Transcript hhidalgo_4th_symp

Detection and attribution
of Climate Change in
Streamflow Seasonality of
the Western US
Hugo Hidalgo1, Tapash Das1,
Dan Cayan1,2, David Pierce1, Tim Barnett1,
Govindasamy Bala3, Art Mirin3, Andrew
Wood4, Celine Bonfils3, Ben Santer3.
1) Scripps Institution of Oceanography
2) United States Geological Survey
3) Lawrence Livermore National Laboratory
4) University of Washington
Acknowledgments
This research is supported
by grants from the Lawrence
Livermore National
Laboratory and the California
Energy Commission
Warming already
has driven
observable
hydroclimatic
change
Observed: Less spring snowpack
TRENDS (1950-97) in
April 1 snow-water content at
western snow courses
Observed: Less snow/more rain
Knowles et al.,
2006
Mote 2003
-2.2 std devs
LESS as snowfall
+1 std dev
MORE as snowfall
Observed: Earlier
greenup dates
Cayan et al. 2001
Observed:
Earlier
snowmelt runoff
Stewart et al. 2005
Slide courtesy of
M. Dettinger
Optimal detection &
attribution
●
●
Detection of climate change is the
process of identifying if an observed
change is significantly different from
what would be expected from natural
internal climate variability (Hegerl et
al. 2006).
Attribution of anthropogenic climate
change is the process of identifying if
the observed change is: a) consistent
with the type of changes obtained
from climate simulations that include
external anthropogenic forcings and
internal variability and b) inconsistent
with other explanations of climate
change (Hegerl et al. 2006).
Optimal detection &
attribution
●
●
Given a variable for detection,
the basic idea is to reduce the
problem of multiples dimensions
(n) to a univariate or lowdimensional problem (Hegerl et
al. 1996).
In this low-dimensional space, a
detection vector is used to
compare the observations with
the expected climate change
pattern.
Previous D&A studies
●
●
Ocean heat content (Barnet et al. 2001; Levitus
et al. 2001; 2005; Reicher et al. 2002; Wong et
al. 2001),
ocean salinity (Wong et al. 1999; Curry et al.
2003),
●
sea level pressure (Gillet et al. 2003),
●
changes in sea ice extent (Tett et al. 1996),
●
●
●
precipitation (Hegerl et al. 2006; Hoerling et al.
2006),
tropopause height (Santer et al 2003)
temperature (Santer et al. 1996a, 1996b;
Hegerl et al. 1996; 1997; 2000; 2006; 2007;
Barnet et al. 1999; Tett et al. 1999; 2001;
Mitchel and Karoly 2001; Stott et al. 2001; Stott
2003; Zwiers and Zhang 2003; Kalroy et al.
2003; Kalroy and Braganza 2005; Stone et al.
2007).
Modeling
●
●
●
●
●
●
Downscaled to 1/8 degree resolution
using method of constructed
analogues (CA) or bias correction
followed by spatial disaggregation
(BCSD)
Precipitation, tmax and tmin used as
input to the variable infiltration
capacity model (VIC; Liang et al.
1994)
The VIC runoff and baseflow were
routed using a computer program by
Lohmann et al. (1996) to obtain daily
streamflow data for the rivers
The JFM streamflow fractions were
computed
Fingerprint was computed
The d parameters were computed for
all model runs (d parameter: trends
projected into fingerprint).
Defining
characteristics of VIC
are the probabilistic
treatment of sub-grid
soil moisture
capacity distribution,
the parameterization
of baseflow as a
nonlinear recession
from the lower soil
layer, and that the
unsaturated
hydraulic conductivity
at each particular
time step is a
function of the
degree of saturation
of the soil (Sheffield
et al. 2004;
Campbell 1974;
Liang et al. 1994)
Modeled vs. “Observed”
monthly streamflow
SACRAMENTO RIVER
SAN JOAQUIN RIVER
COLORADO RIVER
COLUMBIA RIVER
Figure 2. Scatterplots of monthly modeled versus naturalized streamflow for basins in the western US.
“Observed” trends JFM
streamflow fraction
PCM anthro trends JFM
streamflow fraction
PCM solar volcanic
trends JFM streamflow
fraction
FINGERPRINT: FIRST PRINCIPAL
COMPONENT OF THE AVERAGE
JFM FRACTIONS FROM THE
ANTHRO RUNS
DETECTION PLOT (JFM
streamflow fraction)
ANTHRO
PCM
(BCSD)
CONTROL
CCSM3-FV
(CA)
SOLAR &
VOLCANIC
PCM (CA)
NATURALIZED
VIC
CONTROL
PCM
(BCSD)
ENSO and PDO out
FINGERPRINT: FIRST PRINCIPAL
COMPONENT OF THE AVERAGE
JFM FRACTIONS FROM THE
ANTHRO RUNS
DETECTION PLOT (JFM
streamflow fraction)
NATURALIZED
ANTHRO
PCM
CONTROL
CCSM3-FV
VIC
CONTROL
PCM
Center timing of
streamflow
CT detection
ANTHRO
PCM
(BCSD)
CONTROL
CCSM3-FV
(CA)
SOLAR &
VOLCANIC
PCM (CA)
VIC
CONTROL
PCM
(BCSD)
Columbia Out
Detection plot (JFM
streamflow fraction)
Columbia out
ANTHRO
NATURALIZED
PCM
(BCSD) VIC
CONTROL
CCSM3-FV
(CA)
SOLAR &
VOLCANIC
PCM (CA)
CONTROL
PCM
(BCSD)
MIROC runs
Detection plot (JFM
streamflow fraction)
NATURALIZED
ANTHRO
VIC
PCM
CONTROL
CCSM3-FV
SOLAR &
VOLCANIC
PCM
CONTROL
PCM
ANTHRO
MIROC
MIROC, RUN 8
CALI
COLO
COLU
Summer detection plot
ANTHRO
PCM
CONTROL
CCSM3-FV
SOLAR &
VOLCANIC
PCM
NATURALIZED ANTHRO
MIROC
VIC
CONTROL
PCM
CONCLUSIONS
●
●
●
Detection of climate change was
found for the JFM streamflow
fractions over the second half of
the 20th century.
There was a positive detection
even in the case when the ENSO
and PDO signals were
regressed-out.
Attribution of climate change for
JFM streamflow turned out to be
model dependant: PCM implies
it, but for MIROC the jury is out.
Downscaling methods
• CA
• BCSD
– Daily variability
is conserved
– Tends to
produce
textured
weaker trends
– Daily variability
is not
conserved
– Tends to
produce
smooth
stronger trends
TRENDS IN JFM TAVG ( C/PER DECADE)