Browse
Abstract
Postprocessing ensemble weather predictions to correct systematic errors has become a standard practice in research and operations. However, only a few recent studies have focused on ensemble postprocessing of wind gust forecasts, despite its importance for severe weather warnings. Here, we provide a comprehensive review and systematic comparison of eight statistical and machine learning methods for probabilistic wind gust forecasting via ensemble postprocessing that can be divided in three groups: state-of-the-art postprocessing techniques from statistics [ensemble model output statistics (EMOS), member-by-member postprocessing, isotonic distributional regression], established machine learning methods (gradient-boosting extended EMOS, quantile regression forests), and neural network–based approaches (distributional regression network, Bernstein quantile network, histogram estimation network). The methods are systematically compared using 6 years of data from a high-resolution, convection-permitting ensemble prediction system that was run operationally at the German weather service, and hourly observations at 175 surface weather stations in Germany. While all postprocessing methods yield calibrated forecasts and are able to correct the systematic errors of the raw ensemble predictions, incorporating information from additional meteorological predictor variables beyond wind gusts leads to significant improvements in forecast skill. In particular, we propose a flexible framework of locally adaptive neural networks with different probabilistic forecast types as output, which not only significantly outperform all benchmark postprocessing methods but also learn physically consistent relations associated with the diurnal cycle, especially the evening transition of the planetary boundary layer.
Abstract
Postprocessing ensemble weather predictions to correct systematic errors has become a standard practice in research and operations. However, only a few recent studies have focused on ensemble postprocessing of wind gust forecasts, despite its importance for severe weather warnings. Here, we provide a comprehensive review and systematic comparison of eight statistical and machine learning methods for probabilistic wind gust forecasting via ensemble postprocessing that can be divided in three groups: state-of-the-art postprocessing techniques from statistics [ensemble model output statistics (EMOS), member-by-member postprocessing, isotonic distributional regression], established machine learning methods (gradient-boosting extended EMOS, quantile regression forests), and neural network–based approaches (distributional regression network, Bernstein quantile network, histogram estimation network). The methods are systematically compared using 6 years of data from a high-resolution, convection-permitting ensemble prediction system that was run operationally at the German weather service, and hourly observations at 175 surface weather stations in Germany. While all postprocessing methods yield calibrated forecasts and are able to correct the systematic errors of the raw ensemble predictions, incorporating information from additional meteorological predictor variables beyond wind gusts leads to significant improvements in forecast skill. In particular, we propose a flexible framework of locally adaptive neural networks with different probabilistic forecast types as output, which not only significantly outperform all benchmark postprocessing methods but also learn physically consistent relations associated with the diurnal cycle, especially the evening transition of the planetary boundary layer.
Abstract
Major airstreams in tropical cyclones (TCs) are rarely described from a Lagrangian perspective. Such a perspective, however, is required to account for asymmetries and time dependence of the TC circulation. We present a procedure that identifies main airstreams in TCs based on trajectory clustering. The procedure takes into account the TC’s large degree of inherent symmetry and is suitable for a very large number of trajectories
Abstract
Major airstreams in tropical cyclones (TCs) are rarely described from a Lagrangian perspective. Such a perspective, however, is required to account for asymmetries and time dependence of the TC circulation. We present a procedure that identifies main airstreams in TCs based on trajectory clustering. The procedure takes into account the TC’s large degree of inherent symmetry and is suitable for a very large number of trajectories
Abstract
Observations across the North Atlantic jet stream with high vertical resolution are used to explore the structure of the jet stream, including the sharpness of vertical wind shear changes across the tropopause and the wind speed. Data were obtained during the North Atlantic Waveguide and Downstream Impact Experiment (NAWDEX) by an airborne Doppler wind lidar, dropsondes, and a ground-based stratosphere–troposphere radar. During the campaign, small wind speed biases throughout the troposphere and lower stratosphere of only −0.41 and −0.15 m s−1 are found, respectively, in the ECMWF and Met Office analyses and short-term forecasts. However, this study finds large and spatially coherent wind errors up to ±10 m s−1 for individual cases, with the strongest errors occurring above the tropopause in upper-level ridges. ECMWF and Met Office analyses indicate similar spatial structures in wind errors, even though their forecast models and data assimilation schemes differ greatly. The assimilation of operational observational data brings the analyses closer to the independent verifying observations, but it cannot fully compensate for the forecast error. Models tend to underestimate the peak jet stream wind, the vertical wind shear (by a factor of 2–5), and the abruptness of the change in wind shear across the tropopause, which is a major contribution to the meridional potential vorticity gradient. The differences are large enough to influence forecasts of Rossby wave disturbances to the jet stream with an anticipated effect on weather forecast skill even on large scales.
Abstract
Observations across the North Atlantic jet stream with high vertical resolution are used to explore the structure of the jet stream, including the sharpness of vertical wind shear changes across the tropopause and the wind speed. Data were obtained during the North Atlantic Waveguide and Downstream Impact Experiment (NAWDEX) by an airborne Doppler wind lidar, dropsondes, and a ground-based stratosphere–troposphere radar. During the campaign, small wind speed biases throughout the troposphere and lower stratosphere of only −0.41 and −0.15 m s−1 are found, respectively, in the ECMWF and Met Office analyses and short-term forecasts. However, this study finds large and spatially coherent wind errors up to ±10 m s−1 for individual cases, with the strongest errors occurring above the tropopause in upper-level ridges. ECMWF and Met Office analyses indicate similar spatial structures in wind errors, even though their forecast models and data assimilation schemes differ greatly. The assimilation of operational observational data brings the analyses closer to the independent verifying observations, but it cannot fully compensate for the forecast error. Models tend to underestimate the peak jet stream wind, the vertical wind shear (by a factor of 2–5), and the abruptness of the change in wind shear across the tropopause, which is a major contribution to the meridional potential vorticity gradient. The differences are large enough to influence forecasts of Rossby wave disturbances to the jet stream with an anticipated effect on weather forecast skill even on large scales.
Abstract
Different approaches for representing model error due to unresolved scales and processes are compared in convective-scale data assimilation, including the physically based stochastic perturbation (PSP) scheme for turbulence, an advanced warm bubble approach that automatically detects and triggers absent convective cells, and additive noise based on model truncation error. The analysis of kinetic energy spectrum guides the understanding of differences in precipitation forecasts. It is found that the PSP scheme results in more ensemble spread in assimilation cycles, but its effects on the root-mean-square error (RMSE) are neutral. This leads to positive impacts on precipitation forecasts that last up to three hours. The warm bubble technique does not create more spread, but is effective in reducing the RMSE, and improving precipitation forecasts for up to 3 h. The additive noise approach contributes greatly to ensemble spread, but it results in a larger RMSE during assimilation cycles. Nevertheless, it considerably improves the skill of precipitation forecasts up to 6 h. Combining the additive noise with either the PSP scheme or the warm bubble technique reduces the RMSE within cycles and improves the skill of the precipitation forecasts, with the latter being more beneficial.
Abstract
Different approaches for representing model error due to unresolved scales and processes are compared in convective-scale data assimilation, including the physically based stochastic perturbation (PSP) scheme for turbulence, an advanced warm bubble approach that automatically detects and triggers absent convective cells, and additive noise based on model truncation error. The analysis of kinetic energy spectrum guides the understanding of differences in precipitation forecasts. It is found that the PSP scheme results in more ensemble spread in assimilation cycles, but its effects on the root-mean-square error (RMSE) are neutral. This leads to positive impacts on precipitation forecasts that last up to three hours. The warm bubble technique does not create more spread, but is effective in reducing the RMSE, and improving precipitation forecasts for up to 3 h. The additive noise approach contributes greatly to ensemble spread, but it results in a larger RMSE during assimilation cycles. Nevertheless, it considerably improves the skill of precipitation forecasts up to 6 h. Combining the additive noise with either the PSP scheme or the warm bubble technique reduces the RMSE within cycles and improves the skill of the precipitation forecasts, with the latter being more beneficial.
Abstract
We investigate the feasibility of addressing model error by perturbing and estimating uncertain static model parameters using the localized ensemble transform Kalman filter. In particular we use the augmented state approach, where parameters are updated by observations via their correlation with observed state variables. This online approach offers a flexible, yet consistent way to better fit model variables affected by the chosen parameters to observations, while ensuring feasible model states. We show in a nearly operational convection-permitting configuration that the prediction of clouds and precipitation with the COSMO-DE model is improved if the two-dimensional roughness length parameter is estimated with the augmented state approach. Here, the targeted model error is the roughness length itself and the surface fluxes, which influence the initiation of convection. At analysis time, Gaussian noise with a specified correlation matrix is added to the roughness length to regulate the parameter spread. In the northern part of the COSMO-DE domain, where the terrain is mostly flat and assimilated surface wind measurements are dense, estimating the roughness length led to improved forecasts of up to 6 h of clouds and precipitation. In the southern part of the domain, the parameter estimation was detrimental unless the correlation length scale of the Gaussian noise that is added to the roughness length is increased. The impact of the parameter estimation was found to be larger when synoptic forcing is weak and the model output is more sensitive to the roughness length.
Abstract
We investigate the feasibility of addressing model error by perturbing and estimating uncertain static model parameters using the localized ensemble transform Kalman filter. In particular we use the augmented state approach, where parameters are updated by observations via their correlation with observed state variables. This online approach offers a flexible, yet consistent way to better fit model variables affected by the chosen parameters to observations, while ensuring feasible model states. We show in a nearly operational convection-permitting configuration that the prediction of clouds and precipitation with the COSMO-DE model is improved if the two-dimensional roughness length parameter is estimated with the augmented state approach. Here, the targeted model error is the roughness length itself and the surface fluxes, which influence the initiation of convection. At analysis time, Gaussian noise with a specified correlation matrix is added to the roughness length to regulate the parameter spread. In the northern part of the COSMO-DE domain, where the terrain is mostly flat and assimilated surface wind measurements are dense, estimating the roughness length led to improved forecasts of up to 6 h of clouds and precipitation. In the southern part of the domain, the parameter estimation was detrimental unless the correlation length scale of the Gaussian noise that is added to the roughness length is increased. The impact of the parameter estimation was found to be larger when synoptic forcing is weak and the model output is more sensitive to the roughness length.
Abstract
State-of-the-art ensemble prediction systems usually provide ensembles with only 20–250 members for estimating the uncertainty of the forecast and its spatial and spatiotemporal covariance. Given that the degrees of freedom of atmospheric models are several magnitudes higher, the estimates are therefore substantially affected by sampling errors. For error covariances, spurious correlations lead to random sampling errors, but also a systematic overestimation of the correlation. A common approach to mitigate the impact of sampling errors for data assimilation is to localize correlations. However, this is a challenging task given that physical correlations in the atmosphere can extend over long distances. Besides data assimilation, sampling errors pose an issue for the investigation of spatiotemporal correlations using ensemble sensitivity analysis. Our study evaluates a statistical approach for correcting sampling errors. The applied sampling error correction is a lookup table–based approach and therefore computationally very efficient. We show that this approach substantially improves both the estimates of spatial correlations for data assimilation as well as spatiotemporal correlations for ensemble sensitivity analysis. The evaluation is performed using the first convective-scale 1000-member ensemble simulation for central Europe. Correlations of the 1000-member ensemble forecast serve as truth to assess the performance of the sampling error correction for smaller subsets of the full ensemble. The sampling error correction strongly reduced both random and systematic errors for all evaluated variables, ensemble sizes, and lead times.
Abstract
State-of-the-art ensemble prediction systems usually provide ensembles with only 20–250 members for estimating the uncertainty of the forecast and its spatial and spatiotemporal covariance. Given that the degrees of freedom of atmospheric models are several magnitudes higher, the estimates are therefore substantially affected by sampling errors. For error covariances, spurious correlations lead to random sampling errors, but also a systematic overestimation of the correlation. A common approach to mitigate the impact of sampling errors for data assimilation is to localize correlations. However, this is a challenging task given that physical correlations in the atmosphere can extend over long distances. Besides data assimilation, sampling errors pose an issue for the investigation of spatiotemporal correlations using ensemble sensitivity analysis. Our study evaluates a statistical approach for correcting sampling errors. The applied sampling error correction is a lookup table–based approach and therefore computationally very efficient. We show that this approach substantially improves both the estimates of spatial correlations for data assimilation as well as spatiotemporal correlations for ensemble sensitivity analysis. The evaluation is performed using the first convective-scale 1000-member ensemble simulation for central Europe. Correlations of the 1000-member ensemble forecast serve as truth to assess the performance of the sampling error correction for smaller subsets of the full ensemble. The sampling error correction strongly reduced both random and systematic errors for all evaluated variables, ensemble sizes, and lead times.
Abstract
Dropsonde observations from three research aircraft in the North Atlantic region, as well as several hundred additionally launched radiosondes over Canada and Europe, were collected during the international North Atlantic Waveguide and Downstream Impact Experiment (NAWDEX) in autumn 2016. In addition, over 1000 dropsondes were deployed during NOAA’s Sensing Hazards with Operational Unmanned Technology (SHOUT) and Reconnaissance missions in the west Atlantic basin, supplementing the conventional observing network for several intensive observation periods. This unique dataset was assimilated within the framework of cycled data denial experiments for a 1-month period performed with the global model of the ECMWF. Results show a slightly reduced mean forecast error (1%–3%) over the northern Atlantic and Europe by assimilating these additional observations, with the most prominent error reductions being linked to Tropical Storm Karl, Cyclones Matthew and Nicole, and their subsequent interaction with the midlatitude waveguide. The evaluation of Forecast Sensitivity to Observation Impact (FSOI) indicates that the largest impact is due to dropsondes near tropical storms and cyclones, followed by dropsondes over the northern Atlantic and additional Canadian radiosondes. Additional radiosondes over Europe showed a comparatively small beneficial impact.
Abstract
Dropsonde observations from three research aircraft in the North Atlantic region, as well as several hundred additionally launched radiosondes over Canada and Europe, were collected during the international North Atlantic Waveguide and Downstream Impact Experiment (NAWDEX) in autumn 2016. In addition, over 1000 dropsondes were deployed during NOAA’s Sensing Hazards with Operational Unmanned Technology (SHOUT) and Reconnaissance missions in the west Atlantic basin, supplementing the conventional observing network for several intensive observation periods. This unique dataset was assimilated within the framework of cycled data denial experiments for a 1-month period performed with the global model of the ECMWF. Results show a slightly reduced mean forecast error (1%–3%) over the northern Atlantic and Europe by assimilating these additional observations, with the most prominent error reductions being linked to Tropical Storm Karl, Cyclones Matthew and Nicole, and their subsequent interaction with the midlatitude waveguide. The evaluation of Forecast Sensitivity to Observation Impact (FSOI) indicates that the largest impact is due to dropsondes near tropical storms and cyclones, followed by dropsondes over the northern Atlantic and additional Canadian radiosondes. Additional radiosondes over Europe showed a comparatively small beneficial impact.
Abstract
Damaging gusts in windstorms are represented by crude subgrid-scale parameterizations in today’s weather and climate models. This limitation motivated the Wind and Storms Experiment (WASTEX) in winter 2016–17 in the Upper Rhine Valley over southwestern Germany. Gusts recorded at an instrumented tower during the passage of extratropical cyclone “Thomas” on 23 February 2017 are investigated based on measurements of radial wind with ≈70-m along-beam spacing from a fast-scanning Doppler lidar and realistic large-eddy simulations with grid spacings down to 78 m using the Icosahedral Nonhydrostatic model. Four wind peaks occur due to the storm onset, the cold front, a precipitation line, and isolated showers. The first peak is related to a sudden drop in dewpoint and results from the downward mixing of a low-level jet and a dry layer within the warm sector characterized by extremely high temperatures for the season. While operational convection-permitting forecasts poorly predict the storm onset overall, a successful ensemble member highlights the role of upstream orography. Lidar observations reveal the presence of long-lasting wind structures that result from a combination of convection- and shear-driven instability. Large-eddy simulations contain structures elongated in the wind direction that are qualitatively similar but too coarse compared to the observed ones. Their size is found to exceed the effective model resolution by one order of magnitude due to their elongation. These results emphasize the need for subkilometer-scale measuring and modeling systems to improve the representation of gusts in windstorms.
Abstract
Damaging gusts in windstorms are represented by crude subgrid-scale parameterizations in today’s weather and climate models. This limitation motivated the Wind and Storms Experiment (WASTEX) in winter 2016–17 in the Upper Rhine Valley over southwestern Germany. Gusts recorded at an instrumented tower during the passage of extratropical cyclone “Thomas” on 23 February 2017 are investigated based on measurements of radial wind with ≈70-m along-beam spacing from a fast-scanning Doppler lidar and realistic large-eddy simulations with grid spacings down to 78 m using the Icosahedral Nonhydrostatic model. Four wind peaks occur due to the storm onset, the cold front, a precipitation line, and isolated showers. The first peak is related to a sudden drop in dewpoint and results from the downward mixing of a low-level jet and a dry layer within the warm sector characterized by extremely high temperatures for the season. While operational convection-permitting forecasts poorly predict the storm onset overall, a successful ensemble member highlights the role of upstream orography. Lidar observations reveal the presence of long-lasting wind structures that result from a combination of convection- and shear-driven instability. Large-eddy simulations contain structures elongated in the wind direction that are qualitatively similar but too coarse compared to the observed ones. Their size is found to exceed the effective model resolution by one order of magnitude due to their elongation. These results emphasize the need for subkilometer-scale measuring and modeling systems to improve the representation of gusts in windstorms.
Abstract
We investigate the practical predictability limits of deep convection in a state-of-the-art, high-resolution, limited-area ensemble prediction system. A combination of sophisticated predictability measures, namely, believable and decorrelation scale, are applied to determine the predictable scales of short-term forecasts in a hierarchy of model configurations. First, we consider an idealized perfect model setup that includes both small-scale and synoptic-scale perturbations. We find increased predictability in the presence of orography and a strongly beneficial impact of radar data assimilation, which extends the forecast horizon by up to 6 h. Second, we examine realistic COSMO-KENDA simulations, including assimilation of radar and conventional data and a representation of model errors, for a convectively active two-week summer period over Germany. The results confirm increased predictability in orographic regions. We find that both latent heat nudging and ensemble Kalman filter assimilation of radar data lead to increased forecast skill, but the impact is smaller than in the idealized experiments. This highlights the need to assimilate spatially and temporally dense data, but also indicates room for further improvement. Finally, the examination of operational COSMO-DE-EPS ensemble forecasts for three summer periods confirms the beneficial impact of orography in a statistical sense and also reveals increased predictability in weather regimes controlled by synoptic forcing, as defined by the convective adjustment time scale.
Abstract
We investigate the practical predictability limits of deep convection in a state-of-the-art, high-resolution, limited-area ensemble prediction system. A combination of sophisticated predictability measures, namely, believable and decorrelation scale, are applied to determine the predictable scales of short-term forecasts in a hierarchy of model configurations. First, we consider an idealized perfect model setup that includes both small-scale and synoptic-scale perturbations. We find increased predictability in the presence of orography and a strongly beneficial impact of radar data assimilation, which extends the forecast horizon by up to 6 h. Second, we examine realistic COSMO-KENDA simulations, including assimilation of radar and conventional data and a representation of model errors, for a convectively active two-week summer period over Germany. The results confirm increased predictability in orographic regions. We find that both latent heat nudging and ensemble Kalman filter assimilation of radar data lead to increased forecast skill, but the impact is smaller than in the idealized experiments. This highlights the need to assimilate spatially and temporally dense data, but also indicates room for further improvement. Finally, the examination of operational COSMO-DE-EPS ensemble forecasts for three summer periods confirms the beneficial impact of orography in a statistical sense and also reveals increased predictability in weather regimes controlled by synoptic forcing, as defined by the convective adjustment time scale.
