COSMO General Meeting September 2009, Offenbach (DWD), Germany Considerations on convection issues Federico Grazzini, ARPA-SIMC Emilia-Romagna COSMO General Meeting September 2009, Offenbach (DWD), Germany some conclusions so far.. Benefit of aggregating precipitation (Eckert GM 2008, Tesini GM 2009) Overall better results for very high-res models for moderate- high precipitations intensity (Eckert GM 2008) Benefit of high-res models QPF is rather to see in situations with strong synoptic forcing, flow interaction with orography, embedded convection. Strongest impact at scales of 30 to 50 km. Not always convection is a limiting factor. COSMO General Meeting September 2009, Offenbach (DWD), Germany Heavy rain over the Appenines mountains: orographic forcing and embedded convection in a strong southerly flow. Operational precipitation forecasts from different model suites FC+36/48 Averaged precip over warning area E (3000Km2) COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMOI7 Acc total precip Fc+12 Fc+36 29/ UTC COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMOI7 Acc Conv precip Fc+12 Fc+36 29/ UTC COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMOI2 Acc total precip Fc+12 Fc+36 29/ UTC COSMO General Meeting September 2009, Offenbach (DWD), Germany Heavy rain over the Appenines mountains flow: despite relevant presence of convection and instability the COSMOI2 forecast was very accurate, why? Greater loss of predictability occurs over moist convectively unstable regions that are able to propagate energy against the mean flow (absolute instability). From linear analysis of absolute instability theory we see that for U > Ucrit=Nm 2 /(m 2 +k 2 )*sqrt /(m 2 +k 2 ) upstream propagation of gravity waves is inibithed, growing perturbation are swept away by lateral boundary condition, increase of predictability. Hohenegger et al Mon. Wea. Rev. For the choice of m and k (vertical and horizon wave number) we assume that gravity waves induced thermal perturbation are up to 10Km in vertical 50 Km in horizontal. COSMO General Meeting September 2009, Offenbach (DWD), Germany Absolute instability* as defined in Hohenegger et al Mon. Wea. Rev. Heavy rain over the Appenines mountains On this day very high precip were recorded, in the order of mm/24h with peaks during convective events of 30mm/h U > Ucrit Wind velocity U 500 hPa 29/10/ UTC White areas U>Ucrit COSMO I2 integration domain COSMO General Meeting September 2009, Offenbach (DWD), Germany some conclusions so far.. Precipitation scores increase with box size (Eckert GM 2008) Overall better results for very high-res models for moderate- high precipitations intensity (Eckert GM 2008) This suggest that the benefits of very high-res models is rather to see in situations with strong synoptic forcing, flow interaction with orography, embedded convection. Strongest impact at scales of 30 to 50 km. Not always convection is a limiting factor. This is confirmed by experience and case studies. On those cases COSMO2 is often better than COSMO7 due excessive convective precip upwind to the orography of the latter. Some time there is also an excessive feedback of parametrized convection in to the dynamics. What about weakly forced convection ? We present a case study over N-Italy illustrating this situation. COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMO General Meeting September 2009, Offenbach (DWD), Germany Weak synoptic forcing, cold air advection at upper levels north of the Alps COSMO General Meeting September 2009, Offenbach (DWD), Germany A case of non equilibrium convection: building up of CAPE over longer time-scales, high spatial variability of CIN values, convection enhanced by cold pools from previous systems (Done et al QJR Meteorol. Soc.) Floodings outside Parma 27/05/ /05/2009 14:00 Colorno (PR) /05/2009 14:45 Parma T. P. (PR) /05/2009 14:45 Parma (PR) /05/2009 15:15 Casatico Langhirano (PR) /05/2009 15:45 Collagna (RE) /05/2009 16:00 Ligonchio (RE) 30 27/05/2009 16:00 Collagna (RE) 25 Observed precipitation in 15 minutes COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMOI2 FC+12 Precip, CAPECOSMOI2 FC+12 Precip, CIN Precipitation is anticipated and weaker COSMO General Meeting September 2009, Offenbach (DWD), Germany Hourly animation of COSMOI2 forecast of Precip (grey), 2mT,10m wind COSMO General Meeting September 2009, Offenbach (DWD), Germany Expectations from high resolution models Experience is leading to more firm ideas that models are more likely to be accurate in some situations than others. Strong surface influence - > hoped for predictability -> (non-linear) downscaling. Internal organisation -> questionable predictability -> statistical interpretation or nowcasting. These ideas are becoming more objective, especially for convective rainfall, e.g.: Role of mesoscale convergence lines (lots of literature). Role of convective inhibition (CIN) in controlling predictability (Done et al, 2006, Q.J.R.Met.S., CSIP IOP1/18). Role of gravity waves in upstream triggering of convection and hence predictability (Hohenegger et al. 2006, Mon. Wea. Rev.). Role of gravity waves in downstream triggering of convection interaction with CIN (Browning et al, 2005, Q.J.R.Met.S.) and hence organisation. From Peter Clark, RMS Conference Reading 2009 So running at high resolution will not ease the problem of model interpretation COSMO General Meeting September 2009, Offenbach (DWD), Germany Flow dependant use of models ? Conv. Equilibrium: Air Mass convection, widespread convective precipitating single cells (or multi cells) developing simultaneously. Cell organization it is not that complex. In this case parametrized convection could be appropriate Non Equilibrium: CIN is strong, CAPE accumulates, convection develops only in few limited spots but can develop with a strong mesoscale organization (role of cell internal dynamics become more important). Explicit convection should be superior than parametrized. COSMO General Meeting September 2009, Offenbach (DWD), Germany Probability precipitation forecast (PPF): for each grid point a frequency is computed counting how many grid points in the neighborhood (6 grid-points) and in time(2h) are above a defined threshold COSMOI2 PPF valid at 09 UTC 1h TP > 10mm How to improve ?....statistical postprocessing COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMO2 had a very limited diurnal cycle, the afternoon maximum was very weak. DWD introduced turlen=150m together with a modification of the subgrid cloudiness q_krit=1.6, clc_diag=0.5. For Germany this change has a quite positive impact, with more small scale convection, which is in many cases more realistic. Also the diurnal cycle has been improved. However there is still a tendency underestimate the frequency of convective systems in weakly forced cases. Convection is a bit too strong in strongly forced situations. Also the organization, propagation and lifetime of convective systems is often not good enough in non-equillibrium situations, but sometimes it is surprisingly good. In 2009 also the initiation of convection was much better. We still miss some isolated cells or even systems in weakly forced non-equilibrium situations (Seifert, personal communication). At MeteoSwiss they find a negative impact of the reduced mixing length getting to much convection over the Alps.They have now operationally introduced turlen=250m combined with q_krit=1.6, clc_diag=0.5. At ARPA-SIMC we recently upgraded to version 4.9, running I2 with turlen=250m. This had a quite remarkable impact on the diurnal cycle of scattered convection with a substantial increase in number of cells. Distribution and frequency looks realistic now, the intensity seems a bit too strong. NEW Few weeks ago ETH (thanks to Peter Brockhaus and Daniel Luethi) released the IFS convection scheme library to be tested and implemented in COSMO (7). How to improve ?....better physical description of PBL A brief overview of COSMO convection permitting implementations COSMO General Meeting September 2009, Offenbach (DWD), Germany Moist Processes in Future High Resolution NWPModels June 15-17, 2009 Norrkping, Sweden Topological issues: domains often are tailored on the shape of the country. Choice of the domain should be more process oriented. There will always be processes that act on scales smaller than the horizontal resolution, the need of parametrization will not ease with increased resolution. Importance of Areosols although they may add further complexity to the system Worrying that an artificial parameter such as horizontal diffusion has such a strong influence, is that showing the need for different parametrizations? 1D param are still useful at high resol? Is it a problem that we treat convection as fully mature, without intermediate stage? What is the non-mature stage of convection? Data assimilation - Which technique to use? - How to best use radar - Avoid touching large and synoptic waves, modify only small scales Verification COSMO General Meeting September 2009, Offenbach (DWD), Germany Concluding remarks Despite the clear benefit of running NWP models at convective-scale resolutions for the representations of local effects, weakly forced convection is still problematic with a general underestimation by COSMO2 A not correct representation of nocturnal PBL evolution could have a negative effect on diurnal convection preventing or anticipating the triggering. (sensitivity on PBL parameters, sensitivity to turlen and horizontal diffusion) More sophisticated microphysics schemes could help in the correct representation of life cycle and organization of deep convection, especially for the formation of downdrafts and of cold pools. (implementations of the new Seifert and Beheng hierarchical two-moment scheme) Running different implementations of the model may represents a great potential for diagnostic investigation. However it is difficult to exploit this potential due to the limited exchange of data between the different operational centres COSMO General Meeting September 2009, Offenbach (DWD), Germany COSMO PRIORITY TASK ON TACLING DEFICIENCIES ON PARAMETRIZATION OF CONVECTIVE SCHEME Task leader: Federico Grazzini ARPA-SIMC Bologna (Italy) People involved and amount of work needed: F. Grazzini (ARPA-SIMC), case studies evaluation and preparation of diagnostic tools, 2 months P. Mercogliano (CIRA), implemementation of IFS scheme, 5 months W. Interewicz (IMGW Poland), to run experiment and case studies, 4 months A.Seifert and D. Mironov (DWD), scientific support, in total 1 month We are looking for a collaboration also with P. Brockhaus for the implementation of the IFS convection library developed by ETH (Zurich) to be plugged in the COSMO model. Total duration 1 year, March February 2010 Total FTE = 1 Motivation Localised pattern of very high precipitation are often produced by COSMO 7Km in connection with meteorological situations characterised by high thermodynamic instability. An inspection of the fields reveals that in many cases the largest percentage of precipitation is falling as large-scale and is not, as one would expect, coming from the result of the parametrised convection. In such occasions unrealistically high vertical velocities (few m/s at mid-level over 4/5 contiguous grid- points) and localised surface pressure tendencies suggest that the model is resolving explicitly convection at its own scale, producing a spurious upscaling of convection and consequently an overestimate of fluxes. This behaviour could be attributed to a lack of triggering of the deep convection parametrisation scheme. On the other end in 2008 DWD has introduced few modification of the Tiedke89 convective scheme introducing : a water-ice mixed phase, a detrainment of cloud condensate, decreasing of environment humidity threshold for evaporation. These changes were introduced in order to in account of real physical process not yet represented in the former version of the Tidke scheme and already present in more advanced versions like the one developed at ECMWF. These changes resulted in a decrease of unrealistic convective precipitation events. Aim of the package To document the lack of triggering in a more systematic way, also taking advantage of the large numbers of experiments and sensitivities conducted during the QPF projects. We have selected among them a subset of pure convective cases (6 ) that we want to examine in greater detail, not only restricting to precipitation but also to other dynamical fields. To assess the impact of modification introduced in 2008 also in other operational model domains and case studies. Finally to implement the IFS convection code in to the COSMO model and to produce a substantial comparison with the above two versions Milestones March to September Analysis and diagnosis of the fields to be summarised in a memo and/or to be presented at the COSMO General Meeting October to February Implementation of the IFS scheme in the COSMO code, re-run few experiments, assessing the quality of this scheme and to report at the User Seminar i Avevo mandato questo da discutere al User meeting ma poi non se ne fatto niente.non ho rispedito la versione aggiornata COSMO General Meeting September 2009, Offenbach (DWD), Germany Main sensitivities. Why triggering of convection is critically dependent on the time of I.C.? Can we classify conditions where 7Km run is better than 2Km run (o viceversa). Any regime dependency ? Parametrisation vs explicit, how this relate with the flow of energy cascade (dowscale,upscale) ? Mixing length sensitivity Horizontal diffusion sensitivity (controls intensity, cell structure, timing). COSMO General Meeting September 2009, Offenbach (DWD), Germany Convective rate Change to convection scheme introduced with Version 4.4 COSMO General Meeting September 2009, Offenbach (DWD), Germany Total precipitation statistic compared against surface obs. Positive bias in winter Change to convection scheme introduced with Version 4.4 COSMO General Meeting September 2009, Offenbach (DWD), Germany Relative bias Change to convection scheme introduced with version 4.4 COSMO General Meeting September 2009, Offenbach (DWD), Germany Spatial scale (km) Fuzzy Verification: COSMODE COSMOEU good bad Threshold (mm/3h) - = Fraction skill score Upscaling = - COSMO-EU (7km)COSMO-DE (2.8km) Difference COSMO-EU better COSMO-DE better JJA 2007, Verification against Swiss Radar Composite, 3 hourly accumulations COSMO General Meeting September 2009, Offenbach (DWD), Germany Spatial scale (km) good bad Threshold (mm/3h) - = Fraction skill score Upscaling = - COSMO-7 (7km)COSMO-2 (2.2km) Difference COSMO-7 better COSMO-2 better Fuzzy Verification: COSMO2 COSMO7 JJA 2007, Verification against Swiss Radar Composite, 3 hourly accumulations