Simulating the future climate of the Great Lakes using Regional
Climate Models
Frank Seglenieks
Boundary Waters Issues Unit, MSC
Methods of Projecting Hydrologic Impacts of Climate Change, Muskegon, MI
2012-08-27
Page 2
Summary
• Previous estimates of future levels of the Great Lakes
• The Canadian Regional Climate Model
• RCM downscaling – current climate
• RCM downscaling – future climate
• RCM downscaling – future lake levels
• Initial results using NARCCAP
• Future research
Page 3
All methods use output data from Global Circulation Models (GCMs) based on various future carbon scenarios (A1B, A2, etc.).
Unfortunately the GCMs have poor resolution, most don’t even see the lakes and therefore no lake processes are included.
Previous studies of lake levels
Page 4
Michigan – Huron Lake Levels: Decadal Mean
MAX: 177.50 (Oct 1986)
MIN: 175.58 (Mar 1964)
early transient runs(GFTR2, HCTR2, MOTR2, CCTR2)
Lofgren et al. 2002
early 2XCO2 runs(GISS, GFDL, OSU, CCC1)Mortsch et al. 2000
Previous studies of lake levels
Lake level predictions made using GCM data resulted in estimates for much lower lake levels:
Lake Level(m)
Page 5
October 2, 2009:
Study reports Indiana Dunes National Lakeshore threatened by climate change
“Scientists project that Great Lake levels could fall by as much as several feet by 2090.”
November 28, 2008
Climate Change, Water Sharing Could Damage Great Lakes
“Most climate models predict that the water levels in the Great Lakes will fall …. Lake Huron could drop by as much as 4.5 feet.
Previous studies of lake levelsThis has been repeated often in the media:
Page 6
From IPCC AR4
Previous studies of lake levels
In general GCMs predict higher precipitation, evaporation, and runoff over the Great Lakes.
Page 7
The Canadian Regional Climate Model (CRCM) was adapted to study climate change over the Laurentian Great Lakes as part of the International Upper Great Lakes Study (IUGLS).
Need to study the effect of future climates on lake levels to evaluate regulation plans and adaptive management strategies. Hence data was only analyzed on a monthly basis.
The CRCM runs were performed by the Ouranos group in Montreal.
The CRCM
Page 8
The Canadian Regional Climate Model (CRCM) reproduces the main characteristics of the climate system based on dynamical and thermodynamic equations. The Ouranos version uses CLASS 3.3 and a 1-D lake model.
The outside boundary conditions are supplied by the various future climate results of the GCMs, this process is called dynamical downscaling.
The smaller resolution of RCMs allow processes such as lake effect precipitation and precipitation recycling to be modelled (45 km instead of 200+ km).
Future climate results for one run are also available on a much smaller time step (15 min instead of monthly)
The CRCM
Page 9
The CRCM
• CRCM AMMO grid forced around the outside with GCM data.
Page 10
GCM Scenario EnsembleMember
CRCMRun name
period CRCMRun name
period
CGCM A2 1 aey 1961-2000
afb 2040-2070
A2 2 aez 1961-2000
afc 2040-2070
A2 3 afa 1961-2000
afd 2040-2070
A2 4 aet 1961-2100
A2 5 aev 1961-2100
ECHAM5 A2 1 agx 1961-2000
agz 2040-2070
A2 2 ahi 1961-2000
ahk 2040-2070
ARPEGE A1B 1 agw 1961-2000
ahb 2040-2070
The CRCM
• 3 GCMs used with various members
• 8 total future climate sequences
Page 11
CRCM output was used to determine the components of the net basin supply (NBS): lake precipitation, lake evaporation, and incoming runoff for each lake.
The lake precipitation and lake evaporation components of the water budget were directly calculated by the CRCM.
Runoff into each lake is the most difficult component of the water budget to calculate.
The runoff simulated by the CRCM had to be routed down the river system in order to properly simulate the timing of the runoff input into the lakes.
RCM downscaling – current climate
Page 12
Comparison was made between the CRCM results and data from the Great Lakes Environmental Research Laboratory (GLERL) based in Michigan.
GLERL data is based on runs of their Area Ratio Method (ARM).
Environment Canada is working towards running a coupled atmospheric-land surface model called GEM-Surf (formerly MEC/MESH).
RCM downscaling – current climate
Page 13
0 3 6 9 1 2M o n th
-4 0
0
4 0
8 0
1 2 0
1 6 0
Eva
pora
tion
(mm
)
E v a p ora tio n (la n d )
0 3 6 9 1 2M o n th
-4 0
0
4 0
8 0
1 2 0
1 6 0E
vapo
ratio
n (m
m)
E v a p ora tio n (la k e)
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Prec
ipita
tion
(mm
)
C G C M (aey , aez , a fa , a e t, aev )E C H A M 5 (ag x , ah i)A R P E G E (ag w )G L E R L
P recip ita tion (la n d )
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Prec
ipita
tion
(mm
)
P rec ip ita tio n (la k e)
0 3 6 9 1 2M o n th
-4 0
0
4 0
8 0
1 2 0
1 6 0
Evaporation (mm)
E v a p ora tio n (la n d )
0 3 6 9 1 2M o n th
-4 0
0
4 0
8 0
1 2 0
1 6 0
Evaporation (mm)
E v a p ora tio n (la k e)
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Precipitation (mm)
C G C M (aey , aez , a fa , a e t, aev )
E C H A M 5 (ag x , ah i)
A R P E G E (ag w )G L E R L
P recip ita tion (la n d )
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Precipitation (mm)
P rec ip ita tio n (la k e)
Superior
Fig. 6 Mean seasonal cycle for current climate (1961-1990) precipitation and evaporation components for Lake Superior.
RCM downscaling – current climate
• Good agreement between current climate and reference datasets
• Timing issue in lake evaporation
Page 14
RCM downscaling – current climate
• Runoff had to be routed so that it would better match measured inflow.
• This changed the timing of the runoff, but did not change the volume.
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Run
off
(mm
)
L ak e E rie
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Run
off
(mm
)
L ak e M ich igan /H u ron
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Run
off
(mm
)
C G C M (aey , a ez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (ag w )G L E R L
L ak e S u p erior
Page 15
Fig. 9 Mean current climate (1961-1990) NBS seasonal cycle: left panel – unadjusted simulations; right panel – bias corrected simulations.
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e s id u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
Superior
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
N B S (raw )
Michigan - Huron
Erie
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, aev )E C H A M 5 (ag x , ah i)A R P E G E (ag w )G L E R LE C R esid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
RCM downscaling – current climate
• Timing issue seen in NBS of all lakes.
Fig. 9 Mean current climate (1961-1990) NBS seasonal cycle: left panel – unadjusted simulations; right panel – bias corrected simulations.
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e s id u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
Superior
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
C G C M (aey , a ez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
N B S (raw )
Michigan - Huron
Erie
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, aev )E C H A M 5 (ag x , ah i)A R P E G E (ag w )G L E R LE C R esid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
Fig. 9 Mean current climate (1961-1990) NBS seasonal cycle: left panel – unadjusted simulations; right panel – bias corrected simulations.
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e s id u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (ad ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (raw )
Superior
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
C G C M (aey , a ez , a fa , a e t, a e v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )G L E R LE C R e sid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NBS (m
m)
N B S (raw )
Michigan - Huron
Erie
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
C G C M (aey , a ez , a fa , a e t, aev )E C H A M 5 (ag x , ah i)A R P E G E (ag w )G L E R LE C R esid u a l
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-1 0 0
0
1 0 0
2 0 0
3 0 0
NB
S (m
m)
N B S (ra w )
Page 16
Now look at differences between the future (2041-2070) and current time slice (1961-1990). Both time slices use downscaled CRCM data.
As this is dynamical downscaling, time series to not “line-up”.
RCM downscaling – future climate
0 100 200 300
M onth of current c lim ate s im ula tion
0
40
80
120
160
200
Mon
thly
prec
ipita
tion
(mm
)
0 100 200 300
M onth of fu ture c lim ate sim ulation
0
40
80
120
160
200
Mon
thly
prec
ipita
tion
(mm
)
Page 17
Annual total of differences for each component (positive numbers indicate higher future values).
RCM downscaling – future climate
Component Lake CGCM ECHAM5 ARPEGE
Lake Precipitaton Superior 72.0 70.0 12.9(mm over lake) Michigan/Huron 80.3 50.8 43.6
Erie 75.4 48.1 65.5
Lake Evaporation Superior 117.6 77.1 97.9(mm over lake) Michigan/Huron 147.1 103.2 111.1
Erie 163.9 115.0 117.9
Land Precipitaton Superior 63.1 66.2 25.3(mm over land) Michigan/Huron 76.3 55.3 38.0
Erie 69.4 59.3 51.7
Land Evaporation Superior 53.0 46.6 39.2(mm over land) Michigan/Huron 52.3 54.5 29.4
Erie 46.2 59.0 33.9
Runoff Superior 6.7 19.7 -12.2(mm over land) Michigan/Huron 25.5 2.5 8.2
Erie 30.4 -0.4 15.9
Page 18
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
E v ap ora tio n (la n d )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
E v ap o ratio n (la k e)
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
P rec ip ita tio n (lan d )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C G C M (ae y , a e z , a fa , a e t, a e v )E C H A M 5 (a g x , a h i)A R P E G E (a g w )
P rec ip ita tio n (la k e)
RCM downscaling – future climate
• Lake Superior
• Large differences seen in seasonal patterns.
• Generally wetter in the spring and drier in the fall.
Page 19
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C G C M (ae y , a ez , a fa , ae t, aev )E C H A M 5 (ag x , ah i)A R P E G E (a g w )
L a k e S u p er ior
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0D
iffe
renc
e (m
m)
L a k e M ich iga n /H u ro n
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
L a k e E rie
RCM downscaling – future climate
• More runoff during the winter, reduction in spring melt peak, not much change in summer/fall.
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
Lake Michigan/Huron Lake Erie
Page 20
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
N B S (raw )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0D
iffe
renc
e (m
m)
C G C M (aey , aez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0D
iffe
renc
e (m
m)
N B S (raw )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C G C M (aey , aez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )
N B S (a d ju sted )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
N B S (raw )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C G C M (aey , aez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )
N B S (a d ju sted )
Component Lake CGCM ECHAM5 ARPEGE
raw NBS Superior -29.6 38.9 -106.5(mm over lake) Michigan/Huron -10.2 -36.5 -48.9
Erie -22.1 -63.1 -16.1
RCM downscaling – future climate
Superior Michigan/Huron Erie
Page 21
RCM downscaling – future lake levels
• Annual difference in future lake levels show very little overall change.
• Even the range seen in the different models is not that drastic.
Change in lake levels in metres
Page 22
RCM downscaling – future lake levels
• Very important to look at the monthly cycle of lake level, not just the annual differences.
• Exaggerated season cycle seen for most models on most lakes.
• Extreme monthly lake levels are what cause problems.
0 3 6 9 1 2M o n th
-0 .3
-0 .2
-0 .1
0
0 .1
0 .2
Dif
fere
nce
(m)
L ak e E rie
0 3 6 9 1 2M o n th
-0 .6
-0 .4
-0 .2
0
0 .2
Dif
fere
nce
(m)
L ak e M ich ig an /H u ron
0 3 6 9 1 2M o n th
-0 .3
-0 .2
-0 .1
0
0 .1
0 .2
Dif
fere
nce
(m)
C G C M (ae y , a ez , a fa , a e t, ae v )E C H A M 5 (ag x , ah i)A R P E G E (a g w )
L a k e S u p erior
Page 23
RCM downscaling – future lake levels
• Nowhere near as dramatic a change in lake level as seen in earlier studies using GCM results directly.
Michigan – Huron Lake Levels: Decadal Mean
MAX: 177.50 (Oct 1986)
MIN: 175.58 (Mar 1964)
Page 24
Angel and Kunkel analysis
• Angel and Kunkel analyzed over 500 GCMs.
• Most variation seen between GCMs, than members of the same GCM, then downscaling method
Lake Michigan/Huron
-500
-400
-300
-200
-100
0
100
200
1 2 3 4 5
Size of horizontal grid square in degrees
An
nu
al N
BS
dif
fere
nce
(m
m)
AHPS (Angel and Kunkel)
CRCM (MacKay and Seglenieks)
CHARM (Lofgren)
Page 25
Initial results using NARCCAP
• North American Regional Climate Change Assessment Program (NARCCAP) – 6 RCMs, 4 GCMs.
• Ideally 24 different combinations would be available
• Only 12 were actually attempted
• Only 8 actually had enough data to calculate lake levels
Page 26
Initial results using NARCCAP
• One RCM (RCM3) did not resolve the Great Lakes
• Overall, only 6 RCM/GCM combinations are currently available (only 3 had complete data sets for all months):
• CRCM – CCSM
• CRCM – CGCM3 (same as previous analysis)
• HRM3 – GFDL
• HRM3 – HADCM3
• WFRG – CCSM
• WFRG – CGCM3
Page 27
Initial results using NARCCAP
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Pre
cipi
tati
on (
mm
)
E v a p o ra tio n (la n d )
0 3 6 9 1 2M o n th
-5 0
0
5 0
1 0 0
1 5 0
2 0 0P
reci
pita
tion
(m
m)
E v a p o ra tio n (la k e)
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Prec
ipit
atio
n (m
m)
P rec ip ita tio n (la n d )
0 3 6 9 1 2M o n th
0
4 0
8 0
1 2 0
1 6 0
2 0 0
Pre
cipi
tati
on (
mm
)
C R C M
H R M 3
W FR G
G LER L
P rec ip ita tio n (la k e)
C R C M
H R M 3
W FR G
G LER L
• Showing current climate results for Lake Michigan/Huron
Page 28
Initial results using NARCCAP
• Showing differences for Lake Michigan/Huron
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
E v a p o ra tio n (lan d )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0D
iffe
renc
e (m
m)
E v a p o ra tio n (la k e)
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
P recip ita tio n (la n d )
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C R C M
H R M 3
W FR G
P rec ip ita tio n (la k e)
C R C M
H R M 3
W FR G
Page 29
Initial results using NARCCAP
• Showing difference in runoff for all lakes
C R C M
H R M 3
W FR G
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
L a k e E rie
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0D
iffe
renc
e (m
m)
L a k e E rie
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
L a k e M ich ig an /H u ro n
0 3 6 9 1 2M o n th
-8 0
-4 0
0
4 0
8 0
Dif
fere
nce
(mm
)
C R C M
H R M 3
W FR G
L a k e S u p er io r
Page 30
Future research
• Complete analysis of output from NARCCAP RCMs (ie. look at temperature, ice, etc.) and extend to all lakes as well as Ottawa River
• Use the precipitation and temperature data to run the WATFLOOD hydrological model to look at hourly flows
• Analyze for changes of intensity of the future climate (ie. Precip return periods, drought).
• Run a hydrological model that has a full energy balance (ie. MESH)
Page 31
Questions?