Transcript PPS

Comparing GEM Regional,
GEM-LAM 2.5 and RUC Model
Simulations of Mesoscale Features
over Southern Ontario
David Sills, Norbert Driedger and Emma Hung
Cloud Physics and Severe Weather Research Section,
Environment Canada, Toronto, Canada
2009 CMOS Congress
31 May – 4 June, Halifax, NS
Introduction and Motivation
• Variety of NWP models used at the OSPC RSD for
mesoscale analysis and nowcasting guidance:
• REG - regional version of EC’s Global Environmental
Multiscale (GEM) model with 15 km horizontal grid spacing,
• LAM - limited-area version of the GEM model with 2.5 km
horizontal grid spacing, and
• RUC - the US Rapid Update Cycle (RUC) model with
13 km horizontal grid spacing
• LAM’s higher resolution should provide more
accurate solutions in regions of complex topography
• RUC’s 1-hr data assimilation cycle should effectively
‘nudge’ the model solution closer to reality
Methodology
• Focused on features with mesoscale detail over
southern Ontario and surrounding areas:
• Early-season, late-season and summer lake breezes
• Winter land breezes with snow squalls
• Warm / cold fronts
• Low positions
• Others: prefrontal convergence, trofs, lake funnelling, etc.
• used 18 UTC data from June 2008 to May 2009
• Model performance ranked (1st, 2nd, 3rd) based on
subjective comparison of mesoscale features against
observations (sfc winds, radar reflectivity, vis sat)
• Ties were always ranked as 2 (1-2-2, 2-2-3, 2-2-2)
Results - Overall
• 232 mesoscale features were compared from 217 days
• Overall averaged rankings:
LAM – 1.78
RUC – 1.94
REG – 2.19
• LAM ranked higher than REG: 115 events or 49.5%
• RUC ranked higher than REG: 103 events or 44.4%
1s
2s
3s
1s + 2s
LAM
74
136
22
210
RUC
74
97
61
171
REG
11
167
54
178
Results – By Month
• Model rankings have
clear monthly differences
• LAM model superior
Aug-Oct and Feb-Mar,
worse than REG Nov and
Jan
• RUC model superior
Nov-Jan and Apr-May,
worse than REG Oct and
Mar
• No month has REG the
highest ranked model
Results – By Feature Type
(N=32)
• Model ranking also has
clear differences based on
feature type
(N=21)
(N=23)
• LAM superior with early
and late season lake
breezes, worse than REG
for winter land breezes
• RUC superior with low
(N=25)
(N=105)
positions, worse than
REG for early- and lateseason lake breezes
• REG does well with
(N=19)
winter land breezes
Results – By Convection
• Is there a difference for summer convective
environments?
• Very little change for convection vs. no convection
• Model rankings consistent as well
LAM
RUC
REG
Summer – All
Events
Summer Convection
Summer – No
Convection
1.71
1.84
2.23
1.70
1.86
2.20
1.79
1.82
2.32
Case Study – Low on 6 Apr 09
Case Study – Low on 6 Apr 09
Case Study – Low on 6 Apr 09
Late Season Lake Breezes - 15 Oct 09
Late Season Lake Breezes - 15 Oct 09
Late Season Lake Breezes - 15 Oct 09
Winter Land Breezes - 16 Jan 09
Winter Land Breezes - 16 Jan 09
Winter Land Breezes - 16 Jan 09
Conclusions
• Overall, the mesoscale features
generated by the LAM and the RUC
were closer to observations than the
REG, with LAM having the highest
averaged ranking
• There were clear monthly differences
in model rankings, as well in differences
due to feature type
Conclusions Cont’d
• The LAM and RUC ranked about the
same for summer lake breezes and
warm/cold fronts
•The LAM ranked first for early- and lateseason lake breezes, while RUC ranked
first for low positions
• LAM ranked last for winter land breezes,
while RUC ranked last for early- and lateseason lake breezes
Conclusions Cont’d
• There appeared to be little difference
between events with convection and
events without convection
• This is a preliminary investigation – a
more objective approach and larger
sample sizes are needed
• Would a high-resolution LAM with an
hourly data assimilation cycle produce
even better results?
Thank you!