for the period 1955–1985, with a reduced network of 35 stations having data for the longer period 1901–1985. The sparse network has stations mainly on the SH tropical and mid-latitude land masses, with few islands or high-latitude stations. A rotated principal component analysis of the annual mean station pressure anomalies has been used to describe the dominant modes of interannual and interdecadal variations of the SH circulation that can be resolved with the available station network. We show that the sparse station network is able to capture the dominant modes of variability found in the period 1955–1985 using the denser network. The leading mode is associated with the El Niño–Southern Oscillation and is well resolved by the sparse network. This is the leading mode for interannual and interdecadal variations throughout the last century, but there are periods when it has reduced variance and a somewhat different structure, notably 1916–1935. Other modes of interannual variations show marked changes in importance on decadal or longer time-scales.

DECADAL VARIATIONS OF THE SOUTHERN HEMISPHERE CIRCULATION

A data set of monthly mean surface and sea-level pressure observations from a number of stations in the Southern Hemisphere (SH) for the period since 1901 has been used to investigate interannual and interdecadal variations of the SH circulation. A fairly uniform network of 62 stations was selected with data for the period 1955–1985, with a reduced network of 35 stations having data for the longer period 1901–1985. The sparse network has stations mainly on the SH tropical and mid-latitude land masses, with few islands or high-latitude stations. A rotated principal component analysis of the annual mean station pressure anomalies has been used to describe the dominant modes of interannual and interdecadal variations of the SH circulation that can be resolved with the available station network. We show that the sparse station network is able to capture the dominant modes of variability found in the period 1955–1985 using the denser network. The leading mode is associated with the El Nino–Southern Oscillation and is well resolved by the sparse network. This is the leading mode for interannual and interdecadal variations throughout the last century, but there are periods when it has reduced variance and a somewhat different structure, notably 1916–1935. Other modes of interannual variations show marked changes in importance on decadal or longer time-scales.

Process controls and similarity in the us continental‐scale hydrological cycle from eof analysis of regional climate model simulations

AbstractThe surface hydrological output of a regional climate model is investigated with implications for process controls on the spatial‐temporal variability of the water cycle over the continental USA. Principal component analysis was performed on the seasonal and annual hydrological cycles to determine their dominant modes of spatial variability. At both seasonal and annual time‐scales, the first principal component is dominated by precipitation, which controls seasonal wetness and evaporation and accounts for only 52 to 58% of the variability in the continental‐scale hydrological cycle. The second principal component is related to both snowmelt runoff and the time variability of weather (via its influence on the residence time of soil moisture near the land surface) and explains another 22% to 34% of the variability in the hydrological cycle. Based on these findings, a classification of hydroclimatological similarity is proposed in which two areas are similar in their hydroclimatology if their first and second principal components are similar. The classification scheme differs from classical approaches because it is based on dominant modes of variability rather than specific indices such as vegetation or seasonal wetness.

Climatology and natural variability of the global hydrologic cycle in the GLA atmospheric general circulation model

Time average climatology and low‐frequency variabilities of the global hydrologic cycle (GHC) in the Goddard Laboratory for Atmospheres (GLA) general circulation model (GCM) were investigated in the present work. A 730‐day experiment was conducted with the GLA GCM forced by insolation, sea surface temperature, and ice‐snow undergoing climatological annual cycles. Influences of interactive soil moisture on time average climatology and natural variability of the GHC were also investigated by conducting 365‐day experiments with and without interactive soil moisture. Insolation, sea surface temperature, and ice‐snow were fixed at their July levels in the latter two experiments. Results show that the model's time average hydrologic cycle variables for July in all three experiments agree reasonably well with observations. Except in the case of precipitable water, the zonal average climates of the annual cycle experiment and the two perpetual July experiments are alike, i.e., their differences are within limits of the natural variability of the model's climate. Statistics of various components of the GHC, i.e., water vapor, evaporation, and precipitation, are significantly affected by the presence of interactive soil moisture. Even with fixed external conditions, the GHC exhibits a variety of low‐frequency natural variabilities. A long‐term trend is found in the principal empirical modes of variability of ground wetness, evaporation, and sensible heat. Dominant modes of variability of these quantities over land are physically consistent with one another and with land surface energy balance requirements. The dominant empirical modes of ground wetness variability in the annual cycle and perpetual July experiments have similar spatial patterns. The dominant mode of precipitation variability is found to be closely related to organized convection over the tropical western Pacific Ocean. The precipitation variability has timescales in the range of 2 to 3 months and can be identified with the stationary component of the Madden‐Julian Oscillation. The precipitation mode is not sensitive to the presence of interactive soil moisture but is closely linked to both the rotational and divergent components of atmospheric moisture transport. The present results indicate that globally coherent natural variability of the GHC in the GLA GCM has two basic timescales in the absence of annual cycles of external forcings: a long‐term trend associated with atmosphere‐soil moisture interaction which affects the model atmosphere mostly over midlatitude continental regions and a large‐scale 2‐ to 3‐month variability associated with atmospheric moist processes over the western Pacific Ocean.

Principal Component Analysis of Vertical Profiles ofQ1andQ2in the Tropics

Abstract Rotated Principal component analysis (PCA) is applied to the combined vertical profiles of apparent heat source Q1 and apparent moisture sink Q2 from both disturbed and undisturbed periods of the Australian summer monsoon season. The data represent the heating and drying within two radiosonde arrays afforded by the Australian Monsoon Experiment (AMEX), The aim here is to identify dominant modes of variability in combined vertical profiles of Q1 and Q2. Rotation of the principal components (PCs)-done to assure stable, physically meaningful components-yields several PCs, deemed here to be statistically significant. The variation of individual Q1 and Q2 profiles from the mean profile can be expressed as linear combinations of the PCs; therefore, determination of the relative importance of each PC (through examination of its score) during differing convective conditions provides insight into their physical meaning. For instance, the contribution of PC 1 (that mode of variability that explains the max...

Large scale aspects of India-China summer monsoon rainfall

This study investigates the dominant modes of variability in monthly and seasonal rainfall over the India-China region mainly through Empirical Orthogonal Function (EOF) analysis. The EOFs have shown that whereas the rainfall over India varies as one coherent zone, that over China varies in east—west oriented bands. The influence of this banded structure extends well into India. Relationship of rainfall with large scale parameters such as the subtropical ridge over the Indian and the western Pacific regions, Southern Oscillation, the Northern Hemispheric surface air temperature and stratospheric winds have also been investigated. These results show that the rainfall over the area around 40°N, 110°E over China is highly related with rainfall over India. The subtropical ridge over the Indian region is an important predictor over India as well as over the northern China region. '

Phytoplankton pigment patterns and wind forcing off central California

We use a 4‐year time series of high‐resolution coastal zone color scanner imagery to study mesoscale variability in phytoplankton pigment (as a surrogate for biomass) distributions off central California during the spring‐summer upwelling season. We use empirical orthogonal functions to decompose the time series of spatial images into its dominant modes of variability. Similarly, we analyze wind fields derived from the Fleet Numerical Oceanography Center pressure fields. The results from these analyses are used to investigate the coupling between wind forcing of the upper ocean and phytoplankton distributions on mesoscales. Prior to the spring transition, wind stress is dominated by strong northward conditions, and pigment is low and relatively uniform throughout the California Current domain. The first strong southward wind event soon results in a bloom of phytoplankton that is also uniformly distributed offshore. Nearshore values fluctuate relatively little in response to variations in wind forcing. After the spring transition, the wind becomes more steady (southward) and the fluctuations are dominated by changes in the strength of the curl. During southward events, the curl tends to become more positive. Depending on the fluctuations and strength of these curl episodes, filaments begin to form at particular locations. The overall pigment concentration in the California Current drops as a larger fraction of the pigment is distributed within the filaments. The filaments also tend to become shorter during this period, which starts soon after the spring transition and usually reaches its maximum intensity in early summer. Winds (and curl) become much weaker in the mid‐July to August period, and the filaments become much less prominent in terms of concentration and their length increases. Isolated wind events occasionally reinvigorate the filaments. A final bloom in the fall occasionally appears with spatial structure similar to the spring bloom. We conclude that wind forcing, in particular the curl of the wind stress, plays an important role in the distribution of phytoplankton pigment in the California Current. Although the underlying dynamics, especially of the filaments, may be dominated by processes other than forcing by wind stress curl (such as eddies or topographically induced instabilities), it appears that curl may force the variability of the filaments and hence the pigment patterns.

Wind stress curl variability over the North Pacific from the comprehensive ocean‐atmosphere data set

Wind stress curl patterns over the north Pacific, between 3° and 55°N, are calculated from the Comprehensive Ocean‐Atmosphere Data Set (COADS). The mean wind stress curl pattern consists of a basin‐wide band of negative curl south of about 30°N and a basin‐wide band of positive curl to the north. Most of the variance lies in the northern half of the domain, associated with variability in the Aleutian low during winter. The dominant modes of variability are associated with the changes in position and intensity of the Aleutian low and the north Pacific (subtropical) high, each of which extends across the basin, the former in winter and the latter in summer. The seasonal signal accounts for about 40% of the variability in the region of the Aleutian low, about 30% of the variability in the northeast of the basin, and about 60% of the variability in the southwest of the basin between 20° and 30°N. The seasonal signal and first three nonseasonal empirical orthogonal functions (eofs) account for about 60% of the total variability in the mid‐latitude interior Pacific. The first eof represents variations in intensity of the Aleutian low, the second eof represents changes in intensity of the subtropical high and in the latitudinal position of the Aleutian low, and the third eof represents changes in the longitudinal position and orientation of the low. These three eofs together account for about 30% of the nonseasonal variability. On an interannual time scale, a negative (positive) wind stress curl anomaly in winter in the northeast often coincides with a positive (negative) SST anomaly in that region, although the relationship is far from perfect because of the importance of other mechanisms which influence SST variability. There is no distinct relationship between El Nino‐southern oscillation events and wind stress curl anomalies, although weak El Ninos seem to coincide with a positive wind stress curl anomaly across the central north Pacific.

Month-to-Month Variability of the Atlantic Tropical Circulation and Its Relationship to Tropical Storm Formation

Abstract Monthly mean winds have been derived from 200 mb and Analysis of the Tropical Oceanic Lower Layer (ATOLL) winds over the southern North Atlantic, Caribbean, Gulf of Mexico and eastern Pacific during the hurricane seasons (June-November) of 1975 through 1985. After removal of the seasonal cycle, the winds are expressed in terms of empirical orthogonal functions. The dominant mode of variability for the combined 200 mb/ATOLL circulation strongly resembles part of a Walker cell confined near the equator. This mode is strongly correlated with El Nin index (Weare, 1986), and is associated with the.El Nin/Southern Oscillation. A positive (El Nin-like) index tends to be associated with more anticyclonic vorticity at the ATOLL level in the tropics and increases in the vertical shear between about 10° and 30°N. This circulation is unfavorable for tropical storm formation. Correlations are derived between the monthly mean winds and monthly tropical storm frequency in the Atlantic basin. Contemporaneous cor...

Global fluctuations in the distribution of atmospheric mass

Two global data sets have been analyzed to determine the interannual variability of the zonal mean mass distribution. The first is a National Meteorological Center (NMC) global data set of sea level pressures beginning in December 1977 and surface pressures beginning in April 1979. The second is an 11‐year (1972–1982) hybrid set of sea level pressures from the Australian southern hemisphere and U.S. Navy northern hemisphere analyses. The two sets are compared and evaluated by seeing how well they satisfy the constraint of conservation of mass. Anomalies in total mass due to changes in water vapor content of the atmosphere are lost in the data problems. Several inhomogeneities or discontinuities are found and are identified either with changes in analysis procedures in both data sets or the introduction of new orography and operational models at NMC. Consequently only limited periods of data are suitable for determining the dominant modes of variability. Nevertheless, hemispheric mean anomalies in mass in one hemisphere tend to be mirrored in the other, and it is clear that previously deduced fluctuations in hemispheric mean mass are mostly real. Predominant modes of variability of zonal means reveal systematic exchanges of mass (1) almost entirely confined within each hemisphere in isolation, (2) between the extratropics of each hemisphere with little impact on the tropics as an antisymmetric mode, and (3) between the tropics and extratropics in a symmetric mode. The second mode played a strong role in redistributing mass during April–July of 1979 during the First GARP Global Experiment (FGGE) and resulted in an exceptionally deep circumpolar trough over the southern hemisphere and record high pressures over the northern hemisphere. The possible origin of these modes is briefly considered.

Hurricane Climatic Fluctuations. Part II: Relation to Large-Scale Circulation

Correlations are computed between interannual fluctuations of hurricane incidence in the Atlantic basin and large-scale patterns of seasonally-averaged sea-level pressure (SLP; 1899–1978), sea-surface temperature (SST; 1899–1967), and 500 mb heights (Z500; 1946–1978). Dominant modes of interannual variability in average August–September–October (ASO) hurricane incidence are used as measures of overall activity and shifts in activity from region to region. These uncorrelated modes are derived using an empirical orthogonal function (EOF) analysis, as described in Shapiro (1982). The hurricane modes are related to dominant modes of variability in seasonal SLP, SST and Z500, also derived using an EOF analysis. Correlations between the amplitudes of the EOF modes are tested for significance using a measure of artificial skill. May–June–July (MJJ) large-scale SLP anomalies have predictive skill for ASO hurricane activity, significant at the 1.0% level. The correlation predicts about 17% of the variance in activity. Lower SLP precedes more active seasons. Other significant correlations are found: High SST just west of Africa precedes more active seasons, but adds little predictive skill to that of SLP. Relationships between Z500 and hurricane track are consistent with steering concepts, and the results of previous investigators. Weaker westerlies are concurrent with more active seasons.

Temporal and Spatial Scales of the Arctic Circulation

Abstract Daily data for the years 1932–75 are used in a study of the fluctuations in the arctic circulation over time scales of several days to several months. An updated set of normal sea level pressures is constructed, and a semiannual cycle is found in the high-latitude gradient of zonally averaged pressure. July is the only month in which a mean convergence of the low-level flow into the central arctic is indicated. The high-latitude fields of sea level pressure, surface temperature and 700 mb height and temperature are represented in terms of empirical orthogonal functions in order to isolate the dominant modes of variability. The amplitudes of the functions are used to evaluate the daily persistence of arctic pressure anomalies as a function of season and to compare the persistence of arctic and midlatitude pressure anomalies. The month-to-month persistence of arctic pressure anomalies is found to be small, although the monthly persistence does exceed that expected from the lagged autocorrelations o...