Projected Impacts of Climate Change and Implications for River

Flows in the Pacific Northwest

Presented by: Austin Polebitski and Richard Palmer

Department of Civil and Environmental EngineeringUniversity of Washington

November 7, 2006King County

Precipitation

• Anyone notice it raining more lately?

Changing Characteristics of Precipitation

• Outline– Investigate historic trends in precipitation– Explore data from global models to local

situations• Statistical Downscaling• Dynamic Downscaling

– Preliminary Results

Projected Average Monthly Flows –Snoqualmie Falls

0

1000

2000

3000

4000

5000

6000

SeptAugJulJunMayAprMarFebJanDecNovOctMonth

Ave

rage

Flo

w (c

fs)

0

5

10

15

20

25

30

Ave

rage

Pre

cipi

tatio

n (in

ches

)

PrecipitationHistoricECHAM5-2000ECHAM5-2025ECHAM5-2050ECHAM5-2075

Changing Characteristics of Precipitation

• How will precipitation patterns change in a warmed climate?– More Extreme Events?– More Winter and less Summer Precipitation?– More precipitation as Rain rather than Snow?

• How can these questions be answered?– Historic trend evaluation– Use of GCM output

• Dynamical Downscaling• Statistical Downscaling

Historic Trend Evaluation of Daily Precipitation

• Preliminary Results suggest:

– Most significant trend is in November

– Larger percentage of annual rainfall occuring in November, total annual precipitation has remained relatively constant

– Difficult to identify an increase in extreme events to date, models do forecast an increase.

Decadal Trends in November Monthly Precipitation

0

50

100

150

200

250

300

350

400

450

500

Buck Cedar Everett Kent Landburg Palmer Snoq

Station

Ave

rage

Cum

ulat

ive

Prec

ipita

tion

(mm

)

1930194019501960197019801990

Evaluation of Projected Daily Precipitation

• Two methods discussed

– Dynamical Downscaling*

– Statistical Downscaling*

*Downscaling – the process of transforming coarsely-gridded global data to a regionally relevant scale for use in models.

Dynamic Downscaling

• Method under study by Cliff Mass and Eric Salathé of UW-Department of Atmospheric Sciences and the Climate Impact Group

• The method utilizes a mesoscaleforecasting model (MM5) coupled with GCM output to derive future weather and climate patterns

• Can produce output at high resolution and temporal frequency

MM5 Climate Projections –Percent Change in Precipitation

Statistical Downscaling

• Map cdf’s of future climate to historic cdf’sto produce climate impacted datasets (Wood)

• Historical records and site specific information used to downscale

• Precipitation and temperature are the two main outputs from this downscaling method

Percent Change in 20-Year Design Storm from ECHAM5 2000

10%

15%

5%

20%

25%

15%

Average Percent Change in 20 Year Storm

-5.00% 0.00% 5.00% 10.00% 15.00% 20.00% 25.00%

ABERDEEN20NNEBREMERTON

BUCKLEYCEDAR

CULMBACKCUSHMAN_POWER

DARRINGTONELMA

ELWHAEVERETT

FORKSKENT

LANDSBURGLONGMIREMCMILLINMONROE

MUDMTOLYMPIAPALMER

PARADISEPORTANGELES

PORTTOWNSENDQUILCENE

SEATACSFTOLT

SNOQUALMIESTAMPEDE

STARTUP

Percent Change relative to ECHAM5 2000

Impacts Urban Flooding

• ECHAM5 GCM projects increases ‘extreme’precipitation

• More common events (5 year, 10 year, 20 year return events) change more significantly than 100 year event

• To determine impacts to urban systems, output needs to be run through urban models to determine whether code changes might be necessary

Flooding in Natural Systems

• The Distributed-Hydrology-Soil-Vegetation-Model (DHSVM) uses physically based parameters to determine runoff and streamflow for a basin

• Downscaled ECHAM5 output can be used to drive DHSVM

Projected Daily CDF’s of Flow at Snoqualmie Falls

Projected Return Interval Flows for Natural Systems

0

10000

20000

30000

40000

50000

60000

70000

1 2 5 10 20 50 100Return Interval

Flow

(cfs

) HistoricECHAM5-2000ECHAM5-2025ECHAM5-2050ECHAM5-2075

Distribution of Events Exceeding 1-year Return Interval (7000 cfs)

0

100

200

300

400

500

600

700

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Num

ber o

f Eve

nts

Ove

r 700

0 cf

s

Historic2000202520502075

Next Steps

• Complete DHSVM runs for regional basins, i.e. Cedar, Green, Puyallup/White

• Identify key areas of flood inundation interest and possibly perform HEC-RAS models for those areas

• Run climate impacted datasets through Urban-basin models to examine effects of changing design-storms on conveyance