We welcome any methods and ideas, both traditional and machine learning-based, on how to assimilate data, but also on approaches to create and use an ensemble forecast, and how these techniques can vary with the forecast lead-time. Of particular interest will be the perspective of forecasters and the use of ensembles in forecasting extreme weather events.
The conveners invite papers on various issues associated with Data Assimilation and Ensemble Forecasting for weather prediction, such as:
- intercomparison and study of the complementarity between different assimilation techniques: Kalman filtering, variational assimilation, nudging techniques for frequent analysis cycles, etc;
- variational techniques with longer assimilation windows and weak constraint methods to allow for the inclusion of model error estimates;
- ensemble data assimilation systems and flow dependent estimation of background and on-the-fly error statistics;
- representation of uncertainties in initial conditions, model and boundary coupling in Global and Limited-Area Ensemble Prediction Systems;
- verification and calibration methods of Ensemble Prediction Systems;
- use of TIGGE database;
- application of ensemble forecast in different sectors, including energy, health, transport, agriculture, insurance, finance, etc.
Ethiopia is frequently affected by extreme rainfall events with devastating consequences for its society and economy. For example, they caused a deadly flash flood in Dire Dawa town in April 2020, and in Addis Ababa in August 2021. The frequency and severity of extreme rainfall events varies by region and season. Their variability depends mainly on the advection of moist air and the location and intensity of precipitation-bearing systems approaching Ethiopia. The westward propagation of low-pressure systems developing over the Indian Ocean and Arabian Sea as well as southerly moisture flow are widely known precipitation-producing features in this region.
Due to the vulnerability to extreme precipitation, especially over the orographically complex regions of Ethiopia, a very high resolution of Numerical Weather prediction (NWP) models with high quality initial conditions are required to enhance the advance warning time. The accuracy of initial states of the NWP models can be enhanced through advanced variational Data Assimilation (DA) techniques. With this aim, we simulated an extreme precipitation event using the Weather Research and Forecasting (WRF) model over the Horn of Africa centering the target region Ethiopia. The model is set up at approx. 2 km horizontal resolution and 100 vertical levels. The impact of DA is also evaluated using single observation tests and 3DVar simulations at three cycles. The 3DVar simulations are compared with the control run (without DA) and observations from Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS). This shows that the DA improves the results in comparison with the control run. Finally, we combined the Ensemble Transform Kalman Filter (ETKF) with the 3-Dimensional Variational DA.
The hybrid ETKF-3DVar (3DEnsVar) simulations are performed within a Rapid Update Cycle (RUC) with 6 hourly updates and 15 ensemble members. The 3DEnsVar simulations are compared with observations from Global Precipitation Measurement Mission (GPM), Climate Prediction Center morphing method (CMORPH), and European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) products. Results indicate that the 3DEnsVar method is an appropriate DA method to advance extreme rainfall prediction over Ethiopia.
How to cite: Adgeh, T., Schwitalla, T., Warrach-Sagi, K., and Wulfmeyer, V.: Advanced Weather Forecasts for Ethiopia by Optimized Initialization using a 3DVAR hybrid approach, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-458, https://doi.org/10.5194/ems2024-458, 2024.
Climate change poses significant challenges for agriculture, impacting crop production with rising temperatures, altered precipitation patterns, and more frequent extreme weather events. In this context, data assimilation techniques in Numerical Weather Prediction (NWP) models could become essential tools. By integrating real-time observations into forecasts, data assimilation enhances accuracy, providing farmers with valuable insights to adapt to climate change effectively. Thus, observations from high-resolution networks are of interest for non-hydrostatic model simulations both for verification purposes and for data assimilation. Such observations can be influenced by sub-grid scale processes that cannot be represented in the model dynamics; moreover they can be subject to gross errors. Furthermore, initial errors can grow, during a non-hydrostatic simulation, along a large number of independent modes, influencing a variety of dynamic scales. Suitable automatic data quality control techniques are then necessary and their application can improve assimilation results by enabling the representation of weather features that the model can effectively simulate.
The MAGDA (Meteorological Assimilation from Galileo and Drones for Agriculture) Horizon Europe project (https://www.magdaproject.eu), started in 2022, has been developing a toolchain aimed at providing valuable weather and irrigation information to agricultural operators. At the scientific core of MAGDA activity lies the assimilation in non-hydrostatic NWP models of various sources of high-resolution observations, including in situ observations, GNSS (Global Navigation Satellite System), weather radar and meteodrones .
In this work we study the effect of assimilating high-resolution in situ observations in a short-range simulation with the WRF model of a convective event of interest for MAGDA agricultural applications. We aim to assess the impact of preliminary quality control checks on observations being assimilated. In particular, the Spatial Consistency Test based on Optimal Interpolation and Cross Validation can be effective in rejecting data affected by gross errors or large representativeness errors that could otherwise introduce detrimental noise in the model simulation.
This type of application paves the way for the inclusion of low-cost IoT (Internet of Things) sensors in the assimilation procedure, the "metIS hub" used in the Magda project for instance, but it can also be utilized for various other applications using low-cost sensors as observational tools.
How to cite: Lagasio, M., Uboldi, F., Oberto, E., and Milelli, M.: Testing automatic observation quality control for assimilation in a non-hydrostatic simulation in the framework of the MAGDA HE Project, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-262, https://doi.org/10.5194/ems2024-262, 2024.
Numerical weather prediction requires well-defined initial conditions at the interface between land and atmosphere. Weather centres such as the European Centre for Medium-Range Weather Forecasts (ECMWF) currently rely on a variety of data assimilation schemes to analyze different land variables in their operational forecast system.
Recent activities at ECMWF are towards a unified and more consistent land data assimilation system that can provide more accurate initial conditions for the atmospheric forecast. The first step is to replace the current 1D Optimal Interpolation (1D-OI) so far used for single-layer soil and snow temperature analyses and integrate these variables into the most advanced ensemble-based Simplified Extended Kalman Filter (SEKF) applied in operations for multi-layer soil moisture analysis.
The focus is on the technical developments to integrate the first-layer snow and multi-layer soil temperature into the SEKF control vector and on the evaluation of the new implementation with respect to the impact on the atmospheric forecast skill. A sensitivity analysis regarding Jacobians computed from the covariances between screen-level 2-metre temperature observation and soil and snow temperature reveals better representation of small-scale features associated to land heterogeneity when compared to the empirically obtained sensitivity used in the 1D-OI.
A series of NWP experiments were conducted over a three-month summer and winter period to test the benefit of several configurations of the SEKF. Compared to the 1D-OI, the SEKF leads to significant improvements in the 2-metre temperature forecast with seasonal differences in the verification against own analyses and to slightly improved results in the validation using independent synoptic observations. The work presented will lay the foundation for further developments including the integration of additional land variables, such as snow depth and vegetation-related variables, into the SEKF and the investigation of quasi-strong land-atmosphere coupling.
How to cite: Herbert, C., de Rosnay, P., Weston, P., and Fairbairn, D.: Towards unified land data assimilation at ECMWF: Soil and snow temperature analysis in the Simplified Extended Kalman Filter, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-525, https://doi.org/10.5194/ems2024-525, 2024.
ECMWF is now running its extended-range 101-member ensemble, with an approximate grid spacing of 36 km, every day from 00UTC, out to 46 days ahead. This is in conjunction with the 9 km medium-range 51-member ensemble that is run 4 times a day (out to 15 days ahead from 00 and 12 UTC, and 6 days ahead from 06 and 18 UTC). It means that there are effectively 152 ensemble members available from 00UTC in the overlap period, and even more than that if time-lagging is used to include earlier forecasts.
Now that the two ensembles are run every day, work is underway to investigate the potential benefit from combining them and to get a better understanding of how they differ. Many previous studies have already shown that an increase in ensemble size or a combination of ensembles improves skill, especially for longer lead times as members diverge more.
One of the difficulties of blending ensembles at two different resolutions is that each has different biases and representativeness of point observations. Hence the assumption that all members are equally likely is not true since they form two different distributions. That problem could be alleviated by calibrating one to the other or both to observations, which would involve applying statistical methods to a very large ensemble which is not a trivial undertaking.
The work presented here introduces a new, very simple, diagnostic that identifies cyclonic and anticyclonic regions using surface pressure or geopotential height, computed over scales ranging between 500 and 6000km. This has many benefits for assessing ensembles as well as more generally. The prediction of cyclonic and anticyclonic regions is one of the most fundamental aspects of a weather forecast to get correct and hence diagnose. The approach removes the need for any calibration because the scales are much larger than the grid spacing of each ensemble, enabling comparison with analyses processed in the same way. It is also possible to apply standard or spatial ensemble verification metrics, detect any systematic differences in the synoptic pattern between ensembles, and measure predictability over different scales.
Once the diagnostic has been introduced, highlights of comparative behaviour and verification scores will be presented for the individual and combined ensembles using one year of forecasts, along with any selected cases that are helpful for understanding the overall scores.
How to cite: Roberts, N. and Hewson, T.: Combining ensembles using a new synoptic-scale curvature diagnostic, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-349, https://doi.org/10.5194/ems2024-349, 2024.
Precipitation forecasting for heavy-rainfall events over South China in the rainy season is still challenging due to large uncertainties. Convection-permitting ensemble forecasting is expected to address such uncertainties to improve forecasts of heavy rainfall. However, it is not yet clear how to optimally design convection-permitting ensembles by implementing perturbations in initial conditions (ICs). This study investigates the impacts of various IC perturbation methods on convection-permitting ensemble forecasting over South China in the rainy season. Specifically, downscaling, ensemble of data assimilation, time-lagging, and their combination were used to generate IC perturbations for 12-h convection-permitting ensemble forecasting for heavy-rainfall events over South China during the rainy season in 2013–2020. These events were classified as weak- and strong-forcing cases based on synoptic-scale forcing during the presummer rainy season and as landfalling tropical cyclone (TC) cases. Various IC perturbation methods represented different-source IC uncertainties and thus differed in multiscale characteristics of perturbations in vertical structures, horizontal distributions, and time evolution. Combination of various IC perturbation methods evidently increased perturbations or spreads of precipitation in both magnitude and location and thus improved the forecast-error estimation. Such an improvement was most and least evident for TC cases during the early and late forecasts, respectively, and was more evident for strong- than weak-forcing cases beyond 6 h. The growing spatial similarity among various IC perturbations caused the damping of the added value of combining various IC perturbations over each individual type of perturbations. The added value was attributable much more to the increasing magnitude of initial perturbations than to the increasing sources of IC uncertainties. However, the impacts of combining various sources of IC uncertainties on the added value was still nonnegligible although case-dependent for precipitation perturbations in terms of both magnitude and location. Combination of various IC perturbation methods generally improved both the ensemble-mean and probabilistic forecasts with case-dependent improvements. For heavy rainfall forecasting, 1–6-h improvements were most prominent for TC cases in terms of discrimination and accuracy, while 7–12-h improvements were least prominent for weak-forcing cases in terms of reliability and accuracy. In particular, the improvements in predicting weak-forcing cases increased with spatial errors. In contrast, for strong-forcing cases, the improvements were least and most prominent before and beyond 6 h, respectively.
How to cite: Zhang, X.: Impacts of Combining Various Types of Initial Perturbations on Convection-Permitting Ensemble Forecasting over South China during the Rainy Season, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-92, https://doi.org/10.5194/ems2024-92, 2024.
In Catalonia, with around 580 km of coast, most of the population living in the littoral and a lot of socioeconomic activities relying on the sea, accurate wave prediction is essential. The Meteorological Service of Catalonia (SMC) plays a critical role in issuing warnings to the authorities and the population to mitigate potential risks associated with the sea state. Due to the intrinsic uncertainty of the wave models and the probabilistic nature of the early warnings, an Ensemble Prediction System (EPS) is of great help, if not mandatory, to issue these early warnings.
ONA-ENS (multimOdel oceaN wAve Ensemble for the catalaN coaSt) is the multimodel and multiphysics sea wave EPS developed in the SMC for the Balear-Catalan Sea, in the western Mediterranean. It uses the two spectral wave models that are already running in the SMC, SWAN and WW3. In order to take into account the model uncertainty, each model is run with three different parameterizations of the generation and dissipation of the waves. In addition, the initial conditions uncertainty is introduced by using two atmospheric global models, IFS and GFS, that initialize two configurations of the regional model WRF. These atmospheric models produce the wind fields that drive the wave models. Thus, there are four atmospheric members that initialize six wave configurations, creating a 24-member sea wave ensemble. The system is run every day at 00 UTC, with hourly time steps up to five days, at a 3 km spatial resolution.
Here we share the verification results of the ONA-ENS using the data obtained from the 7 buoys of Puertos del Estado available in the Balear-Catalan Sea during 2023. Comparisons against the deterministic operational models at the SMC reveal a reduction in bias. From a probabilistic point of view, comparisons with the ENS ECWAM from ECMWF show an improvement in the dispersion for the first days of the forecast.
How to cite: Assimenidis, S., Bravo, M., Mercader, J., and Moré, J.: ONA-ENS, the multimodel sea wave prediction system of the SMC, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-320, https://doi.org/10.5194/ems2024-320, 2024.
AEMET-γSREPS is a 2.5km multi-NWP models and multi-boundaries LAM-EPS system operating up to 72 hours since 2016 over Iberian Peninsula, Canary Islands and Antarctica Peninsula which is quite extensively used by AEMET forecasting offices. The main γSREPS goal is to improve operationally issued forecasts and warmings with a consistent measure of their predictability.
The γSREPS 20 members come up crossing four regional mesoscale non-hydrostatic convection-permitting NWP models: HARMONIE-AROME (ACCORD-HIRLAM), ALARO (ACCORD-ALADIN), WRF-ARW (NCAR-NOAA) and NMMB (NCEP-NOAA); with five Global NWP models’ boundary conditions: ECMWF-IFS, NCEP-GFS, MétéoFrance-ARPÈGE, JMA-GSM (Japanese) and CMC-GEM (Canadian). Multi-model and multi-boundary approaches have been selected to take into account the NWP model and boundary condition uncertainties respectively. The combination of both prove to hold better skill-spread relationship than other EPS techniques based on uni-NWP models, especially for precipitation and/or convective High Impact Weather (HIW) events.
γSREPS has a big number of spatial and point products available for forecasting offices through an integrated visualisation framework called PANEL. Interestingly, the latter integrates them with other forecasting systems in AEMET such as deterministic HARMONE-AROME and ECMWF IFS and also its ensemble IFS-ENS, allowing forecasters to carry out a “poor man ensemble” conceptual prediction integration in order to issue the best possible predictions and warnings. Static spatial plots comprise mean, maximum and minimum fields, quartiles, probabilities, member “spaghettis” and EFI/SOT (Extreme Forecast Index). Precipitation “spaghetti” product depicting every member’s contour precipitation in a number of distinct thresholds, are very appreciated in forecasting offices because it allows them to evaluate spatial uncertainty and look for possible extreme events represented only by a few members. Moreover, the new dynamic visualisation based on ADAGUC facilitates zoom up to a very local region for detailed forecasts. For point-based products, a new generation of meteograms (gSREPSgrams) which includes the most extreme single member are produced along the probabilistic vertical-profile product, which shows the most unstable member. Both products try to highlight possible extreme events.
The planned foreseeable evolution of AEMET-γSREPS system is to increase the number of members, incorporating more boundary conditions such as the ones from Global ICON, and including more mesoscale NWP models like ICON-LAM and GEM-LAM.
How to cite: Callado-Pallarès, A., Gómez-Navarro, J. J., and Gil-Oliva, D.: AEMET-γSREPS: The Spanish Convection-permitting LAM-EPS on AEMET forecasting offices, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-873, https://doi.org/10.5194/ems2024-873, 2024.
Within the context of the “Global-to-Regional ICON digital twin” (GLORI) project, a convection-permitting ensemble forecasting is established in order to study the predictability of high-impact weather events with high-resolution modeling (up to 500 m) and the influence of the land-surface—atmosphere coupling mechanisms.
At the DWD, ICON-D2-EPS is the limited-area high-resolution component of the ICON modeling system, running as an ensemble of 20 members at 2 km horizontal resolution over Germany and surrounding areas. The perturbed initial conditions are provided by the km-scale ensemble data assimilation system KENDA, run at the same resolution, assimilating a wide range of observations, including radar-derived radar volumes. Boundary conditions are provided by ICON-EPS, the global ensemble with a refinement at 13 km over Europe and is refreshed every 3 hours.
In this work, we employ a nested domain with horizontal resolution of 1-km in the southern region of the ICON-D2, encompassing the Alps mountains. We run a 24-hour forecast simulation starting at 00UTC on the 21st of June 2022, with 20 ensemble members. The choice of the date is crucial as it corresponds to a day when the DWD recorded instances of heavy rain and hail in southern Germany.
In our study, we perform all experiments using a two-moment microphysics scheme. Additionally, we incorporate the standard operational model perturbations and subsequently analyze the influence of various convection schemes on the predictability of processes that lead to convection development. Specifically, we examine the behavior of the convection scheme in two configurations: shallow convection only and deep convection parameterization in the so-called gray-zone-tuning version. By selectively enabling and disabling these schemes, our goal is to evaluate their individual contributions to predictability.
Following this, we implement a tailored variant of the stochastically perturbed parameterization scheme (SPP) in ICON in order to delve into the influence of some uncertain parameters within either microphysics or turbulent parameterization, further advancing our understanding of its effects on the model performance.
How to cite: Parsakhoo, Z., Marsigli, C., Gebhardt, C., Seifert, A., Steinert, T., and Keller, J.: Studies of Convection-Permitting Ensemble Forecasting for ICON-D2 with a 1km Nest over the Alps, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-485, https://doi.org/10.5194/ems2024-485, 2024.
Many factors need to be considered when designing limited-area ensemble models to forecast at short spatio-temporal scales. However, since the main motivation to run them is to capture the inherent forecast uncertainty at small spatio-temporal scales, especially for severe weather cases, representing such uncertainty is key for producing skillful probabilistic forecasts. Arguably, initial and lateral boundary conditions are a major source of uncertainty for limited-area ensemble forecasts. Many strategies have been explored to initialise convective-scale ensembles but a configuration that is objectively better than others has not been found yet, whereas lateral boundary conditions are provided by a coarser ensemble. Also, limited attention has been given to the size of domains for limited-area models. A domain that is too small could prevent the nested model to generate finer scale and more physically realistic structures which arise from local surface forcing.
To this end, at the Met Office, we have started to run some tests to better understand the sensitivity of our convective-scale ensembles to the presence of high-resolution information at initialisation as well as to the size of the domain, this as a function of lead time and regime dependence.
- hourly time-lagged ensemble with initial conditions centred around the high-resolution deterministic to mimic the current operational ensemble (MOGREPS-UK),
- downscaler from a global ensemble on the same domain as the hourly-cycling,
- downscaler from a global ensemble on an increased size domain.
In all the different designs the ensemble consists of 18 members and runs four times a day. However, in the downscaler all the members are generated from a 6-hourly cycle with initial perturbations coming from the same parent ensemble forecast range, whereas in the hourly-cycling three new members are generated every hour, with five additional sets of initial conditions perturbations and high-resolution deterministic analyses.
The aim of this study is then to assess the sensitivity of the probabilistic skill of heavy precipitation forecasts to these three different designs, both through a series of case studies and aggregate statistics over a nine months period (November 2022-August 2023), using ad-hoc verification measures, focussing on the spatial spread-skill relationship. By looking at different cases we will also assess whether there is any link between the probabilistic skill and the synoptic weather patterns, in order to make more informed decisions about which ensemble is the most skilful.
How to cite: Cafaro, C. and Webster, S.: Designing future Met Office regional ensemble prediction systems: the impact of initial and later boundary conditions on spread-skill relationship for heavy precipitation forecasts., EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-247, https://doi.org/10.5194/ems2024-247, 2024.
On 18 August 2022, Europe was hit by an extreme convective phenomenon. In the second half of the night, a derecho formed in the Mediterranean and successively hit several countries such as Spain, France, Italy, Slovenia, Austria and the Czech Republic. This exceptional meteorological event, with wind gusts reaching a maximum of 62.2 m/s, caused extensive damage in its path and led to numerous fatalties in several countries with the highest death tolls reported in Corsica and in the Alps.
The aim of our work is to better understand and document various aspects of the predictability of such extreme phenomena. Using the operational convective ensemble forecasting system of Météo-France AROME-EPS, we built a large ensemble of hundreds of forecasts. The initial states of the forecasts come from the 26 (25 perturbed + 1 unperturbed) analyses of the operational AROME ensemble of data assimilations and the 35 (34 perturbed + 1 unperturbed) lateral boundary conditions come from the operational global ensemble prediction system of Météo-France, leading to a 910-members convective-scale forecast ensemble. The automatic bow echo detection tool developed by Mounier et al (2022) has been used to examine several aspects of the predictability of the derecho. First, we evaluated whether using hundreds of members in an ensemble prediction system improves our ability to anticipate the occurrence of the derecho. We examined several aspects of the predictability of the phenomenon, such as its trajectory, chronology and intensity. We also assessed the ability of this extended ensemble to produce more consistent probabilistic forecasts over time than the operational ensemble forecasting system. Finally, we examined the sensitivity of the forecasts ability to predict such an event to their initial states and to their coupling members.
How to cite: descamps, L. and mounier, A.: On some aspects of the predictability of the 18 August 2022 derecho using a 910-members convective scale ensemble prediction system, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-79, https://doi.org/10.5194/ems2024-79, 2024.
The use of ensemble prediction systems (EPS) enables the quantification of forecast uncertainty. However, the use of EPS is challenging due to the large amount of information it provides. Forecasts from EPS are typically summarised using statistical measures (such as quantiles maps). Although this mathematical representation is effective in capturing the ensemble distribution, it lacks physical consistency, which raises issues for many applications of EPS in an operational context. Following the application of a bow echo detection tool appreciated by forecasters at Meteo-France, we propose two different approaches for providing physically consistent synthesis of French convection-permitting AROME-EPS forecasts.
The first approach is similar to bow echo detection but it is applied to supercell. Supercell can especially produce hail, strong winds or tornadoes. To summarise the risk of supercell in AROME-EPS forecasts, a convolutional neural network has been trained to automatically detect these supercells in AROME-EPS members based on updraft helicity, reflectivity and hail diagnosis. Then, different synthesis plots are produced, based on these detections. A case study will be presented to better understand the usefulness of these plots.
The second approach is a rainfall synthesis. The aim is to automatically classify the members into different classes pre-defined by rainfall climatology. To design a rainfall synthesis, the procedure can be divided into two parts. The first step aims to extract relevant features from each EPS member to reduce the problem dimensionality. Then, clustering is performed based on these features. The originality of our work is to leverage the capabilities of deep learning for feature extraction. For this purpose, we use a convolutional autoencoder (CAE) to learn an optimal low-dimensional representation (also called latent space representation) of the input forecast field. The clustering task is then accomplished using a SOM algorithm. In this work, the algorithm is developed to work on 1-hour accumulated rainfall from AROME-EPS. The rainfall synthesis plot summarises information concerning rainfall positions, the number of members in each class, and rainfall intensities. The rainfall synthesis will be presented for a case study in this presentation.
The two methods proposed are shown to provide an additional and complementary information, useful for facilitating the human expertise. In addition, their design is generic enough to be applied to other events and variables.
How to cite: Mounier, A. and Soulard-Fischer, L.: Extraction of information for forecasters in convective-scale ensembles: supercell detection and rainfall synthesis, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-121, https://doi.org/10.5194/ems2024-121, 2024.
The land-atmosphere coupling approach in current state-of-the-art NWP systems is based on weakly coupled data assimilation systems for individual Earth-system components. Atmospheric and land surface analyses are performed separately, and results are fed back into the next data assimilation window based on a model forecast. This can lead to imbalanced initial conditions or shocks and the observations cannot be fully harnessed for all components when assimilated only into one Earth system component within the same window.
The CopERnIcus climate change Service Evolution (CERISE) project aims to advance coupled surface-atmosphere assimilation in the preparation of the next generations global and regional reanalysis systems. As for the land, ECMWF’s activities are towards a unified Land Data Assimilation System (LDAS) based on the Simplified Extended Kalman Filter (SEKF) that incorporates multi-layer soil moisture analysis in operations and is currently being extended to other variables, making it suitable for improved coupling.
As part of CERISE, a “quasi-strongly coupled data assimilation” is being developed based on "outer land-atmosphere coupling" approach. Aim is to activate the SEKF as part of several 4D-Var outer loops and return the updated land analyses to initialize the atmosphere and the land of next outer loops within the same assimilation window. Initial efforts have focused on infrastructure developments to enable running the SEKF within the 4D-Var non-linear trajectory.
This work presents the preliminary results of scientific activities, numerical experimentation, and preliminary results to identify the proper number of outer loops required for optimal coupled land-atmosphere assimilation by testing different coupling configurations. For variables that are subject to strong diurnal cycles – such as soil and skin temperature - balanced initial conditions between the different outer loops can be advantageous. The new infrastructure has the capability to improve the exploitation of interface observations (e.g. land surface temperature) so that they can simultaneously influence the analysis of multiple Earth system components.
How to cite: Herbert, C., Weston, P., de Rosnay, P., Fairbairn, D., and Pinnington, E.: Exploring outer-loop land-atmosphere coupling, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-530, https://doi.org/10.5194/ems2024-530, 2024.
Globally, 39% of the world's population lives within 100 km of the coast. Eight of the top ten largest cities in the world are located along the coast. In South Korea, 27% of the population also resides in coastal areas. Ocean waves, induced by sea winds, potentially endanger offshore infrastructure and threaten low-lying ecosystems and communities due to coastal erosion and flooding. More accurate prediction of ocean wave patterns, or sea states, is crucial for informed decision-making in mitigating these risks. Recently, many National Meteorological and Hydrological Services (NMHs) have been endeavoring to enhance the accuracy of ocean wave predictions by assimilating in-situ and remote sensing observations. The Korea Meteorological Administration (KMA) also initiated an ocean wave data assimilation system in 2021. However, this system has thus far only been adapted to a global-scale model. This study aims to install and evaluate the impact of ocean wave data assimilation on a regional scale. The regional ocean wave data assimilation system employs optimal interpolation based on WAVEWATCH-Ⅲ version 6.07 with a spatial resolution of 1/30° targeted for the East Asian region. Significant wave height data from five satellites in polar orbit, ocean data buoys, and coastal wave buoys were utilized. Numerical experiments for summer (2023JJA) and winter (2023/24DJF) reveal that the use of data assimilation reduced the root mean square error by 49.2% and 38.6%, respectively, for the initial fields of regional wave models. The assimilated initial fields improved ocean wave predictions by 12 hours in the KMA regional ocean wave model, which is consistent with previous research in the field. Particularly during Typhoon KHANUN in August 2023, there was a tendency to overestimate sea winds, which are input variables of the ocean wave model. Despite the use of overestimated sea winds, the regional ocean wave model utilizing data assimilation reduced the error by up to 94 cm.
How to cite: Oh, S. M., Chang, P.-H., Park, J.-H., Kang, H.-S., Cho, I. H., and Moon, I.-J.: Assessment of impact of data assimilation of the regional ocean wave model, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-65, https://doi.org/10.5194/ems2024-65, 2024.
With the development of machine learning (ML), it provides new means and methods for accurate climate analysis and prediction. This study focuses on summer precipitation prediction using ML algorithms. Based on BCC CSM1.1, ECMWF SEAS5, NCEP CFSv2, JMA CPS2 model data, we conducted the multi-model ensemble (MME) prediction experiment using three tree-based ML algorithms,, the decision tree (DT), the random forest (RF), and the adaptive boosting (AB) algorithm. On this basis, we explored the applicability of ML algorithms to ensemble prediction of seasonal precipitation in China, as well as the impact of different hyperparameters on prediction accuracy. Then, the MME predictions based on optimal hyperparameters were constructed for different regions of China. The results show that all three ML algorithms have an optimal maximum depth less than 2, which means that based on the current amount of data, the three algorithms can only predict positive or negative precipitation anomalies, and extreme precipitation is hard to predict. The importance of each model in the ML-based MME is quantitatively evaluated. The result shows that NCEP CFSv2 and JMA CPS2 have a higher importance in MME for eastern part of China. Finally, summer precipitation in China was predicted and tested from 2019 to 2021. According to the results, the method provides a more accurate prediction of the main rainband of summer precipitation in China. ML-based MME has a mean ACC of 0.3, an improvement of 0.09 over the weighted average MME of 0.21 for 2019-2021, which exhibits a significant improvement over other methods. It shows that ML methods have great potential in improving short-term climate prediction. These results provide an important reference for short-term climate prediction in China. ML-based MME has the potential to accurately forecast the main rainbands of summer precipitation in China.
How to cite: Yang, J.: Multi-model Ensemble Prediction of Summer Precipitation in China Based on Machine Learning Algorithms, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-304, https://doi.org/10.5194/ems2024-304, 2024.
Abstract: With the development of technology, major operational centers and scientific research institutions have carried out research and development of convection permitting ensemble prediction technology and systems. In order to meet the needs of accurate forecasting in the capital, the Convection permitting ensemble prediction system of North China CMA-BJ-EN v1.0 developed by the Institute of Urban Meteorology was officially put into operation in January 2023. This paper introduces the research and development background and key technologies of the system in detail. The results show that the CMA-BJ-EN v1.0 system has a resolution of 3km convective resolution scale, covering 21 members, and can provide hourly probability prediction results of 48h forecast effectiveness in North China; The system is coupled with advanced technologies such as initial conditon perturbation of ensemble data assimilation, stochastic physical process tendency(SPPT) model perturbation, and has complete pre-processing, initial condition perturbation, model perturbation, and post-processing configuration, which can provide a series of probability prediction products; The qualitative and quantitative evaluation of CMA-BJ-EN v1.0 system and its comparison with NCEP global ensemble forecast system show that the system can effectively grasp several major weather processes in North China, and can obtain local refined probability forecast results relative to global ensemble; The statistical results also show that compared with the NCEP GEFS global ensemble forecast, the CMA-BJ-EN v1.0 forecast system can effectively reduce the RMSE of 2m temperature and 10m wind speed, and can effectively improve the probability forecast score of precipitation forecast. The operation of the system can provide users with effective refined probability forecast reference.
Keywords: convection permitting, ensemble prediction system, initial condition perturbation, model perturbation
How to cite: Zhang, H. and Xia, Y.: Introduction of operational convection permitting ensemble prediction system of North China, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-307, https://doi.org/10.5194/ems2024-307, 2024.
Climatologically, typhoons are generated over the western North Pacific (WNP) and move towards the East Asian countries, such as China, Korea, and Japan, etc. The typical track of typhoons in this case is a C-type curve, and they change their track over the ocean in the midlatitude. This turning point of the track of typhoon is called the recurvature point. However, the untypical track typhoon frequently occurred in recent years and the track forecast is difficult.
For example, the 6th typhoon Kanun in 2023, made two sharp recurvature while moved to the north. Kanun has an increase in track error in the early and mid-stage of prediction, which is estimated to be an error caused by the inability to accurately predict the two recurvature point. The maximum error of the models was approximately 400km for the 72-hour forecast and approximately 700km for the 120-hour forecast, respectively. This is larger than the average error of other typhoons in 2023 as well as the overall error average of Kanun. The question of why is the Kanun’s track error is larger than average is raised.
This study is motivated in order to explain the reason of unusual large track error of typhoons. In the first step, we investigate the two unusual track typhoons errors such as, 11th typhoon Hinnamnor in 2022 and 6th typhoon Khanun in 2023. The model track error will be analyzed by calculating the error at the recurvature point of 102 members of the global ensemble model of KIM (Korean Integrated Model, 26 members), UM (Unified Model, 25 members), and ECMWF (The European Centre for Medium-Range Weather Forecasts, 51 members), respectively.
The individual prediction data of the ensemble members rather than the ensemble mean is used. The recurvature point is automatically calculated from the predicted track at each issue time and compared with the recurvature point calculated on the best track. The prediction trend and performance of the ensemble model are identified by analyzing the distribution of the forecast time error and forecast position error at the recurvature point. Furthermore, the prediction field is classified and diagnosed according to the forecast error.
How to cite: Yun, S., Shin, H.-C., Cha, E.-J., Kim, E.-J., Park, J. I., Choi, W. J., and Ha, J.-C.: Diagnostic study of the forecast accuracy of the ensemble models at the recurvature point of the typhoon, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-390, https://doi.org/10.5194/ems2024-390, 2024.
