Search

The forecast probability of tropical cyclone (TC) genesis in the western North Pacific from 2017 to 2020 was investigated using global ensembles from the Japan Meteorological Agency (JMA), the European Centre for Medium-Range Weather Forecasts (ECMWF), the U.S. National Centers for Environmental Prediction (NCEP), and the Met Office in the United Kingdom (UKMO). The time of TC genesis was defined as the time the TCs were first recorded in the best-track data (Case 1) and as the time they reached the intensity of a Tropical Storm (Case 2). The results in Case 1 showed that differences between the forecast probability based on each global ensemble were large, even for a 1-day forecast, and that mean probability were from 18 % to 74 %. The forecasts based on the NCEP had a large frequency bias and overpredicted TC genesis events. The results indicated that the representation of genesis events differed greatly between global ensembles. The effectiveness of multiple ensembles was investigated. The results from the threat score and the false alarm ratio indicated that multiple ensembles had skillful forecasts. When the forecast probability was examined for environmental patterns of synoptic low-level flow, the mean 5-day forecast probability was highest for the pattern in the confluence region. The results also showed that the forecast probability was much larger in Case 2 than in Case 1. In all global ensembles, the mean probability with a lead time of up to 1-week was below 10 % for both Case 1 and 2. This result indicates that even with today's operational forecasting systems, it is difficult to regularly predict TC genesis events with a 1-week lead time with high confidence. These results provide a better understanding of TC genesis forecast products in each global ensemble and will be useful information when multipleensemble products are created. [ABSTRACT FROM AUTHOR]

Ensemble forecast post‐processing is a necessary step in producing accurate probabilistic forecasts. Many post‐processing methods operate by estimating the parameters of a predetermined probability distribution; others operate on a per‐lead‐time or per‐station basis. All of the aforementioned factors either limit the expressive power of the methods in question or require additional models, one for each lead time and station. We propose a novel, neural network‐based method that produces forecasts for all lead times jointly and requires a single model for all stations. We incorporate normalizing spline flows as flexible parametric distribution estimators, which enables us to model complex forecast distributions. Furthermore, we demonstrate the effectiveness of our method in the context of the EUPPBench benchmark, where we conduct 2‐m temperature forecast post‐processing for stations in a subregion of Europe. We show that our novel method exhibits state‐of‐the‐art performance on the benchmark, improving upon other well‐performing entries. Additionally, by providing a detailed comparison of three variants of our novel post‐processing method, we elucidate the reasons why our method outperforms per‐lead‐time‐based approaches and approaches with distributional assumptions. [ABSTRACT FROM AUTHOR]

Inflow forecasts are a basic pre-requisite to support decisions for anticipating reservoir operations. The fact that hydrological forecasts are uncertain, noting that reservoirs often fulfill multiple purposes, bears the risk of non-ideal reservoir operation. Optimization techniques can help to identify the best operational scheme under given inflow. To address the forecast uncertainty, operators can repeat optimization (moving horizon approach) and optimize for inflow forecast ensembles. This article aims to contribute to a better understanding of how the different methods work, how to interpret the results and what the effect of user choices on the optimization results is. The moving-horizon approach and three ensemble optimization methods were applied on a hydropower reservoir in Norway under flood conditions. The functional principle of each method is explained, and advantages and drawbacks of the different methods are discussed with the help of performance indicators. [ABSTRACT FROM AUTHOR]

Heavy rainfall is an important weather that causes flood disasters, and it is also one of the most important natural disasters in our country. With development of the high-resolution numerical models, numerical weather prediction has been the main method for heavy rainfall forecasting. However, the accuracy of the numerical prediction depends on the completeness of the atmospheric motion equations, the accuracy of the initial state, the reasonability of the physical process, and the robustness of the calculation method. As the atmosphere is a nonlinear chaotic system, the small errors in these aspects will cause significant uncertainty in the forecast results. Therefore, the improvement of the rainstorm numerical prediction is closely related to the development of data assimilation, parameterization of physical processes, and ensemble prediction, especially for the role of parameterization schemes of cloud microphysical processes that produce precipitation in numerical models. In addition, in order to improve and perfect the numerical model, the investigation of the evaluation method of the forecast results can also not be ignored and is a crucial part of the numerical prediction. This review describes the development of several key numerical weather prediction techniques. The four-dimensional ensemble variational assimilation method, the microphysics parameterization scheme, and the stochastic kinetic energy backscatter method of ensemble prediction model perturbation are highlighted. An evaluation method of model results based on kinetic energy spectrum analysis is also proposed. Finally, the future research directions in these aspects are summarized.

Flash flooding is a significant societal problem, but related precipitation forecasts are often poor. To address, one can try to use output from convection‐parametrising (global) ensembles, post‐processed to forecast at point‐scale, or convection‐resolving limited area ensembles. The new methodology described here combines both. We apply "ecPoint‐rainfall" post‐processing to the ECMWF global ensemble. Alongside we use 2.2 km COSMO LAM ensemble output (centred on Italy), and also post‐process that, using a scale‐selective neighbourhood approach to compensate for insufficient members and to preserve consistently forecast local details. The two resulting scale‐compatible components then undergo lead‐time‐weighted blending, to create the final probabilistic 6 h rainfall forecasts. Product creation for forecasters, in this way, constituted the "Italy Flash Flood use case" within the EU‐funded MISTRAL project; real‐time delivery of open access products is ongoing. One year of verification shows that, of the five components (2 ×$$ \times $$ raw, 2 ×$$ \times $$ post‐processed and blended), ecPoint is the most skilful. The post‐processed COSMO ensemble adds most value to summer convective events in the evening, when the global model has an underprediction bias. In two typical heavy rainfall case studies we observed underestimation of the largest point totals in the raw ECMWF ensemble, and overestimation in the raw COSMO ensemble. However, ecPoint elevated the ECMWF maxima and highlighted best the most affected areas and merged products seemed to be the most skilful of all. Even though our LAM post‐processing does not include (or arguably need) bias‐correction, this study still provides a unique blueprint for successfully combining ensemble rainfall forecasts from global and LAM systems around the world. It also has important implications for forecast products as global ensembles move ever closer to having convection‐permitting resolution. [ABSTRACT FROM AUTHOR]

We propose a forecasting tool for precipitation based on analogues of circulation defined from 5‐day hindcasts and a stochastic weather generator that we call "HC–SWG." In this study, we aim to improve the forecast of European precipitation for subseasonal lead times (from 2 to 4 weeks) using the HC–SWG. We designed the HC–SWG to generate an ensemble precipitation forecast from the European Centre of Medium‐range Weather Forecasts (ECMWF) and Centre National de la Recherche Météorologique (CNRM) subseasonal‐to‐seasonal ensemble reforecasts. We define analogues from 5‐day ensemble reforecast of Z500 from the ECMWF (11 members) and CNRM (10 members) models. Then, we generate a 100‐member ensemble for precipitation over Europe. We evaluate the skill of the ensemble forecast using probabilistic skill scores such as the continuous ranked probability skill score (CRPSS) and receiver operating characteristic curve. We obtain reasonable forecast skill scores within 35 days for different locations in Europe. The CRPSS shows positive improvement with respect to climatology and persistence at the station level. The HC–SWG shows a capacity to distinguish between events and non‐events of precipitation within 15 days at the different stations. We compare the HC–SWG forecast with other precipitation forecasts to further confirm the benefits of our method. We found that the HC–SWG shows improvement against the ECMWF precipitation forecast until 25 days. [ABSTRACT FROM AUTHOR]

The scale-dependent predictability of the devastating 7·20 extreme rainstorm in Zhengzhou, China in 2021 was investigated via ensemble experiments, which were perturbed on different scales using the stochastic kinetic-energy backscatter (SKEB) scheme in the WRF model, with the innermost domain having a 3-km grid spacing. The daily rainfall (RAIN24h) and the cloudburst during 1600–1700 LST (RAIN1h) were considered. Results demonstrated that with larger perturbation scales, the ensemble spread for the rainfall maximum widens and rainfall forecasts become closer to the observations. In ensembles with mesoscale or convective-scale perturbations, RAIN1h loses predictability at scales smaller than 20 km and RAIN24h is predictable for all scales. Whereas in ensembles with synoptic-scale perturbations, the largest scale of predictability loss extends to 60 km for both RAIN1h and RAIN24h. Moreover, the average positional error in forecasting the heaviest rainfall for RAIN24h (RAIN1h) was 400 km (50–60) km. The southerly low-level jet near Zhengzhou was assumed to be directly responsible for the forecast uncertainty of RAIN1h. The rapid intensification in low-level cyclonic vorticity, mid-level divergence, and upward motion concomitant with the jet dynamically facilitated the cloudburst. Further analysis of the divergent, rotational and vertical kinetic spectra and the corresponding error spectra showed that the error kinetic energy at smaller scales grows faster than that at larger scales and saturates more quickly in all experiments. Larger-scale perturbations not only boost larger-scale error growth but are also conducive to error growth at all scales through a downscale cascade, which indicates that improving the accuracy of larger-scale flow forecast may discernibly contributes to the forecast of cloudburst intensity and position. [ABSTRACT FROM AUTHOR]

An adjoint-free four-dimensional variational (a4dVar) data assimilation (DA) is implemented in an operational ocean forecast system based on an eddy-resolving ocean general circulation model for the Northwestern Pacific. Validation of the system against independent observations demonstrates that fitting the model to time-dependent satellite altimetry during a 10-day DA window leads to substantial skill improvements in the succeeding 60-day hindcast. The a4dVar corrects representation of the Kuroshio path variation south of Japan by adjusting the dynamical balance between amplitude/wavelength of the meander and flow advection. A larger ensemble spread tends to reduce the skill in representing the observed sea surface height anomaly, suggesting that it is possible to use the ensemble information for quantifying the forecast error. The ensemble information is also utilized for modification of the background error covariance (BEC), which improves the accuracy of temperature and salinity distributions. The modified BEC yields the skill decline of the Kuroshio path variation during the 60-day hindcast period, and the ensemble sensitivity analysis shows that changes in the dynamical balance caused by the ensemble BEC result in such skill deterioration. [ABSTRACT FROM AUTHOR]

Accurate and stable prediction of atmospheric PM2.5 concentrations is crucial for air pollution prevention and control. Existing studies usually rely on a single model or use a single evaluation criterion in multi-model ensemble weighted forecasts, neglecting the dual needs for accuracy and stability in PM2.5 forecast. In this study, a novel ensemble forecast model is proposed that overcomes these drawbacks by simultaneously taking into account both forecast accuracy and stability. Specifically, four advanced deep learning models—Long Short-Term Memory Network (LSTM), Graph Convolutional Network (GCN), Transformer, and Graph Sample and Aggregation Network (GraphSAGE)—are firstly introduced. And then, two combined models are constructed as predictors, namely LSTM–GCN and Transformer–GraphSAGE. Finally, a combined weighting strategy is adopted to assign weights to these two combined models using a multi-objective optimization algorithm (MOO), so as to carry out more accurate and stable predictions. The experiments are conducted on the dataset from 36 air quality monitoring stations in Beijing, and results show that the proposed model achieves more accurate and stable predictions than other benchmark models. It is hoped that this proposed ensemble forecast model will provide effective support for PM2.5 pollution forecast and early warning in the future. [ABSTRACT FROM AUTHOR]

Since the Beijing 2022 Winter Olympics was the first Winter Olympics in history held in continental winter monsoon climate conditions across complex terrain areas, there is a deficiency of relevant research, operational techniques, and experience. This made providing meteorological services for this event particularly challenging. The China Meteorological Administration (CMA) Earth System Modeling and Prediction Centre, achieved breakthroughs in research on short- and medium-term deterministic and ensemble numerical predictions. Several key technologies crucial for precise winter weather services during the Winter Olympics were developed. A comprehensive framework, known as the Operational System for High-Precision Weather Forecasting for the Winter Olympics, was established. Some of these advancements represent the highest level of capabilities currently available in China. The meteorological service provided to the Beijing 2022 Games also exceeded previous Winter Olympic Games in both variety and quality. This included achievements such as the "100-meter level, minute level" downscaled spatiotemporal resolution and forecasts spanning 1 to 15 days. Around 30 new technologies and over 60 kinds of products that align with the requirements of the Winter Olympics Organizing Committee were developed, and many of these techniques have since been integrated into the CMA's operational national forecasting systems. These accomplishments were facilitated by a dedicated weather forecasting and research initiative, in conjunction with the preexisting real-time operational forecasting systems of the CMA. This program represents one of the five subprograms of the WMO's high-impact weather forecasting demonstration project (SMART2022), and is also a part of their Regional Association (RA) II Research Development Project (Hangzhou RDP). Therefore, the research accomplishments and meteorological service experiences from this program will be carried forward into forthcoming high-impact weather forecasting activities. This article provides an overview and assessment of this program and the operational national forecasting systems. [ABSTRACT FROM AUTHOR]

To improve the accuracy of short-term wind speed and power forecasts as well as reduce the impact of wind power uncertainty on the grid system, this study attempted to use wind speed observations to select the optimal numerical forecasting ensemble members that closely matched the actual wind speed within the forecast window. Those selected ensemble members then formed an optimized ensemble for training and testing machine learning models. Unlike the conventional method that relied solely on ensemble averaging, this method considered the forecast discrepancies among different ensemble members, avoided the introduction of members with large errors, and thus helped to improve wind speed prediction. The optimal number of ensemble members for wind farms with different altitudes and terrains in Henan and Gansu was determined based on the results of ensemble performance and sensitivity experiments. Comparative analyses demonstrated that the ensemble selection forecasts outperformed ensemble averaging in predicting wind speed fluctuations during different weather processes, closely aligning with actual wind speed observations. The sea-level pressure field estimates generated by the ensemble selection method exhibited a higher level of agreement with ERA5 data. Evaluation of wind speed and power prediction over eleven consecutive months in different wind farms showed that the ensemble selection method improved the accuracy of diurnal variation and monthly average wind speed compared to the original ensemble averaging method. Analysis of the observed data of upslope and downslope winds with different durations and speed changes in two wind farms showed that there were the most wind speed changes of 2 -4 m s-1 within 0-2 h and 2-4 h. Compared to ensemble averaging, the ensemble selection method significantly improved the prediction accuracy for these upslope and downslope winds. Furthermore, using the machine learning algorithm to train the optimal selection ensemble can further reduce the absolute deviation and root mean square error of wind speed, thereby effectively improving the accuracy of power prediction. [ABSTRACT FROM AUTHOR]

Limitations are existed in current ensemble forecasting initial perturbation methods for describing the interactions among various spheres of the Earth system. In this study, a new method is proposed, namely, the coupled conditional nonlinear optimal perturbation (C-CNOP) method, which incorporates multisphere interactions much appropriately. The El Niño-Southern Oscillation (ENSO) is a typical ocean-atmosphere "coupling" (or "interaction") phenomenon. The C-CNOP method is applied to ensemble forecasting of ENSO. It is demonstrated that the C-CNOP method can generate coupled initial perturbations (CPs) that appropriately consider initial ocean-atmosphere coupling uncertainty for ENSO ensemble forecasts. Results reveal that the CPs effectively improve the ability of ENSO ensemble-mean forecasts in both temporal variability of Niño3.4 sea surface temperature anomalies (SSTAs) and spatial variability of ENSO mature-phase SSTAs. Notably, despite the weakest ocean-atmosphere coupling strength in the tropical Pacific occurring during the boreal spring and summer, CPs still capture the uncertainties of this weak coupling when ENSO predictions are initialized at these seasons. This performance of CPs significantly suppresses the rapid increase of ENSO prediction errors due to the high ocean-atmosphere coupling instability during these seasons, and thus effectively extends the lead time of skillful ENSO forecasting. Hence, the C-CNOP method is a suitable initial perturbation approach for ENSO ensemble forecast that can describe initial ocean-atmosphere coupling uncertainty. It is expected that the C-CNOP method plays a significant role in predictions of other high-impact climate phenomena, and even future Earth system predictions. [ABSTRACT FROM AUTHOR]

The nonlinearity nature of land subsidence and limited observations cause premature convergence in typical data assimilation methods, leading to both underestimation and miscalculation of uncertainty in model parameters and prediction. This study focuses on a promising approach, the combination of evolutionary-based data assimilation (EDA) and ensemble model output statistics (EMOS), to investigate its performance in land subsidence modeling using EDA with a smoothing approach for parameter uncertainty quantification and EMOS for predictive uncertainty quantification. The methodology was tested on a one-dimensional subsidence model in Kawajima (Japan). The results confirmed the EDA's robust capability: Model diversity was maintained even after 1000 assimilation cycles on the same dataset, and the obtained parameter distributions were consistent with the soil types. The ensemble predictions were converted to Gaussian predictions with EMOS using past observations statistically. The Gaussian predictions outperformed the ensemble predictions in predictive performance because EMOS compensated for the over/under-dispersive prediction spread and the short-term bias, a potential weakness for the smoothing approach. This case study demonstrates that combining EDA and EMOS contributes to groundwater management for land subsidence control, considering both the model parameter uncertainty and the predictive uncertainty. [ABSTRACT FROM AUTHOR]

Short-term probabilistic forecasts of the trajectory of the COVID-19 pandemic in the United States have served as a visible and important communication channel between the scientific modeling community and both the general public and decision-makers. Forecasting models provide specific, quantitative, and evaluable predictions that inform short-term decisions such as healthcare staffing needs, school closures, and allocation of medical supplies. Starting in April 2020, the US COVID-19 Forecast Hub (https://covid19forecasthub.org/) collected, disseminated, and synthesized tens of millions of specific predictions from more than 90 different academic, industry, and independent research groups. A multimodel ensemble forecast that combined predictions from dozens of groups every week provided the most consistently accurate probabilistic forecasts of incident deaths due to COVID-19 at the state and national level from April 2020 through October 2021. The performance of 27 individual models that submitted complete forecasts of COVID-19 deaths consistently throughout this year showed high variability in forecast skill across time, geospatial units, and forecast horizons. Two-thirds of the models evaluated showed better accuracy than a naïve baseline model. Forecast accuracy degraded as models made predictions further into the future, with probabilistic error at a 20-wk horizon three to five times larger than when predicting at a 1-wk horizon. This project underscores the role that collaboration and active coordination between governmental public-health agencies, academic modeling teams, and industry partners can play in developing modern modeling capabilities to support local, state, and federal response to outbreaks.

Numerical forecast of the sea fog is sensitive to the initial moist stratification within the marine atmospheric boundary layer (MABL). This study develops an online assimilation method to improve the MABL thermal and moist structures in sea fog ensemble forecasts based on the Weather Research and Forecasting model and Grid‐point Statistical Interpolation/EnKF system. It uses the satellite‐retrieved cloud water path (CWP) as the indicator of sea fog and low‐level stratus to determine the best ensemble members at each grid point. The relative humidity and cloud water profiles are extracted from the best members to generate a series of pseudo‐observations, which are assimilated to update all the members by EnKF method. The new method significantly improves the ensemble forecast of five widespread advection fog events over the Yellow Sea, which can be attributed to the decrease of both missed and spurious fog areas. The case study shows that assimilating the information of both humidity and cloud water outperforms assimilating either of them, while the impact of directly assimilating CWP observation is insignificant. The analysis increments of cloud water, thermal and moist structures in MABL together contribute to the correction of forecasted sea fog. The generation of pseudo‐observations can use the dynamic compatibility of the model to alleviate the impact of erroneous data in the observation, leading to the low sensitivity of the new method to CWP retrieval error. Plain Language Summary: Sea fog results in great casualties and property losses. Its numerical forecast is quite challenging, and it is important to create an initial condition with accurate moist stratification. In this work, we develop a method to improve the initial condition of ensemble forecast using satellite retrieved cloud water content based on Weather Research and Forecasting model and GSI/EnKF assimilation system. At each grid point, it uses the satellite‐retrieved cloud water content to determine the best ensemble members. The relative humidity and cloud water profiles are extracted from the best members as the pseudo‐observation, and are used to update all the members by EnKF assimilation. This method has 3 advantages: (a) It only modifies thermal and moist structures within planetary boundary layer without disturbing the free atmosphere; (b) The data volume and computation cost are small; (c) it is not very sensitive to the retrieval error. It significantly improves the ensemble forecast of fog areas. Key Points: This study develops a new assimilation method to improve the nowcasting of sea fog using the satellite‐retrieved cloud water path (CWP)The new method generates and assimilates the 3‐D pseudo‐observation based on both ensemble members and CWP dataIt decreases the model bias in the near‐surface moisture and significantly improves the prediction skill of fog areas [ABSTRACT FROM AUTHOR]

It is well-known in rainfall ensemble forecasts that ensemble means suffer substantially from the diffusion effect resulting from the averaging operator. Therefore, ensemble means are rarely used in practice. The use of the arithmetic average to compute ensemble means is equivalent to the definition of ensemble means as centers of mass or barycenters of all ensemble members where each ensemble member is considered as a point in a high-dimensional Euclidean space. This study uses the limitation of ensemble means as evidence to support the viewpoint that the geometry of rainfall distributions is not the familiar Euclidean space, but a different space. The rigorously mathematical theory underlying this space has already been developed in the theory of optimal transport (OT) with various applications in data science. In the theory of OT, all distributions are required to have the same total mass. This requirement is rarely satisfied in rainfall ensemble forecasts. We, therefore, develop the geometry of rainfall distributions from an extension of OT called unbalanced OT. This geometry is associated with the Gaussian-Hellinger (GH) distance, defined as the optimal cost to push a source distribution to a destination distribution with penalties on the mass discrepancy between mass transportation and original mass distributions. Applications of the new geometry of rainfall distributions in practice are enabled by the fast and scalable Sinkhorn-Knopp algorithms, in which GH distances or GH barycenters can be approximated in real-time. In the new geometry, ensemble means are identified with GH barycenters, and the diffusion effect, as in the case of arithmetic means, is avoided. New ensemble means being placed side-by-side with deterministic forecasts provide useful information for forecasters in decision-making. [ABSTRACT FROM AUTHOR]

Quantiles of precipitation are widely used in ensemble forecast systems. At present, the common practice is to provide precipitation amounts corresponding to different quantiles to users directly, which will make it difficult for users to extract reliable forecast information. Therefore, this study investigates the statistically optimal (using threat score (TS) as a metric) quantiles of precipitation in an ensemble forecast system constructed using the WRF V4.0 model. The main conclusions are as follows: The threat‐score‐optimal quantiles for light rain, moderate rain, heavy rain, rainstorm, and heavy rainstorm forecasts are 40%–60%, 60%–70%, 60%–80%, 70%–80%, and 80%, respectively. Overall, the optimal quantile increases with the rise in precipitation magnitude or the extension of forecast lead time. All the optimal quantile forecast products have higher TS than the corresponding control forecast, ensemble mean forecast, and probability‐matched ensemble mean forecast products. The merged threat‐score‐optimal quantile forecast product formed by combining the optimal quantile forecasts of different precipitation magnitudes shows obvious advantages over other products in statistical verification and case studies, and it shows good potential to be operationally implemented in the future. [ABSTRACT FROM AUTHOR]

Wildfires emit vast amounts of aerosols and trace gases into the atmosphere, exerting myriad effects on air quality, climate, and human health. Ensemble forecasting has been proposed to reduce the large uncertainties in the wildfire air pollution forecast. This study presents the development of a multi‐model ensemble (MME) wildfire air pollution forecast over North America. The ensemble members include regional models (GMU‐CMAQ, NACC‐CMAQ, and HYSPLIT), global models (GEFS‐Aerosols, GEOS5, and NAAPS), and global ensemble (ICAP‐MME). Performance of the ensemble forecast was evaluated with MAIAC and VIIRS‐SNPP retrieved aerosol optical depth (AOD) and AirNow surface PM2.5 measurements during the 2020 Western United States "Gigafire" events (August–September 2020). Compared to individual models, the ensemble mean significantly reduced the biases and produced more consistent and reliable forecasts during extreme fire events. For AOD forecasts, the ensemble mean was able to improve model performance, such as increasing the correlation to 0.62 from 0.33 to 0.57 by individual models compared to VIIRS AOD. The ensemble mean also yields the best overall RANK (a composite indicator of four statistical metrics) when compared to VIIRS and MAIAC AOD. For the surface PM2.5 forecast, the ensemble mean outperformed individual models with the strongest correlation (0.60 vs. 0.43–0.54 by individual models), lowest fractional bias (0.54 vs. 0.55–1.32), highest hit rate (87% vs. 40%–82%), and highest RANK (2.83 vs. 2.40–2.81). Finally, the ensemble shows the potential to provide a probability forecast of air quality exceedances. The exceedance probability forecast can be further applied to early warnings of extreme air pollution episodes during large wildfire events. Plain Language Summary: Wildfires are a major source of air pollution emitting large quantities of particles into the air that adversely affect human health. Predicting wildfire air pollution, however, is challenging. Ensemble forecasting has been proposed to improve the model predictability. We developed a new multi‐model ensemble forecast system of wildfire air pollution over North America, leveraging regional and global atmospheric models by federal agencies and academia. How well the ensemble forecast can predict wildfire pollution was evaluated with observations from satellites and ground monitors. We found that the ensemble mean can significantly reduce the forecast biases and produce more reliable forecasts during extreme wildfire fire events. The ensemble probability forecast of exceedance of the health‐based National Ambient Air Quality Standards for fine particles (PM2.5) can be further applied to early warnings of severe air pollution episodes during large wildfire events. These findings highlight the potential of the ensemble approach to improve the predictability of air pollution during large wildfires. Key Points: Ensemble mean of models provides reliable aerosol optical depth and PM2.5 forecasts with stronger correlation, lower bias, and the highest overall RANKsThe ensemble approach provides a well suited exceedance probability forecast during the 2020 Gigafire eventsThe multi‐model ensemble approach can reduce uncertainties in air quality forecasts and improve model predictability during wildfire events [ABSTRACT FROM AUTHOR]

Taiwan is highly susceptible to global warming, experiencing a 1.4 °C increase in air temperature from 1911 to 2005, which is twice the average for the Northern Hemisphere. This has potentially led to higher rates of respiratory and cardiovascular mortality. Accurately predicting maximum temperatures during the summer season is crucial, but numerical weather models become less accurate and more uncertain beyond five days. To enhance the reliability of a forecast, post-processing techniques are essential for addressing systematic errors. In September 2020, the NOAA NCEP implemented the Global Ensemble Forecast System version 12 (GEFSv12) to help manage climate risks. This study developed a Hybrid statistical post-processing method that combines Artificial Neural Networks (ANN) and quantile mapping (QQ) approaches to predict daily maximum temperatures (Tmax) and their extremes in Taiwan during the summer season. The Hybrid technique, utilizing deep learning techniques, was applied to the GEFSv12 reforecast data and evaluated against ERA5 reanalysis. The Hybrid technique was the most effective among the three techniques tested. It had the lowest bias and RMSE and the highest correlation coefficient and Index of Agreement. It successfully reduced the warm bias and overestimation of Tmax extreme days. This led to improved prediction skills for all forecast lead times. Compared to ANN and QQ, the Hybrid method proved to be more effective in predicting daily Tmax, including extreme Tmax during summer, on extended-range time-scale deterministic and ensemble probabilistic forecasts over Taiwan. [ABSTRACT FROM AUTHOR]

To comprehensively understand the application progress of ensemble forecast technology in influenza forecast based on infectious disease model, so as to provide scientific references for further research. In this study, two keywords of “influenza” and “ensemble forecast” are selected to search and select the relevant literatures, which are then outlined and summarized. It is found that: In recent years, some studies about ensemble forecast technology for influenza have been reported in the literature, and some well-performed influenza ensemble forecast systems have already been operationally implemented and provide references for scientific prevention and control. In general, ensemble forecast can well represent various uncertainties in forecasting influenza cases based on infectious disease models, and can achieve more accurate forecasts and more valuable information than single deterministic forecast. However, there are still some shortcomings in the current studies, it is suggested that scientists engaged in influenza forecast based on infectious disease models strengthen cooperation with scholars in the field of numerical weather forecast, which is expected to further improve the skills and application level of ensemble forecast for influenza.

Solar resource forecasting is an essential step towards smart management of power grids. This study aims to increase the performance of intra-hour forecasts. For this, a novel ensemble model, combining statistical extrapolation of time-series measurements with models based on machine learning and all-sky imagery, is proposed. This study is conducted with high-quality data and high-resolution sky images recorded on the Solar Platform of the West University of Timisoara, Romania. Atmospheric factors that contribute to improving or reducing the quality of forecasts are discussed. Generally, the statistical models gain a small skill score across all forecast horizons (5 to 30 min). The machine-learning-based methods perform best at smaller forecast horizons (less than 15 min), while the all-sky-imagery-based model performs best at larger forecast horizons. Overall, for forecast horizons between 10 and 30 min, the weighted forecast ensemble with frozen coefficients achieves a skill score between 15 and 20%. [ABSTRACT FROM AUTHOR]

We use short‐range ensemble forecasts and ensemble clustering analysis to study the factors affecting the intensification of Hurricane Patricia (2015). Convection‐permitting ensemble forecasts are classified into two groups: 10 spin‐down (SPD) members and 10 spin‐up (SPU) members with intensification rates of <0 and >0 m s−1 for the first 6 hr, respectively. Ensemble clustering analysis found that the wind–pressure relationship was incorrect in the SPD group, indicating that the SPD issue may be partly caused by an imbalance in the initial minimum sea‐level pressure (MSLP) and the maximum wind speed (MWS). The SPD issue appears to be related to three main points: (a) a weaker upper‐level warm core; (b) a drier inner core at low levels; and (c) a dry, cold air intrusion at mid‐levels. In contrast, the SPU group has a stronger upper‐level warm core and a relatively wet inner core at lower levels, as well as a relatively strong secondary circulation. These favorable initial conditions in the SPU group, combined with a greater updraft and stronger convection around the eyewall, result in more latent heating around or in the eyewall that favors the intensification of the tropical cyclone. Comparisons between the SPD and SPU groups suggest that the analyzed ensemble could not accurately capture the relationship between the initial MSLP and MWS, which, combined with the unfavorable thermodynamic conditions at the initial time, resulted in the incorrect evolution of the intensity. Therefore, improving the initial conditions appears to be an effective way to address the SPD issue. Plain Language Summary: Intensity spin‐down (SPD) has been identified as a major issue in numerical hurricane models because maximum wind speed can decrease significantly and artificially in the first few hours of a simulation, thereby degrading the remainder of the intensity forecast. Ensemble clustering analysis was performed to explore the role of the thermodynamic structure of the inner core in predicting the intensification of Hurricane Patricia (2015). Convection‐permitting ensemble forecasts are classified into SPD and spin‐up (SPU) groups based on the intensification rates for the first 6 hr. Several statistically significant differences between the SPD and SPU members were found in the wind–pressure relationship, the upper‐level warm core, moisture in the inner core, and secondary circulation. Particularly, the SPD issue may be partly caused by an imbalance in the initial wind–pressure relationship. Moreover, it also appears to be related to three main points: a weaker upper‐level warm core; a drier inner core at low levels; and a dry, cold air intrusion at mid‐levels. However, the intensification of the hurricane was sensitive to the upper‐level warm core, moisture in the inner core, and secondary circulation. Overall, thermodynamic structure of the inner core plays a key role in predicting the intensification of Hurricane Patricia (2015). Key Points: The spin‐down issue may be partly caused by an imbalance in the initial wind–pressure relationshipIntensification of the hurricane was sensitive to the upper‐level warm core, moisture in the inner core, and secondary circulationThe thermodynamic structure of the inner core plays a key role in predicting the intensification of Hurricane Patricia (2015) [ABSTRACT FROM AUTHOR]

Skilful and localised daily weather forecasts for upcoming seasons are desired by climate-sensitive sectors. Various General circulation models routinely provide such long lead time ensemble forecasts, also known as seasonal climate forecasts (SCF), but require downscaling techniques to enhance their skills from historical observations. Traditional downscaling techniques, like quantile mapping (QM), learn empirical relationships from pre-engineered predictors. Deep-learning-based downscaling techniques automatically generate and select predictors but almost all of them focus on simplified situations where low-resolution images match well with high-resolution ones, which is not the case in ensemble forecasts. To downscale ensemble rainfall forecasts, we take a two-step procedure. We first choose a suitable deep learning model, very deep super-resolution (VDSR), from several outstanding candidates, based on an ensemble forecast skill metric, continuous ranked probability score (CRPS). Secondly, via incorporating other climate variables as extra input, we develop and finalise a very deep statistical downscaling (VDSD) model based on CRPS. Both VDSR and VDSD are tested on downscaling 60 km rainfall forecasts from the Australian Community Climate and Earth-System Simulator Seasonal model version 1 (ACCESS-S1) to 12 km with lead times up to 217 days. Leave-one-year-out testing results illustrate that VDSD has normally higher forecast accuracy and skill, measured by mean absolute error and CRPS respectively, than VDSR and QM. VDSD substantially improves ACCESS-S1 raw forecasts but does not always outperform climatology, a benchmark for SCFs. Many more research efforts are required on downscaling and climate modelling for skilful SCFs. [ABSTRACT FROM AUTHOR]

Soil moisture (SM) affects weather through its impact on surface flux partitioning, influencing vertical atmospheric profiles and circulations driven by differential surface heating. In West Africa, observational studies point to a dominant negative SM‐precipitation feedback, where dry soils help to initiate and maintain convection. In this context, serious concerns exist about the ability of models with parameterised convection to simulate this observed sensitivity of daytime convection to SM. Here, we evaluate the effect of initial SM perturbations in a short‐range ensemble forecast over West Africa, comparing the UK Met Office Global and Regional Ensemble Prediction System (MOGREPS) with parameterised convection (GLOB‐ENS) to its regional convection‐permitting counterpart (CP‐ENS). Results from both models suggest SM perturbations introduce considerable spread into daytime evaporative fraction (EF) and near‐surface temperatures. This spread is still evident on Day 3 of the forecast. Both models also show a tendency to increased afternoon rainfall frequency over negative EF anomalies, reproducing the observed feedback. However, this effect is more pronounced in CP‐ENS than GLOB‐ENS, which illustrates the potential for process‐based forecast improvements at convection‐permitting scales. Finally, we identify persistent biases in rainfall caused by land cover mapping issues in the operational GLOB‐ENS setup, emphasising the need for careful evaluation of different mapping strategies for land cover. [ABSTRACT FROM AUTHOR]

Warm-sector heavy rainfall events over southern China are difficult to accurately forecast, due in part to inaccurate initial fields in numerical weather prediction models. In order to determine an efficient way of reducing the critical initial field errors, this study conducts and compares two sets of 60-member ensemble forecast experiments of a warm-sector heavy rainfall event over coastal southern China without data assimilation (NODA) and with radar radial velocity data assimilation (RadarDA). Yangjiang radar data, which can provide offshore high-resolution wind field information, were assimilated by using a Weather Research and Forecasting (WRF)-based ensemble Kalman filter (EnKF) system. The results show that the speed and direction errors of the southeasterly airflow in the marine boundary layer over the northern South China Sea may primarily be responsible for the forecast errors in rainfall and convection evolution. Targeted assimilation of radial velocity data from the Yangjiang radar can reduce the critical initial field errors of most members, resulting in improvements to the ensemble forecast. Specifically, RadarDA simulations indicate that radial-velocity data assimilation (VrDA) can directly reduce the initial field errors in wind speed and direction, and indirectly and slightly adjust the initial moisture fields in most members, thereby improving the evolution features of moisture transport during the subsequent forecast period. Therefore, these RadarDA members can better capture the initiation and development of convection and have higher forecast skill for the convection evolution and rainfall. The improvement in the deterministic forecasts of most members results in an improved overall ensemble forecast performance. However, VrDA sometimes results in inappropriate adjustment of the initial wind field, so the forecast skill of a few members decreases rather than increases after VrDA. This suggests that a degree of uncertainty remains about the effect of the WRF-based EnKF system. Moreover, the results further indicate that accurate forecasts of the convection evolution and rainfall of warm-sector heavy rainfall events over southern China are challenging. [ABSTRACT FROM AUTHOR]

In order to understand more comprehensively the ensemble forecast results of rainfall with the convective scale ensemble prediction system and thus to further recommend them to the weather forecasters, this study carried out the analysis of the forecast performance of a rainstorm process with a convective scale ensemble prediction system. The results show that: (1) The forecast difference of each ensemble member increases with precipitation magnitude, and the threat score difference between the best and worst performed ensemble member is more than 0.3. (2) Probability-matched mean forecast performs better than control forecast and ensemble mean forecast for both rainstorm and heavy rainstorm precipitation. Ensemble mean is insensitive to extreme precipitation due to the smoothing effect of ensemble member forecast. Therefore, ensemble mean is not suitable for extreme precipitation forecast. (3) From the minimum forecast to the maximum forecast, with the increase of ensemble percentile, the probability of detection, false alarm rate, and frequency bias gradually increase. The forecast at 70% or 80% ensemble percentile performs the best, and it is better than the ensemble mean and probability-matched mean forecast. (4) For the heavy rainstorm precipitation in the west part of northeastern Chongqing, the long-time ensemble probability forecasts with leading-times up to 60 h all successfully predict certain precipitation probability of rainstorm, and the forecasted precipitation from the corresponding best performed ensemble member is close to the observation.

In order to provide more powerful support for forecasters in Sichuan and Chongqing with complicated terrain to carry out short-term (0-12 h) precipitation forecast, a convective scale ensemble forecast operational system is designed based on ensemble Kalman filter data assimilation method (31 ensemble samples) and WRF model with 3 km resolution (model domain: 24.5°-34.5°N, 99°-113°E) and lead time of 12 h, which is started by 3 h cycle. It is urgent to decide how many members should be used for the 12 h ensemble forecast to achieve the most representative probability distribution and optimal ensemble forecast skills. An ensemble forecast experiment is carried out on 16 heavy convective scale precipitation cases occurred in Sichuan and Chongqing with different amount of ensemble members, and the results are analyzed comprehensively. It is concluded that the precipitation forecast skills of the ensemble members for different magnitude of precipitation are roughly the same, so there is little difference in the totally averaged prediction skills of different ensemble size. Talagrand distribution becomes better with the increase of ensemble size first. However, when the ensemble size is larger than 17, the improvement by increasing ensemble size is no longer significant. Meanwhile, the forecast error probability becomes smaller with the increase of ensemble size first, but when the ensemble size reaches 16 to 18, the difference between the forecast error probability and the ideal value tends to be stable, indicating that the improvement by further increasing the ensemble size is no longer significant. The relative area of operational characteristic (AROC) score which represents the prediction probability forecast skills, improves gradually with the increase of ensemble size. However, when the ensemble size is large enough, the improvement by lager ensemble size is no longer significant and the AROC scores tend to be stable. The ensemble size required for stable AROC score increases with the magnitude of precipitation. Overall, when the AROC scores become stable, the ensemble size required for light rain, moderate rain, heavy rain, rainstorm (and heavy rainstorm) are 10, 14, 16 and 18, respectively. Based on the comparative analysis results and considering that there is generally little difference in forecasting skills when the number of members is different by 2, in order to achieve the most representative probability distribution and optimal ensemble forecast skills of precipitation, it is recommended to set the ensemble size of convective scale ensemble prediction system from 16 to 18.

Accurate prediction of photovoltaic (PV) power is the prerequisite for the safe and stable operation of the power grid with high penetration of PV. Despite various machine learning models for forecasting PV power have been developed, their accuracies are generally unstable. Toward this end, this study proposes a novel Stacking ensemble forecast model to improve the precision of day-ahead PV power forecasts. Different from the traditional Stacking model that uses the original training dataset to train the base learners, the proposed model creates multiple sub-training sets from the original training dataset to train the base learners, so as to enhance the diversity of base models and further improve the prediction accuracy. Specifically, in the proposed Stacking ensemble model, four machine learning learners, i.e., generalized regression neural network (GRNN), extreme learning machine (ELM), Elman neural network (ElmanNN), and Long shot-term memory (LSTM) neural network are incorporated, which are trained with the diverse sub-training datasets, and a variety of candidate base models are generated. For those candidate base models, the ones with the best performance are selected and integrated through a meta-model, namely the back-propagation network work (BPNN), to produce the final PV power prediction. The proposed model is evaluated using measured data from a 15kW PV power station in Ashland, Oregon, USA. Results indicate that across three weather scenarios, the performance of the novel Stacking ensemble model consistently outperforms single models and the traditional Stacking ensemble model in terms of the errors for out-of-sample forecasting, which proves the effectiveness of the developed procedure in improving PV power forecasting accuracy.

In this study, the assimilation of tide gauge and altimetry data into a two‐dimensional barotropic numerical model for the southwestern Atlantic continental shelf (SWACS) was developed. To do this, the preoperative 4‐day storm surges ensemble prediction system developed by Dinápoli et al. (2021, Journal of the Royal Meteorological Society147: 557–572) was implemented for the SWACS. This new configuration, called "Model for Storm Surge Simulations" (M3S), considers a curvilinear grid that covers the SWACS with higher resolution along the shoreline (from 2 to 10 km). M3S was forced with an ensemble of 60 members conformed by the combination of perturbations of the eight principal tidal constituents and of the atmospheric products derived from the Global Ensemble Forecast System. Tidal gauge and altimetry data were assimilated in an asynchronous mode using the four‐dimensional‐ensemble square‐root filter (4DEnSRF). The system was developed and validated forecasting two strong positive storm surges. Results show that 4DEnSRF's innovations produce a positive impact upon the forecast skill up to 2 days. Hence, the 4‐day forecast can be divided into two parts: the first 2 days with a stronger dependence on the initial conditions and the other 2 days purely driven by external forcing. It was found that a symmetric assimilation window of 12 hr length produces the best initial condition. Under this configuration, 4DEnSRF removes biases and improves the timing of the M3S forecasted solutions. The largest improvements were observed at the northern SWACS, where more chaotic processes, such as the atmospheric circulation, explain a large part of the sea‐surface height variability. No significant improvements were found at the southern SWACS, which can be attributed to the strong tidal dynamics that characterise the zone. Our results show that the incorporation of 4DEnSRF into M3S can significantly improve the forecast in the SWACS and also the accuracy of the short‐range detection of storm surges. [ABSTRACT FROM AUTHOR]

Tropical cyclone (TC) track forecasting has been considerably improved in recent decades, while TC intensity forecasting remain challenging. In this study, orthogonal nonlinear forcing singular vectors (O‐NFSVs) for emulating the impact of model uncertainties are used to conduct TC ensemble forecasting experiments with the Weather Research and Forecasting (WRF) model, with a focus on improving TC intensity forecasting skill. The O‐NFSVs approach is compared with the traditional stochastic kinetic‐energy backscatter (SKEB) and stochastically perturbed parametrization tendency (SPPT) schemes. The results demonstrate that the O‐NFSVs ensembles generally provide a better representation of the model uncertainties affecting TC intensification, with much better deterministic and probabilistic skills. These results also extend to the ability to forecast TC track, although the perturbations have not been optimized for that specific purpose. The O‐NFSVs are therefore appropriate perturbation structures for describing the uncertainties of the TC intensity and track forecasting and are also favourable for recognizing the rapid intensification process. [ABSTRACT FROM AUTHOR]

In early February 2022, the Sapporo region, north Japan, experienced record‐breaking snowfall. During this period, a mesoscale cyclonic rotation corresponding to a polar mesocyclone (PMC) was observed over the Sea of Japan by a geostationary satellite. This study investigates important processes involved in this heavy snowfall event with attention to the role of the PMC using a 100‐member regional ensemble forecast. The initial conditions for the ensemble forecast are obtained by a regional‐scale ensemble data assimilation system that assimilates conventional observations only. Several ensemble members successfully predict heavy snowfall in the Sapporo region with an 18 hr lead time. Composite analyses reveal that these better ensemble members predict that a southward‐moving PMC with a precipitation system develops from small vortices over the ocean and approaches to the Sapporo region, whereas worse members instead predict precipitation systems shifted to the north of the Sapporo region. Further composite analyses demonstrate that the better ensemble members have environmental characteristics such as a deeper and colder upper‐level trough and large low‐level baroclinicity, which are favorable for the development of PMCs. In particular, the latter is associated with reverse shear in which PMCs tend to move southward. Therefore, these environmental factors are key to predicting the heavy snowfall event in Sapporo associated with the PMC. Plain Language Summary: Heavy snowfalls have profound societal impacts; thus, their accurate prediction is important. In early February 2022, the Sapporo region (with approximately 2 million residents), north Japan, experienced record‐breaking snowfall. This study investigates important processes during this snowfall event using a series of forecasts with small differences in the initial conditions. By comparing better and worse forecasts in terms of the amount of precipitation in the Sapporo region, this study indicates that a small rotating disturbance termed a polar mesocyclone (PMC) and an associated precipitation system resulted in heavy snowfall in the Sapporo region. The better members are characterized by environmental factors that are favorable for the development of PMCs. These factors are key to accurately predicting the heavy snowfall event in Sapporo. Key Points: Record‐breaking snowfall in Sapporo, Japan, in early February 2022 is investigated using a regional ensemble forecastComposite analyses indicate that a polar mesocyclone (PMC) and an associated precipitation system caused heavy snowfallEnsemble members with better forecasting accuracy denote environmental characteristics favorable for the development of PMCs [ABSTRACT FROM AUTHOR]

An extremely heavy rainfall with dual rainbands occurred in the Fujian province on 22 May 2014, causing severe disasters in Fujian. In order to investigate the key forecasting factors and the predictability of this case, the evaluation and sensitivity analysis of precipitation forecasts by the ensemble prediction system which is based on the European Centre for Medium-Range Weather Forecasts (ECMWF) global Model are carried out. The result show that the ECMWF-EPS have better ability to capture the intensity and spatial distribution of the northern rainband, but significantly underestimated the precipitation in the warm area. Through ensemble forecast sensitivity analysis and comparison between good and poor members, the main factors causing forecast deviations in the two rainbands and affecting the predictability of heavy rainfall were revealed. The forecast of rainfall distribution and intensity in the northern rainband was highly sensitive to the predictability of the weather-scale shear line. The westward bias in the forecasted position and the weakened intensity of the shear line were the main causes of the westward and weaker forecast of heavy precipitation in the northern region. Additionally, the forecast of 850 hPa low-level jets, especially the forecasted intensity of zonal winds, which were closely related to the shear line, significantly influenced the intensity forecast of precipitation in the northern region. The forecast of precipitation in the southern warm area, which was far from the shear line, was more sensitive to the wind speed and thermodynamic conditions of the southwestern airflow in the boundary layer of the upstream inflow region. Most ensemble forecast members underestimated the intensity of the southwestern airflow in the coastal boundary layer of South China, which was the main reason for the near omission of heavy rainfall in the southern warm area. This study provides a quantitative correlation between low-level southwest jets along the coast of Guangdong and Fujian and non-typhoon heavy rainfall in Fujian, and explores their impact on heavy precipitation forecasts.

Study region: The Fuhe River Basin in Jiangxi Province, China. Study focus: Global climate change and intensified human activities are making the hydrological processes at Fuhe River Basin experiencing dramatic changes. Although some studies have investigated their individual impacts on basin-scale water resources, their combined effects on hydrology have received little attention. In this study, future scenarios were constructed for three future periods, based on five global climate model outputs (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and a dataset of future land use projections under three shared socioeconomic pathways and representative concentration pathways (SSP-RCPs). Then, the Soil and Water Assessment Tool (SWAT) model was used to assess the relative changes in water balance components and extreme flow frequency under these developed scenarios. Furthermore, the hydrological response assessment methodology was improved from the original multiscenario ensemble flow forecast (MESF) framework, which not only strengthens the connection between climate and land use input changes but also adds more assessment items. New hydrological insights for the region: The flow at the outlet of Fuhe River Basin is expected to increase by approximately 27.1%− 30.2%, 24.7–39.0% and 35.5%− 43.5% in the 2030 s, 2060 s and 2090 s, respectively. Water availability will increase significantly in February, August and October and decrease in November and December. To the end of 21st century, surface runoff will have more than 100% increase. Future floods and droughts will be more frequent and severe under SSP5–8.5.

We introduce a computational framework to forecast cloud index (CI)fields for up to one hour on a spatial domain that covers a city. Such intra-hour CI forecasts are important to produce solar power forecasts of utility scale solar power and distributed rooftop solar. Our method combines a 2D advection model with cloud motion vectors (CMVs)derived from a mesoscale numerical weather prediction (NWP)model and sparse optical flow acting on successive, geostationary satellite images. We use ensemble data assimilation to combine these sources of cloud motion information based on the uncertainty of each data source. Our technique produces forecasts that have similar or lower root mean square error than reference techniques that use only optical flow, NWP CMV fields, or persistence. We describe how the method operates on three representative case studies and present results from 39 cloudy days.

We introduce a computational framework to forecast cloud index (CI)fields for up to one hour on a spatial domain that covers a city. Such intra-hour CI forecasts are important to produce solar power forecasts of utility scale solar power and distributed rooftop solar. Our method combines a 2D advection model with cloud motion vectors (CMVs)derived from a mesoscale numerical weather prediction (NWP)model and sparse optical flow acting on successive, geostationary satellite images. We use ensemble data assimilation to combine these sources of cloud motion information based on the uncertainty of each data source. Our technique produces forecasts that have similar or lower root mean square error than reference techniques that use only optical flow, NWP CMV fields, or persistence. We describe how the method operates on three representative case studies and present results from 39 cloudy days.

The impacts of weather forecast uncertainties have not been quantified in current air quality forecasting systems. To address this, we developed an efficient 2‐D convolutional neural network‐surface ozone ensemble forecast (2DCNN‐SOEF) system using 2‐D convolutional neural network and weather ensemble forecasts, and we applied the system to 216‐hr ozone forecasts in Shenzhen, China. The 2DCNN‐SOEF demonstrated comparable performance to current operating forecast systems and met the air quality level forecast accuracies required by the Chinese authorities up to 144‐hr lead time. Uncertainties in weather forecasts contributed 38%–54% of the ozone forecast errors at 24‐hr lead time and beyond. The 2DCNN‐SOEF enabled an "ozone exceedance probability" metric, which better represented the risks of air pollution given the range of possible weather outcomes. Our ensemble forecast framework can be extended to operationally forecast other meteorology‐dependent environmental risks globally, making it a valuable tool for environmental management. Plain Language Summary: Weather forecasts are intrinsically uncertain, but the impacts of that uncertainty on air quality forecasts are not explicitly quantified in current air quality forecast systems. We proposed here a surface ozone ensemble forecast system, analogous to modern weather ensemble forecast systems, to represent the probability distribution of forecasted surface ozone concentrations given 30–50 possible future weather outcomes. The computation costs of this surface ozone ensemble forecast system were greatly reduced using deep learning techniques that emphasized the spatial patterns of weather. We showed that the surface ozone ensemble forecast system's accuracy met the Chinese operational requirements. However, half of the ozone forecast error was due to weather forecast uncertainties, which cannot be completely eliminated even with perfect pollutant emission estimates and chemistry models. This weather‐induced innate uncertainty in air quality forecasts should be considered for effective air quality management. Key Points: We built a deep‐learning surface ozone ensemble forecast system to quantify pollution risks given the range of possible weather outcomesDeep‐learning models accentuating the spatial patterns of weather effectively represented the ozone‐meteorology relationshipWeather forecast uncertainties contributed 38%–54% of the ozone forecast errors at 24‐hr lead time in Shenzhen [ABSTRACT FROM AUTHOR]

Studies have shown that assimilating the radio occultation (RO) observations, including those from the FORMOSAT‐3/COSMIC (constellation observing systems for meteorology, ionosphere, and climate) (FS3‐C), provides positive impacts on tropical cyclone (TC) forecasts. The FS3‐C's successor, the FORMOSAT‐7/COSMIC‐2 (FS7‐C2), provides denser spatial data coverage over the Tropics and Subtropics, where severe weather systems often occur. This study investigates the impact of FS7‐C2 refractivity profiles on the prediction of TC genesis. A quick observing system simulation experiment is conducted for the period when Hurricanes Helene and Gordon (2006) occurred over the North Atlantic Ocean using a regional ensemble data assimilation system. Though assimilating FS3‐C or FS7‐C2 ROs successfully reproduces Helene's development, assimilating FS7‐C2 ROs better captures the genesis and development of Gordon with abundant moisture and vorticity in Gordon's core region, providing conditions favorable for the development of deep convection. A minimum area‐mean total precipitable water vapor of 54 mm, as well as the existence of mid‐level cyclonic vorticity (e.g., 500 hPa), at the storm core region in the initial condition is required for forecasting Gordon's genesis. Also, the assimilation of FS7‐C2 ROs in our experiments reduces the 500 hPa geopotential error by 22% and improves probabilistic quantitative precipitation forecast compared with assimilating FS3‐C ROs. Two sensitivity tests are conducted to evaluate the impact of low‐level negatively biased FS7‐C2 RO profiles and the removal of FS7‐C2 data below 3 km on Gordon's genesis. The former test does not degrade Gordon's genesis forecast skills due to a dipole error correlation between the background ROs and the moisture field over an observed RO profile near Gordon. The latter test does degrade Gordon's forecast skills but is still better than the assimilation of FS3‐C ROs since the features of low‐level moisture and mid‐level vorticity are preserved to some extent. [ABSTRACT FROM AUTHOR]

This work is devoted to the development of an ensemble of below-cloud scavenging models of pollutant aerosol transport into the atmosphere. Among other factors contributing to the uncertainty of the forecasts of the dispersion and deposition of technogenic gas-aerosol releases in the atmosphere, precipitation scavenging is one of the least studied and, in case of precipitation, can be the dominant mechanism for aerosol deposition. To form the ensemble of below-cloud scavenging models, appropriate experimental data, raindrop-aerosol capture models, raindrop terminal velocity parameterizations, and raindrop size distributions were chosen. The pool of models was prepared and then evaluated to adequately describe the experimental data using statistical analysis. Rank diagrams were used to analyze the adequacy of meteorological ensembles; together with the ensemble distribution construction, they allowed selecting the groups of models with such properties as to produce unbiased estimates and dispersion corresponding to the dispersion of the experimental data. The model calculations of the concentration fraction deposited due to below-cloud scavenging were performed using a log-normal distribution with characteristics corresponding to those observed during the accidents at the Chernobyl NPP and Fukushima-1 NPP. The results were compared with those obtained using the models of the NAME and FLEXPART codes. The results of this work can be used to improve the current approaches applied for modelling the distribution of pollutants in the atmosphere in the case of emergency, enhancing the reliability of forecasts by taking into account uncertainties in the results. The formed multi-model ensemble will be included in the decision support system used in responding to releases of radioactive substances into the atmosphere. [ABSTRACT FROM AUTHOR]

The data‐driven approaches for medium‐range weather forecasting are recently shown to be extraordinarily promising for ensemble forecasting due to their fast inference speed compared to the traditional numerical weather prediction models. However, their forecast accuracy can hardly match the state‐of‐the‐art operational ECMWF Integrated Forecasting System (IFS) model. Previous data‐driven approaches perform ensemble forecasting using some simple perturbation methods, like the initial condition perturbation and the Monte Carlo dropout. However, their ensemble performance is often limited arguably by the sub‐optimal ways of applying perturbation. We propose a Swin Transformer‐based Variational Recurrent Neural Network (SwinVRNN), which is a stochastic weather forecasting model combining a SwinRNN predictor with a perturbation module. SwinRNN is designed as a Swin Transformer‐based recurrent neural network, which predicts the future states deterministically. Furthermore, to model the stochasticity in the prediction, we design a perturbation module following the Variational Auto‐Encoder paradigm to learn the multivariate Gaussian distributions of a time‐variant stochastic latent variable from the data. Ensemble forecasting can be easily performed by perturbing the model features leveraging the noise sampled from the learned distribution. We also compare four categories of perturbation methods for ensemble forecasting, that is, fixed distribution perturbation, learned distribution perturbation, MC dropout, and multi model ensemble. Comparisons on the WeatherBench data set show that the learned distribution perturbation method using our SwinVRNN model achieves remarkably improved forecasting accuracy and reasonable ensemble spread due to the joint optimization of the two targets. More notably, SwinVRNN surpasses operational IFS on the surface variables of the 2‐m temperature and the 6‐hourly total precipitation at all lead times up to 5 days (Code is available at https://github.com/tpys/wwprediction). Plain Language Summary: Ensemble forecasting plays a crucial role in numerical weather prediction (NWP), since a single deterministic model is hard to forecast the chaotic atmosphere conditions. Recent works have begun to explore the data‐driven based ensemble methods due to their rapid prediction speed over traditional NWP. We develop an efficient and effective deep learning model capable of generating large ensemble forecasts with high prediction accuracy and low prediction time cost. The predicted ensemble members have much greater and more reasonable ensemble spread, and better coverage of the ground truth, compared to the prior data‐driven methods. Moreover, our model surpasses the state‐of‐the‐art operational NWP model on the surface atmospheric variables of the 2‐m temperature and the 6‐hourly total precipitation, offering an impressive probability weather prediction baseline. Key Points: A transformer‐based variational model called SwinVRNN is developed for medium‐range weather predictionThe proposed SwinVRNN can effectively generate large ensemble forecasts with great prediction accuracy and reasonable ensemble spreadThe model sets a new state‐of‐the‐art among data‐driven models and surpasses the Integrated Forecast System on key atmospheric variables [ABSTRACT FROM AUTHOR]