Arpae Hydro-Meteo-Climate Service, Bologna, Italy Predictability of severe weather phenomena with a convection-permitting ensemble Chiara Marsigli, Andrea Montani, Tiziana Paccagnella Arpae Hydro-Meteo-Climate Service, Bologna, Italy With thanks also to Virgina Poli, Thomas Gastaldo and Maria Stefania Tesini
Outline Aim of this work The COSMO-IT-EPS ensemble Evaluation of the impact of the perturbations => focus on the precipitation forecast Conclusions
Aim of this work Assess the role of the different perturbations applied to the ensemble: IC perturbations (LETKF) Model perturbations: SPPT, parameter perturbation Study the effect of combining different model perturbations Show which components of the forecast uncertainty are represented in the ensemble spread
COSMO-IT-EPS 2.2 km 65 v.l. LETKF (Arpae) EnKF DA (COMET) COSMO-ME-EPS (COMET) 10 km 45 v.l. ICs BCs 10 km 40 v.l. parameterized convection COSMO-IT-EPS (Arpae) 2.2 km 65 v.l. explicit convection
Impact of model physics perturbation on autumn precipitation Model perturbations: Exp1: no model perturbation (CTRL) Exp2: SPPT Exp3: SPPT + Parameter Perturbation October 2015
6h precipitation - verification over boxes of 0.2 x 0.2 deg average > 5mm maximum > 5mm
Evaluation of the ensemble spread Compute dFSS (FSS between all pairs of ensemble members) Compute SAL between all pairs of ensemble members
Evaluation of ensemble spread using the SAL metric What is the perturbation influencing? Precipitation intensity Precipitation structure Localisation of the precipitation Use a spatial verification measure: SAL (Wernli et al 2008) 3 independent components: Structure Amplitude Location Used here not for verification but for evaluating the similarity between fields, only forecasts
SAL-spread metric 10/10/15 no physics pert SPPT SPPT + PP
Impact of perturbed ICs from LETKF on thunderstorm prediction summer 2016
20 June 2016 20 June 2016 ITEPS only physics Red areas: radar estimate Probability maps +7h +8h 1 10 30 50 70 90 %
20 June 2016 ITEPS physics + IC +1h +2h +3h +4h +5h +6h +7h +8h Red areas: radar estimate Probability maps +7h +8h 1 10 30 50 70 90 %
25 June 2016 ITEPS only physics Red areas: radar estimate Probability maps +5h +6h 1 10 30 50 70 90 %
25 June 2016 ITEPS physics + IC +1h +2h +3h +4h +5h +6h Red areas: radar estimate Probability maps +5h +6h 1 10 30 50 70 90 %
2 July 2016 – ITEPS only physics Red areas: radar estimate Probability maps +10h +11h 1 10 30 50 70 90 %
2 July 2016 – ITEPS physics + IC Red areas: radar estimate Probability maps +10h +11h 1 10 30 50 70 90 %
radar estimate of hourly precipitation 20 June 2016 radar estimate of hourly precipitation
forecasted hourly precipitation – physics only - member 1
forecasted hourly precipitation – physics + IC - member 1
radar estimate of hourly precipitation 20 June 2016 radar estimate of hourly precipitation
forecasted hourly precipitation – physics only - member 1
forecasted hourly precipitation – physics only - member 2
Concluding remarks The skill of the precipitation forecast is improved, with a better discrimination ability, by the model-perturbed ensembles Adding parameter perturbation has no significant impact on the skill over the entire period Spread in precipitation localisation and structure is highlighted by the SAL-spread metric Perturbed initial conditions have a positive impact during the first hours of the forecast Operational use: spatial and temporal aggregation dependent on the predictability of the phenomena
forecasted hourly precipitation – physics only - member 3
forecasted hourly precipitation – physics + IC - member 2
forecasted hourly precipitation – physics + IC - member 3