AS1.2 | Forecasting the weather
EDI
Forecasting the weather
Co-organized by NH1/NP5
Convener: Yong Wang | Co-conveners: Aitor Atencia, kan dai, Lesley De Cruz, Daniele NeriniECSECS
Orals
| Mon, 15 Apr, 14:00–15:45 (CEST), 16:15–18:00 (CEST)
 
Room 0.11/12
Posters on site
| Attendance Tue, 16 Apr, 10:45–12:30 (CEST) | Display Tue, 16 Apr, 08:30–12:30
 
Hall X5
Orals |
Mon, 14:00
Posters on site
Tue, 10:45
Forecasting the weather, in particular severe and extreme weather has always been the most important subject in meteorology. This session will focus on recent research and developments on forecasting techniques, in particular those designed for operations and impact oriented. Contributions related to nowcasting, meso-scale and convection permitting modelling, ensemble prediction techniques, and statistical post-processing are very welcome.
Topics may include:
 Nowcasting methods and systems, use of observations and weather analysis
 Mesoscale and convection permitting modelling
 Remote sensing and data assimilation
 Ensemble prediction techniques
 Ensemble-based products for severe/extreme weather forecasting
 Seamless deterministic and probabilistic forecast prediction
 Post-processing techniques, statistical methods in prediction
 Use of machine learning, data mining and other advanced analytical techniques
 Impact oriented weather forecasting
 Presentation of results from relevant international research projects of EU, WMO, and EUMETNET etc.

Key Words: Forecast technique, nowcasting, ensemble prediction, statistics, AI

Orals: Mon, 15 Apr | Room 0.11/12

Chairpersons: Yong Wang, Aitor Atencia
14:00–14:05
5-minute convener introduction
14:05–14:35
|
EGU24-13853
|
solicited
|
Highlight
|
Virtual presentation
A Research Agenda for the Evaluation of AI-Based Weather Forecasting Models 
Imme Ebert-Uphoff, Jebb Q. Stewart, and Jacob T. Radford and the CIRA-NOAA team

Over the past few years purely AI-driven global weather forecasting models have emerged that show increasingly impressive skill, raising the question whether AI models might soon compete with NWP models for selected forecasting tasks. At this point these AI-based models are still in the proof-of-concept stage and not ready to be used for operational forecasting, but entirely new AI-models emerge every 2-3 months, with rapidly increasing abilities. Furthermore, many of these models are orders of magnitude faster than NWP models and can run on modest computational resources enabling repeatable on-demand forecasts competitive with NWP. The low computational cost enables the creation of very large ensembles, which better represent the tails of the forecast distribution, which, if an ensemble is well calibrated, allows for better forecasting of rare and extreme events.

However, these AI-based weather forecasting models have not yet been rigorously tested by the meteorological community, and their utility to operational forecasters is unknown. In this presentation we propose several studies to address the above issues, grouped into two central foci:

(1) Nature of AI models: AI-based models have very different characteristics from NWP models. Thus, in addition to applying evaluation procedures developed for NWP models, we need to develop procedures that test for AI-specific weaknesses. For example, NWP models and their physics backbone guarantee certain properties - such as dynamic coupling between fields - that AI-based models are not required to uphold. Developing suitable tests is based on a fundamental understanding of the AI-based models.

(2) Forecaster Perspective: Evaluation of weather forecasting models should be performed with respect to particular applications of weather forecasts, and it is critical to have research meteorologists and operational forecasters involved in the evaluation process. Our initial evaluation of AI-based models in CIRA weather briefings revealed that these models have characteristics that make interpretation of their forecasts fundamentally different from the physics-based NWP model predictions meteorologists are familiar with. For example, the increasing “blurriness” of AI-based predictions with longer lead times is not a reflection of weaker atmospheric circulations, but rather a reflection of uncertainty. Evaluations aimed at specific meteorological phenomena and atmospheric processes will allow the community to make informed decisions in the future regarding in what environments and for which applications AI-based weather forecasting models may be safe and beneficial to use.

In summary, AI-based weather forecasts have different characteristics from familiar dynamically-based forecasts, and it is thus important to have a robust research plan to evaluate many different characteristics of the models in order to provide guidelines to operational forecasters and feedback to model developers. In this abstract we propose a number of characteristics to evaluate, present results we already obtained, and suggest a research plan for future work.

How to cite: Ebert-Uphoff, I., Stewart, J. Q., and Radford, J. T. and the CIRA-NOAA team: A Research Agenda for the Evaluation of AI-Based Weather Forecasting Models, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-13853, https://doi.org/10.5194/egusphere-egu24-13853, 2024.

14:35–14:45
|
EGU24-5373
|
ECS
|
Highlight
|
On-site presentation
Convective environments in AI-models - What have AI-models learned about atmospheric profiles? 
Monika Feldmann, Louis Poulain-Auzeau, Milton Gomez, Tom Beucler, and Olivia Martius
The recently released suite of AI-based medium-range forecast models can produce multi-day forecasts within seconds, with a skill on par with the IFS model of ECMWF. Traditional model evaluation predominantly targets global scores on single levels. Specific prediction tasks, such as severe convective environments, require much more precision on a local scale and with the correct vertical gradients in between levels. With a focus on the North American and European convective season of 2020, we assess the performance of Panguweather, Graphcast and Fourcastnet for convective available potential energy (CAPE) and storm relative helicity (SRH) at lead times of up to 7 days.
Looking at the example of a US tornado outbreak on April 12 and 13, 2020, all models predict elevated CAPE and SRH values multiple days in advance. The spatial structures in the AI-models are smoothed in comparison to IFS and the reanalysis ERA5. The models show differing biases in the prediction of CAPE values, with Graphcast capturing the value distribution the most accurately and Fourcastnet showing a consistent underestimation.
By advancing the assessment of large AI-models towards process-based evaluations we lay the foundation for hazard-driven applications of AI-weather-forecasts.

How to cite: Feldmann, M., Poulain-Auzeau, L., Gomez, M., Beucler, T., and Martius, O.: Convective environments in AI-models - What have AI-models learned about atmospheric profiles?, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5373, https://doi.org/10.5194/egusphere-egu24-5373, 2024.

14:45–14:55
|
EGU24-15431
|
ECS
|
Highlight
|
On-site presentation
Nowcasting of extreme precipitation events: performance assessment of Generative Deep Learning methods 
Gabriele Franch, Elena Tomasi, Rishabh Umesh Wanjari, and Marco Cristoforetti

Radar-based precipitation nowcasting is one of the most prominent applications of deep learning (DL) in weather forecasting. The accurate forecast of extreme precipitation events remains a significant challenge for deep learning models, primarily due to their complex dynamics and the scarcity of data on such events. In this work we present the application of the latest state-of-the-art generative architectures for radar-based nowcasting, focusing on extreme event forecasting performance. We analyze a declination for the nowcasting task of all the three main current architectural approaches for generative modeling, namely: Generative Adversarial Networks (DGMRs), Latent Diffusion (LDCast), and our novel proposed Transformer architecture (GPTCast). These models are trained on a comprehensive 1-km scale, 5-minute timestep radar precipitation dataset that integrates multiple radar data sources from the US, Germany, the UK, and France. To ensure a robust evaluation and to test the generalization ability of the models, we concentrate on a collection of out-of-domain extreme precipitation events over the Italian peninsula extracted from the last 5 years. This focus allows us to assess the improvements these techniques offer compared to extrapolation methods, evaluating continuous (MSE, MAE) and categorical scores (CSI, POD, FAR), ensemble reliability, uncertainty quantification, and warning lead time. Finally, we analyze the computational requirements of these new techniques and highlight the caveats that must be considered when operational usage of these methods is envisaged. 

How to cite: Franch, G., Tomasi, E., Wanjari, R. U., and Cristoforetti, M.: Nowcasting of extreme precipitation events: performance assessment of Generative Deep Learning methods, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-15431, https://doi.org/10.5194/egusphere-egu24-15431, 2024.

14:55–15:05
|
EGU24-7536
|
On-site presentation
Nowcasting with Transformer-based Models using Multi-Source Data  
Çağlar Küçük, Apostolos Giannakos, Stefan Schneider, and Alexander Jann

Rapid advancements in data-driven weather prediction have shown notable success, particularly in nowcasting, where forecast lead times span just a few hours. Transformer-based models, in particular, have proven effective in learning spatiotemporal connections of varying scales by leveraging the attention mechanism with efficient space-time patching of data. This offers potential improvements over traditional nowcasting techniques, enabling early detection of convective activity and reducing computational costs. 

In this presentation, we demonstrate the effectiveness of a modified Earthformer model, a space-time Transformer framework, in addressing two specific nowcasting challenges. First, we introduce a nowcasting model that predicts ground-based 2D radar mosaics up to 2-hour lead time with 5-minute temporal resolution, using geostationary satellite data from the preceding two hours. Trained on a benchmark dataset sampled across the United States, our model exhibits robust performance against various impactful weather events with distinctive features. Through permutation tests, we interpret the model to understand the effects of input channels and input data length. We found that the infrared channel centered at 10.3 µm contains skillful information for all weather conditions, while, interestingly, satellite-based lightning data is the most skilled at predicting severe weather events in short lead times. Both findings align with existing literature, enhancing confidence in our model and guiding better usage of satellite data for nowcasting. Moreover, we found the model is sensitive to input data length in predicting severe weather events, suggesting early detection of convective activity by the model in rapidly growing fields. 

Second, we present the initial attempts to develop a multi-source precipitation nowcasting model for Austria, tailored to predict impactful events with convective activities. This model integrates satellite- and ground-based observations with analysis and numerical weather prediction data to predict precipitation up to 2-hour lead time with 5-minute temporal resolution.  

We conclude by discussing the broad spectrum of applications for such models, ranging from enhancing operational nowcasting systems to providing synthetic data to data-scarce regions, and the challenges therein.

How to cite: Küçük, Ç., Giannakos, A., Schneider, S., and Jann, A.: Nowcasting with Transformer-based Models using Multi-Source Data , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7536, https://doi.org/10.5194/egusphere-egu24-7536, 2024.

15:05–15:15
|
EGU24-19257
|
ECS
|
On-site presentation
Dynamic Locally Binned Density Loss 
Jan Prosi, Sebastian Otte, and Martin V. Butz

In the field of precipitation nowcasting recent deep learning models now outperform traditional approaches such as optical flow [1,2]. Despite their principled effectiveness, these models and their respective training setups suffer from particular shortcomings.  For instance, they often rely on pixel-wise losses, which lead to blurred predictions by which the model expresses its uncertainty [2]. Additionally, these losses can negatively impact training dynamics by overly penalizing small spatial or temporal discrepancies between predictions and actual observations, i.e., the double penalty problem [3]. Generative methods such as discriminative losses or diffusion models do not suffer from the blurring effect as much [1, 4]. However, training these methods is complicated because training success is highly sensitive to the network architecture as well as to the learning setup and its parameterization [5].

Previous research has shown that spatial verification methods such as the fractions skill score offer an easy-to-implement alternative to solve the problem of pixel-wise losses [6, 7]. However, the fact that each pixel within the neighborhood of a spatial kernel is weighted equally poses a limiting factor to their performance and potential. Inspired by theories of cognitive modeling and in relation to the fractions skill score loss, we introduce a dynamic locally binned density (DLBD) loss: Forecasting target is not the actual precipitation in a grid cell but a target distribution, which encodes the density of binned precipitation values in a locally weighted area of grid cells. The loss is then determined via the cross-entropy of the predicted and the target distribution. We show that our novel prediction loss avoids the double penalty problem.  It thus diminishes the negative impact of small spatial offsets. Moreover, it enables the learning model to gradually shift focus towards progressively more accurate predictions.

We achieve best performance by simultaneously training on multiple concurrent forecasting targets that cover different local extents. We schedule the weighting of the loss terms such that the focus shifts from larger to smaller neighborhoods over the course of training. This way, the DL model first learns density dynamics and basic precipitation shifts. Later, it focuses on minimizing small spatial deviations, tuning into the local dynamics towards the end of training.  Our DLBD loss is easy-to-implement and shows great performance improvements.  We thus believe that DLBD losses can also be used by other forecasting architectures where the current forecasting loss precludes smooth loss landscapes.

 


1: Leinonen et al. 2023: Latent diffusion models for generative precipitation nowcasting with accurate uncertainty quantification
2: Espeholt et al. 2022: Deep learning for twelve hour precipitation forecasts
3: Grilleland et al. 2009: Intercomparison of spatial forecast verification methods.
4: Ravuri et al. 2021: Skilful precipitation nowcasting using deep generative models of radar
5: Mescheder et al. 2018: Which training methods for GANs do actually converge?
6: Roberts et al. 2008: Scale-selective verification of rainfall accumulations from high resolution forecasts of convective events.
7: Lagerquist et al. 2022: Can we integrate spatial verification methods into neural-network loss functions for atmospheric science?

How to cite: Prosi, J., Otte, S., and Butz, M. V.: Dynamic Locally Binned Density Loss, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19257, https://doi.org/10.5194/egusphere-egu24-19257, 2024.

15:15–15:25
|
EGU24-5571
|
ECS
|
On-site presentation
SHADECast: Enhancing solar energy integration through probabilistic regional forecasts 
Alberto Carpentieri, Doris Folini, Jussi Leinonen, and Angela Meyer

Surface solar irradiance (SSI) is a pivotal component in addressing climate change. As an abundant and non-fossil energy source, it is harnessed through photovoltaic (PV) energy production. As the contribution of PV to total energy production grows, the stability of the power grid faces challenges due to the volatile nature of solar energy, predominantly influenced by stochastic cloud dynamics. To address this challenge, there is a need for accurate, uncertainty-aware, near real-time, and regional-scale SSI forecasts with forecast horizons ranging from minutes to a few hours.

Existing state-of-the-art SSI nowcasting methods only partially meet these requirements. In our study, we introduce SHADECast [1], a deep generative diffusion model designed for probabilistic nowcasting of cloudiness fields. SHADECast is uniquely structured, incorporating deterministic aspects of cloud evolution to guide the probabilistic ensemble forecast, relying only on previous satellite images. Our model showcases significant advancements in forecast quality, reliability, and accuracy across various weather scenarios.

Through comprehensive evaluations, SHADECast demonstrates superior performance, surpassing the state of the art by 15% in the continuous ranked probability score (CRPS) over diverse regions up to 512 km × 512 km, extending the state-of-the-art forecast horizon by 30 minutes. The conditioning of ensemble generation on deterministic forecasts further enhances reliability and performance by more than 7% on CRPS.

SHADECast forecasts equip grid operators and energy traders with essential insights for informed decision-making, thereby guaranteeing grid stability and facilitating the smooth integration of regionally distributed PV energy sources. Our research contributes to the advancement of sustainable energy practices and underscores the significance of accurate probabilistic nowcasting for effective solar power grid management.

 

References

[1] Carpentieri A. et al., 2023, Extending intraday solar forecast horizons with deep generative models. Preprint at ArXiv. https://arxiv.org/abs/2312.11966 

How to cite: Carpentieri, A., Folini, D., Leinonen, J., and Meyer, A.: SHADECast: Enhancing solar energy integration through probabilistic regional forecasts, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5571, https://doi.org/10.5194/egusphere-egu24-5571, 2024.

15:25–15:35
|
EGU24-19321
|
ECS
|
On-site presentation
Viability of satellite derived irradiance data for ML-based nowcasts 
Pascal Gfäller, Irene Schicker, and Petrina Papazek

Photovoltaic (PV) power production is increasingly becoming a central pillar in the shift to renewable power sources. The use of solar irradiance has great potential, as it is practically limitless and globally provides magnitudes more energy to the Earth than currently or foreseeable required. Solar irradiance as a power source does, however come with certain downsides. Besides the effects of seasonality and day-night-cycles on its usable potential, it´s broad use suffers mostly from uncertainty through its volatility. The actual extent of solar irradiance at the surface of the Earth is strongly influenced by a variety of atmospheric phenomena, most prominently clouds and atmospheric turbidity. The forecasting of near-future solar irradiance can thereby be beneficial in the estimation of PV power production in itself and with the goal of maintaining a stable equilibrium in electrical grids.

To achieve nowcasts on a larger grid scope, forecasting of solar irradiance from satellite data can substitute forecasting of power output for individual sites. Satellite data, in contrast to ground-based data sources or NWP model estimates, is less reliant on the proper workings of a wide range of externalities. General-purpose spatiotemporal neural networks can be adapted to this task and provide predictions within a very short timeframe, with no requirement of HPC-infrastructure. A sparse model relying on a single satellite-based data source has less points of failure that could affect its forecasting performance and can be very efficient, but this sparsity could also reduce the achievable predictive accuracy. Benefits of smaller and simpler forecasting pipelines therefore may need to be balanced with requirements in terms of accuracy.

To gather more meaningful and reliable results, a variety of spatiotemporal neural networks is implemented and tested to provide a more meaningful foundation. The models were selected and evaluated with respect to their different architectural patterns and designs, to get a notion of architectures beneficial to this task and achieve a more generalizable argument concerning the use satellite data as the sole basis of solar irradiance nowcasting.

In an attempt of improving the viability of satellite-based nowcasting a commonly occurring flaw in near-real-time satellite data sources, missing or skipped frames, solutions to mitigate issues in operational nowcasting are considered. In place of ad-hoc preprocessing such as interpolation of missing data frames, an attempt to condition the models to missing frames is made.

How to cite: Gfäller, P., Schicker, I., and Papazek, P.: Viability of satellite derived irradiance data for ML-based nowcasts, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19321, https://doi.org/10.5194/egusphere-egu24-19321, 2024.

15:35–15:45
|
EGU24-8449
|
ECS
|
On-site presentation
Radiation fog nowcasting with XGBoost using station and satellite data 
Michaela Schütz, Jörg Bendix, and Boris Thies

The research project “FOrecasting radiation foG by combining station and satellite data using Machine Learning (FOG-ML)” represents a comprehensive effort to advance radiation fog prediction using machine learning (ML) techniques, with focus on the XGBoost algorithm. The nowcasting period is up to four hours into the future.

The initial phase of the project involved developing a robust classification-based model that could accurately forecast the occurrence of radiation fog, a challenging meteorological phenomenon. Radiation fog is particularly difficult to predict because it depends on a complex interplay of factors such as ground cooling, humidity, and minimal cloud cover. It often forms rapidly and in local areas. This required careful analysis of the chronological order of the data and consideration of autocorrelation to increase the effectiveness of model training.

Building upon this foundation, the next two phases concentrated on improving the model’s forecasting performance for visibility classes (step 2) and for absolute visibility values (step 3). The main focus was then on a nowcasting period of up to two hours. This nowcasting period is critical in fog prediction as it directly impacts transportation planning and safety. The use of ground-level observations in step 2 and integration of satellite data in step 3 provided a rich dataset that allowed for more nuanced model training and validation.

In the latest phase of research, satellite data has been incorporated to further refine the prediction model, especially regarding the fog formation and dissipation. Satellite imagery provides additional variables of atmospheric data that are not readily available from ground-based observations. This integration aims to address one of the inherent limitations in fog forecasting methods, particularly in areas where ground-based observations are sparse.

Throughout the different stages, the project emphasized the need for thorough data processing and validation. This included the implementation of cross-validation techniques to assess the generalizability of the models and the use of various metrics to gauge their predictive power. This has also included the incorporation of trend information, which has proven to be crucial for forecasting with XGBoost. Our research has also shown that not only the overall performance, but also the performance of the transitions (fog formation and resolution) should be analyzed to get a complete picture of the model performance. This finding was consistent throughout the entire study, regardless of classification-based forecast or regression-based forecast.

We have been able to significantly improve the performance of our nowcasting model with each step. We will be presenting the key findings and latest results from this research at EGU24.

All results from step 1 can be found in “Current Training and Validation Weaknesses in Classification-Based Radiation Fog Nowcast Using Machine Learning Algorithms” from Vorndran et al. 2022. All results from step 2 can be found in “Improving classification-based nowcasting of radiation fog with machine learning based on filtered and preprocessed temporal data” from Schütz et al. 2023.

How to cite: Schütz, M., Bendix, J., and Thies, B.: Radiation fog nowcasting with XGBoost using station and satellite data, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-8449, https://doi.org/10.5194/egusphere-egu24-8449, 2024.

Coffee break
Chairpersons: Daniele Nerini, Lesley De Cruz
16:15–16:45
|
EGU24-17158
|
solicited
|
Highlight
|
On-site presentation
AIFS – ECMWF’s Data-Driven Probabilistic Forecasting  
Zied Ben Bouallegue, Mihai Alexe, Matthew Chantry, Mariana Clare, Jesper Dramsch, Simon Lang, Christian Lessig, Linus Magnusson, Ana Prieto Nemesio, Florian Pinault, Baudouin Raoult, and Steffen Tietsche

In just two years, the idea of an operational data-driven system for medium-range weather forecasting has been transformed from dream to very real possibility. This has occurred through a series of publications from innovators, which have rapidly improved deterministic forecast skill. Our own evaluation confirms that these forecasts have comparable deterministic skill to NWP models across a range of variables. However, on medium-range timescales probabilistic forecasting, typically achieved through ensembles, is key for providing actionable insights to users. ECMWF is building on top of these recent works to develop a probabilistic forecasting system, AIFS. We will showcase results from our progress towards this system and outline our roadmap to operationalisation.

How to cite: Ben Bouallegue, Z., Alexe, M., Chantry, M., Clare, M., Dramsch, J., Lang, S., Lessig, C., Magnusson, L., Prieto Nemesio, A., Pinault, F., Raoult, B., and Tietsche, S.: AIFS – ECMWF’s Data-Driven Probabilistic Forecasting , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17158, https://doi.org/10.5194/egusphere-egu24-17158, 2024.

16:45–16:55
|
EGU24-6155
|
ECS
|
On-site presentation
Enhanced Foundation Model through Efficient Finetuning for Extended-Range Weather Prediction 
Shan Zhao, Zhitong Xiong, and Xiao Xiang Zhu

Weather forecasting is a vital topic in meteorological analysis, agriculture planning, disaster management, etc. The accuracy of forecasts varies with the prediction horizon, spanning from nowcasting to long-range forecasts. The extended range forecast, which predicts weather conditions beyond two weeks to months ahead, is particularly challenging. This difficulty arises from the inherent variability in weather systems, where minor disturbances in the initial condition can lead to significantly divergent future trajectories.

Numerical Weather Prediction (NWP) has been the predominant approach in this field. Recently, deep learning (DL) techniques have emerged as a promising alternative, achieving performance comparable to NWP [1, 2]. However, their lack of embedded physical knowledge often limits their acceptance within the research community. To enhance the trustworthiness of DL-based weather forecasts, we explore a transformer-based framework which considers complex geospatial-temporal (4D) processes and interactions. Specifically, we select the Pangu model [3] with a 24-hour lead time as the initial framework. To extend the prediction horizon to two weeks ahead, we employ a low-rank adaptation for model finetuning, which saves computation resources by reducing the number of parameters to only 1.1% of the original model. Besides, we incorporate multiple oceanic and atmospheric indices to capture a broad spectrum of global teleconnections, aiding in the selection of important features.

Our contributions are threefold: first, we provide an operational framework for foundation models, improving their applicability in versatile tasks by enabling training rather than limiting them to inference stages. Second, we demonstrate how to leverage these models with limited resources effectively and contribute to the development of green AI. Last, our method improves performance in extended-range weather forecasting, offering enhanced prediction skills, physical consistency, and finer spatial granularity. Our methodology achieved reduced RMSE on T2M, Z500, and T850 for 0.13, 139.2, and 0.52, respectively, compared to IFS. In the future, we plan to explore other settings, such as predicting precipitation and extreme temperatures.

REFERENCES
[1] Nguyen, Tung, et al. "ClimaX: A foundation model for weather and climate." arXiv preprint arXiv:2301.10343 (2023).
[2] Lam, Remi, et al. "Learning skillful medium-range global weather forecasting." Science (2023): eadi2336.
[3] Bi, Kaifeng, et al. "Accurate medium-range global weather forecasting with 3D neural networks." Nature 619.7970 (2023): 533-538.

How to cite: Zhao, S., Xiong, Z., and Zhu, X. X.: Enhanced Foundation Model through Efficient Finetuning for Extended-Range Weather Prediction, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6155, https://doi.org/10.5194/egusphere-egu24-6155, 2024.

16:55–17:05
|
EGU24-14325
|
Virtual presentation
Improving NWP post-processing at the Bureau of Meteorology 
(withdrawn)
Benjamin Owen, James Canvin, Thomas Gale, Timothy Hume, Robert Johnson, Daniel Mentiplay, Anja Schubert, Belinda Trotta, and Jennifer Whelan
17:05–17:15
|
EGU24-9528
|
ECS
|
On-site presentation
Ensemble forecast post-processing based on neural networks and normalizing flows 
Peter Mlakar, Janko Merše, and Jana Faganeli Pucer

Ensemble weather forecast post-processing can generate more reliable probabilistic weather forecasts compared to the raw ensemble. Often, the post-processing method models the future weather probability distribution in terms of a pre-specified distribution family, which can limit their expressive power. To combat these issues, we propose a novel, neural network-based approach, which produces forecasts for multiple lead times jointly, using a single model to post-process forecasts at each station of interest. We use normalizing flows as parametric models to relax the distributional assumption, offering additional modeling flexibility.We evaluate our method for the task of temperature post-processing on the EUPPBench benchmark dataset. We show that our approach exhibits state-of-the-art performance on the benchmark, improving upon other well-performing entries. Additionally, we analyze the performance of different parametric distribution models in conjunction with our parameter regression neural network, to better understand the contribution of normalizing flows in the post-processing context. Finally, we provide a possible explanation as to why our method performs well, exploring per-lead time input importance.

How to cite: Mlakar, P., Merše, J., and Faganeli Pucer, J.: Ensemble forecast post-processing based on neural networks and normalizing flows, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-9528, https://doi.org/10.5194/egusphere-egu24-9528, 2024.

17:15–17:25
|
EGU24-7753
|
ECS
|
On-site presentation
On the usefulness of considering the run-to-run variability for an ensemble prediction system 
Hugo Marchal, François Bouttier, and Olivier Nuissier

The run-to-run variability of numerical weather prediction systems is at the heart of forecasters' concerns, especially in the decision-making process when high-stakes events are considered. Indeed, forecasts that are brutally changing from one run to another may be difficult to handle and can lose credibility. This is all the more true nowadays, as many meteorological centres have adopted the strategy of increasing runs frequency, some reaching hourly frequencies. However, this aspect has received little attention in the literature, and the link with predictability has barely been explored.

In this study, run-to-run variability is investigated through 24h-accumulated precipitations forecasted by AROME-EPS, Météo-France's high resolution ensemble, which is refreshed 4 times a day. Focusing on the probability of some (warning) thresholds being exceeded, results suggest that how forecasts evolve over successive runs can be used to improve their skill, especially reliability. Various possible aspects of run sequence have been studied, from trends to rapid increases or decreases in event probability at short lags, also called "sneaks" or "phantoms", as well as the persistence of a non-zero probability through successive runs. The added value provided by blending successive runs, known as lagging, is also discussed.

How to cite: Marchal, H., Bouttier, F., and Nuissier, O.: On the usefulness of considering the run-to-run variability for an ensemble prediction system, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7753, https://doi.org/10.5194/egusphere-egu24-7753, 2024.

17:25–17:35
|
EGU24-21772
|
On-site presentation
Calibration of Convective-scale Hourly Precipitation Based on the Frequency-Matching Method 
Xiaoshi Qiao, Shizhang Wang, and Mingjian Zeng

Calibration of convective-scale hourly precipitation based on the frequency-matching method was carried on using CMPASS observation and CMA-MESO 3km forecast data. The character of hourly precipitation bias was studied.The effect of frequency-matching method (FMM) on the bias correction of CMA-MESO 3km hourly precipitation forecasts was analyzed. In the bias characteristic analysis, the differences in precipitation intensity in different regions of the country and the differences in precipitation in different months were considered. The whole country was divided into 7 sub-regions for monthly analysis. In the bias correction based on the frequency-matching method, the daily variations of precipitation bias and the impact of increasing and decreasing precipitation values on the corrected precipitation scores were analyzed. The results show that CMA-MESO 3km forecasts have a wet bias in light rainfall in the cold season, while a dry bias dominates in moderate to heavy rainfall. In the warm season, except for the Tibet region, the hourly precipitation forecast bias of CMA-MESO 3km shows significant daily variations, with more precipitation in the afternoon and less at night and in the morning, especially for heavy rainfall. Therefore, whether to consider the daily variations of precipitation bias in the use of FMM correction mainly reflects in the summer, especially at night and in the morning. Considering the daily variations of precipitation bias is beneficial to improving the forecast skills (TS scores) for nighttime and morning in the summer. Further analysis shows that the positive contribution of FMM correction to forecast scores mainly comes from the increase in frequency adjustment, especially for heavy rainfall. However, for light rainfall with wet bias, FMM often results in negative contribution. Therefore, FMM has a significant improvement effect on heavy rainfall in winter and nighttime rainfall in summer. The reason for this result is that the hit rate of CMA-MESO hourly precipitation forecast is low, and the false alarm rate is generally high, especially for heavy rainfall. In this case, the increased precipitation significantly increases the hit rate, while the false alarm rate increases to a lesser extent, thereby improving the precipitation scores.

How to cite: Qiao, X., Wang, S., and Zeng, M.: Calibration of Convective-scale Hourly Precipitation Based on the Frequency-Matching Method, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-21772, https://doi.org/10.5194/egusphere-egu24-21772, 2024.

17:35–17:45
|
EGU24-12420
|
ECS
|
On-site presentation
Can convection permitting forecasts solve the tropical African precipitation forecasting problem? 
Felix Rein, Andreas H. Fink, Tilmann Gneiting, Philippe Peyrille, James Warner, and Peter Knippertz

Forecasting precipitation over Africa, the largest landmass in the tropics, has been a long standing problem. The unique conditions of the West African monsoon result in large and long lasting mesoscale convective systems. Global numerical weather prediction (NWP) models have gridsizes in the 10s of kilometers, particular when run in ensemble mode, leaving convection to be parameterized. This often results in precipitation being forecast on too large scales, in the wrong places, and with too weak intensity, ultimately leading to little to no skill in tropical Africa.


It has been argued that convection permitting (CP) NWP forecasts would cure some of the problems described above but those have only recently become feasible in an operational setting, although ensembles are still deemed to be too expensive. Here, we systematically compare regional deterministic CP and global ensemble forecasts in the region over multiple rainy seasons for the first time. We analyze CP forecasts from AROME and Met Office Tropical African Model, and seven global ensemble forecasts from the TIGGE archive, both individually and as a multi-model ensemble. In order to create an uncertainty estimate, we create neighborhood ensembles from CP forecasts at surrounding grid points, which allows for a fair comparison to the ensembles and a probabilistic climatology. Considering both precipitation occurrence and amount, we use the Brier score (BS) and the continuous ranked probability score (CRPS), along with their decompositions in discrimination, miscalibration and uncertainty, for evaluation.


Using neighborhood methods, deterministic forecasts are turned into probabilistic forecasts, allowing a fair comparison with ensembles. All numerical forecasts benefit from Neighborhoods, improving their BS and CRPS in terms of both miscalibration and discrimination. We find all individual forecasts to have skill over most of tropical Africa, with some ensemble models lacking skill in some regions and the multi model showing the most overall skill. The CP forecasts TAM and AROME outperform non-CP forecasts mainly in the region of the little dry Season and the Soud. However, large areas of low skill in terms of CRPS remain and even with high resolution, numerical models still struggle to predict precipitation in tropical Africa. 

How to cite: Rein, F., Fink, A. H., Gneiting, T., Peyrille, P., Warner, J., and Knippertz, P.: Can convection permitting forecasts solve the tropical African precipitation forecasting problem?, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-12420, https://doi.org/10.5194/egusphere-egu24-12420, 2024.

17:45–17:55
|
EGU24-21770
|
Virtual presentation
A Local Data Assimilation Method (Local DA v1.0) and its Application in a Simulated Typhoon Case 
Shizhang Wang and Xiaoshi Qiao

Integrating the hybrid and multiscale analyses and the parallel computation is necessary for current data assimilation schemes. A local data assimilation method, Local DA, is designed to fulfill these needs. This algorithm follows the grid-independent framework of the local ensemble transform Kalman filter (LETKF) and is more flexible in hybrid analysis than the LETKF. Local DA employs an explicitly computed background error correlation matrix of model variables mapped to observed grid points/columns. This matrix allows Local DA to calculate static covariance with a preset correlation function. It also allows using the conjugate gradient (CG) method to solve the cost function and allows performing localization in model space, observation space, or both spaces (double-space localization). The Local DA performance is evaluated with a simulated multiscale observation network that includes sounding, wind profiler, precipitable water vapor, and radar observations. In the presence of a small-size time-lagged ensemble, Local DA can produce a small analysis error by combining multiscale hybrid covariance and double-space localization. The multiscale covariance is computed using error samples decomposed into several scales and independently assigning the localization radius for each scale. Multiscale covariance is conducive to error reduction, especially at a small scale. The results further indicate that applying the CG method for each local analysis does not result in a discontinuity issue. The wall clock time of Local DA implemented in parallel is halved as the number of cores doubles, indicating a reasonable parallel computational efficiency of Local DA.

How to cite: Wang, S. and Qiao, X.: A Local Data Assimilation Method (Local DA v1.0) and its Application in a Simulated Typhoon Case, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-21770, https://doi.org/10.5194/egusphere-egu24-21770, 2024.

17:55–18:00
Discussion

Posters on site: Tue, 16 Apr, 10:45–12:30 | Hall X5

Display time: Tue, 16 Apr 08:30–Tue, 16 Apr 12:30
Chairpersons: Aitor Atencia, Lesley De Cruz, Daniele Nerini
X5.1
|
EGU24-7091
The Development of precipitation model modifed with ECMWF IFS and XGBoost and its performance verification 
Eunju Cho, Yeon-Hee Kim, Seungbum Kim, and Young Cheol Kwon

This study was conducted to develop a modified precipitation model for its amount and existence by combining machine learning method, Extreme Gradient Boosting(XGBoost), with ECMWF IFS(Integrated forecasting system) and, finally, estimate the related performance.

According to the analysis of regional precipitation characteristic, prior to its development, the ratio of precipitation existence was various on a basis of a forecast’s district and its season. These different patterns on each district makes it necessary to develop the regional and seasonal model respectively.

And, the first attempt at the machine learning showed the importance of each feature as input-variables, as a result of which cloud physics-related features, for example large-area precipitation, total precipitation, visibility and what not, proved so significant. However, the insufficient amount of these feature’s data seemed to result in overfitting. And therefore, the feasible features, except for cloud physics-related things, of IFS data were used. In addition, auxiliary features and their gradient for every lead-time were calculated and added: relative vorticity, divergence, equivalent potential temperature, main 6 patterns for Korean summer and so on. The number of features amounted to around 144 with which for the 9-year training set, 2013~2021, based learning to be conducted regionally, followed by using validation-set of 2022.

As a result of validation for precipitation existence and its amount up to 135 hours ahead on the 10 regions at 00UTC in summer of 2022, Critical Success Index(CSI) was more improved by 10.3% than before. Accuracy(ACC) for each lead-time rose by 6% and its fluctuation also decreased. And the correction by this machine learning alleviated the overfitting trend of precipitation forecast amount produced by the original model, and improved correlation and linearity between observation and forecast. In particular, while the machine learning prevailed over the original model up to 100 hours ahead, from then on, both of them showed similar performance or that of the former was downward slightly. If the above-mentioned cloud physics features are used to further sharpen machine learning technique, its performance should be enhanced more and more.

How to cite: Cho, E., Kim, Y.-H., Kim, S., and Kwon, Y. C.: The Development of precipitation model modifed with ECMWF IFS and XGBoost and its performance verification, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7091, https://doi.org/10.5194/egusphere-egu24-7091, 2024.

X5.2
|
EGU24-6873
Does more frequent Very Short-Range Forecast provide more useful information? 
Joohyung Son, Jongseong Kim, and Seongjin Kim

The Very Short-Range Forecast (VSRF) for precipitation from the Korea Meteorological Administration (KMA) is released every 10 minutes, providing forecasts for the next 6 hours at 10-minute intervals. However, when the forecast is provided to the public, it is updated at 10-minute interval, but only provides up to 6 hours at every hour. Consequently, from the public's perspective, forecasts for specific times may change every 10 minutes. While this allows users to access the latest updates, it also poses a challenge in terms of reduced reliability due to constantly changing predictions.

This study aims to assess the prediction performance and variability between forecasts released at 10-minute intervals and those at 1-hour intervals. We evaluated with the Very Short-Range Forecast numerical model KLAPS in VSRF and seek to determine which approach offers more valuable information from the public's standpoint. The assessment focuses on two distinct types of precipitation. The first involves convective showers, which sporadically appear over short durations, driven by atmospheric instability during the Korean Peninsula's summer. The second relates to systematic precipitation associated with a frontal boundary accompanying a medium-scale low-pressure system. For convective showers, the 1-hour interval exhibits better performance and continuity, particularly as the forecast time extends. In the case of systematic precipitation, the 1-hour interval remains superior, though the skill is not as prominent as with convective showers. This highlights that an abundance of information doesn't always equate to high-quality information.

How to cite: Son, J., Kim, J., and Kim, S.: Does more frequent Very Short-Range Forecast provide more useful information?, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6873, https://doi.org/10.5194/egusphere-egu24-6873, 2024.

X5.3
|
EGU24-6856
Very Short-Range Precipitation Forecast in Korea Meteorological Administration 
ho yong lee, Jongseong Kim, Joohyung Son, and Seong-Jin Kim

Korea Meteorological Administration (KMA) has been providing the public with an hourly precipitation forecast updated every 10 minutes for the next 6 hours since 2015. This forecasts, named as the Very Short-Range Forecast (VSRF), differs from other longer forecasts ? such as short-range and medium-range forecasts issued by forecasters. The VSRF is automatically generated by a system based on two different models: MAPLE (McGill Algorithm for Precipitation nowcasting by Lagrangian Extrapolation) and KLAPS (Korea Local Analysis and Prediction System). 

MAPLE, based on Variational Echo Tracking (VET) from radar observations, has an intrinsic disadvantage: its performance decreases rapidly. On the other hand, numerical weather prediction systems like KLAPS are not initially as effective as MAPLE due to model balancing factors such as spin-up, but they maintain initial skill for a slightly longer period. Therefore, to provide the best predictions to the public, it is necessary to merge the two models properly. KMA conducted tests to determine the optimal way to utilize both models and established weights for each model based on their performance and precipitation tendencies. According to a 4-year evaluation, MAPLE outperforms for up to 2 hours, while KLAPS performs better after 4 hours. Consequently, the two models were merged with a hyperbolic tangent weight applied between 2 and 4 hours, and we named it as the best guidance. 

The best guidance was verified against precipitation observed by 720 raingauges over South Korea during the summer seasons from 2020 to 2023. It demonstrated better skill compared to both MAPLE and KLAPS. The average threat scores, with a rain intensity threshold of 0.5 mm/h throughout the forecast period, were 0.40 for the best guidance, 0.38 for MAPLE, and 0.35 for KLAPS.

The best guidance depends on both MAPLE and KLAPS. Therefore, KMA is actively working to improve the performance of each model. Additionally, a very short-range model based on AI is currently under development and running in semi-operations.

How to cite: lee, H. Y., Kim, J., Son, J., and Kim, S.-J.: Very Short-Range Precipitation Forecast in Korea Meteorological Administration, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6856, https://doi.org/10.5194/egusphere-egu24-6856, 2024.

X5.4
|
EGU24-19377
|
ECS
Advancing Spatiotemporal Rainfall Nowcasting through Deep Learning Techniques 
Ahmed Abdelhalim, Miguel Rico-Ramirez, Weiru Liu, and Dawei Han

For weather forecasters and hydrologists, predicting rainfall in the short term – minutes to a few hours – is crucial for a range of applications. While traditional nowcasting methods excel in operational settings, they face limitations in predicting convective storm formation and high-intensity events. Enter deep learning, a powerful tool transforming numerous fields. Convolutional neural networks, in particular, have shown promise in improving nowcasting accuracy. These networks can learn complex patterns and relationships within data, like the intricate tapestry of rainfall variations observed in historical radar sequences. However, capturing long-term dependencies in this data remains a challenge, resulting in fuzzy nowcasts and underestimating high-intensity events. This study proposes a novel deep learning model that goes beyond simple extrapolation, effectively capturing both the spatial correlations and temporal dependencies within rainfall data. Our hybrid convolutional neural network architecture tackles this challenge through three key components: Decoder & Encoder: These modules focus on unraveling the intricate spatial patterns of rainfall and a temporal Module to learn the subtle long-term evolutions and interactions between rain cells over time. By capturing these temporal dependencies, the model can produce more accurate forecasts. To evaluate the model performance, it is compared against both deep learning and optical flow baselines. This presentation will introduce the model and provide a summary of its performance in spatiotemporal rainfall nowcasting.

Keywords: deep learning; spatiotemporal encoding, rainfall nowcasting; radar; optical flow

How to cite: Abdelhalim, A., Rico-Ramirez, M., Liu, W., and Han, D.: Advancing Spatiotemporal Rainfall Nowcasting through Deep Learning Techniques, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19377, https://doi.org/10.5194/egusphere-egu24-19377, 2024.

X5.5
|
EGU24-6545
Improving precipitation nowcasting using deep generative models: a case-study and experiences in R2O  
Kirien Whan, Charlotte Cambier van Nooten, Maurice Schmeits, Jasper Wijnands, Koert Schreurs, and Yuliya Shapovalova

Precipitation nowcasting is essential for weather-dependent decision-making. The combination of radar data and deep learning methods has opened new avenues for research. Deep learning approaches have demonstrated equal or better performance than optical flow methods for low-intensity precipitation, but nowcasting high-intensity events remains a challenge. We use radar data from the Royal Netherlands Meteorological Institute (KNMI) and explore various extensions of deep learning architectures (i.e. loss function, additional inputs) to improve nowcasting of heavy precipitation intensities. Our model outperforms other state-of-the-art models and benchmarks and is skilful at nowcasting precipitation for high rainfall intensities, up to 60-min lead time. 

Transferring research to operations is difficult for many meteorological institutes, particularly for new applications that use AI/ML methods. We discuss some of these challenges that KNMI is facing in this domain. 

How to cite: Whan, K., Cambier van Nooten, C., Schmeits, M., Wijnands, J., Schreurs, K., and Shapovalova, Y.: Improving precipitation nowcasting using deep generative models: a case-study and experiences in R2O , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6545, https://doi.org/10.5194/egusphere-egu24-6545, 2024.

X5.6
|
EGU24-9935
|
ECS
Machine and Deep Learning algorithms to improve weather forecasts over a complex orography Mediterranean region 
Luca Furnari, Umair Yousuf, Alessio De Rango, Donato D'Ambrosio, Giuseppe Mendicino, and Alfonso Senatore

The rapid development of artificial intelligence algorithms has generated considerable interest in the scientific community. The number of scientific articles relating to applying these algorithms for weather forecasting has increased dramatically in the last few years. In addition, the recent operational launch of products such as GraphCast has put this area of research even more in the spotlight. This work uses different Machine Learning and Deep Learning algorithms, namely ANN (Artificial Neural Network), RF (Random Forest) and GNN (Graph Neural Network), with the aim to improve the short-term (1-day lead time) forecasts provided by a physically-based forecasting system. Specifically, the CeSMMA laboratory, since January 2020, has been producing daily forecasts accessible via the https://cesmma.unical.it/cwfv2/ webpage related to a large portion of southern Italy. The NWP (Numerical Weather Prediction) system is based on the WRF (Weather Research and Forecasting) model, with boundary and initial conditions provided by the GFS (Global Forecasting System) model. The AI algorithms post-process the NWP output, applying correction factors achieved by a two-year training considering the observations of the dense regional monitoring network composed of ca. 150 rain gauges.

The results show that the AI is able to improve daily rainfall forecasts compared to ground-based observations. Specifically, the ANN reduces the average MSE (Mean Square Error) by approximately 29% and the RF by 21% with respect to the WRF forecast for the whole study area (about 15’000 km2). Moreover, the GNN applied to a smaller area (considering only 22 rain gauges) further reduces the MSE by 35% during the heaviest rainfall months.

In addition to improving the performance of the forecast, the AI-based post-processing provides reasonable precipitation spatial patterns, reproducing the main physical phenomena such as the orographic enhancement since it is not a surrogate model and benefits from the original physically-based forecasts.

 

Acknowledgements. This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ‘Innovation Ecosystems’, building ‘Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work reflects only the authors’ views and opinions, neither the Ministry for University and Research nor the European Commission can be considered responsible for them.

How to cite: Furnari, L., Yousuf, U., De Rango, A., D'Ambrosio, D., Mendicino, G., and Senatore, A.: Machine and Deep Learning algorithms to improve weather forecasts over a complex orography Mediterranean region, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-9935, https://doi.org/10.5194/egusphere-egu24-9935, 2024.

X5.7
|
EGU24-16617
Forecasting extreme events with the crossing-point forecast  
Zied Ben Bouallegue

The crossing-point forecast (CPF) is a new type of ensemble-based forecast developed at the European Centre for Medium-Range Weather Forecasts. The crossing point refers to the intersection between the cumulative probability distribution of a forecast and the cumulative probability distribution of a model climatology. Originally, the CPF has emerged as a consistent forecast with the diagonal score, a weighted version of the continuous ranked probability score targeting high-impact events. Ranging between 0 and 1, the CPF can serve as an index for high-impact weather and thus directly be compared with the well-established extreme forecast index. The CPF is also interpretable in terms of a return period and conveys a sense of a “probabilistic worst-case scenario”.  Using a recent example of an extreme event affecting Europe, we illustrate and discuss the performance and specificities of this new type of forecast for extreme weather forecasting.

Ben Bouallegue, Z (2023).  Seamless prediction of high-impact weather events: a comparison of actionable forecasts. arXiv:2312.01673

How to cite: Ben Bouallegue, Z.: Forecasting extreme events with the crossing-point forecast , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-16617, https://doi.org/10.5194/egusphere-egu24-16617, 2024.

X5.8
|
EGU24-12855
|
Project IMA: Lessons Learned from Building the Belgian Operational Seamless Ensemble Prediction System 
Lesley De Cruz, Michiel Van Ginderachter, Maarten Reyniers, Alex Deckmyn, Idir Dehmous, Simon De Kock, Wout Dewettinck, Ruben Imhoff, Esteban Montandon, and Ricardo Reinoso-Rondinel

 

In recent years, several national meteorological services (NMSs) have invested considerable resources in the development of a seamless prediction system: rapidly updating forecasts that integrate the latest observations, covering timescales from minutes to days or longer ahead (e.g. DWD's SINFONY; FMI's ULJAS, MetOffice's IMPROVER and Geosphere Austria's SAPHIR) [1]. This move was motivated mainly by rising expectations from end users such as hydrological services, local authorities, the renewable energy sector and the general public. The development of seamless prediction systems was made possible thanks to the increasing availability of high-resolution observations, continuing advances in numerical weather prediction (NWP) models, nowcasting algorithms, and improved strategies to combine multiple information sources optimally. Moreover, the rise of AI/ML techniques in forecasting and nowcasting can further reduce the computational cost to generate frequently updating seamless operational forecast products.

 

We present the journey of building the Belgian seamless prediction system at the Royal Meteorological Institute of Belgium, with the working title "Project IMA". IMA uses both the deterministic INCA-BE and the probabilistic pysteps-BE systems to combine nowcasts with the ALARO and AROME configurations of the ACCORD NWP model. In the lessons learned along the way, we focus on what is often omitted, moving from research to operations, and integrating what we learn from operations back into research. We discuss the benefits of integrating new developments within the free and open-source software (FOSS) pysteps [2]. Our experience shows that using and contributing to FOSS not only leads to more transparency and reproducible, open science; it also enhances international collaboration and can benefit other users, including developing countries, bringing us a step closer to the ambitious goal of Early Warnings for All by 2027 [3].

 

References

 

[1] Bojinski, Stephan, et al. "Towards nowcasting in Europe in 2030." Meteorological Applications 30.4 (2023): e2124.

[2] Imhoff, Ruben O., et al. "Scale‐dependent blending of ensemble rainfall nowcasts and numerical weather prediction in the open‐source pysteps library." Quarterly Journal of the Royal Meteorological Society 149.753 (2023): 1335-1364.

[3] WMO, "Early warnings for all: Executive action plan 2023-2027", 8 Nov 2022,  https://www.preventionweb.net/quick/75125.

How to cite: De Cruz, L., Van Ginderachter, M., Reyniers, M., Deckmyn, A., Dehmous, I., De Kock, S., Dewettinck, W., Imhoff, R., Montandon, E., and Reinoso-Rondinel, R.: Project IMA: Lessons Learned from Building the Belgian Operational Seamless Ensemble Prediction System, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-12855, https://doi.org/10.5194/egusphere-egu24-12855, 2024.

X5.9
|
EGU24-5849
|
ECS
Towards seamless rainfall and flood forecasting in the Netherlands: improvements to and validation of blending in pysteps 
Ruben Imhoff, Michiel Van Ginderachter, Klaas-Jan van Heeringen, Mees Radema, Simon De Kock, Ricardo Reinoso-Rondinel, and Lesley De Cruz

Flood early warning in fast responding catchments challenges our forecasting systems. It requires frequently updated, accurate and high-resolution rainfall forecasts to provide timely warning of rainfall amounts that will reach a catchment in the coming hours. The Netherlands is a typical example, with polder systems below sea level, a high level of urbanization and catchments with short response times. The need for better short-term rainfall forecasts is clearly present, but this is generally not feasible with numerical weather prediction (NWP) models alone. Hence, an alternative rainfall forecasting method is desirable for the first few hours into the future.

Rainfall nowcasting can provide this alternative but quickly loses skill after the first few hours. A promising way forward is a seamless forecasting system, which tries to optimally combine rainfall products from nowcasting and NWP. In this study, we applied the STEPS blending method to combine rainfall forecasts from ensemble radar nowcasts with those from the Harmonie-AROME configuration of the ACCORD NWP model in the Netherlands. This blending method is part of the open-source nowcasting initiative pysteps. To make blending possible in an operational setup, including the needs of involved water authorities, we made several adjustments to the blending implementation in pysteps, for instance:

  • We reduced the computational time by using a faster preprocessing and advection scheme.
  • We improved the noise initialization (needed for generating ensemble members) to allow for stable forecasts, also when one or both product(s) contain(s) no rain.
  • We enabled a dynamic disaggregation of the 1-hour resolution NWP forecasts to match the temporal resolution of the radar nowcast.

We operationalized the updated blending framework in the flood forecasting platforms of the involved water authorities. Given a forecast duration of 12 hours for the blended forecast and a 10-minute time step, average computation times are 3.4 minutes for a deterministic run and 12.3 minutes for an ensemble forecast with 10 members on a 4-core machine. Preprocessing takes approximately 10 minutes and only needs to occur when a new NWP forecast is issued. We tested the implementation for an entire, rainy summer month (July 15 to August 15, 2023) and analyzed the results over the entire domain. The results demonstrate that the blending method effectively combines radar nowcasts with NWP forecasts. Depending on the statistical score considered (such as RMSE and critical success index), the blending method performs either better or on par with the best-performing individual product (radar nowcast or NWP). A consistent finding is that the blending closely tracks the nowcast quality during the initial 1 to 2 hours of the forecast (in this study, the nowcast had lower errors than NWP during the first 2 – 2.5 hours), after which it gradually transitions into the NWP forecast. At longer lead times, the seamless product retains local precipitation structures and extremes better than the NWP product. It does this by leveraging information from the radar nowcast and the stochastic perturbations. Based on these results, a seamless forecasting approach can be regarded as an improvement for the involved water authorities.

How to cite: Imhoff, R., Van Ginderachter, M., van Heeringen, K.-J., Radema, M., De Kock, S., Reinoso-Rondinel, R., and De Cruz, L.: Towards seamless rainfall and flood forecasting in the Netherlands: improvements to and validation of blending in pysteps, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5849, https://doi.org/10.5194/egusphere-egu24-5849, 2024.

X5.10
|
EGU24-19699
|
ECS
Evaluation of seamless forecasts for severe weather warnings  
Verena Bessenbacher, Jonas Bhend, Lea Beusch, Daniele Nerini, Colombe Siegenthaler, Christoph Spirig, and Lionel Moret

At MeteoSwiss, NWP and ML-based models are run operationally on a daily basis to provide weather forecasts and weather warnings for the general public. These forecasts come from various models that differ in lead times, initialization frequency, spatial resolution, and extents. We aim at combining those sources into a probabilistic, gridded weather forecast that is seamless in space and time. Creating a seamless forecast needs careful post-processing so as not to introduce cut-offs or unphysical behavior at the seams between the model runs. This includes using multiple forecast sources and forecast initializations (called lagged ensembles) and combining these using comprehensive blending methods. 

The first minimal viable product of a seamless forecast is currently being produced at MeteoSwiss, and will soon be available to the forecasters in real time. 

We evaluate the merit of these forecasts in terms of warning thresholds for rain and wind gusts. To do so, we compare reforecasts and observations from ground stations as well as rain radar observations from a set of past severe weather events over Switzerland. We benchmark the seamless forecast with individual forecast sources and post-processed products to evaluate the added value of seamlessly combining different forecast sources into one blended product. We furthermore plan to compare different methods for blending between sources soon.

How to cite: Bessenbacher, V., Bhend, J., Beusch, L., Nerini, D., Siegenthaler, C., Spirig, C., and Moret, L.: Evaluation of seamless forecasts for severe weather warnings , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19699, https://doi.org/10.5194/egusphere-egu24-19699, 2024.

X5.11
|
EGU24-3548
Improving the Completion of Weather Radar Missing Data with Deep Learning 
Aofan Gong, Haonan Chen, and Guangheng Ni

Weather radars commonly suffer from the data-missing problem that limits their data quality and applications. Traditional methods for the completion of weather radar missing data, which are based on radar physics and statistics, have shown defects in various aspects. Several deep learning (DL) models have been designed and applied to weather radar completion tasks but have been limited by low accuracy. This study proposes a dilated and self-attentional UNet (DSA-UNet) model to improve the completion of weather radar missing data. The model is trained and evaluated on a radar dataset built with random sector masking from the Yizhuang radar observations during the warm seasons from 2017 to 2019, which is further analyzed with two cases from the dataset. The performance of the DSA-UNet model is compared to two traditional statistical methods and a DL model. The evaluation methods consist of three quantitative metrics and three diagrams. The results show that the DL models can produce less biased and more accurate radar reflectivity values for data-missing areas than traditional statistical methods. Compared to the other DL model, the DSA-UNet model can not only produce a completion closer to the observation, especially for extreme values, but also improve the detection and reconstruction of local-scale radar echo patterns. Our study provides an effective solution for improving the completion of weather radar missing data, which is indispensable in radar quantitative applications.

How to cite: Gong, A., Chen, H., and Ni, G.: Improving the Completion of Weather Radar Missing Data with Deep Learning, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3548, https://doi.org/10.5194/egusphere-egu24-3548, 2024.

X5.12
|
EGU24-5909
|
ECS
Impact of Spatial Density of Automatic Weather Station Data on Assimilation Effectiveness in WRF-3DVar Model 
Zeyu Qiao, Bu Li, Aofan Gong, and Guangheng Ni

Implementing the 3-Dimensional Variational (3DVar) data assimilation technique using high-density automatic weather station (AWS) observations substantially improves the precipitation simulation and forecast capabilities in the Weather Research and Forecasting (WRF) model. Given the impact of spatial distribution and quantity of observation data on assimilation effectiveness, there is a growing need to assimilate the most efficient amount of observation data to improve the precipitation forecast accuracy, especially in the context of the proliferation of data from diverse sources. This study investigates the impacts of spatial density of assimilated data on enhancing model predictions, focusing on a squall line event in Beijing on 2 August 2017 which has approximately 2400 AWSs in the simulation domain. Seven experiment groups assimilating varying proportions of AWS data (3.125, 6.25, 12.5, 25, 50, 75, and 100 percent of total AWSs) were conducted, comprising 10 experiments per group. The results were then compared with the experiment without data assimilation (CTRL) and the observations. Results show that while the WRF model roughly captured the evolution of this event, it overestimated the precipitation amount with significant deviations in precipitation locations. A general positive correlation was observed between the spatial density of assimilated data and the enhancement in model performance. However, there is a notable threshold beyond which additional data ceases to enhance forecast accuracy. The model performs best when the ratio of the number of assimilated AWSs to the model simulated area reaches 1/40 km-2. Moreover, significant variations in improvement effects across experiments within the same group indicate the substantial impact of spatial distribution of assimilated AWSs on forecast outcomes. This study provides a reference for devising more efficient and cost-effective data assimilation strategies in numerical weather prediction.

How to cite: Qiao, Z., Li, B., Gong, A., and Ni, G.: Impact of Spatial Density of Automatic Weather Station Data on Assimilation Effectiveness in WRF-3DVar Model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5909, https://doi.org/10.5194/egusphere-egu24-5909, 2024.

X5.13
|
EGU24-18548
|
ECS
Enhancing Regional NWP Model with GNSS Zenith Total Delay Assimilation: A WRF and WRFDA 3D-Var Approach in the Greater Region of Luxembourg 
Haseeb Ur Rehman, Felix Norman Teferle, Addissu Hunegnaw, Guy Schumann, Florian Zus, and Rohith Muraleedharan Thundathil

Compared to alluvial floods, flash or pluvial floods are difficult to predict because they result from intense and brief periods of extreme precipitation. Luxembourg has a history of being impacted by floods, with notable occurrences in January 2011, May 2016, December 2017, January 2018, February 2019, and February 2020. However, July 2021 stands out as the most severe flood year on record in the region. In this study we are aiming to develop, a high-resolution numerical weather prediction (NWP) model for effective local heavy rainfall prediction in a nowcasting scenario and provide real time for flood simulation. The modeling relies on the Weather Research and Forecasting (WRF) model, which incorporates local Global Navigation Satellite System (GNSS) data assimilation and local precipitation observations to simulate small-scale, high-intensity convective precipitation.

As part of this, we will also test run the LISFlood flood model in an operational inundation forecast mode, meaning that the flood model will be run with the WRF precipitation forecasts as inputs.

The WRF model was configured for the Greater Region, utilizing a horizontal grid resolution of 12 km and incorporating high-resolution static datasets. Meteorological data i.e. July 13 -14 2021, from the Global Forecast System (GFS) were employed in the model setup as initial boundary condition. Zenith Total Delay (ZTD) data collected from various GNSS stations (112) across Germany and Luxembourg were assimilated into the model. Additionally, observational datasets including Surface Synoptic Observations (SYNOP), Upper Air Data, Radiosonde measurements (TEMP), and Tropospheric Airborne Meteorological Data Reporting (TAMDAR) were assimilated. Following this integration, an sensitivity analysis of various meteorological parameters such as precipitation, surface temperature (T2), and relative humidity was performed.

 

Keywords: NWP, WRF, Flash flood, LISFlood, Weather forecast, High-Resolution, GNSS, ZTD

How to cite: Rehman, H. U., Teferle, F. N., Hunegnaw, A., Schumann, G., Zus, F., and Muraleedharan Thundathil, R.: Enhancing Regional NWP Model with GNSS Zenith Total Delay Assimilation: A WRF and WRFDA 3D-Var Approach in the Greater Region of Luxembourg, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-18548, https://doi.org/10.5194/egusphere-egu24-18548, 2024.

X5.14
|
EGU24-7291
|
ECS
Improvements in fog predictions via a modified reconstruction of moisture distribution using the Weather Research and Forecasting(WRF) model 
Eunji Kim, Soon-Young Park, Jung-Woo You, and Soon-Hwan Lee

Since fog is an important weather phenomenon affecting the traffic safety, accurate fog forecasting should be attained to minimize meteorological disasters. Most fog forecasts determine only the presence or absence of fog based on less visibility than 1 km, which is known as the visibility diagnostic method. During this process, fog could be predicted by the visibility calculated in the numerical weather prediction (NWP) model using the cloud liquid water content (LWC) near the surface. In this study, we investigated to increase the accuracy of fog forecast by optimizing the reconstruction of moisture distribution method, which can simulate the intensity of fog as well as the presence or absence of fog. The performances of the fog simulations were examined by modifying the relative humidity threshold at a height of 2 m and the stability parameters which affect turbulence and also one of the important criteria for fog occurrence. When we applied the optimize parameters to fog prediction in the winter seasons, the probability of detection (POD) has been increased significantly from 0.21 to 0.54. These improvements were attributed to the corrected relative humidity threshold and the stability parameters. Although the false alarm rate (FAR) remained almost unchanged, the critical success index (CSI) has been improved slightly lesser than those of the POD. When we analyzed the life cycle of fog, it takes time for the NWP model to simulate water droplets in the fog-developing stage. Therefore, the accuracy of the fog simulation is intimately related to the reconstruction of moisture distribution. The NWP model, however, showed a better performance in the process of fog dissipation than the reconstruction of moisture distribution method that was sensitive to temperature and turbulence. In conclusion, the reconstruction of moisture distribution led to a considerable improvement of the fog prediction in the generation and development stage since we used the optimized humidity threshold. It is also expected that accurate fog prediction could be achieved in the future by considering the aerosol effects, which is another importance factor for the fog generation.

How to cite: Kim, E., Park, S.-Y., You, J.-W., and Lee, S.-H.: Improvements in fog predictions via a modified reconstruction of moisture distribution using the Weather Research and Forecasting(WRF) model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7291, https://doi.org/10.5194/egusphere-egu24-7291, 2024.

X5.15
|
EGU24-7086
Development of stadium-specific numerical forecast guidance for weather forecast for the 2024 Gangwon Winter Youth Olympic Games 
Yeon-Hee Kim, Eunju Cho, Sungbin Jang, Junsu Kim, Hyejeong Bok, and Seungbum Kim

The 2024 Gangwon Winter Youth Olympic Games (GANGWON 2024) will be held in the province of Gangwon in the Republic of Korea from January 19 to February 1, 2024, which already hosted the Olympic Winter Games PyeongChang 2018. In order to successfully host these first Winter YOG to be held in Asia, which will be held for the first time in Asia, it is necessary to provide customized weather information for decision-making in game operation and support in establishing game strategies for athletes and their teams. Accordingly, the Korea Meteorological Administration develops point-specific numerical forecast guidance for major stadiums and provides it to the field to support successful hosting of YOG and improvement of performance. Numerical forecast guidance is the final data delivered to consumers or forecasters as post-processed numerical model data that has been corrected by applying altitude correction and statistical methods to produce highly accurate forecasts. For a total of 13 forecast elements (temperature, minimum/maximum temperature, humidity, wind direction/speed, precipitation, new snow cover, sky conditions, precipitation probability, precipitation type), we developed user-customized numerical forecast guidance specialized for competition points  (Gangneung Olympic Park, Pyeongchang Alpensia Venue, Biathlon Center, Olympic Sliding Center departure/arrival, Wellyhilli departure/arrival, High1 departure/arrival). Through the process of Perfect Prognostic Method (PPM), Model Output Statistics (MOS), optimization, and optimal merging, the systematic errors inherent in the numerical model are removed, and the optimal data (BEST) with improved forecasting performance is provided as customized numerical forecast guidance specific to stadium locations.  In the prediction performance evaluation for the period of December 2023, the accuracy (improvement rate) compared to the average of available models was temperature 1.49℃ (18%), humidity 12% (25%), wind speed 1.87m/s (33%), and visibility 12.8km (17%).

How to cite: Kim, Y.-H., Cho, E., Jang, S., Kim, J., Bok, H., and Kim, S.: Development of stadium-specific numerical forecast guidance for weather forecast for the 2024 Gangwon Winter Youth Olympic Games, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7086, https://doi.org/10.5194/egusphere-egu24-7086, 2024.

X5.16
|
EGU24-9659
Application Research of Multi-source New Detection Data in Snow Depth Prediction for Beijing Winter Olympics 
Jia Du, Bo Yu, Yi Dai, Sang Li, Luyang Xu, Jiaolan Fu, Lin Li, and Hao Jing

According to the demand of the Winter Olympic Organizing Committee for snow depth prediction, the application of multi-source new data in snow depth was studied based on densely artificial snow-depth measurement, microscopic snowflake shape observation and PARSIVEL data. The specific conclusions are as follows: (1) Most of the Snow-Liquid-Ratio(SLR) in Beijing competition zone was between 0.69 and 1.43 (unit: cm/mm, the same below), while that in Yanqing zone was between 0.53 and 1.17. But 7.5% of the SLRs in Yanqing zone exceeded 3.5, which all occurred in the same period of the key service time of 2022 Beijing Winter Olympics, making it more difficult to predict new snow depth. (2) The higher the SLR, the lower the daily minimum surface temperature and lowest air temperature.  Plate or column ice crystals, rimed snowflakes, and dendritic snowflakes were observed, whose corresponding SLRs increased. The average falling speed of particles falling below 2m/s can be used as an indicator of phase transfer. (3) The vertical distributions of temperature and humidity with SLR <1 or >2 were summarized. It was found that when the cloud area coincided with the dendritic growth zone with height close to Yanqing zone, the SLR would be more than 2, higher than that of Beijing zone. (4) A weather concept model generating large SLR was extracted. Snow in Beijing is often accompanied by easterly winds in boundary layer, which is easy to form a wet and ascending layer in the lower troposphere due to the blocking of western mountain. In the late winter season, helped by the temperature’s profile, it tends to produce unrimed dendritic snowflakes, leading to a great SLR.

How to cite: Du, J., Yu, B., Dai, Y., Li, S., Xu, L., Fu, J., Li, L., and Jing, H.: Application Research of Multi-source New Detection Data in Snow Depth Prediction for Beijing Winter Olympics, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-9659, https://doi.org/10.5194/egusphere-egu24-9659, 2024.

X5.17
|
EGU24-14541
Status and Plan of Standard Verification System for the NWP model in Korea Meteorological Administration 
Sora Park, Hyeja Park, Haejin Lee, Saem Song, Jong-Chul Ha, and Young Cheol Kwon

The Korea Meteorological Administration (KMA) has established and operated a standard verification system of the operational NWP models to evaluate the predictive performance of NWP model and compare them with other NWP models operated by domestic and foreign organization. This secures the objectivity of the verification results by applying the verification standards (WMO-No.485) presented by World Meteorological Organization (WMO), and being able to compare the performance with the numerical forecasting models of other institutions under the same conditions. The NWP models to be verified is a global, a regional, very short-range, and an ensemble prediction system and verification against analyses and observations are performed twice a day (00 UTC, 12 UTC). In addition to standard verification, precipitation, typhoon and various verification indexes (CBS index, KMA index, jumpiness index) are verified and used to evaluate the utilization of NWP models. The Korea Integrated Model (KIM), which is developed for Korea’s own NWP model, has been in operation since April 2020. Since the start of operation, the RMSE of 500hPa geopotential height (in Northern Hemisphere) has decreased every year, showing that forecast performance is improving. In addition, it can be seen that the 72-hour prediction accuracy for 12-hour accumulated precipitation (1.0 mm or more) in the Korean Peninsula area (75 ASOS stations) is also improving. As such, this study intends to discuss the predictive performance of the numerical forecast model based on the standard verification system and plans to improve the verification system in the future. 

How to cite: Park, S., Park, H., Lee, H., Song, S., Ha, J.-C., and Kwon, Y. C.: Status and Plan of Standard Verification System for the NWP model in Korea Meteorological Administration, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14541, https://doi.org/10.5194/egusphere-egu24-14541, 2024.

X5.18
|
EGU24-17339
|
ECS
Evaluation of the Impact of Uniform and Non-Uniform Resolution Implementations in Numerical Weather Prediction Models over the Accuracy of Short-Term Wind Prediction 
Ruhi Deniz Yalcin, M. Tugrul Yilmaz, and İsmail Yucel

The increasing integration of renewable energy resources to the national grids necessitates
accurate prediction of power generation from those sources in terms of secure operation of
electricity grid system and energy trading. Electricity generation of renewable energy power
plants such as wind and solar are inherently affected by weather conditions. The wind condition
particularly is affected by surface characteristics such as orography and vegetation, therefore it is
the one of the near surface atmospheric variables having the strongest local variability. The high-
resolution Numerical Weather Prediction (NWP) models are utilized to take the local conditions
into account. WRF model is the one of the most common NWP models having been widely
investigated by various researchers. On the other hand, The Model for Prediction Across Scales
(MPAS) is a relatively new NWP model utilizing non-uniform mesh structures, developed by the
National Center for Environmental Predictions (NCEP). However, there are limited studies in the
literature which compare the prediction performance of WRF and MPAS model in terms of
surface wind speed. This study evaluates the prediction accuracy of near surface wind of two
downscaled NWP models namely, WRF-ARW and MPAS. Both models are configured with
almost identical physics suites and initialized with 3 hourly 00-UTC initialization of Global
Forecast System (GFS) data. The model outputs are obtained at 10 minutes interval for 48 hours
horizon. Hourly averaged model results are compared with observations from 104 on-site
meteorological stations located in Turkiye having different complexity in terms of correlation
coefficient and RMSE.

How to cite: Yalcin, R. D., Yilmaz, M. T., and Yucel, İ.: Evaluation of the Impact of Uniform and Non-Uniform Resolution Implementations in Numerical Weather Prediction Models over the Accuracy of Short-Term Wind Prediction, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17339, https://doi.org/10.5194/egusphere-egu24-17339, 2024.

X5.19
|
EGU24-14232
Statistics of the Subgrid Cloud of an Idealized Tropical Cyclone at Convection-Permitting Resolution 
Xu Zhang and Jian-Wen Bao

Large-eddy simulations of an idealized tropical cyclone (TC) were conducted as benchmarks to provide statistical information about subgrid convective clouds at a convection-permitting resolution over a TC convection system in different stages. The focus was on the vertical and spatial distributions of the subgrid cloud and associated mass flux that need to be parameterized in convection-permitting models. Results showed that the characteristics of the subgrid clouds varied significantly in various parts of the TC convection system. Statistical analysis revealed that the subgrid clouds were mainly located in the lower troposphere and exhibited shallow vertical extents of less than 4 km. The subgrid clouds were also classified into various cloud regimes according to the maximum mass flux height. Local subgrid clouds differed in mass-flux profile shape and magnitude at various regimes in the TC convection system.

How to cite: Zhang, X. and Bao, J.-W.: Statistics of the Subgrid Cloud of an Idealized Tropical Cyclone at Convection-Permitting Resolution, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14232, https://doi.org/10.5194/egusphere-egu24-14232, 2024.

X5.20
|
EGU24-18938
Forecasting tropical high-impact rainfall events using a hybrid statistical dynamical technique based on equatorial waves 
Samantha Ferrett, Gabriel Wolf, John Methven, Tom Frame, Christopher Holloway, Oscar Martinez-Alvarado, and Steve Woolnough

Recent work within the WCSSP FORSEA project and its successor FORWARDS has demonstrated that a hybrid statistical-dynamical forecasting technique combining model ensemble forecasts of equatorial waves with climatological rainfall statistics conditioned on wave phase and amplitude can provide additional skill in predicting high impact weather. The underlying rationale for the technique is twofold. Firstly that high impact rainfall events in the tropics are commonly associated with presence of equatorial waves; and secondly that while global models can adequately predict the evolution of dynamical structure of equatorial waves on time-scales of several days they do not predict the relationship between waves and rainfall well. In tests using the Met Office Global and Regional Forecasting System (MOGREPS) the hybrid forecast is found to outperform model rainfall forecasts from both the global and regional convection permitting versions of MOGREPS, however a weighted blend of the MOGREPS forecasts and the hybrid forecast was found to have the highest skill and further improvements in the method may be obtained by taking into consideration the effects of wave-superposition and interaction. To ascertain whether forecasts can be further improved by better predictions of wave amplitude and phase we compare to hypothetical best-case hybrid forecast computed using wave amplitudes and phases taken from reanalysis. This best-case scenario indicates that errors in forecasting all wave types diminish the hybrid forecast's skill, with the most significant reduction observed for Kelvin waves, suggesting that a significant improvement in the prediction of the propagation of equatorial waves would have a significant impact on rainfall prediction in the tropics. 

How to cite: Ferrett, S., Wolf, G., Methven, J., Frame, T., Holloway, C., Martinez-Alvarado, O., and Woolnough, S.: Forecasting tropical high-impact rainfall events using a hybrid statistical dynamical technique based on equatorial waves, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-18938, https://doi.org/10.5194/egusphere-egu24-18938, 2024.

X5.21
|
EGU24-10809
Analysis of a rare severe convective weather event in spring in Beijing of China 
Bo Yu

On May 17, 2019, a rare severe convective weather occurred in Beijing, accompanied by local heavy rainstorm, hail, thunderstorm and gale. This severe convective weather occurred significantly earlier than normal years, bringing great challenge to the forecast. Using multiple observation data and radar four-dimensional variational assimilation products to analyze the triggering and development evolution of this severe convection. Under the conditions of no obvious weather scale system and local high potential unstable energy, the eastward advancement of the sea breeze front was the main factor triggering strong convection. As the northwest wind in the air increasing, the environmental conditions became stronger vertical wind shear, which was beneficial for the storm to maintain for a longer period of time. The supercell was the main cause of the convective weather. During the development of storms, they split into two parts and moved counterclockwise. The southern echo gradually weakened as it moved northward, while the northern echo moved southward, strengthening and developing into a super cell accompanied by a mesocyclone. The significant fluctuations in the height of the 0 ℃ layer within a small range resulted in different melting rates of hail during its descent, leading to the formation of spiky hail.

How to cite: Yu, B.: Analysis of a rare severe convective weather event in spring in Beijing of China, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10809, https://doi.org/10.5194/egusphere-egu24-10809, 2024.

X5.22
|
EGU24-9382
|
ECS
A study on the enhancement mechanism of the eastward moving convective clouds over the Tibetan plateau and their enhancement in the secondary terrain
(withdrawn after no-show)
xiaofang wang
X5.23
|
EGU24-13809
Investigating Model Forecast Bias for Rainfall Produced by Slantwise Ascent above Cold High 
Ning Hu and Jiaolan Fu

Moderate to heavy rain produced by slantwise ascent of moist air above the cold high is prevalent in cold season in East China. The slantwise ascent is usually characterized by a southwest moist flow aroused by the so-called southern branch trough of 500hPa level to the south of the Qinghai-Tibet Plateau, while the cold high is usually formed by cold air damming, which is familiar to weather forecasters due to topographic feature of East China. The routine short-range forecast skill for this kind of precipitation of weather forecasters is usually limited by model performance. Through large sample model verification, our study indicates that, for the rainfall produced by southwesterly moist flow ascending above the cold high, the ECMWF model always underestimates the rainfall amount on the northeastern part of the rainfall belt, which could be taken as a systematic bias of the state-of-the-art global model. Our case studies indicate that the underestimation of rainfall amount is related to the weaker slant ascent of moist southwest flow forecast by ECMWF model than observation or reanalysis. The southwest flow above the northeastern flow induced by the cold high forms strong wind shear and warm-moist advection, which favors the occurrence of conditional symmetric instability producing strong slantwise ascent not well reflected by global model.

How to cite: Hu, N. and Fu, J.: Investigating Model Forecast Bias for Rainfall Produced by Slantwise Ascent above Cold High, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-13809, https://doi.org/10.5194/egusphere-egu24-13809, 2024.