Numerical weather prediction models inherently suffer from systematic forecast biases that degrade ensemble forecast performance, necessitating effective correction methods. In this study, two online bias correction schemes, BC and BC-EOF, were developed using the raw biases computed from the 24-240-h forecast biases averaged over the previous 10 days' forecast and the leading empirical orthogonal function (EOF) of these biases, respectively. Both schemes were evaluated using the China Meteorological Administration Global Ensemble Prediction System (CMA-GEPS). Systematic biases in potential temperature in the CMA-GEPS demonstrate the most pronounced linear growth, with the leading EOF mode explaining more than 50% of the variance and the associated principal component exhibiting a near-linear trend. Therefore, two bias correction schemes were applied to the potential temperature. Using 21-member ensemble experiments over boreal winter and summer periods, the ensemble performance and mass conservation impacts of the two schemes were analyzed. Both BC and BC-EOF significantly reduce the temperature and geopotential height biases, improving the probability forecast skill, forecast accuracy, and spread-skill relationship. Additionally, BC outperforms BC-EOF in temperature forecasting, whereas BC-EOF excels in terms of geopotential height improvements. Moreover, the localized adjustments in BC may introduce spurious heat sources or sinks and exacerbate mass loss through thermodynamic interactions, whereas BC-EOF better preserves mass conservation. Overall, the findings demonstrate that the two schemes can improve the GEPS performance, with BC-EOF balancing the forecast skill improvement with physical consistency.

As an important approach to improving prediction accuracy,the post-process error correction of climate model products plays an indispensable role in global operational climate systems.To enhance the prediction precision of numerical climate prediction models,this study applies the Convolutional Neural Network(CNN)approach to conduct post-process correction on key operational prediction products of the third-generation climate operational prediction system of the China Meteorological Administration(CMA),i.e.,CMA-CPSv3.The targeted products include monthly 2 m air temperature,precipitation over China,and the El Niño-Southern Oscillation(ENSO)index during the period 2001-2023.Using reanalysis data from the National Centers for Environmental Prediction(NCEP)as the observational benchmark,a dedicated correction model has been developed through deep learning training of a multi-layer CNN architecture.After model construction,changes in the model performance before and after correction are evaluated during an independent test period.Results indicate that the CNN model significantly improves the prediction accuracy of climate model products.For temperature and precipitation predictions in China,the correlation coefficient of 1-7 months lead predictions is increased by 0.1-0.5.Among these improvements,the Root Mean Square Error(RMSE)of temperature is decreased by 0.5-1.0℃,representing a reduction rate of 20%—30%.For precipitation,the correlation coefficient is increased by 0.1-0.2(an increase of 10%—20%),and the RMSE is decreased by 0.1-1.0 mm/d(a reduction rate of 3%—30%),with the RMSE reduction rate reaching 30%—50%in Eastern and Southeastern China.For the ENSO index,the correlation skill for forecasts with a lead time of 1-7 months is enhanced by 5%—7%,and the RMSE at a lead time of 7 months is reduced by 50%,suggesting that the model effectively addresses the issue of excessive oscillation amplitude of the ENSO index in the original CMA-CPSv3 model.Furthermore,this study explicitly identifies a limitation of the CNN model,i.e.,excessive intensity smoothing,when applied to the correction of extreme climate events,and proposes multi-dimensional directions for future optimization.It thus provides a technical solution that integrates scientific rigor and practical applicability for operational post-processing of CMA's climate models.

The "spin-up" problem—where convection-permitting models require hours to develop realistic clouds from large-scale initial fields—critically limits short-term severe weather forecasting. Cloud analysis offers a potential solution by directly incorporating hydrome-teor information from remote sensing observations. In this study, we leverage multi-source remote sensing data, including three-dimensional mosaic radar reflectivity, hourly aver-aged FY-2G satellite black-body temperature (TBB), and FY-2G total cloud water products, within a stepwise cloud-analysis initialization scheme. The scheme is implemented in a convective-scale ensemble forecasting system (CMA-Meso, 3 km resolution) for a heavy rainfall event. For each ensemble member, three-dimensional hydrometeor increments are independently generated from these remote sensing retrievals and gradually introduced over the first ten time steps, ensuring smooth coordination with the model's dynam-ic-thermal framework. Results demonstrate that the remote sensing-driven cloud analysis substantially enhances ensemble system performance across multiple dimensions: (i) spin-up time is significant-ly reduced, with precipitation forecasts exhibiting reasonable structure from the initial forecast hour; (ii) deterministic forecast accuracy improves systematically, with reduced RMSE for geopotential height, temperature, and wind fields across all levels; (iii) proba-bilistic forecasting skill is enhanced, evidenced by improved CRPS and AROC for surface elements and precipitation thresholds; (iv) ensemble reliability is optimized, with spread better matching forecast errors. Mechanistic analysis reveals that these improvements stem from physically coordinated hydrometeor-latent heat initial perturbations and sub-sequent cloud-radiation feedbacks that continuously regulate thermal-dynamic structures. This study establishes that assimilating diverse remote sensing data via cloud analysis is an effective approach for addressing spin-up challenges in convective-scale ensemble prediction.

The “spin-up” problem, in which convection-permitting models require hours to develop realistic clouds from large-scale initial fields, critically limits short-term severe weather forecasting. Cloud analysis can serve as a feasible approach to directly assimilate hydrometeor information from remote sensing retrievals. In this study, we leverage multi-source remote sensing data, including three-dimensional mosaic radar reflectivity, hourly averaged FY-2G satellite brightness temperature (black-body temperature, TBB), and FY-2G total cloud water products, within a stepwise cloud analysis initialization scheme. The scheme is implemented in a convective-scale ensemble forecasting system (CMA-Meso, 3 km resolution) for a heavy rainfall event. For each ensemble member, three-dimensional hydrometeor increments are independently generated from these remote sensing retrievals and gradually introduced over the first ten time steps, ensuring smooth coordination with the model’s dynamic thermal framework. Quantitatively, the scheme reduces near-surface Continuous Rank Probability Score (CRPS) errors, improves the overall predictive skill by 2.6–7.9% (maximum at the 12 h spin-up period), and increases ensemble spread by 2–5.8%, mitigating under-dispersion. Probabilistic precipitation forecasts show uniform area under the relative operating characteristic curve (AROC) improvements across all thresholds, 1.16–5.77% for light rain, 3.03–8.97% for moderate rain, and 6.00–12.07% for heavy rain, with these maxima consistently occurring at the 12 h spin-up time. Although Brier scores are marginally larger, these AROC gains confirm the enhanced discrimination of convective rainfall. At 500 hPa, CRPS reductions of 7.1–15.6% emerge after 24 h (largest 15.6% for geopotential height at 24 h), zonal wind CRPS is reduced by 2.2% at 12 h, and ensemble spread increases by 3.1–7.0% for all three variables. These improvements, particularly the pronounced benefits during the initial 12 h, demonstrate that the remote sensing-driven cloud analysis effectively shortens spin-up. Mechanistically, the gains arise from physically coordinated hydrometeor-latent heat perturbations and subsequent cloud radiation feedback that continuously regulate thermal-dynamic structures. This study establishes that assimilating diverse remote sensing data via cloud analysis is an effective approach for overcoming spin-up challenges in convective-scale ensembles.

Initial perturbations play a crucial role in determining the performance of an ensemble prediction system (EPS). This study comprehensively compares three types of initial perturbations—generated by ensemble data assimilation (EDA), singular vectors (SV), and their hybrid EDA–SV—using the China Meteorological Administration (CMA) global forecast model. In addition to conventional ensemble verification metrics, diagnostic tools such as kinetic energy (KE) spectrum analysis and spatial filtering are employed. Compared with the SV-based perturbations currently used operationally in the CMA global EPS (CMA-GEPS), EDA-based perturbations exhibit more smaller-scale structures and higher global perturbation KE, particularly in the Tropics. Ensemble forecasting experiments reveal that the EDA method provides superior ensemble spread and perturbation KE in the Tropics during the early forecast period. However, the SV method performs better in the extratropics throughout the forecast period and in the Tropics during the mid-to-late forecast period. The EDA–SV approach improves the overall performance of CMA-GEPS, yielding better spread–error relationships and enhanced forecast skill compared with SV and EDA methods. Results from spatially filtered ensembles further show that EDA–SV combines the subsynoptic-scale and mesoscale advantages of EDA-based perturbations in the Tropics, with the large-scale and synoptic-scale strengths of SV-based perturbations across the globe. This synergy leads to superior performance across spatial scales and lead times in both tropical and extratropical regions. Consequently, the EDA–SV method is planned for implementation in the next upgrade of the CMA-GEPS.

Ensemble prediction systems (EPSs) provide valuable tools for predicting atmospheric blocking, a meteorological phenomenon capable of triggering extreme weather and climate events with substantial socioeconomic impacts. As an operational global medium-range EPS independently developed by the China Meteorological Administration (CMA), CMA-GEPS has not yet been systematically evaluated for its ability to forecast atmospheric blocking. To address this gap, this study adopts the blocking index proposed by Tibaldi and Molteni (1990) to assess the performance of CMA-GEPS in simulating Northern Hemisphere blocking. Using 500-hPa geopotential height ensemble forecasts from CMA-GEPS and ERA5 (fifth generation ECMWF atmospheric reanalysis) data during the winter of 2024/25, the predictive skill of blocking is evaluated in terms of occurrence frequency and probabilistic forecasts, by means of Hovmöller diagrams, relative operating characteristic (ROC) curves, and Brier scores. Results indicate that CMA-GEPS can reasonably reproduce the longitudinal variations in blocking frequency. However, it cannot consistently predict blocking with high probabilities, as discrepancies between high-probability forecast regions and ERA5-based observed blocking regions widen with increasing lead times. Further detailed evaluations are conducted for blocking emerging over the Ural Mountains and the Okhotsk Sea, given their significant impacts on specific weather and climate events in China. ROC curve results demonstrate that CMA-GEPS can discriminate blocking occurrences from non-occurrences for lead times up to 360 hours over the Ural Mountains, and only up to 216 hours over the Okhotsk Sea. Brier scores also confirm better probabilistic forecast skill over the Ural Mountains. Overall, this work provides useful guidance and increased confidence for forecasters in applying blocking predictions from CMA-GEPS.

Understanding elevation-dependent precipitation over the Eastern Pamir Plateau (EPP), a primary water source for adjacent arid regions, is essential for alpine water resource management. Using a dense rain-gauge network (2010–2019) and reanalysis data, this study reveals the spatial characteristics and physical mechanisms that drive warm-season elevation-dependent precipitation across the EPP. Regional precipitation follows an S-shaped vertical profile, where the maximum precipitation altitude (MPA) centers at 2400–2800 m with precipitation gradients of +10 and − 9 mm/100 m below and above the peak, respectively. A gradient reversal to +11 mm/100 m occurs above 3600 m, leading to enhanced precipitation at high elevations. Instead of being driven by a uniform orographic effect, this regional profile reflects a spatial combination where distinct sub-basins dominate specific elevation zones. Specifically, the initial precipitation increase below the MPA is driven by the combination of the Kashgar River Basin (KRB) and Yarkant River Basin (YRB); the KRB trumpet-shaped topography promotes low-to-mid-level moisture convergence, within which increasing convective available potential energy drives the precipitation increase. Concurrently, the mechanical lifting of easterly airflows along the windward slopes of the YRB sustains a continuous precipitation increase up to the elevation of 2800 m (+13 mm/100 m). Above 2800 m, regional precipitation decreases with elevation, driven by water vapor depletion across all basins. Ultimately, the high-elevation precipitation reversal above 3600 m is exclusively driven by the localized alpine system of the Tashkurgan River Basin, where 500 hPa southwesterly moisture transport, canyon dynamic convergence, and low lifting condensation level conditions combine to jointly favor high-elevation condensation. These findings reveal the non-linear orographic precipitation pattern and its associated mechanisms, providing a new perspective for understanding precipitation processes in arid alpine regions.

Comprehensively representing model uncertainties with a consideration of randomness and nonlinearity in physics parameterizations is a crucial issue in convection‐allowing ensemble prediction systems (CAEPSs). In this study, the nonlinear forcing singular vector (NFSV) for a nonlinear representation of model uncertainties and the stochastically perturbed parameterization tendencies (SPPT) scheme for a stochastic representation of model uncertainties, are evaluated and compared in the China Meteorological Administration (CMA)‐CAEPS with a horizontal resolution of 3 km. A conditional nonlinear–stochastic perturbation method is also used to consider both stochastic and nonlinear representations of model uncertainties. Three experiments were carried out over South China for a month (1–30 May 2020), one with an SPPT scheme, one with a NFSV scheme and another one with a nonlinear–stochastic perturbation using a combination of SPPT and NFSV schemes. The combination of SPPT and NFSV schemes is compared to the NFSV and SPPT scheme to investigate whether the conditional nonlinear–stochastic model perturbation method, which combines nonlinear and stochastic schemes, can better represent model uncertainty than the SPPT and NFSV approaches. The results show that the nonlinear perturbation NFSV scheme has a certain advantage over the SPPT scheme, and further combining the NFSV and SPPT schemes improves overall probabilistic forecasting skill and has an advantage over using only the NFSV or SPPT scheme, which may imply the positive impact of using the nonlinear model perturbation scheme in the CAEPSs, and that considering both a stochastic and a nonlinear representation of model uncertainties contributes to a more comprehensive representation of model error in CAEPSs. This discovery sheds light on the design and development of model perturbation strategies for future convective‐allowing ensemble prediction.

Multiscale ensemble members can more accurately characterize the dynamic background error covariance structures in hybrid three-dimensional variational (Hybrid-3DVAR) assimilations, contributing to improving the analysis accuracy of meso- and small-scale systems. On the basis of multiscale ensemble members from the China Meteorological Administration Global Ensemble Prediction System (CMA-GEPS) and Regional Ensemble Prediction System (CMA-REPS), this study analyses multiscale spread characteristics and integrates them to construct a dynamic background error covariance suitable for Hybrid-3DVAR analysis, aiming to capture the flow dependence of background error covariance in assimilation analysis. Furthermore, a multiscale Hybrid-3DVAR assimilation scheme is designed via the CMA mesoscale prediction system (CMA-MESO V6.0), and four groups of analysis-forecast experiments are conducted for the heavy rainfall events in North China, including 3DVAR, single-scale Hybrid-3DVAR, and sequential multiscale Hybrid-3DVAR. The key findings include the following: (1) The 12-h forecast ensemble members from the CMA-GEPS and CMA-REPS exhibit similar spread structures, both of which effectively capture the flow-dependent characteristics of weather system errors. However, global ensemble errors demonstrate smoother spatial distributions, whereas regional members show larger spread with finer error structures in areas of intense convective development. (2) Horizontal and vertical correlation coefficients calculated from global ensemble members exhibit smoother patterns and larger correlation scales, whereas those from regional ensemble members are more localized with smaller correlation scales and contain more spurious noise in long-distance correlations than global ensemble members. (3) The sequential multiscale Hybrid-3DVAR assimilation experiment exhibits the smallest analysis and forecast biases and root mean squared error (RMSE) in the wind, humidity, and temperature fields. It also shows higher Equitable Threat Scores (ETS) for rainfall forecasts across all intensity levels, along with a notable reduction in false alarms for rainstorms with greater magnitudes.

To develop approximate initial perturbation schemes for a convection-permitting ensemble prediction system (CPEPS), we studied the characteristics of multiscale singular vector (SV) downscaling perturbations and observed perturbations and revealed the advantages of combined perturbations over single perturbations. The results indicate that multiscale SV downscaling perturbations exhibit flow-dependent features with larger magnitudes across various wavelengths. In contrast, the observed perturbations are distributed throughout the entire domain, with circular patterns around the sounding observation stations and smaller magnitudes. The combination of multiscale SV perturbations and observed perturbations, referred to as the combined perturbations, effectively integrates the characteristics of both, thereby producing the largest energy spectra and initial perturbation patterns with broader spatial distributions. Baroclinic instability and moist convection are the main mechanisms driving the development of these perturbations. Multiscale SV downscaling perturbations show the highest perturbation energy at both initial and forecast times, as well as the best spread–skill relationships and probabilistic prediction ability for all variables. Although the observed perturbations have overall lower prediction skills, they still positively impact certain precipitation cases in southern China, highlighting their essential role. Compared with the best-performing multiscale SV perturbations, the combined perturbations further enhance the perturbation energy, probabilistic prediction scores, as well as spread–skill relationships for all the variables over the first 24 h. Overall, this study demonstrates the distinct impacts of various initial perturbation methods in CPEPS and provides insights for developing improved initial perturbation schemes.

The ongoing rise in greenhouse gas emissions is leading to a sharp increase in global surface temperatures and more frequent extreme weather events, which has intensified the fluctuation range of daily extreme temperatures and increased the difficulty of prediction. Research on forecasting changes in daily extreme temperature can provide reliable scientific data for assessing future disaster risks and support decision-making. Due to limitations in the performance and sensitivity, current global climate models (GCM) exhibit considerable uncertainty in predicting extreme temperatures, increasing the difficulty of predicting future trends. It is necessary to correct the direct prediction results of GCMs to obtain more reliable prediction results. Therefore, Siberian sea level pressure and the sea surface temperature of the Indian Ocean, both of which have significant impacts on the daily extreme temperature changes in China, are selected as physical factors for correction. Two methods, emergent constraints and Pareto optimal ensemble, are employed to correct GCM' predictions of daily extreme temperature changes in China under the SSP1-2.6 scenario for the middle of the 21st century. A comparison of results before and after correction reveals that both methods could effectively reduce the inter-model uncertainty of future daily extreme temperature changes. Among them, Pareto optimal ensemble scheme, which integrates three-variable factors-daily extreme temperature in China, Siberian sea level pressure, and the Indian Ocean sea surface temperature, proves most effective in minimizing inter-model uncertainty. The range of multi-model predictions of daily maximum (minimum) temperature changes in China for the mid-21st century, as corrected by the three-variable Pareto optimal ensemble scheme, is narrowed to 1.26 ℃ to 2.10 ℃ (1.12℃ to 2.06 ℃). The uncertainty range is reduced by approximately 36.8% (32.9%) compared to the uncorrected results. Moreover, the signal-to-noise ratio of the predicted daily extreme temperature changes increase in most areas of China, rising from below 1 without correction to above 1. At the same time, corrected results based on three-variable Pareto optimal ensemble scheme show significant regional differences, adjusting the magnitude of warming differentially over the Qinghai-Xizang Plateau, Northwest China, and Sichuan Basin. Overall, employing physical constraint derived from selected constraint factors to correct predictions of future daily extreme temperature changes in China is shown to be useful and feasible.

Industrial anomaly detection suffers from limited data, making cross-domain generalization particularly challenging. Generalist Anomaly Detection (GAD) aims to train a unified model on a source domain that can effectively detect anomalies in unseen target domains. In the initial semantic feature space, strong entanglement between anomalies and object categories or defect types hinders effective generalization across domains. Recent works address this issue by projecting features into a residual space; however, such methods primarily increase cross-domain overlap for normal features, while anomalous features remain specific to object categories, defect types and data domains, leading to poor alignment and generalization. To address this limitation, we propose Value-order Decomposition (VOD), a simple yet effective technique that bridges three types of generalization gaps across object categories, defect types (including real and synthetic defects), and data domains. VOD disentangles and suppresses object-category-, defect-type-, and domain-specific information, promoting alignment within normal and abnormal samples while preserving their separability, thereby enabling robust generalization across the three gaps. Leveraging the strong alignment between real and synthetic defects within the same object, we perform anomaly detection using only normal and synthetic-abnormal reference, and effectively generalize to unseen real defect types. Experiments on diverse industrial and medical benchmarks demonstrate that our method, using a simple cut-and-paste anomaly simulation strategy, achieves strong generalization across the three gaps.

Introduction As the "Water Tower of Central Asia," the Pamir Plateau is critical for regional water security. Research on its extreme precipitation is therefore vital for water resource assessment, hydrological modeling, and disaster risk planning.Methods Using daily observational data and the Peak Over Threshold method, we constructed extreme precipitation series across the region. Forty-six statistical functions were applied to select the optimal fit for return period analysis at each station, enabling the estimation of precipitation amounts across eight return periods.Results (1) Extreme precipitation exhibits non-synchronous variation with annual precipitation. While about two-thirds of stations show increasing daily extremes, 42.9% show opposite trends between the two. Spatially, extremes are lowest in the southwestern plateau and highest in the Fergana Basin. (2) The contribution of extreme precipitation to multiyear totals ranges from 24.0% to 40.0%, peaking at 88.13% in some years, and shows a significant negative correlation with annual precipitation. Stations above 3,400 m record higher amounts, suggesting a maximum precipitation belt above the previously recognized 2,000-3,500 m range. (3) Stations with annual precipitation below 150 mm exhibit the strongest extremes, where a single 50-100 year return period event can approach or exceed half of the local annual total. (4) The Wakeby and Gen.Pareto distributions show the widest applicability for the region, each optimal for 36% of stations. Extreme precipitation increases 2-3-fold as return periods extend from 2 to 100 years, reaching a maximum of 68.1 mm.Discussion These findings highlight significant flood risks and ecological vulnerability, particularly in arid areas with fragile ecosystems. The identified spatial patterns and the lack of a universally applicable distribution function underscore the complexity of extreme precipitation in the Pamir Plateau. This work provides a critical foundation for improved water resource management and climate adaptation strategies in Central Asia.

In the context of global climate change, scientific projections of future photovoltaic power potential (PVpot) under extreme climate conditions are crucial for providing quantitative support for designing climate-resilient energy strategies. In this study, we developed a projection model based on the extreme forecast index (EFI) method using data from the sixth version of the Model for Interdisciplinary Research on Climate (MIROC6). Applying this model, we investigated the variation in projected extreme climate conditions and associated PVpot during 2025–2060. Particular attention was paid to the critical dual-carbon transition window periods of 2030 and 2060. Benchmarked against the ERA5 reanalysis dataset, the MIROC6 historical dataset was found to be able to reasonably reproduce the climate for multiple variables, including 2 m temperature, 10 m wind speed (W10m) and surface-downwelling shortwave radiation (Rad). The spatial distribution and temporal evolution of EFI from 2025 to 2060 indicated the probability of extreme high-temperature events across most parts of China, with the probability showing a statistically significant increasing trend over the study period. The highest probability with EFI over 0.5 was found in Xizang and Yunnan Province, along with eastern Inner Mongolia and Northeast China. A significant declining trend in the probability of extreme high winds was projected for W10m across most of China (p < 0.05), except for parts of the west. Superimposed on this trend were marked interannual variations in the EFI's spatial distribution. Notably, the probability of extreme winds was projected for Southern China and parts of Northeast China in 2030, whereas only weak signals of extreme wind remained in limited areas by 2060. Rad and PVpot featured predominantly negative EFI values between −0.4 and −0.1 across most parts of China, indicating non-negligible probabilities of extreme low solar radiation and PVpot conditions during the critical window periods of the dual-carbon goals. Temporal evolution analysis revealed that although EFI Rad and EFI PVpot remain negative, they show a significant increasing trend over time in most regions, suggesting a gradually weakening risk of extreme low PVpot towards 2060. The robustness of our findings was confirmed by sensitivity analysis of the distributions of model climate and EFIs across the model historical periods.

Traditional ensemble forecasts are based on deterministic models, and their performance is directly affected by the forecast quality of the deterministic models they rely on. The China Meteorological Administration Global Forecast System (CMA-GFS) was upgraded from v3.3 to v4.0 in 2023, resulting in significant improvements in both its horizontal resolution and forecast performance. This upgrade also provides a crucial foundation for optimizing the CMA Global Ensemble Prediction System (CMA-GEPS) with the CMA-GFS as the forecast model. To investigate the impact of the forecast model update on CMA-GEPS forecasts in wintertime, the operational CMA-GEPS v1.3 is taken as the baseline in this paper, and two ensemble forecast experimental systems are constructed using CMA-GFS v3.3 and v4.0, respectively, with the only difference lying in the forecast model. Two groups of 31 d ensemble forecast comparative experiments for the winter season of 2024 are carried out, and analyses are implemented from three perspectives, i.e., perturbation characteristics, statistical scores, and a representative case study. Results show that after the forecast model is updated from CMA-GFS v3.3 to v4.0, the ability of ensemble perturbations from CMA-GEPS to capture the forecast errors is slightly improved, and the perturbations grow faster. Moreover, the ensemble spread (SPD) of most elements verified increases significantly. Except that the ensemble mean Root Mean Square Errors (RMSE) of upper-level temperature in the tropical region deteriorate, the RMSEs of other variables remain nearly unchanged or improved. Overall, the gap between SPD and the ensemble mean RMSE becomes narrow, suggesting an enhanced ensemble reliability. Apart from the upper-level temperature in the tropical region, both the Continuous Ranked Probability Score (CRPS) and Outlier primarily decline, indicating improved probabilistic forecast skills. Over China, the forecast skill of light rain is nearly unchanged, and forecasts of moderate rain and heavy rain are improved to some extent. In summary, the forecast model upgrade generally improves the wintertime forecast performance of CMA-GEPS, which is further confirmed through the analysis of a representative cooling weather process in the northeastern region of China in late November. However, the increased SPD introduced by the new forecast model exacerbates the over-dispersive problem of the geopotential height forecast from CMA-GEPS during the early forecast period. Therefore, even if the initial perturbation technology, model perturbation strategy, horizontal resolution, and ensemble size of the ensemble prediction system are kept unchanged, it is still necessary to optimize the perturbation parameters after updating the forecast model to achieve a comprehensive improvement of the CMA ensemble forecast skill. In the future, impacts of the forecast model upgrade on CMA-GEPS in other seasons will be further explored.

Madden–Julian oscillation (MJO) is one of the dominant sources of extended-range atmospheric predictability. However, many atmospheric numerical models fail to predict MJO owing to limitations in representing air–sea interaction processes. This study investigated the role of sea surface temperature (SST) forcing in improving MJO predictability using China Meteorological Administration uncoupled Global Ensemble Prediction System (CMA-GEPS), extending forecasts from 16 to 35 days for the first time by applying different SST forcing schemes. Three SST forcing schemes are examined: (1) CTL SST, which uses fixed SST forcing, (2) E-Folding SST, which adjusts initial analyses toward observed climatology, relying primarily on the historical observations, and (3) Two-Tiered SST, which combines the bias-corrected SST analysis and predictions from the CMA coupled prediction system. The results confirmed that each SST forcing scheme is reasonably designed, with the Two-Tiered scheme providing better SST forcing information and more realistic equatorial wave features. For MJO prediction, the Two-Tiered scheme exhibits superior skill, particularly with strong MJO initialization, extending predictability to 20.6 days, i.e., 4.7 days longer than the CTL run and 3.4 days longer than the E-Folding test. Furthermore, the Two-Tiered scheme outperforms CTL and E-Folding over the western Indian Ocean, but it underperforms over the Maritime Continent. It exhibits poorer eastward propagation characteristics and larger MJO errors due to larger SST forcing errors, which mainly from the analysis-related errors of the coupled model. Larger SST forcing errors increase the water vapor biases and amplify MJO prediction errors, indicating the need for optimized SST forcing. Additionally, the Two-Tiered scheme significantly enhances anomaly correlation coefficient (ACC) skills for 500-hPa geopotential height, with the improvement closely linked to the enhancement in MJO forecast skill. Overall, even the one-way SST forcing that includes multi-faceted information of the coupled model can provide guidance for extended-range prediction with atmospheric-only numerical models.

Extreme temperature events have significant impacts on human society and economic activities,yet the prediction still involves considerable uncertainties,making the use of ensemble forecasting methods crucial.The Pangu-Weather Global Ensemble Prediction System(PGW-GEPS)was developed by integrating the Pangu-Weather(PGW)with perturbed initial conditions of the China Meteorological Administration Global Ensemble Prediction System(CMA-GEPS).Using the 2022 extreme heat wave event in Zhejiang and the 2024 cold wave event in Inner Mongolia as two cases,the forecasting performances of PGW-GEPS and CMA-GEPS on these two extreme temperature events are evaluated and compared based on multiple assessment metrics.The results indicate that,for both the Zhejiang heat wave and Inner Mongolia cold wave event,PGW-GEPS exhibits forecast accuracy and uncertainty representation capabilities comparable to CMA-GEPS.Both systems effectively capture the increase in 2 m air temperature forecast uncertainty with longer lead times and its subsequent decrease as the forecast initialization approaches the observation period.However,for the Zhejiang heat wave,PGW-GEPS shows deficiencies in forecasting the shear line and exhibits larger forecast errors in the medium range.A comparative analysis of the kinetic energy spectra of these two events further reveals that PGW-GEPS exhibits an attenuation phenomenon below the sub-synoptic scale.In summary,the AI(Artificial Intelligence)-based PGW-GEPS demonstrates its forecasting capability for extreme temperature events.Its forecast accuracy for 3-10 d extreme temperature prediction is comparable to that of CMA-GEPS,while its computational speed is advantageous.However,PGW-GEPS still faces challenges in capturing rapidly evolving meso-micro scale weather systems,and further improvement in forecasting sub-synoptic systems is required.This study provides valuable insights into the application of artificial intelligence models in ensemble forecasting.

ABSTRACT Accurate representation of forecast uncertainty is essential for effective ensemble forecasting. To quantify the strengths and weaknesses of the China Meteorological Administration (CMA) convection‐permitting ensemble prediction system (CPEPS) in representing spread–skill relationships and to identify avenues for improvement, this study applied multidimensional diagnostic metrics—temporal evolution, spatial distribution, spread–skill of perturbations applied at different scales, and point‐to‐point distributions of the spread–skill relationships between the ensemble mean root mean square error (RMSE) and ensemble spread—to operational CMA‐CPEPS data for the period January–June 2025. The results revealed that the CMA‐CPEPS is under‐dispersive for most variables except 500‐hPa geopotential height. Temperature exhibits a poorer spread–skill relationship than that of wind variables, and near‐surface variables show a poorer relationship than that of variables at mid‐ and low‐tropospheric levels. Spread–skill relationships vary regionally, with larger spreads and higher spread–skill ratios over China north of 30° N but smaller spreads and under‐dispersion over China south of 30° N. However, the higher spread–skill ratio over northern China does not yield uniformly larger correlation coefficients between the RMSE and ensemble spread. The point‐to‐point distributions effectively explain the observed six‐month averaged spread–skill relationships. Moreover, contributions of perturbations at different scales on the relationships differ across variables and forecast lead times. Based on these diagnostics, we have identified potential areas for improvements to enhance the spread–skill relationships in the CMA‐CPEPS. The diagnostic methods can be extended to evaluations of other ensemble prediction systems, providing a foundation for continuously improving EPSs.

Abstract High-precision time-frequency traceability is the foundation for critical infrastructure in modern communication, energy, finance, and scientific research. However, traditional remote time-frequency transfer technologies face challenges with low automation levels, complex operational processes, and a lack of built-in security mechanisms for data exchange. This paper proposes and implements a new method based on digital calibration certificates (DCCs) for remote automated time-frequency calibration and synchronization. This method establishes a closed-loop control system, using DCCs as the core medium to achieve high-precision, automated remote time-frequency transfer. To validate this method, the remote experiment continuously lasted for over 100 days. This experiment successfully synchronized hydrogen and cesium atomic clocks at different locations with UTC(NIM) to a nanosecond level. The results demonstrate that this method achieves full automation of the time-frequency traceability calibration process while ensuring the integrity and authority of traceability data through the use of DCCs.

Effective representation of model uncertainty is crucial for improving the forecast skill of convection-permitting ensemble prediction system. The Stochastic Perturbed Parameterization Tendencies (SPPT) scheme is one of the primary approaches used to represent model uncertainty, and its effect is controlled by three parameters: Perturbation magnitude, temporal correlation scale, and horizontal perturbation scale. There have been few studies on the optimization of these three parameters for the operational 3 km CMA-REPS (Regional Ensemble Prediction System of China Meteorological Administration) v4.0. Based on CMA-REPS, this study selects 13 heavy rainfall cases in North China in 2024 to conduct SPPT parameter sensitivity experiments. The forecast skill for upper-air and surface variables, precipitation, and perturbation energy growth are analyzed. First, using a smaller magnitude (with a standard deviation of 0.35) and dropping attenuation of the perturbations within the boundary layer most effectively enhances the forecast skill of the variables. Second, a 3 h temporal scale is conducive to improving the forecast skill within the initial 12 h, whereas a 6 h time scale performs better after 24 h of integration. A horizontal scale of 500 km yields the best overall performance. Compared with a 1000 km scale, it improves the spread and consistency for most variables. Further reducing the scale to 200 km can improve light and moderate rain forecasts within the initial 12 h but leads to a decline in forecast skill after 18 h. Third, spatiotemporal scales significantly influence the perturbation energy growth. The 3 h temporal scale promotes the perturbation energy growth across scales within the initial 12 h, while the 6 h scale is more favorable for perturbation growth after 18 h. The 500 km horizontal scale is most beneficial for the development of difference kinetic energy and difference latent energy. Although the 200 km horizontal scale can initially enhance low-level perturbation energy and promote smaller-scale perturbation growth during convectively active periods, it results in the minimal development of larger- and meso-scale components, as well as perturbation in the middle and late periods of integration. In conclusion, a 0.35 standard deviation with unattenuated boundary layer perturbations, a 6 h temporal scale, and a 500 km horizontal scale are recommended.