National Oceanic and Atmospheric Administration’s (NOAA’s) Environmental Modeling Center (EMC) is a lead developer of operational Numerical Weather Prediction (NWP) systems at the National Weather Service (NWS), which are used for the protection of life and property and the enhancement of the economy. EMC transitions to operations and maintains more than 20 numerical prediction systems that are used by NWS, NOAA, other United States (U.S.) federal agencies, and various other stakeholders. These systems are developed through a close collaboration with academic, federal and commercial sector partners. EMC maintains, enhances and transitions-to-operations numerical forecast systems for weather, ocean, climate, land surface and hydrology, hurricanes, and air quality for the U.S. and global domains.

NOAA’s operational predictions are transitioning to the Unified Forecast System (UFS) framework in order to simplify the operational prediction suite of modeling systems. The UFS is being designed as a community-based, comprehensive atmosphere-ocean-sea-ice-wave-aerosol-land coupled Earth modeling system with coupled data assimilation and ensemble capabilities, organized around applications spanning from local to global domains and predictive time scales ranging from sub-hourly analyses to seasonal predictions.  Disparate legacy operational applications that have been developed and maintained by EMC in support of various stakeholder requirements are being transitioned to the UFS framework. The transition started several years ago and is planned to continue over the next few years. Fewer resulting applications will consolidate NCEP’s Production Suite that shares a set of common scientific components and technical infrastructure.  This streamlined suite is expected to accelerate the transition of research into operations and simplify maintenance of operational systems.

This talk describes major development and operational implementation projects at EMC over the last couple of years including for example a new UFS-based hurricane application, recent advances in the use of satellite data and a new verification system. We will present EMC plans for the next few years, within the overall NOAA strategy, and how planned efforts link with other modeling efforts within NOAA, in the broader U.S. and international community.

How to cite: Stajner, I., Gross, B., Tallapragada, V., Levit, J., Montuoro, R., Mehra, A., Kleist, D., and Yang, F.: NOAA’s Environmental Modeling Center Update: Transitioning to Unified Forecast System Applications for Operations, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-12029, https://doi.org/10.5194/egusphere-egu24-12029, 2024.

Machine Learning (ML) became pervasive in every domain of the research, providing opportunities of modeling phenomena that were difficult to capture using known equations. From small models running on student computers, to giant LLMs trained on the whole Internet, ML models come in all shapes and sizes. To the meteorological community, one branch of this research stands out as revolutionary - ML-based global weather models.

ML-based global weather models lie on the opposite end of the spectrum compared to numerical weather prediction (NWP) models. Instead of representing the physics in a form of equations and solving these equations on the model grid, ML models are purely data-driven - even if they managed to represent physics internally, the inference of that physics would remain a black box.

Yet, these models underwent significant advancement in the past year - and three of them stand out - GraphCast (Google), ClimaX (Microsoft) and MetNet (Google). The former two, open-sourced for research purposes, are being tested currently at Belgingur. Having many years of experience with running and deploying NWP weather models, we notice how working with these models differs from working with the new class of ML-based (or data-driven) models.

This talk discusses essential differences between working with NWP and ML-based weather models. What we can, and what we cannot control? What does the process of working with such an ML model look like? What is the main advantage of an ML model run in production? What are the main obstacles in deploying an ML model and running it operationally?

With the current pace of the growth of Machine Learning models, we will be encountering them in our everyday work sooner or later. Knowing the challenges and opportunities of them will help us understand how to use them to our advantage.

How to cite: Stanisławska, K. and Rognvaldsson, O.: Running global Machine Learning weather models - challenges, observations and conclusions, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-15614, https://doi.org/10.5194/egusphere-egu24-15614, 2024.

The U.S. Navy is building a coupled thermosphere-ionosphere prediction system.  As part of this project, we are developing a new dynamical core (DyCore) extending from the ground to the exobase (~500 km).  The DyCore must be able to handle large variations in both temperature and composition, which motivates a new Horizontally Explicit Vertically Implicit (HEVI) time integrator.  Unlike traditional linear Implicit-EXplicit (IMEX) methods commonly used in numerical weather prediction (NWP), HEVI does not require a fixed reference state.  Our DyCore combines HEVI with a Specific Internal Energy Equation (SIEE) and a Spectral Element Method (SEM) spatial discretization to form a robust, whole-atmosphere model for the neutral atmosphere.  We present results for two test cases using the proposed DyCore: an idealized heating/cooling test extending into the middle thermosphere and a perturbation experiment yielding nonhydrostatic baroclinic instability. The idealized heating/cooling test, which is compared to corresponding results from the hydrostatic Navy Global Environmental Model (NAVGEM), demonstrates that HEVI is more robust than traditional linear IMEX methods.  The baroclinic instability test shows that HEVI, when combined with a banded lower-upper (LU) direct solve, is efficient and allows a large timestep.  These numerical results suggest that our HEVI-enabled DyCore is a good candidate for the proposed thermosphere-ionosphere prediction system.

This work was funded by the Office of Naval Research Marine Meteorology and Space Weather program.

How to cite: Kelly, J. F., Giraldo, F. X., Reinecke, P. A., Alves, F., Barton, C. A., and Eckermann, S. D.: Horizontally Explicit Vertically Implicit (HEVI) Time-Integrators for a Non-Hydrostatic Whole Atmosphere Models, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-2375, https://doi.org/10.5194/egusphere-egu24-2375, 2024.

Aerosols and droplets are the main factors of visibility reduction by scattering and absorbing light. For visibility predictions in operational NWP models, hydrometeors are often considered to be the dominant factor in the total extinction, whereas aerosol effects are usually simplified or omitted in models developed for relatively clean regions. In China, also many NWP studies related to visibility forecasting during haze-fog processes have been conducted, primarily focusing on severely polluted periods before 2018. These studies often employed visibility parameterizations that considered either aerosol extinction alone or hydrometeor extinction alone. Therefore, the significance of incorporating both aerosol and hydrometeor extinction into visibility forecasting during haze-fog processes remains uncertain, particularly under recent rapid changes in aerosol concentration, composition, and hygroscopicity in China.

In this study, we first use the 3-D meteorology fields from the Shanghai Meteorological Service WRF-ADAS Real-Time Modeling System (WARMS) to drive the Community Multiscale Air Quality (CMAQ) model. In this version, CMAQ is used in an off-line mode and visibility is diagnosed by combining extinctions due to hydrometeors and aerosols. Satellited derived NOx emissions using the Daily Emissions Constrained by Satellite Observations (DECSO) algorithm have been incorporated to give more up-to-date emissions. We analyze the results of a one-month forecasting period during the winter of 2021-2022 to assess the model's performance and understand the impact of hydrometeor and aerosol extinction on operational visibility forecasting. We find that for the city of Shanghai, aerosol extinction has a minor impact on the model’s performance when forecasting visibility below 1 km but becomes crucial for predictions spanning 1-10 km. Comparison against observations shows that the model well captures the general contributions from various chemical constituents with nitrate as the most important factor in aerosol extinction (~60%). Furthermore, our assessment of the North China Plain (NCP) highlights that in highly polluted areas aerosols could be significant for visibility below 1 km. Finally, we conduct case studies with the fully coupled WRF-Chem model and compare results with the offline WARMS-CMAQ system. Aerosol effects on fog and visibility forecasting due to feedbacks between aerosols, radiation, and cloud physics will be discussed.

How to cite: Xie, Y., Wang, X., Gao, Y., Chen, B., van der A, R., Ding, J., Gu, W., Zhou, M., and Wang, H.: Improving Visibility Forecasting during Haze-fog Processes in Shanghai and Eastern China: the Significance of Aerosol and Hydrometeor, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3686, https://doi.org/10.5194/egusphere-egu24-3686, 2024.

Data-driven machine learning methods for weather forecasting have experienced a steep progress over the last years, with recent studies demonstrating substantial improvements over physics-based numerical weather prediction models. Beyond improved forecasts, the major advantages of purely data-driven models are their substantially lower computational costs and faster generation of forecasts, once a model has been trained. However, in contrast to ensemble forecasts from physical weather models, most efforts in data-driven weather forecasting have been limited to deterministic, point-valued predictions only, making it impossible to quantify forecast uncertainties which is crucial for optimal decision making in applications.

Our overarching aim is to evaluate and compare methods for creating probabilistic forecasts from data-driven weather models. The uncertainty quantification (UQ) approaches we compare are either based on generating ensemble forecasts from data-driven weather models via perturbations to the initial conditions, or based on statistical post-hoc UQ methods. The perturbation-based methods either leverage initial conditions from the ECMWF IFS ensemble, add random Gaussian noise to the deterministic initial conditions, or add random field perturbations based on past observations (Magnusson et al., 2009). The post-hoc approaches operate on deterministic forecasts and quantify forecast uncertainty using established post-processing methods, namely distributional regression networks (Rasp and Lerch, 2018) and isotonic distributional regression (Walz et al., 2022; Henzi et al., 2021).

Using forecasts from Pangu-Weather (Bi et al., 2023), we evaluate these UQ methods over Europe for selected user-relevant weather variables, such as wind speed at 10 m, temperature at 2 m, and geopotential height at 500 hPa. We focus on daily initialised Pangu-Weather forecasts for 2022 with a forecast horizon of up to 7 days and compare their performance against ECMWF IFS ensemble forecasts. Our results suggest that Pangu-Weather predictions combined with UQ approaches yield improvements over the ECMWF ensemble forecasts for lead times of up to 5 days in terms of the Continuous Ranked Probability Score. However, it strongly depends on the variable of interest which of the UQ methods performs best, none of the different UQ methods performs best over all variables and lead times. Post-hoc UQ methods tend to perform better for shorter lead times, while initial condition perturbations are superior for longer lead times, with in particular the random field method showing promising results.

References:

• Bi, K., Xie, L., Zhang, H., Chen, X., Gu, X. and Tian, Q. (2023). Accurate medium-range global weather forecasting with 3D neural networks. Nature, 619, 533–538.
• Henzi, A., Ziegel, J. F. and Gneiting, T. (2021). Isotonic distributional regression. Journal of the Royal Statistical Society Series B: Statistical Methodology, 83, 963–993.
• Magnusson, L., Nycander, J. and Källén, E. (2009). Flow-dependent versus flow-independent initial perturbations for ensemble prediction. Tellus A: Dynamic Meteorology and Oceanography, 61, 194.
• Rasp, S. and Lerch, S. (2018). Neural networks for postprocessing ensemble weather forecasts. Monthly Weather Review, 146, 3885–3900.
• Walz, E.-M., Henzi, A., Ziegel, J. and Gneiting, T. (2022). Easy Uncertainty Quantification (EasyUQ): Generating Predictive Distributions from Single-valued Model Output. Preprint, available at https://arxiv.org/abs/2212.08376.

How to cite: Horat, N., Bülte, C., Quinting, J., and Lerch, S.: Uncertainty quantification for data-driven weather models, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5395, https://doi.org/10.5194/egusphere-egu24-5395, 2024.

The Global Navigation Satellite System (GNSS) ground-based network in Europe is a comparatively dense network that provides valuable humidity information through Zenith Total Delays (ZTDs) and tropospheric gradients. ZTDs include information on column water vapor, while tropospheric gradients provide information on moisture distribution. Recently, we developed the tropospheric gradient operator (Zus et al., 2023) and implemented it in the Weather Research and Forecasting (WRF) model (Thundathil et al., 2023, under review).

We have conducted ZTD and tropospheric gradient assimilation experiments over a couple of periods, which lasted for two months. We will present our latest test period, the Benchmark Campaign organized within the European COST Action, in May and June 2013. Data from more than 250 GNSS stations in central Europe covering Germany, the Czech Republic, and part of Poland and Austria were assimilated during this period. The data assimilation (DA) system used a rapid update cycle of 3-dimensional variational DA with 6-hourly cycles for two months.

Our research methodology involved configuring a 0.1 x 0.1-degree mesh in the WRF model with 50 vertical levels up to 50 hPa for Europe. Model forcing was done with the European Centre for Medium-Range Weather Forecasts (ECMWF) operational analysis. We conducted three runs, which included the assimilation of conventional datasets from ECMWF (or control run), ZTD added on top of the control run, and ZTD and gradients on top of the control run. We observed a significant reduction of the root mean square errors; we observed a 42 % and 16 % reduction for ZTDs and gradients in the ZTD assimilation run, which further reduced to 43 % and 21 % for ZTDs and gradients in the ZTD and gradient assimilation. Validation with the atmospheric reanalysis ERA5 and radiosondes revealed improvements in the lower troposphere.

We conducted an additional sensitivity experiment using a sparsely distributed GNSS network. This process involved reducing the station density from roughly 0.5 degrees to 1 degree by replacing the original network with one consisting of 100 stations. We found that the improvement in the humidity field with the assimilation of ZTD and gradients from the sparse station network (1-degree resolution) is roughly the same as in the humidity field with the assimilation of ZTD only from the dense station network (0.5-degree resolution). Therefore, the assimilation of gradients in addition to ZTDs is particularly interesting in regions with a few GNSS stations. It may also be considered a cost-effective way to increase the density of networks.

After preliminary testing of the GNSS ZTD plus gradient assimilation with WRF, we are ready to move to convective-scale assimilation using an ensemble-based approach over different regions and seasons. We will be presenting initial results from our high-resolution simulations.

References

Zus, F., Thundathil R., Dick G., and Wickert J. "Fast Observation Operator for Global Navigation Satellite System Tropospheric Gradients." Remote Sensing 15, no. 21 (2023): 5114.

Thundathil, R. M., Zus, F., Dick, G., and Wickert, J. "Assimilation of GNSS Tropospheric Gradients into the Weather Research and Forecasting Model Version 4.4.1", Geoscientific Model Development Discussion [preprint], in review, 2023.

How to cite: Thundathil, R., Zus, F., Dick, G., and Wickert, J.: Impact of GNSS tropospheric gradient assimilation and sensitivity analysis, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5656, https://doi.org/10.5194/egusphere-egu24-5656, 2024.

Existing weather models are known to have poor skill at forecasting rainfall over East Africa, where there are regular threats of drought and floods. Improved precipitation forecasts could reduce the effects of these extreme weather events and provide significant socioeconomic benefits to the region. We present a novel machine learning based method to improve precipitation forecasts in East Africa, using postprocessing based on a conditional generative adversarial network (cGAN). This addresses the challenge of realistically representing tropical rainfall in this region, where convection dominates and is poorly simulated in conventional global forecast models. We postprocess hourly forecasts made by the European Centre for Medium-Range Weather Forecasts Integrated Forecast System at 6-18h lead times, at 0.1^o resolution. We combine the cGAN predictions with a novel neighbourhood version of quantile mapping, to integrate the strengths of both machine learning and conventional postprocessing. Our results indicate that the cGAN substantially improves the diurnal cycle of rainfall, and improves rainfall predictions up to the 99.9^th percentile of rainfall. This improvement persists when evaluating against the 2018 March-May season, which had extremely high rainfall, indicating that the approach has some ability to generalise to more extreme conditions. We explore the potential for the cGAN to produce probabilistic forecasts and find that the spread of this ensemble broadly reflects the predictability of the observations, but is also characterised by a mixture of under- and over-dispersion. Overall our results demonstrate how the strengths of machine learning and conventional postprocessing methods can be combined, and illuminate what benefits machine learning approaches can bring to this region.

How to cite: Antonio, B., McRae, A., MacLeod, D., Cooper, F., Marsham, J., Aitchison, L., Palmer, T., and Watson, P.: Postprocessing East African rainfall forecasts using a generative machine learning model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-2897, https://doi.org/10.5194/egusphere-egu24-2897, 2024.

The Advanced Weather Research Forecast model (WRF-ARW) is used to investigate the potential impacts of assimilating the FengYun-4A (FY-4A) Geostationary Interferometric Infrared Sounder (GIIRS) long-wave temperature sounding channel data on prediction of Typhoon In-Fa (2021). In addition, a series of data assimilation experiments are conducted to demonstrate the added value of the FY-4A/GIIRS data assimilation for typhoon forecasts. It is shown that the higher spectral resolution and broader coverage of GIIRS radiance data can positively impact the model analysis and forecasts with larger temperature and moisture increments at the initial time of simulations, thus producing the better simulation for typhoon warm core aloft, vortex wind structure and spiral rainfall band. Moreover, the assimilation of the GIIRS data can also lead to better storm steering flows and consequently better typhoon track forecasts. Overall, the assimilation of FY-4A/GIIRS temperature sounding channel data shows some added values to improve the track and storm structure forecasts of Typhoon In-Fa.

How to cite: Zhang, L., Niu, Z., Weng, F., Dong, P., Huang, W., and Zhu, J.: Impacts of Direct Assimilation of the FY-4A/GIIRS Long-Wave Temperature Sounding Channel Data on Forecasting Typhoon In-Fa (2021), EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6940, https://doi.org/10.5194/egusphere-egu24-6940, 2024.

Convective-scale ensembles are routinely used in operational centres around the world to produce probabilistic precipitation forecasts, but a lack of spread between members is providing forecasts that are frequently overconfident. This deficiency can be corrected by increasing spread, increasing forecast accuracy or both. A recent development in the Met Office forecasting system is the inclusion of Large-Scale Blending (LSB) in the convective-scale data assimilation scheme. This method aims to reduce the synoptic-scale forecast error in the analysis by reducing the influence of the convective-scale data assimilation at scales that are too large to be constrained by the limited domain. These scales are instead initialised using output from the global data assimilation scheme, which we expect to reduce the forecast error and, thus, improve the spread-skill relationship. In this study, we have quantified the impact of LSB on the spread-skill relationship of hourly precipitation accumulations by comparing forecast ensembles with and without LSB over a 17-day summer trial period. This trial found modest but significant improvements to the spread-skill relationship as calculated using metrics based on the Fractions Skill Score. Skill is improved for a lower precipitation centile by an average of 0.56% at the largest scales, but a corresponding degradation of spread limits the overall correction. The spread-skill disparity is reduced the most in the higher centiles due to a more muted spread response, with significant reductions of up to 0.40% obtained at larger scales. Case study analysis using a novel extension of the Localised Fractions Skill Score demonstrates how spread-skill improvements transfer to smaller-scale features, not just the scales that have been blended. There are promising signs that further spread-skill improvements can be made by implementing LSB more fully within the ensemble.

How to cite: Gainford, A.: Improvements in the spread-skill relationship of precipitation in a convective-scale ensemble through blending, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-9034, https://doi.org/10.5194/egusphere-egu24-9034, 2024.

In modern forecasting it is now a common technique to use an ensemble of forecasts generated by Numerical Weather Prediction (NWP) models. This necessitates a statistical approach be taken when using these weather predictions to inform decision-making and leveraging probabilities in the production of forecasts. It is often required to take the spread of predictions made by NWPs in the ensemble and reduce these to a single value, a pseudo-deterministic forecast, analogous to a forecast made be a traditional deterministic NWP, in order to allow end users to make binary decisions often defined at a definite threshold. These values may be representative of a single physical parameter modelled (e.g. road surface temperature) or may combine multiple parameters in a physically consistent manner (e.g. the road surface temperature coupled to the depth of water on the road for calculating road state), and are used by stakeholders in a number of sectors often to inform safety critical decision making. Therefore, it is important to ensure that the methodology used to reduce the ensemble of predictions to a pseudo-deterministic forecast is as accurate as possible and can retain information related to the ensemble spread , whilst ensuring consistency in parameters through the spatial and temporal domain.

The Surface Transport Forecast (STF) system produces forecasts for different transport surfaces in response to NWP outputs. The STF system is architected such that it runs simultaneously for each member of the NWP forecast ensemble, producing a corresponding ensemble of STF predictions. This enables the computation of a pseudo-deterministic forecast, which retains the maximum amount of information provided by the NWP ensemble.

To reduce the STF ensemble to a pseudo-deterministic forecast a Kernel Density Estimation (KDE) is utilised to build Probability Density Functions (PDFs), which can be readily interrogated using standard statistical techniques. It is found that pseudo-deterministic forecasts, which are consistent across a combination of physical modelled parameters, can be determined using covariant techniques, ensuring the ensemble is reduced as late as possible in the forecast production keeping the maximum benefit provided by the forecast spread. We will present the numerical and computational implementation of the described method in our STF system. Further, we will analyse the pseudo-deterministic forecasts produced and verify the validity of results at specific locations using multiple years of road observations.

How to cite: Wiggs, J., Eyles, J., and Lake, A.: Creating a Pseudo-Deterministic Forecast for Surface Transport from an NWP Ensemble with Consistency Across Multiple Variables using KDE, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-11209, https://doi.org/10.5194/egusphere-egu24-11209, 2024.

A new land surface package developed at Environment and Climate Change Canada (ECCC) has been evaluated in the context of the medium-range global deterministic numerical weather prediction (NWP) system. The evaluation is performed by comparison of NWP forecasts against near-surface and

atmospheric analyses. The new land surface package includes i) new databases to specify soils and vegetation characteristics, ii) improved initialization of land surface variables by the assimilation of space-based remote sensing observations, and iii) a more sophisticated land surface scheme.

Evaluation for the screen-level air temperature and humidity indicates that the new land surface package resulted in smaller STDEs and larger temporal correlation between forecasts and analyses comparing to the current operational configuration. The improvement is greater for humidity than for air temperature.

Upper-air evaluation indicates that the impact of the new land surface package on the Planetary boundary layer (PBL) is substantial but more mixed, with large spatial variability in terms of its effect.

This study also investigated the physical and statistical links between near-surface and upper-air forecast errors at the medium range.

How to cite: Alavi, N., Belair, S., Carrera, M., Abrahamowicz, M., Bilodeau, B., Simjanovski, D., Charpentier, D., Badawy, B., and Leroyer, S.: Impact of a new land surface package in Canadian  numerical weather prediction system on the medium range weather forecast in the lower and upper atmosphere, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-12345, https://doi.org/10.5194/egusphere-egu24-12345, 2024.

The UFS-R2O Project, which began in July 2020 as a five-year plan with deliverables for the first three years funded, has made significant progress in developing the medium-range and sub-seasonal to seasonal (MRW/S2S) predictions, a regional, high-resolution hourly-updating and convection-allowing ensemble system for prediction of short range severe weather (CAM/SRW), and a Hurricane Application developing a very high-resolution Hurricane Analysis and Forecast System (HAFS) with storm following moving nests.  The Project is organized with Application Teams and Development Teams interacting with each other to reflect the cross-cutting nature of the UFS components and infrastructure. It  fostered successful collaborations between the National Centers for Environmental Prediction (NCEP) Environmental Modeling Center, several NOAA research labs, the National Center for Atmospheric Research (NCAR), the Naval Research Lab (NRL), and multiple universities and cooperative institutes.  Most sIgnificant outcomes of the project thus far are the implementation of the HAFSv1 ahead of the schedule, and the development of a six-way global coupled (atmosphere/ ocean/ land/ sea-ice/ wave/ aerosol) modeling system, both within the UFS framework, major accomplishments from the community modeling perspective.  

The UFS-R2O Project has entered into its second phase (2023-2024), albeit with reduced funding, to continue the momentum built during the first phase.  While the first three years of the project were focused on engineering and infrastructure, Phase II is primarily targeting systematic testing and evaluation of the prototype UFS configurations for selecting the candidates for potential transition to operations in the next few years.  In addition, Phase II of the project includes a new Seasonal Forecast System (SFS) Application Tean established to develop SFS v1 that will replace the legacy Climate Forecast System (CFSv2) currently in operations since 2011.

This presentation describes the outcomes of the UFS R2O Project for the first three years, and highlights the progress and plans for the Phase II.

How to cite: Tallapragada, V., Whitaker, J., and Kinter, J.: NOAA's Unified Forecast System Research to Operations (UFS R2O) Project Phase II - Accomplishments, Progress and Future Plans, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-2930, https://doi.org/10.5194/egusphere-egu24-2930, 2024.

The Geophysical Fluid Dynamics Laboratory (GFDL)’s System for High‐resolution prediction on Earth‐to‐Local Domains (SHiELD) model typically uses the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analyses to initialize its medium-range global forecasts. Both initial condition (IC) and forecast model have an impact on model prediction skills. The quality of the IC is partially determined by the model short-range forecast used as first guess in data assimilation. 

A data assimilation (DA) system has been developed for the global SHiELD to demonstrate the prediction skills of the model initialized from its own analysis. The DA system largely leverages the advanced DA techniques used in GFS and assimilates all the observations assimilated in GFS. Compared to the SHiELD forecasts initialized from GFS analysis, SHiELD forecast skill is significantly improved by using its own analysis. Tremendous improvement was found in the Southern Hemisphere with positive impact lasting up to 10 days. The DA system is also useful in identifying and understanding model errors. The most noticeable model error detected by the DA system originates from the TKE-EDMF boundary layer scheme. The model error leads to insufficient ensemble spread, which could not be fully addressed by the multiplicative inflation and stochastic physics schemes used in the system. Including two versions of the TKE EDMF scheme in the ensemble can alleviate the systematic model error, which further improves forecast skills. The use of the interchannel correlated observation errors for Infrared Atmospheric Sounding Interferometer (IASI) and Cross-track Infrared Sounder (CrIS) was also investigated, which improves the forecast skill up to day 5 and further reduces the impact of the model error in the marine stratocumulus region. Further understanding of the model error associated with the TKE-EDMF scheme will be presented. 

How to cite: Tong, M., Harris, L., Zhou, L., Gao, K., Kaltenbaugh, A., and Xiang, B.: Improved Weather Predictions Through Data Assimilation for GFDL SHiELD, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-13187, https://doi.org/10.5194/egusphere-egu24-13187, 2024.

Belgingur Ltd. has created a novel weather forecasting framework, called Weather On Demand – WOD, that is deployable in the cloud and on in-house hardware and which can be customised for any location world-wide at a very short notice.

The WOD framework is a distributed system for:

• Running the WRF weather model for data-assimilation and forecasts by either triggering scheduled or on-demand jobs.
• Gathering upstream weather forecasts and observations from a wide variety of sources.
• Processing data for long to medium-term storage.
• Making results available through APIs.
• Making data files available to custom post-processors.

Much effort is put into starting processing as soon as the required data becomes available and in parallel when possible.

Recent additions to the WOD system include the potential of:

• Optional use of the hybrid data assimilation techniques of the WRF Data Assimilation system [1, 2].
• Set up a multi-domain dispersion forecast of volcanic ash and gases.
• Use of the Verif [3] verification package to compare forecasts, both upstream and WOD, to observations.
• Using different sources of initial data to that of the boundary forcing data.

On-going developments focuses on the use of in-situ UAV profiles and radar data as input to the WOD data assimilation system.

We have further started experimenting with using global models, both conventional NWP models as well as novel ML models (cf. abstract no. EGU24-15614).

References:

[1] Xuguang Wang, Dale M. Barker, Chris Snyder, and Thomas M. Hamill, 2008: A hybrid ETKF–3DVAR data assimilation scheme for the WRF model. Part I: Observing system simulation experiment. Mon. Wea. Rev., 136, 5116–5131.

[2] Xuguang Wang, Dale M. Barker, Chris Snyder, and Thomas M. Hamill, 2008: A Hybrid ETKF–3DVAR Data Assimilation Scheme for the WRF Model. Part II: Real Observation Experiments. Mon. Wea. Rev., 136, 5132–5147.

[3] Nipen, T. N., R. B. Stull, C. Lussana, and I. A. Seierstad, 2023. Verif: A Weather-Prediction Verification Tool for Effective Product Development. Bulletin of the American Meteorological Society 104, 9; 10.1175/BAMS-D-22-0253.1.

How to cite: Rognvaldsson, O. and Stanislawska, K.: The Weather On Demand weather forecast framework - Recent developments and outlook, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-16714, https://doi.org/10.5194/egusphere-egu24-16714, 2024.

The prediction of excessive rainfall using numerical weather prediction (NWP) models is unequivocally difficult owing to the myriad of complexities that must be resolved (e.g., parent storm dynamics, microphysics) in order to forecast the placement and intensity of rainfall correctly. However, machine learning (ML) has provided a new avenue by which we can generate predictions of excessive rainfall with sufficient lead time to inform decision makers and planners to the threat of inclement weather. ML techniques are able to decode known long-standing relationships between environmental predictors and convective hazards from long historical records, and they have demonstrated tremendous value in predicting weather hazards at longer lead times (e.g., Hill et al. 2023). Further, continued effort by the meteorological community to explain ML models and their forecasts is building trust between developers and end users. As a result, their use in meteorological hazard forecasting is expanding, particularly into the medium range (e.g., 4-8 days) when forecasters are reliant on relatively coarse NWP models to create forecasts.

In this work, we are using Random Forests (RFs) to generate daily probabilistic forecasts of excessive rainfall at 1-8 day lead times. The RFs are trained using output from the Global Ensemble Forecast System and historical observations of excessive rainfall. Environmental parameters like precipitable water and CAPE, as well as modeled precipitation, are spatiotemporally arranged so the RFs can learn spatial and diurnal patterns that associate with excessive rainfall. The RF models are evaluated against a spatio-temporally varying climatology and show skill out to 7 days, and routinely outperform human-based forecasts past a 1-day lead time. In this presentation, we will highlight performance characteristics of the RFs into the medium-range (e.g., out to 8 days) and discuss the implications of excessive rainfall definitions in RF model training. Additionally, we will present an ensemble prediction framework that provides estimates of uncertainty and ranges of forecast solutions that operational forecasters desire at extended lead times.

How to cite: Hill, A. and Schumacher, R.: Medium-Range Excessive Rainfall Prediction with Machine Learning, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-11708, https://doi.org/10.5194/egusphere-egu24-11708, 2024.

Atmospheric predictability is intrinsically limited by the upscale growth of initial small-scale, small-amplitude errors. For practical predictability, model error and initial-condition uncertainty also contribute significantly. The accurate representation and interactions of these factors within numerical weather prediction systems determine the extent to which forecast uncertainty is correctly modeled. An improved understanding of upscale error-growth mechanisms and their flow dependence in numerical weather prediction models has several implications: it enables more focused model verification and development, aids in recognizing limitations in emerging forecasts systems like machine-learning-based approaches, and may indicate when the intrinsic limit of predictability has been reached.

Studying the flow dependence of error growth requires a local perspective, which is not provided by the traditional spectral perspective on upscale error growth. We here take a complementary approach and apply a feature-based perspective. We have developed an automated algorithm to identify error features in gridded data and track their spatial and temporal evolution. Errors are considered in terms of potential vorticity (PV) and near the tropopause, where they maximize. A previously derived PV-error tendency equation is evaluated to quantify the different contributions to error-growth experiments with the global prediction Model ICON from the German Weather Service. Errors in these experiments grow from differences in the seeding of a stochastic convection scheme. In a suite of experiments, this source of uncertainty competes with initial-condition uncertainty of varying magnitude. Evaluation of the process-specific error-growth rates allow the detailed quantification of the upscale-growth mechanisms. For this purpose, we integrate the growth rates over the respective area associated with an error feature. Examination of the combined growth rates of all features in an upscale-error-growth experiment reproduces a previously found three-stage multi-scale upscale-growth paradigm. Illustration the importance of flow dependence, the growth rates from a single feature can substantially differ from the overall average. Further highlighting this importance, intrinsic limits of predictivity can be identified for some features even in the presence of substantial initial-condition uncertainty. The presentation will conclude with a comparison of error evolution in conventional numerical weather prediction systems to a data-driven, machine-learned model.

How to cite: Schmidt, S., Riemer, M., and Selz, T.: A feature-based framework to investigate atmospheric predictability., EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17438, https://doi.org/10.5194/egusphere-egu24-17438, 2024.

Ensemble forecasts play a pivotal role in weather prediction, providing valuable insights into the inherent uncertainty of atmospheric processes. Strategies in ensemble construction involve generating multiple simulations by perturbing initial conditions, model parameters, or both. This diverse set of forecasts allows meteorologists to capture a range of possible future scenarios, acknowledging the inherent complexity of the atmosphere. Model resolution is a critical factor, influencing the representation of small-scale features and improving the overall accuracy of ensemble predictions. Additionally, forecast range-related issues address the challenge of extending predictions beyond a few days, where uncertainties tend to grow. Combining advanced statistical techniques with cutting-edge modeling technologies helps refine ensemble forecasts, enhancing our ability to anticipate and mitigate the impacts of weather-related events on society and the environment.

The investigation based operational global ensemble forecast system from NCEP, CMC, ECMWF and CMA to focus on the analyses of ensemble design that combined to the data assimilation for initial condition perturbation and various stochastic physical perturbations. The impact of model resolutions (both horizontal and vertical) will be addressed to the different atmospheric characteristics, such as forecast uncertainty, reliability and resolution. The forecast capability and predictability to the extreme events will be discussed from single model ensemble and multi-model ensemble. Finally, the 1st-moment and 2nd-moment ensemble forecast calibration will be demonstrated from traditional statistical method and machine learning based ensemble reforecasts

How to cite: Zhu, Y.: An assessment of prediction and predictability through the state-of-the-art global ensemble forecast systems, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6802, https://doi.org/10.5194/egusphere-egu24-6802, 2024.

Irrigation is one agricultural practice that contributes to maintain an optimal soil water content for crop development. Currently, farmers find this practice as an essential method for adapting to climate change. The Earth science community identified some irrigation effects beyond soil moisture and plant growth impact, as multiple studies found an influence on atmospheric variables such as 2 m temperature, relative humidity and even precipitation. Moreover, the effect of irrigation on the Earth’s system has been studied on various temporal and geographical scales and with different climate and land surface models. However, there are few studies that simulated the effect of irrigation on higher resolutions on a regional scale. Therefore, the aim of this study is to include the representation of irrigation processes in the operational ICON-nwp in Limited Area Mode on the EURO-CORDEX domain. The implementation of the current irrigation parameterization in ICON-nwp coupled with TERRA is an adaptation of the CHANNEL scheme developed by Valmassoi et al. (2020). We found suitable to include this scheme in the land surface and atmosphere interface of the icon-nwp-2.6.6-nwp0 version. The present study consists of four sensitivity experiments with different irrigation water amounts, namely 2.6 mmd-1, 6.7 mmd-1 and two fixed soil water contents, field capacity and saturation. All experiments have the same irrigation frequency (1 day), length (24 hours), and simulation period (May to August). The model settings for the experiments are 3 km resolution, 75 vertical levels and ICON boundary and initial conditions. The results from the difference between experiments and the control run demonstrate that ICON captures the irrigation effect on land surface atmospheric variables. As expected, soil moisture content increased on different magnitudes in all experiments. Moreover, 2 m temperature values dropped on average -0.74 K in irrigated areas. Likewise, energy fluxes were sensible to the different irrigation amounts.

How to cite: Roque, J., Valmassoi, A., and Keller, J.: Irrigation parameterization in the Operational Numerical Weather Prediction model ICON-nwp, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-20366, https://doi.org/10.5194/egusphere-egu24-20366, 2024.

In 2016, the U. S. Congress created the Joint Technology Transfer Initiative (JTTI) program in the Office of Oceanic and Atmospheric Research (OAR), the research wing of the National Oceanic Atmospheric Administration (NOAA).  Within OAR, the Weather Program Office (WPO) is responsible for managing the JTTI program.  The main mission of this program is to continuously develop and transition the mature weather technologies from the research community to the National Weather Service (NWS) operations.  

JTTI selects promising Research to Operations (R2O) transition projects through two types of competitions: one for the external community (non-NOAA) that includes private, academic sectors and non-profit organizations through Notices of Funding Opportunities; and the other for the NOAA scientific community.  Additionally, the JTTI program collaborates with the NWS Office of Science and Technology Integration (OSTI) and provides funding for the Unified Forecasting System - R2O project and testbed activities.  JTTI-funded R2O projects cover three main frameworks within the NWS forecasting operations: the observational, modeling, and products and services frameworks.  The topics covered include data assimilation; convective scale weather modeling; stochastic physics; ensemble model building; hydrologic modeling; post-processing of model output on time scales ranging from hourly to subseasonal; high impact weather forecasting tools; artificial intelligence/machine learning; and social behavioral and economic science. To date, the JTTI program has funded 155 R2O projects and transitioned 20 projects to the NWS operations.  In this paper, we present the JTTI implementation process in NOAA and share some of the successful R2O stories.

How to cite: Kondragunta, C., Pratt, A., Garrett, K., Kurkowski, N., Sellers, W., and Kunkel, V.: Crossing the Valley of Death : Transitioning Weather Research to Operations in NOAA, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-20553, https://doi.org/10.5194/egusphere-egu24-20553, 2024.