High resolution forecast of heavy precipitation with Lokal Modell: analysis of two case studies in the Alpine area

Transcription

1 Natural Hazards and Earth System Sciences, 5, , 5 SRef-ID: /nhess/ European Geosciences Union 5 Author(s). This work is licensed under a Creative Commons License. Natural Hazards and Earth System Sciences High resolution forecast heavy with Lokal Modell: analysis two case studies in Alpine area M. Elementi, C. Marsigli, and T. Paccagnella ARPA-SIM (Servizio Idro-Meteorologico della Regione Emilia-Romagna) Bologna, Viale Silvani 6, 4122 Bologna, Italy Received: 4 April 5 Revised: 18 May 5 Accepted: 2 June 5 Published: 2 August 5 Part Special Issue HYDROPTIMET Abstract. Norrn Italy is frequently affected by severe conditions ten inducing flood events with associated loss properties, damages and casualties. The capability correctly forecast se events, strongly required for an efficient support to civil protection actions, is still nowadays a challenge. This difficulty is also related with complex structure field in Alpine area and, more generally, over Italian territory. Recently a new generation non hydrostatic meteorological models, suitable to be used at very high spatial resolution, has been developed. In this paper performance non hydrostatic Lokal Modell developed by COSMO Consortium, is analysed with regard to a couple intense events occurred in Piemonte region in Norrn Italy. These events were selected among reference cases Hydroptimet/INTERREG IIIB project. LM run at operational resolution 7 km provides a good forecast general rain structure, with an unsatisfactory representation distribution across mountain ranges. It is shown that inclusion new prognostic equations for cloud ice, rain and snow produces a remarkable improvement, reducing in upwind side and extending intense rainfall area to downwind side. The unrealistic maxima are decreased towards observed values. The use very high horizontal resolution (2.8 km) improves general shape field in flat area Piemonte region but, keeping active moist convection scheme, sparse and more intense rainfall peaks are produced. When convective is not parametrised but explicitly represented by model, this negative effect is removed. 1 Introduction The aim this paper is numerical investigation two meteorological cases which can be considered representative intense events frequently affecting Norrn Italy during autumn season. The synoptic description se case studies is presented in paper by Milelli et al. (5) 1. The new generation non-hydrostatic high resolution models permits a better description mesoscale processes and associated events. In presence complex topography, as in Alpine area, se models should also improve representation interaction between meteorological structures and orography. Orographic forcing plays in fact a major role in localisation and intensification in this region, where moist Mediterranean flows undergo to sudden uplift. The capability numerical models to forecast correctly local and intense is still nowadays limited. This is true even at short time-range, up to 48 h, due to loss atmospheric predictability going down to small spatial and temporal scales. With regard to high-resolution modelling (1 to km), it is not completely satisfactory just to reproduce precipitating structures, but a correct localisation in space and time is required toger with realistic peak values. This is even more crucial when hydrological models need to be coupled in a full hydro-meteorological chain. A representative review state---art in field high resolution modelling intense in Alpine area can be found in Special Issue Quarterly Journal Royal Meteorological Society devoted to MAP (Mesoscale Alpine Programme) results (Bougeault et al., 3). Correspondence to: M. Elementi (melementi@smr.arpa.emr.it) 1 Milelli, M., Llasat, C., and Ducrouq, V.: The cases June, November 2 and September 2 as examples Mediterranean floods., Nat. Hazards Earth Syst. Sci., submitted, 5.

2 594 M. Elementi et al.: High resolution simulations heavy 4 E 8 E 12 E 16 E E ªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª ª 48 N 44 N 4 N ª Fig. 1. Integration domain and orography field (in decameters) LM used operationally by ARPA SIM (LAMI). The meteorological model employed for numerical simulations presented in this work is Lokal Modell (LM), operationally managed by ARPA SIM ( regional HydroMeteorological Service Emilia-Romagna) since 1. LM is developed by COSMO Consortium (COnsortium for Small-scale MOdelling), which coordinates cooperation Germany, Italy, Switzerland, Greece and Poland ( LM is a fully-compressible (non-hydrostatic) primitive equation model without any scale approximation. Due to unfiltered set equations, vertical momentum equation is not approximate, allowing a better description nonhydrostatic phenomena such as moist convection, breeze circulations and some kind mountain-induced waves. The Italian implementation LM (LAMI), managed by ARPA SIM in framework an agreement among UGM (Ufficio Generale di Meteorologia), ARPA SIM and ARPA Piemonte, consists two runs a day (at : and 12: UTC) for 72 h with a spatial horizontal resolution 7 km and 35 levels in vertical on domain represented in Fig. 1. Hourly boundary conditions are provided by DWD (Deutscher Wetterdienst) global model GME ( one-way nesting ). Initial conditions are obtained through a mesoscale data assimilation system based on nudging technique (Schraff and Hess, 3). LM contains a full set physical parametrisations, described in Cosmo Newsletter number 4 (4). One major problems Limited-Area Models, in forecasting intense in presence a mountain range, is tendency to overestimate rainfall in upwind areas, with a related drying effect in downwind regions. This error heavily reduced usefulness Quantitative Precipitation Forecast (QPF) as input to hydrological forecasting models. In order to improve forecast distribution, some recent improvements LM have been tested. The operational set-up model at time experiments included only two micro-physical species: wa- 4 E 8 E 12 E 16 E ª ª ª 48 N 44 N 4 N Fig. 2. Integration domain used for simulations at 7 km horizontal resolution (top panel) and at 2.8 km horizontal resolution (bottom panel). ter vapour and liquid water. The addition a prognostic treatment rain and snow (Gassmann, 2; Baldauf and Schulz, 4) has been tested, toger with recently developed cloud ice scheme (Doms, 2). To test impact on forecast higher horizontal resolution, simulations se case studies have been performed also at 2.8 km. The boundary conditions for se runs have been provided every hour by 7-km model runs through a one-way nesting procedure. The paper is organised as follows: in Sect. 2 a short description case studies is given, toger with model configuration used. In Sect. 3 results simulations are presented, comparing forecasts obtained with different model set-up. Conclusions are drawn in Sect Case studies and model configuration The case studies presented in this paper have been selected within Hydroptimet Project (an EU/INTERREG IIIB, area MEDOCC project), whose aim was to improve short

3 M. Elementi et al.: High resolution simulations heavy E 9 E E 11 E 12 E 13 E PRECIPITAZIONE CUMULATA SU 24 ORE RIS. 7 KM Friday 15 November 2 UTC OFFNB Forecast t+18 VT: Friday 15 November 2 18UTC 8hPa v-component wind/u-component wind/ u-component wind/v-component wind 4 E 6 E 8 E E 12 E 14 E 16 E 1 46 N 44 N precipitazione E cumulata 11 E(mm) 12 E 24 h13 E 6-6 fino al 16/11/ N 4 N 4 E 6 E 8 E E 12 E 14 E 16 E 25.m/s Fig. 4. Wind at 8 hpa forecast at +18 h by OPE 7KM run starting at : UTC 15 Nov. 2 (first event). PREC CUM 8 E 24 ORE 9 E E +3 RIS. 13 E 2.8 KM Fig. 3. Cumulated rain field (mm) between 6: UTC 15 Nov. 2 and 6: UTC 16 Nov. 2 (first event) for simulations OPE 7KM (top panel) and OPE 2p8KM (bottom panel). Observational data are shown in medium panel, colour scale is same or two figures; empty diamonds represent stations where no is observed. range forecast hydro-meteorological events leading to natural hazards, in order to provide guidance during Civil Protection alert conditions. The two cases are characterised by severe events affecting Norrn Italy and mainly Piemonte region: first event occurred between 14 and 16 November 2; observed cumulated over 24 h starting from 15 November 2 6: UTC is shown in middle panel Fig. 3; exceeding mm/24 h were recorded over most Norrn 1 Piemonte, where sparse peaks were also above 1 mm, and along Sourn border Piemonte. On 12 November synoptic situation showed a weak ridge over Central Mediterranean Sea and a trough stretching from North-western Europe to Atlantic coastline Iberian Peninsula. During day, axis trough gradually moved eastward, and upper level winds became south-westerly over all Norrn Italy. This situation permained during following day, because high pressure area present over Eastern Europe, which was blocking eastward motion trough. A minimum sea level pressure was moving from Ireland to Iberian Peninsula and a secondary minimum formed in afternoon 15th over Balearic Islands, which gradually moved towards Ligurian Gulf. second event occurred between 24 and 26 November 2; observed cumulated over 24 h starting from 25 November 2 6: UTC is shown in middle panel Fig. 5; exceeded mm/24 h over a vast Alpine area, with peaks above mm over Norrn Piemonte and also above 1 mm in Liguria region. The synoptic situation was characterised by a deep low f Irish coastline, that, from 23 November, began to expand southward, reaching North-African coastline and directing moist air towards Norrn Italy. Due to long duration sourly flow, height freezing level was constantly increasing and relative humidity recorded was always very high, close to % up to m. A more complete meteorological description se case studies is given in paper by Milelli et al. (5) 1. Both cases are characterised by presence a large scale forcing which determines a moist south-westerly flow

4 596 M. Elementi et al.: High resolution simulations heavy PREC CUM 8 E 24 ORE 9 E E +3 RIS. 13 E7 KM precipitazione E cumulata 11 E(mm) 12 E 24 h13 E 6-6 fino al 25/11/ Table 1. Analysed simulations for first event. 7 km resolution Initial time and date Type Label : UTC 15 Nov. 2 OPE OPE 7KM : UTC 15 Nov. 2 QI QI 7KM : UTC 15 Nov. 2 PROGN PROGN 7KM 2.8 km resolution Initial time and date Type Label : UTC 15 Nov. 2 OPE OPE 2p8KM : UTC 15 Nov. 2 QI QI 2p8KM : UTC 15 Nov. 2 PROGN PROGN 2p8KM : UTC 15 Nov. 2 PRO EXP PRO EXP 2p8KM Table 2. Analysed simulation for second event. PREC CUM 8 E 24 ORE 9 E E +3 RIS. 13 E 2.8 KM Fig. 5. Cumulated rain field (mm) between 6: UTC 24 Nov. 2 and 6: UTC 25 Nov. 2 (second event) for simulations OPE 7KM (top panel) and OPE 2p8KM (bottom panel). Observational data are shown in medium panel, colour scale is same or two figures; empty diamonds represent stations where no is observed. over Mediterranean Sea directed towards Western Alps. The heavy is mostly due to uplift moist air by steep orography, while a convective enhancement, due to local effects, is embedded in large scale systems. An improvement QPF for se cases can be expected from a more accurate modelling physical processes which determine distribution. The LM simulations at 7 km horizontal resolution presented in this paper are carried out in a configuration somehow different from operational one. The initial condition and 3-hourly boundary conditions are supplied by 1 7 km resolution Initial time and date Type Label : UTC 24 Nov. 2 OPE OPE 7KM 2.8 km resolution Initial time and date Type Label : UTC 24 Nov. 2 OPE OPE 2p8KM operational ECMWF (European Centre for Medium-range Wear Forecasts) global model. The mesoscale assimilation cycle (nudging) is not applied. The model version operational in 4 (3.5), released by COSMO in September 3, is used. The 7-km runs domain is shown in top panel Fig. 2. The 2.8-km simulations, whose domain is shown in bottom panel Fig. 2, are nested on 7-km simulations having same features. The features simulations are in Table 1 for first case study and in Table 2 for second one. The terms used in se tables are explained here: # Initial time and date: represents starting point simulation; # Type: OPE: operational setting with prognostic cloud water, no cloud ice, diagnostic rain and snow; QI: cloud ice added as prognostic variable, diagnostic rain and snow;

5 M. Elementi et al.: High resolution simulations heavy E 9 E E 11 E 12 E 13 E PRECIPITAZIONE CUMULATA SU 24 ORE RIFE RIS. 7 KM 8 E 9 E E 11 E 12 E 13 E PRECIPITAZIONE CUMULATA SU 24 ORE PROGN RIS. 7 KM 1 1 Fig. 6. Cumulated rain field (mm) between 6: UTC 15 Nov. 2 and 6: UTC 16 Nov. 2 for QI 7KM run (left panel) and PROGN 7KM one (right panel). PROGN: cloud ice added as prognostic variable, rain and snow also treated as prognostic variables; PRO EXP: Tiedtke scheme precipitating convection not used; cloud ice added as prognostic variable, rain and snow also treated as prognostic variables. # Label: name used in text to refer to simulation; The parametrisation precipitating convection, based on Tiedtke scheme (Tiedtke, 1989), is actived in all 7- km simulations presented here. Moving to 2.8-km resolution, where an explicit representation cumulus convection should be allowed, Tiedtke scheme has been switched f in one first case runs (PRO EXP 2p8KM run, nested on PROGN 7KM run). Every simulation, both at 7 km both at 2.8 km resolution, lasts for 36 h. 3 Results 3.1 Impact resolution Referring to first case, in Fig. 3 forecast rain field is shown for OPE 7KM (top panel) and for OPE 2p8KM (bottom panel) runs; is cumulated over 24 h, from 6: UTC to 6: UTC. The observed for same period is also shown in medium panel. This accumulation period allows to compare forecast against observations over whole area where heavy s were recorded, since Swiss observations are available only cumulated over this 24-h period. The 7-km simulation produces a good forecast as regards structure rain field and its localisation, but is overestimated over Western Alps, between 9 and degrees East. This is linked to above mentioned difficulty models to provide adequate values rain field across mountain ridges. Anor difference with observations is underestimation over flat part Piemonte region in Western Po Valley. Considering that middle-lower tropospheric flow is mainly from south (see horizontal wind at 8 hpa forecast by OPE 7KM simulation at +18 h, Fig. 4), it is evident overestimation is in upwind side Alps (Norrn Piemonte and Lombardia regions), while in downwind side (mainly in Switzerland) re is an underestimation. When resolution is increased up to 2.8 km, it is remarkable breaking-up rain field with respect to correspondent 7-km simulation (comparing bottom panel with top one in Fig. 3). At this resolution an appreciable overestimation is also detectable over Liguria. It is natural to link this breaking-up field to finer resolution representation orography, even if associated improvement forecast, if any, is difficult to be evaluated. This difficulty is due to inadequacy observational network. The availability a good estimate field from meteorological radars will help, in future, to face this problem. This breaking-up effect can also be partially produced by convective scheme, not anymore suitable at this resolution. With regard to second case, results relative to OPE 7KM run and OPE 2p8KM run are shown in Fig. 5; is cumulated over 24 h, from 6: UTC to 6: UTC. In top panel 7-km forecast is shown, while 2.8-km one is shown in bottom panel. The observations for same period are reported in medium panel. The areas interested by intense are correctly forecast by Lokal Modell at 7 km, but is still overestimated in some regions, particularly over Western Alps between 8 and 9 degrees East. The same breaking-up structure when going to 2.8 km is observed (Fig. 5, bottom panel). As for first case, use higher resolution determines an increment values on highest mountains, where forecast rainfall maxima exceed in some points mm over 24 h. This feature highlights well known necessity to furr adapt numerical models to make m really suitable to be run at a such high resolutions as those associated to cloud resolving scales.

6 598 M. Elementi et al.: High resolution simulations heavy PREC CUM 8 E 24 ORE 9 E 211 E E E KM, CON 13 E QI (RIFE) 7 1 PREC CUM 8 E 24 ORE 9 E 211 E E E KM, RAIN 13 E PROGN E 9 E E 11 E 12 E 13 E PREC CUM 24 ORE RIS. 2.8 KM, CON PROGN., CONV. ESPL. 7 1 Fig. 7. Cumulated rain field (mm) between 6: UTC 15 Nov. 2 and 6: UTC 16 Nov. 2 for QI 2p8KM run (top panel), for PROGN 2p8KM one (bottom left panel) and for PRO EXP 2p8KM one (bottom right panel). longitude media 8.73, 7 km 6 E 8 E E 12 E ª 48 N 46 N 44 N 319. Fig. 8. The red line represents longitudinal cross-section ( shadowing is orography field and colour scale is in decameter). 3.2 The use new prognostic variables As already mentioned, LM has been recently improved by including new prognostic equations to better describe evolution, eir rain or snow. A cloud ice prognostic equation has been also included into model. These new equations should allow for a more physical description all processes related to QPF. The important aspect is that is allowed to be advected in surrounding boxes, while previously it was forced to fall out in same column where it was produced. These model changes are so in direction to improve forecast distribution over mountain ranges. This new features have been tested only for first case study In left panel Fig. 6 forecast by QI 7KM simulation (where cloud ice is added as prognostic variable, as defined in Sect. 2) is shown. LM simulation is starting at : UTC 15 November 2 and is cumulated over 24 h from +6 up to +3 h. A comparison with correspondent from OPE 7KM run (top panel Fig. 3) shows a general increase, leading to a furr increase unrealistic high values. The same is true also for correspondent 2.8 km simulations (top panel Fig. 7). On or hand, in PROGN 7KM simulation (where rain and snow are also treated as prognostic variables) a remarkable reduction on Western Alps and on Sourn France (Fig. 6, right panel) is observed. In particular, overestimation on upwind side Alps is greatly reduced, with a correspondent reduction drying effect in downwind area. Furrmore, structure forecast rain field is smoor, going in direction reducing local and unrealistic maxima (grid point storms). Similar considerations can be extended also to PROGN 2p8KM run (bottom left panel Fig. 7), where a significant improvement is also evident as regards increase in Western Po Valley. In bottom right panel Fig. 7 PROGN EXP 2p8KM run with explicit precipitating convection (i.e. without Tiedtke scheme) is shown. The impact this change is not dramatic. The field is generally smoor and localised maxima are reduced or removed. The differences among fields obtained by above described simulations, are hereafter analysed in greater detail considering a longitudinal cross-section both

7 Orography (km) Precipitation Orography (km) (mm) Precipitation (mm) Orography (km) Precipitation (mm) Precipitation (mm) Orography (km) Precipitation (mm) Precipitation (mm) M. Elementi et al.: High resolution simulations heavy 599 M. Elementi et al.: Latitude Latitude OROGRAPHY AND PRECIPITATION OROGRAPHY AND PRECIPITATION 2.8 KM ESPL PROGN TD RUN 3.5 LM VERSION 6-6 UTC FROM 15/11/2 OBSERVATIONS FROM 6 UTC OF 15/11/2 TO 6 UTC OF 16/11/2 (ORO 2.8) longitude Latitude media Latitude Latitude Latitude Fig. 9. LongitudinalFig. cross-sections 9. Longitudinal cross sections orography in inkm (blue curve) curve) and and in mm (red curve), cumulated in mm between (red curve), cumulated between 6 UTC 15/11/2 and 6 UTC 16/11/2, for QI 7KM run (top left panel), PROGN 7KM run (top right panel), QI 2p8KM 6: UTC 15 Nov. run (medium 2 and left panel), 6: PROGN UTC 2p8KM run 16(medium Nov. right 2, panel), for PRO EXPQI 2p8KM 7KM run (bottom run (top panel) left and for panel), observations PROGN (green 7KM run (top right panel), curve) interpolated on 2.8 km longitudinal cross section. The latitude is in abscissa. QI 2p8KM run (medium left panel), PROGN 2p8KM run (medium right panel), PRO EXP 2p8KM run (bottom panel) and for observations (green curve) interpolated on 2.8 km longitudinal cross-section. The latitude is in abscissa., leading to a furr increase unrealistic high values. The same is true also for correspondent overestimation on upwind side France (Fig. 6, right panel) is observed. In particular, 2.8 km simulations (top panel Fig. 7). On or hand, Alps is greatly reduced, with a correspondent reduction model orography in and PROGN 7KM forecast simulation rainfall (where rain along and snow are line drying effect in downwind area. Furrmore, also treated as prognostic variables) a remarkable reduction structure forecast rain field is smoor, going in shown in Fig. 8. The cross-sections on relative Western Alps toand ondifferent Sourn direction reducing local and unrealistic simulations are shown in Fig. 9, where orography is plotted in blue and in red. In bottom right panel Fig. 9 observed s, plotted in green, interpolated over same cross-section, are shown as a reference. With regard to QI 7KM simulation (top left panel), it is evident how highest and unrealistic values orography are forecast in upwind regions south two mountain peaks, while almost no is forecast downwind. Considering PROGN 7KM simulation (top right panel), forecast is more spread out across mountain range, with a realistic decrease maxima and a compensating increase minima. This is even more true for 2.8-km resolution runs (Fig. 9, medium left panel for QI 2p8KM run run and medium right panel for PROGN 2p8KM run). The unrealistic maxima to south mountain peaks, which reach even higher values due to increased resolution, are drastically reduced to more realistic values by adding prognostic in model. In PRO EXP 2p8KM (bottom left panel Fig. 9) it is confirmed smoor distribution and it is worth noting correct distribution across Apennines ridge ( sourn mountain peak in cross section). In order to furr highlight positive impact prognostic treatment and also to allow a comparison with observations, distribution over two boxes 1 degree per 1 degree is considered. orography The boxes are respectively upwind (Fig., blue one) and downwind ( red one) Alpine ridge in region where observed was most intense. In Fig. 11 distributions in upwind (left panel) and in downwind (right panel) boxes are shown. The observations (blue histogram) are compared with forecast values obtained by QI 7KM run (green histogram) and by PROGN 7KM run (red histogram). In upwind box, use prognostic removes presence forecast values exceeding mm, which are not observed, and tends to reduce values in range mm. The whole distribution forecast by PROGN 7KM run is more similar to observed one with respect to QI 7KM one.

8 6 M. Elementi et al.: High resolution simulations heavy 8 M. Elementi et al.: Fig.. Location two two considered boxes boxes 1 degree 1 degree x 1 degree, 1 degree, one inone indownwind downwind area (red) area and (red) and or inor upwind upwind area (blue), area (blue), over over orography orography field infield decameters. in decameters. maxima In (grid downwind point storms). box (Fig. Similar 11, right considerations panel), PROGN can be forecast extended also to PROGN distribution 2p8KM fits well run (bottom observed left panel one. In Fig. particular, 7), where while a significant QI 7KM improvement run forecasts is also evident as values regardsare mainly increase concentrated in in first Western bin ( Pomm), Valley. in In PROGN bottom7km rightrun panel re Fig. is a 7significant PROGN number EXP 2p8KM points where run wiexplicit forecastprecipitating values are bigger convection than (i.e. mm. without Tiedtke Considering scheme) is 2.8 shown. km resolution The impact runs this (Fig. change 12), is not occurrence dramatic. The belonging field is generally to first smoor two ranges and ( localised and mm) maxima in upwind are reduced area (left or removed. panel) is underestimated by both PROGN 2p8KM run and QI 2p8KM run. The very high values (over mm) forecast by QI The run differences and not observed, among are completely removed fields obtained by PROGN by above 2p8KM described run. Even simulations, if PROGN are hereafter 2p8KM analysed run still overestimates in greater detail considering a longitudinal in cross section mm range, re both is a general model orography improvement and forecast whole distribution. rainfall along line shown in Fig. 8. The cross sections relative to In downwind box (right panel), behaviour is very different simulations are shown in Fig. 9, where orography similar to that 7-km runs, distribution forecast by PROGN 2p8KM run being rar is plotted in blue and in red. In bottom right panel Fig. 9 observed s, plotted in similar to observed one. Since explicit convection green, interpolated over same cross section, are shown in PRO EXP 2p8KM did not produce a dramatic impact as a reference. With regard to QI 7KM simulation (top with respect to PROGN 2p8KM run in Alpine area, left panel), it is evident how highest and unrealistic correspondent histogram is not shown. values are forecast in upwind regions south two mountain peaks, while almost no is forecast downwind. Considering PROGN 7KM 4 Conclusions simulation (top right panel), forecast is The more performance spread out across non-hydrostatic mountain range, limited-area with a realistic model Lokal decrease Modell maxima in forecasting and a compensating heavy increase over complex terrain minima. has been evaluated on some case studies in framework Hydroptimet/INTERREG IIIB project. Results This obtained is evenfor more two trueheavy for rainfall 2.8 kmevents resolution that affected runs (Fig. Norrn 9, medium Italy in November left panel2 for have QI been 2p8KM presented. run run anda medium numberright LM panel runs forhas been PROGN performed 2p8KMon run). The two unrealistic cases at two horizontal maxima resolutions, to south 7 and 2.8mountain km. Initial andwhich boundary reachconditions even higher are values provided due toby interpolating increased peaks, resolution, operational are drastically ECMWF IFS reduced model. to more The 7-km realistic simulation values by with adding same prognostic setting operational in 4, in butmodel. without In mesoscale PRO EXP assimilation 2p8KM (bottom cycle, isleft considered panel Fig. reference 9) it is confirmed run (OPE 7KM). smoor The reference distribution runs for two cases andpro- vide it is worth a good noting forecast correct distribution overall rain structure, but across spatial distribution Apennines ridge ( sourn is mountain badly reproduced peak in across cross section). mountain ranges. Precipitation maxima are increased to unrealistic values and displaced to upwind side Inorder Alps, towhile furr an highlight artificial drying positive effect impact is produced on prognostic lee side. treatment and also to allow a comparison The increase with observations, horizontal resolution distribution up to 2.8 km maintains over general two shape boxes 1 degree per 1 degree fields, but is considered. induces a crumbling The boxes are respectively high upwind (Fig. regions, with blue associated one) and sparse downwind and more ( intense red one) peaks. The Alpine problems ridge related in to region incorrect where distribution observed was across most intense. mountain ranges are worsened by higher resolution. The impact on QPF inclusion new prognostic equations In Fig. for 11 cloud distributions ice, rain and snow have been presented in upwind only for (left panel) first case and study. in downwind The inclusion (right panel) cloud boxes as are prognostic shown. The variable observations (with diagnostic (blue histogram) rain and snow) are compared shows a with general increase forecast values, obtained by mostly QI 7KM over run (green Alpine histogram) area, leading and to by a furr PROGN increase 7KM unrealistic run (red histogram). high values In forecast upwind. box, This use is true for prognostic both 7 and 2.8 km removes simulations. presence When, in addition forecast to values cloud exceeding ice, also rain

9 M. M. Elementi et al.: et al.: 9 9 M. Elementi et al.: High resolution simulations heavy 61 Fig. Fig Percentage distribution s in in upwind upwind (left (left panel) panel) and and in downwind (right (right panel) panel) areas areas for for QI 7KM QI 7KM run run (green (green histogram), for for PROGN PROGN 7KM 7KM run run (red (red histogram) and and for for observed (blue (blue histogram). Fig. 11. Percentage distribution s in upwind (left panel) and in downwind (right panel) areas for QI 7KM run (green histogram), for PROGN 7KM run (red histogram) and for observed (blue histogram). Fig. 12. Percentage distribution s in upwind (left panel) and downwind (right panel) areas for QI 2p8KM run Fig. Fig Percentage distribution s in in upwind upwind (left (left panel) panel) and and downwind (right (right panel) panel) areas areas for for QI 2p8KM QI 2p8KM run run (green histogram), for PROGN 2p8KM run (red histogram) and for observed (blue histogram). (green (green histogram), for for PROGN PROGN 2p8KM 2p8KM run run (red (red histogram) and and for for observed observed (blue (blue histogram). and snow are treated as prognostic variables, re is a remarkable which which are improvement are not not observed, QPF, and with and tends a reduction tends to to reduce reduce pre- tions PRO PRO are EXP under EXP 2p8KM testing, 2p8KM did focussing did not not produce onproduce optimal a dramatic a use impact LMimpact as with with cloud resolving scales is still a challenging task. Several op- mm, mm, values values cipitation in in inrange - upwind side mm. mm. The The mountain whole whole distribution range and an regards respect respect intense to to PROGN 2p8KM events 2p8KM inrun run Alpine region. Alpine Alpine The area, area, extentionforecast intense by by PROGN rainfall 7KM area 7KM torun run is downwind more is more similar side. similar to tothe observed unrealistic one one maxima with with arespect decreased to to QI towards QI 7KM 7KM observed one. one. values. use correspondent QPF as anhistogram input integrated is not is not shown. shown. hydro-meteorological modelling chain is also under evaluation and results will be This positive effect permains also when Tiedtke presented in a forthcoming paper. convection scheme is switched f in 2.8 km run. The 4 4Conclusions In explicit In downwind representation box box (Fig. (Fig. 11, convective 11, right right panel), processes panel), PROGN helps in Acknowledgements. The work presented here was partially funded forecast forecast reducing sparse and distribution localised fits fits well well maxima, observed leading one. one. by The Theperformance INTERREG IIIB/MEDOCC non hydrostatic Hydroptimet Project. limited area The In particular, In while while QI QI 7KM 7KMrunrun forecasts authors thank MeteoSwiss and or Italian Regions for to a smoor field. model model Lokal Lokal Modell Modell in forecasting heavy heavy over observed data provided. over values values are These are mainly mainly positive concentrated results areinhighlighted first first bin bybin (- considering (- mm), mm), ina in complex terrain terrain has has been been evaluated on on some some case case studies studies PROGN longitudinal 7KM 7KM cross-section run run re re is aissignificant a forecast number rainfall number crossing points points Edited in in by: framework R. Romero Hydroptimet/INTERREG IIIB IIIB where where Western forecast forecast Povalues Valley values are basin are bigger and bigger than than surrounding mm. mm. mountain Reviewed project. project. by: Results V. Results Homar obtained and anor forreferee two two heavy heavy rainfall rainfall events events Considering ridges. The positive kmimpact resolution runs prognostic runs (Fig. (Fig. 12), treatment 12), occurrence belonging to to first first two two ranges ranges presented. maxima, especially at both 2.8 km resolution. oc- that that affected affected Norrn Italy Italy in November in 2 2 have have been been is clearly visible in reduction upwind (- (- and and - - mm) mm) in in upwind upwind area area (left (left panel) panel) is underestimated by by both both PROGN 2p8KM 2p8KMrunrun and and QI QI 2p8KM 2p8KM A number A number LM LMruns runs has has been been performed on on two two is un- The new model features, and explicit representation precipitating convective processes when moving to 2.8 km run. run. The The very very high high values values (over (over mm) mm) forecast forecast by by cases cases at two at two horizontal resolutions, 7 and 7 and km. km. Initial resolution, give a substantial help in improving distribution run and and not forecast not observed, are are completely over mountain removed ranges. by by and and boundary conditions are are provided by by interpolating Initial QI QI run PROGN 2p8KM 2p8KM run. run. Even Even if if PROGN PROGN 2p8KM 2p8KM run run still still operational ECMWF IFS IFS model. model. The The 7 km 7 kmsimulation These heavy events are still under investiga- overestimates tion to evaluate impact inor model - - changes mm mm range, on range, with with same same setting setting operational in 4, in 4, but but without without re re QPF is aisgeneral produced a general improvement by LM. The improvement whole whole distribution. precipita- downwind forecast when box box (right increasing (right panel), panel), horizontal behaviour resolution is very is very up sim- to sim- run run (OPE (OPE 7KM). 7KM). The The reference runs runs for for two two cases cases mesoscale assimilation cycle, cycle, is considered is reference In Intion ilar ilar to that to that 7 km 7 km runs, runs, distribution forecast forecast by by PROGN PROGN 2p8KM 2p8KM run run being being rar rar simi- simitiolalar to to observed one. one. Since Since explicit explicit convection in in provide provide a good a good forecast forecast overall overall rain rain structure, but but spatial spatial distribution is badly is badly reproduced across across mountain ranges. ranges. Precipitation maxima maxima are are

10 62 M. Elementi et al.: High resolution simulations heavy References Baldauf, M. and Schulz, J. P.: Prognostic in Lokal- Modell (LM) DWD, COSMO Newsletter, No. 4, , 4. Bougeault, P., Houze, R. A., Rotunno, R., and Volkert, H. (Guest Eds.): MAP (Mesoscale Alpine Programme), Quart. J. Royal Meteorol. Soc., 588, Vol. 129, Part B (MAP), , 3. COSMO Newsletter No. 4, p., available at cscs.ch/, 4. Doms, G.: The LM cloud ice scheme. COSMO Newsletter, No. 2, p , available at 2. Gassmann, A.: 3D-transport. COSMO Newsletter, No. 2, p , available at 2. Schraff, C. and Hess, R.: A description nonhydrostatic regional model LM. Part III: Data Assimilation, available at http: // 3. Tiedtke, M.: A comprehensive mass flux scheme for cumulus parametrization in large-scale models, Monthly Wear Review, 117, 8, , 1989.