This study utilizes observations from the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument aboard the Thermosphere Ionosphere Mesosphere Energetics and Dynamics (TIMED) satellite . The SABER instrument is a 10-channel broadband limb-scanning infrared radiometer that measures atmospheric emissions in the 1.27 to 17 µm spectral range, providing vertical profiles of kinetic temperature, geopotential height, and volume emission rates (VER, in units of W m⁻³ or photons cm⁻³ s⁻¹) of various atmospheric constituents from approximately 10 to 120 km altitude . SABER employs a limb-scanning technique, viewing the Earth's atmosphere tangentially through the limb scanning to obtain vertical profiles with high vertical resolution (approximately 2 km). The instrument scans the limb from the surface to approximately 400 km altitude, with a horizontal sampling resolution of about 400 km along the satellite track. The TIMED satellite operates in a 625 km circular orbit with a 74° inclination, providing near-global coverage from 83° S to 83° N latitude, with the latitudinal coverage varying seasonally due to the satellite's yaw cycle . A known limitation of SABER is the inability to observe both polar regions simultaneously due to the yaw maneuver cycle, which restricts continuous high-latitude coverage.

This orbital configuration enables SABER to sample each latitude and local time approximately every 60 d, providing temporal and spatial coverage for climatological studies. For this investigation, we used SABER Level 2A version 2.0 data products spanning from January 2002 to December 2023, encompassing over two decades of continuous observations. The primary data products employed include: (1) kinetic temperature profiles derived from CO₂ 15 µm limb emission measurements, with an estimated precision of 1–2 K in the mesosphere ; (2) OH_VER derived from the 1.6 and 2.0 µm OH Meinel band emissions, providing information about the OH airglow layer structure and intensity ; and (3) geopotential height profiles that enable accurate altitude registration of the measurements. The SABER temperature measurements are suited for gravity wave studies due to their high vertical resolution and global coverage. SABER's temperature sensitivity of 1–2 K enables detection of gravity wave signatures in the middle atmosphere . Similarly, the OH_VER measurements provide information about the mesospheric airglow layer, which serves as a tracer for atmospheric dynamics and chemistry . The long-term stability and consistency of SABER measurements have been validated through comparisons with other satellite instruments and ground-based observations, confirming their suitability for climatological trend analysis . We note that SABER retrieval errors increase with altitude, particularly above 90 km where the CO₂ emission signal weakens. However, no systematic temporal trends in retrieval precision have been identified in validation studies spanning the mission lifetime , and the focus of this study on the OH peak altitude (∼85–87 km) lies within the altitude range where SABER temperature retrievals are most reliable. Nevertheless, residual long-term drifts in instrument sensitivity cannot be entirely excluded, and results with marginal statistical significance should be interpreted with this caveat in mind.

The hydroxyl (OH) airglow layer in the mesosphere represents a tracer for atmospheric dynamics and chemistry, originating from the exothermic reaction between atomic hydrogen and ozone: H + O₃ → OH^* + O₂, where OH^* denotes vibrationally excited hydroxyl . The resulting vibrationally excited OH molecules emit characteristic infrared radiation in the Meinel band system, which SABER detects at 1.6 and 2.0 µm wavelengths . This study focuses on the peak of the OH emission layer, following established methodologies that have demonstrated the peak region's sensitivity to atmospheric perturbations and its reduced susceptibility to retrieval uncertainties compared to integrated column measurements .

The OH_VER profiles are processed through a multi-step quality control and analysis procedure. First, individual profiles are screened for data quality using the SABER data quality flags, removing profiles with instrument anomalies, cloud contamination, or retrieval convergence issues. Profiles with unrealistic VER values (negative emissions or values exceeding 1×10-6 W m⁻³) is excluded from the analysis . We apply altitude registration corrections using co-measured geopotential heights to ensure consistent altitude referencing. The identification of the OH emission peak follows the methodology established by for characterizing the OH emission layer. The OH peak identification algorithm employs a multi-step quality control approach: (1) smooth the VER profile using a 3-point running mean to reduce noise, (2) identify the altitude of maximum VER within the expected mesospheric range (75–95 km), (3) ensure that the identified peak represents a genuine OH emission maximum by requiring the peak value to exceed half of the profile maximum, and (4) verify the peak structure by checking that the VER decreases monotonically on both sides of the identified maximum . Profiles where the peak cannot be unambiguously identified or where multiple peaks of similar magnitude exist are flagged and excluded from further analysis; approximately 5 %–10 % of profiles are discarded through this quality control procedure. The peak identification algorithm extracts two key parameters for each profile: (1) the peak OH_VER intensity (W m⁻³), representing the maximum emission rate, and (2) the peak altitude (km), indicating the height of maximum emission.

The VER provides a direct measure of OH emission intensity that is suitable for long-term trend analysis and comparison with atmospheric models . Following peak identification, the data undergo temporal and spatial binning to create regular time series suitable for climatological analysis. Individual profiles are aggregated into monthly means within 10° latitude bins, extending from 50° S to 50° N. This binning strategy balances the need for adequate statistical sampling with sufficient spatial resolution to capture latitudinal variations in OH behavior. Only latitude bins containing at least 20 individual profiles per month are retained to ensure statistical robustness .

Gravity wave Ep per unit mass (Ep) represents a measure of wave activity and provides direct quantification of the energy available for momentum deposition and atmospheric mixing . The calculation of gravity wave potential energy (Ep) from SABER temperature measurements follows established methodologies developed for limb-sounding satellite observations . The approach involves extracting gravity wave temperature perturbations from the background atmospheric state and converting these perturbations to Ep. Following the methodology detailed in and extended by , the processing begins with the extraction of temperature perturbations from SABER kinetic temperature profiles. Background temperature profiles are obtained by removing large-scale wave signatures: planetary waves and Kelvin waves with zonal wavenumbers 0–6 are removed using a spatial Fourier decomposition applied to ascending and descending orbit nodes at each latitude, following the approach detailed by and adapted by and .

Temperature perturbations are calculated as T′(z)=T(z)-T‾(z), where T(z) is the observed temperature profile and T‾(z) is the background temperature obtained from the running mean filter. Additional filtering is applied to isolate gravity wave activities, with the temperature perturbations undergoing vertical wavelength filtering to retain perturbations with vertical wavelengths between 3 and 20 km. This wavelength range is chosen to capture gravity waves reliably detected by SABER's vertical resolution while excluding short-scale instrumental noise (wavelengths <3 km) and planetary wave signals (wavelengths >20 km) . Temperature perturbations are isolated using continuous wavelet transform (CWT) with a Morlet wavelet, providing superior time-frequency localization for gravity wave characterization as described by . The Ep per unit mass is calculated using the fundamental relationship : 1Ep=12gN2T′T02 where g is the gravitational acceleration (9.81 m s⁻²), N is the buoyancy frequency (Brunt-Väisälä frequency), T′ is the gravity wave temperature perturbation amplitude, and T0 is the background temperature. The buoyancy frequency is calculated from the background temperature profile using: 2N2=gT0dT0dz+gcp where dT0/dz is the background temperature gradient, z is altitude, and cp is the specific heat of air at constant pressure (approximately 1004 J kg⁻¹ K⁻¹). The Ep calculations are performed within 5 km thick altitude layers centered on the OH peak altitude. This layer thickness was chosen to provide a representative measure of gravity wave activity in the immediate vicinity of the OH layer while maintaining sufficient vertical resolution to capture altitude-dependent variations and ensuring adequate statistical sampling of wave perturbations. The temperature perturbation amplitude T′ is calculated as the root-mean-square (RMS) value of the filtered perturbations within each altitude layer, following the statistical approach validated by for stratospheric gravity wave analysis.

Quality control procedures ensure the reliability of Ep estimates. Profiles with unrealistic buoyancy frequencies (N2<0, indicating convective instability, or N2>1×10-3 s⁻², exceeding physically plausible values for the mesosphere) are excluded; this criterion removes approximately 3 % of profiles, predominantly at high altitudes where temperature lapse rates approach the adiabatic value. Ep values exceeding 1000 J kg⁻¹ are flagged as outliers and excluded; this threshold corresponds to a temperature perturbation amplitude of approximately 15 K (for typical mesospheric N and T0 values), which represents the upper range of observed gravity wave amplitudes in the MLT region and removes fewer than 1 % of remaining profiles. We note that N2 can occasionally exceed g/cp in the MLT due to steep temperature gradients near the mesopause; however, the filtering criteria described above are applied conservatively. A screening based on instrument uncertainty rather than fixed thresholds could provide an alternative approach, but the SABER Level 2A data products do not include profile-by-profile uncertainty estimates that would enable such screening. The final Ep dataset undergoes the same spatial and temporal binning procedure as the OH data, creating monthly mean time series in 10° latitude bins for climatological analysis. We emphasise that our Ep values represent gravity wave potential energy per unit mass evaluated at the OH peak altitude. Because atmospheric density decreases exponentially with altitude, Ep per unit mass increases with altitude even for constant wave amplitude; this property is discussed further in Sect. .

This study employs a multiple linear regression (MLR) approach, following the methodology previously used by and widely used in middle atmospheric trend studies . The MLR framework quantifies contributions from specific physical drivers while isolating the long-term trend. While the F10.7 index is not a perfect proxy for solar EUV radiation, it remains the most widely used and validated proxy for long-term atmospheric studies due to its continuous record and strong correlation with solar EUV emissions . The MLR analysis begins with the construction of time series representing primary sources of atmospheric variability. Solar activity is represented by using the 10.7 cm solar radio flux (F10.7 index), obtained from the National Research Council of Canada, which serves as a proxy for solar ultraviolet radiation that drives mesospheric photochemistry . The F10.7 index is suited for atmospheric studies because it correlates with solar EUV emissions that control photochemical processes in the mesosphere and lower thermosphere .

Figure reveals the temporal evolution of mesospheric OH airglow and gravity wave activity over the 22-year SABER observation. The top panel shows the OH_VER, which exhibits latitudinal and seasonal patterns consistent with the underlying photochemical processes governing OH production and loss in the mesosphere. The OH_VER displays seasonal variations with maximum values occurring during local December–January–February (DJF) and June–July–August (JJA) months at mid-latitudes, reaching peak emission rates of approximately 8–10 ×10-7 W m⁻³ in both hemispheres. This seasonal pattern likely reflects the coupling between atomic hydrogen and ozone concentrations, both of which are modulated by seasonal changes in atmospheric dynamics and photochemistry . The latitudinal structure of OH_VER shows an asymmetry between hemispheres, with higher emission rates observed in the Northern Hemisphere (NH), particularly during DJF months. This hemispheric asymmetry can be attributed to differences in planetary wave activity and stratosphere-mesosphere coupling, which affect the transport of atomic oxygen and hydrogen species that control OH chemistry . The equatorial region exhibits relatively stable OH emission rates throughout the year, with values typically ranging between 4–6 ×10-7 W m⁻³, consistent with the reduced seasonal variability expected in tropical latitudes where solar zenith angle variations are minimal.

The middle panel displays the Ep at the OH maximum altitude, providing characterization of gravity wave activity in the mesosphere. The Ep exhibits temporal and spatial variability, with values ranging from less than 36 Jkg-1 to over 48 Jkg-1. Features include increased gravity wave activity during local winter months at mid-latitudes, consistent with tropospheric wave generation and propagation conditions during these periods . This seasonal pattern is consistent with findings from , who demonstrated that gravity wave Ep over South America shows seasonal dependence, with maximum values during JJA months when tropospheric forcing is largest. The spatial distribution of gravity wave activity shows hemispheric asymmetries, with higher Ep values in the NH. We note that our Ep values, measured at the OH peak altitude (∼87 km), do not show the pronounced winter mid-latitude peaks near 50° N/S reported by and at lower altitudes (50–100 km on a fixed grid). This difference arises because: (i) our analysis evaluates Ep at a single altitude (the OH peak) rather than across the full 50–100 km range where the strong stratospheric jet effect on gravity wave filtering is most pronounced; (ii) our latitude coverage extends only to ±50°, whereas the strongest winter GW activity occurs poleward of 50° where the polar night jet is strongest; and (iii) gravity wave Ep per unit mass at the OH peak altitude is modulated by local mesospheric conditions (temperature, stability) that differ from the stratospheric environment .

The equatorial region shows higher Ep values but with inter-annual variability, likely associated with the QBO and other tropical atmospheric phenomena . The solar cycle modulation is visible in both OH_VER and Ep throughout the observation period. The solar maximum around 2014–2015 corresponds to increased OH concentrations globally, reflecting the increased production of atomic hydrogen through solar Lyman-α radiation . Ep exhibits latitude-dependent solar cycle variations, indicating that solar forcing modulates gravity waves through multiple pathways: altered atmospheric stability and background winds . The bottom panel shows the altitude of the OH maximum, which varies between approximately 75 and 90 km with latitudinal and temporal patterns. The OH peak altitude exhibits a seasonal cycle, with higher altitudes during local DJF months in the NH and JJA months in the Southern Hemisphere (SH), consistent with seasonal changes in atmospheric temperature structure and the vertical distribution of atomic oxygen .

Figure shows the seasonal decomposition analysis, providing characterization of the annual variations and the residual inter-annual variability. The seasonal cycles reveal the annual rhythms that characterize mesospheric variability. For OH_VER, the mid-latitude stations (50° N and 50° S) show DJF and JJA maxima, reaching peak values of 4–5 ×10-7 µm at mid-latitudes. The tropical station (10° N) exhibits smaller seasonal variations. The Ep seasonal cycles show larger amplitudes, particularly at mid-latitudes, where DJF and JJA values can exceed their opposite seasons by factors of 2–3, with maxima reaching 7–15 Jkg-1. This reflects the seasonal dependence of gravity wave generation in the troposphere, with increased wave activity during these months when storm tracks are most active. The NH shows higher Ep values compared to the SH, particularly during DJF months, reflecting orographic and convective wave sources. Additionally, differences in the gravity wave environment between the two hemispheres, including the distribution of topographic features, jet stream characteristics, and convective activity. This may contribute to the observed hemispheric asymmetry in Ep . The residual time series (Fig. b) reveals inter-annual variability that remains after removing the seasonal patterns, showing evidence for solar cycle modulation, ENSO influences, and long-term trends. The OH_VER residuals show variations with evidence of solar cycle modulation, while the Ep residuals exhibit higher frequency variability with larger amplitude fluctuations.

OH_VER seasonal cycles show strong latitudinal and hemispheric variations. At 50° latitude, both hemispheres exhibit maxima during local winter (DJF in NH, JJA in SH) with amplitudes of 0.4–0.5 ×10-7 µm, driven by solar zenith angle and atmospheric dynamics. The seasonal amplitude decreases equatorward: ∼ 0.2–0.3 ×10-7 µm at 30° (NH winter maximum) and minimal variation at 10° with near-symmetric hemispheric behavior. The Ep seasonal cycles display larger relative amplitudes, with mid-latitude (50°) maxima of 7–15 Jkg-1 which is 3–4 times the seasonal minima. Residual variability reveals inter-annual fluctuations including enhanced OH_VER during the 2014–2015 solar maximum and persistent hemispheric asymmetries throughout the observation period.

The linear trend analysis (Fig. ) shows latitudinal patterns in the long-term evolution of both OH_VER and Ep over the 22-year observation period. We note that a 22-year time span, while among the longest continuous satellite records available for mesospheric studies, covers only approximately two solar cycles and is generally considered short for robust trend analysis. Moreover, the solar cycles covered here (cycles 23 and 24) had markedly different magnitudes, with cycle 24 being considerably weaker than cycle 23, which may affect the separation of solar-driven variability from long-term trends in the MLR analysis . These limitations should be kept in mind when interpreting the trend and solar response results presented below. Statistical significance of the trends is indicated by blue stars (p<0.05) and red stars (p≥0.05) overlaying the profiles, as detailed in the figure caption. For OH_VER (Fig. a–b), the annual trends are not statistically significant at the 95 % confidence level at most mid-latitude bins (25–50°), where the 95 % confidence intervals overlap zero despite negative central values. Near the equator (±15°), trends are weak and also not statistically significant. The magnitude of the OH trend central values ranges from -0.2 to +0.4 ×10-9 µm per year. Seasonal decomposition reveals distinct DJF and JJA trend patterns: JJA trends are predominantly positive across all latitudes, particularly in NH subtropics, while DJF trends show hemispheric asymmetry – positive in the SH, variable in the NH – though individual latitude bins should be evaluated against their respective confidence intervals.

Ep trends (Fig. c–d) differ from OH_VER, exhibiting stronger latitudinal gradients and larger relative amplitudes. Near the equator (±15°), the trends in potential energy (Ep) are weak and statistically insignificant, remaining close to zero. In contrast, statistically significant positive trends emerge in the mid-latitudes of both hemispheres (25 to 50°). We note that part of the observed Ep trend may reflect the upward shift of the OH emission layer (0.02–0.06 km yr⁻¹; see Fig. ): because Ep per unit mass increases with altitude as atmospheric density decreases, a rising OH peak altitude produces an apparent increase in Ep even without a real change in gravity wave activity. This effect is discussed quantitatively in Sect. .

In the NH, the Ep trend strengthens from approximately 0.03 J kg⁻¹ yr⁻¹ at 25°N to a peak of about 0.05 J kg⁻¹ yr⁻¹ between 30° N and 40° N, before decreasing to around 0.04 J kg⁻¹ yr⁻¹ at 50° N. A similar pattern is observed in the SH, where the trend peaks at approximately 0.05 J kg⁻¹ yr⁻¹ around 40° S. A seasonal analysis reveals distinct patterns. During June–July–August (JJA), trends are consistently positive across all latitudes. However, during December–January–February (DJF), the patterns show hemispheric asymmetry. The NH mid-latitudes exhibit strong positive trends, reaching up to 8–10 J kg⁻¹ yr⁻¹, whereas the SH trends are smaller and more variable.

The solar response analysis (Fig. ) shows the relationship between solar variability and mesospheric processes across all latitudes. As with the trend analysis, statistical significance is indicated by blue and red stars overlaying the profiles (see figure caption for details). For OH_VER (Fig. a–b), the solar response is positive across all latitudes with an equatorial maximum reaching values of 0.8–1.0 ×10-7 µm per 100 sfu (solar flux units). JJA responses are larger than DJF responses, particularly in the NH. The hemispheric asymmetries in solar response are present during DJF months, with the SH showing larger responses than the NH. The Ep solar response (Fig. c–d) shows patterns with both positive and negative responses depending on latitude and season. The annual response shows positive values in tropical regions and negative values at mid-latitudes. The seasonal decomposition of Ep solar response shows differences between JJA and DJF patterns. JJA responses are positive across most latitudes, while DJF responses show hemispheric asymmetries with positive responses in the SH and negative responses in the NH at mid-latitudes.

The hemispheric asymmetries in solar responses (Fig. ) show contrasting behavior between OH and Ep. For OH (Fig. a–b), both hemispheres show positive solar responses, with the SH exhibiting larger responses at mid-latitudes during DJF months. The Ep solar responses (Fig. c–d) show hemispheric differences during the DJF months. The SH displays positive solar responses at most latitudes, while the NH shows negative solar responses at mid-latitudes during DJF.

Figures and examine the responses of mesospheric OH and gravity wave activity to two primary modes of tropical atmospheric variability: the QBO and the ENSO. The QBO response analysis (Fig. ) reveals patterns of mesospheric coupling to stratospheric wind oscillations. Statistical significance of the responses is indicated by blue stars (p<0.05) and red stars (p≥0.05) overlaying the overall profiles, as shown in the figure captions.

The OH response to the QBO in Fig. a shows opposite responses between the equator and midlatitudes. Around the equator (±15°), QBO30 produces positive responses (+2.8×10-9 µm m s⁻¹), while QBO50 induces negative values near -1.0×10-9 µm (m s⁻¹). In contrast, both hemispheres at 25–50° show weaker and opposite-signed trends: in the NH, QBO30 responses are negative (-0.3 to -0.8×10-9 µm m s⁻¹) while QBO50 yields positive (+0.2 to +0.9×10-9 µm m s⁻¹). The SH shows a similar pattern, with QBO30 negative (-0.9 to -1.4×10-9 µm m s⁻¹) and QBO50 positive (∼+0.9 to +1.1×10-9 µm m s⁻¹). Thus, equatorial OH is modulated with opposite phase between QBO30 and QBO50, while midlatitudes show a smaller but consistent opposing trend. Around the equator, variability is evident, with QBO30 responses shifting between positive and negative depending on season, particularly during JJA. In the NH midlatitudes, QBO50 trends dominate as positive anomalies, whereas in the SH midlatitudes, both QBO30 and QBO50 yield small but oppositely signed contributions, with seasonality introducing scatter. Overall, the seasonal structure indicates that the equatorial trend is the largest and most sensitive to phase, while midlatitudes remain weaker and seasonally modulated.

Ep responses (Fig. c) to the QBO also show equatorial–midlatitude contrasts. At the equator (±15°), Ep regressions are mixed, with QBO30 alternating between slightly negative (-0.11 to -0.08 J kg⁻¹ m s⁻¹) and positive (+0.06 to +0.09 J kg⁻¹ m s⁻¹), while QBO50 remains near zero to slightly negative (-0.07 to -0.20 J kg⁻¹ m s⁻¹). In the NH 25–50° N, Ep responses to QBO30 are positive, peaking around +0.25 J kg⁻¹ (m s⁻¹), while QBO50 trends slightly negative (-0.01 to -0.20 J kg⁻¹ m s⁻¹). In the SH midlatitudes, QBO30 is positive (+0.05 to +0.09 J kg⁻¹ m s⁻¹), with QBO50 negative (-0.1 to -0.2 J kg⁻¹ m s⁻¹). This indicates that Ep is modulated by the QBO in the NH midlatitudes, particularly by QBO30.

Figure d shows strong seasonal and latitudinal dependence in QBO modulation. Equatorial Ep remains near-zero or negative year-round, whereas NH mid-latitudes display seasonally dependent responses: positive under QBO30 during JJA, negative or weak under QBO50. In the SH midlatitudes, Ep responses are less consistent but show positive anomalies under QBO30, especially during JJA, and negative or near-zero under QBO50. Seasonality thus enhances the midlatitude asymmetry, with NH responses larger and more coherent than SH.

The ENSO response analysis (Fig. ) shows ENSO responses of OH and Ep across latitude bands. As indicated by the blue and red stars in the figure, statistical significance varies with latitude (see figure caption for details). Figure a indicates that OH ENSO responses are negative across all latitudes, with the largest values in the SH (-3.72×10-7 µm MEI at 50° S, -2.67×10-7 µm MEI at 35° S) and smaller values in the NH (-2.15×10-7 µm MEI at 25° N, weakening to -0.18×10-7 µm MEI at 50° N). Around the equatorial region (±15°), OH decreases are present (-2.11×10-7 µm MEI at the equator, -2.12×10-7 µm MEI at 15° S). Figure b separates seasonal signals, showing that both DJF (orange) and JJA (blue) responses are negative throughout, with larger negative anomalies during JJA, especially in the tropics and subtropics, while DJF shows weaker responses. Figure c depicts Ep ENSO response, which is positive across latitudes. Near the equator (±15°), the response is small (0.03–0.04 J kg⁻¹ MEI), increasing toward higher latitudes: 0.21 J kg⁻¹ MEI at 50° N and 0.17 J kg⁻¹ MEI at 50° S. Figure d shows the seasonal breakdown of Ep responses: both DJF (green) and JJA (purple) display positive anomalies across most latitudes, with larger responses in JJA, particularly in the equatorial and SH regions, whereas DJF is more moderate.

Figure provides the analysis of the correlation patterns between deseasonalized and detrended OH_VER and Ep time series and various atmospheric indices. The correlation analysis for OH_VER (Fig. c) shows positive correlations across all latitudes with maximum values of 0.6–0.7 (95 % confidence interval [CI]: 0.5–0.8, estimated via Fisher's z-transformation) in tropical regions. The ENSO correlations show a latitudinal structure with maximum values in tropical and subtropical regions, reaching correlations of 0.3–0.4 (95 % CI: 0.1–0.5). The QBO correlations are weaker, with the 30 mb index exhibiting larger correlations than the 50 mb index in most regions. The Ep correlations (Fig. d) show different patterns compared to OH_VER. The solar correlations are weaker and more variable, with both positive and negative values depending on latitude. The ENSO and QBO correlations for Ep show latitudinal variations with larger correlations in tropical and subtropical regions. The QBO correlations show differences between the 30 and 50 mb indices. The ENSO correlations show hemispheric asymmetries, with larger correlations in the NH.

The variance decomposition analysis (Fig. ) shows the percentage of variance explained by each predictor as a function of latitude. In the equatorial region (±15°), OH variance (Fig. a) is influenced by solar forcing (≈ 6–8 %) with small contributions from QBO and ENSO, while Ep variance (Fig. b) shows contributions from QBO (up to ∼5 %) and modest ENSO impact. Correlations (Fig. c–d) in the equator show: OH correlates positive with solar but near-zero with QBO and ENSO, whereas Ep shows positive correlation with QBO (30 and 50 mb) and weaker links to solar and ENSO. In the NH (25–50° N), OH variance is dominated by solar (∼10–15 %) with little ENSO or QBO contribution, while Ep variance is split across solar, QBO, and the long-term trend, each explaining a few percent. OH is positively correlated with solar forcing, while Ep shows positive correlations with QBO and trend. In the SH (25–50° S), OH variance is solar-driven (10–15 %), with ENSO explaining a small fraction, whereas Ep variance shows trend and QBO contributions (up to 5–7 %). OH links mainly to solar, while Ep shows positive associations with QBO and trend.

Figure presents the altitude-latitude structure of the OH and Ep climatology along with the spatial distribution of linear trends. In Fig. a (OH_VER), the emission layer is largest near the mesopause around 85–90 km, forming a maximum at the equator (±15°) with relatively uniform latitudinal distribution but a weakening in the NH midlatitudes (25–50° N). In the SH midlatitudes (25–50° S), OH emissions appear somewhat more extended vertically but weaker compared to the equatorial peak. In Fig. b (Ep, the gravity wave potential energy per unit mass), the climatology shows a broad maximum above 85 km, largest near the equator, extending well into the upper mesosphere and lower thermosphere. The equatorial region (±15°) shows the largest and vertically thickest values of Ep, while both hemispheres (25–50° N and 25–50° S) show weaker enhancements that do not extend as high as at the equator. The SH shows slightly larger midlatitude Ep values than the NH. In Fig. c (linear trend slope of OH_VER), the equatorial region (±15°) is dominated by positive trends between 80–95 km, indicating an intensification of OH emissions. In contrast, both hemispheres (25–50° N and 25–50° S) display weaker and less consistent signals: trends are largely not statistically significant, with some patches of negative slopes below 80 km and near 90 km. The SH exhibits more coherent negative trends below 80 km, whereas the NH shows smaller-scale variability without statistical significance.

The altitude trends and solar responses of the OH peak (Fig. ) show the vertical structure changes of the mesospheric OH layer over the observation period. Statistical significance is indicated by blue stars (p<0.05) and red stars (p≥0.05) overlaying the overall profiles, as detailed in the figure caption. In Fig. a, the linear trend in the OH emission peak altitude is positive across all latitudes, with values around 0.02–0.04 km yr⁻¹ near the equatorial region (±15°). In the NH midlatitudes (25–50° N), the overall trend increases slightly, peaking near 0.05 km yr⁻¹ during JJA, while DJF trends remain at ∼0.02 km yr⁻¹. In the SH midlatitudes (25–50° S), the trends are also positive but more seasonally contrasted: JJA trends are larger (up to ∼0.05–0.06 km yr⁻¹), whereas DJF trends are closer to ∼0.01–0.02 km yr⁻¹. In Fig. b, the solar response (km 100 sfu) shows a latitude-dependent asymmetry. Near the equator (±15°), the overall response is small, close to zero, but seasonally distinct – negative during JJA and positive to near zero during DJF. In the NH midlatitudes (25–50° N), the overall solar response remains small but slightly positive, while JJA responses are near zero and DJF responses show negative values. By contrast, in the SH midlatitudes (25–50° S), the overall solar response is negative, dominated by negative values during JJA, while DJF responses are closer to zero or negative.

The seasonal analysis employs harmonic decomposition techniques to extract annual oscillation (AO) and SAO components from the deseasonalized time series, following the methodology established by for OH airglow emission analysis. Sinusoidal functions are fitted to monthly climatological data to quantify amplitude and phase characteristics. We decompose seasonal cycles into AO and SAO components via least-squares harmonic fitting, where AO captures solar-driven annual variations and SAO represents semiannual oscillations linked to equinoctial dynamics and stratospheric circulation. The amplitude represents the magnitude of seasonal variation relative to the annual mean, while the phase indicates the timing of maximum values throughout the year, providing insights into the physical mechanisms controlling seasonal variability in mesospheric OH chemistry and gravity wave activity.

The seasonal climatology (Fig. ) reveals the annual and semi-annual patterns that drive mesospheric variability. In Fig. a, OH_VER shows a semiannual cycle with equatorial maxima (±15°) around April–May and October–November, while minima occur during solstitial months. In the NH midlatitudes (25–50° N), OH values are weaker overall, with enhancements in late spring and autumn. In the SH midlatitudes (25–50° S), OH intensities are lower, though peaks are seen in May–June and November–December. In Fig. b, Ep exhibits largest values at the equator, with broad enhancements during March–April and again in September–October, consistent with semiannual variability. In the NH (25–50° N), Ep remains relatively weak year-round, with a faint enhancement in Boreal autumn. The SH (25–50° S) shows similarly weak Ep values, with no distinct seasonal maxima compared to the equatorial dominance. In Fig. c, the OH peak altitude displays a semiannual oscillation at the equator, reaching higher altitudes (∼83–85 km) in April–May and October–November, and lower altitudes (∼78–80 km) during solstitial months. In the NH (25–50° N), the peak altitude is elevated mainly in April–May, while in the SH (25–50° S), the main enhancement occurs during October–November. Overall, the equatorial region controls the dominant semiannual cycle, while the midlatitudes show weaker, seasonally shifted responses.

Figure shows the harmonic analysis which estimates the amplitude and phase characteristics of AO and SAO components. In the semiannual oscillation (SAO) panels (left column), OH relative amplitude (panel a) peaks at the equator (±15°) around 10–12 %, while in the SH midlatitudes (25–50° S) the amplitude reaches comparable values near 25° S but decreases further poleward. In the NH (25–50° N), amplitudes are weaker, staying below 8 %. The SAO phase indicates maxima occurring earlier in the year at higher latitudes and shifting toward later days near the equator. Ep relative amplitude (Fig. c) also peaks in the SH (25–50° S) with values near 4–5 %, while equatorial amplitudes are weaker (∼2–3 %), and NH values remain low (<2 %). The SAO altitude amplitude (Fig. e) is largest at the equator, approaching ∼1 km, and decreases steadily toward both hemispheres, with values around 0.6 km in the midlatitudes; phases are nearly flat across latitudes.

In the annual oscillation (AO) panels (right column), OH relative amplitude (Fig. b) shows larger signals in the SH (25–50° S), peaking above 15–20 %, while equatorial values are lower (∼5–7 %), and the NH midlatitudes (25–50° N) show amplitudes around 10 %. Phases indicate hemispheric asymmetry, with maxima occurring later in the year in the NH compared to the SH. Ep relative amplitude (Fig. d) exhibits alternating peaks across latitude, with larger amplitudes in the SH (∼3–4 %), weaker equatorial values (<2 %), and NH amplitudes (∼2 %). Finally, altitude amplitude (Fig. f) shows equatorial suppression (∼0.2–0.3 km) compared to enhancements in the midlatitudes: ∼0.6 km in the SH (25–50° S) and ∼0.4–0.5 km in the NH (25–50° N). AO phases are more variable, reflecting hemispheric differences and less coherent timing compared to the SAO.

The decomposition analysis separates temperature-driven effects from residual non-thermal processes in OH variability. The methodology integrates SABER satellite observations with geophysical indices representing atmospheric drivers (F10.7, Kp, MEI, QBO) using a two-stage MLR framework. After aggregating data into 10° latitude bins and removing seasonal periodicities, the deseasonalized anomalies are modeled using: 9Y(t)=C0+C1⋅t+∑i=2NCi⋅Pi(t)+ϵ(t) where Y(t) represents OH or temperature, C1 is the linear trend coefficient, Pi(t) are the geophysical proxies, and ϵ(t) is residual variability. The decomposition leverages the temperature dependence of OH chemistry. The temperature-driven component of OH trends is calculated as: 10TrendOH,Temp=OH‾⋅β⋅TrendTemp where β is the temperature sensitivity coefficient, the residual trend is then isolated as: 11TrendOH,Res=TrendOH,Obs-TrendOH,Temp

This decomposition is applied to both trends and responses to each proxy, providing quantitative separation of thermal versus non-thermal mechanisms. Figure presents the decomposition of OH trends and the correlation between OH and temperature trends. In Fig. a, the observed OH trend is negative across all latitudes. At the equator (±15°), OH trends are close to zero, showing weak decreases. In the NH (25–50° N), the trend is negative, ranging between -1 and -3×10-10 µmyr-1, while in the SH (25–50° S) the decline reaches values near -4 to -5×10-10 µmyr-1. Figure b shows that temperature trends at the OH peak altitude are negative across latitudes, with the cooling (∼-0.15 Kyr-1) near the equator, weakening toward the midlatitudes in both hemispheres, where values range around -0.07 to -0.1 Kyr-1. Figure c shows the decomposition of the OH trend: the temperature-driven component (green) accounts for most of the negative signal, especially at the equator (±15°) and in the SH (25–50° S). A positive residual component (purple) emerges across latitudes, partly offsetting the cooling effect. This residual is larger in the midlatitudes of both hemispheres, suggesting additional drivers beyond temperature sustain OH emission levels despite the cooling background.

The correlation coefficient between the OH and temperature trends (Fig. d) ranges from -0.25 to 0.5. Peak values of 0.4–0.5 occur in several latitude bands, consistent with the temperature dependence of the H + O₃ reaction rate. The correlation reaches a minimum of -0.25 near 25° N, indicating an anti-correlation in the Northern Hemisphere subtropics. This sign reversal points to non-thermal processes, such as changes in the Brewer–Dobson circulation or hemispheric asymmetries in wave-driven transport, that oppose the thermochemical response. Away from this minimum, the correlation returns to positive values but remains below 0.5 at all latitudes, indicating that non-thermal contributions persist across the globe. The correlation analysis in Fig. d thus complements the decomposition in Fig. c, confirming that thermal effects account for only part of the OH trend and that non-thermal processes dominate in the Northern Hemisphere subtropics.