Zonal mean wind, the Indian monsoon, and July drying in the western Atlantic subtropics

Authors


Abstract

[1] A fully closed zonal momentum budget is decomposed to explain the occurrence of zonal mean easterlies at subtropical latitudes in July. Eddy momentum fluxes from stationary eddies, most prominently the western sector of the Indian monsoon Tibetan High, are the primary mechanism governing the negative tendency of zonal mean momentum near 20°N–30°N. This strengthening of the zonal mean easterlies in July is significantly correlated with the concurrent strengthening of the North Atlantic Subtropical High (NASH) and the rainfall deficit in the western North Atlantic (WATL). Interannual variations of the Indian monsoon reflect changes in the strength of these zonal mean easterlies, with downstream teleconnections on the westward displacement of the NASH and precipitation in the WATL. An increase in rainfall in India from June to July corresponds to a decrease in rainfall in the WATL.

1. Introduction

[2] The annual march of the seasons is driven by smoothly varying changes in solar insolation as the sun crosses the equator leading to corresponding changes in surface heating. The seasonality of precipitation in the tropical Americas generally reflects these changes in radiative forcing. Deep, moist convection migrates meridionally with the sun, characteristic of monsoon systems over the Americas with a maximum and minimum in rainfall during the summer and winter, respectively [Vera et al., 2006]. Over the western Atlantic (WATL; region defined in Figure 2) however, there are two identifiable peaks in precipitation during the summer rainy season. This bimodal distribution is characterized by a July minimum in precipitation in between the late spring and late summer peaks (Figure 1), and is an instance of a singularity [Wang and Xu, 1997; Mapes et al., 2005], or subseasonal kink in the annual cycle of precipitation.

Figure 1.

Pentad averaged precipitation over the WATL domain (20°N–30°N, 60°W–100°W) for TRMM 3B42 (dashed line) and CMAP (solid line). Climatologies are based on data from 1979 to 2009 for CMAP and 1998–2009 for TRMM.

[3] An intensification of the North Atlantic Subtropical High (NASH) in July seems linked, and is the most widely accepted mechanism attributed to the origin of the midsummer drying in the Atlantic [Hastenrath, 1976, 1978; Granger, 1985; Giannini et al., 2000; Small et al., 2007; Gamble et al., 2007; Gamble and Curtis, 2008]. The NASH may modulate precipitation in the WATL through a variety of processes, both indirect and direct. Directly, sea level pressure (SLP) anomalies lead to changes in low-level divergence through Ekman pumping. The increase in subsidence that accompanies an increase in SLP may lead to changes in the vertical stratification of temperature and/or moisture, leading to changes in convective instability [Knaff, 1997]. The NASH can also directly affect precipitation through changes in water vapor fluxes since moisture transport patterns roughly follow SLP contours. The NASH enhancement in the WATL in boreal summer (hereafter, “summer”) strengthens the easterly trades or Caribbean Low-Level Jet (CLLJ), whose northward branch is the primary source of water vapor for the central United States in summer [Mestas-Nuñez et al., 2007]. The NASH may influence WATL precipitation indirectly through a wind–evaporation–sea surface temperature (WES) feedback [Xie and Philander, 1994; Xie, 1996]: stronger easterly trade winds associated with a stronger NASH lead to greater evaporative cooling (latent heat flux) and lower SSTs, which suppresses convection and may reinforce the NASH further. In a modeling study, Wang et al. [2007] demonstrated that the role of the Atlantic Warm Pool (the warm body of water encompassing the WATL) was to reduce the strength of the NASH and thus the CLLJ. While such process level mechanisms of rainfall reduction still need elaboration, the large-scale intensification of the NASH in the WATL around July seems key.

[4] While the intensification of the NASH appears to be a proximate cause of the July drying, it begs the next question: What causes the midsummer strengthening of the NASH? Several studies [e.g., Rodwell and Hoskins, 2001; Seager et al., 2003; Miyasaka and Nakamura, 2005; Nigam and Chan, 2009] have examined the gross seasonality of the NASH and its Pacific counterpart, for instance, to explain why the subtropical anticyclones are strongest in summer, not winter. However, little research has been done at the refined level of examining subseasonal fluctuations of the NASH on the time scale of interest here (i.e., June to July changes).

[5] The July climatology of the NASH is depicted in Figure 2, which shows the 850 mbar geopotential height (Z850) and wind fields (Figure 2a) along with the corresponding 850 mbar relative vorticity (Figure 2b). The low-level circulation over the North Atlantic is dominated by anticyclonic flow around the subtropical high maintained by Sverdrup vorticity balance: Heating contrasts at the eastern boundary of the Atlantic drive equatorward flow and positive planetary vorticity advection. This B effect is balanced by vortex tube shrinking (negative tendency) associated with subsidence and low-level divergence off the coast. Radiative cooling of low-level clouds further enhances this subsidence [Seager et al., 2003; Miyasaka and Nakamura, 2005], and remote dynamical influences, like Rossby waves emitted from monsoons, may also contribute [Rodwell and Hoskins, 1996; Rodwell and Hoskins, 2001].

Figure 2.

(a) July Z850 contours overlain with the vector wind field. The contour interval is 40 m. (b) The 850 mbar relative vorticity in July. Right-hand and left-hand boxes represent the “EATL” and “WATL” domains, respectively. (c) Time series of 850 mbar vorticity from the EATL and WATL boxes. Daily data are smoothed by a 5 day boxcar average, and all data are from MERRA climatology.

[6] This eastern basin driven view is consistent with the mean eastward location of the NASH, but the southwest corner of the NASH's anticyclonic vorticity extends westward into the WATL in midsummer (Figure 2b). This protrusion is clearly visible in Figure 2c, as the annual cycle of 850 mbar vorticity over the WATL (left box in Figure 2b) has a well-defined midsummer minimum, with a precipitous negative anticyclonic tendency from June to July (Figure 2c). On the other hand, the East Atlantic (EATL, right box in Figure 2b) exhibits a corresponding maximum in vorticity in midsummer (Figure 2c). These out-of-phase changes in the subtropics are consistent with an east-west displacement of the entire NASH circulation, rather than a local expansion on its western periphery. If so, it implicates larger than basin-scale factors in governing the midsummer strengthening of the NASH in the WATL.

[7] We hypothesize the westward displacement of the NASH may be due to a transition from westerly to easterly mean zonal flow in the subtropics in July, resulting in anomalous westward advection of the NASH's anticyclonic vorticity. The planetary zonal mean zonal wind turns easterly near 20°N–30°N in midsummer, which could cause an east-west displacement of the NASH. Figure 3a shows a latitude-time climatology of the vertically and zonally averaged {[u]} (notation listed in Table 1), and a time series of {[u]} averaged across 20°N–30°N is also shown in Figure 3b. Zonal mean easterlies protrude to 30°N in July and August. In Figure 3b, there are two critical points where the zonal flow in the subtropics changes sign: from westerly to easterly (beginning of July) and then back to westerly (middle of September). The timing of these changes correspond to the timing of 850 mbar vorticity changes across the Atlantic basin (Figure 2c). This climatological time signature, and its interannual consistency (section 4), are the basis for our working hypothesis that an east-west seesaw of anomalous vorticity advection is driven by sign changes of the zonal mean flow. The zonal wind at 850 mbar over the Atlantic basin follows the planetary and vertically averaged zonal wind {[u]} (Figure 3b). Since the latter has a more tractable budget, we focus on it hereafter. If our working hypothesis is correct, then we need to understand the story of {[u]} in summer.

Figure 3.

(a) Latitude-time plot of {[u]}cos(ϕ). (b) Time series plot of {[u]}cos(ϕ) averaged across 20°N–30°N (black line) and U850 averaged across 20°N–30°N in the Atlantic sector (red line). The zero line is dashed in both plots and data are from MERRA climatology.

Table 1. Notational Convention and Common Acronyms Used in This Text
SymbolMeaning
Square bracketsplanetary zonal mean
AsteriskZonal eddy (deviation)
Curly bracketsvertical mean
Overbartime mean (monthly)
Primetemporal eddy (deviation)
WATLwestern North Atlantic
EMFDeddy momentum flux divergence
SWTHSouthwest Tibetan High
JMJJuly minus June

[8] This paper is organized as follows: Section 2 briefly describes the observational data used in this study. Section 3 deconstructs a zonal momentum budget to identify the source of midsummer {[u]} anomalies. Section 4 explores the interannual relationship between {[u]}, the Indian monsoon, and implications for the NASH and July drying in the WATL region. Finally, a summary and discussion are given in section 5.

2. Data

[9] NASA's recently released Modern Era Retrospective-Analysis for Research and Applications (MERRA) is the primary data set used in this study [Bosilovich et al., 2006]. Data are collected at daily intervals from 1979 to 2006 at 18 equally spaced pressure levels from 1000 to 50 mbar. Climatologies are defined as the 28 year average from 1979 to 2006. Meteorological fields are projected onto a global horizontal grid, with regular grid spacing of 1.25° in longitude and latitude. The NCEP/NCAR Global Reanalysis version 1 (NCEP-GR1 [Kalnay et al., 1996]) is also consulted in section 4 for intercomparison with MERRA and to bolster the significance of our interannual correlations. NCEP-GR1 has horizontal grid spacing of 2.5° in longitude and latitude. Daily data is collected at 14 unequally spaced pressure levels from 1000 to 50 mbar and vertically interpolated to match MERRA's resolution. Monthly CMAP precipitation data [Xie and Arkin, 1997] over the same 28 year period is also used in section 4.

[10] When vertically integrating, we neglect data below the 950 mbar level in order to mitigate some of the ambiguities of constructing planetary zonal means from a small sample of data points (in the case of MERRA) and from using data extrapolated below the ground (in the case of NCEP-GR1). In constructing a zonal momentum budget in section 3, we use a local Cartesian pressure coordinate system noting the following conversions: dy = a dϕ, where a is the mean radius of the earth (a = 6.37 × 106 m) and ϕ is latitude. (While the equations of motion more appropriately apply to angular momentum, accounting for the effects of spherical geometry exhibit negligible differences at subtropical latitudes in our calculations.) Derivatives are numerically estimated based on a center differencing scheme using three-point, Lagrangian interpolation.

[11] The notational convention used throughout the text is shown in Table 1. Square brackets indicate a planetary zonal mean over the globe and an overbar indicates a time (monthly) average. Asterisks and primes designate deviations (eddies) from these zonal and time means, respectively. Curly brackets represent a vertical average. Frequently used acronyms are also listed in Table 1 for the reader's reference.

3. Zonal Momentum Budget

3.1. Closed MERRA [u] Budget

[12] A momentum budget is deconstructed in this section in order to explicate the extension of zonal mean easterlies into the subtropics in midsummer. Wind tendencies are available in MERRA in addition to standard zonal and meridional wind fields. The total tendency is divided among 5 source terms so that the Eularian rate of change of the zonal mean wind can be written as:

equation image

[13] The first term on the RHS in equation (1) is the zonal wind tendency due to atmosphere dynamics, the sum of the pressure gradient force, Coriolis force, and advection terms. Note that mountain torques are implicit in this term (the residual of the zonal mean pressure gradient where mountains intrude into the atmosphere). Term I will be further broken down in section 3.2.

[14] Term II is due to the gravity wave drag scheme and term III represents the turbulent transport of momentum (mostly in the planetary boundary layer), including surface friction. Together, these two terms constitute physical drag of zonal momentum. The fourth term, includes all moist processes, chiefly vertical mixing of momentum by convection.

[15] Finally DUDTANA, is the “analysis tendency” of the zonal wind, an artificial forcing term introduced during the Incremental Analysis Update (IAU [Bloom et al., 1996]) used in MERRA. DUDTANA reflects the difference between a short-term forecast and the observed (analyzed) zonal wind field at a given time. Its climatology reflects systematic errors in the model.

[16] Figure 4 shows an annual time series of the major terms in equation (1), averaged across 20°N–30°N. The observed {[u]} tendency (LHS of equation (1)) has a decreasing trend from winter to summer with maximum negative values in June which jump sharply to positive values in the middle of July, implying a minimum in {[u]} at this time (consistent with Figure 3). {[DUDTDYN]} goes negative in late May with a minimum around July, as large-scale dynamics exert westward force during the warm season. The annual cycle of {[DUDTTRB]} resembles {[u]} in Figure 3 but is of opposite sign since frictional drag opposes the direction of the flow. In the subtropics, surface drag acts as a momentum source in midsummer, providing positive zonal momentum from the solid earth to the atmosphere when the mean zonal flow goes negative in July and August (Figure 3). {[DUDTANA]} is substantially positive throughout the year, implying systematic model errors. This is consistent with an easterly bias at 20°N–30°N in the free-running (AMIP) native GEOS-5 simulations (not shown). These results are also summarized in Table 2, which lists the value of each term averaged over June and July.

Figure 4.

Daily time series of the major terms in equation (1) for vertical mean tendencies of the zonal wind in MERRA at 20°N–30°N. Daily data are from MERRA climatology and are smoothed by a 5 day boxcar average.

Table 2. Individual Zonal Wind Tendency Fields in MERRA for 20°N–30°N for June to Julya
TermValue
  • a

    Values are vertically and zonally averaged. The summation of all individual tendency terms as well as the actually tendency is also shown. Units are in m/s/month.

DUDTDYN−4.91
DUDTGWD−0.05
DUDTTRB−1.45
DUDTMST−0.14
DUDTANA3.15
Sum−3.40
LHS of equation (1)−3.63

[17] The summation of all 5 tendencies is equal to the total rate of change of {[u]} and is on the order of negative 3–4 m s−1 mo−1 yielding a net acceleration from 1 June to 1 August of around negative 7 m s−1 (Figure 3). Equation (1) represents a fully closed zonal momentum budget in MERRA as the net tendency from the RHS approximately equals the observed tendency from June to July. {[DUDTDYN]} is the largest term at approximately −5 m s−1 mo−1 and is the primary source for the negative {[u]} tendency in summer. {[DUDTTRB]} is a negative source tendency for {[u]} acting to reduce westerly momentum in June by friction. The tendencies from moist physics and gravity wave drag are negligible and can be ignored. {[DUDTANA]} on the other hand is not negligible, and its positive value on the order of 3 m s−1 mo−1 acts to partially offset the negative tendency from {[DUDTDYN]}. As DUDTANA represents unknown model biases, it unfortunately cannot be readily diagnosed. Nevertheless, since {[DUDTDYN]} is the only significant negative tendency, it is decomposed further in section 3.2 to identify the role of eddies as the primary source of the negative mean zonal wind acceleration in summer.

3.2. DUDTDYN in Detail

[18] The zonal mean of DUDTDYN can be written as:

equation image

[19] Term I is the horizontal convergence of zonal momentum by the mean meridional circulation (MMC). This includes both the Coriolis deflection of the Hadley Cell as well as advection of the mean zonal wind by the mean meridional wind. Term II is the vertical advection of zonal momentum by the MMC. Term III is eddy momentum flux divergence (EMFD). (Throughout this text, we are including the minus sign in term III when referencing EMFD). Terms IV and V are the vertical convergence of zonal momentum by the zonal eddies and the mountain torque term. MERRA's pressure level data sets omit grid points under ground so that the zonal mean pressure gradient can be computed on individual isobaric levels. As a check, the mountain torque term was calculated using surface pressure and elevation data following Lorenz and Hartman [2003], and the two techniques yielded nearly identical results.

[20] The dominant terms in equation (2) are EMFD and the horizontal convergence of momentum by the MMC (terms I and III), so for brevity we show only show those in Figure 5 (the values of all terms are listed in Table 3). At upper levels (∼200 mbar) around 20°N–25°N, there is a peak in negative tendency from EMFD which is offset by positive tendency from the MMC. As eddies decelerate the wind aloft, the departure from geostrophic balance drives an MMC as airflows poleward down the meridional pressure gradient. The Coriolis force acting on this resultant MMC results in a compensating eastward acceleration near 20°N–25°N. Furthermore, mass continuity dictates that any mean meridional flow at upper levels is offset by an equal and opposite flow at low levels (neglecting net pole-equator mass shifts). This vertical compensation is identifiable in Figure 5, as the MMC induces a negative acceleration at low levels underneath the positive acceleration aloft.

Figure 5.

Latitude-pressure plot of the primary terms in equation (2) (terms I and III) in June and July. A cosine(ϕ) weighting function is applied, and data are from MERRA climatology.

Table 3. Vertically and Zonally Averaged Values of Each Term in Equation (2) at 20°N–30°N for June to July as Well as for the Actual Model Dynamical Tendencya
TermValue
  • a

    The summation of the five individual source terms is also given. Units are in m/s/month.

10.61
20.96
3−7.32
40.58
50.62
Sum−4.55
DUDTDYN−4.91

[21] The mutual adjustment between eddies and the MMC is well known [e.g., Holton, 2004] and the strong cancellation between the two terms is clearly demonstrated in the Transformed Eularian Mean formulation of the zonal mean circulation [Andrews and McIntyre, 1976; Andrews et al., 1987]. Ultimately, it is the small imbalances between EMFD and the Coriolis torque on the MMC that force zonal mean flow changes at any one level. One of the difficulties in a budget analysis such as this one, however, is the ability to clearly identify the term responsible for the imbalance. This issue is mitigate by considering the vertically averaged form of equation (2). Table 3 lists the column mean values of each of the 5 forcing terms from equation (2) averaged across 20°N–30°N from June to July. EMFD stands out as the single dominant term, as all other terms are an order of magnitude smaller. Hence EMFD is responsible for column mean forcing of the zonal mean zonal wind on the order of −7 ms−1 mo−1 in the subtropics in midsummer. Moreover, the summation of all terms on the RHS of equation (2) is nearly equivalent to the respective {[DUDTDYN]} value, indicating that the model's actual dynamical tendency can be accurately decomposed into its individual source terms from daily resolution pressure level outputs in MERRA. The vertically averaged form of equation (2) can now be expressed as:

equation image

[22] After substituting into our original tendency equation (equation (1)), the MERRA {[u]} budget can be simplified as:

equation image

3.3. The 200 mbar EMFD Distribution

[23] It is evident from above that the EMFD term is the cause for the negative acceleration of the vertically integrated [u] in the northern summer subtropics. Since EMFD acts primarily at upper levels (Figure 5), by extension, the eddy momentum torque at upper levels drives a tendency of the same sign at lower levels, via mass compensation in the MMC. This constraint simplifies our analysis greatly, allowing EMFD to be studied in more detail at upper levels.

[24] Though eddy momentum fluxes are only of formal interest to the [u] budget in its zonal average, analysis of the spatial distribution of u*v* sheds insight into the important regional processes responsible for this zonal mean forcing. Eddy momentum flux can also be further distilled down in time by making use of the simple temporal decomposition:

equation image

total stationary transient.

[25] Here, the total eddy flux for a given month is the product of daily u and v wind fields, the stationary eddy flux is the product of monthly u and v wind fields, and the transient eddy flux is simply the arithmetic difference of the two. In summer, the transient contribution is minimal and the EMFD tendency is well captured by monthly mean wind fields. Hence, we focus on the contribution of stationary eddies to the [u] budget. These summer stationary eddies (or waves) in the subtropics are forced by longitudinal heating gradients characteristic of a monsoon structure in the Northern Hemisphere [Chen, 2003, 2010].

[26] Figure 6a shows contours of 200 mbar stationary eddy momentum flux u*v* averaged over June and July overlain with the total wind field. A map of EMFD is also shown in Figure 6b, with the latitudinal average calculated across 20°N–30°N in Figure 6c. The primary contribution to the zonal mean EMFD is from around 25°E longitude (Figure 7b), which corresponds to the western edge of the Tibetan High (Figure 7a; hereafter, TH). On its northwest corner (∼35°N), the southwesterly flow around the TH interacts strongly with the oncoming westerlies, causing an area of enhanced positive momentum flux over the eastern Mediterranean. To the south (∼20°N), there is negative eddy momentum flux from southeasterlies on the southwest corner of the TH. Hence, near 25°N, there is large EMFD in the meridional direction as southwesterlies pump positive zonal momentum to the north and southeasterlies pump negative zonal momentum to the south. While similar reasoning of the opposite sense holds for EMFD on the eastern flank of the TH (∼100°E), the negative tendency from eddies near 25°E is only partially offset by the positive tendency to the east, due to the anomalously positive eddy momentum flux over the Mediterranean.

Figure 6.

(a) Contoured 200 mbar stationary eddy momentum fluxes (u*v*) in June and July overlain with total vector winds and (b) EMFD (−∂/∂y) of the eddy momentum fluxes in Figure 6a. (c) The longitudinal distribution of EMFD in Figure 6b calculated across 20°N–30°N along with the total value in each sector. Data are from MERRA climatology.

Figure 7.

Map depicting the various important regions and definitions used in this study upon which interannual indices and correlations are based.

[27] There is also a negative, albeit smaller in magnitude, tendency due to EMFD from the Tropical Upper Tropospheric Troughs (TUTTs) in the eastern Atlantic and Pacific basins. TUTTs are cold core lows that extend across the subtropical regions of the oceanic basins [Whitfield and Lyons, 1992]. Due to the meridional shear gradient (westerlies increasing poleward), the positive tilt of the TUTTs result in positive eddy momentum fluxes, with northeasterlies upstream of the trough axis and southwesterlies downstream. Situated above the subtropical anticyclones in summer [White, 1982], TUTTs are thought to be maintained by land-sea contrasts in heating and precipitation leading to cooling and subsidence over the eastern ocean basins. The generation of subsidence that leads to the formation of the stationary TUTT feature may originate from deep monsoon heating over the continents [Rodwell and Hoskins, 2001; Chen et al., 2001; Liu et al., 2004] or from shallow heating cooling couplets across the west coast of North Africa and North America [Miyasaka and Nakamura, 2005]. The presence of TUTTS situated above the subtropical highs are no accident, but rather both are forced by deep subsidence driven by east basin cooling, with the positive vorticity aloft balanced primarily by the zonal advection of negative vorticity [Miyasaka and Nakamura, 2005].

[28] At upper levels, the juxtaposition of the warm core Indian monsoon anticyclone (TH) and the cold core Atlantic cyclone (TUTT) to the west appears to be especially important to the [u] budget in the subtropics, causing the large positive eddy momentum flux over the Mediterranean (Figure 6b). This regional contribution from Eurasia to the zonal mean 200 mbar EMFD is greater than the Pacific and Atlantic Oceans by a factor of at least 3 (Figure 6c). Given this strong dependence, we examine the interannual relationship between the Indian monsoon and the EMFD below, with implications on {[u]} and the westward advection of the NASH.

4. Interannual Connections

4.1. Indian Monsoon and EMFD

[29] The timing and intensity of the July drying in the WATL may be dynamically linked to the Indian monsoon through feedbacks on the zonal momentum budget in the subtropics. We suggest the interannual variability of the Indian monsoon and in particular, the Tibetan High (TH), will drive the negative zonal mean eddy torque at 200 mbar (and therefore {[u]}, see section 3), an idea based on the climatological importance of EMFD on the southwest corner of the TH (SWTH) in driving the zonal mean (Figure 6).

[30] To test this idea, we employ the summer Indian Monsoon Index (IMI) defined by Wang et al. [2001]. The IMI is calculated as the 850 mbar zonal wind (U850) difference between box 1 and box 2 in Figure 7. The IMI is a measure of the intensity of low-level inflow around and into the Indian monsoon trough, so a large and positive value indicates a “wet” year. Other circulation indices of the monsoon strength exist [Webster and Yang, 1992; Wang et al., 2001], and the choice of which one best represents monsoon precipitation patterns may rest in the particular subregion of the broader south Asian monsoon (SAM) considered [Wang and Fan, 1999]. Since, it is the western sector of the TH over the Middle East that is important in terms of EMFD in midsummer (Figure 6), we focus on the monsoon heating immediately to its east (i.e., over India) as depicted by the IMI. The two major convective centers of the SAM, convection over the Bay of Bengal–India–Arabian Sea and that over the South China Sea and Philippine Sea, are poorly correlated on interannual time scales [Wang and Fan, 1999], thus Wang et al. [2001] advocate the use of two separate indices: the IMI and the Western North Pacific index (the latter is defined analogously to the IMI, except for west Pacific). Furthermore, Ailikun and Yasunari [1998] showed that the index defined by Webster and Yang [1992] is associated with convection over the western Pacific rather than that over India.

[31] Using MERRA data from 1979 to 2006, the mean June plus July IMI for each year is calculated. (The onset of the monsoon occurs sometime in June as is maturely established by July.) Figure 8 shows upper-lower tercile differences of wet minus dry IMI (9 year averages for each tercile) for eddy geopotential height at 850 mbar and 200 mbar, overlain with the total vector wind of wet years. During wet IMI years, there are anomalously low 850 mbar heights and cyclonic flow over and to the west of the enhanced convective activity over India, the Arabian Peninsula, and extending in to the Mediterranean, consistent with the heating-driven dynamics discussed by Gill [1980] and Rodwell and Hoskins [1996]. At upper levels, the TH is enhanced west of 80°E, with maximum positive anomalies near the eastern Mediterranean Sea. This finding is in agreement with previous diagnostic studies [Chaudhari et al., 2010; Rajeevan, 1993] showing 200 mbar anticyclonic anomalies associated with excessively warm and wet Indian monsoon years. Interestingly, the broad-scale anticyclonic circulation of the TH is not enhanced, with positive 200 mbar height anomalies only found west of India during wet years. In fact, the TH and the monsoon trough at 850 mbar over south China are diminished during wet IMI years, supporting the proposition that independent circulation indices are needed to describe the entire SAM [Wang et al., 2001].

Figure 8.

June and July eddy geopotential height and vector wind differences of good minus bad IMI years. Black box indicates the SWTH region shown in Figure 7. Data are from MERRA climatology.

[32] The enhanced anticyclonic flow centered over the Mediterranean during wet IMI years causes anomalous EMFD divergence (−∂uv/∂y) on the extreme southwest corner of the TH as strong southeasterlies (negative uv) turn to weak southwesterlies (positive uv) between 20°N and 30°N in the box shown (Figure 8a). This SWTH region corresponds to the western edge of the climatological EMFD maximum in Figure 6b, indicating a westward displacement of this EMFD maximum during wet IMI years. Guided by these composite mean circulation differences, Figure 9 shows interannual scatterplots of the SWTH EMFD (box in Figures 7 and 8) versus the IMI in MERRA and NCEP-GR1 for 1979–2006. All values are averaged over June and July.

Figure 9.

(a and b) Interannual scatterplots of June to July IMI versus the EMFD on the southwest corner of the Tibetan High (SWTH; see Figure 7). (c and d) Interannual scatterplots of EMFD at the SWTH versus the planetary zonal mean at the same latitude band. Figures 9a and 9c are for MERRA; Figures 9b and 9d are for NCEP-GR1. Years are color coded by ENSO 3.4 anomaly.

[33] A stronger Indian summer monsoon significantly correlates with (and presumably is the cause of) a larger negative 200 mbar EMFD tendency (Figures 9a and 9b), which is the primary term governing {[u]} tendency and the protrusion of easterlies into the subtropics in midsummer (section 3). In other words, years with a strong cyclonic flow at low levels over India have an enhanced upper level anticyclone to the west, causing anomalous EMFD as westerly (positive) u momentum is pumped northward across the 30°N latitude line. Moreover, local values of SWTH EMFD are positively correlated to the respective zonal mean value (Figures 9c and 9d), indicating that the SWTH is an important region governing interannual variability, in addition to the climatological mean (Figure 6b), of the 200 mbar zonal mean eddy torque in the subtropics.

[34] Individual years are also color coded in Figure 9 based on the simultaneous (June and July) Niño 3.4 anomaly [Trenberth, 1997], with warm (red), cold (blue), and neutral (black) color coding based on the respective tercile category (This same color coding is continued in Figure 11). No clear relationships to the phase of ENSO are seen, indicating that the relationships here are not just aspects of ENSO. The interaction between ENSO and the Indian monsoon has a rich research history [e.g., Walker, 1924; Bjerknes, 1969; Pant and Parthasarathy, 1981]. The negative relationship between warm ENSO events and monsoon precipitation [Yasunari, 1990; Webster and Yang, 1992; Kirtman and Shukla, 2000] varies with time, and in particular, has diminished in recent decades, possibly due to global warming [Kumar et al., 1999; Ashrit et al., 2001]. These findings may explain the lack of a relationship in Figure 9, since our relatively small sample of years start at 1979.

[35] In comparing the two different data sets, MERRA has slightly higher correlation coefficients than NCEP-GR1 and there is some considerable disagreement in the details of individual values of EMFD. The degree to which these reflect robust circulation biases or are an artifact of different horizontal resolution should be investigated further. Fortunately, interannual differences are qualitatively similar among both reanalyses and all correlations are statistically significant at the 95% confidence level.

4.2. The {[u]}-Based Composites

[36] To test our underlying hypothesis outlined in section 1, that midsummer zonal mean easterlies in the subtropics cause westward advection of the NASH with implications for midsummer precipitation changes, we focus on interannual comparisons of July minus June (JMJ) differences. JMJ differences are a useful metric, since this time signature captures the rainfall reduction (Figure 1), the corresponding strengthening of the NASH (Figure 2), and the negative acceleration of {[u]} (Figure 3) in the subtropics (20°N–30°N). Interannual differences of {[u]} at 20°N–30°N for JMJ are indexed and tercile composites of individual fields are constructed based on this index. The upper tercile in this context refers to the 9 years (from 1979 to 2006) for when JMJ {[u]} is largest and most negative.

[37] Upper minus lower tercile composites for precipitation, SLP, and 200 mbar EMFD based on this JMJ {[u]} index are shown in Figure 10. The zonal wind at 850 mbar (U850) is also shown (Figure 10b), to confirm that it tracks {[u]} as in climatology (see Figure 3). Precipitation data used in Figure 10 are from CMAP and wind and SLP data are from NCEP-GR1 (MERRA yields similar results). Data are first coarsened from its native 2.5° horizontal resolution to 10°, since the interest here is in large-scale differences in magnitude between low and high years, where the statistical significance preferentially resides. For U850, the spatial averaging is increased to 60° in longitude, which is roughly the zonal scale of the Atlantic. This coarsening isolates the importance of basin-scale zonal wind changes in the Atlantic, which are the basis of our hypothesis. (At small scales, the wind changes are dominated by pressure differences that can be inferred from Figure 10c.)

Figure 10.

Composite differences of various fields for good minus bad {[u]} index years. Fields are for July minus June differences, and stippling indicates regions of 95% significance using a Student's t test. Data are from NCEP-GR1.

[38] Composite differences for all grid points are shown in Figure 10, with local statistical significance (at the 95% confidence level based on a t test) indicated by hatching (note that we omit the Pacific in Figure 10 for conciseness, as its inclusion does not change the conclusions drawn therefrom). Figure 10a shows that there is enhanced negative EMFD over the SWTH (box in Figures 7 and 8) when {[u]} JMJ change in the subtropics is most negative. The SWTH region corresponds to the climatological maximum in negative EMFD tendency identified previously, and is the primary source of the negative {[u]} acceleration in midsummer, forcing both the mean climatology (Figure 6) as well as its interannual variability (Figure 9). These composite differences of EMFD are insensitive to the whether {[u]} (Figure 10a) or the IMI (Figure 8a) is used as the base index, indicating that interannual variability of the Indian monsoon, EMFD on the SWTH, and the negative acceleration of {[u]}, are robustly connected processes.

[39] The composite JMJ difference of the low-level zonal wind (U850) is shown in Figure 10b. When the vertical mean {[u]} decreases from June to July in the subtropics, so does U850, with significant negative anomalies stretching westward across the Atlantic in the 20°N–30°N latitude band. But in the midlatitudes around 45°N, JMJ U850 anomalies are positive over the Atlantic, implying a large-scale enhancement of the anticyclonic circulation (consistent with Figure 10c). There is also a region of positive U850 anomalies southwest of the Indian subcontinent extending westward into Africa. This region around 10°N–20°N may reflect enhanced southwesterlies during years when the monsoon is more active.

[40] SLP differences (Figure 10c) show that the entire NASH is strengthened when {[u]} is most negative, with large positive anomalies over the WATL. July SLP falls over the Arabian Peninsula are consistent with the monsoon precipitation (heating) anomalies to the east: the Indian monsoon is wetter, with increased precipitation in July over India and Pakistan (Figure 10d). Figure 10d also demonstrates that when there is a large and negative {[u]} JMJ change, the corresponding change in precipitation over the WATL (box in Figure 7) is significant and negative.

4.3. WATL and Indian Monsoon Rainfall

[41] As a consistency check of our key finding in section 4.2 that a wetter Indian monsoon in July corresponds to a drier WATL (Figure 10d), we examine JMJ precipitation composites (upper minus lower tercile) based on all of the important indices outlined above. Namely, JMJ differences of {[u]}, SLP in the WATL, precipitation in the WATL, and the IMI are used as the base indices for targeting precipitation differences in Figure 11. The findings across all indices (four panels in Figure 11) are strikingly similar: there are wet anomalies in Indian coincident with dry anomalies in the WATL. The magnitude of the difference between India and the WATL also remains steady across the different base indices. It appears the moistening of the monsoon is remotely connected to the drying of the WATL in July. The consistency of this inverse relationship suggests midsummer changes in precipitation in India and the WATL are robustly related features whose interannual variability might be linked through the midsummer enhancement of subtropical easterlies and the intensification of the NASH.

Figure 11.

July minus June precipitation composite differences based on various indices. Only grid points at the 95% confidence level (using a t test) are shown. Data are from NCEP-GR1. See Figure 7 for definition of indices.

[42] The aforementioned processes governing the July drying (JMJ rainfall) are also correlated to interannual variations of mean monthly rainfall. In particular, the spatial pattern of June rainfall composited against the June indices in Figures 10 and 11 reveals a similar dipole pattern: a wet India corresponds to an enhanced NASH and dry WATL (not shown). This might suggest that if there is an early and strong monsoon, the negative {[u]} momentum, NASH strengthening, and drying in the WATL all occur early in June. However, when July minus June IMI is large and positive (i.e., a late monsoon onset), the subsequent drying in the WATL occurs later as well.

[43] Given the high degree of serial correlation and spatial interdependence often present in meteorological fields, Livezey and Chen [1983] suggest that field (global) significance testing is required to ascertain statistical significance. To address this, we performed a series of Monte Carlo experiments (following Wolter et al. [1999]) of the precipitation differences presented in Figure 11. 10,000 random time series each 28 years long (corresponding to the 1979–2006 base period) were created using a random number generator. These time series were subsequently sorted into upper and lower terciles. The JMJ precipitation data was then composited based on these indices from the randomly generated time series. Precipitation differences of upper minus lower tercile were then retested for local significance at each grid point. Finally, the number of significant grid points at the 95% confidence level (using the t test) from the 10,000 experiments were arranged in a decreasing sequence with the 500th grid points marking the field significance threshold (α = 0.05). For each base index in Figure 11, the number of significant grid points is greater than the relevant Monte Carlo criterion, indicating that the above composite differences of precipitation are meaningful.

4.4. Correlation Analysis

[44] Interannual time series of several key indices are examined here to extend the conclusions drawn from sections 4.2 and 4.3. There is a significant difference in JMJ SLP and precipitation over the WATL region (shown in Figure 7) dependent on the strength of the zonal mean easterlies in the subtropics (Figure 10). Individual time series for JMJ SLP and JMJ precipitation in the WATL are constructed and compared to the respective {[u]} time series. Figure 12 shows scatterplots of JMJ differences for each individual year from 1979 to 2006.

Figure 12.

(a and b) Interannual scatterplots of SLP versus precipitation in the WATL domain. Scatterplots of WATL (c and d) SLP versus {[u]} at 20°N–30°N as well as (e and f) WATL precipitation versus {[u]}. Figures 12a, 12c, and 12e are for MERRA; Figures 12b, 12d, and 12f are for NCEP-GR1. Values are for July minus June differences and years are color coded by ENSO 3.4 anomaly.

[45] There is a significant negative correlation (r = −0.8) between SLP and precipitation (Figures 12a and 12b), perhaps unsurprisingly as the midsummer strengthening of the NASH has previously been cited as the proximate reason for the rainfall deficit in the WATL (see section 1). JMJ {[u]} also significantly correlates with SLP: years with strong easterlies also have a large and positive SLP rise in July (Figures 12c and 12d). Furthermore, interannual variability of {[u]} also correlates significantly with precipitation in the WATL (Figures 12e and 12f). It appears that the strength of the midsummer easterlies, as viewed by JMJ {[u]} changes, is a key indicator of interannual variations of the NASH and hence, July drying in the WATL. Reassuringly, the same analysis for MERRA and NCEP-GR1 give consistent results (Figures 12a, 12c, and 12e and Figures 12b, 12d, and 12f, respectively) and as mentioned previously, these year-to-year changes are not significantly correlated with the Niño 3.4 anomaly (color coding in Figure 12).

[46] The correlations presented in Figure 12 and the composite differences in Figures 10 and 11 point to a significant relationship between the Indian monsoon, {[u]}, and WATL precipitation on an interannual time scale. Figure 13 shows time series of these 3 key indices for JMJ to better understand the frequency of the interannual variability driving these correlations. NCEP-GR1 {[u]} data and IMI index are shown in Figure 13 because of the longer record length (beginning in 1948), though results using MERRA are similar for the same period (post-1979). Figure 13a demonstrates that in years when {[u]} JMJ is large and negative, JMJ precipitation in WATL also tends to be large and negative while the IMI tends to be strong and positive (equivalent to stronger monsoon in July). These interrelationships appear to arise largely from random year-to-year variability. In other words, they are not solely a result of a particularly strong correlation from one or two years (e.g., the 1982–83 and 1997–98 El Niños), or of a trend over the entire record length. There is some indication of decadal variability, however, which is better seen by low-pass filtering the total time series shown in Figure 13a.

Figure 13.

Interannual time series of the key indices discussed in this text for the (a) total time series and (b) low-pass filtered time series to remove frequencies with periods less than 7 years. All values are for July minus June differences. Precipitation is in units of mm/d, {[u]} is in m/s, and IMI is a normalized quantity. See Figure 7 for further details.

[47] Figure 13b shows time series of the same indices but with a low-pass digital filtered applied to isolate periods longer than 7 years. Correlations between the different indices are categorically higher for the filtered than unfiltered (total) time series. For instance, the correlation between {[u]} and the IMI increases from −0.4 for the unfiltered to −0.7 for the filtered time series. The correlation between the IMI and {[u]} increases form +0.4 to +0.6, whereas the correlation between IMI and precipitation in the WATL increase from −0.4 to −0.6 for the filtered time series. The IMI and {[u]} indices also have less interannual variability during the period of approximately 1970–1990. These findings might reflect the importance of a low-frequency oscillation(s) in dictating the observed teleconnections presented here.

5. Summary and Discussion

[48] The July rainfall deficit in the western North Atlantic (WATL) has been previously attributed to the western enhancement of the North Atlantic Subtropical High (NASH). However, little has been offered to explain the NASH enhancement itself. We suggest the extension of zonal mean easterlies into the subtropics in July is an essential mechanism governing the western displacement of the NASH in midsummer, and therefore may be a key processes governing precipitation in the WATL.

[49] This diagnostic study revealed several key connections lending support to our hypothesis which are summarized in the schematic in Figure 14. Heating from precipitation related to the Indian monsoon is a primary processes governing eddy momentum fluxes in the subtropics in summer, with enhanced eddy momentum flux divergence (EMFD) when the precipitation and Tibetan High are enhanced in July. This EMFD in turn drives the negative acceleration of the vertically averaged zonal mean zonal wind {[u]} in the subtropics in summer. Therefore, the Indian monsoon, and in particular, the intraseasonal change in its strength from June to July appears to be a key process governing the negative acceleration of zonal mean momentum from June to July.

Figure 14.

(top) Summary schematic of the proposed teleconnections linking the Indian monsoon with July drying in the WATL. Implied causation flows from right to left. (bottom) Linear correlation coefficients of the various processes, with blue arrows indicating direct pathways and red and black arrows indicating indirect connections. Correlations are based on July minus June differences for the 1979–2006 period. For wind and SLP data, correlations presented are based on the average MERRA and NCEP-GR1 coefficients.

[50] These midsummer variations of {[u]} are significantly correlated with both the westward displacement of the NASH and the associated rainfall deficit in the WATL, pointing to the strength of the zonal mean easterlies as an indicator of the strength of the July drying in the WATL. The key steps connecting the Indian monsoon with WATL precipitation are shown in Figure 14 along with the respective correlation for each step. Interestingly, the remote correlation between the Indian monsoon and the July drying in the WATL is significant and negative, in addition to the intermediary processes linking the two. The increase in precipitation from June to July in India corresponds to a decrease in precipitation from June to July in the WATL. The role of the monsoon in driving zonal mean easterlies implies that this inverse relationship is not a spurious association, but rather, the Indian monsoon appears to be an important factor governing the variability of midsummer precipitation in the WATL. Model experiments could further demonstrate the role of monsoon heating on driving midsummer {[u]} changes and whether monsoon precipitation might affect precipitation in the subtropical Americas.

[51] The findings of this observational study lend support to the new hypothesis that the midsummer enhancement of the planetary mean easterlies plays a critical role in the midsummer drying of the subtropical Americas. The westward advection of the NASH's anticyclonic vorticity across the Atlantic by subtropical easterlies seems key to this drying. However, future work is needed to understand the specific mechanism(s) linking {[u]} to precipitation. {[u]}, with its distinctive sharp time signature (Figure 3a), will contribute to advective tendency signals with that same time signature of all fields in the northern subtropics. But whether that leads to moisture and/or precipitation changes depends on the zonal gradients being advected, and on the dynamical processes that link the advected fields with weather. Additional studies that examine these dynamical processes in more detail, like changes in Ekman divergence due to the westward displaced NASH, are needed to elucidate the downstream implications of summer {[u]} changes on precipitation processes. While the WATL is emphasized here, these principles surely must act elsewhere in the vicinity of western subtropical ocean basins and should be investigated further.

[52] For instance, the Caribbean Sea and easternmost Pacific both have similar midsummer dry periods, but with subtle differences in the timing of the relative rainfall minimum. Our July minus June differences may be weak in these regions because of this timing issue. The rainfall minimum in the different regions might also involve different sets of physical processes, such as local air-sea interactions in the Pacific [Magaña et al., 1999; Magaña and Caetano, 2005] or aerosol concentration in the Caribbean [Angeles et al., 2010]. Teasing apart the physical processes and dynamical mechanisms will require further studies of historical variability and individual years, as well as modeling efforts.

Acknowledgments

[53] This material is based upon work supported by the National Science Foundation under grant 0731520. The authors are grateful for the constructive comments by several anonymous reviewers.