Due to limitations in observational data and analytical methods, considerable uncertainties remain in studies of long-term regional variations in surface solar radiation (SSR) and their driving factors. In this study, we first examined the long-term trends of SSR over China and its subregions since the mid-twentieth century, based on a homogenized and reconstructed SSR dataset (SSRIH20CR). Building on this, we further investigated the potential driving factors influencing SSR changes over China. The findings indicate that while total cloud cover (TCC) has a notable correlation with the interannual variability of SSR, but its contribution to the long-term trend is relatively limited (accounting for only 4.1
Aridity over the Tibetan Plateau (TP) is jointly controlled by moisture supply and atmospheric evaporative demand, yet how their relative importance varies across regions, elevations, and future climate scenarios remains unclear. Using ERA5 reanalysis (1979–2023) and bias-corrected REMO simulations driven by three global climate models, we investigate historical and projected aridity index (AI) changes and their climatic drivers. During 1979–2023, wetting occurred across 60
Despite the widespread decline in near-surface wind speed (NWS), climate models still show large uncertainties in reproducing historical NWS trends and projecting future changes over the Tibetan Plateau (TP). In this study, historical NWS simulations from 22 Coupled Model Intercomparison Project Phase 6 (CMIP6) High-Resolution Model Intercomparison Project (HighResMIP) datasets, consisting of 11 paired higher-resolution (HR) and lower-resolution (LR) models, were evaluated against observations over the TP during 1979–2014. Observations show a pronounced decline in TP NWS, with an annual trend of − 0.159 m/s per decade, mainly driven by spring decreases (− 0.235 m/s per decade). Both HR and LR ensemble means reproduced the spatial distribution of NWS but substantially underestimated the declining trend, with annual trends of only − 0.024 and − 0.020 m/s per decade, respectively, indicating a widespread “stilling” bias in CMIP6 models. The relatively small difference between HR and LR simulations further suggests that increasing horizontal resolution alone provides limited improvement in reproducing historical NWS decline over the TP. Based on multiple statistical metrics, seven optimal simulations were selected to construct an optimal model ensemble mean. Future projections under Shared Socioeconomic Pathway (SSP) 5–8.5 indicate a continued NWS decline over the TP during 2015–2049, with an annual trend of approximately − 0.02 m/s per decade, accompanied by continued regional warming. Random forest analysis further suggests that the underestimation of historical NWS decline is likely related to model biases in key dynamical and thermodynamical processes, particularly snow variability and surface sensible heat flux. These findings highlight the robustness of future TP NWS decline while emphasizing that improving physical process representation may be more important than simply increasing spatial resolution for reducing persistent model biases.
The diabatic heating over the Southeast Asian low-latitude highlands (SEALLH) is an important precursor to the Meiyu onset date (MOD), yet its predictive utility remains unclear. This study evaluates the SEALLH diabatic heating anomalies as predictors for hindcasting the MOD via the interannual increment approach. Five categories of the machine learning (ML) models are trained with the CMIP6 historical simulations and validated with the ERA5 data. The hindcasts are made with two strategies: (1) directly hindcasting the MOD interannual increments (Group 1), and (2) interannual increments derived from two consecutive MOD hindcasts (Group 2). In both groups, the top-performing models better capture the physical linkages among large-scale circulation, moisture convergence, and the MOD than the bottom-performing models. However, the top-performing models in Group 1 exhibit a major physical inconsistency in the 500-hPa zonal wind anomalies. This inconsistency is largely corrected in Group 2. Consequently, top-performing models in Group 2 achieve superior hindcast skill, with a temporal correlation coefficient (TCC) of 0.62 compared with 0.52 in Group 1. The Extra Trees model exhibits robust performance for both current-year and one-year-ahead hindcasts. These findings demonstrate the effectiveness of SEALLH diabatic heating for the MOD hindcast and highlight the critical role of physical consistency in the ML-based climate prediction.
This study characterizes spatiotemporal patterns of climate extremes in South America (SA) and assesses oceanic mode influences using daily MERGE and SAMeT data (2000–2024). Eight ETCCDI indices reveal a north–central precipitation gradient linked to the ITCZ, with dry spell maxima over central-eastern Brazil. Notable exceptions include southern Chile (orographic rainfall) and northeastern Argentina/western southern Brazil, where the SALLJ influences intense multi-day precipitation extremes. Temperature extremes show widespread warming with increasing tropical nights northward and higher frost frequency southward. Observed changes during 2000–2024 suggest drying tendencies in central-eastern SA and wetting in the northwest, though these may reflect interdecadal variability rather than long-term trends. ENSO emerges as the dominant statistical modulator, with composites and lag-correlations revealing a robust north–south dipole with regionally asymmetric signals: La Niña produces stronger drying over southeastern SA than El Niño produces wetting, consistent with Rossby wave propagation and SALLJ-related modulation, as documented in previous studies. Tropical Atlantic Variability (TAV) exerts its most robust influence over eastern Amazonia via ITCZ displacement, with weak signals elsewhere. The Combined Index produces weaker signals than ENSO alone, consistent with ENSO as the clearly dominant statistical driver of variability. These results underscore the need for mode-aware, region-specific climate services across SA.
This study evaluates the skill of NOAA’s Seamless System for Prediction and Earth System Research (SPEAR) and the NCEP Climate Forecast System version 2 (CFSv2) in forecasting spring and summer precipitation anomalies over the contiguous United States (CONUS) during 1991–2022. Using anomaly correlation coefficient, percentage bias, bootstrap significance testing, and quantile-based categorical metrics, we assess seasonal-mean precipitation and wet extremes of seasonal-mean precipitation, defined as exceedances of the 80th and 90th percentile thresholds. Coupled singular value decomposition (SVD) is further used to diagnose tropical Pacific sea surface temperature (SST)–CONUS precipitation covariance and the role of ENSO-related teleconnections. Both systems exhibit modest and regionally coherent skill during spring, with SPEAR showing statistically significant SPEAR advantages over CFSv2 occurring only in region- and lead-time dependent parts of CONUS, rather than consistently across the domain, while summer skill is substantially weaker in both models, especially for upper-tail seasonal precipitation anomalies. SVD results show that both models reproduce an ENSO-like leading coupled mode, but ensemble-mean SVD produces larger squared covariance fractions than observations because ensemble averaging filters internally generated variability and emphasizes the forced SST-related signal; individual-member SVD results are closer to observations. Associated 500-hPa geopotential height and vertical-motion diagnostics indicate that both systems capture some aspects of the spring Pacific–North American teleconnection pathway, but the circulation response is weaker, displaced, or less coherent in summer. Overall, SPEAR and CFSv2 share common limitations in warm-season precipitation prediction, suggesting that improved representation of SST-forced teleconnections, regional atmospheric variability, and additional climate drivers beyond ENSO is needed to improve seasonal rainfall forecasts.
Compound heatwave drought events (CDHWs) have intensified in recent decades, yet their historical spatiotemporal evolution and the role of large-scale climate drivers remain insufficiently understood. This study constructs a globally consistent CDHWs dataset (1981–2024) and systematically analyzes five El Niño-Southern Oscillation (ENSO)-sensitive hotspots: South Asia, Southeast Asia, tropical South America, Australia, and southern Africa. We quantify CDHWs frequency, duration, and severity, revealing both global intensification and marked regional heterogeneity. By applying decadal sliding-window correlations alongside 0–12 month lagged correlation diagnostics, we examine the temporal evolution, interannual variability, and lag structure of ENSO-CDHWs linkages. Results show globally asymmetric ENSO modulation, with El Niño generally amplifying CDHWs, but regional responses vary substantially. ENSO-CDHWs associations varied over time and across regions, with clearer short-lag correlations in Southeast Asia and tropical South America, peaking at lags of 1–3 months. These findings provide novel insights into the evolving spatiotemporal dynamics of compound extremes and their climate-mode modulation, offering a foundation for ENSO-informed early-warning systems and regional adaptation planning.
The Niño-3.4 index is a primary predictor of tropical cyclone (TC) rapid intensification (RI) in seasonal prediction, but fails to fully capture the evolution characteristics of El Niño–Southern Oscillation (ENSO) events. To address this limitation, a new metric—the ENSO decaying rate—is introduced in this study, which quantifies ENSO decaying flavors. When combined with the conventional Niño-3.4 index, the ENSO decaying rate improves the predictability of boreal autumn TC RI mean occurrence position on interannual timescales. This improvement arises because the ENSO decaying rate better captures the evolution of large-scale oceanic and atmospheric conditions that subsequently influence RI occurrence in boreal autumn. Furthermore, during the short-decaying El Niño transition years, the TC heat potential favors RI of autumn TCs in the South China Sea and the Philippine Sea. The ENSO decaying rate, rather than the conventional Niño-3.4 index, captures the TC heat potential conditions. Our findings demonstrate a strong association between the ENSO decaying rate and subsequent RI mean occurrence longitude in autumn on interannual timescales. This offers a potential pathway for improving the RI seasonal prediction skill.
Accurate prediction of the Bay of Bengal (BoB) summer monsoon (BoBSM) onset is critical for agriculture planning and water resource management across Asia, yet the physical sources of its subseasonal predictability remain poorly understood. Using ECMWF subseasonal-to-seasonal (S2S) reforecasts, we show that the BoBSM onset can be skillfully predicted up to 15 days in advance, and further investigate the physical mechanisms underlying this forecast skill. Results indicate that the primary source of predictability originates from sea surface temperature anomalies (SSTAs) in the tropical Indo-Pacific and North Atlantic Oceans, which jointly modulate the pre-monsoon mid-to-upper-tropospheric meridional temperature gradient (MTG) over the BoB. Tropical Indo-Pacific SSTAs provide the most robust source of predictability. During El Niño decay, El Niño-related teleconnections induce basin-wide Indian Ocean warming, enhance tropospheric diabatic heating, and maintain a favorable thermal structure that modulates the MTG before monsoon onset. Moreover, North Atlantic SSTAs contribute additional predictability through two distinct atmospheric teleconnection pathways. Tropical North Atlantic warming excites equatorial Kelvin-wave responses that are skillfully reproduced by the forecast system, whereas the North Atlantic tripole pattern forces a mid- to high-latitude Rossby wave train extending into Eurasia. Predictability associated with the Rossby-wave pathway declines rapidly beyond two pentads because of reduced forecast skill in representing the North Atlantic tripole pattern, making errors in North Atlantic teleconnections a major source of uncertainty in extended-range forecasts of BoBSM onset. These results highlight the dominant role of three-ocean SST forcing and identify extratropical teleconnection errors as a key limitation of current S2S prediction systems.
Utilizing ERA5 reanalysis datasets, this study investigates how the ozone valley over the Tibetan Plateau (OVTP) influences rainfall in East Asian summertime and elucidates the mechanisms involved. By defining an OVTP intensity index (OVTPI) with the area-averaged ozone content over the TP in May and June, the lead regression analysis was conducted for East Asian precipitation during June and July. It is shown that enhanced OVTP is primarily followed by an increase in precipitation in the southeastern coastal China during July, with a local anomalous convergence center for the vertically accumulated water vapor transport. Analysis of the general circulation patterns associated with the enhanced OVTP indicate that the South Asian High (SAH) extends eastward during June-July, and an anomalous high-pressure system develops above the Korean Peninsula. Located to the south of this anticyclone, the anomalous easterlies provide favorable moisture conditions for increased precipitation in neighboring regions. Further, a positive land-atmosphere feedback process is established in June over the TP, where the enhanced ozone valley can permit increased solar radiation to reach the ground. Consequently, the SAH extends eastward in July as a wave like train forms with the TP warming. The anomalous low-pressure system is excited downstream of the Tibetan Plateau, and the enhanced uplift explains an enhancement of precipitation in Southeast China coast from June to July. High-skill chemistry-climate models also validate the linkage between the OVTP and East Asian rainfall. This study reveals that the ozone over the TP can induce changes in atmosphere thermodynamics and dynamics locally and downstream, which might be a useful predictor for seasonal rainfall forecast.
Vietnam’s Mekong Delta plays a pivotal role in national and global food security. However, agriculture-based livelihoods in the delta are highly susceptible to climate extremes, highlighting the need for more localized climate change information. Hence, this study applied univariate and multivariate bias correction methods to raw climate model simulations from the Coupled Model Intercomparison Project Phase 6 under eight Shared Socioeconomic Pathways (SSP) to develop localized climate projections. The univariate method employed is the Quantile Delta Mapping (QDM) algorithm, whereas the multivariate approaches include the N-dimensional probability density function transform (MBCn) and the Rank Resampling for Distributions and Dependences (R2D2). The performance of these bias correction techniques was evaluated based on their ability to reduce model biases in marginal distributions and to reproduce multidimensional characteristics, including inter-variable, inter-site, and temporal dependence structures. Successive cross-validation results indicate that the bias-corrected fields exhibit substantial improvements in marginal distributions, even at high quantile values. Moreover, multivariate methods demonstrate superior performance in capturing multivariate characteristics compared to their univariate counterpart. For future projections, this study combined the trend-preserving feature of the QDM algorithm with sliding window settings to better preserve the underlying climate change signals from raw climate model outputs. At the annual scale, projections for the near to far future consistently indicate a warmer and wetter response relative to the baseline period across most SSP-based scenarios. However, rainfall during the dry season (months) is projected to decrease with considerable uncertainty. The findings provide valuable insights for climate-sensitive planning and decision-making in the delta. Multivariate bias correction schemes provide substantial improvements in the representation of multidimensional climate characteristics. Localized climate change scenarios, tailored to the needs of end-users, were developed for Vietnam’s Mekong Delta. Vietnam’s Mekong Delta is projected to experience a warmer and wetter climate overall, although dry-season rainfall is likely to decrease compared to the baseline period.
Moist heatwaves, defined by the concurrence of high maximum air temperature and relative humidity, pose increasing risks to human health in tropical regions, yet their representation in global climate models remains limited by coarse spatial resolution and systematic biases. In this study, a hybrid statistical downscaling method (Bias Correction Constructed Analogues with Quantile mapping, BCCAQ) is applied to 17 CMIP6 models to generate 0.1° gridded daily maximum temperature and relative humidity over the southern Greater Mekong Subregion. The downscaled outputs are evaluated against observations for 1981−2014 and used to assess future changes for the near future (2027−2066) and far future (2067−2100) under SSP1−2.6 and SSP5−8.5. The downscaling substantially improves the representation of mean and seasonal climate conditions (added value up to ≈ 0.98), although biases persist toward the upper quantiles. Future projections indicate a robust and spatially coherent warming-drying signal, with maximum temperature rising by up to 3.49 °C by the late century under SSP5−8.5, which together amplifies moist heatwave characteristics relative to the historical baseline. This intensification follows a clear hierarchy: frequency increases only modestly, duration rises more substantially, and temperature-based accumulated intensity shows the strongest and most rapid amplification, with far-future trends reaching 33.33 °C-days per decade under SSP5−8.5 and model agreement of ≥ 94
In this study, we investigate contrasting behaviors in the structure, intensity, and propagation of equatorial Kelvin (EK), equatorial Rossby (ER), and mixed Rossby-gravity (MRG) waves over the tropical Pacific between boreal winter and summer. The EK and MRG waves propagate faster in boreal summer than in winter and exhibit stronger convective amplitude during summer. The ER wave shows little seasonal difference in phase speed ( 5 m s−1) but has markedly stronger convection in winter. Horizontally, the EK and MRG waves exhibit only slight seasonal structural differences. The characteristic “swallowtail” convective envelope of ER waves arises from the bottom-up evolution of convective vertical structure along the wave’s propagation direction, and its seasonal asymmetry about the equator is driven by the northward migration of the ITCZ in boreal summer. Vertically, the EK wave exhibits a more pronounced “boomerang” structure in boreal summer due to stronger and deeper convective heating. The ER wave undergoes a fundamental transition from equivalent barotropic structure in boreal winter to first baroclinic mode in summer, governed by the weakening of background westerly vertical shear. The MRG wave similarly transitions from a near-barotropic structure in boreal winter to a boomerang-like first baroclinic structure in summer, attributable to the combined effects of weakened background westerly vertical shear and stronger convective heating. These findings demonstrate that equatorial wave characteristics are systematically modulated by the seasonally evolving background state, providing observational benchmarks for evaluating climate model simulations.
Subseasonal to seasonal (S2S) scale prediction, especially precipitation prediction, depends predominantly on initial conditions. To examine the impacts of atmospheric and land initial conditions on the predictions of the Madden-Julian Oscillation (MJO) and related S2S precipitation, we conduct a novel study using coupled atmosphere-land simulations in the Energy Exascale Earth System Model (E3SM). Our findings indicate that reanalysis-based atmospheric and land initial conditions yield improved S2S precipitation simulations compared with those using a long-term spin-up equilibrium state. The impacts of initial conditions on precipitation simulation persist for approximately 40 and 50 days in the MJO and global regions, respectively, and are strongly associated with outgoing longwave radiation in the MJO region and surface latent heat flux at the global scale. Although atmospheric initial conditions exert a dominant influence on MJO simulation, improved land initial conditions provide an important secondary source of predictability by better representing land–atmosphere coupling over the Maritime Continent. More realistic surface moisture fluxes and surface temperature can modulate boundary-layer moistening and MJO-related convection, thereby contributing to improved MJO prediction. These findings have important implications for S2S precipitation prediction and provide crucial insights for the further development of Earth system models.
A common approach to studying the Madden–Julian Oscillation (MJO) relies on indices derived from the first two leading empirical orthogonal function (EOF) modes, under the implicit assumption that these modes do not change over time. However, recent changes in the MJO characteristics imply the possibility of changes in the MJO leading modes, which have not received much attention. Here, we find that the leading EOF modes of MJO convection have, in fact, undergone significant changes since 1979. The changes manifest as a weakening of MJO variability over the tropical south Indian Ocean and a westward shift of the variability center over the southwestern Pacific. The westward shift of MJO variability over the western Pacific is closely associated with changes in the background circulation and moisture, which are attributed to the La Niña-like sea surface temperature changes. Meanwhile, the weakening of MJO variability over the southern Indian Ocean is attributed to the combined reductions in EOF-related moisture variability and vertical velocity variability. Specifically, the decline in moisture variability stems from weakened horizontal advection via eddy–eddy interactions and weakened vertical advection of background moisture driven by the intraseasonal vertical motion anomalies; meanwhile, the reduced vertical velocity variability is attributed to the decreased efficiency of atmospheric moistening via vertical advection. These findings indicate that the information carried by the MJO indices that are based on fixed EOF modes, as well as the associated climate impacts with the leading EOF modes, varies over time, suggesting that caution is warranted when employing EOF-based MJO indices to study the long-term trends of MJO activity and its response to climate change.
Heatwaves (HWs) pose an increasing threat by significantly impacting agriculture, public health, and the economy. Traditional HW assessments based solely on maximum temperature (Tmax) overlook the combined influence of temperature and humidity on human thermal stress. This study employs the Heat Index (HI), which integrates Tmax and relative humidity (RH), to identify and classify HW and severe heatwave (SHW) events across India from 1981 to 2023, focusing on the pre-monsoon (March–May) and early monsoon (June–August) seasons. Using Steadman’s approach, we further distinguish dry heatwaves (DHWs) and moist heatwaves (MHWs) based on the Tmax–RH relationship. The analysis reveals prominent DHWs and MHWs hotspots across western, northwestern, coastal, central, and Indo-Gangetic Plain regions. BSISO-related convection and circulation anomalies modulate HW types, with dry phases (positive OLR) favoring DHWs and wet phases (negative OLR) promoting MHWs through enhanced convection and moisture transport. Trend dynamics indicate that Tmax predominantly drives HWs during March–June, while RH amplifies SHWs during July–August. On average, India experiences 9–12 HWs and 1–2 SHWs annually, with frequency peaking in June. HW durations reach 75–90 days in arid regions and exceed 25 days along eastern coastal regions, both showing increasing trends. Intensity values range from 55 to 64 °C, with rising trends over Rajasthan, Gujarat, Uttar Pradesh, and Bihar, and declines along the eastern coast. The findings underscore the urgent need for region-specific mitigation strategies, including heat action plans, early warning systems, and public health preparedness, to reduce the escalating impacts of HW and SHW events across India.
Extratropical cyclones play a critical role in midlatitude climate dynamics, but their accurate simulation remains a challenge in climate models. While enhanced resolution simulations have been widely used to study these systems in the Northern Hemisphere, their performance in the Southern Hemisphere, particularly for intense cyclones, remains less explored. This study evaluates the impact of increased horizontal resolution on the simulation of extratropical cyclones over a key hotspot in southeastern South America, comparing HighResMIP models (25–50 km) with standard CMIP6 models (100–250 km). Using an objective feature-tracking algorithm applied to T42 vorticity at 850 hPa, we analyze spatial and frequency distributions, structure, and life cycle of extratropical cyclones, comparing against ERA5 reanalysis. Results show that finer resolution models reduce biases in cyclone number and track density, significantly improving the detection of intense cyclones, whereas coarser resolution models underestimate their frequency by more than 50
The troposphere-stratosphere coupling is a fundamental driver of winter atmospheric variability, yet the relative importance of different tropospheric precursors remains a subject of intense debate. This study identifies a significant decadal strengthening in the dynamical link between the December Siberian High (SH) and the January stratospheric polar vortex (SPV) over the past few decades. Our analysis reveals that the SH has emerged as a robust and independent precursor to SPV weakening, characterized by a marked increase in upward-propagating planetary wave activity. Notably, we find that this intensified SH-SPV coupling is decoupled from Ural Blocking (UB) activity, which has traditionally been considered the primary mediator of troposphere-stratosphere interactions. This direct SH-to-SPV pathway provides critical predictive skill for the intraseasonal reversal of East Asian surface temperatures. Our findings highlight the necessity of monitoring SH variability—beyond traditional blocking indices—to improve sub-seasonal forecasts of stratospheric disturbances and associated cold extremes.
A previous study has reported similarity in the variability of tropical high cloud cover in the present climate and in climate change (interannual and centennial scales) in the tropics, evaluated based on percentage cloud cover, which is a standard output of the Coupled Model Intercomparison Project (CMIP). However, physically interpreting intermodel comparisons of simulated clouds using the standard output of cloud cover is not necessarily straightforward. The present study uses data from the Atmospheric Model Intercomparison Project (AMIP), AMIP future4K and AMIP p4K to evaluate the feedback of high cloud cover (a change of high cloud cover per unit surface warming) based on the standard output of cloud cover and cloud cover evaluated using a satellite simulator under a common definition in general circulation models based mainly on modeled cloud radiative properties. We reveal that the feedback of high cloud cover at the interannual scale is consistently negative, as in the previous study. However, the feedback of high cloud cover at the centennial scale is nearly zero, different from that at the interannual scale. The reason lies in the previous study’s definition of high cloud cover by the standard output of cloud cover to be consistent with the result of cloud cover evaluated using a satellite simulator at the interannual scale. However, this definition does not necessarily hold in different climate states, such as future climates. The reduction of high cloud cover evaluated with the standard output of cloud cover in the future climate states tends to be more remarkable compared to the changes of high cloud cover based on cloud cover evaluated using a satellite simulator. Furthermore, the relationship between feedback of high cloud cover and the feedback of outgoing longwave radiation is stronger when using cloud cover evaluated using a satellite simulator than when using the standard output of cloud cover. This study also reveals large uncertainties in the projection of high cloud cover over the tropics in a perturbed climate (increase of sea surface temperature), even when using cloud cover evaluated using a satellite simulator, because of an almost equal number of models showing increased and decreased high cloud covers due to warming. This study demonstrates the need to include satellite simulator data as an endorsed output of major MIPs for the in-depth analysis of clouds, such as to determine the major causes of modeled uncertainties in future changes of high cloud cover, based on CMIP data.
