PSL Publication Report

Abstract

The High-Resolution Rapid Refresh (HRRR) model is a convection-allowing model used in operational weather forecasting across the contiguous United States (CONUS). To provide a computationally efficient alternative, we introduce HRRRCast, a data-driven emulator built with advanced machine learning techniques. HRRRCast includes two architectures: a ResNet-based model (ResHRRR), our main architecture, and a graph neural network–based model (GraphHRRR). ResHRRR utilizes convolutional neural networks enhanced with squeeze-and-excitation blocks and feature-wise linear modulation and supports probabilistic forecasting via the denoising diffusion implicit model (DDIM). To better handle longer lead times, we train a single model to predict multiple lead times (1, 3, and 6 h) and then use a greedy rollout strategy during inference. When evaluated on composite reflectivity over the full CONUS domain using ensembles of 3–10 members, ResHRRR outperforms HRRR forecast at the light rainfall threshold (20 dBZ) and achieves competitive performance at moderate thresholds (30 dBZ). Our work advances the pioneering StormCast model described in Pathak et al. by 1) training on the full CONUS domain, 2) training on multiple lead times to improve long-range performance, 3) using analysis data for training instead of the +1-h postanalysis data inadvertently used in StormCast, and 4) incorporating future Global Forecast System (GFS) weather states as inputs and adding a downscaling component that significantly improves long-lead forecast accuracy. Grid-based, neighborhood-based, and object-based verification metrics confirm improved storm placement, lower-frequency bias, and enhanced success ratios compared with HRRR. Additionally, HRRRCast’s ensemble forecasts maintain sharper spatial detail and reduced blurriness than deterministic baselines, with power spectra more closely matching HRRR analyses. Overall, HRRRCast represents a step toward efficient, data-driven regional weather prediction with competitive accuracy and ensemble capability.

Significance Statement

This study introduces HRRRCast, a data-driven emulator for the High-Resolution Rapid Refresh (HRRR) model, designed for regional, convection-allowing forecasting over the contiguous United States (CONUS) at 6-km resolution. HRRRCast uses a SE-ResNet-based architecture and diffusion modeling for generating probabilistic forecasts. Trained on HRRR analysis data with Global Forecast System (GFS) synoptic input—including future GFS states—it supports multilead time prediction (1, 3, 6 h) in a single model. HRRRCast outperforms HRRR in composite reflectivity skill at 20 dBZ and achieves competitive performance at 30 dBZ. It produces sharper forecasts, reduced bias, and more reliable ensembles, offering a scalable, cost-effective alternative to physics-based models for regional ensemble forecasting.

Abstract

As machine learning becomes ever more prevalent within Earth and atmospheric science, clear and consistent descriptions of models, observations, and observation-based datasets, particularly reanalyses, are increasingly vital. Reanalyses remain foundational for climate and weather research, but advancements in data assimilation and model nudging methods, as well as increasingly complex physical parameterization options, mean that not all variables within reanalyses are equally constrained by observations. Because machine learning models are often trained and evaluated on such datasets, imprecise terminology and inadequate documentation can lead to a loss of information content, mislead users unfamiliar with data nuances, lead to the training of flawed machine learning models, and ultimately result in model evaluations that do not realistically describe performance relative to observations. This essay argues for more careful use of the term “reanalysis,” emphasizing that it should be reserved for datasets that explicitly blend observations with models through data assimilation. It highlights the rise of “reanalysis adjacent” datasets, as well as the growing disconnect between data producers and increasingly interdisciplinary users, particularly within the machine learning community. It offers guidance for dataset producers and users, alongside recommendations to enhance transparency, including renewed use of variable classification systems, better documentation of variable-specific uncertainties, and greater community-wide emphasis on data transparency. Without such efforts, Earth science datasets may be applied indiscriminately, regardless of fitness for purpose. Ensuring trustworthy and interpretable data are essential for maintaining the scientific integrity of Earth system modeling in the machine learning age.

Abstract

The Arctic sea-ice surface energy budget is characterized by two radiative states driven by the presence or absence of opaque clouds that warm the surface. Ground-based observations from the 1998 SHEBA campaign collected over sea-ice suggested that the two radiative states correspond to distinct surface pressure values. Using satellite data, we search for such relationship over 13 years across all sea-ice-covered Arctic basins in winter and spring when radiative states influence sea-ice growth and can trigger early melt. We find a consistent, statistically significant pressure influence across basins, with higher surface pressure associated with the transmissive state and lower pressure with the opaque state. The pressure offset is weaker (∼4 hPa) than the SHEBA-based estimate (∼15 hPa), likely caused by an anomalous circulation during spring 1998. These results indicate that surface pressure weakly but systematically modulates radiative-state occurrence, helping to explain small regional differences in surface cloud radiative warming.

Plain Language Summary

Over Arctic sea ice, the atmosphere commonly settles into two modes: clearer or thin-cloud days that let heat escape to space through infrared radiation, and opaque-cloud days that trap heat near the surface. A field campaign called SHEBA suggested that surface air pressure strongly influences which mode occurs. We examined this relationship using 13 years of satellite observations above the Arctic sea-ice from January to June, when clouds can limit winter ice growth or trigger earlier spring melt. We find a consistent and statistically robust connection: higher surface pressure is linked to clearer days, while lower pressure is linked to cloudier days. The typical pressure difference is smaller, about 4 hPa, than suggested by SHEBA, about 15 hPa, likely because spring 1998 experienced unusual circulation. This weak but systematic influence of surface pressure helps explain small regional differences in cloud-related surface warming.

Abstract

Identifying sources of ice nucleating particles (INPs) in the central Arctic is important for understanding controls on the phase of Arctic mixed-phase clouds (AMPCs). We show meltwater samples collected from the Arctic pack ice in July 2020 contain biological INPs active at relatively warm temperatures (T ≥ −10°C). We attribute the meltwater INPs to organisms and processes unique to a meltwater habitat. Concentrations of biological INPs active at T ≥ −10°C on filters deployed downwind of meltwater sites showed an enhancement associated with surface proximity. We hypothesize that time over the melt-pond-covered pack ice may have influenced the higher concentrations of biological INPs on the aerosol filters. More work to resolve emission mechanisms from melt ponds is necessary for understanding the extent of this potential source, which may increase importance as melt seasons extend spatiotemporally.

Plain Language Summary

Understanding what controls cloud properties in the Arctic is important for predicting weather and climate in the region. In this study, we found that, compared to seawater, meltwater on top of sea ice contains high concentrations of ice nucleating particles (INPs), which help form ice in clouds. The meltwater INPs were biological and could trigger ice formation at relatively warm temperatures (above −10°C). Air samples taken near meltwater areas had more of these particles compared to air from farther away, and the concentration of INPs in the air seemed linked to time spent over the ice rather than over open ocean. This suggests that the melting sea ice surface may be an important source of biological INPs. More research is needed to understand how these particles are emitted from meltwater and how big a role they play in the radiation budget as Arctic melt seasons grow longer and larger.

Abstract

The surface energy budget (SEB) is a central regulator of Arctic climate and sea ice evolution, yet its processes remain poorly constrained due to sparse observations and complex, coupled surface-atmosphere interactions. This study leverages year-long measurements from the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) to provide the most comprehensive assessment to date of the central Arctic SEB and its modulation by atmospheric variability. Ship- and ice-based observations from October 2019 to September 2020 were used to directly measure or tightly constrain each term of the SEB, leading to exceptional energetic closure with the seasonal snow and ice mass balance. The analysis reveals strong seasonal transitions in atmosphere-surface energy transfer that are modulated by the atmospheric state and constrained by the ability of the surface temperature to respond. Classification of the atmosphere into its two dominant radiative states—the semi-transparent (ST) and opaque (OP)—highlights the central role of synoptic-scale variability in clouds. The ST atmospheric state dominated the long winter ice growth season, with limited cloudiness supporting persistent surface radiative cooling and ice growth. The OP state, associated with liquid-containing or thick ice clouds, became dominant in spring, with the combination of increased solar heating and cloud surface longwave warming driving ice and snow melt. Eddy covariance versus bulk approaches for deriving surface turbulent heat fluxes provide vastly different perspectives on the role of turbulence in modulating the SEB. These results establish a high-quality benchmark dataset for Arctic SEB studies and demonstrate how the balance of atmospheric radiative states exerts a first-order control on the annual evolution of the sea ice. The findings have broad implications for advancing observing technologies, understanding Arctic amplification, improving climate models, and predicting future sea ice change.

Abstract

Prior to the onset of the South Asian monsoon, the Arabian Sea experiences a warming phase during which the Arabian Sea mini warm pool (ASMWP) becomes one of the world's warmest oceanic regions, characterized by sea surface temperatures (SSTs) exceeding 30°C. To understand the role of this warming in monsoon evolution, we performed an SST sensitivity experiment using the Weather Research and Forecasting (WRF) atmospheric model. Our case study focuses on the 2023 monsoon, which was characterized by high initial SST that declined faster than in normal years. We found that if the SST declines more slowly than normal, the off-shore precipitation increases by more than 100% around the ASMWP, but decreases in the northern Arabian Sea and the western coast of India. To understand the precipitation differences, we separated the components of the integrated vapor transport (IVT) and found that the changes in both water vapor and wind affect precipitation. The analysis revealed that the changes in water vapor are due to (a) stronger evaporation and precipitation in the ASMWP, and (b) moisture advection outside the ASMWP. It is also shown that the pressure adjustment mechanism can explain the changes in wind speed. With warmer SST conditions, the atmosphere pressure drops and causes wind convergence, thereby creating a weak cyclonic wind anomaly that redistributes the water vapor. Finally, we examined the vertically integrated buoyancy in the context of an idealized plume model. In the ASMWP, temperature changes contribute to increases in buoyancy below 850 hPa, but humidity changes contribute to far more increases in buoyancy between 800 and 600 hPa, which enhance the convection and lead to more precipitation.

Plain Language Summary

Before the South Asian monsoon, the Arabian Sea mini warm pool (ASMWP) becomes one of the world's warmest oceanic regions. To understand the impact of the warm pool on the monsoon, we performed numerical simulations focusing on the 2023 monsoon when the ocean temperature was high and declined faster. Based on the simulations, we found that the precipitation increases by 100% in the warm pool if the ocean temperature declines more slowly, but the precipitation outside the warm pool decreases when the ocean gets warmer. To understand the precipitation differences, we analyzed the output from the numerical model and found that the changes in humidity and wind are important. Our analysis revealed that the changes in humidity are due to (a) stronger evaporation and precipitation in the warm pool, and (b) changes in wind speed outside the warm pool that redistribute the water vapor. The changes in wind speed are due to a weak cyclone generated by the ocean temperature differences in the mini warm pool. In the vertical direction, the humidity changes lead to the instabilities of the atmosphere that contribute to more precipitation.

Abstract

Anthropogenic climate change is unfolding rapidly, yet its regional manifestation can be obscured by internal variability. A primary goal of climate science is to identify the externally forced climate response from among the noise of internal variability. Separating the forced response from internal variability can be addressed in climate models by using a large ensemble to average over different possible realizations of internal variability. However, with only one realization of the real world, it is a major challenge to isolate the forced response directly in observations. In the Forced Component Estimation Statistical Method Intercomparison Project (ForceSMIP), contributors used existing and newly developed statistical and machine learning methods to estimate the forced response over 1950–2022 within individual realizations of the climate system. Participants used neural networks, linear inverse models, fingerprinting methods, and low-frequency component analysis, among other approaches. These methods were trained using large ensembles from multiple climate models and then applied to observations. Here, we evaluate method performance within large ensembles and investigate the estimates of the forced response in observations. Our results show that many different types of methods are skillful for estimating the forced response in climate models, though the relative skill of individual methods varies depending on the variable and evaluation metric. Methods with comparable skill in models can give a wide range of estimates of the forced response pattern in observations, illustrating the epistemic uncertainty in forced response estimates. ForceSMIP gives new insights into the forced response in observations, its uncertainty, and methods for its estimation.

Significance Statement

The Forced Component Estimation Statistical Method Intercomparison Project (ForceSMIP) aims to reduce uncertainty in estimates of the climate response to anthropogenic and other external forcing and to evaluate statistical and machine learning methods designed to estimate the forced response from individual realizations of the climate system. New and existing statistical and machine learning methods are evaluated within climate models, for which the forced response is known. Applying these methods to observations gives an estimate of the real-world forced response. The observational forced response estimate agrees with climate models on the large-scale features, but it also shows discrepancies that give insights into responses that may not be simulated well by climate models. In some regions with large internal variability, such as the North Atlantic Ocean, it remains difficult to determine the relative contributions of anthropogenic forcing and internal variability to historical changes.

Abstract

Northeast Pacific marine heatwaves occur year-round but are shaped by seasonally-varying dynamics including El Niño-Southern Oscillation (ENSO) and North Pacific atmosphere-ocean interactions. Using a data-driven cyclostationary linear inverse model constructed from 64 years of monthly sea surface temperature and height reanalyses, we demonstrate that longer-lived marine heatwaves preferentially begin in winter, when ENSO teleconnections and oceanic memory most strongly influence event amplification. Springtime subsurface storage and subsequent fall/winter reemergence of thermal anomalies drive extended persistence through wintertime re-intensification. While intense events can also begin in summer, without strong ENSO and reemergence dynamics they rarely persist. The relative importance of tropical forcing versus North Pacific upper-ocean dynamics varies by region and season, producing distinct marine heatwave “flavors” linked to different ENSO types or to internal North Pacific processes alone. These findings reveal how seasonal phase-locking drives marine heatwave evolution, with implications for predictability and marine ecosystem impacts.

Abstract

Despite decades of research, the Madden-Julian Oscillation (MJO) remains challenging for global prediction systems, partly because its representation is sensitive to convection-environment interactions. We evaluate a new prognostic convective closure in NOAA's Unified Forecast System (UFS) that incorporates large-scale moisture-flux convergence (MFC) as a source alongside buoyancy. By allowing convection to respond directly to developing convergence anomalies, the scheme increases sensitivity to large-scale moistening and produces more coherent MJO preconditioning. Using the UFS Seasonal Forecast System and composite diagnostics across nine MJO events, we show that the mixed closure generates a stronger and more realistic buildup of MFC over the central and eastern Pacific days before peak convection. This enhanced preconditioning, which is weaker in the traditional closure, is accompanied by improved MFC-precipitation coherence and a more organized large-scale MJO structure. These results illustrate how convective-closure formulation can influence the large-scale feedbacks critical for MJO representation.

Plain Language Summary

The Madden-Julian Oscillation (MJO) is a large region of tropical clouds and rainfall that slowly moves around the globe and strongly influences weather patterns worldwide. However, it remains difficult for weather and climate models to simulate and predict. One reason is that models must represent how tropical convective clouds interact with their surrounding environment, which occurs on scales smaller than the model grid. In this study we test a new way of representing these processes in NOAA's UFS. The new method allows convection to respond more directly to large-scale patterns of atmospheric moisture convergence. Using simulations of nine MJO events, we find that this approach produces a more realistic buildup of convergence before major rainfall develops. This leads to a more organized representation of the large-scale MJO circulation, highlighting how the treatment of convection can influence global weather patterns.

Abstract

This study created future streamflow projections using dynamically downscaled precipitation and temperature projections from the Intermediate Complexity Atmospheric Research (ICAR) model and the Variable Infiltration Capacity (VIC) model. These more physically realistic projections were compared with VIC simulations based on the widely utilized localized constructed analog (LOCA) statistical downscaling method to evaluate how downscaling choices influence water supply estimates. ICAR’s annual streamflow change at Lees Ferry, Arizona, was higher than LOCA in the ensemble mean due to equal influences from precipitation and temperature. However, annual streamflow changes at Lees Ferry were not significantly different between downscaling methods despite significant differences in precipitation and temperature as there was still a substantial spread and there were no systematic differences between downscaling methods across Earth system models. Similar streamflow projections were the result of different seasonal and spatial patterns. For example, while ICAR projected less cool-season precipitation than LOCA at high elevations (>3000 m), summer precipitation projections, which are often characterized as more uncertain, differed substantially between downscaling methods and increased ICAR’s streamflow projections relative to LOCA. Furthermore, downscaling decisions were more likely to affect streamflow projections at the local scale rather than the regional scale as 25%–32% of the smaller catchments within the upper basin showed significant differences between downscaling methods in annual streamflow changes. Finally, this study considers the uncertainty and confidence in the process representation of precipitation and temperature, emphasizing how spatial and seasonal differences in these variables—and the ways their underlying processes are represented—directly influence streamflow projections.

Abstract

The Atlantic Tradewind Ocean–Atmosphere Mesoscale Interaction Campaign (ATOMIC) and Elucidating the Role of Clouds Circulation Coupling in Climate Campaign (EUREC⁴A) were joint U.S./EU campaigns in January–February 2020 that sought to investigate atmospheric shallow convection and air–sea interaction during the winter season in the northwest tropical Atlantic. This study evaluates bulk surface turbulent heat fluxes and the mean state parameters used to calculate them, such as 2-m surface air temperature T_a and humidity q_a, 10-m wind speed U₁₀, and sea surface temperature (SST), from 15 platforms during the two joint field campaigns. These four state parameters were adjusted for all platforms to a calibrated reference ATOMIC platform, the NOAA Ship Ronald H. Brown. Results show a reduction in the mean latent heat flux H_L of 11.50 W m⁻² and an increase in the mean sensible heat flux H_S of 1.2 W m⁻². Heat fluxes from the fifth generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5) and Objectively Analyzed Air–Sea Fluxes (OAFlux) satellite-derived blended datasets were evaluated with the heat fluxes calculated from the corrected in situ state parameters. ERA5 and OAFlux had higher wind speeds than ATOMIC resulting in 12.51 W m⁻² more H_L for ERA5. However, OAFlux had 10.47 W m⁻² less H_L which is attributed to its 0.93 g kg⁻¹ wet bias in q_a. OAFlux and ERA5 H_S were higher than observations (3.81 and 4.49 W m⁻², respectively). The diurnal cycle of ERA5 agreed well with ATOMIC observations of U₁₀, but phase differences, particularly in the afternoon, were evident for T_a and q_a.

Abstract

Universities are increasingly becoming centers for activities that bridge the gap between weather and climate information. These activities integrate emerging research with science-informed decision-making to contribute to the development of official climate services. Yet there is relatively little literature describing how they balance the competing demands of stakeholder engagement, academic incentives, and the need for actionable science. Here, we contribute to the conversation by examining Western Water Assessment (WWA)—a boundary organization and university-based climate information provider operating in Colorado, Wyoming, and Utah for over two decades. Originally motivated by improving the usability of seasonal forecasts for water managers, WWA has broadened its scope in response to evolving stakeholder needs as well as extreme events in its region. This transformation has been facilitated by both its more flexible funding model and its position within a university, both of which supported WWA in cultivating longer-term relationships and maintaining institutional stability. This paper highlights the strengths and challenging experiences faced by a university-based climate information provider through the lens of the history and evolution of WWA. The strengths of the university setting, such as access to innovation, credibility, stability, flexibility, and neutrality, on balance outweigh the negatives of misaligned incentive structures and limited capacity to meet demand. Overall, this paper underscores the value of integrating research and outreach within a robust university framework and seeks to provide useful insights into how university-based climate information provision can support societal resilience.

Abstract

The atmospheric response to the solar eclipse of 8 April 2024 in North America is investigated with a specific focus on the marine atmospheric boundary layer (MABL). We leverage measurements collected during the Third Wind Forecast Improvement Project (WFIP3), including Doppler lidars, sonic anemometers, and thermodynamic profiler data to investigate the atmospheric response across sites that experienced partial eclipse conditions with nearly 90% obscuration. Using these measurements, we examine eclipse-induced changes in key meteorological parameters, such as temperature, wind speed, and turbulent fluxes. Most previous eclipse studies have been conducted over land, whereas this study provides new observations for both coastal and marine environments, offering additional insight into eclipse-driven variability in the MABL. The findings confirm a notable decrease in downwelling shortwave radiation during the eclipse, which results in rapid cooling of surface air. The temperature reduction ranges from to in coastal regions and from to over the ocean. This analysis suggests that the MABL’s higher thermal inertia compared to coastal regions moderates the temperature decrease during the eclipse. Wind speed exhibits a more complex behavior, as it is influenced by both the MABL and preexisting synoptic conditions. Although a reduction in wind speed is observable up to approximately 140 m above ground level (AGL) at more inland sites, at other locations closer to the coast, this reduction is constrained to the lowest 100 m AGL. Turbulence parameters retrieved from sonic anemometers, such as turbulence kinetic energy, turbulent heat flux, and friction velocity, decrease during the eclipse at coastal sites, accompanied by a brief transition of atmospheric stability from unstable to neutral or weakly stable conditions. For the open-ocean sites, the variability in turbulence statistics and atmospheric stability is minimal during the occurrence of the eclipse.

Abstract

This study explores the ability of data-driven models and an operational, process- based SNOW-17 snow model to estimate historical snow water equivalent (SWE). Current operational streamflow forecasts issued by the California-Nevada River Forecast Center use SWE simulated from SNOW-17. Three machine learning methods of varying complexity: Multiple- Linear Regression, Random Forest Regression, and Long Short-term Memory (LSTM) were tested. SWE from LSTM, the best performing data-driven model, was then compared to SNOW-17 for operational context to understand their respective skills and biases, thereby highlighting opportunities to provide improved SWE information for streamflow forecasts. New hydrological insights for the region: LSTM outperformed other data-driven methods and SNOW- 17, with cross-basin median KGE of 0.76 and 0.45 in LSTM and SNOW-17, respectively during 2011–2016. This demonstrates LSTM’s ability to capture non-linear processes and delayed hydrologic responses, enabled by its internal memory structure. Both models performed better at higher elevations, more so in SNOW-17. Biases in SNOW-17 were similar across different test periods, while LSTM was sensitive to training data. However, basin-to-basin variability in KGE and bias was larger in SNOW-17. In addition, snowmelt duration was shorter and snow disappeared earlier in SNOW-17. Overall, LSTM’s relatively high skill in estimating basin-wide SWE indicates strong potential for integration into operational hydrologic forecasting systems, with possible benefits for improving streamflow simulations.

Abstract

This study evaluates the capability of the local ensemble transform Kalman filter within the Joint Effort for Data Assimilation Integration (JEDI) framework to directly assimilate radar observations for improving short-range ensemble forecasts of landfalling hurricanes. Experiments are performed using a model configuration that adopts the physics suite of the Hurricane Analysis and Forecast System (HAFS) for Hurricane Ida (2021). Four hourly cycling data assimilation (DA) experiments are conducted and are compared against a control forecast without DA. The baseline DA experiment assimilates only conventional observations, while the other three additionally assimilate radar reflectivity from the Multi-Radar Multi-Sensor system, radial velocity from WSR-88D radars, or both. Results show that assimilating radial velocity substantially improves the inner-core wind structure, reduces the ensemble spread in track forecasts during the DA cycling, and enhances the accuracy of landfall location forecasts. Assimilating reflectivity improves rainfall forecasts, particularly for outer rainbands prior to landfall, while having a relatively minor impact on intensity. Postlandfall rainfall prediction also benefits from radial velocity assimilation, mainly through improved track forecasts. Quantitative verification using equitable threat scores shows that radar DA improves forecasts of total rainfall accumulation. Radial velocity assimilation contributes most to moderate rainfall prediction, while reflectivity assimilation offers greater skill in forecasting extreme rainfall. These findings underscore the value of incorporating radar data into ensemble hurricane forecasting and provide practical guidance for future implementation of radar DA within the HAFS–JEDI framework for operational applications.

Abstract

Background: To make strategic decisions about safety and resource allocation, managers must understand when prescribed burn windows of opportunity occur, how they vary throughout the year, and how to anticipate them. For two burn types—pile and broadcast—we characterize the temporal and spatial distributions of past “goburn” days identified from a database of 15,468 burn permits from 2011 to 2024 in the Northern California Geographic Area Coordination Center’s region. We assess the mean temporal evolution of daily anomalies of local- and synoptic-scale environmental variables that occurred in the 2 weeks surrounding those ignitions. We then estimate the mean number of historical burn windows for both burn types in each month using a novel approach to define our prescription criteria. This approach defines the criteria based on ranges of environmental conditions present on past go-burn days.

Results: Broadcast burns exhibit a strong seasonal signal with the greatest number of ignitions in June and October. Pile burns were four times more numerous than broadcast burns and occurred most often in November–December. Ignition locations varied by month and burn type; e.g., piles ignited in the fall most often occurred in the Sierra while piles ignited in the spring most often occurred in the Foothills ecoregion. Composite local- and synoptic-scale variables indicated preferences for certain environmental conditions depending on the month. Conditions preceding February broadcast burns exhibited warming and drying near the surface and anomalous ridge patterns in the upper atmosphere. Pile burns in October were associated with cooler and wetter surface conditions and anomalous trough patterns aloft. The expected number of burn windows varied by month and burn type (e.g., broadcast burn windows were, on average, three times more likely than those of piles in October).

Conclusions: We found that ample opportunities exist in northern California for burning throughout the year, yet those opportunities heavily depend on burn type. Findings herein are useful for determining favorable environmental patterns that create burn windows and for identifying priority locations and times to implement broadcast and pile burns.

Abstract

The Third Wind Forecast Improvement Project (WFIP3) is a multi-institutional field campaign designed to advance the understanding and prediction of the offshore atmospheric boundary layer along the US east coast. Extending from February 2024 through August 2025, WFIP3 combines long-term coastal and offshore measurements with targeted modeling and forecasting efforts. This data paper presents the WFIP3 event log, a curated record of 578 d of meteorological phenomena and field observations that complements the campaign's extensive high-frequency datasets. The event log provides both manually documented daily weather discussions and automatically derived indicators of atmospheric processes – including low-level jets, wind ramps, extreme wind veer, and weak wind conditions – based on observations from scanning lidars deployed at three coastal and offshore sites. The dataset offers structured metadata, standardized time and site identifiers, and consistent terminology to facilitate its integration with WFIP3's observational and modeling data products. The log supports diverse applications, from model evaluation and forecast verification to the selection of case studies on offshore boundary-layer dynamics. The WFIP3 event log is publicly available through the US Department of Energy's Wind Data Hub, providing the research community with a transparent and enduring contextual reference for the interpretation and use of WFIP3 measurements.

Abstract

It is often difficult to attribute uncertainty and errors in atmospheric models to designated model components. This is because sub-grid parameterised processes interact strongly with the large-scale transport represented by the explicit model dynamics. We carry out experiments with prescribed large-scale dynamics and different sub-grid physics suites. This dataset has been constructed for the Model Uncertainty Model Intercomparison Project (MUMIP), in which each suite forecasts sub-grid tendencies at a 22km grid. The common dynamics is derived from a convection-permitting benchmark: an ICON DYAMOND experiment (2.5km grid). We compare four different physics suites for atmospheric models in an Indian Ocean experiment. We analyse their joint PDFs of precipitation and associated physics tendencies for a full month, where precipitation is used to diagnose uncertainty of convective activity. We find that all physics suites produce very similar precipitation amounts, with very high correlations between models, i.e., > 0.95 at the native grid. However, the convection-permitting benchmark is more dissimilar from each of the physics suites, with correlations of ≈ 0.80. Similarly, we show that the vertically averaged physics tendencies in the free-troposphere are highly similar between the four physics suites, yet different if reconstructed for the benchmark. The water vapour sink is very closely linked with precipitation in the four physics suites. This suggests that the coarse-grid models are overconfident.

We hypothese is that variation in unresolved convective structures can lead to variation in the dynamics, following a given amount of latent heating at fine grids, but not in our physics suites. The difference appears to be caused by the explicit interactions between gravity waves, occurring at fine grids only Groot et al. (2024) We assess whether their non-linear feedback from convective precipitation systems explains our joint PDFs of precipitation. The slightly exponential curve supports the interaction mechanism. These findings are further evidence for a non-linear feedback between convective organisation/aggregation and dynamics. This feedback has been studied earlier in a real-case study with ICON by looking from the fixed-physics rather than the fixed-dynamics perspective. Our current results may indicate that sub-grid physics with stochastic physics perturbations emulate convective organisation effects.

Abstract

The surface energy budget governs Arctic sea-ice growth/melt, yet observations are sparse, and reanalysis data sets suffer from systematic biases. Here, we train a neural network with observational data to bias-correct hourly ERA5 fluxes over Arctic ice-covered regions (≥70°N; sea-ice concentration >80%) for 1994–2024. Training data cover two full seasonal cycles and different sea-ice regimes. The neural network reduces RMSE for net shortwave radiation by ∼40%, downward longwave radiation by ∼16% and the total surface energy budget by ∼55%, eliminating the wintertime warm bias of ∼4 K in ERA5. Wintertime surface cooling is reduced by ∼50%, yielding thermodynamic ice-growth estimates of ∼80–120 cm, consistent with SMOS–CryoSat satellite thickness increases and in contrast to the 150–200 cm growth implied by ERA5. Our bias-corrected data capture the observed clear/cloudy states of the winter boundary layer and can be used to study Arctic climatology, evaluate climate models and drive sea-ice-ocean models.

Plain Language Summary

The Arctic Ocean is warming rapidly, and its shrinking sea-ice cover is strongly influenced by exchange of heat between the atmosphere and ice surface. Correctly estimating this energy exchange is crucial for understanding how quickly sea-ice grows in winter and melts in summer. However, direct surface flux measurements over Arctic sea-ice are sparse, and widely used ERA5 reanalysis data contain persistent errors in surface fluxes and near-surface temperature. We use observations from major Arctic field campaigns to train a machine-learning model that corrects these errors. The resulting data set SEBai substantially improves estimates of surface fluxes and 2 m temperature. It reduces errors by 40% for sunlight absorbed at the surface, and more than half for the total surface energy budget. It removes the 4°C winter warm bias in ERA5. These improvements point to a reduced wintertime cooling and weaker summertime heating, suggesting that Arctic sea-ice may be more vulnerable to climate change than ERA5 implies. SEBai also reproduces the observed clear and cloudy winter states that strongly affect thermal infrared radiation but are poorly represented in ERA5. SEBai provides a more reliable baseline for climate model evaluation, Arctic climate studies, and driving sea-ice-ocean models.

Abstract

To develop a dataset of actual evapotranspiration (ET_a) for operational drought monitoring, we used a model of the advection–aridity approach to the complementary relationship between ET_a and evaporative demand and North American Land Data Assimilation System phase 2 (NLDAS-2) atmospheric forcing variables. We first ran the model on an uncalibrated basis, daily from 1980 to 2014 across the contiguous United States (CONUS). We observed strong seasonal differences (overestimation) relative to ET_a from the fourth generation of Global Land Evaporation Amsterdam Model (GLEAM4), particularly in the peak summer season. This motivated calibrating model parameters to maintain a long-term water balance using streamflow from the Hydro-Climatic Data Network in 651 minimally disturbed catchments from the Catchment Attributes and Meteorology for Large-Sample Studies (CAMELS) dataset. Though calibrated ET_a still exceeded GLEAM4 over most of the CONUS, absolute differences were reduced to <1 mm day⁻¹, with regional variations. On a mean annual basis, its spatial distribution aligned with expectations, exhibiting inverse relationships with elevation and latitude, a minimum in the desert Southwest, and maxima in the upper Midwest and along the Gulf Coast and Atlantic coast in the summer. The response of calibrated ET_a to past droughts defined in the U.S. Drought Monitor shows spatial similarity with GLEAM4 and other land surface models driven by NLDAS-2 but with higher skill across the southeastern and western United States and Great Plains and lower skill in the Intermountain West. The development of the 2011 Texas drought was well represented by ET_a anomalies. These results suggest that, although technically simple, our complementary-relationship-derived ET_a has utility in operational drought monitoring.

Significance Statement

Reliable estimates of actual evapotranspiration (ET_a) are critical for monitoring and understanding drought, yet existing datasets often require complex modeling frameworks that do not permit decomposition of the meteorological drivers of ET_a and drought. We develop a long-term, contiguous United States (CONUS)-wide ET_a dataset based on a physically grounded, computationally efficient model using the complementary relationship approach. After regional and seasonal calibration using streamflow data from minimally disturbed catchments and independent estimates of potential evaporation, our ET_a closely aligns with independent benchmarks. The dataset demonstrates particular strength in capturing regional drought signals, especially in the southeastern and western United States and Great Plains. These findings highlight the value of a parsimonious ET_a modeling approach for improving operational drought monitoring capabilities.

Abstract

Quantifying the impact of land use changes on the threat of flash-floods is a critical consideration in flood hazard planning and risk reduction, and is an area of active research. Here, a coupled Weather Research and Forecasting model hydrological extension package (i.e., WRF-Hydro) modeling approach is applied to simulate flash-flooding processes for short-duration, localized, intense precipitation events. To better understand the effect of urbanization on flash floods, a series of numerical experiments is performed surrounding Ellicott City, Maryland, a location which has experienced both significant heavy rainfall events and suburban development over the past several decades. Two intense rainfall events occurring on 30 July 2016 and 27 May 2018 are investigated, respectively, to first calibrate the hydrologic model performance and then quantify the sensitivity of flash flooding to varying degrees of urbanization. Performing the same experiments using observed historical land use states is of more limited insight, as the thrust of suburban development in the Ellicott City region significantly predates satellite-derived land use datasets. Results confirm that urbanization produces larger river streamflow, higher water stages, faster hydrologic responses to achieve peak flow discharge, and shorter recession limbs, even for very intense, short-duration events. The collective findings suggest that WRF-Hydro is applicable for both watershed flash flood prediction and hypothesis testing, and demonstrates potential utility to urban development decision-makers in locations such as Ellicott City, which could face future increases in catastrophic flooding.

Abstract

This study examines ensemble forecast experiments in which model prognostic variables are nudged toward reanalysis values in the tropics to assess the effects of tropical errors on week 3–4 predictions of two long-lasting extreme precipitation events in California during winter 2022–23. For both events, the first spanning late December to mid-January and the second spanning late February to early March, nudging yields significantly improved predictions for the large-scale flow over the North Pacific; however, the impacts on the California precipitation forecast differ between the events. Comparison of the results for the two events highlights that subseasonal prediction of California precipitation extremes requires accurately forecasting predictable signals linked to tropical forcing, as well as extratropical synoptic-scale dynamics. For the December–January event, which is poorly predicted in the nonnudged “control” forecast, nudging results in especially large improvements in the week 3–4 California precipitation forecast combined with decreases in ensemble spread. These improvements correspond to improved prediction of the onset, persistence, and phasing of the North Pacific wave pattern. This pattern fosters landfall of successive cyclones and atmospheric rivers over California. For the February–March event, nudging yields improved prediction of the amplitude and persistence of a blocking ridge over the eastern Pacific but mixed results for the California precipitation downstream. In this latter event, the ensemble spread in the nudged forecasts remains large, and the precipitation forecast accuracy depends strongly on the representation of midlatitude synoptic-scale Rossby wave breaking near the U.S. West Coast on the eastern flank of the blocking ridge.

Significance Statement

We analyze weather model experiments to assess the effects of forecast errors in the tropics (10°S–10°N) on predictions of two extreme precipitation events in California during winter 2022–23. We specifically focus on forecasts generated 3 weeks before the start of each event. Correcting tropical errors produces large improvements in forecasts of the weather patterns associated with each event. However, the impacts of tropical error corrections on the California precipitation forecast differ between the events, reflecting contrasting influences of tropical processes. For the earlier event, the precipitation forecast is improved substantially; for the later event, the forecast is only modestly improved due to errors in midlatitude weather conditions that are not strongly controlled by tropical processes.

Abstract

Two wave buoys deployed in the direct path of Hurricane Milton show a remarkable reduction in significant wave height and peak wave period as the eye of the storm passes and wind speeds are briefly reduced. The rapid adjustment of the waves to the lower wind speeds is initially unexpected, but is explained by the observed changes in the scalar and directional spectra. The scalar spectra show saturation of the high frequency tail in all regions (and at all wind speeds). The directional spectra confirm the radiation of low frequency energy outwards from the storm, such that the most energetic waves never propagate into the eye. Existing parametric models for wave development confirm that waves outside of the eye experience enhanced fetch associated with the translation speed and size of the storm. Inside the eye, the waves are consistent with the fully developed limit at the locally reduced wind speed.

Plain Language Summary

It is well-known that hurricanes have a small region with reduced winds inside the center “eye” of the storm, but little is known about ocean waves within this region. Recent buoy measurements across the eye show abruptly smaller waves inside the eye. Waves in the open ocean typically change slowly by accumulating energy from the wind over large distances and long times, so the abruptness is surprising. How do waves have enough time to adjust to the brief reduction in wind speed? This paper shows that the apparent adjustment is really two distinct processes for two different types of waves. The shorter waves have almost no adjustment (i.e., no change) because they are at a limiting steepness for all of the wind speeds encountered, including the relatively lower wind speeds within the eye. The longer waves do not really adjust either; they are simply absent within the eye because they are formed outside of it and always directed away from it. The absence of longer waves explains most of the abrupt change in total wave height. The analysis concludes by comparing the observations with a simplified model for wave development.