Search

Skillful subseasonal forecasts are crucial for various sectors of society but pose a grand scientific challenge. Recently, machine learning-based weather forecasting models outperform the most successful numerical weather predictions generated by the European Centre for Medium-Range Weather Forecasts (ECMWF), but have not yet surpassed conventional models at subseasonal timescales. This paper introduces FuXi Subseasonal-to-Seasonal (FuXi-S2S), a machine learning model that provides global daily mean forecasts up to 42 days, encompassing five upper-air atmospheric variables at 13 pressure levels and 11 surface variables. FuXi-S2S, trained on 72 years of daily statistics from ECMWF ERA5 reanalysis data, outperforms the ECMWF's state-of-the-art Subseasonal-to-Seasonal model in ensemble mean and ensemble forecasts for total precipitation and outgoing longwave radiation, notably enhancing global precipitation forecast. The improved performance of FuXi-S2S can be primarily attributed to its superior capability to capture forecast uncertainty and accurately predict the Madden-Julian Oscillation (MJO), extending the skillful MJO prediction from 30 days to 36 days. Moreover, FuXi-S2S not only captures realistic teleconnections associated with the MJO but also emerges as a valuable tool for discovering precursor signals, offering researchers insights and potentially establishing a new paradigm in Earth system science research. This paper introduces FuXi-S2S, a machine-learning model that outperforms conventional numerical weather prediction models at subseasonal timescales globally, extending the skillful Madden–Julian Oscillation prediction form 30 days to 36 days. [ABSTRACT FROM AUTHOR]

Warm-sector heavy rainfall along the south China coast poses significant forecasting challenges due to its localized nature and prolonged duration. To improve the prediction of such high-impact weather events, high-resolution numerical weather prediction (NWP) models are increasingly used to more accurately represent topographic effects. However, as these models' grid spacing approaches the scale of convective processes, they enter a "gray zone", where the models struggle to fully resolve the turbulent eddies within the atmospheric boundary layer, necessitating partial parameterization. The appropriateness of applying convection parameterization (CP) schemes within this gray zone remains controversial. To address this, scale-aware CP schemes have been developed to improve the representation of convective transport. Among these, the multi-scale Kain–Fritsch (MSKF) scheme enhances the traditional Kain–Fritsch (KF) scheme, incorporating modifications that facilitate its effective application at spatial resolutions as high as 2 km. In recent years, there has been an increase in the application of machine learning (ML) models across various domains of atmospheric sciences, including efforts to replace conventional physical parameterizations with ML models. This work introduces a multi-output bidirectional long short-term memory (Bi-LSTM) model intended to replace the scale-aware MSKF CP scheme. This multi-output Bi-LSTM model is capable of simultaneously predicting the convection trigger while also modeling the associated convective tendencies and precipitation rates with a high performance. Data for training and testing the model are generated using the Weather Research and Forecast (WRF) model over south China at a horizontal resolution of 5 km. Furthermore, this work evaluates the performance of the WRF model coupled with the ML-based CP scheme against simulations with the traditional MSKF scheme. The results demonstrate that the Bi-LSTM model can achieve high accuracy, indicating the promising potential of ML models to substitute the MSKF scheme in the gray zone. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

In numerical weather prediction (NWP) models, physical parameterization schemes are the most computationally expensive components, despite being greatly simplified. In the past few years, an increasing number of studies have demonstrated that machine learning (ML) parameterizations of subgrid physics have the potential to accelerate and even outperform conventional physics-based schemes. However, as the ML models are commonly implemented using the ML libraries written in Python, very few ML-based parameterizations have been successfully integrated with NWP models due to the difficulty of embedding Python functions into Fortran-based NWP models. To address this issue, we developed a coupler to allow the ML-based parameterizations to be coupled with a widely used NWP model, i.e., the Weather Research and Forecasting (WRF) model. Similar to the WRF I/O methodologies, the coupler provides the options to run the ML model inference with exclusive processors or the same processors for WRF calculations. In addition, to demonstrate the effectiveness of the coupler, the ML-based radiation emulators are trained and coupled with the WRF model successfully. [ABSTRACT FROM AUTHOR]

Significant advancements in the development of machine learning (ML) models for weather forecasting have produced remarkable results. State-of-the-art ML-based weather forecast models, such as FuXi, have demonstrated superior statistical forecast performance in comparison to the high-resolution forecasts (HRES) of the European Centre for Medium-Range Weather Forecasts (ECMWF). However, a common limitation of these ML models is their tendency to generate increasingly smooth predictions as forecast lead times increase, which often results in the underestimation of intensities of extreme weather events. To address this challenge, we developed the FuXi-Extreme model, which employs a denoising diffusion probabilistic model (DDPM) to enhance finer-scale details in the surface forecast data generated by the FuXi model in 5-day forecasts. An evaluation of extreme total precipitation (TP), 10-meter wind speed (WS10), and 2-meter temperature (T2M) illustrates the superior performance of FuXi-Extreme over both FuXi and HRES. Moreover, when evaluating tropical cyclone (TC) forecasts based on International Best Track Archive for Climate Stewardship (IBTrACS) dataset, both FuXi and FuXi-Extreme shows superior performance in TC track forecasts compared to HRES, but they show inferior performance in TC intensity forecasts in comparison to HRES. [ABSTRACT FROM AUTHOR]

Increasing distributed rooftop solar photovoltaic generation in the southern California coast necessitates accurate solar forecasts. In summertime mornings marine boundary layer stratocumulus commonly covers the southern California coast. The inland extent of cloud cover varies primarily depending on the temperature inversion base height (IBH, i.e. boundary layer height) and topography as confirmed using radiosonde sounding measurement and satellite irradiance data. Most operational numerical weather prediction models consistently overestimate irradiance and underpredict cloud cover extent and cloud thickness, presumably due to an underprediction of IBH. A thermodynamic method was developed to modify the boundary layer temperature and moisture profiles to better represent the boundary layer structure in the Weather and Research Forecasting model (WRF). Validation against satellite global horizontal irradiance (GHI) demonstrated that the best IBH ensemble improves GHI accuracy by 23% mean absolute error compared to the baseline WRF model and is similar to 24-h persistence forecasts for coastal marine layer region. The spatial error maps showed deeper inland cloud cover. Validation against ground observations showed that IBH ensembles were able to outperform persistence forecast at coastal stations. [ABSTRACT FROM AUTHOR]

The New Goddard shortwave (SW) radiation scheme of the Weather Research and Forecasting (WRF) numerical weather prediction model leads to positive biases in the clear-sky downwelling SW radiation (also referred to as global horizontal irradiance, GHI). Clear-sky GHI is attenuated primarily by four atmospheric constituents: (i) ozone (ii) background gases (e.g., trace gases), (iii) precipitable water and, (iv) aerosols. The effect of each constituent in the New Goddard SW scheme is isolated here by subtracting from the GHI predicted for an atmosphere that lacks one constituent, the GHI predicted for an atmosphere with all the constituents. Compared with the WRF’s Rapid Radiative Transfer Model for Global Circulation Models (RRTMG), the main contributions to the clear-sky irradiance bias in the New Goddard SW scheme come from modeling issues with the absorptions by water vapor and ozone. Enhancing the absorption due to water vapor continuum and using the RRTMG’s ozone profiles in the New Goddard SW scheme improved the agreement with the WRF’s RRTMG predictions for both GHI and direct normal irradiance. These results are further confirmed with the REST2 radiative transfer model. [ABSTRACT FROM AUTHOR]