QBO deepens MJO convection

The study investigates how the stratospheric Quasi-Biennial Oscillation (QBO) modulates the Madden–Julian Oscillation (MJO), the leading mode of tropical intraseasonal variability. The MJO exerts widespread impacts on global weather and climate extremes, yet its maintenance and propagation mechanisms remain incompletely understood and are often poorly represented in models. Recent observations show that MJO activity and its eastward propagation across the Indo-Pacific warm pool are stronger during easterly QBO (EQBO) winters than during westerly QBO (WQBO) winters, but the physical mechanism has been unclear and models struggle to reproduce the relationship. A prevailing hypothesis posits that QBO-induced temperature anomalies near the tropical tropopause alter upper-troposphere–lower-stratosphere (UTLS) static stability, modulating the vertical development of deep convection within MJO envelopes and the associated longwave (LW) cloud–radiation feedback. This work tests, with observations, whether deep convective systems within MJO envelopes are systematically deeper and radiatively more influential during EQBO winters.

Prior work identified the QBO as a key interannual mode in the equatorial stratosphere with potential to influence tropical convection and, more recently, the MJO. Studies have suggested that QBO-related UTLS temperature anomalies (including a strengthened cold cap near ~100 hPa during EQBO) can destabilize the upper troposphere, facilitating deeper convection above MJO events. LW cloud–radiation feedback has been recognized as a crucial maintenance mechanism for the MJO, affecting its scale, strength, and ability to cross the Maritime Continent. Modeling studies and idealized frameworks proposed that QBO-induced stability changes could uniquely affect the MJO because of its sensitivity to LW cloud–radiation feedback, but robust observational confirmation of deeper MJO convection and altered radiative effects under different QBO phases had been lacking. Furthermore, GCMs generally fail to reproduce the observed QBO–MJO linkage, even with QBO nudging, highlighting the need for observational constraints.

The study combines multiple satellite and reanalysis datasets to quantify cloud, precipitation, radiative, and thermodynamic properties of deep convective systems within MJO envelopes under different QBO states, focusing on boreal winter (DJF). Key elements include:

• Period and regions: Primary focus on DJF 1980–2021 (41 winters) and, for satellite products, 2002–2021 (19 winters). Analyses emphasize the Maritime Continent (MC: 100°E–150°E, 15°S–5°N), with comparisons to the Indian Ocean (IO) and West Pacific (WP).
• Event selection and compositing: QBO phase defined by zonal-mean 50-hPa wind (U50) averaged over 5°S–5°N: EQBO if U50 < −7 m s−1, WQBO if U50 > 7 m s−1, otherwise neutral. ENSO state by DJF Niño3.4; winters with Niño3.4 > 0.5 (“El Niño”) are excluded to form Non–El Niño composites. MJO identified via the Real-time Multivariate MJO (RMM) index; core analyses emphasize phases 4–5 (MJO over MC), with all amplitudes and a subset with 1 ≤ amplitude < 2 for robustness.
• Cloud–precipitation regimes (CPRs): Use the Pr6x1 hybrid regimes (k-means clustering) from Jin et al. (2021) based on MODIS CTP–COT joint histograms (42 cloud bins) combined with IMERG precipitation histograms (6 bins). Four deep convective regimes are analyzed: Core (regimes 1–2) and Anvil (regimes 4 and 6). Regime occurrence is quantified by relative frequency of occurrence (RFO). Hovmöller diagrams (15°S–5°N-mean) track propagation characteristics.
• Datasets: MODIS (Terra/Aqua) Level-3 daily 1° products for CTP, COT, joint histograms; IMERG V06B Final for precipitation; CERES SYN1deg for column radiative flux divergence and OLR; NCEP/CPC Merged IR for brightness temperature (BT); MERRA-2 for atmospheric temperature and wind profiles.
• Radiative metrics: 24-h mean total (SW+LW) column radiative flux divergence and OLR are computed for CPR grid cells, centered on local noon (±12 h), averaged across matched Terra/Aqua overpasses.
• Neutral buoyancy level (NBL) analysis: Compute NBL (equilibrium level) of non-entraining, pseudo-adiabatic plumes using MetPy’s EL function, with MERRA-2 temperature and humidity on 34 pressure levels (to 30 hPa). Surface conditions replace the lowest level where appropriate. Assess the fraction of grid cells with NBL ≤100 hPa during WQBO and EQBO composite days under MJO phases 4–5 (and 1 ≤ amplitude < 2).
• Temperature modification experiment: Apply the mean EQBO–WQBO temperature difference above 100 hPa over the MC to perturb profiles (subtract from EQBO; add to WQBO) and recompute NBL distributions to isolate the role of UTLS temperature anomalies.
• Statistical testing: Differences between EQBO and WQBO composites are evaluated using t-tests; p-values reported for RFO, CTP, BT, precipitation, and radiative metrics. Correlations of T50–100 anomalies with U50 and Niño3.4 are assessed to motivate the Non–El Niño focus.
• Seasonality and regionality: Examine seasonal composites of EQBO–WQBO UTLS temperature differences and seasonal cycles of Core regime RFO and mean CTP to explain why effects are strongest in DJF and over the MC.

• QBO–temperature linkage and ENSO filtering: DJF T50–100 anomalies over the MC correlate strongly with U50 (r ≈ 0.699) and weakly with Niño3.4 (r ≈ 0.069). When considering both, EQBO–WQBO contrasts in T50–100 are large during La Niña/neutral winters but reduced in El Niño winters; extreme El Niño winters lack EQBO cases. Analyses therefore focus on Non–El Niño winters.
• UTLS mean-state differences: In WQBO winters, positive temperature anomalies occur from ~40–100 hPa; in EQBO winters, negative temperature anomalies peak near ~70–90 hPa and extend below 100 hPa, weakening static stability near the cold-point tropopause. This favors deeper convection in EQBO and inhibits it in WQBO.
• MJO propagation signature: Hovmöller diagrams of Core and Anvil RFO show more frequent and continuous eastward propagation across the MC into the WP during EQBO winters, while WQBO winters exhibit stagnation over the MC; QBO-neutral winters show intermittent propagation.
• Enhanced deep convective occurrence during EQBO: Over the MC during MJO phases 4–5, Core RFO increases from 12.2% (WQBO) to 15.9% (EQBO), and Anvil RFO from 25.8% to 31.0%; differences are statistically significant (p < 0.005). Similar contrasts hold when restricting to moderate MJO amplitudes (1 ≤ amplitude < 2).
• Deeper clouds in EQBO: For Core (regimes 1–2) and Anvil (regimes 4, 6) regime grid cells in the MC under moderate MJO amplitudes, cloud-top pressure (CTP) and brightness temperature (BT) distributions are systematically lower in EQBO than WQBO with significance ≥97.5% (regime-dependent p-values reported). This indicates higher cloud tops and colder emitting temperatures in EQBO.
• Neutral buoyancy levels: Fraction of grid cells with NBL ≤100 hPa is 5.1% (WQBO) vs 18.4% (EQBO). Applying the mean EQBO–WQBO UTLS temperature difference reduces the EQBO–WQBO gap in NBL ≤100 hPa by ~36% when subtracted from EQBO profiles (13.3% → 8.5%) and by ~18% when added to WQBO profiles (13.3% → 10.9%), demonstrating that UTLS temperature anomalies materially affect the tallest plumes.
• Stronger LW cloud–radiation feedback in EQBO: CERES diagnostics show that OLR is reduced more in EQBO than WQBO for all Core and Anvil regimes, yielding lower column radiative flux divergence (i.e., greater column energy retention). Some Core regime grid cells even exhibit net radiative energy gain. Differences in cloud optical depth are less robust, implying the radiation budget changes are mainly due to higher cloud tops.
• Precipitation differences: Over the MC, grid-mean precipitation differences between EQBO and WQBO for regimes 1, 2, 4, 6 are statistically insignificant, suggesting enhanced LW feedback (often measured by rain rate to OLR anomaly ratio) without corresponding increases in rain rate. In the WP, regime 1 precipitation is higher in EQBO (p = 0.014), and LW feedback contrasts are weaker because OLR reductions coincide with increased precipitation.
• Seasonality and regionality: EQBO–WQBO UTLS temperature differences are largest in DJF and confined to the deep tropics; deep convective Core regimes are most frequent and reach highest altitudes over the MC/WP in DJF. This co-location explains why the QBO–MJO connection is strongest in DJF and over the MC.

The findings provide observational evidence that EQBO-induced UTLS cooling reduces static stability near the tropopause, allowing deeper growth of deep convective systems within MJO envelopes. Higher cloud tops lower OLR and enhance LW cloud–radiation feedback, which supports stronger and more persistent MJO convective envelopes and facilitates their eastward propagation across the Maritime Continent. The consistency between deeper clouds, reduced OLR, increased column radiative energy retention, and increased occurrence of Core/Anvil regimes under EQBO aligns with the proposed QBO temperature stratification mechanism. The seasonality and regionality of both UTLS temperature anomalies and the population of deep convective systems explain why the QBO–MJO connection is most evident in DJF and over the MC. While enhanced LW feedback strengthens the MJO’s maintenance, competing effects (e.g., changes in vertical velocity profiles) may exist and warrant further study. The results offer targets for model evaluation and development toward capturing the QBO–MJO coupling.

This study demonstrates, using satellite observations and reanalysis, that the QBO deepens MJO-associated convection: EQBO winters exhibit systematically higher cloud tops (lower CTP and BT) of deep convective and anvil clouds, stronger suppression of OLR, and enhanced LW cloud–radiation feedback within MJO envelopes. These changes are accompanied by increased occurrence and improved eastward propagation of deep convective systems across the MC. A plume NBL analysis and a temperature-modification experiment confirm that UTLS temperature anomalies linked to the QBO materially contribute to the deepest convection. The QBO–MJO connection is strongest in DJF and over the MC, where UTLS temperature anomalies and deep convection are both maximized. Future work should: (i) examine additional mean-state variables beyond UTLS temperature, (ii) investigate ENSO’s nonlinear interference with QBO signals in the UTLS, (iii) leverage geostationary satellite data to resolve diurnal/lifecycle evolution of convection, (iv) extend records for increased statistical power, and (v) use these observational constraints to improve GCM representations of the QBO–MJO relationship.

• Sample-size constraints: Excluding El Niño winters (Niño3.4 > 0.5) to isolate QBO effects reduces the number of composite days (e.g., 59 WQBO and 44 EQBO days for some MC analyses), though key results remain statistically significant.
• Observational sampling: Reliance on polar-orbiting satellites (Terra/Aqua) limits diurnal coverage and may miss parts of convective lifecycles; regime assignments are centered on local noon.
• Attribution scope: The analysis focuses on QBO-induced UTLS temperature changes; other contributing mechanisms (e.g., wave-induced ice cloud modulation) cannot be ruled out.
• Regional variability: Some contrasts (e.g., in the IO and WP) are weaker or less statistically robust than in the MC, suggesting spatial heterogeneity.
• Dataset limitations: Known IMERG biases for light precipitation are minor here but still present; MODIS/CERES/NCEP IR and MERRA-2 reanalysis have inherent uncertainties.
• Model generalization: While observations provide strong evidence, translating these findings to model improvements requires careful tuning and verification across different GCM frameworks.