1 Introduction

The atmospheric meridional circulation (MC) plays a greater role in the poleward transport of momentum and heat than wave-induced eddy fluxes. It links the equatorial trough and subtropical ridge making it an important feature for study in the context of climate variability (Chiang and Vimont 2004; Diaz and Bradley 2004). Research has focused on its acceleration, multi-decadal variability and poleward expansion under global warming (Chen et al. 2002; Held and Soden 2006; Zhou and Wang 2006; Frierson et al. 2007; Lu et al. 2007; Hu and Zhou 2010; Stachnik and Schumacher 2011; Tao et al. 2016; Staten et al. 2018; Watt-Meyer et al. 2019; Grise and Davis 2020). Principal Component analysis of atmospheric streamfunction (over height and latitude) reveals a primary mode comprised of equatorially symmetric meridional overturning and a secondary mode with seasonal asymmetric rotors (Dima and Wallace 2003; Feng et al. 2011) featuring inter-decadal variability linked to sea surface temperature (Ma and Li 2008; Feng et al. 2013; Quan et al. 2014; Guo et al. 2016).

Besides the streamfunction metric, tropopause height has been used to infer changes in the extent and intensity of the MC (Seidel et al. 2008; Davis and Birner 2013) and impacts of global warming. Model simulations have linked the MC expansion to greenhouse forcing (Hu et al. 2013), aerosols and stratospheric ozone (Kang et al. 2011), variability of the polar vortex (Garfinkel et al. 2015), the El Niño–Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) (Allen and Kovilakam 2017).

MC expansion contributes to drier conditions in subtropical latitudes (20–35°) and a retreat of extratropical storm tracks and jet streams (Hu and Fu 2007; Archer and Caldeira 2008; Fu and Lin 2011; Bender et al. 2012; Nguyen et al. 2013; Amaya et al. 2018), impacting communities and ecosystems mostly in the northern subtropics (Liu et al. 2020). Although MC widening, as defined by the outer edge of overturning and tropospheric streamfunction, is widely accepted (Davis and Rosenlof 2012; Lucas et al. 2014), less is known about changes of intensity and links to regional ocean-atmosphere coupling.

Among the characteristics of the MC is conservation of angular momentum particularly in the upper limb where zonal winds are weak. If the extra-tropical atmosphere is in radiative equilibrium and eddies do not penetrate equatorward, the MC will exhibit minor longitudinal structure. The overturning is stronger on the winter side due to surplus solar heating in the summer hemisphere. Thermal forcing of the meridional (V) overturning circulation is derived from (Satoh 1994): dV = ((g H ρ ⁄ 2 T) dT ⁄ dy) dt. A scale analysis of the mean state (shown in Fig. 1a, b) yields: g = 10 m s-2, H = depth of convective heating = 6 103 m, ρ = layer density = 0.5 kg m-3, T = temperature = 270 K, dT ⁄ dy = meridional gradient = 1 K ⁄ 106 m, over one day dt = 105 sec. Lower inflow and upper outflow forms a conveyor that feeds tropical uplift: dV ⁄ dy = 5 ⁄ 2 106 = dW ⁄ dz = 3 10−2 ⁄ 6 103, neglecting dissipation. Naturally, the idealized symmetric forcing is altered by land-sea distribution and continental monsoons.

Fig. 1
figure 1

S-N height sections of (1960–2023) long-term mean MC (total vectors max 2 m/s, temperature departure red contours > 1 C): a globally averaged all longitudes, b inter-America 40–90 W with elevation profile 70–75 W; vertical motion exaggerated 30-fold. c Streamfunction SVD time-scores forming the MC index: PC-1 total (blue) PC-2 anomaly red. At right is the inter-America domain and area of deepest convection over the Amazon Basin (blue circle). Note that height sections use a linear P-scale

Full size image

Here the focus is on the MC across the inter-Americas (Fig. 1c) from June to October when convection shifts into the Caribbean (Jury and Rodriguez 2011; Maldonado et al. 2017; Chikamoto et al. 2020; Liu et al. 2020) during the quiescent phase of the Amazon Monsoon (Vera et al. 2006). Empirical evidence is gathered on spatial structure, temporal variability and teleconnections, and statistical analyses are interpreted to address scientific questions on: (i) how does the inter-America MC differ from global (40–90 W vs. all), (ii) are air-sea interactions in-phase across the east Pacific and west Atlantic, (iii) how do strong and weak MC regimes affect the Caribbean climate in boreal summer, (iv) what ocean-atmosphere processes and teleconnections emerge, (v) what is the long-term trend of the inter-America MC. The paper proceeds to describe data and methods, followed by empirical statistics subdivided into singular value decomposition, point-to-field correlations, composites of fast minus slow MC, and linear trends. Outcomes are consolidated in a summarizing discussion.

2 Data and methods

The MC across latitudes 34–35 N is described via NCEP1 reanalysis (Kalnay et al. 1996) meridional wind, vertical motion and streamfunction over the period 1960–2023. Long-term means are calculated and a comparison is made for all longitudes ‘global’ and 40–90 W ‘inter-America’ (Fig. 1c). Low resolution NCEP1 fields identify key features of the MC since 1960; comparisons with NCEP2 found little difference. Atmospheric field data from ERA5 (Hersbach et al. 2020) at 25 km resolution include surface zonal winds, sea-level air pressure (SLP), and rainfall. SST is analyzed at 100 km resolution via Hadley Centre reanalysis, and total column ozone is derived from ESA multi-satellite reanalysis (VanDerA et al. 2015) as a measure of entrainment. Table 1 lists the dataset characteristics.

Table 1 Listing of datasets employed
Full size table

Singular value decomposition (SVD) with varimax rotation was applied to monthly NCEP1 streamfunction along a meridional height section averaged 40–90 W within 34–35 N latitude, 1000 − 100 hPa height and 1960–2023 time. The SVD calculation (using the IRI Climate Library analysis tool) was done independently for the total and anomalous streamfunction to describe the seasonal and standardized components of the MC. The SVD loading patterns exhibit opposing rotors, representing equatorially symmetric cells consistent with the long-term mean. The variance captured in these modes is: 71% for PC-1 total and 35% for PC-2 anomaly. The 768 month record has an average and standard deviation for PC-1 = 1.0 and 0.24 compared with PC-2 = 0.0 and 1.0. The total streamfunction is subordinate to anomaly influence (Fig. 1c) and is dominated by an annual crest in Dec-Apr corresponding with the Amazon Monsoon. The SVD time-scores of PC-1 total and PC-2 anomaly are averaged (in equal proportions) to form an index of the inter-America MC. Combining the total and anomaly values is advantageous because it retains seasonal influences on inter-annual fluctuations. High and low values indicate the opposing rotors spin up and down, respectively. The MC index is constrained by static SVD patterns animated by seasonal and asymmetric shifts of the overturning circulation.

Wavelet spectral analysis (using the Climate Explorer KNMI analysis tool) is applied to the inter-America MC time-score to determine significant temporal cycles within the 64 year record. The time-score is regressed onto boreal summer Jun-Oct fields of HAD SST, ERA5 10 m zonal (U) wind, SLP, and rainfall, and ESA multi-satellite ozone over the area 34–35 N, 130 W-0E. The time-score is cross-correlated with a variety of known climate variables (1960–2023) and require |r| > 0.20 to reach 95% confidence with ~ 100 degrees of freedom. Ranking the SVD time-score, the top-10 and bottom-10 Jun-Oct seasons are identified (Table 2a) to construct S-N height section composites of fast minus slow NCEP1 MC: zonal wind, relative vorticity, divergence, air temperature, specific humidity, and depth sections of GODAS ocean salinity and sea temperature. Lastly the Jun-Oct MC averaged 40–90 W is analyzed for linear trend 1960–2023 using the Jun-Oct V and W wind components to create vectors in a S-N height section.

Table 2 (a) top 10 high and low MC index years for Jun-Oct seasons used in the composite analysis (Figs. 5a-c, 6a-c and a-c and 7a and a), (b) correlation of the Inter-america MC time-score (Fig. 2c) with precribed climate indices N = 768, namely: Atlantic Multi-decadal Oscillation, Arctic Oscillation, Madden-Julian Oscillation at 120 W Pacific, North Atlantic Oscillation, Nino3 SST, Pacific North-America, and Southern Oscillation Index; as defined by the Climate Prediction Center and extracted from the Climate Explorer KNMI website. Significance thresholds are |r| > 0.20 for 95% confidence; weak correlations with other climate indices are omitted
Full size table

3 Results

3.1 Long-term mean MC

The global Hadley circulation over all longitudes and time 1960–2023 is compared with the 40–90 W inter-America cells in Fig. 1a, b. Combined V and W wind components represent the global cells as rectangular and asymmetric, with equatorial convergence and uplift centered on 7 N due to the northward land distribution. The subsident airflow on the subtropical edge is located at 28 S and 28 N (Fig. 1a). The southern Hadley cell is wider than northern, partly due to the SW monsoon over the Indian Ocean. The global meridional overturning has inflow 1000 − 900 hPa and outflow 250 − 150 hPa. Air temperature departures from the (all-longitude) tropical mean profile show + 2ºC from 1000 − 300 hPa with northward bias in the lower troposphere.

The long-term mean inter-America MC (Fig. 1b) is quite different (cf. Li et al. 2023). The opposing rotors are shallow and do not make a full circle. The layers of inflow and outflow are mixed and equatorial convergence and uplift is slanted southward over the Amazon Basin. The subtropical edge of the southern MC reflects overturning with sinking motion over 30 S, but subsidence in the northern cell is disconnected over 25 N. Air temperature departures from the tropical mean profile show two layers of + 2ºC inferred to be solar-forced in the layer 1000 − 700 hPa and convective-forced in the layer 600 − 300 hPa. Unlike the global Hadley circulation, the MC in longitudes 40–90 W is symmetrical about the equator due to the presence of South America with its varied topography and Amazon Monsoon. Low-level inflow and uplift over the inter-Americas is near the equator because the land distribution is centered on 5 S (Fig. 1b, c).

3.2 Inter-America MC variability

The outcome of SVD is presented in Fig. 2a, b in the form of S-N height sections of streamfunction averaged 40–90 W for PC-1 total and PC-2 anomaly. The loading patterns feature opposing quasi-symmetric rotors. The total streamfunction has cells centered on 650–700 hPa at 15–18 S and 750–800 hPa at 20–22 N. The upper jet streams intrude with inverse loadings at 200 hPa at 34 S and 35 N, creating a quadrature pattern. The anomaly loading pattern features broader cells centered on 700 hPa at 22 S and 650–700 hPa at 27 N, without nodes in the upper level. The streamfunction SVD loading patterns show that part of the southern MC intersects the Altiplano.

Fig. 2
figure 2

S-N height sections of SVD loading pattern for monthly streamfunction 40–90 W: a PC-1 total with rotary icons, b PC-2 anomaly, c combined time-score used to distinguish fast and slow MC, d wavelet spectral energy shaded from 90–99% confidence (blue to red) within the cone of validity (grey mask). Symbols in c) identify anomalous cases in Jun-Oct 2011 and 2016

Full size image

The combined inter-America MC time score (Fig. 2c) forms the basis for statistical work. It exhibits annual spikes as the rotors shift across latitudes during active and passive phases of the Amazon Monsoon. Spells of weak MC are noted in 2010–2011. Wavelet spectral analysis of the combined SVD time score (Fig. 2d) illustrates steady pulsing of the annual cycle, with a hiatus in the 1990s when quasi-biennial oscillations prevailed. Three year cycling was strong in the 1960s, growing to 4 year cycles in the 1980s and 5 year cycles in the 2010s. In the period 1990–2020 a 7–8 year inter-decadal cycle was sustained as quasi-biennial oscillations waned.

Temporal cross-correlations of the MC index with known climate variables are presented in Table 2b. Values are weak with respect AMO, AO, NINO3, SOI, PNA, (–0.223, 0.269, –0.289, –0.244, + 0.207, respectively) and slightly higher with MJOp (–0.342) and NAO (+ 0.334). It can be inferred that the inter-America MC tends to accelerate during La Nina, but correlations are weak and suggest the inter-America MC operates independently from ENSO.

To verify the sense of rotation, Fig. 3a, b presents S-N height sections in boreal summers with low − 0.316 (2011) and high + 0.417 (2016) MC index, in the recent period with improved data assimilation. There is little structural difference in the southern hemisphere, except for anomalous poleward outflow above 300 hPa in 2011 (Fig. 3a). The northern MC is reversed and exhibits anomalous poleward rising motion 800 − 300 hPa at 15–25 N in Jun-Oct 2011. In the high year 2016, the northern MC has vigorous overturning with anomalous inflow and subsidence 1000 − 500 hPa at 10–25 N and anomalous rising motion 800 − 500 hPa at 0–5 N. Thus, a high index for the inter-America MC in boreal summer reflects accelerated overturning between the Caribbean and Amazon.

Fig. 3
figure 3

S-N height Sects. 40–90 W of anomalous Jun-Oct MC in: a 2011 with − 0.316 index, and (b) 2016 with + 0.417 index; vertical motion exaggerated 30-fold (vector max 1 m/s). Rotors identify changes in the northern MC in anomalous boreal summers

Full size image

3.3 Point-to-field correlations

Using the combined time-score (Fig. 2c), field regressions were calculated with SST, SLP, U wind, rainfall and ozone in Fig. 4a-d and focus on Jun-Oct seasonal responses 1960–2023. Sea surface temperatures show cooling with respect to high inter-America MC index, not only across the tropical north Atlantic 10–80 W, 5–25 N but in the east Pacific cold tongue; both basins are in-phase. Trade winds are stronger over the Caribbean 10–20 N, 30–100 W during accelerated MC as expected (Wang 2007), and driven by a ridge of higher pressure 20–25 N, 30–70 W (according to regression). Rainfall declines across the tropical Atlantic and east Pacific (Fig. 4c), indicating weakened convection in transient easterly waves during boreal summer. Increased rainfall is noted over South America with anomalous westerly winds off Peru, beneath an accelerated inter-America MC. The SST and rain signatures do not reflect the ENSO dipole (Jury 2023), indicating that anomalous MC generates zonally widespread features that may be sustained via air-sea interactions. An accelerated MC lifts low ozone air in the tropics (10 S) and promotes stratospheric injections of high ozone air in the subtropics (Fig. 4d). These features create a zonal banding of total column ozone which identify entrainment associated with the MC.

Fig. 4
figure 4

Point-to-field correlation maps with respect to the Jun-Oct 1960–2023 MC time-score (Fig. 2c) for: (a) SST, (b) U wind with arrow icons and zone of high pressure labelled, (c) rainfall, (d) total column ozone. Zone of r > 0.6 for sea level air pressure is labelled in b). Note that color bar is reversed in c) and range differs in d). Neutral shading refers to statistically insignificant results

Full size image

3.4 Atmosphere-ocean composites

Using the 10 highest and 10 lowest inter-America MC time-scores (Table 2a), composite fast minus slow circulation illustrates the forcing of the boreal summer climate as S-N height sections in Fig. 5a-c. The MC differences reveal low level inflow 0–20 N and converging uplift 900 − 500 hPa over 5 S (Amazon Basin). Equatorward subsidence is prominent over 800 − 400 hPa at 25–30 N. Zonal wind differences reflect a deep layer of easterlies 500–800 hPa at 10–15 N overtopped by westerlies. The induced shear is favourable to equatorial MJO, but unfavourable to easterly waves in the Caribbean (as indicated by Fig. 4c). Except for the Amazon (600 hPa, 5 S), humidity differences indicate widespread drying under an accelerated inter-America MC, particularly over the Caribbean 1000 − 500 hPa, 10–27 N. Analyzing Caribbean hurricanes from the NOAA Hurdat database: 11 occurred under high MC while 22 occurred under low MC index.

Fig. 5
figure 5

S-N height sections 40-90W of composite high minus low Jun-Oct MC: a meridional vectors (max 0.5 m/s), b zonal wind (m/s), c specific humidity difference(g/kg); based on years listed in Table 2. Small vectors and neutral shading denote statistically insignificant results

Full size image

Continuing the composite analysis of fast minus slow MC, the S-N height Sects. 40–90 W Fig. 6a-c, reveal a quadrature pattern in relative vorticity in the 100–400 hPa layer. An accelerated MC conspires with the subtropical jets in Jun-Oct (cf. Fig. 5b) to produce alternating centers of action in the upper troposphere at 25 S, 12 S, 10 N, 25 N; which maintain subsidence in the outer limb. The Jun-Oct composite relative divergence pattern exhibits 950 hPa positive and 150 hPa negative differences over the Caribbean that support subsidence 10–25 N. Over the Amazon (5 S) inflow to the equatorial trough is shallow and divergence appears in the layer 700 − 600 hPa. Air temperature differences show radiative cooling across the MC most prominently in the 500 − 200 hPa layer at 25 S and 25 N and also over the Caribbean 1000 − 800 hPa (cf. Yanai and Tomita 1998). On the southern edge 30 S and above 200 hPa, NCEP1 temperature differences are positive with respect to fast minus slow inter-America MC in boreal summer.

Fig. 6
figure 6

S-N depth Sects. 40-90 W of composite high minus low Jun-Oct MC: (a) relative vorticity (10−6 s−1), (b) relative divergence (10−6 s−1, smoothed), and (c) air temperature difference (C, detrended); based on years listed in Table 2. Curved line in a) highlights upper quadrature pattern, where – cyclonic south, + cyclonic north. Neutral shading denotes statistically insignificant results

Full size image

The desiccation of atmospheric moisture (as seen in Fig. 5c) has a profound effect on the subsurface ocean, as seen in the high minus low MC index S-N depth composite (Fig. 7a). Salinity is 0.25 ppt above normal, as expected, but the signal extends to 100 m depth indicating a long fetch for air-sea interactions. Increased salinity supports turbulent mixing that deepens the layer of sea temperature differences < − 0.5 C to 150 m from 5 to 20 N, far greater than could be attributed to trade wind evaporation. The salinity signal offers feedback to sustain the inter-America MC beyond seasonal time scales.

Fig. 7
figure 7

a S-N depth Sects. 40–90 W of composite high minus low Jun-Oct ocean salinity (shaded) and sea temperatures < -0.5 (dashed blue). b S-N height Sects. 40–90 W of linear trend in the Jun-Oct 1960–2023 MC (vector, max 0.03 m s−1/yr) with elevation profile. Small vectors and neutral shading denote statistically insignificant results

Full size image

3.5 Linear trends

Calculating the linear trend in the inter-America MC (S-N height section, Fig. 7b) a robust feature emerges: the northern MC has been accelerating in boreal summer since 1960. Subsidence strengthened in the 700 − 300 hPa layer from 10 to 25 N while rising motions accelerated in the 600 − 300 hPa layer at 5 N, together forming a tight rotor centered on 500 hPa at 10 N. Another feature of the long-term trend was the southward slant of the southern MC connecting subtropical subsidence and uplift over the Amazon. The Jun-Oct 2016 anomaly pattern (Fig. 2b) looks quite similar. If such a trend persists and is attributable to natural multi-decadal oscillations and anthropogenic climate change, then transient easterly waves over the Caribbean may become less frequent. One might expect long-term drying trends over the Antilles Islands, but recent studies (Jury 2022) have shown that low level moistening from warmer SST sustains rainfall over larger islands and hilly coasts bordering the Caribbean.

4 Summary

Atmospheric convection in the Caribbean sector of the inter-Americas is modulated by trade-wind subsidence and the subtropical easterly waves from June to October when SSTs exceed 27 C (Jury et al. 2007; Herrera et al. 2015; Maldonado et al. 2017; Moraes et al. 2022). Northward migration of the equatorial trough and MC occurs during the quiescent phase of the Amazon monsoon. A MC index was formulated via SVD applied to meridional height section NCEP1 streamfunction 40–90 W, 1960–2023. The combined PC-1 total and PC-2 anomalous time-scores (cf. Fig. 1c) quantified MC intensity over the inter-Americas. Although atmospheric convection in boreal summer is modulated by Pacific ENSO, sustained MC anomalies tend to be decoupled from external forcing (cf. Table 2b). Results here linked intensification of the MC to stronger trade winds, marine evaporation and cooling, and suppressed atmospheric convection over the Caribbean. This outcome was anticipated, yet the east Pacific SST and rain response was in-phase with the tropical north Atlantic, inferring that the Walker Cell-induced ENSO dipole (Diaz and Bradley 2004) may be subordinate during periods of anomalous MC.

The Hadley circulation over the inter-Americas was found to be more symmetric than global due to land-sea distribution. Characteristics of an intensified MC in boreal summer (Jun-Oct) included low-level easterlies and upper level westerly winds in latitudes 10–15 N which would promote equatorial MJO but inhibit subtropical easterly waves. Composite humidity and rainfall declined during intensified MC, inducing higher salinity and cooler sea temperature to 150 m depth across the Caribbean 5–25 N in Jun-Oct season. Composite divergence and air temperature differences exhibited mid-tropospheric outflow and radiative cooling, while the tropopause showed convergence and warming during MC acceleration over the inter-Americas. Long-term trends revealed intensification of the northern MC in boreal summer: rising over the Amazon, subsiding over the Caribbean, inter-connected by lower and upper airflows at 10 N. The faster meridional overturning appears linked to climate change and multi-decadal oscillations but seems decoupled from ENSO and its zonal circulations.

Annual pulsing of the MC conspires with inter-decadal fluctuations to produce many of the anomalous features seen here. The zonally banded ozone correlation pattern (cf. Fig. 4d) indicated an anomalous southward shift of the equatorial and subtropical belts during periods of MC intensification in boreal summer. Further work will consider austral summer, Walker-Hadley interactions, and trends in continental monsoons.