AS1.3

Detecting and predicting heavy precipitation for the next few hours is of great importance in weather related decision-making and early warning systems. Although great progress has been achieved in convective-permitting numerical weather prediction (NWP) over the past decades, video prediction models based on deep neural networks have become increasingly popular over the last years for precipitation nowcasting where NWP models fail to capture the quickly varying precipitation patterns. However, previous video prediction studies for precipitation nowcasting showed that heavy precipitation events are barely captured. This has been attributed to the optimization on pixel-wise losses which fail to properly handle the inherent uncertainty.  Hence, we present a novel video prediction model, named CLGAN, embedding the adversarial loss is proposed in this study which aims to generate improved heavy precipitation nowcasting. The model applies a Generative Adversarial Network (GAN) as the backbone. Its generator is a u-shaped encoder decoder network (U-Net) equipped with recurrent LSTM cells and its discriminator constitutes a fully connected network with 3-D convolutional layers. The Eulerian persistence, an optical flow model DenseRotation and an advanced video prediction model PredRNN-v2 serve as baseline methods for comparison. The models performance are evaluated in terms of application-specific scores including root mean square error (RMSE), critical success index (CSI), fractions skill score (FSS) and the method of object-based diagnostic evaluation (MODE). Our model CLGAN is superior to the baseline models for dichotomous events, i.e. the CSI, with a threshold of heavy precipitation (8mm/h), is significantly higher, thus revealing improvements in accurately capturing heavy precipitation events. Besides, CLGAN outperforms in terms of spatial scores such as FSS and MODE. We conclude that the predictions of our CLGAN architecture match the stochastic properties of ground truth precipitation events better than those of previous video prediction methods. The results encourage the applications of GAN-based video prediction architectures for extreme precipitation forecasting.

How to cite: Ji, Y., Gong, B., Langguth, M., Mozaffari, A., Mache, K., Schultz, M., and Zhi, X.: GAN-based video prediction model for precipitation nowcasting, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12086, https://doi.org/10.5194/egusphere-egu22-12086, 2022.

Rainfall plays a significant role in agricultural farming and is considered one of the major natural sources for all living things.  The increase in greenhouse emissions and change in climatic conditions have an adverse effect on the rainfall patterns. Therefore, it becomes crucial to analyze the changing patterns and to forecast rainfall  to mitigate natural disasters that could be caused by the unexpected heavy rainfalls. This paper aims to compare the performance of seven states of the art time series models namely Moving Average(MA), Naïve Forecast(NF), Simple Exponential(SE), Holt’s Linear(HL), Holt’s Linear Additive(HLA), Autoregressive Integrated Moving Average(ARIMA), Seasonal Autoregressive Integrated Moving Average(SARIMA) for the prediction of rainfall. The historical monthly rainfall data from six different stations in United Arab Emirates (UAE) was obtained to assess the performance of seven techniques. Experimental results show that ARIMA outperforms all the prediction models with a mean square error (RMSE) of 9.49 followed by Holt’s Linear model with an RMSE value of 9.91. The performance of all the models is comparable and shows promising performance in rainfall prediction. This also shows the ability of these models to predict the rainfall in arid regions like the UAE

How to cite: Baig, F., Sherif, M., Ali, L., Khan, W., and Faiz, M. A.: Predicting Rainfall using Data-Driven Time Series Approaches, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11143, https://doi.org/10.5194/egusphere-egu22-11143, 2022.

In this study, we present a deep learning-based method to provide seamless high-frequency wind speed forecasts for up to 30 hours ahead. For each selected site, our method generates an ensemble forecast with an update frequency of 10 to 15 minutes(depending on the observation site’s update-frequency). The main objective in this machine learning based post-processing method is to optimally exploit highly resolved NWP models and particularly utilize their multi-level meteorological parameters to integrate the three-dimensionality of weather processes. Further key objectives of this research are to consider different spatial and temporal resolutions and different topographic characteristics of the selected sites. We evaluate the best praxis for efficiently post-processing both the 10-meter wind speed at selected Austrian meteorological observation sites and wind speed on hub height of wind turbines in wind farms.

The method is based on an artificial neural network (ANN), particularly a long-short-term-memory (LSTM) adopted to process several differently structured inputs simultaneously (i.e., different gridded inputs along with observed time-series) and generate ensemble output. An LSTM layer models recurrent steps in the ANN and is, thus, useful for time-series, such as meteorological observations.

Our ensemble forecast method is evaluated for a case study in 2021 using several years of training, including extreme weather event for the selection of sites. The utilized data includes the meteorological observations, gridded nowcasting data as well as NWP data from ECMWF IFS and AROME at several pressure/altitude levels. Hourly runs for 12 test locations (selected TAWES sites covering different topographic situations in Austria) and two wind turbine sites in different seasons are conducted. The obtained results indicate that the model succeeds in learning from inputs while remaining computationally efficient. In most cases the ANN method yields high forecast-skills and is compared to available methods such as the raw NWP model output, climatology, and persistence.

How to cite: Schicker, I., Papazek, P., and DeWit, R.: High-frequency ensemble wind speed forecasting using deep learning, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11240, https://doi.org/10.5194/egusphere-egu22-11240, 2022.

In front of determinism limitations, ensemble forecasting provides competitive advantage assessing uncertainty and helping weather information users in decision-making. Analog ensemble method (AnEn) is one of the most intuitive and computationally cheap ensemble methods that leverages a single deterministic model integration to produce probabilistic information. This method builds an ensemble forecast from a set of past observations of the target variable, neatly selected from a historical training dataset. For a given location, the most similar past forecasts to the current prediction are identified and the associated  past observations are nominated  as members of the analog ensemble forecast. However, The  AnEn forecasting quality is tightly affected by the process of skillful analogs selection in the training data which depends on predictor’s weighting among other factors. This work presents a new weighting strategy based on machine learning techniques (XGBoost, Random Forest and Linear regression) and assesses the impact of its application on the AnEn performance  for 10m wind speed  and 2m temperature forecasting over 13 Moroccan airports in the short term forecasting framework (24 hours). To achieve this, hourly forecasts from the operational mesoscale AROME model and the verifying observations covering 5 year period (2016-2020) are used.  The predictors include 2m temperature, 2m relative humidity, 10m wind speed and direction, mean sea level pressure and surface pressure,  meridonal and zonal components of 10m wind. The basic configuration of Delle Monache et al. (2013) -DM13- where all the predictor’s weights are equal to one is used here as a benchmark. The best weights are computed independently from one airport to another. Since the proposed predictor-weighting strategies can accomplish both the selection of relevant predictors as well as finding their optimal weights, and hence preserve physical meaning and correlations of the used weather variables, the AnEn performances are improved by up to 50 % for bias and by 30% for RMSE for most airports. This improvement varies as function of lead-times and seasons compared to AROME and DM13’s configuration. Results show also that AnEn performance is geographically dependent where a slight worsening is found for some airports.

Keywords : Analog Ensemble,  Machine Learning, Predictors Weighting Strategies, 2m Temperature, 10m Wind Speed, XGBoost, Linear Regression, Random Forest, Ensemble Forecasting.

How to cite: Alaoui, B., Bari, D., and Ghabbar, Y.: New AI based weighting strategy for 2m temperature and 10m wind speed forecasting over Moroccan airports  using the analog ensemble method., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2450, https://doi.org/10.5194/egusphere-egu22-2450, 2022.

For hydrological runoff simulations in hydropower applications, accurate analyses and short-term forecasts of precipitation are of utmost importance. Traditionally, radar-based extrapolations are used for very short-term time scales (approx. 0 - 2 hours ahead). However, during recent years, convection-permitting NWP models have become better at very high spatial and temporal resolution forecasts (e.g. through radar assimilation, RUC configurations). Such models have the advantage of capturing the complex and non-linear evolution of precipitation systems like fronts or thunderstorms in a more physically accurate way than extrapolations, but they are also prone to inaccuracies in precipitation distribution. The aim of this paper is to employ machine learning to combine the strengths of the conventional radar extrapolation (localization and movement of existing storms) with the benefit of the model’s ability to predict storm evolution.  Results show that even a relatively simple sequential deep neural network is able to outperform both, the operational nowcasting and NWP model forecasts. However, the results are highly sensitive to variable selection, loss function, and localization features have a large impact on performance, which is also discussed.

How to cite: Kann, A., Atencia, A., Scheffknecht, P., and Giannakos, A.: AI-based blending of conventional nowcasting with a convection-permitting NWP model, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12384, https://doi.org/10.5194/egusphere-egu22-12384, 2022.

Thanks to recent advances in multisensory observation systems and high-resolution numerical weather prediction (NWP) models, a wealth of information is available to feed and improve operational weather forecasting systems. At the same time, end users such as the renewable energy sector and hydrological services require increasingly detailed and timely weather forecasts that take into account the latest observations.

However, data assimilation in NWP models cannot yet leverage the full spatial or temporal resolution of today's observation systems. Moreover, the combined assimilation and model run takes significantly more time than an extrapolation-based nowcast, and cannot match its accuracy at short lead times. Therefore, many National Meteorological Services (NMSs) are moving towards seamless prediction systems. Seamless prediction aims to make optimal use of today’s rapidly available, high-resolution multisensory observations, nowcasting algorithms and state-of-the-art convection-permitting NWP models. This approach integrates multiple data and model sources to provide a single, frequently updating deterministic or probabilistic forecast for lead times from minutes to days.

We present the seamless ensemble prediction system of the Royal Meteorological Institute of Belgium, called Project IMA (Japanese for "now" or "soon"). It provides rapidly updating seamless forecasts for the next 5 minutes to 24 hours. The nowcasting component is based on two systems: (1) the open-source probabilistic precipitation nowcasting scheme pySTEPS, which now features a scale-dependent blending with NWP ensemble forecasts (also presented in this session) and (2) an ensemble of INCA-BE nowcasts using two different NWP models, for other meteorological variables. The short-range NWP component consists of a multimodel lagged Mini-EPS of two convection-permitting configurations of the ACCORD system: AROME and ALARO, running at 1.3km resolution. It features a 3-hourly DA cycle and provides high-frequency precipitation output to facilitate the blending of precipitation nowcasts and forecasts. The system runs robustly using our NodeRunner tool based on EcFlow, ECMWF's operational work-flow package. We will give an overview of the development (past and future), some lessons learned, and use cases for Project IMA.

How to cite: De Cruz, L., Deckmyn, A., Degrauwe, D., Dehmous, I., Delobbe, L., Dewettinck, W., Goudenhoofdt, E., Imhoff, R., Reyniers, M., Smet, G., Termonia, P., Van den Bergh, J., Van Ginderachter, M., and Vannitsem, S.: Project IMA: Building the Belgian Seamless Prediction System, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12529, https://doi.org/10.5194/egusphere-egu22-12529, 2022.

Early hydrological hazard warning demands precise weather forecasts to accurately predict the timing and the location of intense precipitation events which can cause severe floods/landslides and present risks to urban and natural environments. Extrapolation of precipitation by radar rainfall products at high space and time scales with short lead times outperforms forecasts of numerical weather prediction. Therefore, developing and improving of rainfall nowcasts systems are essential. Rainfall nowcasting is the process of forecasting precipitation field movement and evolution at high spatial and temporal resolutions with short lead times(<6h) in which the advection of the precipitation fields is estimated by extrapolating real-time remotely sensed observations. Radar rainfall nowcasting is increasingly applied because of the high potential of radar products in short-term rainfall forecasting due to their high spatiotemporal resolutions (typically, 1 km and 5 min). It consists of two procedures in tracking precipitation features to calculate the velocity from a series of consecutive radar images and propagating the most recent precipitation observation into the future using the obtained velocity. Optical flow represents one of the most used methods for tracking the motion fields from consecutive images. Deep learning techniques are those machine learning methods that utilise deep artificial neural networks. Deep learning has become one of the most popular and rapidly spreading methods in different scientific disciplines including water-related research. Deep learning applications in radar-based precipitation nowcasting is still in its early stage with many knowledge gaps and their full potential in rainfall nowcasting requires more investigation. This work evaluates the performance of a deep convolutional neural network (called rainnet) and three optical flow algorithms (called Rainymotion Sparse, Rainymotion Dense, Rainymotion DenseRotation) compared with Eulerian Persistence to assess their predictive skills in nowcasting. Synthetic precipitation scenarios have been created with different motion fields (linear and rotational motions), velocities, intensities, sizes, and locations. The models have been evaluated to forecast different precipitation processes that contribute mainly to model errors such as constant and accelerated linear and rotational motions, growth and decay in both size and intensity. Different verification metrics have been used to evaluate the skill of the forecasts.

Keywords: radar rainfall nowcasting; deep learning; optical flow; extrapolation; rainnet; rainymotion

How to cite: Abdelhalim, A., Rico-Ramirez, M., and Han, D.: Evaluation of radar rainfall nowcasting techniques to forecast synthetic storms of different processes, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10595, https://doi.org/10.5194/egusphere-egu22-10595, 2022.

Terrain with different shapes and ground surface properties has extremely complex impacts on atmospheric motion, and the forecast uncertainty and complexity caused by terrain brings great challenges to disaster prevention and mitigation. Therefore, it is essential to design a new-style model topography disturbance model for ensemble prediction system specifically to solve the prediction uncertainty caused by complex terrain. In this paper, on the basis of combing the current models and methods for dealing with different terrain uncertainty, and considering the non-uniformity of terrain gradient, the key element of describing terrain complexity, an orthogonal terrain disturbance method based on terrain gradient is designed and proposed, and the obtained high-resolution orthogonal terrain disturbance is superimposed on the static terrain height of the model to generate different ensemble members, so as to describe the uncertainty in the terrain generation process of high-resolution numerical model. At the same time, a comparative study is carried out with the ensemble forecast of model terrain disturbance between using the new-style method and using different terrain interpolation schemes or smoothing schemes. The preliminary test shows that: first of all, the ensemble dispersion of terrain height disturbance based on the new-style method is closely related to the terrain gradient. The area with small terrain gradient has smaller terrain disturbance ensemble dispersion, while the area with large terrain gradient has larger ensemble dispersion, which shows that the new scheme is more reasonable. Furthermore, compared with the model terrain disturbance schemes with different interpolation or smoothing methods, the dispersion of the new-style method is larger, and the skill of the new-style method becomes more and more obvious with the increase of model resolution. Thirdly, from the comparative study of the forecast effect of high-level and low-level weather elements, the new-style method ensemble forecast has obvious improvement on the forecast effect of low-level variables, especially in areas with complex terrain or large terrain gradient. The possible reason is that the new method can more objectively describe the terrain uncertainty. Fourthly, compared with the ensemble forecast results of different interpolation and smoothing methods, the new-style terrain disturbance scheme can improve the precipitation probability forecast skill and reduce the ensemble average root mean square error, and improve the ensemble average forecast of upper-air elements and near-surface elements. Lastly, the test of the number of ensemble members shows that the prediction effect of new-style terrain disturbance scheme with less members is equivalent or better than that of the interpolation or smoothing terrain disturbance scheme with more members. In summary, the new-style terrain perturbation theory based on terrain gradient in this paper provides a technical reference for the development of complex terrain convection-allowing scale ensemble forecast, which has important theoretical value and application prospect.

Key words: complex terrain，ensemble prediction，convection-allowing scale，topographic perturbation，topographic gradient

How to cite: Chaohui, C., Yi, L., Hongrang, H., Kan, L., and Yongqiang, J.: Preliminary study of a new-style terrain disturbance method based on gradient inhomogeneity in convection-allowing scale ensemble prediction system, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-13244, https://doi.org/10.5194/egusphere-egu22-13244, 2022.

Mountain lee waves are a kind of gravity waves often associated with adverse weather phenomena, such as turbulence that can affect the aviation safety. Not surprisingly, turbulence events have been related with numerous aircraft accidents reports. In this work, several mountain lee wave events in the vicinity of the Adolfo Suarez Madrid-Barajas airport (Spain) are simulated and analyzed using HARMONIE-AROME, the high-resolution numerical model linked to the international research program ACCORD-HIRLAM. Brightness temperature from the Meteosat Second Generation (MSG-SEVIRI) has been selected as observational variable to validate the HARMONIE-AROME simulations of cloudiness associated with mountain lee wave events. Subsequently, a characterization of the atmospheric variables related with the mountain lee wave formation (wind direction and speed, static stability and liquid water content) has been carried out in several grid points placed in the windward, leeward and over the summits of the mountain range close to the airport. The characterization results are used to develop a decision tree to provide a warning method to alert both mountain lee wave events and associated lenticular clouds. Both HARMONIE-AROME brightness temperature simulations and the warnings associated with mountain lee wave events were satisfactory validated using satellite observations, obtaining a probability of detection and percent correct above 60% and 70%, respectively.  

How to cite: Díaz Fernández, J., Bolgiani, P., Santos Muñoz, D., Sastre, M., Valero, F., Farrán, J. I., González Alemán, J. J., and Martín Pérez, M. L.: Characterization and warnings for mountain waves using HARMONIE-AROME, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2471, https://doi.org/10.5194/egusphere-egu22-2471, 2022.

To quantify the intensity of squall line in mid-latitudes, the author recently proposed a squall line intensity assessment method based on cold pool, which provides a measure of squall line intensity.

The disturbance potential temperature density is calculated by using the potential temperature, water vapor and all kinds of water condensate output from the numerical weather forecast model, and the boundary of the cold pool is judged according to the disturbance potential temperature density less than -2K. Based on the contour surface buoyancy, the high surface buoyancy is calculated according to the disturbance potential temperature density, and then the strength of the cold pool is calculated. In this method, the intensity of squall line is analyzed comprehensively by principal component analysis, combined with the weather phenomena accompanied by squall line occurrence, such as cold pool intensity, surface wind speed, ground pressure variation, surface temperature variation, simulated radar echo and so on. The above analysis is the local intensity on different grid points when the squall line occurs, and the overall squall line intensity is obtained by accumulating the local intensity in the squall line range.

The method is verified by the model output data of a squall line process occurred in northern Jiangsu on May 16, 2013. The results show that the distribution of the local squall line intensity is coupled with the surface wind field and heavy precipitation. The intensity evolution of the overall squall line reaches the peak in a short time and then decreases, which corresponds to the life history of the birth, development, maturity and dissipation of the squall line, and also reflects the characteristics of the short life history of the squall line developing rapidly and then dissipating. This method provides technical support for the forecast of squall line and the emergency plan issued by meteorological department.

Acknowledgements. This research was supported by the National Natural Science Foundation of China (Grant Nos. 41975128 and 42075053).

Keywords: squall line, intensity, assessment method, disturbance potential temperature density

How to cite: Yang, R., Jiang, Y., Chen, C., He, H., Li, Y., and Huang, H.: An Assessment Method of Squall Line Intensity Based on Cold Pool, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-13532, https://doi.org/10.5194/egusphere-egu22-13532, 2022.