Short-term water level forecasts for the Laurentian Great Lakes
using coupled atmosphere, land-surface and lake models Vincent
Fortin 1, Murray Mackay 1, Merc Casas-Prat 1, Frank Seglenieks 2,
Sarah Dyck 2, Frdric Dupont 2, Franois Roy 1, and Gregory C. Smith
1 1 Meteorological Research Division, Environment Canada 2
Meteorological Service of Canada, Environment Canada European
Geosciences Union General Assembly 2015 Vienna | Austria | 12 17
April 2015 EGU.eu
Slide 2
2 Outline Environment Canada's weather and environmental
prediction systems Improving weather and lake level forecast skill
through two-way atmosphere-lake coupling Skill of 48-h water level
forecasts over the Great Lakes Extending the lead time to 30 days
Summary and future work
Slide 3
3 Environment Canada's weather prediction systems A suite of
prediction systems based on the GEM atmospheric model GEPS: 16 days
GDPS: 10 days GEPS: 32 days, 60 km, 20 member ensemble global, 1
run/week 60 km, 20 member ensemble, global, 2 runs/day 25 km,
deterministic, global, 2 runs/day REPS: 3 days, 15 km, 20 member
ensemble, North America, 2 runs/day RDPS: 2 days, 10 km,
deterministic, North America, 4 runs/day HRDPS: 2 days, 2.5 km,
deterministic, Canada, 4 runs/day Pan Am: 1 day, 250 m,
deterministic, Greater Toronto Area, 1 run/day 32 days
Operational
Slide 4
4 Gulf of St. Lawrence Prediction System GEM RDPS (10 km)
atmospheric model with ISBA land-surface model NEMO+CICE (5 km)
ocean-ice model GEM GDPS (25 km) atmospheric model See poster
EGU2015-6866 by Pellerin et al.: Evaluation of a new Gulf of St.
Lawrence coupled environmental prediction system based on the GEM
atmospheric model and on the NEMO-CICE ocean-ice model See poster
EGU2015-6866 by Pellerin et al.: Evaluation of a new Gulf of St.
Lawrence coupled environmental prediction system based on the GEM
atmospheric model and on the NEMO-CICE ocean-ice model
Slide 5
5 Great Lakes Prediction System Dupont et al. (2012) Water
Quality Research Journal of Canada WATROUTE routing model GEM RDPS
(10 km) atmospheric model with ISBA land-surface model NEMO+CICE (2
km) ocean-ice model Pan Am demo: 250 m weather forecast for Greater
Toronto Area GEM GDPS (25 km) atmospheric model
Slide 6
6 Great Lakes watershed and domain of Pan Am demo Lake Superior
Lake Michigan Lake Huron Lake Erie Lake Ontario Chicago (Calumet
Harbor) Harbor Beach Buffalo / Niagara Falls Toronto 250 km Toledo
Stannard Rock Lighthouse Marquette
Slide 7
7 Lake Superior supplies June 2004 - Dec 2009 Overlake
evaporation prediction Deacu et al. (2012) Journal of
Hydrometeorology Stannard Rock location, Lake Superior Hourly
latent heat flux at Stannard Rock (6-18h forecasts, dec. 2008)
Observed flux (eddy cov.) Forecasts: Atmos. model (op. config) Lake
model (improved flux param.) Matching fluxes in atmos. & lakes
Observed supplies (change in storage + outflow) Watershed
simulation: Based on operational config. With precipitation
analysis Improved flux param. for lakes Matching fluxes in atmos.
& lakes
Slide 8
8 2 1 0 -2 Impact of coupling on winter 2014 2m temperature
forecasts 32103210 Bias improved by 1C at night, 0.5C during the
day Standard deviation improved by 0.1 C Uncoupled atmospheric
forecastCoupled forecast
Slide 9
Environmental Prediction Applications of the Great Lakes
Prediction System Surface current and temperature forecasts Search
and rescue Particle tracking (drifting boat, oil spill) Water
temperature [C]Surface currents [m/s] Summer of 2008 (system
running daily since the fall of 2014) Water level forecasts Storm
surge and coastal inundations Optimization of hydropower
production
Slide 10
How well can we forecast wind-driven anomalies in water level?
Source: Ohio Department of Natural Resources
http://geosurvey.ohiodnr.gov/ A week in the life of Lake Erie
Slide 11
Forecast issued on Oct 30 th, 2014 for Chicago Forecast issued
On Oct 31 th Obs. Model 0h 24h 48h +80 cm +60 cm +40 cm +20 cm +80
cm +60 cm +40 cm +20 cm Obs. Model Storm surge forecast for
Chicago
Slide 12
Storm surge forecast for Chicago: beyond the first peak
Observed water level EC forecast 00-24h forecast 24-28h forecast US
NOAA forecast 00-24h forecast 24-28h forecast
Slide 13
Seiche forecasts for Buffalo Environment Canada's Great Lakes
Prediction System (NEMO ocean model, 2-km resolution) U.S. National
Weather Service Great Lakes Operational Forecast System (POM ocean
model, 5-km resolution) Observed water level 00-24h forecast 24-28h
forecast Observed water level 00-24h forecast 24-28h forecast
Slide 14
1.9 cm 3.1 cm 3.2 cm 4.9 cm 5.4 cm 8.6 cm 8.2 cm 10.1 cm RMSE
do not significantly increase over time: longer deterministic
forecasts possible? RMSE as a function of lead time: analyzed
period Oct 23rd Dec 7th 2014
Slide 15
15 Towards a Great Lakes Ensemble Prediction System WATROUTE
routing model (1km) GEM REPS (15 km) atmospheric model with
improved land-surface model (SVS) NEMO (2 km) ocean model Monthly
outlook, 20 members GEM GEPS (60 km) atmospheric model See poster
At CGU/AGU spring meeting in Montreal next month! See poster At
CGU/AGU spring meeting in Montreal next month!
Slide 16
16 Summary and Future Work Long-term water level prediction
very sensitive to coupling strategy between atmosphere and lakes
flux coupling required in order to close the water balance but
better fluxes also lead to better weather forecasts Skillful
forecasts of water level anomalies caused by wind stress can now be
obtained two days in advance At longer time scales, tributary flow
becomes more important, hence precipitation and snowmelt will
require an ensemble prediction framework