data/samples/articles_10.json

[
  {
    "DOI": "10.1016/J.SCITOTENV.2026.181790",
    "Title": "Local and distal processes determine precipitation isotope records in the Great Plains USA",
    "Year": 2026,
    "Abstract": "Nuanced characterizations of moisture source dynamics and local hydrometeorological processes are essential for interpreting long-term records of stable isotopes in precipitation. Here, we analyze over two decades of stable isotope records from a site in the Great Plains of United States, revealing a distinct seasonal contrast in 18O variability between warm (MarchNovember) and cold (DecemberFebruary) periods. During the warm season, isotopic enrichment was largely driven by enhanced convective activity and sub-cloud evaporation under high VPD conditions. Back-trajectory diagnostics indicate that continental moisture sources dominate precipitation at the study site, while Gulf of Mexico transport via the Great Plains low-level jet exerts a disproportionate influence on 18O and d-excess variability. During extreme precipitation years, isotopic signatures reflect the combined effects of atmospheric circulation anomalies and local aridity. The 2012 drought year exhibited elevated 18O and reduced d-excess consistent with enhanced kinetic fractionation under dry conditions, whereas the wet year 2019 showed isotopic enrichment associated with intensified Gulf-sourced moisture transport under humid conditions. These findings demonstrate how precipitation 18O in the Great Plains integrates both local evaporative demand and large-scale moisture transport processes. Given ongoing challenges in representing humidity trends and regional hydroclimate dynamics in climate models, improved characterization of moisture sources and isotope variability is critical for evaluating model projections and interpreting long-term climate change in semi-arid continental regions."
  },
  {
    "DOI": "10.1016/J.ATMOSENV.2026.121995",
    "Title": "The impact of model retraining frequency on predictive performance in air pollution forecasting",
    "Year": 2026,
    "Abstract": "High-resolution PM 2.5 forecasts are increasingly produced with machine-learning models, yet practical guidance on how often these models should be retrained and validated remains limited. This study quantifies the impact of retraining frequency bias correction air-quality prediction skill across multiple cities contrasting emission sources meteorological regimes, using The Goddard Earth Observing System composition forecast (GEOS-CF) fused in-situ observations. Site-specific trained at least two years hourly data, alternative update schedules (618-month baselines 612-month cycles) evaluated RMSE, R 2 , SHAP-based feature importance. Bias-corrected consistently improves GEOS-CF by more than 107% in reduces RMSE over 75%, annual providing largest gains (13% increase 12% reduction RMSE) relative to frequent updates. SHAP analysis shows that dominant predictors their importance vary city, combinations boundary-layer height, aerosol optical depth, humidity, wind, nitrogen oxides driving levels, demonstrating a single global pre-trained model is inadequate locally tuned required. Together, results define minimum data requirements, preferred intervals, need for site-specific bias-corrected offering concrete design rules operational forecasting systems."
  },
  {
    "DOI": "10.1029/2025JD045335",
    "Title": "A Climatology of Mesoscale Convective System Hazards in the United States and Their Representation in a ConvectionPermitting Model",
    "Year": 2026,
    "Abstract": "Abstract Mesoscale convective systems (MCSs) are large, organized convective storms that frequently produce flash floods and other severe hazards such as damaging winds, hail, and tornadoes. Developing an observationally based MCS hazard climatology is important for establishing a baseline to evaluate the representation of these events in numerical models. This study constructs such a climatology using a 13year MCS data set, storm reports, and atmospheric reanalysis. MCSrelated and nearstorm environmental variables are extracted and used to train objectbased machine learning (ML) models. Three models are developed to predict flash floods, severe (including all wind, hail and tornado events), and significantsevere events, with the latter representing higherimpact hazards. The flash flood and severe models perform well in distinguishing hazardproducing MCSs from nonproducing ones, while the significantsevere model shows limited skill, likely due to sample size constraints. The flash flood and severe models are then applied to the full MCS archive to reconstruct a more complete warm season hazard climatology, addressing the potential underreporting in storm reports and gaps in flash flood reports during early years. The study also examines the application of these models to convectionpermitting model simulations. While the spatial distribution of simulated MCS hazards generally aligns with observations, event frequencies differ considerably. These discrepancies are attributed to biases in both the derived input variables and the representation of MCS properties within the model. , Plain Language Summary Mesoscale convective systems (MCSs) are large, organized thunderstorms that produce hazardous weather, including flash floods, damaging winds, hail, and tornadoes. This study creates a 13year climatology of MCSrelated hazards by developing machine learning models trained on observational and atmospheric data. Three separate models were built: one for flash floods, one for all types of severe weather (wind, hail, and tornadoes), and one for the most intense severe events. The models for flash floods and general severe weather worked well, while the model for the most intense events had more difficulty, likely because these cases are rarer. The study also tested whether the same prediction models could be used with output from highresolution model simulations. While the simulations showed similar storm patterns in space, the number of predicted hazardous events did not always match realworld observations. This mismatch was due to differences in both the simulated weather conditions and how the storms themselves were represented in the model. Overall, the study shows that machine learning can help fill gaps in severe weather records and improve our understanding of storm hazards. , Key Points Developed objectbased machine learning (ML) models to predict mesoscale convective system (MCS)produced hazards Created a more complete MCS hazard climatology for the warm season by correcting historical data gaps and storm underreporting ML applied to ConvectionPermitting Models reveals biases in simulated MCSs and predictors lead to inaccurate hazard frequency"
  },
  {
    "DOI": "10.1029/2025WR040312",
    "Title": "Integrating Satellite Retrievals, Numerical Models, and Machine Learning for Global Groundwater Recharge Estimation",
    "Year": 2026,
    "Abstract": "Abstract Knowledge of the groundwater recharge rate determines whether aquifer use is sustainable. However, accurately measuring recharge globally presents significant challenges due to the complexity of subsurface processes and the lack of direct observational methods. This study addresses these challenges by developing a methodology that integrates satellite data, numerical models, and machine learning to estimate groundwater recharge globally. The methodology involves two steps. First, we run a numerical model, Hydrus1D, to simulate soil moisture fluxes in the unsaturated zone by solving the Richards equation in the vertical direction for 235 different points representing various climates and soil types across the globe. Second, using Hydrus1D inputs and outputs, we train a supervised ensemble machinelearning model, specifically a Gaussian Process Regression model, as an emulator to mimic Hydrus1D. This enables us to process satellite observations efficiently to estimate annual recharge flux globally. Inputs for the model include NASA's SMAP soil moisture and GPM precipitation observations, ERA5 climate reanalysis data, and soil hydraulic properties. Rainfall, unsaturated hydraulic conductivity, and soil moisture are identified as the most significant predictors of groundwater recharge. The approach effectively captures global recharge patterns, particularly in regions with high rainfall, though it shows some limitations in arid areas with minimal recharge and heavily irrigated areas. We confirm the reasonableness of recharge estimates by comparing them with observed changes in subsurface water storage from the GRACE satellite mission. The method effectively captures the observed trends in water storage, demonstrating the model's capability to estimate recharge using largescale satellite and reanalysis data. , Plain Language Summary We need to know how much groundwater is replenished to ensure its sustainable use, but measuring the recharge that replenishes groundwater globally is difficult, if not impossible. This study develops a new method that combines satellite data, computer models, and machine learning to estimate groundwater recharge at a global scale. First, a numerical model simulates how water moves through the soil under a wide variety of climate and soil conditions. Then, a machine learning model is trained to mimic the numerical model and learn from data to estimate annual recharge rates efficiently. The model uses inputs like soil moisture and rainfall data from satellite observations, climate records, and soil properties. The key factors influencing recharge are rainfall, soil moisture, and metrics of how easily water moves through soil. The method accurately maps global recharge patterns but has some limitations in dry and irrigated areas. The results showed that the new method effectively captures realworld recharge patterns, making it a valuable tool for understanding groundwater sustainability. , Key Points A new method integrates satellite data, physicallybased models, and machine learning to estimate global groundwater recharge Rainfall, soil moisture, and unsaturated hydraulic conductivity are key predictors of groundwater recharge The method accurately captures global recharge patterns and water storage trends, aligning with independent studies"
  },
  {
    "DOI": "10.1016/J.ATMOSRES.2026.108937",
    "Title": "Widespread extreme precipitation events over Iran: Large-scale patterns and their associated global indices",
    "Year": 2026,
    "Abstract": "Widespread extreme precipitation events (WEPEs) have intensified globally in recent decades, leading to severe hydrological and socio-economic impacts. Iran has experienced several destructive WEPEs, particularly during 2019 2020. Using Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG-F) data 20002024, (EPs) were classified as WEPEs when their spatially connected affected area exceeded 10% of Iran's land area. A total 153 identified across all seasons. Owing the dominant contribution winter spring events, subsequent large-scale analyses focused on these two seasons, comprising 135 (80 55 spring). Evaluation against observations from 39 synoptic stations IMERG-identified WEPE days, indicates that IMERG-F reliably captures frequency intensity EPs supporting its use long-term analysis. Cluster analysis using Ward's method delineated three spatial sub-regions identified: R1 (western northwestern Iran), R2 (eastern, central southeastern Iran) R3 (southern Iran). exhibited highest while event duration contributed comparably R3, reaching up 500-h/25-year (~21-day/25-year). Composite reveal over R1, are associated with upper deep trough Red Sea, eastern Saudi Arabia, Sea extending toward respectively. Increased instability dynamical forcing substantially strengthen upwards motion, moisture transport is dominated by strong integrated vapor Persian Gulf. Lead-lag composite 200 hPa geopotential height anomalies a clear evolution preceding Iran. Furthermore, closely Circumglobal Wave Train (CGT) both North Atlantic Oscillation (NAO) influence indirectly modulating CGT variability."
  },
  {
    "DOI": "10.1029/2025JD045001",
    "Title": "Urban Impacts on Precipitation in the Greater Ho Chi Minh City Metropolitan Area, Vietnam",
    "Year": 2026,
    "Abstract": "Abstract Urban impacts on precipitation have been extensively studied in midlatitude developed countries. However, despite the recent rapid urban development, such studies focusing on Southeast Asian cities remain limited. This study evaluated urban impacts on precipitation in the Greater Ho Chi Minh City Metropolitan area (GHCM) by multicase numerical simulations using the Weather Research and Forecasting model (WRF) for 217 precipitation events during the rainy seasons from 2013 to 2022. We conducted each simulation with and without 2020 urban land use (URB and NO_URB), and under two different initial and boundary conditions to enhance the robustness of the results. Results show that daily precipitation over the central urban area was up to 30% greater in URB than in NO_URB, with statistical significance ( p < 0.05), especially between 15:00 and 22:00 LT. URB also exhibited higher sensible heat flux (+250 W m 2 ), higher convective boundary layer height (+350 m), and lower surface pressure (0.2 hPa) than NO_URB at 14:00 LT. These thermodynamic changes caused greater Convective Available Potential Energy and smaller Convective Inhibition than those of NO_URB, indicating intensified atmospheric instability due to urbanization. In addition, gridscale horizontal water vapor convergence was stronger in URB than in NO_URB. This convergence is caused by the lower surface pressure in URB than NO_URB. As a result, water vapor accumulated within the convective boundary layer. These results indicate that the urban area strengthened atmospheric instability and gridscale horizontal water vapor convergence, resulting in greater precipitation than the nonurban case. , Plain Language Summary Urban effects on rainfall have been well studied in developed countries in the midlatitudes. However, even though cities in Southeast Asia have grown rapidly in recent years, there are still only a few studies on this region. This study examined how urbanization affects rainfall in the Greater Ho Chi Minh City Metropolitan area (GHCM), using numerical simulations for 217 rainy events from 2013 to 2022. For each event, we ran two simulations: one that included the land use of urban areas as of 2020, and one that included urban areas removed. Our results showed that, in the urban scenario, daily rainfall in the central urban area was up to 30% greater than that in the nonurban scenario, especially between 15:00 and 22:00 LT. A key difference from cities in midlatitude regions is that this increase in rainfall lasted for a longer time in GHCM. We found two main reasons for the increased rainfall in the urban scenario. First, urbanization caused the atmosphere to become more unstable. Second, urbanization caused the horizontal moisture convergence in the urban area, and as a result, water vapor accumulated within the convective boundary layer. , Key Points Under synopticscale undisturbed conditions, rainfalls over the Greater Ho Chi Minh City Metropolitan Area increase because of urbanization Urbaninduced strong atmospheric instability and horizontal water vapor convergence were key drivers of the urbaninduced rainfall increase A set of 434 simulation cases enabled statistically robust assessment of urban impacts on precipitation"
  },
  {
    "DOI": "10.1029/2025JD045573",
    "Title": "MaddenJulian Oscillation and Atmospheric Rivers: New Insights on Water Source and Transport for Extreme Rainfall Over the Western U.S.",
    "Year": 2026,
    "Abstract": "Abstract Atmospheric rivers (ARs) were first documented by Zhu and Newell for transporting global water vapor. ARs contribute to extreme rainfall, especially over the Western United States. The primary water vapor source of ARs is from the tropical ocean, where convective systems bring the moist flux upward from the surface to the troposphere. Previous studies have investigated ARs in connection to the MaddenJulian Oscillation (MJO) using the Realtime Multivariate MJO (RMM) Index, which is based on Empirical Orthogonal Function (EOF) analysis of outgoing longwave radiation and upperlevel wind. The question of what is the physical mechanism connecting the MJO and ARs remains unclear. This study aims to provide new insights into the effect of MJO convection as a water vapor source for ARs by investigating the direct connection between the MJO convection and ARs using the Largescale Precipitation Tracking (LPT) developed by Kerns and Chen. We track the MJO largescale precipitation and ARs in time and space using satellite data and reanalysis data from 2000 to 2024. We find that largescale convection of the MJO serves as a major water vapor source for ARs during the boreal winter (DecemberMarch) when the MJO LPT systems extended further in the westcentral Pacific. During these months, ARs are twice as likely to occur when the MJO convection is active. ARs are stronger when they are physically connected to MJO convection. These stronger ARs are more likely to lead to increased extreme rainfall and flood risk along the U.S. West Coast. , Plain Language Summary Atmospheric rivers (ARs), vehicles of water vapor transport from the tropics to the midlatitudes, play a key role in extreme rainfall and flooding on the West Coast of the United States. Past research has linked the ARs to the MaddenJulian Oscillation (MJO) using common statistical indices, but these indices cannot capture the actual physical connection between the two, especially the intensity and duration of ARs to the MJO convection. To better understand the real connection of the two physical systems, we use a tracking method that follows the MJO largescale precipitation in connection to ARs from 2000 through 2024. We find that explicitly tracking the location of the MJO precipitation for individual MJO events provides new insights on the role of the MJO, as the main water source, in AR intensity and duration. In particular, we find that ARs that physically overlap with MJO rainfall last longer, are larger, and are more intense. These ARs are also more likely to lead to flooding events in California, Oregon, and Washington during the boreal winter (DecemberMarch). , Key Points Largescale convection linked to the MaddenJulian Oscillation (MJO) is a major water vapor source for wintertime atmospheric rivers (ARs) ARs are about twice as likely to occur when MJO convection is active. They tend to be stronger when physically connected to MJO convection These stronger ARs, associated with MJO convection, are linked to extreme rainfall and heightened flood risk along the U.S. West Coast"
  },
  {
    "DOI": "10.1029/2025GL120807",
    "Title": "Tropical TCV as a Process Diagnostic: Connecting Probability to Processes in kmScale Models Via Moisture Budget Statistics",
    "Year": 2026,
    "Abstract": "Abstract Observations show a bimodal frequency distribution in total column vapor (TCV) over tropical oceans, with convective rainfall predominantly produced on the moist side of the frequency minimum between two modal peaks. Here we show a kmscale model of the tropics with explicit convection produces a bimodal TCV distribution, whereas the same model with parameterized convection does not. The parameterized model also fails to realistically confine rainfall to a moist mode. Using concepts from statistical mechanics we relate TCV frequency and tendency, and isolate process contributions to tendency in TCV phasespace. Where bimodality is lacking, we find an incorrect relationship between moisture flux convergence and TCV in environments with little or no rainfall. The resulting lack of a strong gradient in TCV tendency with respect to TCV is inconsistent with that expected to maintain a TCV frequency minimum. Our results demonstrate value in the TCV distribution as a process diagnostic. , Plain Language Summary It has been observed that the tropical atmosphere at any given time is mainly split into two regions: a moist region where columns contain high quantities of water vapor, and a drier region with less vapor. These regions are separated by a narrow margin zone with moderate amounts of moisture. In other words, moist and dry columns are both common, but marginal columns are less frequent. Models of the tropical atmosphere need to capture this distribution because moisture affects cloud and rainfall patterns, with implications for weather forecasting and for climate simulation. For a model of the entire tropical atmosphere on a kmscale grid, we show that a realistic column moisture distribution can be limited by model subgrid physics. Failure to realistically separate moist and dry regions of the distribution is associated with rainfall biases and with weaker redistribution of moisture by winds. , Key Points The tropical moisture distribution is bimodal in observations, but reproducing the bimodality in kmscale models depends on model physics The lack of bimodality is associated with excessive probability of light rainfall, specifically in dry to moderately moist columns In columns with low rain, the model without bimodality has a weak relationship between moisture and flux contributions to moisture tendency"
  },
  {
    "DOI": "10.1029/2025JD044800",
    "Title": "Precipitation Efficiency by Storm Type in kmScale Climate Simulations and Satellite Observations",
    "Year": 2026,
    "Abstract": "Abstract Precipitation efficiency (PE), the fraction of cloud condensate converted into surface precipitation, is a key metric for understanding Earth's hydroclimate but has been challenging to observe and evaluate in models. Using satelliteretrieved ice water path (IWP) and precipitation estimates, we assess PE in a 4km convectionpermitting climate simulation across six North American storm types: smallscale, intermediate, and mesoscale convective systems, tropical cyclones, extratropical cyclones, and atmospheric rivers. The simulation reproduces observed PE characteristics and IWPprecipitation relationships across all storm types. The agreement between IWPbased and an alternative PE estimation method leveraging saturation adjustment is better for convective storms that produce a lot of cloud ice. PE varies by storm type, with tropical cyclones exhibiting the highest PE and least sample variability. PE also evolves over storm lifetimes, showing lower values during initiation and decay. Our results support the use of kmscale models for realistic PE analysis. , Plain Language Summary Precipitation efficiency (PE) measures how much condensate in clouds reaches the ground as rain or snow. Understanding PE is important because even if models produce the same amount of precipitation, differences in efficiency reflect different cloud and precipitation processes that affect weather and climate. However, PE has been difficult to measure, limiting how well weather and climate models can be evaluated. In this study, we use satellite data measuring ice in clouds and surface precipitation to estimate PE. We compare these estimates with results from a highresolution climate model that simulates weather at a grid spacing of 4 km. We examine six common storm types in North America, including thunderstorms, hurricanes, and atmospheric rivers. We find that PE varies by storm type. Hurricanes have the highest PE and show the least variability between cases. The model reproduces these patterns and the relationship between cloud ice and precipitation. We also compare two methods for estimating PE in the model: one based on cloud ice growth and another using detailed condensation physics. These methods agree best for convective storms producing lots of cloud ice. Our results show that finescale climate models can realistically simulate how clouds produce precipitation, improving prediction across scales. , Key Points Precipitation efficiency is accurately captured by kilometerscale model simulations when compared to satellite data Agreement between precipitation efficiency calculation methods is strongest for convective storms Precipitation efficiency varies with storm type and life stage, with tropical cyclones exhibiting the highest values"
  },
  {
    "DOI": "10.1002/JOC.70099",
    "Title": "La Nina Impacts on Southeastern African Climate: The Influence of Event Duration",
    "Year": 2025,
    "Abstract": "ABSTRACT The multiyear La Nina event of 20202023, which brought with it several climate disasters across the globe, sparked both mainstream and scientific interest in La Nina events, which typically have received less attention than El Nino. In southern Africa, there is a general expectation in the scientific community and among user groups that La Nina events result in cool and wet summers. However, such impacts do not always occur and the full diversity of La Nina impacts, including multiyear events, has not been systematically explored. Here, various temperature and rainfall characteristics occurring during three categories of La Nina eventssingle, double, and triple year events areinvestigated for the period 19702023. Spaceparameter bubble plots are used to display anomalies in midlate summer rainfall, heavy rain and dry spell frequencies, and in mean temperature and extreme heat days across four domains in southeastern Africa. Despite being relatively populated with important port cities and agriculture, not much work has focused on this region. Doubleyear La Nina summers were the least consistent in exhibiting expected wet characteristics. During tripleyear events, the subtropical domains, southern Mozambique and eastern South Africa, showed the greatest tendency towards wet conditions, although a few seasons deviated significantly in rainfall distribution from their multiyear counterparts. In contrast, the tropical domains, southern Tanzania and northern Mozambique, more consistently were cooler than average than the subtropical domains. These findings highlight the diversity of La Nina impacts on summer conditions in southeastern Africa and show that La Nina events are not necessarily associated with cooler and wetter than average summers over the region."
  }
]