Model Uncertainty Representation in a Convection-Permitting Ensemble—SPP and SPPT in HarmonEPS

Andræ, Ulf; Ollinaho, Pirkka; Hally, Alan; Hämäläinen, Karoliina; Kauhanen, Janne; Ivarsson, Karl-Ivar; Yazgi, Daniel

doi:10.1175/mwr-d-21-0099.1

Cited by 9 publications

(5 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The MEPS ensemble consists of a deterministic and 14 perturbed members. Stochastic perturbations are introduced to lateral boundary forcing, observations, analysed model state, and certain parameters in model physics description as documented in Frogner et al (2019) and Frogner et al (2022). The ensemble is distributed in time such that five new members are launched every hour.…”

Section: Metcoop Ensemble Prediction System (Meps)mentioning

confidence: 99%

Variational Bias Correction of Polar-Orbiting Satellite Radiances in Convective-Scale Data Assimilation

Eresmaa

2024

Tellus A: Dynamic Meteorology and Oceanography

View full text Add to dashboard Cite

Maintenance of robust bias correction is a major challenge in the assimilation of meteorological polar-orbiting satellite data into limited-area Numerical Weather Prediction (NWP) systems. This article presents a variant of the variational bias correction algorithm suitable for use in convection-resolving systems. Stable bias correction requires continuous and representative sampling of predictor variables such as satellite view angles and air-mass properties. In convection-resolving NWP systems, the sampling is often compromised because of small computing domains, short assimilation time windows, and large diurnal variation in data availability.The proposed variant is designed around the assumption of one recurring daily analysis hour at which a given satellite provides comprehensive data coverage inside the computing domain. The idea is to allow the variational algorithm to adjust the bias correction coefficients at that analysis hour only, and otherwise keep the updated coefficients constant during the analysis. The time of the daily coefficient update is to be specified separately for each satellite, taking account of the satellite orbit parameters. The proposal is an alternative to the widely-adopted operational practice where independent streams of coefficients are maintained and updated separately at each analysis hour.The proposal is evaluated by data assimilation experiments in the context of a stateof-the-art Northern-European limited-area NWP system. In comparison with the operational setup, the proposed method is found to slightly improve the satellite radiance data fit to the NWP model background. Nevertheless, verification against independent data sources indicates no solid and statistically significant impact on forecast system performance.

show abstract

Section: Metcoop Ensemble Prediction System (Meps)mentioning

confidence: 99%

Variational Bias Correction of Polar-Orbiting Satellite Radiances in Convective-Scale Data Assimilation

Eresmaa

2024

Tellus A: Dynamic Meteorology and Oceanography

View full text Add to dashboard Cite

show abstract

“…SPPT often appears to generate not enough spread in ensemble forecasts, preventing them from producing reliable probabilistic forecasts, see e.g. Christensen et al (2017); Frogner et al (2022).…”

Section: Equation (mentioning

confidence: 99%

Additive Model Perturbations Scaled by Physical Tendencies for Use in Ensemble Prediction

Tsyrulnikov¹,

Astakhova²,

Gayfulin³

2023

Preprint

View full text Add to dashboard Cite

Imperfections and uncertainties in forecast models are often represented in ensemble prediction systems by stochastic spatio-temporal perturbations of model equations. A new technique for this purpose termed Additive Model-uncertainty perturbations scaled by Physical Tendencies (AMPT) is proposed in this article. AMPT employs the previously developed Stochastic Pattern Generator (Tsyrulnikov and Gayfulin, 2017) to generate pseudo-random space and time correlated 4D perturbation fields. The perturbations are independent for different model variables and scaled by the local-area averaged modulus of physical tendency in the respective model variable. AMPT attempts to address weak points of the popular model perturbation scheme known as Stochastically Perturbed Parametrization Tendencies (SPPT). AMPT is capable of producing non-zero perturbations even at grid points where physical tendency is zero and avoids perfect correlations in the perturbation fields in the vertical and between different variables. Due to the non-local link from physical tendency to the local perturbation magnitude, AMPT can generate significantly greater perturbations than SPPT without causing instabilities. Relationships between biases and spreads caused by AMPT and SPPT were studied in an ensemble of forecasts. The non-hydrostatic, convection permitting forecast model COSMO was used. In ensemble prediction experiments, AMPT perturbations led to statistically significant improvements (as compared to SPPT) in probabilistic performance scores such as spread-skill relationship, CRPS, Brier Score, and ROC area for near-surface temperature, and mixed scores for precipitation.

show abstract

“…In some models, with their specific configurations of model perturbation schemes, SPPT appeared to generate not enough spread in ensemble forecasts, preventing them from producing reliable probabilistic forecasts, see e.g. Christensen et al (2017); Frogner et al (2022). One possible reason for that is the limitation on the magnitude of perturbation, Eq.…”

Section: Motivationmentioning

confidence: 99%

Additive Model Perturbations Scaled by Physical Tendencies for Use in Ensemble Prediction

Tsyrulnikov,

Astakhova,

Gayfulin

2023

Tellus A: Dynamic Meteorology and Oceanography

View full text Add to dashboard Cite

Imperfections and uncertainties in forecast models are often represented in ensemble prediction systems by stochastic perturbations of model equations. In this article, we present a new technique to generate model perturbations. The technique is termed Additive Model-uncertainty perturbations scaled by Physical Tendencies (AMPT). The generated perturbations are independent between different model variables and scaled by the local-area-averaged modulus of physical tendency. The previously developed Stochastic Pattern Generator is used to generate space and time-correlated pseudo-random fields. AMPT attempts to address some weak points of the popular model perturbation scheme known as Stochastically Perturbed Parametrization Tendencies (SPPT). Specifically, AMPT can produce non-zero perturbations even at grid points where the physical tendency is zero and avoids perfect correlations in the perturbation fields in the vertical and between different variables. Due to a non-local link from physical tendency to the local perturbation magnitude, AMPT can generate significantly greater perturbations than SPPT without causing instabilities. Relationships between the bias and the spread caused by AMPT and SPPT were studied in an ensemble of forecasts. The non-hydrostatic, convection-permitting forecast model COSMO was used. In ensemble prediction experiments, AMPT perturbations led to statistically significant improvements (compared to SPPT) in probabilistic performance scores such as spread-skill relationship, CRPS, Brier Score, and ROC area for near-surface temperature. AMPT had similar but weaker effects on near-surface wind speed and mixed effects on precipitation.

show abstract

Model Uncertainty Representation in a Convection-Permitting Ensemble—SPP and SPPT in HarmonEPS

Cited by 9 publications

References 31 publications

Variational Bias Correction of Polar-Orbiting Satellite Radiances in Convective-Scale Data Assimilation

Variational Bias Correction of Polar-Orbiting Satellite Radiances in Convective-Scale Data Assimilation

Additive Model Perturbations Scaled by Physical Tendencies for Use in Ensemble Prediction

Additive Model Perturbations Scaled by Physical Tendencies for Use in Ensemble Prediction

Contact Info

Product

Resources

About