Response of tropical global ocean temperature to the Sun's quasi-decadal UV radiative forcing of the stratosphere

Authors


Abstract

[1] The quasi-decadal oscillation (QDO) in the Earth's climate system fluctuated in phase with the 11-year period signal in solar total irradiance (STI) variability throughout the 20th century. The QDO was associated with global average upper ocean temperature variability dominated by the tropical global average from 20°S to 20°N. Earlier diagnosis of the tropical global average oceanic thermal budget found the anomalous warming tendency driven not by the radiative forcing (∼0.15 W m−2) from the 11-year period signal in STI but by the larger anomalous quasi-decadal sensible plus latent heat flux (∼0.5 W m−2) from the overlying troposphere. Now we diagnose the corresponding thermal budget of the tropical global average atmospheric temperature variability, finding it largest in the lower stratosphere (∼0.8°K), decreasing downward into the lower troposphere (∼0.15°K) and the upper ocean (∼0.05°K). These diagnostics find quasi-decadal temperature variability in the tropical lower stratosphere arising from a thermal balance between anomalous radiative forcing (1.0–1.5 W m−2) by ozone absorption of the 11-year signal in solar UV irradiance variability (modified by absorption of mean solar IR irradiance by volcanic aerosol variability) and variable longwave radiation and vertical advection. The latter two processes altered the vertical gradient of equivalent potential temperature in the tropical troposphere that allowed mean vertical circulation to drive an anomalous warming tendency in the lower troposphere. The latter matched the amplitude and phase of the downward quasi-decadal sensible plus latent heat flux anomaly that drove the anomalous warming tendency in the tropical global ocean.

1. Introduction

[2] The quasi-decadal oscillation (QDO) near 11-year period was one of six principal signals observed in global patterns of sea surface temperature (SST) and sea level pressure (SLP) during the 20th century [Allan, 2000; White and Tourre, 2003]. It was characterized by global patterns of SST and SLP variability similar to those characterizing the El Niño–Southern Oscillation (ENSO) signals ranging in period from 3 to 7 years and the quasi-biennial oscillation (QBO) near 2.3-year period [White and Tourre, 2003; White et al., 2003b]. The QDO was associated with global average temperature variability in the upper ocean (∼0.1°K) that was dominated by the tropical global average from 20°S to 20°N, both averages fluctuating in phase with the nominal 11-year signal in the solar total irradiance (STI) variability [White et al., 1997, 1998, 2001a; White and Tourre, 2003].

[3] Recently, White et al. [2003a] diagnosed the thermal budget of this tropical global average upper ocean temperature variability (20°S to 20°N), finding the surface radiative forcing (i.e., 0.10–0.15 W m−2) by the nominal 11-year signal in STI variability [Lean et al., 1995] unable to explain corresponding changes in tropical global average diabatic heat storage (〈DHS〉) in the upper ocean; that is, the transient Stefan-Boltzmann radiation model produced amplitudes too small and phase differences too large. Instead, they found the 〈DHS〉 tendency driven by anomalous sensible plus latent heat flux of ∼0.5 W m−2. Thus the tropical atmosphere appeared to heat the underlying tropical ocean via a turbulent transfer of heat across the air-sea interface on quasi-decadal period scales.

[4] This finding was unexpected, but appeared to be consistent with a body of work conducted over the past decade by Labitzke and van Loon [1997a, 1997b, 1999] and Haigh [1994, 1999, 2003], who found a relatively large quasi-decadal signal in tropical lower stratosphere temperature variability of O (1.0°K) driven by the radiative forcing from the stratosphere ozone absorption of the nominal 11-year signal in the solar ultraviolet (UV) irradiance variability. This tropical global average temperature signal observed in the lower stratosphere was an order of magnitude larger than that (∼0.1°K) observed in the upper ocean [White et al., 1997, 1998, 2001a, 2003a]. Labitzke and van Loon [1999] explained this simply; that is, it occurred because the percent amplitude in the 11-year signal of the solar UV irradiance variability (i.e., 1–1.5% of the mean UV irradiance) was more than an order of magnitude larger than that of the STI variability (i.e., ∼0.04% of the mean total irradiance). The principal question answered in the present study is whether this relatively large quasi-decadal temperature variability in the tropical lower stratosphere drove corresponding temperature variability in the tropical lower troposphere and upper ocean.

2. Data and Methods

[5] We utilize 9 atmosphere variables to diagnose the thermal budget of the quasi-decadal tropical global average temperature variability in the tropical atmosphere. These variables are taken from the National Centers for Environmental Prediction (NCEP) and Department of Energy (DOE) Reanalysis II data set [Kanamitsu et al., 2002] over the 25 years from 1975 through 1999. The NCEP/DOE reanalysis corrects for deficiencies in model initial conditions, boundary conditions, and cloud, radiation, and precipitation parameterizations [Roads et al., 1998] by assimilating daily satellite temperature and water vapor profiles in the troposphere (from 50 hPa to the sea surface) from the TIROS operational vertical sounder (TOVS) [Rao et al., 1990]. The TOVS data were available daily on a 300 km grid over the globe, while the assimilated Reanalysis II data used in the present study were available monthly on a 2° longitude by 2° latitude grid.

[6] We also utilize monthly SST, sensible heat flux (QS), latent heat flux (QL), atmospheric temperature (T), atmospheric wind (V), atmospheric divergent wind (VD), atmospheric vertical (pressure) velocity (ω), and the horizontal and vertical eddy heat flux divergence from the NCEP/DOE Reanalysis II data set. These available data allow us to analyze air-sea temperature differences but not air-sea specific humidity differences. We also utilize upper ocean temperature data objectively analyzed from all available temperature profiles extending from the sea surface to 400 m depth [White, 1995; White et al., 2001b].

[7] Monthly anomalies of oceanic and atmospheric variables are constructed by subtracting individual monthly mean estimates from long-term monthly means over the 25-year record from 1975 through 1999. We isolate the nominal 11-year period signal by band-pass filtering monthly anomalies in each data set utilizing half-power points at 8- and 14-year periods. This filter has a frequency response function that is flat with steep sides and negligible side lobes [Kaylor, 1977], yielding ∼2.5 cycles of quasi-decadal variability over the 25-year record.

[8] The nominal 11-year signal in STI variability during the 25-year record from 1975 through 1999 displayed peaks in 1981 and 1991, with an amplitude of ∼0.5 W m−2 from 1975 through 1999 (Figure 1a). The corresponding 11-year signal in the solar UV irradiance variability fluctuated with a larger amplitude ranging from 1.5 to 2.0 W m−2 (see Appendix A), The radiative forcing averaged across the tropics from 20°S to 20°N was smaller than this by a factor of ∼3. These peaks in quasi-decadal solar forcing nearly coincided with the volcanic eruptions of El Chichon in April of 1982 and Mt. Pinatubo in July of 1991, which ejected volcanic aerosols into the lower stratosphere. During 2–3 years following each eruption, the volcanic aerosols in the tropical lower stratosphere partially absorbed the mean solar infrared (IR) irradiance while shielding the troposphere below [Robock, 2000; Douglass and Clader, 2002]. We investigate the influence that the variable volcanic aerosol absorption of mean solar IR irradiance exerted on the stratospheric ozone absorption of variable solar UV irradiance (see Appendix A).

Figure 1.

(a) Time sequences of tropical global average quasi-decadal SST anomalies (20°S to 20°N) and of the 11-year signal in the STI variability from 1975 through 1999. (b) Animation sequence from 1986 through 1995 of meridional vertical sections displaying global zonal-averaged quasi-decadal variability of upper ocean temperature extending from the sea surface to 400 m depth and from 40°S to 40°N. (c) Same as Figure 1b but for global zonal-averaged quasi-decadal anomalies of atmospheric temperature extending upward from the sea surface to 20 km height and from 40°S to 40°N. Color contours are given by the color bar at bottom in units of 0.02 K for upper ocean temperature anomalies and 0.10 K for air temperature anomalies.

3. Animation Sequences of Meridional Vertical Sections of Zonal Average Quasi-Decadal Variability in the Atmosphere and Upper Ocean

[9] We examine animation sequences of meridional vertical sections of zonal average quasi-decadal variability of upper ocean temperature (〈T′O〉) from 400 m depth to the sea surface, and of atmosphere temperature (〈T′A〉), vertical velocity (〈ω′〉), and zonal wind (〈u′〉) from the sea surface to 20 km height, from 40°S to 40°N over one cycle of the QDO from 1986 through 1995 (Figures 1 and 2) . The first two animation sequences are plotted together with time sequences of the quasi-decadal signals in tropical global average SST variability (20°S to 20°N) and in STI variability (Figure 1a), the latter displaying peaks centered near 1980/81 and 1990/91, and a trough near 1986 [White et al., 1997].

Figure 2.

(a) Same as Figure 1 but for global zonal-averaged quasi-decadal anomalies of atmospheric vertical pressure velocity in the negative pressure direction (positive upward with the sign reversal). (b) Same as Figure 1 but for global zonal-averaged quasi-decadal anomalies of zonal surface wind (positive eastward). Color contours are given by the color bar at bottom in units of 2 × 10−4 hPa s−1 for vertical velocity anomalies and in units of 0.15 m s−1 for zonal wind anomalies.

[10] The animation sequence of 〈T′O〉 displays minimum (maximum) variability (∼0.1°K) in the tropical upper ocean, confined to the near-surface mixed layer and upper portion of the main pycnocline above 200 meters depth during 1986/87 (1991/92), uniform in sign from 15°S to 25°N (Figure 1b). These minimum and maximum 〈T′O〉 approximately coincided with respective minimum and maximum quasi-decadal STI variability, as observed earlier [White et al., 1998; White et al., 2003a]. The apparent upward propagation of warm 〈T′O〉 from 1987 to 1991 belies the Rossby wave reflection of off-equatorial thermocline depth anomalies onto the equator at the western boundary of the Pacific Ocean and their subsequent slow eastward propagation as equatorial coupled waves into the eastern equatorial ocean observed in three dimensions [e.g., White et al., 2003b]. The vertical integral of temperature variability in the near-surface mixed layer above the main pycnocline forms the basis of the diabatic heat storage variability (〈DHS′〉), the tendency of which was balanced by sensible plus latent heat flux variability when averaged over the tropical global ocean from 20°S to 20°N [White et al., 2003a].

[11] The animation sequence of 〈T′A〉 displays minimum (maximum) variability (∼1.0°K) in the tropical lower stratosphere above 13 km height during 1986/87 (1991/92), uniform in sign from 40°S to 40°N (Figure 1c). Minimum (maximum) lower stratosphere 〈T′A〉 began penetrating downward into the lower troposphere in 1986 (1991) during minimum (maximum) radiative forcing by the nominal 11-year signal in STI variability (Figure 1a), modified by weaker radiative forcing from quasi-decadal variability in volcanic aerosol production (Appendix A). Maximum 〈T′A〉 in the tropical lower stratosphere was half an order of magnitude larger than that of ∼0.2°K in the lower troposphere (Figure 1c) and an order of magnitude larger than 〈T′O〉 of ∼0.1°K in the near-surface mixed layer in the upper ocean (Figure 1b).

[12] The animation sequence of −〈ω′〉 (Figure 2a) displays minimum (maximum) variability during the cool-to-warm (warm-to-cool) transition from 1988 through 1990 (1993 through 1995) (Figure 1c). Normal zonal average Hadley cell circulation has ascending motion from ∼10°S to ∼10°N and descending motion from ∼10° latitude to ∼35°latitude in both hemispheres [Peixoto and Oort, 1992]. From 1988 through 1990, Hadley cell circulation in the Northern Hemisphere was weaker than normal, with anomalous descent (i.e., negative) from the equator to ∼10°N and anomalous ascent (i.e., positive) from 10°N to 25°N, while that in the Southern Hemisphere was displaced equatorward, with anomalous ascent from the equator to ∼5°S and anomalous descent from ∼ 5°S to ∼20°S (Figure 2a). From 1993 through 1995, the Hadley cell circulation in the Northern Hemisphere became stronger than normal, and that in the Southern Hemisphere was displaced poleward (Figure 2a). Averaged across the tropics from 20°S to 20°N, the −〈ω′〉 variability was on average downward (upward) during the transition from 1988 through 1990 (1993 through 1995).

[13] The animation sequence of 〈u′〉 (Figure 2b) displayed maximum (minimum) variability nearly straddling the equator from 20°S to 20°N near the tropopause (i.e., near 13 km height) during the cool-to-warm (warm-to-cool) transition from 1988 through 1990 (1993 through 1995) (Figure 1c). During the first two years of both transitions, 〈u′〉 of opposite sign occurred in the lower troposphere (i.e., below 2.5 km height), corresponding to zonal trade wind intensification (weakening) during the transition from 1988 through 1989 (1993 through 1994), as observed [White et al., 2003a].

4. Time Sequences of Tropical Global Average Quasi-Decadal Temperature Variability

[14] Now we examine time sequences of quasi-decadal variability of tropical global average (20°S to 20°N) lower stratosphere temperature at 15 km height (〈T′15km〉), upper troposphere temperature at 10 km height (〈T′10km〉), middle troposphere temperature at 5 km height (〈T′5km〉), surface air temperature at 2 m height (〈T′2m〉), sea surface temperature (〈SST′〉), air-sea temperature difference (〈T′2m − SST′〉), sensible plus latent heat flux (−〈Q′S + Q′L〉), and upper ocean temperature (〈T′O〉) from 1977 through 1999 (Figures 3a–3h). Examination of these time sequences finds the tropical troposphere and upper ocean temperature variability (i.e., 〈T′10km〉, 〈T′5km〉, 〈T′2m〉, 〈T′O〉) reaching maximum value during 1980 and 1990 (Figures 3b, 3c, 3d, and 3h). However, tropical lower stratosphere temperature variability (i.e., 〈T′15km〉) lagged tropical troposphere and upper ocean temperature variability by ∼1 year (Figure 3a), indicating that the latter did not arise simply from variable longwave radiative forcing from the tropical lower stratosphere. Even so, the amplitude of these temperature signals decreased downward from the tropical lower stratosphere to the tropical upper ocean by an order of magnitude (i.e., from ∼0.8°K for (〈T′15km〉. to ∼0.3°K for〈T′10km〉, to ∼0.15°K for 〈T′5km〉 and 〈T′2m〉, and to ∼0.05°K for 〈T′O〉), suggesting that the latter derived from the former. The question is; what are the thermal processes responsible for this apparent downward transfer of heat from the tropical lower stratosphere to the tropical upper ocean?

Figure 3.

Time sequences from 1977 through 1999 of quasi-decadal anomalies of tropical global average (20°S to 20°N) (a) temperature in the lower stratosphere at 15 km height (〈T′15km〉); (b) temperature in the upper troposphere at 10 km height (〈T′10km〉); (c) temperature in the middle troposphere at 5 km height (〈T′5km〉); (d) surface air temperature at 2 m height (〈T′2m〉); (e) sea surface temperature (〈SST′〉); (f) air-sea temperature difference (〈T′2m − SST′〉); (g) sensible plus latent heat flux (−〈Q′S + Q′L〉), positive downward; and (h) depth average upper ocean temperature (〈T′O〉). Note that the ordinates for 〈T′15km〉 and 〈T′10km〉 range over ±1.0°K and ±0.4°K, respectively, while those for 〈T′O〉, 〈SST′〉, 〈T′2m〉, and 〈T′5km〉 range over ±0.2°K and that for 〈T′2m − SST′〉 ranges over ±0.1°K. Temporal lag cross correlation (i) between 〈T′O〉 and −〈Q′S + Q′L〉 and (j) between〈T′O〉 and 〈T′2m − SST′〉, with the 90% confidence level (0.80) determined for 5 effective degrees of freedom [Snedecor and Cochran, 1980].

[15] Further examination of these time sequences finds both 〈T′2m − SST′〉 and −〈Q′S + Q′L〉 variability fluctuating in phase with one another and leading 〈T′O〉 by ∼90° of phase (Figures 3f, 3g, and 3h). These temporal phase relationships and the ∼0.5 W m−2 amplitude for −〈Q′S + Q′L〉 are entirely consistent with the response of the 〈DHS′〉 tendency to −〈Q′S + Q′L〉 observed earlier by White et al. [2003a]. Because 〈T′2m〉 display twice the amplitude of 〈SST′〉, we are not surprised to find positive downward −〈Q′S + Q′L〉 fluctuating with positive 〈T′2m − SST′〉 (Figures 3f and 3g). These temporal phase relationships are confirmed quantitatively by computing their temporal lag cross correlations (Figures 3i and 3j).

5. Source of Tropical Global Average Quasi-Decadal Sensible Plus Latent Heat Flux Variability

[16] The animation sequence of horizontal maps of quasi-decadal air-sea temperature difference variability (T′2m − SST′) from 30°S to 30°N (Figure 4a) finds T′2m warmer than SST′ across most of the tropical global ocean from 20°S to 20°N during the cool-to-warm transition from 1987 through 1990 (Figures 1b and 1c). During this cool-to-warm transition, the quasi-decadal zonal surface wind variability (ZSW′) was negative over most of the tropical global ocean (Figure 4b), associated with a strengthening the global average trade winds. So, negative ZSW′ tended to instigate upward quasi-decadal (Q′S + Q′L), while positive (T′2m − SST′) tended to instigate downward (Q′S + Q′L) [Cayan, 1992a, 1992b]. The question is: which effect was most important in the tropical global average?

Figure 4.

Animation sequences from 1986 through 1995 of horizontal maps displaying quasi-decadal anomalies of (a) the air-sea temperature differences (T′2m − SST′) and (b) the zonal surface wind (ZSW′) across the global ocean from 30°S to 30°N. Color contours are given by the color bar at the bottom in units of 0.02°K for (T2m − SST′) and 0.1 m s−1 for ZSW′, with yellow to red (light blue to dark blue) indicating positive (negative) anomalies.

[17] To answer this question, we examine the global tropical distribution of quasi-decadal −(Q′S + Q′L) from 30°S to 30°N in the middle of this cool-to-warm transition during 1988 (Figure 5a). This distribution shows positive (downward) −(Q′S + Q′L) dominating most of the tropical global ocean except in the central tropical Pacific Ocean. This global pattern of positive −(Q′S + Q′L) can be seen, for the most part, to have been matched by the global pattern of positive (T′2m − SST′) from 20°S to 20°N (Figure 5b). On the other hand, negative ZSW′ across much of the central tropical Pacific Ocean from 20°S to 20°N, indicating trade wind intensification there (Figure 5c), can be seen to have been responsible for the negative (upward) −(Q′S + Q′L) displayed there. Yet outside this central tropical Pacific domain, across most of the Indian, western Pacific, eastern Pacific, and Atlantic oceans, positive (T′2m − SST′) (Figure 5b) accounted for most of the positive (downward) sensible plus latent heat flux variability from 20°S to 20°N (Figure 5a). These results are consistent with the temporal lag cross correlation between time sequences of tropical global average 〈T′2m − SST′〉 and −〈Q′S + Q′L〉 from 1977 through 1999 (not shown).

Figure 5.

Horizontal maps in 1988 displaying quasi-decadal anomalies of (a) (minus) sensible plus latent heat flux [−(Q′S + Q′L)] (positive downward), (b) air-sea temperature difference (T′2m − SST′), (c) zonal surface wind (ZSW′), (d) surface air temperature at 2 m height (T′2m), and (e) sea surface temperature (SST′). Each map extends across the global ocean from 30°S to 30°N. Color contours are 1.0 W m−2, 0.02°K, 0.1 m s−1, 0.05°K, and 0.05°K, respectively.

[18] These positive (T′2m − SST′) across most of the tropical global ocean in 1988 (Figure 5b) arose principally because T′2m during this cool-to-warm transition (Figure 5d) was larger (more positive) than SST′ (Figure 5e) over more than 70% of the tropical global ocean from 20°S to 20°N. The fact that the magnitude of the tropical global average atmospheric temperature anomalies increased upward from the lower troposphere to the lower stratosphere by more than half an order of magnitude throughout the quasi-decadal cycle (Figures 3a, 3b, 3c, and 3d) suggests that the thermodynamics governing T′2m across most of the tropical global ocean in 1988 derived primarily from atmosphere processes. Moreover, because these mostly positive (T′2m − SST′) across the tropical global ocean in 1988 coincided with the cool-to-warm transition from 1987 through 1990 (Figure 1c), when the magnitude of vertical motion variability from 20°S to 20°N was maximum (Figure 2a), we hypothesize that the anomalous diabatic cooling of the tropical lower troposphere (required to explain the anomalous downward sensible plus latent heat flux into the upper ocean) was balanced by a warming tendency from the anomalous vertical thermal advection in response to radiative forcing of the tropical lower stratosphere by the 11-year signal in the solar UV irradiance variability. Testing this hypothesis requires diagnosing the thermal budget of the tropical global atmosphere.

6. Diagnosis of the Thermal Budget of the Quasi-Decadal Signal in the Tropical Atmosphere

[19] We diagnose the thermal budget of the QDO averaged both zonally and globally across the global tropical atmosphere from 20°S to 20°N as a function of height from the sea surface to 20 km by estimating the terms in the following linearized thermal budget [Peixoto and Oort, 1992, p. 52] utilizing the NCEP/DOE Reanalysis II data set as indicated in section 2, that is,

equation image

In this expression, T is temperature, V is wind velocity, θE is the equivalent potential temperature taking into account the increase in potential temperature from the condensation of available moisture; ω is vertical (pressure) velocity in the p direction (negative upward); (equation image) is horizontal eddy thermal flux; (equation image) is vertical eddy thermal flux; QD/(ρACPA) is diabatic heating in units of °K s−1; CPA is the specific heat of air; ρA is the density of air; where the overbar and prime indicate the long-term monthly mean and quasi-decadal anomalies, respectively. Vertical velocity anomalies are computed as the vertical integral of horizontal divergence anomalies, with that at the sea surface taken to be zero [White and Chen, 2002].

[20] We estimate the anomalous diabatic heating on the right-hand side of equation (1) as the residual from the measured terms on the left-hand side. The anomalous diabatic heating represents the combination of anomalous net radiational heating and latent heat release [Roads et al., 1998]. The vertical integral of diabatic heating anomaly through the troposphere column equals the net heat flux into the column across the air-sea interface and the tropopause near 13 km height.

[21] We compared the terms on the left-hand side of equation (1), finding them dominated by the anomalous vertical thermal advection throughout the tropical troposphere and lower stratosphere, the other terms more than half an order of magnitude smaller. This yields the dominant thermal balance of the quasi-decadal signal across the tropical troposphere and lower stratosphere; that is,

equation image

The evolution of the terms in equation (2), zonally averaged along the meridional vertical section from 40°S to 40°N and from the sea surface to 20 km height is given by animation sequences from 1986 through 1991 (Figure 6), bracketing the cool-to-warm transition from 1987 through 1990 in the QDO (Figure 1c). These animation sequences display the principal balance between the anomalous diabatic heating (Figure 6c) and the anomalous vertical thermal advection in the tropical atmosphere from 20°S to 20°N, the latter composed of two parts, that is, the mean vertical advection of anomalous equivalent potential temperature (Figure 6b) and the anomalous vertical advection of mean equivalent potential temperature (Figure 6a).

Figure 6.

(a) Animation sequence from 1986 through 1991 of meridional vertical sections displaying globally zonal-averaged quasi-decadal estimates of the anomalous vertical advection of mean equivalent potential temperature (ω′(equation image/equation imageequation imageE/∂p) on the left-hand side of equation (2), extending from the sea surface to 20 km height and from 40°S to 40°N. (b) Same as Figure 6a but for the mean vertical advection of anomalous equivalent potential temperature (equation image (equation image/equation imageE)(∂θ′E/∂p) on the left-hand side of equation (2). (c) Same as Figure 6a but for the anomalous diabatic heating QD/(ρACPA) on the right-hand side of equation (2). (d) Same as Figure 6a but for the model net radiative forcing of the lower stratosphere (above ∼13 km height) by the 11-year signal in solar UV irradiance variability, modified by that from volcanic aerosol variability, computed as (S′UV + V′IR) in equation (3). Color contours are given by the color bar at the bottom in units of 0.5 × 10−7 °K s−1 for the heat budget terms in Figures 6a–6c and of 0.5 W m−2 for the net radiative forcing in Figure 6d.

6.1. Tropical Lower Stratosphere

[22] Examination of the anomalous thermal budget in the tropical lower stratosphere above ∼13 km height from 20°S to 20°N finds the anomalous diabatic heating on the right-hand side of equation (1) (Figure 6c) balanced principally by the anomalous vertical advection of mean equivalent potential temperature on the left-hand side throughout the record (Figure 6a). When averaged across the latitude domain from 40°S to 40°N, both terms were minimum (maximum) during the cool (warm) phase of the quasi-decadal signal in 1986 (1991) (Figure 1c). However, when averaged across the tropics from 20°S to 20°N, both terms were minimum during the cool-to-warm transition from 1987 through 1990 (Figure 1c). This anomalous diabatic cooling accounted for the minimum cool temperature anomalies in the lower stratosphere in 1986 and its prolongation in the tropics from 20°S to 20°N during the first half of the cool-to-warm transition from 1987 through 1989 (Figure 1c).

[23] This indirect estimation of the anomalous diabatic heating in the tropical lower stratosphere (Figure 6c) allows us to estimate its anomalous net radiative forcing, assuming latent heat release to have been negligible above the tropopause [Roads et al., 1998]. Remember that the NCEP/DOE reanalysis model was not driven by radiative forcing from the stratospheric ozone absorption of anomalous solar UV irradiance (S′UV) or from the anomalous stratosphere volcanic aerosol absorption of the mean solar IR irradiance (V′IR) [Kanamitsu et al., 2002]. However, the reanalysis model did assimilate the observed temperature changes in the lower stratosphere from TOVS [Rao et al., 1990] that would be expected to respond to such anomalous radiative forcing. Thus we can indirectly estimate the anomalous net radiative forcing (S′UV + V′IR) (Figure 6d) by recognizing that the diabatic heating anomaly in the tropical lower stratosphere (Figure 6c) represented the difference between (S′UV + V′IR) and the anomalous longwave radiation response to space and to the troposphere below; that is,

equation image

This integral is taken from the bottom of the tropical lower stratosphere near ∼13 km height upward to height z, where 4σequation image13km3T13km is anomalous longwave radiation downward into the troposphere and 4σequation image3T′ is anomalous longwave radiation upward into the middle stratosphere from height z, with σ the Stefan-Boltzmann constant.

[24] This indirect estimation of (S′UV + V′IR) in equation (3) yields minimum negative values in 1986/87 and maximum positive values in 1991/92 across the tropics from 20°S to 20°N (Figure 6d), consistent with the phase of minimum and maximum anomalous net radiative forcing estimated from both sources in Appendix A. Its amplitude is estimated to have been 1.0–1.5 W m−2 when averaged over the lower stratosphere from 13 to 20 km height and over the tropics from 20°S to 20°N (Figure 6d). This is only slightly larger than more direct and independent estimate of 0.8–1.2 W m−2 derived in Appendix A.

6.2. Tropical Troposphere

[25] In the tropical troposphere below ∼13 km height, the magnitude of the anomalous vertical advection of mean equivalent potential temperature (Figure 6a) was comparable with that of the mean vertical advection of anomalous equivalent potential temperature (Figure 6b), both achieving maximum variability from 20°S to 20°N during the cool-to-warm transition from 1987 through 1990. Together, these two mechanisms (Figures 6a and 6b) on the left-hand side of equation (2) yielded a warming tendency in the tropical lower troposphere when averaged from 20°S to 20°N (light blue to dark blue color contours) that balanced the anomalous diabatic cooling (Figure 6c) on the right-hand side of equation (2). This anomalous diabatic cooling of the tropical lower troposphere from 1987 through 1990 represented below-average net radiative forcing and latent heat release associated with an increase in anomalous temperature of the troposphere (Figures 1c, 3c, and 3d) and the downward −〈Q′S + Q′L〉 across the air-sea interface into the tropical upper ocean (Figure 3g).

[26] To quantify the source of anomalous quasi-decadal tropical global average sensible plus latent heat flux into the upper ocean, we compute the spatial average of the temporal lag regression coefficients between −(Q′S + Q′L) and all corresponding terms in equation (1) at each grid point over the tropical global ocean from 20°S to 20°N for 25 years from 1975 through 1999 as a function of height from the sea surface to 20 km (Figures 7a–7g). These regression coefficients are multiplied by the standard deviation of anomalous sensible plus latent heat flux anomalies to yield the correct units of °K s−1.

Figure 7.

(a) Vertical sections of temporal lag regression coefficients between tropical global average quasi-decadal −〈Q′S + Q′L〉 (positive downward) and corresponding terms of the atmospheric thermal budget in equation (1): (a) equation image · ∇T′, (b) V′ · ∇equation image, (c) ω′(equation image/equation imageEequation imageE/∂p), (d) equation image(equation image/equation imageE)(∂θ′E/∂p), (e) ∇ · (equation image)′, (f) ∂(equation image)′/∂p, and (g) QD/(ρACPA). These regression coefficients extend over the height of the troposphere from 0 to 20 km height, with temporal lag regression computed for the 25 years from 1975 through 1999. Usual regression coefficients are multiplied by the root-mean-square (RMS) of −〈Q′S + Q′L〉 so that resulting regression coefficients yield both magnitude and phase information in units of °K s−1, with a contour interval of 0.2 × 10−7 °K s−1.

[27] This average temporal lag regression finds positive (downward) −〈Q′S + Q′L〉 and negative 〈Q′D/(ρACPA)〉 in the lower troposphere (0–5 km height) fluctuating in phase, with the latter lagging the former by ∼1 year (Figure 7g). This nominal phase alignment is supported by magnitude agreement, that is, integrating 〈QD′〉 over the troposphere from 0 to 13 km height, where the contribution in the upper troposphere from 6 to 13 km is relatively small because of the reduction in air density with height, yields an anomalous net flux of heat across the air-sea interface of ∼0.5 W m−2, which is the same as the magnitude of the −〈Q′S + Q′L〉 measured from bulk formulae (Figure 3g). Thus the gain in anomalous diabatic heat in the tropical upper ocean during the cool-to-warm transition from 1987 through 1990 at the rate of ∼0.5 W m−2 [White et al., 2003a] was matched by a loss of anomalous diabatic heat in the lower troposphere (0–5 km height) at the rate of ∼0.5 W m−2 via the anomalous sensible plus latent heat flux across the air-sea interface.

[28] The anomalous diabatic cooling in the tropical lower troposphere (Figure 7g) on the right-hand side of equation (1) was balanced by the warming tendency from the anomalous vertical thermal advection in the tropical lower troposphere (Figures 7c and 7d) on the left-hand side. Here, we find the latter warming tendency (i.e., negative on the left hand side of the equation) dominated by the mean vertical advection of anomalous equivalent potential temperature (Figure 7d), modified by a weaker contribution from the anomalous vertical advection of mean equivalent potential temperature (Figure 7c). The other terms in equation (1) (Figures 7a, 7b, 7e, and 7f) displayed much weaker covariance with the sensible plus latent heat flux anomalies, consistent with the dominant balance given in equation (2). Thus the change in the vertical gradient of the equivalent potential temperature in the presence of the mean vertical circulation of the Hadley cell produced the anomalous warming tendency in the tropical lower troposphere that balanced the diabatic cooling associated with the downward anomalous sensible plus latent heat flux into the tropical upper ocean.

7. Discussion and Conclusions

[29] White et al. [1997] found global average SST variability from 40°S to 60°N associated with the QDO dominated by tropical global average SST variability from 20°S to 20°N, both fluctuating in phase with the 11-year signal in STI variability throughout the 20th century [Lean et al., 1995]. Subsequently, White et al. [1998] compared the quasi-decadal signal in tropical global average upper ocean temperature 〈T′O〉 variability with that expected from a transient form of the Stefan-Boltzmann radiation model. They found the surface radiative forcing (i.e., 0.10–0.15 W m−2) associated with the 11-year signal in STI variability producing amplitudes too small and phase lags too large to account for the observed anomalous 〈T′O〉 tendency. Subsequently, White et al. [2003a] diagnosed the tropical global average diabatic heat storage (〈DHS〉) budget of the quasi-decadal signal in 〈T′O〉 variability over the 25 years from 1975 through 1999, finding the anomalous 〈DHS′〉 warming tendency averaged across the tropical global ocean from 20°S to 20°N driven by downward −〈QS + QL〉 anomalies of ∼0.5 W m−2, the latter peaking during the transition between peaks and troughs in the 11-year signal in STI variability. In the present study, we sought to understand the source of these quasi-decadal −〈Q′S + Q′L〉 anomalies.

[30] We began by examining animation sequences of the meridional vertical sections of quasi-decadal zonal average temperature anomalies in the upper ocean and atmosphere from 40°S to 40°N over one cycle of the QDO variability from 1986 through 1995 [White et al., 2003a]. We found the warm and cool phases of the QDO in the tropical upper ocean from 20°S to 20°N associated with warm and cool temperature anomalies in the atmosphere, respectively, that were uniform in sign across the tropics. We found tropical global average temperature anomalies from 20°S to 20°N increasing downward from ∼0.80°K in the lower stratosphere to ∼0.25°K in the upper troposphere, to 0.15°K in the lower troposphere, and to ∼ 0.05°K in the upper ocean. Furthermore, we found the quasi-decadal signal in tropical lower troposphere temperature anomalies larger than that in the corresponding SST anomalies by ∼0.1°K, yielding positive anomalous tropical global average air-sea temperature differences (〈T′2m − SST′〉) that correlated nearly one to one with the downward anomalous tropical global average sensible plus latent heat flux (−〈Q′S + Q′L〉) from 1975 through 1999. Thus the 〈Q′S + Q′L〉 that drove the warming tendency in 〈T′O〉 and 〈DHS′〉 in the tropical upper ocean [White et al., 2003a] was in turn driven by larger tropical global average temperature (and presumably specific humidity) anomalies in the overlying troposphere.

[31] To understand how this quasi-decadal tropical global average temperature variability in the troposphere was generated, we diagnosed the thermal budget of the quasi-decadal signal in the tropical atmosphere from 20°S to 20°N using the NCEP/DOE Reanalysis II data set [Kanamitsu et al., 2002] over the 25 years from 1975 through 1999. This diagnosis found the principal balance occurring between the anomalous diabatic heating and the anomalous vertical thermal advection throughout the column. This indirect estimation of the tropical global average diabatic heating allowed us to begin understanding how the tropical global average temperature in the lower stratosphere and troposphere came to warm and cool during corresponding tropical warm and cool phases of the QDO in the tropical upper ocean and lower troposphere [White et al., 2003a, 2003b].

[32] Earlier, Labitzke and van Loon [1997a, 1997b, 1999] found the quasi-decadal signal in temperature variability in the tropical lower stratosphere fluctuating in phase with the nominal 11-year signal in STI variability, with stratospheric ozone absorbing the UV portion of the STI spectrum. Following this lead, we equated the anomalous quasi-decadal tropical global average diabatic heating to the difference between an unknown shortwave radiative forcing and the known anomalous longwave radiation to space and to the troposphere below. This computation yielded indirect estimates of the radiative forcing by stratospheric ozone absorption of the 11-year signal in the solar UV irradiance variability, modified by the radiative forcing by anomalous partial volcanic aerosol absorption of mean solar IR irradiance, as proposed by Labitzke and van Loon [1999]. This indirect estimate of net radiative forcing fluctuated in phase with that estimated more directly (Appendix A) with an amplitude (i.e., 1.0–1.5 W m−2) that was only ∼20% larger than that estimated more directly (i.e., 0.8–1.2 W m−2).

[33] In the tropical lower troposphere, we found the quasi-decadal signal in tropical global average diabatic heating variability fluctuating out of phase with that in −〈Q′S + Q′L〉, with its vertical integral over the height of the troposphere estimated to have been ∼0.5 W m−2, nominally the same as that of −〈Q′S+ Q′L〉 estimated from bulk formulae [White et al., 2003a]. This indicated that the gain of 〈DHS′〉 in the tropical upper ocean from 20°S to 20°N observed by White et al. [2003a] was balanced by the loss of anomalous diabatic heat in the tropical troposphere, the latter transferred downward across the air-sea interface via a turbulent transfer of heat and moisture.

[34] Finally, we examined the thermal budget terms that balanced of anomalous tropical global average diabatic cooling in the lower troposphere and downward turbulent transfer of heat across the air-sea interface during the cool-to-warm transition. We found the anomalous vertical gradient in equivalent potential temperature between the tropical lower troposphere and lower stratosphere driven by the radiative forcing of the latter by stratospheric ozone absorption of the 11-year signal in the solar UV irradiance variability (modified by the variable volcanic aerosol absorption of mean solar IR irradiance). This allowed the mean vertical advection of anomalous equivalent potential temperature associated with the normal Hadley cell circulation to drive an anomalous warming tendency in the tropical lower troposphere that balanced the anomalous diabatic cooling there, and drove the downward sensible plus latent heat flux across the air-sea interface into the tropical upper ocean.

[35] This work expanded upon that of Labitzke and van Loon [1997a, 1997b, 1999] and Haigh [1994, 1999, 2003], who focused on the response of the tropical lower stratosphere and troposphere during peaks and troughs in the 11-year signal in STI variability and it UV component. Here we found quasi-decadal anomalies of the vertical and horizontal circulation, the temperature lapse rate, and the air-sea temperature difference averaged across the tropical troposphere from 20°S to 20°N maximum during the transition between peaks and troughs in the 11-year signal in the solar UV irradiance variability, and between cool and warm phases of the QDO in the tropical upper ocean and lower troposphere [White et al., 2003a, 2003b]. Of all the potential thermal mechanisms for balancing the anomalous diabatic heating in the lower troposphere during these transitions, it was the change in temperature lapse rate that was the most important, not the change in vertical velocity.

[36] Earlier, Douglass and Clader [2002] and Gleisner and Thejll [2003] found the volcanism of El Chichon in April of 1982 and Mt. Pinatubo in July of 1991 projecting principally onto the 3–5 year period ENSO. We found these two volcanic eruptions projecting also onto the 11-year period signal QDO. We found their influence advancing peaks and troughs in the radiative forcing of the lower stratosphere by the 11-year signal in solar UV irradiance variability by ≤1 year over the 25-year record from 1975 through 1999, and increasing their magnitude by another ∼2/3. In the future, we may be able to separate these two effects by diagnosing the lower stratosphere thermal budget using monthly (i.e., unfiltered) data, as suggested by Haigh [2003].

Appendix A:: Net Radiative Forcing of the Lower Stratosphere

[37] We estimate the amplitude and phase of the quasi-decadal signal in the radiative forcing of the lower stratosphere (S′UV) by the 11-year signal in solar UV irradiance variability from consideration of the 11-year signal in the STI variability [Lean et al., 1995]. We assume the 11-year signal in S′UV fluctuated in phase with the radiative forcing associated with STI variability, but with different amplitude. Estimating this amplitude begins by recognizing that while the mean UV portion of the solar spectrum represents ∼10% of the mean STI (∼1366 W m−2), the amplitude of its 11-year signal was 1.0–1.5% of the mean solar UV irradiance from 1979 through 1999 [Haigh, 2003]. This percentage amplitude was 25–50 times larger than that of the 11-year signal in STI variability, which ranged from 0.03% to 0.04% of the mean STI over the same period. This translates into an amplitude for the 11-year signal in solar UV irradiance variability that was ∼ 4 times larger than the 0.4–0.5 W m−2 amplitude of the 11-year signal in STI variability (Figure 1a). However, the resulting radiative forcing of the tropical lower stratosphere (i.e., S′UV) was less than this by a factor ∼3 because of averaging over the Earth's tropical strip from 20°S to 20°N. Thus the amplitude of the 11-year signal in S′UV ranged from ∼0.5–0.7 W m−2 over this period (thick line in Figure A1a).

Figure A1.

(a) Time sequence of the anomalous radiative forcing (S′UV) by stratospheric ozone absorption of anomalous solar UV irradiance variability, with monthly anomalies (thin line) and the band-pass-filtered quasi-decadal anomalies (thick line) displayed from 1900 through 1999. (b) Same as Figure A1a but for the anomalous radiative forcing (V′IR) by the variable volcanic aerosol absorption of mean solar IR irradiance in the lower stratosphere. (c) Same as Figure A1a but for the tropical global average temperature variability at 15 km height (〈T′15km〉) in the lower stratosphere from 1979 through 1999.

[38] Labitzke and van Loon [1999] and Robock [2000] observed positive temperature anomalies across the tropical and extratropical lower stratosphere during the 2–3 years following each volcanic eruption in response to the radiative forcing (V′IR) by the partial absorption by volcanic aerosols of the mean solar IR radiation and of longwave radiation from below. Here, we estimate the amplitude and phase of quasi-decadal variability in V′IR (thick line in Figure A1b) from monthly estimates given by Sato et al. [1993] (thin line in Figure A1b). Its amplitude ranged from 0.3–0.5 W m−2 from 1979 through 1999, which was ∼2/3 of S′UV (thick lines in Figures A1a and A1b). Recently, Gleisner and Thejll [2003] concluded that the radiative forcing of the lower stratosphere by the quasi-decadal variability in volcanic aerosols during the most recent epoch of volcanism could be isolated from that of the 11-year signal in solar UV irradiance variability. Here, we find the V′IR associated with the volcanic eruptions of El Chichon in April of 1982 and Mt. Pinatubo in July of 1991 (thin line in Figure A1b) projecting a significant proportion of their variance onto the quasi-decadal variability from 1979 through 1999 (thick line in Figure A1b) with an amplitude smaller but on the same order as that of S′UV.

[39] The available observational and model evidence indicates that radiative forcing by the 11-year signal in solar UV irradiance variability accounted for most of the quasi-decadal signal in 〈T′15km〉 from 1979 through 1999 [e.g., Labitzke and van Loon, 1999; Haigh, 2003]. However, we find the radiative forcing by quasi-decadal variability in volcanic aerosol production modifying the former, so that the amplitude of net radiative forcing ranged from 0.8–1.2 W m−2 (thick lines in Figures A1a and A1b). In the main text, we indirectly estimate the amplitude of this net radiative forcing by diagnosing the lower stratosphere thermal budget (Figure 6d), yielding residual estimates ranging from 1.0–1.5 W m−2 when averaged from 20°S to 20°N and over the 13–20 km height of the available lower stratosphere (Figure 6d). These indirect estimates are ∼20% larger than 0.8–1.2 W m−2 estimated more directly, any differences certainly within the error limits of both estimates.

[40] When we compare the amplitude and phase of the quasi-decadal variability in S′UV and V′IR (thick lines in Figures A1a and A1b) with those of the tropical global average temperature variability in the lower stratosphere (〈T′15km〉) [Kanamitsu et al., 2002] (thick line in Figure A1c), we find peaks in 〈T′15km〉 in 1981 and 1991 lagging peaks in S′UV by ∼1 year and leading peaks in V′IR by 1–2 years. Thus the net radiative forcing by (S′UV + V′IR) explains why peaks in 〈T′15km〉 in 1981 and 1991, and troughs in 1986/87 and 1997 (thick line in Figure A1c) were advanced ≤1 year in time from corresponding peaks and troughs in the 11-year signal in S′UV (thick line in Figure A1a).

Acknowledgments

[41] Appreciation is extended to Ted Walker, who, as programmer, is responsible for the analyses conducted in this study. We extend our thanks to Andrea Fincham, who is responsible for drafting the figures. White is supported by the National Aeronautics and Space Administration (NASA) program “Living with a Star” under contract NAG5-12465. White is also supported by the Scripps Institution of Oceanography of the University of California, San Diego.

Ancillary