With the rapid development of deep learning weather prediction (DLWP) models like GenCast, rigorous evaluation of their physical consistency is essential. This study investigates the dynamical fidelity of GenCast against ECMWF IFS-HRES and IFS-ENS using comprehensive kinetic energy (KE) and difference kinetic energy (DKE) spectra over 2021. Unlike the physically consistent error growth in IFS-ENS, GenCast exhibits weak planetary-scale growth and a persistent, flattened KE tail at high wavenumbers starting from the first forecast step. These mesoscale artifacts persist across multiple GenCast variants and AIFS-ENS, indicating a broader challenge for noise-conditioned generation. Helmholtz decomposition further reveals white-noise-like variance rather than balanced dynamics. Spatially, weak interactions between large-scale and mesoscale wind fields suggest a misrepresentation of topography-flow interactions. Furthermore, analyses of KE gradient (divided by del KE divided by) revealed that GenCast fails to reproduce the sharp, filamentary structures, instead generating broad, isotropic, and noisy patterns. These findings suggest that current noise injection mechanisms in DLWPs produce noisy artifacts mimicking variance without reproducing realistic error growth physics. Improving these mechanisms is vital for developing physically consistent DLWPs.

Soil moisture is a key driver of climate variability and extremes such as heatwaves and wildfires, and accurate prediction at seasonal timescales is therefore essential. Skillful forecasts at these scales, however, remain a significant challenge. While multi-model ensemble (MME) climate forecasts consistently outperform individual models for atmospheric variables, their soil moisture outputs require careful interpretation when combined across models because of differences in land surface model (LSM) structures and soil configurations. To bridge this gap, we developed a seasonal soil moisture prediction system using the Joint UK Land Environment Simulator driven by NCEP CFSv2 meteorological forecasts, and integrated monthly temperature and precipitation forecasts from the APCC MME—which exhibits superior seasonal prediction skill to single dynamical models—into the system’s meteorological forcing. Despite correcting only temperature and precipitation—the two variables consistently available across all participating MME models—the approach yielded substantial improvements in soil moisture forecast skill. Retrospective hindcast experiments for 1991–2016, initialized in February, show that the MME-corrected forecast (J-MME) substantially outperforms the CFSv2-driven baseline (J-CFSv2) in predicting boreal spring–summer soil moisture. The global average coefficient of determination ( R ^2 ) between forecast and reanalysis increased by 0.10–0.12 at three- to four-month lead times. The improvements were driven by precipitation correction in water-limited regions and temperature correction in energy-limited high-latitude regions. This approach also extended the effective lead time for statistically significant predictions and improved the detection of historical drought events, demonstrating that even limited integration of MME information into an LSM framework can yield meaningful gains in seasonal soil moisture and drought forecasting.

Abstract In this study, we developed a TabNet‐based machine learning model to predict tropical cyclone (TC) rapid intensification (RI) in the Western North Pacific. The most significant challenge in predicting RI is the severe class imbalance between rapid and non‐rapid intensification cases, typically 4.2:1 ratio based on 1977–2021 records. To overcome this, the synthetic minority oversampling technique (SMOTE) with ratios of 4:1, 3:1, 2:1, and 1:1, and three loss functions: cross entropy, balanced cross entropy, and focal loss were examined. The combination of balanced cross entropy and 2:1 SMOTE achieved the highest performance. Leveraging the TabNet's interpretability, sea surface temperature was identified as the most important factor, followed by temperature and specific humidity at 850 hPa. However, performance showed clear year‐to‐year variation. The year 2020 exhibited relatively poor performance likely due to the sparsity of RI cases rather than abnormal weather conditions, which highlights the important role of sample size in our model. In conclusion, TabNet combined with data augmentation and optimized loss functions significantly improves forecast performance of RI of TCs.

This study provides a systematic evaluation of historical simulations of monthly surface air temperature and precipitation from CMIP6 and CMIP5 models. By utilizing an error decomposition framework that separates the Mean Squared Error (MSE) into mean bias, variance, and correlation components, we quantify the specific sources of error in each variable. The results demonstrate that CMIP6 models exhibit an apparent improvement in temperature simulations, primarily driven by a reduction in mean bias, whereas precipitation shows marginal improvements. A long-term error analysis reveals that while absolute errors have decreased in recent decades, the normalized errors have increased due to the decrease in observational variability. This suggests that models may struggle to capture the reducing magnitude of natural variability. This study highlights that while representation of the mean state of surface temperature has improved, structural discrepancies in precipitation patterns remain a critical challenge for model development.

Accurate medium-range weather forecasts are essential for disaster preparedness and resource management. They are traditionally achieved by numerical weather prediction (NWP) models, but such models are costly to run operationally. Recent advances in machine-learning-based weather prediction (MLWP) offer a promising alternative to NWP for improving forecast efficiency. This study evaluates the prediction skills of four state-ofthe-art MLWP models-FengWu, FuXi, GraphCast, and PanguWeather-and their ensemble mean (MLWPEM), initialized with operational analysis from Korea Integrated Model (KIM) as well as ERA5 reanalysis. An emphasis is placed on forecasts of extreme events over East Asia-such as heavy rainfall, typhoons, heat waves, and cold spells in 2023-up to 10 days ahead. When initialized with operational KIM analysis, MLWP forecasts exhibit performance comparable to, or better than, operational KIM forecasts in predicting 500-hPa geopotential and 2-m air temperature. Even for rainfall forecasts, MLWP models-particularly GraphCast-outperform operational KIM forecasts at short lead times (1-3 days), but their performance decreases markedly for intense rainfall events (>= 30 mm/6 h). Forecast skills are further improved when considering MLWP-EM. Case studies show that KIM-initialized MLWP-EM forecasts outperform operational KIM forecasts in predicting temperature extremes, but underestimate rainfall amounts during heavy rainfall events and typhoons. Overall forecast skills increase when MLWP models are initialized with ERA5 reanalysis, highlighting the importance of initial conditions even in MLWP. These results demonstrate the operational potential of MLWP-EM forecasts initialized with operational analysis for medium-range forecast, while underscoring persistent challenges in accurately representing extreme precipitation.

Pakistan, with its long history of flooding and rapid urban expansion, has suffered severe consequences for both its population and economy, exemplified by the catastrophic flood of 2022. Understanding evolving flood patterns and their impacts is critical, particularly in the context of climate change. This study introduces the Flood-focused Extent of Water - Pakistan (FEWPakistan), a high-resolution, long-term flood dataset spanning 35 years (1988-2022). The dataset was developed using Landsat satellite imagery and a newly devised flood detection methodology that integrates the Normalized Difference Water Index and the Normalized Difference Snow Index. Further, Global Human Settlement Layer data has been integrated to quantify the human impacts of flooding over time. Our analysis reveals a clear trend of intensifying flood risks and damage over the past decade, with the proportion of affected urban populations annually rising from 10.5% to 13.4%. In contrast, rural areas experienced smaller increases in flood exposure, with peaks of 8.3%. Approximately 70% of the variability in flood exposure is attributable to the compound effects of urban population growth and expansion of inundated area. These findings underscore the escalating challenges posed by flooding in rapidly urbanizing regions and the urgent need for targeted adaptation and mitigation strategies.

Since the late 20th century, an emerging atmospheric teleconnection pattern, the trans-Eurasian heatwave-drought train, has intensified remarkably during summer, correlating with a surge in concurrent heatwave-drought events from Eastern Europe to East Asia. Tree-ring proxies, spanning three centuries, reveal that the recent intensity of this pattern is unprecedented in the historical records. In contrast, the circumglobal teleconnection, which historically dominated the continental-scale Eurasian heatwave occurrences, has shown no discernible trend amid global warming. Consequently, this emerging pattern signifies a radical shift in Eurasian heatwave-drought climatologies. The mechanism involves Rossby wave propagation linked to warming sea surface temperatures in the Northwestern Atlantic and enhanced Sahel precipitation, both amplified recently by overlapping effects of anthropogenic warming and natural variability. Land-atmosphere interactions driven by soil moisture deficits further intensified the pattern regionally. Climate models predict that anthropogenic forcings will continue to strengthen the pattern throughout this century.

Quantifying the timing of hydroclimatic changes due to global warming is crucial for water resources management. This study investigates the hydroclimatic changes based on the Time of Emergence (TOE) analysis in the western United States, with a focus on the impacts of climate change and aridification. Utilizing the Community Earth System Model Version 2 Large Ensemble (CESM2-LE), we investigate projected changes in temperature, precipitation, evapotranspiration, runoff, snow, and Total Water Storage (TWS). Despite precipitation increases in the future, our results project a robust decrease in TWS by the end of the 21st century, mainly driven by higher evapotranspiration and reduced snow. Total Water Storage, which is an integrated measure of all terrestrial hydrological processes, exhibits a more pronounced climate change signal compared to precipitation. Significant regional variations also emerge in the TOE of TWS, with states around the interior and high-elevation regions experiencing changes faster, as early as in the 2030 s, than those on the Pacific Coast or in southern regions. The study highlights the Upper Colorado River Basin as an emerging aridification hotspot, emphasizing the need for targeted research and adaptive water resource management strategies. This research underscores the importance of jointly considering TWS and the aridity index to comprehensively assess hydrologic regime shifts under global warming. Plain Language Summary: Climate change is leading to shifts in temperature, rainfall, evapotranspiration, river flow, snow, and other factors. These changes collectively affect what we call the 'hydroclimate,' which includes changes in groundwater and aridity. Our study finds that as climate change progresses, there will be a noticeable decrease in groundwater, particularly in the interior states. Additionally, we observe an expansion of dry areas, especially in the Upper Colorado River Basin due to reduced snow. This information is crucial for policymakers who need to plan and prepare for future water resource management in the face of a warming climate.

The East Asian Jet (EAJ), a key upper-level westerly wind that controls monsoon circulation and precipitation patterns within the East Asian summer monsoon system, has undergone multiple regime shifts since the mid-20th century. While transitions in the late-1970s and mid-1990s are well documented, a more recent shift in the early 2000s remains less well characterized. Here we identify a regime shift of the EAJ occurring in 2002/2003 through striking changes, manifested by an abrupt intensification in both the mean state and interannual variance of the jet. This shift coincides with significant surface warming over southern China and widespread drying across East Asia. Our analysis identifies two key features characterizing this transition: strengthening of the Tibetan Plateau High (TPH) and emergence of enhanced SM–T coupling variability over Inner East Asia. After 2003, coupling strength exhibits significant correlation with surface energy partitioning and lower-tropospheric thermal structure. This coupling variability is closely linked to jet fluctuations, indicating that enhanced jet variability is strongly mediated by land–atmosphere (L–A) coupling processes. In addition, the intensified TPH additionally contributes to jet intensification through modifications in upper-tropospheric circulation patterns. These results highlight the critical role of L–A interactions in EAJ variability and provide insights into potential linkages between drought conditions under global warming and monsoon circulation changes mediated through upper-level atmospheric flow.

Alaska is experiencing simultaneous trends of increased winter wetness and heightened summer fire risk due to global warming, leading to more frequent wildfires and greater unpredictability in fire behavior in recent decades. Large-ensemble simulations show that warming drives distinct seasonal changes: in winter, an intensified ridge over the western U.S. enhances moisture transport to Alaska, increasing precipitation while promoting vegetation growth near the Alaska Range. In summer, rising temperatures intensify the fire weather index signaling greater wildfire potential and increase lightning activity. Although the links among these complex seasonal changes remain difficult to validate, temporal overlap—enhanced vegetation growth followed by more fire-conducive weather, and associated increase in lightning could collectively heighten wildfire risk. The robustness of our large-ensemble simulations provides compelling evidence for these cascading effects. Extreme lightning-driven events, such as the Swan Lake Fire, represent the emerging pattern in Alaska’s evolving fire regime. The concurrent rise in winter wetness and summer fire conditions underscore the urgent need for adaptive fire management strategies that address these interconnected climate drivers.

Recent studies have shown that the observed global warming trend over recent decades provides efficient constraints not only for future global mean temperature increases (ΔTgm) across Earth system models but also for changes in several climate variables that include significant ΔTgm-related uncertainty. However, ΔTgm-related emergent constraints (ECs) cannot reduce the uncertainty unrelated to ΔTgm. Here, to overcome this limitation, we develop an EC method and apply it to future changes in the annual maximum daily precipitation in order to reduce uncertainty therein. An EC for precipitation sensitivity based on historical extreme precipitation biases is combined with the constrained ΔTgm. This combined EC decreases the variance of the global mean precipitation by 42%, an improvement from only using temperature (resulting in 26% reduction), and the variance of regional precipitation by ≥ 30% in 24% of the globe (whereas ≥ 30% reduction is only seen in 2% of the globe with the temperature-related EC).

The Indo-Gangetic Plain has experienced a substantial rise in relative humidity in recent decades, with implications for human health and well-being. Here we use atmospheric reanalysis and large-ensemble climate model simulations to assess changes since the 1960s. Relative humidity increased by 10.3 +/- 0.3 percent, mainly due to a 2.9 +/- 0.1 grams per kilogram rise in specific humidity and a slight decrease in air temperature (-0.2 +/- 0.1 degrees Celsius). Aerosol-induced surface cooling played a crucial role in enabling this moistening. Decomposition analysis reveals that specific humidity accounts for 95% of the increase, with cooling explaining the rest. Future projections show contrasting trends. High-emission scenarios peak and then decline after the 2040s, as greenhouse gas warming overtakes weakening aerosol effects. In contrast, low-emission scenarios maintain stable or slightly increasing humidity. These findings reveal how aerosols and greenhouse gases exert opposing influences on humidity and underscore the need for coordinated climate strategies in this vulnerable region.

Unraveling drivers of the interannual variability of tropical land carbon cycle is critical for understanding land carbon-climate feedbacks. Here we utilize two generations of factorial model experiments to show that interannual variability of tropical land carbon uptake under both present and future climate is consistently dominated by terrestrial water availability variations in Earth system models. The magnitude of this interannual sensitivity of tropical land carbon uptake to water availability variations under future climate shows a large spread across the latest 16 models (2.3 ± 1.5 PgC/yr/Tt H2O), which is constrained to 1.3 ± 0.8 PgC/yr/Tt H2O using observations and the emergent constraint methodology. However, the long-term tropical land carbon-climate feedback uncertainties in the latest models can no longer be directly constrained by interannual variability compared with previous models, given that additional important processes are not well reflected in interannual variability but could determine long-term land carbon storage. Our results highlight the limited implication of interannual variability for long-term tropical land carbon-climate feedbacks and help isolate remaining uncertainties with respect to water limitations on tropical land carbon sink in Earth system models.

Dataset for Multi-Task Learning for "Simultaneous Retrievals of Passive Microwave Precipitation Estimates and Rain/No-Rain Classification". Global Precipitation Measurement (GPM) dual-frequency precipitation radar (DPR) and Goddard Profiling Algorithm (GPROF) were provided from NASA Global Precipitation Measurement Precipitation Data Directory (https://gpm.nasa.gov/data/directory). The original data used for this study have been supplied by JAXA’s GSMaP. The source code for preprocessing and model training is available at https://doi.org/10.5281/zenodo.7627112.

Three-dimensional (3D) hetero-integration technology is poised to revolutionize the field of electronics by stacking functional layers vertically, thereby creating novel 3D circuity architectures with high integration density and unparalleled multifunctionality. However, the conventional 3D integration technique involves complex wafer processing and intricate interlayer wiring. Here we demonstrate monolithic 3D integration of two-dimensional, material-based artificial intelligence (AI)-processing hardware with ultimate integrability and multifunctionality. A total of six layers of transistor and memristor arrays were vertically integrated into a 3D nanosystem to perform AI tasks, by peeling and stacking of AI processing layers made from bottom-up synthesized two-dimensional materials. This fully monolithic-3D-integrated AI system substantially reduces processing time, voltage drops, latency and footprint due to its densely packed AI processing layers with dense interlayer connectivity. The successful demonstration of this monolithic-3D-integrated AI system will not only provide a material-level solution for hetero-integration of electronics, but also pave the way for unprecedented multifunctional computing hardware with ultimate parallelism.

Abstract The clustering effect of β-SiC nanoparticles on the electrical conductivity of polypropylene matrix composites was investigated through a new multiscale modeling framework where density functional theory-based first-principles calculation and electron hopping-based numerical homogenization are integrated. According to parametric studies for particle dispersion states, the electrical conductivity of the nanocomposites clearly depends on the dispersion/agglomeration of the nanoparticles. Due to the work function of β-SiC, agglomerated particles made a greater contribution to improving electrical conductivity when compared to well-dispersed particles. In addition, the microstructure-conductivity relationship was determined using the clustering density. The proposed framework was validated with the reported experimental literature.

The impact of climate change on typhoon is of great concern in the East Asia. In particular, typhoon heavy rainfall has a destructive impact on our society and economy since they are many megacities along the coastal regions. Although observations suggest significant changes in typhoon heavy rainfall, the contribution of anthropogenic forcing has not been determined.In this study, we show that anthropogenic global warming has a substantial impact on the observed changes in typhoon heavy rainfall in the western North Pacific region. Observational data show that, in general, typhoon heavy rainfall has increased (decreased) in coastal East Asia (tropical western North Pacific) during latter half of 20th Century and onward. A similar spatial distribution is found in the “Anthropogenic fingerprint”, difference between Earth systems with and without human-induced greenhouse gas emission, from a set of large ensemble climate simulations. This provides evidence to support that the significant increase in the frequency of typhoon heavy rainfall along coastal East Asia is not explained solely by natural variability. Further, the results show that since mid-1970s, the signal of “Anthropogenic fingerprint” has been increasing rapidly and departs from natural variability in early-2000s.Reference:Utsumi, N., & Kim, H. (2022). Observed influence of anthropogenic climate change on tropical cyclone heavy rainfall. Nature Climate Change, 12(5), 436–440. https://doi.org/10.1038/s41558-022-01344-2