2. Materials and Methods
The data of the NOAA-CIRES-DOE 20th Century Reanalysis version 3 (20CRv3) on a 1° × 1° grid for the period 1900–2015 were used [
56,
57]. It assimilated observational data from the International Surface Pressure Databank (ISPD version 4.7), which is the world’s largest collection of atmospheric pressure observational data [
58]. These surface atmospheric pressure observations are made available through international collaboration with the support of the Atmospheric Circulation Reconstructions over the Earth (ACRE) initiative and the Global Climate Observing System and World Climate Research Programme.
In 20CRv3, the boundary conditions of the sea surface temperature (SST) until 1981 are the daily average data of Simple Ocean Data Assimilation with Sparse Input version (SODAsi.3) [
59], seasonally adjusted using HadISST2.2 climatology for the period 1981–2010 [
60]. In HadISST2.3, regions where sea ice had ever been reported are populated with HadISST2.2 daily averages since 1963 and HadISST2.1 monthly averages interpolated to daily averages for the period 1850–1962. For the period 1981–2015, the average daily SST data from HadISST2.2 were used. The boundary conditions for sea ice concentration were taken from the HadISST2.3 monthly average sea ice data [
57,
60,
61].
The 20CRv3 system uses the NCEP GFS v14.0.1 model with a T254 grid (approximately 75 km at the equator) with 64 vertical levels up to 0.3 mb and 80 individual ensemble elements [
56]. The ensemble filter was used as described in [
56] but with the addition of an updated 4D incremental analysis [
62].
The monthly average data of air temperature at a height of 2 m from the surface (TAS), atmospheric pressure at sea level (SLP), zonal and meridional components of wind speed at a height of 10 m from the surface (SWS), the total amount of precipitation per month on the surface (TAP) and total cloud cover (TCC) were taken from 20CRv3. For the entire investigated time period (1900–2015), at each grid point, the average annual cycles of changes in the hydrometeorological characteristics under consideration were calculated. After that, the calculated annual cycles were subtracted from the initial data at each grid point to obtain anomalies relative to the annual cycle (hereinafter referred to as anomalies).
To characterize TAS anomalies during El Niño and La Niña, an index called the extended oceanic Niño index (EONI) is used [
63]. This index represents the average values of TAS anomalies in the equatorial Pacific Ocean (5° N–5° S, 170°–80° W), including the region adjacent to the Isthmus of Panama. This region has much more vessel-based data than the central Pacific regions used to estimate traditional El Niño indices such as Niño3.4, Niño3 and Niño4. The region over which the EONI is calculated also partially includes the region of the Niño1+2 index.
Methods to identify the modes of climatic variability have been studied by different authors; for example, principal component analysis (PCA) with various modifications and improvements based on nonlinearity and the theory of dynamical systems was applied in [
64]. We applied PCA, also called empirical orthogonal function (EOF) analysis, to find the EOF1 of the variability of TAS and SLP anomalies at interannual periods inherent in the GAO and ENSO (
Figure 1). For this purpose, preliminary band-pass filtering was performed with a Butterworth filter for 2–7 years of the series of TAS and SLP anomalies at each node of the 20CRv3 grid. In addition, to suppress the strong variability of the TAS and SLP anomalies at high latitudes compared to the tropics, the time series at each grid point were normalized to their standard deviations. After that, PCA was applied to the obtained filtered and normalized fields.
The Global Atmospheric Oscillation (GAO) can be characterized by the GAO index (GAO1), which is calculated as the sum of the normalized values of SLP anomalies in 10 geographic regions that coincide with extremes (maxima and minima) in the global EOF1 field of SLP anomaly variability at interannual periods (
Figure 1b). The GAO index is calculated using the following formula: GAO1 = P(5° S–5° N, 145°–155° E) + P(5° S–5° N, 55°–65° E) + P(5° S–5° N, 35°–25° W) + P(55°–65° N, 95°–85° W) + P(65°–55°S, 95°–85° W) − P(5° S–5° N, 95°–85° W) − P(45°–55° N, 175°–165° W) − P(45°–55° N, 15°–5° W) − P(55°–45° S, 15°–5° W) − P(55°–45° S, 175°–165° W), where P is the average of SPL anomalies in regions with the given coordinates [
54]. When calculating GAO1, the normalized values of SLP anomalies in each region are used to equalize their contributions to the final index. Since the GAO1 values are dimensionless, for the convenience of working with this index, it is normalized to its standard deviation.
Since the SPL anomalies in the equatorial regions of the Pacific Ocean (5° S–5° N, 145°–155° E) and (5° S–5° N, 95°–85° W) are used to calculate the equatorial Southern Oscillation index, in order to separate the GAO from the SO, these regions can be excluded from GAO1. Thus, the GAO index without SO is calculated from the SLP anomalies in eight regions using the following formula: GAO2 = P(5° S–5° N, 55°–65° E) + P(5° S–5° N, 35°–25° W) + P(55°–65° N, 95°–85° W) + P(65°–55° S, 95°–85° W) − P(45°–55° N, 175°–165° W) − P(45°–55° N, 15°–5° W) − P(55°–45° S, 15°–5° W) − P(55°–45° S, 175°–165° W), where P is the average of SPL anomalies in the regions with the given coordinates [
54]. Thus, the GAO2 index includes the series of SLP anomalies in eight regions normalized to their standard deviations; that is, SLP anomalies in all eight participating regions make comparable contributions to the GAO2 index. This GAO index without SO (GAO2) is normalized to its standard deviation and is used as the GAO index in this work.
The average oscillation amplitudes of the anomalies of the studied hydrometeorological characteristics between the opposite phases of the GAO were calculated as follows. If the GAO2 values were greater (less) or equal to +0.5 (−0.5) for 5 months or more, then this period of time was attributed to the positive (negative) phase of the GAO. Thus, for the studied period (1900–2015), 30 positive and 25 negative phases of the GAO were identified. The average values of the anomalies of the studied characteristics were calculated for each time interval selected in this way. The mean values and variances were calculated for the obtained two sets of mean values of the anomalies of the studied characteristics at positive and negative phases of the GAO. They were used to calculate the oscillation amplitudes of the anomalies of the studied characteristics between the average positive and average negative phases of the GAO, and using Student’s t-test, their statistical significance was assessed.
Wavelet analysis is useful for analyzing nonstationary signals. A cross-wavelet analysis of the calculated indices GAO2 and EONI was performed using the Morlet wavelet function [
65]. Therefore, the complex form of the continuous wavelet transform C(a,b) is C(a,b) = C(a,b)
Re + C(a,b)
Im at scales a and positions b. To estimate the cross-correlations of the two time series x and y based on the calendar time and time scales of oscillations, the products of their real wavelet components were calculated as follows: C
xy(a,b) = C
x(a,b)
Re × C
y(a,b)
Re [
55,
66,
67].
Additionally, the magnitude-squared wavelet coherence, which is a measure of the correlation between signals x and y in the time–frequency plane, was calculated [
68]. The wavelet coherence of two time series x and y is: |S(C*
x(a,b)C
y(a,b))|
2/S(|C
x(a,b)|
2)·S(|C
y(a,b)|
2), where C
x(a,b) and C
y(a,b) denote the continuous wavelet transforms of x and y at scales a and positions b. The superscript * is the complex conjugate, and S is a smoothing operator in time and scale. For real-valued time series, the wavelet coherence is complex-valued if a complex-valued analytic Morlet wavelet is used.
For the wavelet coherence calculation, the “wcoherence” function of MATLAB ver. R2017b was used (
https://www.mathworks.com/help/wavelet/ref/wcoherence.html (accessed on 20 October 2021)). All other data mining was performed by using programs written by the authors of the manuscript. The wavelet coherence is computed using the analytic Morlet wavelet. Due to the inverse relationship between frequency and period, a plot that uses the sampling interval is the inverse of a plot that uses the sampling frequency. For areas where the coherence exceeds 0.5, plots that use the sampling frequency display arrows to show the phase lag of y with respect to x. The arrows are spaced in time and scale. The direction of an arrow corresponds to the phase lag on the unit circle. For example, a vertical arrow indicates a π/2 or quarter-cycle phase lag. The corresponding lag in time depends on the duration of the cycle.
3. Results and Discussion
To assess the links between the GAO and ENSO, a cross-wavelet analysis of the calculated indices GAO2 and EONI was performed (
Figure 2a). The cross-wavelet diagram of GAO2 and EONI (
Figure 2a) allows the identification of temporal variations that determine their large cross-correlations. The red regions in this wavelet diagram indicate an almost complete coincidence between the phases of the GAO2 and EONI variations. The areas marked in blue correspond to the out-of-phase behavior of these indices. As can be seen, the red areas cover a significant part of the GAO2 and EONI cross-wavelet diagram on time scales from 2 to 7 years. However, there are calendar intervals in which the links between GAO2 and EONI are weakened in some periods of fluctuations and strengthened in other periods. The wavelet coherence of GAO2 and EONI (
Supplementary Materials Figure S1) shows similar results to the cross-wavelet diagram (
Figure 2a).
The time series of GAO2 and EONI with Butterworth band-pass filtering from 2 to 7 years and sliding cross-correlations between them with a window of 8 years (
Figure 2b) demonstrate strong links between the GAO and ENSO on interannual time scales. The observed weakening of these links in the 1940s can be attributed to the paucity of observational data during World War II. When calculating cross-correlations with a window of 8 years, the length of the series involved in obtaining individual correlation values was 96 values (8 times 12 months). However, after applying band-pass filtering, these 96 values can no longer be considered independent, and therefore, the given values of the correlations are purely illustrative and not evaluative. Since even white noise can become cyclical after applying band-pass filtering, the results presented in
Figure 2b should be considered an addition to the cross-wavelet diagram, which is shown in
Figure 2a, and were not subjected to band filtering of the rows after the data were obtained.
Figure 3 maps the oscillation amplitudes of the mean air temperature anomalies at the surface (TAS) between the opposite phases of the GAO and Student’s
t-criterion. TAS anomalies form a global structure that is symmetrical about the equator, taking into account the location of the continents. Along the equator of the central and eastern parts of the Pacific Ocean, positive TAS anomalies are observed, which are typical for El Niño. Weaker values of above-zero anomalies extend from the equator to the north and south along the coasts of the Americas. Reaching high latitudes, they intensify and form two regions of positive TAS anomalies located symmetrically relative to the equator over Alaska and the Amundsen Sea. The wavelet coherence between GAO2 and TAS anomalies in the Amundsen Sea region (55°–70° S, 150°–100° W) (
Supplementary Materials Figure S2) is higher than that between GAO2 and TAS anomalies in the Alaska region (55°–70° N, 170°–120° W) (
Supplementary Materials Figure S3), which may be the result of the ocean–atmosphere interaction in the Amundsen Sea region.
Negative TAS anomalies are located in the western part of the Pacific Ocean with two centers in the middle latitudes of its northern and southern parts. The wavelet coherence between GAO2 and TAS anomalies in northern (30°–45° N, 160° E–160° W) (
Supplementary Materials Figure S4) and southern (20°–35° S, 160° E–150° W) (
Supplementary Materials Figure S5) regions of the western Pacific Ocean middle latitudes is high at periods of 2–7 years throughout the 1900–2015 time interval. However, this is not surprising, since these regions are close to the tropics of the Pacific Ocean, where ENSO events are observed.
TAS anomalies in the Indian Ocean region—Hindustan Peninsula, Southeast Asia, the west of the Indonesian archipelago, Australia, Africa and the Arabian Peninsula—are mostly positive, with the exception of the south and southeast of the Indian Ocean off the coast of Australia, as well as the Tibetan Plateau region, where negative TAS anomalies are observed. The wavelet coherence between GAO2 and TAS anomalies in the Indian Ocean region (0°–20° S, 50°–90° E) (
Supplementary Materials Figure S6) has a gap in its high values in the 1940s, which may be the result of the lack of observations in this region during World War II. In addition, the wavelet coherence between GAO2 and TAS anomalies in the Indian Ocean region is higher at periods of about 2–4 years than at 5–7 years. The wavelet coherence between GAO2 and TAS anomalies in the Hindustan Peninsula region (10°–30° N, 75°–85° E) (
Supplementary Materials Figure S7) has higher values than that between GAO2 and TAS anomalies in Australia (20°–35° S, 120°–150° E) (
Supplementary Materials Figure S8) or Southern Africa (15°–30° S, 20°–30° E) (
Supplementary Materials Figure S9) regions.
In the tropical region of the Atlantic Ocean, negative TAS anomalies are observed, which transform into positive anomalies in the northern and southern subtropics and then revert to negative anomalies in middle and high latitudes. However, the wavelet coherence between GAO2 and TAS anomalies in high latitudes of the Southern Atlantic region (70°–60° S, 60°–30° W) (
Supplementary Materials Figure S10) has a lot of gaps at 2–7-year periods over the 1900–2015 calendar interval. The above-mentioned positive TAS anomalies are located in the northwestern part of North America. At the same time, negative TAS anomalies are observed in the southern and eastern parts of North America. The South American continent is predominantly occupied by positive TAS anomalies, with the exception of its southern part. The wavelet coherence between GAO2 and TAS anomalies in the South American continent region (15° S–5° N, 70°–50° W) (
Supplementary Materials Figure S11) is higher than that between GAO2 and TAS anomalies in the southern and eastern parts of the North American region (25°–45° N, 105°–70° W) (
Supplementary Materials Figure S12).
In the northern part of Eurasia and the Arctic, negative TAS anomalies are observed. In the southern part of Eurasia, including the Mediterranean, Black and Caspian Seas, positive TAS anomalies are observed, with the exception of the previously mentioned region of the Tibetan Plateau. However, the wavelet coherence between GAO2 and TAS anomalies in the northern part of Eurasia and the Arctic (60°–80° N, 10°–150° E) (
Supplementary Materials Figure S13) and the Black and Caspian Sea regions (35°–45° N, 25°–50° E) (
Supplementary Materials Figure S14) is low at 2–7-year periods and unstable throughout the 1900–2015 calendar interval. TAS anomalies in Antarctica are predominantly positive, but they are not very reliable due to the extremely small number of observations in this region.
The global spatial structure of the oscillation amplitudes of mean sea level pressure (SLP) anomalies between the opposite phases of the GAO has symmetry, both relative to the equator and relative to the 90° W meridian (
Figure 4). A vast area of positive SLP anomalies is observed in the low latitudes of South America, the Atlantic Ocean, Africa, the Indian Ocean region, the Indonesian archipelago, Australia and the western part of the Pacific Ocean. The wavelet coherence between GAO2 and SLP anomalies in this region (30° S–30° N, 50° W–170° E) has high values at interannual periods throughout the 1900–2015 calendar interval (
Supplementary Materials Figure S15). This is because the GAO2 index includes SLP anomalies in (5° S–5° N, 55°–65° E) and (5° S–5° N, 35°–25° W) regions.
In the middle latitudes, this large region of positive SLP anomalies is bordered by regions of negative SLP anomalies, which form an X-shaped structure with the center of intersection at the equator of the central-eastern part of the Pacific Ocean. The main centers of negative SLP anomalies are located in the middle latitudes of the Pacific, Atlantic and Indian Oceans. The wavelet coherence between GAO2 and SLP anomalies in middle latitudes of the northern (45°–55° N, 175°–165° W) (
Supplementary Materials Figure S16) and southern (55°–45° S, 175°–165° W) (
Supplementary Materials Figure S17) regions of the Pacific Ocean and in middle latitudes of the northern (45°–55° N, 15°–5° W) (
Supplementary Materials Figure S18) and southern (55°–45° S, 15°–5° W) (
Supplementary Materials Figure S19) regions of the Atlantic Ocean has high values at interannual periods, but a lot of gaps are evident during the 1900–2015 calendar interval. In the center and east of Asia, negative SLP anomalies are also observed, but they are much weaker than the centers of negative SLP anomalies over the oceans, which may serve as evidence of the leading role of the ocean–atmosphere interaction in the physical mechanism of the GAO.
At high latitudes to the north and south of the center of the intersection between the branches of the X-shaped structure of negative SLP anomalies, there are two regions of positive SLP anomalies. The first region covers most of North America and the entire Arctic, and the second one—the region of the Amundsen, Bellingshausen and Weddell Seas—covers all of Antarctica. The wavelet coherence between GAO2 and SLP anomalies in the Amundsen, Bellingshausen and Weddell Sea region (70°–55° S, 130°–70° W) (
Supplementary Materials Figure S20) is higher than in the North American region (40°–70° N, 110°–80° W) (
Supplementary Materials Figure S21), which may be the result of the ocean–atmosphere interaction.
In many respects, the fields of the oscillation amplitudes of the mean TAS (
Figure 3) and SLP (
Figure 4) anomalies between opposite GAO phases turned out to be similar to those of the EOF1 of the interannual variability of TAS and SLP anomalies (
Figure 1). However, the application of the principal component method (
Figure 1) implies that the analysis includes all data of the TAS and SLP anomalies, including those corresponding to the neutral states of the GAO and ENSO. The spatial structures of the TAS and SLP anomalies inherent in the neutral states of the GAO and ENSO are inhomogeneous. Therefore, when used in the analysis, they reduce the calculated values of cross-correlations, for example, in [
43]. On the contrary, the composite analysis used to obtain the fields of the oscillation amplitudes of the mean TAS (
Figure 3) and SLP (
Figure 4) anomalies between the opposite GAO phases does not include neutral states of the GAO and ENSO and thus only accounts for differences between more homogeneous TAS and SLP anomalies associated with opposite phases of the GAO (El Niño and La Niña). Therefore, in order to suppress the inhomogeneity of the neutral states of the GAO, when constructing the fields of principal components (
Figure 1), band-pass filtering from 2 to 7 years was applied, which contributes to the isolation of the GAO variability.
The main differences between the EOF1-field of the interannual variability of the TAS anomalies presented in
Figure 1a and the field of oscillation amplitudes of mean TAS anomalies between opposite phases of the GAO (
Figure 3) are observed in the tropics of the Atlantic Ocean. Perhaps this is because the wave property of the GAO propagates from west to east [
44,
45,
46,
47,
54], which is also shown below. The PCA method displays the later phase of the GAO, when the TAS anomalies in the tropics of the Pacific Ocean have already influenced the tropics of the Atlantic Ocean, causing the TAS anomalies of the regions to adopt the same sign. There are no such significant differences in the fields of the SPL anomalies (
Figure 1b and
Figure 4), with the exception of the regions of North America and Eurasia. However, over the continents, GAO anomalies are much less pronounced and more unstable than over the oceans because the GAO is a process of interaction between the ocean and the atmosphere. Without the inertia (memory) of the ocean, high-frequency atmospheric disturbances rapidly destroy the spatial structure of the GAO.
Figure 5 shows the field of oscillation amplitudes of the mean wind speed anomalies near the surface (SWS) between the opposite phases of the GAO, and
Figure 6 and
Figure 7 show the fields of its zonal and meridional components. In the western and central parts of the near-equatorial region of the Pacific Ocean, anomalies of the westerly wind are observed, characteristic of weakened Pacific trade winds.
Figure 4 shows negative anomalies of the SPL in the Aleutian minimum region. The SWS anomalies correspond to these negative SLP anomalies and form the cyclonic SWS anomaly in the North Pacific Ocean. In the region of the South Pacific that is symmetrically located relative to the equator, a cyclonic anomaly of the SWS is also observed but is slightly smaller in size. It is possible that its smaller spatial scales are associated with the adjacent positive SLP anomalies and the anticyclonic SWS anomaly in the region of the Amundsen, Bellingshausen and Weddell Seas. In the North American region, however, such large-scale anomalies of the SLP and SWS are not observed, which confirms the decisive role of the interaction between the ocean and the atmosphere in the formation of the GAO.
In the tropics of the Indian Ocean, SWS anomalies are observed, characteristic of the reverse of the Walker circulation cell in this region, with the largest east wind anomalies off the coast of Sumatra. However, these easterly wind anomalies do not cause the SAT anomalies characteristic of the Indian Ocean Dipole (IOD) (
Figure 3). As shown in [
55], the largest temperature anomalies characteristic of IOD occur during the positive phase of the GAO at depths of 50–100 m due to a rise in the thermocline off the coast of Sumatra. In the middle latitudes of the southern part of the Indian Ocean, cyclonic anomalies of SWS are observed, corresponding to negative anomalies of SPL in this region.
In the middle and high latitudes of the North Atlantic, there is a cyclonic SWS anomaly centered over the British Isles (
Figure 5). It corresponds to the NAO-like structure of the SPL anomalies (
Figure 4); however, it shifts somewhat to the southwest in comparison with the spatial structure of the NAO, with the center of negative SLP anomalies in the Icelandic minimum region with a positive NAO phase. Thus, the northeastern wind anomalies arising during the positive phase of the GAO impede the transfer of heat and moisture from the North Atlantic to the northern part of Eurasia and the Arctic and cause negative SAT anomalies in this vast region (
Figure 3). Cyclonic SWS anomalies in the Aleutian Minimum region with northerly wind anomalies in its western part, covering the region of northeast Asia, also contribute to negative SAT anomalies in northeast Asia. The northeastern wind anomalies and negative SAT anomalies observed in the Greenland Sea and Labrador Sea during the positive phase of the GAO may cause an intensification of deep convection in this region.
While the northern part of the cyclonic SWS anomaly centered over the British Isles causes negative TAS anomalies in northern Eurasia during periods of positive GAO phases, the southern part of this cyclonic anomaly, due to southwestern wind anomalies, contributes to positive TAS anomalies in southern Europe; Mediterranean, Black and Caspian Seas; and Central Asia region. The region of cyclonic anomalies of the SWS is also located almost symmetrically relative to the equator in the high and middle latitudes of the southern part of the Atlantic Ocean. In its western part, southerly wind anomalies cause negative TAS anomalies in the Weddell Sea region, thereby contributing to the occurrence of deep convection in its ice-free part.
The largest amplitudes of fluctuations in the total amount of precipitation (TAP) between the positive and negative GAO phases are observed in the tropical belt of the Earth (
Figure 8). During the positive phase of the GAO in the tropics of the Atlantic Ocean, the regions of the Indian subcontinent, South Africa, the Indonesian archipelago, Australia and the tropics of South and Central America, less precipitation falls than during the periods of the negative GAO. In [
69], using a climate model of intermediate complexity, a physical mechanism was identified, which, according to the authors, is the reason for the regional decrease in precipitation at the edges of convective zones during El Niño (positive phase of GAO). According to this mechanism, a warm troposphere increases the moisture content of the surface boundary layer, which is necessary for convection to occur. In regions where moisture is abundant, warm and humid air rises to support rainfall, but this increases the moisture gradient relative to neighboring areas. Thus, a decrease in the amount of precipitation occurs at the edges of the convective zones. This mechanism is the main cause of drought in some tropical regions during El Niño. Positive TAP anomalies during the positive GAO phase are observed in the near-equatorial zone of the Pacific Ocean, and in some regions, negative TAP anomalies are observed at the edges of this zone (
Figure 8). Positive TAP anomalies are observed in the middle latitudes of the Atlantic, Indian and Pacific Oceans, with the exception of the western part of the Pacific Ocean. During the positive GAO phase, positive TAP anomalies are also observed in Europe and Central Asia.
TAP anomalies correlate well with total cloud cover (TCC) anomalies (
Figure 9). Positive TCC anomalies are located in the tropics of the central part of the Pacific Ocean, in the region where the planetary convection region shifts during El Niño. Negative TCC anomalies are located in the tropical regions of Indonesia, Australia, the Indian Ocean, the Indian subcontinent, Africa, the Atlantic and South America. Positive TCC anomalies during the positive phase of the GAO are observed in the middle latitudes of the Atlantic Ocean, North and South America, and Eurasia, where positive TAP anomalies are also located.
The physical mechanism of the GAO can be briefly formulated as follows: during the negative phase of the GAO (
Figure 10a and
Figure 11a), negative SAT anomalies are observed in the tropics of the Indian and Atlantic Oceans, the African continent and the Indonesian archipelago (
Figure 11a–c). In this vast region of negative TAS anomalies, the SPL begins to increase, and in the surrounding middle latitudes, negative SPL anomalies appear—the transition of the GAO to its positive phase (
Figure 10b–f). The positive SPL anomalies arising in the region of the Indonesian archipelago (
Figure 10c–e) cause the displacement of the western SWS anomalies, shifting from this region to the east to the central part of the equatorial region of the Pacific Ocean (
Figure 12c–e), thereby causing a weakening of the Pacific trade winds and triggering the process of El Niño initiation, which further develops due to the Bjerknes feedback and the interaction of Kelvin and Rossby waves. The teleconnections of the strengthened El Niño due to changes in the Walker circulation in turn cause an increase in TAS in the tropics of the Indian Ocean and the African continent (
Figure 11e,f), which leads to negative SPL anomalies in these regions and the transition of the GAO back to its negative phase.
As a hypothesis, the eastward propagation of GAO SLP and TAS anomalies (
Figure 10 and
Figure 11) can be explained by the influence of the Chandler wobble on the movement of the poles of the Earth. The Chandler wobble excites tidal waves (pole tides) in the atmosphere and oceans, propagating from west to east in middle latitudes in antiphase in both hemispheres [
70,
71]. The analyses of satellite data on the sea surface temperature (SST) and sea surface height (SSH) anomalies of the Pacific Ocean carried out in [
63,
72] showed that the North Pacific pole tide, after its reflection from the west coast of Central America, causes positive SST and SSH anomalies in the equatorial region of the Pacific Ocean. Thus, this pole tide can be considered an El Niño trigger [
72,
73]. The continents are not an insurmountable obstacle to the propagation of atmospheric pole tides from west to east. Consequently, since the GAO is affected by the Chandler wobble, it can be expected that the spatial features in the extratropical components of the GAO should also propagate from west to east. Based on this west-to-east spread of the GAO, a predictor index was proposed in [
54,
66,
67] to predict El Niño and La Niña events with a lead time of about one year. Nevertheless, it should be noted that further studies are required to verify the statement that the eastward propagation of anomalies connected with the GAO is induced by the pole tides.
Seasonality plays an important role in GAO correlations and remote responses to ENSO. Strong correlations were found at the 1.5-year period between variations in the Indian Ocean Dipole (IOD) and ENSO [
55], which is the result of the seasonality of these phenomena. In [
67], it was shown that predicting ENSO based on the eastward propagation of the GAO could help to overcome the well-known spring predictability barrier (SPB) of ENSO forecasting. Additionally, based on various data, in [
63,
74], the power spectra of GAO and ENSO indices revealed numerous peaks not only at periods caused by the seasonality but also at the sub-harmonics of the 14-month Chandler wobble in the movement of the Earth’s poles and the super-harmonics of the 18.6-year lunisolar nutation of the Earth’s rotation axis and 11-year sunspot activity. Due to the incommensurability of their periods, all of these external forces act on the climate system at different points in time. As a result, the shape of the power spectra of GAO and ENSO indices looks very complex. It reveals numerous peaks located at multiples of the above periodicities, as well as their sub- and super-harmonics. However, the detected peaks may be random due to the short time series of available observations and may disappear in the future.
The interaction of oceanic baroclinic waves (equatorial Kelvin and Rossby waves and off-equatorial Rossby waves), described by the theory of a “delayed oscillator” [
75], plays an important role in the physical mechanism of ENSO. The movement of oceanic Kelvin and Rossby waves with a deep or elevated thermocline and their reflection from the boundaries in the west and east of the Pacific Ocean forms a cycle of oscillations, which is expressed in the change in the surface temperature of the equatorial part of the Pacific Ocean. In [
76], it was shown that these oceanic baroclinic waves form a single dynamical system whose average period is 4 years (the effective period varies between 1.5 and 7 years). This quadrennial quasi-stationary wave (QSW) is coupled to an annual QSW formed by a first baroclinic, fourth meridional mode equatorial Rossby wave. The annual QSW is resonantly forced by easterlies, while the quadrennial QSW is partly forced by ENSO, which is a component of the GAO.
ENSO is characterized based on the date that the events are mature. Their time lag, defined relative to the central value of successive intervals of 4 years, e.g., 1996–2000, affects their evolution and, for a given amplitude, their variability. It determines the dynamics of the quadrennial QSW in the tropical Pacific since ENSO always occurs at the end of the eastward phase propagation of that QSW [
77]. This new approach allows specifying the modes of interaction between the atmosphere and the oceanic baroclinic waves of the tropical Pacific Ocean. The Kelvin wave transfers the warm waters of the western tropical Pacific to the east. The stimulation of evaporative processes in the eastern Pacific favors a rise in the thermocline and, consequently, the westward recession of the quadrennial QSW.
4. Conclusions
Modern studies of El Niño–Southern Oscillation (ENSO) teleconnections show an apparent impact of ENSO on various climatic processes occurring in various parts of the Earth, that is, the global influence exerted by ENSO. However, they also show that some climatic processes occurring in various regions of the Earth, which are sometimes quite remote from the tropics of the Pacific Ocean, lead to the onset of El Niño or La Niña events and, as data suggest, may have an impact on ENSO.
The recently discovered Global Atmospheric Oscillation (GAO) includes the ENSO itself, as well as forward and reverse teleconnections between ENSO and anomalies of air temperature at the surface, atmospheric pressure at sea level and wind speed on time scales from 2 to 7 years in different regions of the Earth. The GAO is a planetary mode of interannual climatic variability. The spatial structure of GAO anomalies spreads from west to east, both in the tropical belt of the Earth and in middle and high latitudes.
In this work, the amplitudes of fluctuations of global anomalies of air temperature at the surface, sea-level pressure, wind speed at the surface, total precipitation and total cloudiness between the opposite phases of the GAO are investigated. The global structures of the fields of oscillation amplitudes of these anomalies and their high interconnection with each other are shown. The influence of the GAO on the anomalies of the studied characteristics in different regions of the Earth is described. Student’s t-test was used to estimate the probability of these anomalies.
The physical mechanism by which the transition from the negative to the positive phase of the GAO affects the onset of El Niño events is described. Moreover, the further development of El Niño and the intensification of its anomalies, in turn, affect the transition from the positive to negative phase of the GAO.
It was found that the GAO causes record temperature anomalies in the Arctic and in the north of Eurasia, which can have a significant impact on permafrost and, as a result, on housing, railways, and automobile and pipeline infrastructure in this region of Russia.
The analysis of the wavelet coherence between the GAO index and surface air temperature and sea-level pressure anomalies in different regions of the Earth show that this coherence is higher over the oceans than over the continents. This may be evidence of the important role of the ocean–atmosphere interaction in the physical mechanism of the GAO.
The results show indications that the positive GAO phase has an enhancing effect on deep convection in the Greenland Sea, the Labrador Sea and the Weddell Sea. However, additional studies are required to confirm the effect of the GAO on deep convection in these regions.