Global Forecast Research Articles

FuXi: a cascade machine learning forecasting system for 15-day global weather forecast

Over the past few years, the rapid development of machine learning (ML) models for weather forecasting has led to state-of-the-art ML models that have superior performance compared to the European Centre for Medium-Range Weather Forecasts (ECMWF)’s high-resolution forecast (HRES), which is widely considered as the world’s best physics-based weather forecasting system. Specifically, ML models have outperformed HRES in 10-day forecasts with a spatial resolution of 0.25∘. However, the challenge remains in mitigating the accumulation of forecast errors for longer effective forecasts, such as achieving comparable performance to the ECMWF ensemble in 15-day forecasts. Despite various efforts to reduce accumulation errors, such as implementing autoregressive multi-time step loss, relying on a single model has been found to be insufficient for achieving optimal performance in both short and long lead times. Therefore, we present FuXi, a cascaded ML weather forecasting system that provides 15-day global forecasts at a temporal resolution of 6 hours and a spatial resolution of 0.25∘. FuXi is developed using 39 years of the ECMWF ERA5 reanalysis dataset. The performance evaluation demonstrates that FuXi has forecast performance comparable to ECMWF ensemble mean (EM) in 15-day forecasts. FuXi surpasses the skillful forecast lead time achieved by ECMWF HRES by extending the lead time for Z500 from 9.25 to 10.5 days and for T2M from 10 to 14.5 days. Moreover, the FuXi ensemble is created by perturbing initial conditions and model parameters, enabling it to provide forecast uncertainty and demonstrating promising results when compared to the ECMWF ensemble.

Learning skillful medium-range global weather forecasting.

Global medium-range weather forecasting is critical to decision-making across many social and economic domains. Traditional numerical weather prediction uses increased compute resources to improve forecast accuracy but does not directly use historical weather data to improve the underlying model. Here, we introduce GraphCast, a machine learning-based method trained directly from reanalysis data. It predicts hundreds of weather variables for the next 10 days at 0.25° resolution globally in under 1 minute. GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets, and its forecasts support better severe event prediction, including tropical cyclone tracking, atmospheric rivers, and extreme temperatures. GraphCast is a key advance in accurate and efficient weather forecasting and helps realize the promise of machine learning for modeling complex dynamical systems.

Data-Driven Global Subseasonal Forecast for Intraseasonal Oscillation Components

As a challenge in the construction of a “seamless forecast” system, improving the prediction skills of subseasonal forecasts is a key issue for meteorologists. In view of the evolution characteristics of numerical models and deep-learning models for subseasonal forecasts, as forecast times increase, the prediction skill for high-frequency components will decrease, as the lead time is already far beyond the predictability. Meanwhile, intraseasonal low-frequency components are essential to the change in general circulation on subseasonal timescales. In this paper, the Global Subseasonal Forecast Model (GSFM v1.0) first extracted the intraseasonal oscillation (ISO) components of atmospheric signals and used an improved deep-learning model (SE-ResNet) to train and predict the ISO components of geopotential height at 500 hPa (Z500) and temperature at 850 hPa (T850). The results show that the 10–30 day prediction performance of the SE-ResNet model is better than that of the model trained directly with original data. Compared with other models/methods, this model has a good ability to depict the subseasonal evolution of the ISO components of Z500 and T850. In particular, although the prediction results from the Climate Forecast System Version 2 have better performance through 10 days, the SE-ResNet model is substantially superior to CFSv2 through 10–30 days, especially in the middle and high latitudes. The SE-ResNet model also has a better effect in predicting planetary waves with wavenumbers of 3–8. Thus, the application of data-driven subseasonal forecasts of atmospheric ISO components may shed light on improving the skill of seasonal forecasts.

Ensemble-based analysis of heavy rainfall–snowfall associated with mesoscale precipitation bands within an extratropical cyclone over northeastern China

This paper investigates the key synoptic-scale factors that affected the forecasting of mesoscale rainfall and snowfall and their associated uncertainties in a heavy rain–snow event in northeastern China on 18–20 November 2020, using ensemble-based sensitivity analysis based on global ensemble forecasts from the European Centre for Medium-Range Weather Forecasts. The heavy precipitation event was attributed to an extratropical cyclone and experienced two stages, with the snowfall stage having a better precipitation forecast skill than the rainfall stage. The mesoscale rainfall and snowfall were caused by a mesoscale rainband over Liaoning Province and two mesoscale snowbands over Heilongjiang Province, respectively, and they showed some differences with respect to their forecast skill and related key synoptic-scale factors contributing to the precipitation centers. The precipitation amount in the two different stages was correlated significantly with the midlevel trough and sensitive to the location and intensity of the low-level vortex (surface cyclone), and particularly the low-level jets and the associated water vapor transport. However, some differences were confirmed in the two different stages: the weaker midlevel trough and accompanying weaker low-level temperature trough in the rainfall stage were related to increased precipitation because the midlevel trough was far away from the control area, while the stronger midlevel trough and accompanying stronger low-level temperature trough were associated with increased precipitation in the snowfall stage. In addition to the synoptic-scale low-level jet (SLLJ), the precipitation in the rainfall stage was also affected by a boundary layer jet (BLJ) over the ocean, while only SLLJs were present in the snowfall stage. The uncertainty of the precipitation forecast was derived mainly from the uncertainty in the strength and location of the SLLJs and BLJ. Notably, the intensity of northeasterly winds west of the low-level vortex may affect the predictability of heavy snowfall.

IoT-based wireless data acquisition and control system for photovoltaic module performance analysis

The development of photovoltaic (PV) technology has led to an increasing demand for efficient and reliable monitoring systems that can ensure the optimal performance of PV modules. In particular, remote monitoring systems are critical for PV systems installed in distant locations where manual monitoring is difficult and expensive. In this article, we introduce a low-cost wireless monitoring system that employs NodeMCU boards, Raspberry Pi, and Internet of Things (IoT) technologies to monitor and analyze the operational and environmental data of PV modules. By using free and open-source software packages including MQTT Mosquitto Broker, InfluxDB, Node-RED, and Grafana, the system can collect, store, and visualize data, offering an effective means of studying a PV system performance from anywhere. The proposed system is found to work efficiently under different testing conditions; providing precise measurements for different parameters, including voltage, current, power, ambient temperature, humidity, panel temperature, PV coolant temperature at inlet and outlet, and spectral irradiance. In addition to the data collected by the set of sensors, the system provides another source of environmental data using a web-based service called "OpenWeatherMap" that provides a complete global forecast. Alerts and thresholds have been successfully implemented in our system, providing an added advantage of being able to detect and respond to any anomalies in the collected data. Furthermore, the system provides two operational controls: wireless ON/OFF control and wireless control of a servo motor which can be utilized in various applications improving overall system performance.

On the use of consistent bias corrections to enhance the impact of AeolusLevel‐2B Rayleigh winds on National Oceanic and Atmospheric Administration global forecast skill

AbstractThe operational Aeolus Level‐2B (L2B) horizontal line‐of‐sight (HLOS) retrieved Rayleigh winds, produced by the European Space Agency (ESA), utilize European Centre for Medium‐Range Weather Forecasts (ECMWF) short‐term forecasts of temperature, pressure, and horizontal winds in the Rayleigh–Brillouin and M1 correction procedures. These model fields or backgrounds can contain ECMWF model‐specific errors, which may propagate to the retrieved Rayleigh winds. This study examines the sensitivity of the retrieved Rayleigh winds to the changes in the model backgrounds, and the potential benefit of using the same system, in this case the National Oceanic and Atmospheric Administration's Finite‐Volume Cubed Sphere Global Forecast System (FV3GFS), for both the corrections and the data assimilation and forecast procedures. It is shown that the differences in the model backgrounds (FV3GFS minus ECMWF) can propagate through the Level‐2B horizontal line‐of‐sight Rayleigh wind retrieval process, mainly the M1 correction, resulting in differences in the retrieved Rayleigh winds with mean and standard deviation of magnitude as large as 0.2 m·s−1. The differences reach up to 0.4, 0.6, and 0.7 m·s−1 for the 95th, 99th, and 99.5th percentiles of the sample distribution with maxima of ∼1.4 m·s−1. The numbers of the large differences for the combined lower and upper 5th, 1st, and 0.5th percentile pairs are ∼6,100, 1,220, and 610 between 2.5 and 25 km height globally per day respectively. The ESA‐disseminated Rayleigh wind product (based on the ECMWF corrections) already shows a significant positive impact on the FV3GFS global forecasts. In the observing system experiments performed, compared with the ESA Rayleigh winds, the use of the FV3GFS‐corrected Rayleigh winds lead to ∼0.5% more Rayleigh winds assimilated in the lower troposphere and show enhanced positive impact on FV3GFS forecasts at the day 1–10 range but limited to the Southern Hemisphere.

A robust error correction method for numerical weather prediction wind speed based on Bayesian optimization, variational mode decomposition, principal component analysis, and random forest: VMD-PCA-RF (version 1.0.0)

Abstract. Accurate wind speed prediction is crucial for the safe and efficient utilization of wind resources. However, current single-value deterministic numerical weather prediction methods employed by wind farms do not adequately meet the actual needs of power grid dispatching. In this study, we propose a new hybrid forecasting method for correcting 10 m wind speed predictions made by the Weather Research and Forecasting (WRF) model. Our approach incorporates variational mode decomposition (VMD), principal component analysis (PCA), and five artificial intelligence algorithms: deep belief network (DBN), multilayer perceptron (MLP), random forest (RF), eXtreme gradient boosting (XGBoost), light gradient boosting machine (lightGBM), and the Bayesian optimization algorithm (BOA). We first predict wind speeds using the WRF model, with initial and lateral boundary conditions from the Global Forecast System (GFS). We then perform two sets of experiments with different input factors and apply BOA optimization to tune the four artificial intelligence models, ultimately building the final models. Furthermore, we compare the aforementioned five optimal artificial intelligence models suitable for five provinces in southern China in the wintertime: VMD-PCA-RF in December 2021 and VMD-PCA-lightGBM in January 2022. We find that the VMD-PCA-RF evaluation indices exhibit relative stability over nearly a year: the correlation coefficient (R) is above 0.6, forecasting accuracy (FA) is above 85 %, mean absolute error (MAE) is below 0.6 m s−1, root mean square error (RMSE) is below 0.8 m s−1, relative mean absolute error (rMAE) is below 60 %, and relative root mean square error (rRMSE) is below 75 %. Thus, for its promising performance and excellent year-round robustness, we recommend adopting the proposed VMD-PCA-RF method for improved wind speed prediction in models.

Integrating LEO and GEO Observations: Toward Optimal Summertime Satellite Precipitation Retrieval

Abstract Reliable quantitative precipitation estimation with a rich spatiotemporal resolution is vital for understanding the Earth’s hydrological cycle. Precipitation estimation over land and coastal regions is necessary for addressing the high degree of spatial heterogeneity of water availability and demand, and for resolving the extremes that modulate and amplify hazards such as flooding and landslides. Advancements in computation power along with unique high spatiotemporal and spectral resolution data streams from passive meteorological sensors aboard geosynchronous Earth-orbiting (GEO) and low Earth-orbiting (LEO) satellites offer exciting opportunities to retrieve information about surface precipitation phenomena using data-driven machine learning techniques. In this study, the capabilities of U-Net–like architecture are investigated to map instantaneous, summertime surface precipitation intensity at the spatial resolution of 2 km. The calibrated brightness temperature products from the Global Precipitation Measurement (GPM) Microwave Imager (GMI) radiometer are combined with multispectral images (visible, near-infrared, and infrared bands) from the Advanced Baseline Imager (ABI) aboard the GOES-R satellites as main inputs to the U-Net–like precipitation algorithm. Total precipitable water and 2-m temperature from the Global Forecast System (GFS) model are also used as auxiliary inputs to the model. The results show that the U-Net–like algorithm can capture fine-scale patterns and intensity of surface precipitation at high spatial resolution over stratiform and convective precipitation regimes. The evaluations reveal the potential of extracting relevant, high spatial features over complex surface types such as mountainous regions and coastlines. The algorithm allows users to interpret the inputs’ importance and can serve as a starting point for further exploration of precipitation systems within the field of hydrometeorology.

Estimating Full Longwave and Shortwave Radiative Transfer with Neural Networks of Varying Complexity

Abstract Radiative transfer (RT) is a crucial but computationally expensive process in numerical weather/climate prediction. We develop neural networks (NN) to emulate a common RT parameterization called the Rapid Radiative Transfer Model (RRTM), with the goal of creating a faster parameterization for the Global Forecast System (GFS) v16. In previous work we emulated a highly simplified version of the shortwave RRTM only—excluding many predictor variables, driven by Rapid Refresh forecasts interpolated to a consistent height grid, using only 30 sites in the Northern Hemisphere. In this work we emulate the full shortwave and longwave RRTM—with all predictor variables, driven by GFSv16 forecasts on the native pressure–sigma grid, using data from around the globe. We experiment with NNs of widely varying complexity, including the U-net++ and U-net3+ architectures and deeply supervised training, designed to ensure realistic and accurate structure in gridded predictions. We evaluate the optimal shortwave NN and optimal longwave NN in great detail—as a function of geographic location, cloud regime, and other weather types. Both NNs produce extremely reliable heating rates and fluxes. The shortwave NN has an overall RMSE/MAE/bias of 0.14/0.08/−0.002 K day−1 for heating rate and 6.3/4.3/−0.1 W m−2 for net flux. Analogous numbers for the longwave NN are 0.22/0.12/−0.0006 K day−1 and 1.07/0.76/+0.01 W m−2. Both NNs perform well in nearly all situations, and the shortwave (longwave) NN is 7510 (90) times faster than the RRTM. Both will soon be tested online in the GFSv16. Significance Statement Radiative transfer is an important process for weather and climate. Accurate radiative transfer models exist, such as the RRTM, but these models are computationally slow. We develop neural networks (NNs), a type of machine learning model that is often computationally fast after training, to mimic the RRTM. We wish to accelerate the RRTM by orders of magnitude without sacrificing much accuracy. We drive both the NNs and RRTM with data from the GFSv16, an operational weather model, using locations around the globe during all seasons. We show that the NNs are highly accurate and much faster than the RRTM, which suggests that the NNs could be used to solve radiative transfer inside the GFSv16.

Downscaled seasonal forecasts for the California Current System: Skill assessment and prospects for living marine resource applications

Ocean forecasting is now widely recognized as an important approach to improve the resilience of marine ecosystems, coastal communities, and economies to climate variability and change. In particular, regionally tailored forecasts may serve as the foundation for a wide range of applications to facilitate proactive decision making. Here, we describe and assess ~30 years of retrospective seasonal (1–12 month) forecasts for the California Current System, produced by forcing a regional ocean model with output from a global forecast system. Considerable forecast skill is evident for surface and bottom temperatures, sea surface height, and upper ocean stratification. In contrast, mixed layer depth, surface wind stress, and surface currents exhibit little predictability. Ocean conditions tend to be more predictable in the first half of the year, owing to greater persistence for forecasts initialized in winter and dynamical forecast skill consistent with winter/spring influence of the El Niño–Southern Oscillation (ENSO) for forecasts initialized in summer. Forecast skill above persistence appears to come through the ocean more than through the atmosphere. We also test the sensitivity of forecast performance to downscaling method; bias correcting global model output before running the regional model greatly reduces bias in the downscaled forecasts, but only marginally improves prediction of interannual variability. We then tailor the physical forecast evaluation to a suite of potential ecological applications, including species distribution and recruitment, bycatch and ship-strike risk, and indicators of ecosystem change. This evaluation serves as a template for identifying promising ecological forecasts based on the physical parameters that underlie them. Finally, we discuss suggestions for developing operational forecast products, including methodological considerations for downscaling as well as the respective roles of regional and global forecasts.

Predictability of the Most Long‐Lived Tropical Cyclone Freddy (2023) During Its Westward Journey Through the Southern Tropical Indian Ocean

AbstractThis study aimed to explore the predictability of the most long‐lived tropical cyclone (TC) Freddy in 2023 while it traversed westward across the southern tropical Indian Ocean during the first 18 days of its existence. Global ensemble forecasts revealed southward track deflection and intensity underestimation of Freddy. We identified three key factors contributing to the limited predictability of Freddy, which are associated with the Mascarene High, Storm Dingani, and Freddy itself. The large track errors of Freddy can be attributed to the underestimated strength of the Mascarene High, the more northeastern position of Dingani, and the presence of excessively large or small sizes of Freddy. These findings were further validated through a high‐resolution regional model. Specifically, Freddy's track and intensity most closely matched the observations when these three factors were most closely represented. It underscores the pivotal role played by the interaction between TCs and multi‐scale systems in TC forecasts.

Retrospective evaluation of real-time estimates of global COVID-19 transmission trends and mortality forecasts.

Since 8th March 2020 up to the time of writing, we have been producing near real-time weekly estimates of SARS-CoV-2 transmissibility and forecasts of deaths due to COVID-19 for all countries with evidence of sustained transmission, shared online. We also developed a novel heuristic to combine weekly estimates of transmissibility to produce forecasts over a 4-week horizon. Here we present a retrospective evaluation of the forecasts produced between 8th March to 29th November 2020 for 81 countries. We evaluated the robustness of the forecasts produced in real-time using relative error, coverage probability, and comparisons with null models. During the 39-week period covered by this study, both the short- and medium-term forecasts captured well the epidemic trajectory across different waves of COVID-19 infections with small relative errors over the forecast horizon. The model was well calibrated with 56.3% and 45.6% of the observations lying in the 50% Credible Interval in 1-week and 4-week ahead forecasts respectively. The retrospective evaluation of our models shows that simple transmission models calibrated using routine disease surveillance data can reliably capture the epidemic trajectory in multiple countries. The medium-term forecasts can be used in conjunction with the short-term forecasts of COVID-19 mortality as a useful planning tool as countries continue to relax public health measures.

Analysis of the Gálvez–Davison Index for the Forecasting Formation and Evolution of Convective Clouds in the Tropics: Western Cuba

The Gálvez–Davison Index (GDI) is an atmospheric stability index recently developed to improve the prediction of thunderstorms and shallower types of moist convection in the tropics. Because of its novelty, its use for tropical regions remains largely unexplored. Cuba is a region that suffers extreme weather events, such as tropical storms and hurricanes, some of them worsened by climate change. This research analyzes the effectiveness of the GDI in detecting the potential for convective cloud development, using forecast data from the Weather Research and Forecasting (WRF) model for Western Cuba. To accomplish this, here, we evaluated the performance of the GDI in ten study cases from the dry and wet seasons. As part of our study, we researched how GDI correlates with brightness temperatures (BTs) measured using GOES-16. In addition, the GDI results with the WRF model are compared with results using the Global Forecast System (GFS). Our results show a high correlation between the GDI and BT, concluding that the GDI is a robust tool for forecasting both synoptic and mesoscale convective phenomena over the region studied. In addition, the GDI is able to adequately forecast stability conditions. Finally, the GDI values computed from the WRF model perform much better than those from the GFS, probably because of the greater horizontal resolution in the WRF model.

Forecasting daily extreme temperatures in Chinese representative cities using artificial intelligence models

Accurate daily extreme temperature forecasts are important for public safety. However, traditional Numerical Weather Prediction (NWP) methods for temperature forecasting become increasingly costly and time-consuming as forecast resolution improves. To overcome these challenges, recent studies have embraced artificial intelligence (AI) techniques. In this paper, we select five AI models: Multiple linear regression, Support Vector Regressor, Gradient Boosting Regression Tree (GBRT), Long Short-Term Memory, and multilayer perceptron (MLP) to forecast extreme temperatures in nine representative Chinese cities, which are Harbin, Xian, Beijing, and Lhasa in the North and Shanghai, Guiyang, Guangzhou, Shenzhen, and Haikou in the South. To assess the performances, we compare AI models with two benchmarks: the persistence method by using the previous day's temperature extremes directly, and the prediction from the NWP method of the Global Forecast System (GFS) provided by the National Centers for Environmental Prediction. Among the 7 models forecasting daily extreme temperatures in nine cities, the MLP model outperforms the best, followed by the GBRT model. Compared with the persistence forecast, all 5 AI models and GFS show forecasting ability for daily maximum temperature. However, the GFS performs worse than the persistence model for daily minimum temperature. Specifically, for the MLP model, the average Mean Absolute Errors (MAEs) for forecasting daily maximum and minimum temperatures in the 9 cities are 1.61 °C and 1.14 °C, reducing the GFS model's MAEs for daily maximum and minimum temperatures forecasts by 15% and 45%, respectively. Overall, AI models show promise in enhancing temperature forecasting accuracy.

Efficient coastal inundation early-warning system for low-lying atolls, dealing with lagoon and ocean side inundation in Tarawa, Kiribati

Tarawa is a low-lying atoll in the Gilbert Island group, capital of the Republic of Kiribati and home of nearly 70.000 inhabitants. With limited land area, rapid population growth and urbanization, strong interannual sea level variability induced by ENSO and sea level rise, Tarawa is highly vulnerable to coastal flooding. In this context, Early Warning Systems are a proven cost-effective climate adaptation measure to strengthen community resilience. In virtually enclosed atolls, the water level experienced at the shore is compounded by tides, sea level anomaly, storm surge and the contribution of waves. While wave setup and runup, are the primary components driving inundation along the ocean-facing shorelines, sea level anomaly, wind setup and wave pumping through the atoll rim contributes more inside lagoons. In this paper we present an efficient process-based approach to forecast flooding events along both, the ocean and lagoon coasts of atoll islands. With the intention of being highly scalable to other island countries, the system has been designed as a lightweight and accurate tool, that provides actionable and user-friendly water level predictions 7 days in advance. Publicly available global forecast products are ingested by a high-resolution wave model and tailor-made metamodels to translate ocean forcings to water levels at the shore. In absence of a comprehensive topography dataset, extreme value distributions of 27-year hourly water levels were evaluated every 500 m along the coast to define and communicate different levels of warnings according to the recurrence interval of the forecasted event. The long-term wave climate, nearshore wave transformation, and water levels produced under this work increase Tarawa's risk knowledge and support informed investments and future development strategies. While this system will significantly enhance ocean services in Kiribati, improving baseline data still remains a critical need for government and communities to support informed decision making to better cope with increased coastal hazards.

Recommendations to improve the interpretation of global flood forecasts to support international humanitarian operations for tropical cyclones

AbstractInternational humanitarian organisations increasingly turn to forecast teams to support the coordination of efforts to respond to disasters caused by hazards such as tropical cyclones and large‐scale fluvial floods. Such disasters often occur where there is limited local capacity or information available to support decision making and so global forecasting capacity is utilised to provide impact‐based flood forecast bulletins. A multidisciplinary team joined together to provide forecast bulletins and expertise for such events through the UK Foreign and Commonwealth Development Office (FCDO). This paper captures the successes and challenges from two cyclones: Hurricane Iota in Central America (November 2020) and Cyclone Eloise in Mozambique (January 2021). Recommendations to improve global forecasting systems are made which will benefit the international community of researchers and practitioners involved in disaster prediction, anticipatory action and response. These include the need for additional data and expertise to support the interpretation of global models, clear documentation to support decision makers faced with multiple sources of information, and the development of user relevant metrics to assess the skill of global models. We discuss the value of effective partnerships and improving synergies between global models and local contexts, highlighting how global forecasting can help build local forecasting capability.

Machine Learning-Based Calibrated Model for Forecast Vienna Mapping Function 3 Zenith Wet Delay

An accurate estimation of zenith wet delay (ZWD) is crucial for global navigation satellite system (GNSS) positioning and GNSS-based precipitable water vapor (PWV) inversion. The forecast Vienna Mapping Function 3 (VMF3-FC) is a forecast product provided by the Vienna Mapping Functions (VMF) data server based on the European Centre for Medium-Range Weather Forecasts (ECMWF)-based numerical weather prediction (NWP) model. The VMF3-FC can provide ZWD at any time and for any location worldwide; however, it has an uneven accuracy distribution and fails to match the application requirements in certain areas. To address this issue, in this study, a calibrated model for VMF3-FC ZWD, named the XZWD model, was developed by utilizing observation data from 492 radiosonde sites globally from 2019–2021 and the eXtreme Gradient Boosting (XGBoost) algorithm. The performance of the XZWD model was validated using 2022 observation data from the 492 radiosonde sites. The XZWD model yields a mean bias of −0.03 cm and a root-mean-square error (RMSE) of 1.64 cm. The XZWD model outperforms the global pressure and temperature 3 (GPT3) model, reducing the bias and RMSE by 94.64% and 58.90%, respectively. Meanwhile, the XZWD model outperforms VMF3-FC, with a reduction of 92.68% and 6.29% in bias and RMSE, respectively. Furthermore, the XZWD model reduces the impact of ZWD accuracy by latitude, height, and seasonal variations more effectively than the GPT3 model and VMF3-FC. Therefore, the XZWD model yields higher stability and accuracy in global ZWD forecasting.

Synthesis‐Style Auto‐Correlation‐Based Transformer: A Learner on Ionospheric TEC Series Forecasting

AbstractAccurate 1‐day global total electron content (TEC) forecasting is essential for ionospheric monitoring and satellite communications. However, it faces challenges due to limited data and difficulty in modeling long‐term dependencies. This study develops a highly accurate model for 1‐day global TEC forecasting. We utilized generative TEC data augmentation based on the International Global Navigation Satellite Service (IGS) data set from 1998 to 2017 to enhance the model's prediction ability. Our model takes the TEC sequence of the previous 2 days as input and predicts the global TEC value for each hourly step of the next day. We compared the performance of our model with 1‐day predicted ionospheric products provided by both the Center for Orbit Determination in Europe (C1PG) and Beihang University (B1PG). We proposed a two‐step framework: (a) a time series generative model to produce realistic synthetic TEC data for training, and (b) an auto‐correlation‐based transformer model designed to capture long‐range dependencies in the TEC sequence. Experiments demonstrate that our model significantly improves 1‐day forecast accuracy over prior approaches. On the 2018 benchmark data set, the global root mean squared error (RMSE) of our model is reduced to 1.17 TEC units (TECU), while the RMSE of the C1PG model is 2.07 TECU. Reliability is higher in middle and high latitudes but lower in low latitudes (RMSE < 2.5 TECU), indicating room for improvement. This study highlights the potential of using data augmentation and auto‐correlation‐based transformer models trained on synthetic data to achieve high‐quality 1‐day global TEC forecasting.

Improved regional forecasting of an extreme Arctic cyclone in August 2016 with WRF MRI‐4DVAR

AbstractCycling data assimilation and forecast experiments in August 2016 together with a case study of an intense Arctic cyclone (AC16) are performed. Initial conditions from newly developed Multi‐Resolution Incremental Four‐Dimensional Variational (MRI‐4DVAR) and Three‐Dimensional Variational (3DVAR) data assimilation along with forecasts from the polar version of the Weather Research and Forecasting (Polar WRF) model, mimicking operational configurations, are applied. The tasks are to evaluate MRI‐4DVAR performance during a 20‐day cycling run, to investigate the impacts of initial conditions on the forecast skill of AC16, and to identify the factors impacting AC16's predictability. The results from the 20‐day cycling period demonstrate the robustness and reliability of MRI‐4DVAR for data assimilation and subsequent forecast skill. Multiple processes, including mergers of Arctic cyclones, mergers of vortices, vertical coupling between low‐level and upper‐level circulations, baroclinic processes and jet stream forcing, contributed to the generation and development of AC16. Compared to the initial conditions from 4DVAR, 3DVAR produced amplified polar vortices, stronger baroclinic instability, intensified upper‐level jet streams and a stronger low‐level frontal zone, causing the overdevelopment of AC16 in 3DVAR‐based forecasts. For MRI‐4DVAR, the accurate prediction of AC16 5–7 days ahead is likely due primarily to the more accurate representation of upper‐level atmospheric fields, that was facilitated by better satellite radiance assimilation with MRI‐4DVAR that also produced a balanced initial model state. It is concluded that the high‐resolution Polar WRF which is optimized for Arctic conditions combined with 4DVAR facilitated the improved prediction of AC16 compared to the Global Forecast System (GFS) operational deterministic global forecast.

Recent Impact of COSMIC-2 with Improved Radio Occultation Data Assimilation Algorithms

Abstract A Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) follow-on constellation, COSMIC-2, was successfully launched into equatorial orbit on 24 June 2019. With an increased signal-to-noise ratio due to improved receivers and digital beam-steering antennas, COSMIC-2 is producing about 5000 high-quality radio occultation (RO) profiles daily over the tropics and subtropics. The initial evaluation of the impact of assimilating COSMIC-2 into NOAA’s Global Forecast System (GFS) showed mixed results, and adjustments to quality control procedures and observation error characteristics had to be made prior to the assimilation of this dataset in the operational configuration in May 2020. Additional changes in the GFS that followed this initial operational implementation resulted in a larger percentage of rejection (∼90%) of all RO observations, including COSMIC-2, in the mid- to lower troposphere. Since then, two software upgrades directly related to the assimilation of RO bending angle observations were developed. These improvements were aimed at optimizing the utilization of COSMIC-2 and other RO observations to improve global weather analyses and forecasts. The first upgrade was implemented operationally in September 2021 and the second one in November 2022. This study describes both RO software upgrades and evaluates the impact of COSMIC-2 with this most recently improved configuration. Specifically, we show that the assimilation of COSMIC-2 observations has a significant impact in improving temperature and winds in the tropics, though benefits also extend to the extratropical latitudes.