Detectability of Streamflow Timing Trends

1
Maurer, E.P.1, I.T. Stewart1, C. Bonfils2, P.D.
Duffy3 and D. Cayan4 1Santa Clara University,
2U.C. Merced, 3Lawrence Livermore National Lab,
4Scripps Institution of Oceanography, UCSD and
WRD, USGS (
Detectability of trends towards earlier
streamflow in the Sierra Nevada
4
Where will streamflow timing change with
different ?T?
2
Obligatory VIC Graphic
Abstract We examine the seasonal timing of flows
on four major rivers in California, and how these
are affected by climate variability and change.
We measure seasonal timing of soil runoff and
river flows by the center timing (CT), defined
as the day when half the annual flow has passed a
given measurement point. We use a
physically-based surface hydrologic model driven
by meteorological input from a global climate
model to quantify the year-to-year variability in
CT resulting from natural internal climate
variability (the internal oscillations of the
climate system). We find that estimated 50-year
trends in CT due to natural internal climate
variability often exceed the trends in CT
observed over the last 50 years. Thus, although
observed trends in CT may be statistically
significant, they are not necessarily a result of
external influences on climate such as increased
greenhouse gases. To estimate when CT changes
might be expected to exceed levels possible from
natural climate variability, we calculate the
sensitivity of CT to increases in temperature
ranging from 1 to 5 degrees. We find that at
elevations between 2000 2800 m are most
sensitive to temperature increases in this range,
and can experience changes in CT exceeding 45
days. As temperatures rise, so do the elevations
that are most sensitive to further increases in
temperature. Based on these sensitivities, we
estimate that changes in CT will exceed those
possible from natural climate variability by the
mid- or late 21st century, depending on rates of
future greenhouse gas emissions.
VIC modeling results shift 1961-1990
temperatures by fixed values, calculate CT

• VIC Model is driven with GCM-simulated
  (bias-corrected, downscaled) P, T and reproduces
  Q for historic period

binned results

• VIC Model Features
• Developed over 15 years
• Energy and water budget closure at each time step
• Multiple vegetation classes in each cell
• Sub-grid elevation band definition (for snow)
• Subgrid infiltration/runoff variability

3
Streamflow Timing at Key Locations PCM control
run results
CT shift for individual VIC grid cells under
specified temperature shifts relative to
1961-1990. Stewart et al. (2005) points (against
basin average elevation) in the SSJ basin are
added as diamonds, and for these diamonds red
indicates significance at the 90 level.
Sacramento-San Joaquin Basin Key points (inflows
to major reservoirs)
1
Preliminary PCM Control Run Analysis
Parallel Climate Model (PCM) Before using
Control Run (constant 1870 atmosphere) How does
it simulate 20th Century in California?
1950-1999 streamflow timing trends at these 4
points (based on VIC modeling)
1950-1976 period for one grid cell 2 m Surface
Air Temperature PCM 20c3m" run1" IPCC AR4
experiment OBS is gridded monthly
observations Sept-Jan good interannual
variability small biases Feb-Aug PCM
underestimates interannual variability low bias
in temperature simulation
No significant trends. Because these sites
include rain dominated area their timing is less
sensitive to historic inter-annual temperature
variability than other areas.
Incremental change in CT for an incremental
change in T. Each bar charts the increase in CT
beyond that already experienced with the next
lowest temperature shift. Whiskers and bar
representing 10, 25, 75 and 90 percentile
elevations within each basin.

• Impacts through 2C focused North of Lake Tahoe
• Maximum impact in 2000-2800 m range
• For up to 2C rise peak impact is in 2000-2400m
  range
• Above that, peak impact shifts to 2400-2800m range

What is the variability in 50-year streamflow
timing trends in California?
629 years of control PCM simulated CT dates for
Feather R.
5
When will these ?Ts and ?CTs happen?
Projected Changes in Temperature Relative to
1961-1990 (from Hayhoe et al. and Cayan et al.)

• Cumulative distribution functions for CT trend
  (days/50 years) for PCM control run.
• Q10 is the value not exceeded in 10 of the trend
  segments.
• Q10 varies from 17-19 days for these sites.
• Based on this control run a 50-year trend in CT
  would need to shift 17-19 days earlier to achieve
  statistical confidence level of 90

Biases are different at different points, with
PCM overestimating interannual variability at
other locations. Overall for all California
variability is close to observed.
This spatially variable GCM bias means raw output
is not useful for hydrology bias correction and
downscaling is needed
Ensemble mean of 11 GCMs
Projected Changes in Timing Relative to 1961-1990
(from Maurer, 2006)
Downscaling GCM Output
P (scale) and T (shift) factor time series
developed Factors interpolated to 1/8 grid cell
centers (about 150 km2 per grid cell)
At GCM scale, CDFs of Precipitation and
Temperature for each month are developed for
Observations and GCM for climatological period.
Quantiles for GCM are mapped onto CDF for
Observations
Applied to entire 629-year control run
But havent past studies shown that streamflow
timing is changing?
Yes. Past study by Stewart et al. found a CT
shift of 17.7 and 20.5 days earlier over the 1948
to 2002 period for two of their three sites that
obtained 90 confidence within the Sacramento-San
Joaquin basin. Those were smaller basins in
snow-dominated areas, not inflows to managed
water system. Can we identify hypsometric
characteristics of basins that will be most
vulnerable to streamflow timing shifts under
warming temperatures?

• ?CTs for these 2 lower elevation basins will
  statistically significant levels my early-to-mid
  21st century under lower emissions, or
  mid-to-late 21st century under higher emissions.
• ?CTs for higher elevation basins will be delayed,
  but could eventually exhibit greater changes than
  lower elevation basins under higher emissions.
• A lower emissions future avoids much of the
  impact on timing for areas above 2400 m

Fig A. Wood
Mean and variance of observed data are reproduced
for climatological period Temperature trends into
future in GCM output are preserved Relative
changes in mean and variance in future period GCM
output are preserved, mapped onto observed
variance
Cayan, D., E. Maurer, M. Dettinger, M. Tyree, K.
Hayhoe, C. Bonfils, P. Duffy, and B. Santer,
2006, Climate scenarios for California,
California Climate Change Center publication no.
CEC-500-2005-203-SF Hayhoe, K., Cayan, D., Field,
C., Frumhoff, P., Maurer, E., Miller, N., Moser,
S., Schneider, S., Cahill, K., Cleland, E., Dale,
L., Drapek, R., Hanemann, R.M., Kalkstein, L.,
Lenihan, J., Lunch, C., Neilson, R., Sheridan,
S., and Verville, J. 2004, Emissions pathways,
climate change, and impacts on California,
Proceedings of the National Academy of Sciences
(PNAS) 101 (34), 1242212427. Maurer, E.P.,
2006, Uncertainty in hydrologic impacts of
climate change in the Sierra Nevada, California
under two emissions scenarios, Climatic Change
(in press) Stewart, I. T., D. R. Cayan, and M. D.
Dettinger, 2004, Changes in snowmelt runoff
timing in western North America under a 'business
as usual' climate change scenario, Climatic
Change, 62, 217-232