Log In | Users | Register
Edit | Attach | New | Raw | Diff | Print | | Tools

myCOSMO-NExT Case: 2012 08 01

INTRODUCTION

On August 1st 2012 an MCS affected regions of eastern France and most of Switzerland. The convection initiated as isolated cells along a prefrontal trough/ convergence line in the Loire valley during mid-afternoon. Over the next couple hours, the convection rapidly grew upscale towards the east through outflow boundary merging and interaction with the nearby topography. The resulting MCS took on an asymmetric leading-line / trailing stratiform configuration as it advected further east over Switzerland during the overnight hours. Observations also seemed to show the development of an apparent weak Mesoscale Convective Vortex (MCV) circulation within the MCS after several hours. While this MCS was only of moderate intensity, several numerical models including the COSMO 7 km and COSMO 2 km models had significant trouble in adequately simulating this convective event. It will be interesting to see how the COSMO-1 km model is able to handle this particularly tricky convective case...

ppt presentation @ ECSS 2013 conference : http://www.essl.org/ECSS/2013/programme/presentations/101.pdf

The ability of the COSMO model to correctly simulate this convective case or not seems heavily dependent on its ability to correctly determine the initial conditions in the convective initiation zone, namely the Roanne region of France in the upper Loire Valley. Indeed, the initial cells that develop in this region between 14:00 and 14:30 UTC on satellite/radar imagery eventually grow upscale and translate eastwards over the next several hours to form a well developed MCS over most of Switzerland. The initial conditions which seemed the most crucial to get correct on this day and that COSMO-7 and COSMO-2 apparently had trouble with, were low-level humidity (2m Td, 850hPa theta-e), low-level temperature (2m T, 850hPa theta-e) and as a consequence a realistic assessment of the Lifting Condensation Level (LCL), the Level of Free Convection (LFC) and of the airmass instability (MLCAPE, MUCAPE). Indeed, since part of the upscale growth process of the MCS seems to have been strongly dependant on outflow boundary interactions between the initial storms and the surrounding topography, if the model misses these initial storms it will have a very difficult time in capturing the subsequent evolution of the MCS as it translates eastward...

I will be delving into the COSMO-1 data for this case and will keep you all informed of what I find... smile

Version 1

COSMO-1 forecast: 01.08.2012, 06 UTC run

* valid @ 12 UTC (H+6hr): pre-convective environment

  • 2m temperature : @ 12 UTC, both COSMO-1 and COSMO-2 overestimate this parameter over most of central France (Massif Central region) by about 2 to 4°C. COSMO-1 is slightly better with respect to COSMO-2 around the Vichy and Roanne region close to where the first convective cells initiated based on satellite/radar. At 12 UTC, the surface 2m temperature in the Roanne region was 30°C. The reason for this temperature overestimation by the models is unknown at the moment, but a hypothesis is that it could be related to an exagerated katabatic foehn effet on the northern slopes of the COSMO modeled Massif Central topography.

  • 2m dewpoint temperature : @ 12 UTC, both COSMO-1 and COSMO-2 greatly underestimate this parameter when compared to the synoptic observations over a large portion of the Massif Central region by -2 to -6°C, and locally even by -6 to -8°C close to the genesis region of the subsequent first convective cells (upper Loire valley). The forecast improvement COSMO-1 brings compared to COSMO-2 for this parameter in this case appears non-existant. At 12 UTC, the actual surface 2m dewpoint temperature in the Roanne region was 16°C. The reason for this dewpoint underestimation by the models is unknown at the moment, but a hypothesis is that it could be related to an exagerated katabatic foehn effet on the northern slopes of the COSMO modeled Massif Central topography.

  • Level of Free Convection (LFC) : @ 12 UTC, both COSMO-1 and COSMO-2 significantly overestimate the height of the LFC with regards to a realistic assessment based on modified proximity soundings using 12 UTC Roanne surface (2m) T and Td observations. This is not surprising since both models had a very difficult time in correctly estimating the critical boundary layer RH in the convective genesis region (overestimated 2m T and underestimated 2m Td). While the COSMO-1 values for the LFC at 12 UTC (2800-4000 m amsl) appear slightly less erroneous than the LFC values forecast by COSMO-2 (3000-4200 m amsl), the difference is not really significant nor sufficient for convection to subsequently initiate in the higher resolution model. When compared to the modified proximity soundings with the surface Roanne 2m T and 2m Td observations (LFCs 2200-2500 m amsl), the proximity soundings with the COSMO-2 model surface 2m T and 2m Td values over this region overestimate the LFCs anywhere between 600-1700 m (see proximity soundings below). With such elevated LFCs in the models, it is hence not suprising that they were not able to convect over this region despite the important low-level forcing provided by both the surrounding topography and the pre-frontal convergence zone.

  • Mean-Layer (ML) and Most Unstable (MU) Convective Available Potential Energy (CAPE) : @ 12 UTC, as a direct consequence of over/under estimated COSMO modeled low-level temperature/humidity information respectively, both COSMO-2 and COSMO-1 significantly underestimated the airmass instability in the pre-convective environment region where the first thunderstorm cells initiated near Roanne. COSMO-1 appears to perform slightly better than COSMO-2 in this regard, with COSMO-1 modeled values of MLCAPE between 200-600 J/kg compared to COSMO-2 modeled values of MLCAPE between 50-400 J/kg. The actual mean-layer based airmass instability in this region (MLCAPE) at this time based on modified proximity sounding data appears to have been anywhere between 1000-1300 J/kg. Concerning MUCAPE values @ 12 UTC, COSMO-1 modeled values approach 400-1300 J/kg which is apparently quite an underestimation since in reality even mean-laver based values equal or surpass these values. Regarding convective initiation, these data clearly show the double penalty that results when models poorly estimate low-level temperature and humidity information (in this case overestimating T and underestimating Td)... not only is it more difficult for models to initiate the convection (higher LFC) but if the model does manage to convect, the available modeled CAPE will be much lower than in reality, resulting in weaker modeled convection that will probably have a more difficult time in sustaining itself through the domain or in initiating new convection downstream... a direct result of weaker updrafts (sustainment) and downdrafts (initiating new cells through outflow merging)... Of course, the same double penalty holds true in the opposite direction if low-level temperature is underestimated and the low-level humidity is overestimated (convective initiation too easy and convection too strong/persistent).

  • MSLP field/low-level wind/moisture convergence zones : @ 12 UTC, Both COSMO1 and COSMO2 models are able to quite accurately resolve the pre-frontal convergence zone oriented north-northwest to south-southeast and clearly detectable via the 12 UTC synoptic surface station observations over central France. In terms of accuracy, both the orientation AND location of this pre-frontal convergence zone (wind & pressure) are well resolved. It is therefore quite strange that the modeled surface/low-level convergence zone was not able to pool moisture along it as one would expect with such a feature. Looking at the COSMO1 2m temperature and dewpoint fields, it appears that excessive drying and warming takes place just east of the "Monts d'Auvergne" near Clermont-Ferrand (obs: T33°/Td15°; COSMO1: T35°/Td08°) as well as just south of Roanne (obs: T30°/Td16°; COSMO1: T33°/Td09°) in the upper Loire Valley. Given modeled low-level wind directions in these regions, it appears quite possible/probable that this excessive "drying/warming" is due at least partly to exagerated katabatic foehn effects along the lee sides of the Massif Central topography. Other possible sources of this discreptancy which may be worth investigating would be possible model deficiencies in estimating the surface moisture flux.

  • COSMO convective 3hr QPF sums : via an analysis of the 3hr COSMO forecasted precip sums between 12-15 UTC, it can be quite clearly seen that the various model shortcomings described above led both COSMO-2 and COSMO-1 to completely or nearly complemently miss the convective initiation in the Upper Loire Valley, so critical to the subsequent MCS development and upscale growth. A nuance concerning the model performance is nevertheless warranted if we consider that COSMO-1 began timid convective initiation in approximately the correct region (not far from Roanne) between 14 and 15 UTC and was able to maintain a weak convective precipitation signal for a little over an hour in this region (see model precip fields). This weak signal located in the correct region tends to confirm the fact that both COSMO models handled the position of the pre-frontal convergence line quite nicely but that what was lacking for upscale growth was an adequate representation of the low-level thermodynamic fields (temperature and humidity).

Temp fcst6r COSMO1 France 12UTC 01082012.jpg

Temp fcst6r COSMO2 France 12UTC 01082012.jpg

Temp obs France 12UTC 01082012.jpg

Temp obs2 France 12UTC 01082012.jpg

Td fcst6r COSMO1 France 12UTC 01082012.jpg

Td fcst6r COSMO2 France 12UTC 01082012.jpg

LFC fcst6hr COSMO2 France 12UTC 01082012.jpg

LFC fcst6hr COSMO1 France 12UTC 01082012.jpg

ModProxRAOB COSMO2 Roanne 12UTC 01082012.jpg

ModProxRAOB surfobs Roanne 12UTC 01082012.jpg

LFC fcst6hr COSMO2obs Roanne 12UTC 01082012.jpg

MLCAPE fcst6hr COSMO2 France 12UTC 01082012.jpg

MLCAPE fcst6hr COSMO1 France 12UTC 01082012.jpg

MUCAPE fcst6hr COSMO1 France 12UTC 01082012.jpg

Sfcconv MSLP winds fcst6hr COSMO2 France 12UTC 01082012.jpg

Sfcconv MSLP fcst6hr COSMO1 France 12UTC 01082012.jpg

Sfcconv 10mwinds fcst6hr COSMO1 France 12UTC 01082012.jpg

QPF3hr COSMO2 15UTC 01082012.jpg

QPF3hr COSMO1 15UTC 01082012.jpg

QPF1hr COSMO1 15UTC 01082012.jpg

QPF1hr COSMO1 16UTC 01082012.jpg

* COSMO-1 : @15 UTC (H+9hr) : during convective initiation in upper Loire Valley

  • 2m temperature : @ 15 UTC, similar overall 2m T overestimation (2 to 4°C) by COSMO-1 with respect to the observations over Massif Central and Rhône/Saône river valleys. This overestimation is even locally higher (5 to 8°C) over upper Saône river valley. At this time, in the Roanne region the surface 2m temperature was about 32°C.

  • 2m dewpoint temperature : @ 15 UTC, when compared to observations, COSMO-1 underestimates 2m Td by -3 to -5°C over much of the Massif Central region, locally up to -5 to -7°C close to the genesis region of the first convective cells (Roanne region). At this time, in the Roanne region the surface 2m dewpoint was around 16°C.

  • Level of Free Convection (LFC) : @ 15 UTC, due to continued PBL heating in the afternoon, surface temperatures continued to rise a bit (32°C) in the Roanne region compared to 12 UTC (30°C) but 2m dewpoints remained around 16°C according to regional observations, leading to a slightly higher LFC in this region. While COSMO-2 and COSMO-1 model 2m temperatures at 15 UTC were comparable to the values at 12 UTC in the immediate vicinity of Roanne, the models advected the dry modeled air located over Roanne at 12 UTC further north at 15 UTC, dropping modeled dewpoints north of the city. This resulted in rising COSMO-2 modeled LFC values to between 3900-5100 m amsl in this region and rising COSMO-1 modeled LFC values to between 3900-4700 amsl. These modeled LFC values are clearly too high when compared to LFC values obtained using proximity soundings coupled with observed surface temperature and dewpoint information.

  • Mean-Layer (ML) and Most Unstable (MU) Convective Available Potential Energy (CAPE) : @ 15 UTC, no significant difference exists between the MLCAPE pattern and values of COSMO-2 compared to the ones of COSMO-1. Both models forecast approximately 200-800 J/kg of MLCAPE both to the east and west of the Loire Valley. A similar pattern is evident in the MUCAPE field with values approching 400-1200 J/kg in COSMO-1. Very little MLCAPE/MUCAPE is forecast around the immediate vicinity of Roanne, which appears to be a reflection of the very low 2m dewpoint values forecast by the models in this immediate vicinity. As a result, as was the case for 12 UTC, the modeled CAPE values at 15 UTC are significantly underestimated in this region.

  • MSLP field/low-level wind/moisture convergence zones : @ 15 UTC, COSMO1 continues to rather accurately depict the orientation and location of the pre-frontal convergence zone as apparent on the 15 UTC surface synoptic map of central France and as is clearly visible on HRV Meteosat satellite imagery (see map and satellite images below). At this time, the first convective cells had already initiated as could be observed via satellite and radar imagery. These initial cells were located in proximity to Roanne and its surrounding topography. Further to the northwest along the northern portion of the convergence line, while moisture pooling (cumulus congestus) was apparent on HRV satellite imagery, this shallow convection (turkey towers) had trouble evolving into thunderstorm activity for reasons most likely related to the height of the LFC over relatively flat ground. Along that portion of the convergence line, COSMO1 appears to better resolve the low-level humidity as forecasted 2m dewpoint values coincide rather well with sfc observations. This is apparently much less the case further south just east of Roanne since COSMO1 is not able to simulate the actual convection that initiates in this region at this time, most surely linked to the model's inaccurate estimation of the LFC there (poorly resolved low-level humidity).

  • COSMO convective 3hr QPF sums : via an analysis of the 3hr COSMO forecasted precip sums between 15-18 UTC, it can be seen that COSMO-2 continues to be unable to initiate any type of convective precipitation signal during this time frame, time frame during which the MCS was well underway over the Saône Valley. In this respect, COSMO-1 (as hinted to earlier in the previous QPF discussion) was able to initiate a very weak QPF signal just east of Roanne and was able to sustain it for about an hour and a half before the signal tends to weaken and evaporate as it translates northeastwards. As stated earlier, the presence of this signal in the approximate correct location attests of COSMO's ability to correctly position the pre-frontal convergence zone but its very weak and transcient nature attests of COSMO's inability in this case to correctly simulate the low-level thermodynamic structure of the PBL, so important for a correct assessment of the LFC.

Temp fcst9hr COSMO1 France 15UTC 01082012.jpg

Temp obs France 15UTC 01082012.jpg

Td fcst9hr COSMO1 France 15UTC 01082012.jpg

Temp obs2 France 15UTC 01082012.jpg

LFC fcst9hr COSMO2 France 15UTC 01082012.jpg

LFC fcst9hr COSMO1 France 15UTC 01082012.jpg

MLCAPE fcst9hr COSMO2 France 15UTC 01082012.jpg

MLCAPE fcst9hr COSMO1 France 15UTC 01082012.jpg

MUCAPE fcst9hr COSMO1 France 15UTC 01082012.jpg

Sfcconv 10mwind sfcobs France 15UTC 01082012.jpg

Sfcconv MSLP fcst9hr COSMO1 France 15UTC 01082012.jpg

Sfcconv 10mwind fcst9hr COSMO1 France 15UTC 01082012.jpg

Sfcconv HRVsat MCHrad 14UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1430UTC 01082012.jpg

Sfcconv HRVsat MCHrad 15UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1530UTC 01082012.jpg

Radar 1600UTC.jpg

Radar 1630UTC.jpg

Radar 1700UTC.jpg

Radar 1730UTC.jpg

QPF3hr COSMO2 18UTC 01082012.jpg

QPF3hr COSMO1 18UTC 01082012.jpg

QPF1hr COSMO1 16UTC 01082012.jpg

QPF1hr COSMO1 17UTC 01082012.jpg

QPF1hr COSMO1 18UTC 01082012.jpg

* COSMO-1 : @ 18 UTC (H+12hr) : during MCS upscale growth ( with outflow boundary interactions) and just prior to/during western CH Jura convective initiation

  • 2m temperature : @ 18 UTC, this field across most of eastern France and Switzerland coincides rather well between COSMO-1/COSMO-2 and observations at this time (COSMO-1 between 0 and +2°C). Some overestimation is nevertheless still apparent over parts of the Saône river valley and in Franche-Comté Jura region. In the immediate proximity of La Dôle where the first Jura convective cells initiated, a +1 to +3°C model temperature anomaly is apparent. At this time, the observed surface 2 m temperature in Geneva was 29°C and in Besançon it was 28°C.

  • 2m dewpoint temperature : @ 18 UTC, while COSMO-1 and COSMO-2 seem to have humidified the airmass in proximity to the western Jura region at this time compared to earlier in the day (12 and 15 UTC), they didn't raise the 2m dewpoints sufficiently compared to observations in the immediate vicinity of La Dôle, region where the first Jura convective cells initiated. Both COSMO models performed similarly and were rather accurate in terms of 2m Td over the flatter terrain behind and west of the Jura chain (0 to -2°C anomaly) but both models were much too dry over the Jura peaks (-4 to -6°C anomaly) in the convective genesis region as well as over the Besançon region (-1 to -7°C anomaly) . At this time, the observed surface 2m dewpoint in Geneva was 15°C and in Besançon it was 19°C.

  • Level of Free Convection (LFC) : @ 18 UTC, the forecasted LFC values for the COSMO-2 model over the Jura peaks was around 3600 m amsl and around 4200 m amsl over the flatter Besançon region to the northwest. For the COSMO-1 model, the forecasted LFCs were around 3500 m amsl for both the Jura peak region and the Besançon region. These modeled LFC values are again much too high compared to the LFC values obtained from the modified 12 UTC Payerne proximity sounding using 18 UTC observed 2m temperature and dewpoint information from Geneva and Besançon (Proximity sounding LFCs 2100-2700 m amsl). This data sheds some light to why the models failed to convect shortly after 18 UTC over the western Jura. Firstly, the modeled LFC values were much too high and secondly, the models did not produce the initial convection over the Loire valley responsable for low-level convective outflow that translated eastward towards the Jura and that provided additional low-level lift in this region.

  • Mean-Layer (ML) and Most Unstable (MU) Convective Available Potential Energy (CAPE) : @ 18 UTC, modeled MLCAPE values over the western Jura region were on the order of 400-800 J/kg in COSMO-2 and between 400-1200 J/kg in COSMO-1. Modified Payerne proximity soundings based on T and Td observations from Geneva and Besançon result in MLCAPE on the order of 700-2000 J/kg. The underestimation of the instability was greatest over the Jura peaks and over the Besançon region where the greatest 2m dewpoint anomalies were found between the observations and the model values. Elsewhere in the model domains, the pattern of the modeled MLCAPE/MUCAPE fields appear credible as highest values seemed to be pooled along sfc wind/moisture convergence zones which are rather accurately depicted in the models (especially the pre-frontal convergence zone located over the French Rhône/Saône valleys at this time)... However, while the patterns of the modeled CAPE fields in these regions seem rather well estimated, the values themselves appear underestimated.

  • MSLP field/low-level wind/moisture convergence zones : @ 18 UTC, significant convective upscale growth was well underway as could be observed by visualizing the developing MCS on satellite and radar imagery. The initial thunderstorm cells that had developed along the pre-frontal convergence line in the Upper Loire Valley translated eastwards with it. As these convective cells matured and began to produce convective downdrafts, a distinctive outflow boundary formed ahead of them which was clearly visible on both HRV Meteosat imagery and MCH radar imagery (see sat/rad images below). This outflow boundary was confirmed by the inbound radial velocities measured and observed by our La Dôle radar (see images). By 18 UTC, this convective outflow boundary had already reached the western Jura topography and the first convective cells in this region were already underway. The increased low-level wind/moisture convergence resulting from the outflow boundary interaction with the topography undoubtedly aided the air parcels in this immediate region to reach their LFC. Since neither COSMO2 nor COSMO1 was able to resolve the initial convection, neither was able to produce the increased low-level lift provided by this convective outflow. Moreover, given the fact that both COSMO models significantly underestimated the low-level dewpoints in this region, it is not sure whether the outflow boundary lift would have sufficed for the air parcels to reach their LFC in the drier modeled airmass anyway. As a matter of fact, the 15 UTC COSMO2 run analysis which resolved the initial convection west of the Jura as a result of latent heat nudging was not able to sustain this convection eastwards, hinting that the modeled low-level air was not moist enough compared to reality.

  • COSMO convective 3hr QPF sums : via an analysis of the 3hr COSMO forecasted precip sums between 18-21 UTC, one notices a much larger difference in QPFs between COSMO-2 and COSMO-1 than during the previous 6hrs of this event. While the QPF differences between the 2 models were minimal up to 18 UTC, COSMO-1 was able to simulate a more realistic convective precipitation signal than COSMO-2 between 18-21 UTC, primarily over the elevated topography of the Alpine region. The reason for this better performance is most likely related to COSMO-1's slightly better representation of the LFC compared to COSMO-2 since the low-level thermodynamic variables were a bit better simulated in the higher resolution model. As a result, while COSMO-1 still significantly overestimated the height of the LFC, air parcels forced over the highest peaks of the Alps were nevertheless able to attain it contrary to what occurred in COSMO-2. While COSMO-1's precip signal was more realistic than COSMO-2's signal, it was still significantly underestimated since the overall convective coverage was more scattered and less organized in the model forecast than in reality.

Temp fcst12hr COSMO1 France 18UTC 01082012.jpg

Temp obs France 18UTC 01082012.jpg

Td fcst12hr COSMO1 France 18UTC 01082012.jpg

Temp obs2 France 18UTC 01082012.jpg

Temp obs climap.jpg

Td obs climap.jpg

LFC fcst12hr COSMO2 France 18UTC 01082012.jpg

LFC fcst12hr COSMO1 France 18UTC 01082012.jpg

ModProxRAOB COSMO2 GVA BES 18UTC 01082012.jpg

ModProxRAOB surfobs GVA BES 18UTC 01082012.jpg

LFC fcst6hr COSMO2obs GVA BES 18UTC 01082012.jpg

MLCAPE fcst12hr COSMO2 France 18UTC 01082012.jpg

MLCAPE fcst12hr COSMO1 France 18UTC 01082012.jpg

MUCAPE fcst12hr COSMO1 France 18UTC 01082012.jpg

Sfcconv MSLP winds fcst12hr COSMO2 France 18UTC 01082012.jpg

Sfcconv 10mwind fcst12hr COSMO1 France 18UTC 01082012.jpg

Sfcconv MSLP fcst12hr COSMO1 France 18UTC 01082012.jpg

Sfcconv HRVsat MCHrad 16UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1710UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1730UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1745UTC 01082012.jpg

Sfcconv HRVsat MCHrad 18UTC 01082012.jpg

Sfcconv HRVsat MCHrad 1830UTC 01082012.jpg

Sfcconv HRVsat MCHrad 19UTC 01082012.jpg

Radar 1730UTC.jpg

Radar 1800UTC.jpg

Radar 1830UTC.jpg

Radar 1900UTC.jpg

Radar 1930UTC.jpg

Radar 2000UTC.jpg

Radar 2030UTC.jpg

Radar 2100UTC.jpg

Radar 2130UTC.jpg

Radar 2200UTC.jpg

Radar 2230UTC.jpg

Radar 2300UTC.jpg

Radar 2330UTC.jpg

Radar 0000UTC.jpg

QPF COSMO2 21UTC 01082012.jpg

QPF COSMO1 21UTC 01082012.jpg

QPF1hr COSMO1 19UTC 01082012.jpg

QPF1hr COSMO1 20UTC 01082012.jpg

QPF1hr COSMO1 21UTC 01082012.jpg

QPF1hr COSMO1 22UTC 01082012.jpg

QPF24hr COSMO1 06UTC 02082012.jpg

QPF24hr COSMO2 12UTC 02082012.jpg

QPE24hr radar 12UTC 02.08.2012.jpg

Raingagedata 24hr 01.08.2012.jpg

....

Version 1

COSMO-1 forecast: 01.08.2012, 15 UTC run

Overall quality of forecast : For the most part, compared to the 6 UTC run, the 15 UTC run does not show huge differences in forecasted parameters through the 21 UTC timeframe. As a result, only a few specific model fields will be shown for this run that were deemed significant. Moreover, comparing this 15 UTC run of COSMO-1 with the 15 UTC COSMO-2 run is rather difficult since from what I can gather, the COSMO-1 COSMO-NExT runs do not include radar data assimilation (LHN).

* COSMO-1 : @15-18 UTC (H+ 0-3hr) : during convective initiation in upper Loire Valley and during MCS upscale growth/western CH Jura convective initiation

  • General comments : due to the absence of radar data assimilation (LHN) no convective signal is apparent during the first few hours of the model integration in proximity to the Massif Central, contrary to what was observable in the 15 UTC COSMO-2 run (see model fields below). However, given the fact that the instability fields in this COSMO-1 run at 15 and 18 UTC over central and eastern France appear very similar to the ones in the comparable 15UTC COSMO-2 run, it is not sure whether LHN in this COSMO-1 run would of created longer lasting modeled convection than in the comparable COSMO-2 run. As mentioned previously, the COSMO-2 LHN induced convection that was present in the first few hours of the 15 UTC model integration evaporated entirely before reaching the western Jura chain due to underestimated low-level humidity values over the Massif Central/eastern France and hence undestimated CAPE fields and overestimated LFCs. Another interesting observation is that even without LHN, the 6 UTC run of COSMO-1 was able to create a weak convective precip signal by 16 UTC in the approximately correct location in the Upper Loire Valley whereas the 15 UTC COSMO-1 run was not able to create this same signal very early on in the model integration... which I imagine is a direct result of the spin-up problem.

QPF3hr 15UTCCOSMO2 18UTC 01082012.jpg

QPF3hr 15UTCCOSMO1 18UTC 01082012.jpg

QPF3hr COSMO1 18UTC 01082012.jpg

MUCAPE fcst03hr COSMO2 France 18UTC 01082012.jpg

MUCAPE fcst03hr COSMO1 France 18UTC 01082012.jpg

MUCAPE fcst12hr COSMO1 France 18UTC 01082012.jpg

MUCAPE fcst15hr COSMO1 France 21UTC 01082012.jpg

* COSMO-1 : @18-24 UTC (H+ 3-9hr) : during MCS upscale growth/western CH Jura convective initiation and later stages of MCS lifecycle

  • General comments : during this timeframe, while COSMO-1 forecasted more convective precipitation than during the early stages/developing stage of the MCS, the values forecast remained well below the observed accumulations and generally confined to the Alpine topography. Much of convective activity which concerned the Lake Geneva and the western/central/eastern Plateau area was either not resolved or poorly resolved by the model. The same tendency in the convective precipitation signal was observed for this timeframe with the COSMO-2 and COSMO-1 6 UTC runs. This is again an indication that the model airmasses remained capped (high CIN values) over the flatter terrain while the model LFCs were reached over the highest Alpine topography. Also noteworthy during this time frame is the apparent development and presence of a weak Mesoscale Convective Vortex (MCV) within the MCS which undoubtedly aided to further sustain and structure the convective activity (clearly observable on radar images/animation during this time frame). The latent-heat and Coriolis forced MCVs are still often poorly resolved by limited area models and this, of course, did not aid in an accurate representation of the convective event either.

QPF3hr 15UTCCOSMO1 21UTC 01082012.jpg

QPF3hr 15UTCCOSMO1 24UTC 01082012.jpg

QPF3hr COSMO1 21UTC 01082012.jpg

QPF3hr COSMO1 24UTC 01082012.jpg

QPF COSMO2 21UTC 01082012.jpg

MUCAPE fcst6hr 15UTCCOSMO1 France 21UTC 01082012.jpg

CONCLUDING REMARKS

While it can be observed that both the COSMO-2 and COSMO-1 models had a particularly difficult time simulating this convective event (MCS) over eastern/central France and Switzerland, the COSMO-1 model seems to have done a slightly better job, especially in the later portions of the MCS's lifecycle and over the Alpine topography. While both models successfully resolved the pre-frontal convergence line (location and orientation) over central France, both models did a poor job in accurately depicting the low-level thermodynamics variables (temperature and humidity) in the PBL, overestimating T and greatly underestimating Td (greatly exagerated drying). The origin of this dry anomaly is unknown but an educated guess leads the author to believe exaggerated modeled katabatic warming/drying originating from a southerly wind creating foehn effects on the lee side of the Massif Central may be the cause. It seems that this exaggerated effect in the COSMO models contaminated a large part of the "warm sector" airmass in proximity of the pre-frontal convergence line over central/eastern France leading to large overestimations of the LFC heights over much of the area in question. As a result, both models missed the initial convection over central France and subsequent thunderstorm outflow interactions to the east which were instrumental in the upscale growth of the MCS. Moreover, the models tended to convect only over the highest Alpine topography where the modeled LFCs were reached. While deep moist convection is explicitly resolved in both COSMO-2 and COSMO-1 models, the complex topography within their domains and hence the large uncertainties in their representation of low-level humidity makes them still heavily reliant on real-time radar data assimilation techniques (ex: LHN) in order to improve their simulation of convective events, as this case seems to support. Unfortunately, while LHN was not active in the COSMO-1 simulation, the radar assimilation in the COSMO-2 simulation was of limited use in this case since the modeled low-level troposphere was erroneously too dry, effectively evaporating the LHN convective precip signal within the first couple hours of the model integration. Also noteworthy for this particular case was that the modeled low-level humidity values were so anomalous in the convective genesis region of the model domain that ensemble information proved to be of very limited use. Indeed, Daniel Leuenberger and Luca Weber ran an interesting COSMO-LEPS reanalysis of this event using the COSMO-2 model with modified lateral boundary conditions from the IFS and while several members of the COSMO-LEPS were able to produce transcient and isolated precipitation signals, no coherent precip structures resembling the actual MCS with regards to intensity and/or location were produced by the ensemble. Lastly, the presence of an MCV within the MCS itself during the latter portion this convective event undoubtedly aided in sustaining the system into the early morning hours of the next day as the MCS translated eastwards overnight. Better representation of MCVs in limited area models in the future may also prove useful in more accurately forecasting MCS upscale growth and evolution.

SUGGESTIONS

This specific case stresses the importance of incorporating/ingesting/inserting high quality initial conditions at a high temporal resolution into the model assimilation chain, especially regarding low-level temperature and humidity information. This is of course not always an easy task, particularly amidst the very heterogeneous terrain of Western Europe where low-level humidity can vary very quickly in time and space for various reasons (ex: upslope, katabatic effects, altitude). Unfortunately, obtaining accurate modeled convective parameters (LFC,CAPE,LI) depends heavily on an accurate estimation of this low-level humidity information (2m Td, 850 hPa ThetaE?). For example, a 1 or 2 degree difference in the mean value of the low-level (first 50-100 hPa layer) dewpoint can sometimes increase/decrease MLCAPE values by a factor of 2 or more.

In order to alleviate these important shortcomings, it seems advisable to continue to integrate surface observations as often as possible into "RUC" type nowcasting models (ex: COSMO-2, COSMO-1), more specifically at least once every hour or more. In this respect, it may prove quite useful for the ORGMeteoSwiss APN group to further exploit "observation-based analysis and forecasting techniques" in the context of the development of the COSMO-NExT model philosophy and more fully incorporate this technology into the next version of the intranet Model Browser or into the INCA nowcasting system that is already being utilized/exploited at ORGMeteoSwiss and that is showing promising results. Similar to the obs-based analysis/forecasting tools (Mesoscale Analysis Page) used by forecasters at the Storm Prediction Center (SPC) in Oklahoma City, it is believed that ORGMeteoSwiss could develop a similar beneficial platform for their forecasters by exploiting these nowcasting methods. Using both COSMO-2 and COSMO-1 model data and current SwissMetNet? observational data to calculate rapidly updated analysis and forecast fields at a high spatial and temporal resolution crucial would allow a more accurate calculation of crucial derived/diagnostic fields such as MLCAPE, MLCIN in order to more accurately diagnose pre-convective environments and hence the forecast of convective initiation and MCS evolution.

While the implementation of such systems demand significant resource allocation, the benefits are many. It could also allow the creation of an APN sub-group specifically dedicated to this development and just as importantly, to the needs of the forecasters in general concerning the creation of new model fields. This would allow the forecasters to exploit the improved COSMO model data and our dense observational SwissMetNet? data as efficiently as possible and allow them to provide important feedback to APN. It may also provide an interesting marketing opportunity (monetary if paid service or visibility if free-access) for ORGMeteoSwiss by providing a state-of-the-art forecasting webtool to our external clients/users.

SPC Mesoscale Analysis Page (forecaster/user nowcasting platform) : http://www.spc.noaa.gov/exper/mesoanalysis/new/viewsector.php?sector=19

INCA nowcasting system : http://www.inca-ce.eu/index.php?option=com_content&view=category&layout=blog&id=38&Itemid=130

VERA analysis tool : http://www.univie.ac.at/IMG-Wien/vera/en/

Other possible ways to improve modeled convective initiation/sustainment/upscale growth may be to add satellite and lightning information into the assimilation chain. During the European Conference on Severe Storms (ECSS) in Helsinki (June 2013), several presentations stressed the importance of assimilating both satellite and lightning data in nowcasting models in order to more accurately simulate mesoscale convective systems (MCSs). Regarding lightning data, A WRF simulation of the 29 June 2012 derecho case in the US carried out with lightning data shows that it more accurately simulates the derecho event than if run without lightning data, with regards to both reflectivity structure and wind speed outflow.

Version 2

COSMO-1 forecast: 01.08.2012, 06 UTC run

Overall quality of forecast : Version 2 of this COSMO-1 06 UTC run can be qualified as similar in quality to version 1. While version 2 seems to have better represented the low-level humidity values (somewhat more realistic dewpoints) and better estimated instability fields (more realistic CAPE values), this step in the right direction was not enough to improve the 1hr precipitation forecast field footprint of the evolving MCS.

* COSMO-1 : @15-18 UTC (H+ 9 to 12hr) : during convective initiation in upper Loire Valley and during MCS upscale growth/western CH Jura convective initiation

* Version2_06utcrun_MUCAPE_15utc_COSMO1.png:
Version2_06utcrun_MUCAPE_15utc_COSMO1.png

  • Version2_06utcrun_MUCAPE_18utc_COSMO1.png:
    Version2_06utcrun_MUCAPE_18utc_COSMO1.png

  • Version2_06utcrun_ppn12hr_00utc_COSMO1.png:
    Version2_06utcrun_ppn12hr_00utc_COSMO1.png

  • General comments : Version 2 was not able to sustain the initial cells it creates in approximately the correct location near Roanne. As in version 1, the convective precip signal is rapidly evaporated in this area, hinting that the modeled atmosphere was still too dry there. As for the overall modeled precip signal of the MCS, as in version 1, version 2 picks out the Alpine signal but fails to detect the Swiss Jura and Plateau signal. Version 2 does a slightly better job than version 1 in detecting the northern part of the MCS over the Besançon region of France. This Besançon region is where version 2 improved the representation of humidity and instability significantly and may explain the better performance there. Had LHN been included in the assimilation chain of version 2, this would have perhaps helped to improve the results to some extent...

Version 2

COSMO-1 forecast: 01.08.2012, 15 UTC run

Overall quality of forecast : As for version 1, version 2 of the 15 UTC run is not very useful because of the lack of LHN in the assimilation chain and hence the presence of the spin-up problem (15 UTC is the time of convective initiation near Roanne). Version 2 of this run is even worse than version 1 with a lower instability footprint and weaker precipitation footprint of the MCS than for version 1. As a result, the rating/grade attributed to version 2 for this case is based on the 6 UTC run only. If a version 3 is conducted, it would be better to obtain the 12 UTC run rather than the 15 UTC run.

For a condensed summary of the assessment for this case as well as all other myCOSMO-NExT cases refer to the myCOSMO-NExT Overview page.

Note: This page is both world-readable and world-writeable.

Attach
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:49
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:49
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 13:12
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 11:58
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:59
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 13:15
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:07
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:07
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 18:02
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 18:02
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 15:47
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 15:47
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 17:25
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 17:25
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:58
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 13:14
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 15:58
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 13:14
jpg
version 1 uploaded by LionelPeyraud on 20 Nov 2013 - 15:51
jpg
version 1 uploaded by LionelPeyraud on 20 Nov 2013 - 15:51
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 18:09
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:34
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:25
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 16:01
jpg
version 1 uploaded by LionelPeyraud on 14 Oct 2013 - 18:10
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 15:58
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 15:58
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 15:58
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 15:57
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:08
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:07
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:07
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:44
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:44
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:05
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:06
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:06
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:06
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:07
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:07
jpg
version 1 uploaded by LionelPeyraud on 20 Nov 2013 - 15:44
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:28
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:24
jpg
version 1 uploaded by LionelPeyraud on 20 Nov 2013 - 15:44
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:07
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:41
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:08
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:09
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:06
jpg
version 1 uploaded by LionelPeyraud on 18 Nov 2013 - 16:41
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:09
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:32
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:33
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:34
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:32
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:33
jpg
version 1 uploaded by LionelPeyraud on 03 Oct 2013 - 16:33
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:05
jpg
version 1 uploaded by LionelPeyraud on 22 Nov 2013 - 14:40
jpg
version 1 uploaded by LionelPeyraud on 22 Nov 2013 - 14:41
jpg
version 1 uploaded by LionelPeyraud on 22 Nov 2013 - 14:41
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:51
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:51
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:51
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:52
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:52
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:52
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:53
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 13:53
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:04
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:04
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:04
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:05
jpg
version 1 uploaded by LionelPeyraud on 21 Nov 2013 - 14:07
jpg
version 1 uploaded by LionelPeyraud on 19 Nov 2013 - 14:18
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:51
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:07
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:06
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 13:13
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:08
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:08
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:09
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:09
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:52
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:52
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:53
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:53
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:54
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:54
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:54
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:57
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 13:13
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:08
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 15:48
jpg
version 1 uploaded by LionelPeyraud on 16 Oct 2013 - 13:12
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 15:18
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 14:37
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 14:43
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 15:02
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 15:41
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:28
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 17:34
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 17:49
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:05
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:39
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:39
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:39
jpg
version 1 uploaded by LionelPeyraud on 10 Sep 2013 - 15:44
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 17:51
jpg
version 2 uploaded by LionelPeyraud on 29 Aug 2013 - 18:17 ...
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:07
jpg
version 1 uploaded by LionelPeyraud on 29 Aug 2013 - 18:29
png
version 1 uploaded by LionelPeyraud on 23 Jul 2015 - 14:53
png
version 1 uploaded by LionelPeyraud on 23 Jul 2015 - 14:54
png
version 1 uploaded by LionelPeyraud on 23 Jul 2015 - 14:54
spacer
This site is managed by the Center for Climate Systems Modeling (C2SM).
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors. Ideas, requests, problems? Send feedback!