Pacific Northwest Hydrologic and Climate Change Scenarios for the 21st Century

A Brief Introduction to New Products and Overview of Downscaling Approaches

Climate Science in the Public Interest

a webinar sponsored by the

Climate Impacts GroupUniversity of Washington

March 24, 20101:00-2:00pm

Webinar Tips

• Mute phones to reduce background noise (*6 or “mute” button). Do not put call on “hold” if your organization has hold music.

• Type questions in the chat box (lower left). We will track the questions and answer as time permits at the end of the presentation.

• Additional opportunities for Q&A:• Post-webinar survey• Email CIG: cig@uw.edu

• Presentation will be posted on the webinar meeting page on the CIG web page (www.cses.uw.edu/cig)

Purpose

• Who is the Climate Impacts Group?

• Part 1: Introduction to the Columbia Basin Climate Change Scenarios Project products and statistical downscaling methods

• Part 2: A broader look at the role of dynamical regional climate modeling in scenario development at the CIG

• Future workshops, webinar opportunities

Collective expertise includes (but is not limited to):

• Statistical and dynamical downscaling of global climate model projections

• Macro and fine scale hydrologic modeling• Water resources impacts assessment• Terrestrial and aquatic ecosystems impacts

assessment • Adaptation planning and outreach

Objectives • Increase regional resilience to climate variability and change

• Produce science useful to (and used by!) the decision making community

An integrated research team studying the impacts of climate variability and climate change in the PNW and western US

The Climate Impacts GroupThe Climate Impacts Group

Part 1

Responding to Evolving Stakeholder Needs for 21st Century Hydrologic Scenarios:An Overview of the Columbia Basin Climate Change Scenarios Project

Alan F. HamletMarketa McGuire Elsner

CSES Climate Impacts GroupDepartment of Civil & Environmental Engineering

University of Washington

Climate Science in the Public Interest

Climate Change Planning Needs are Evolving

Stakeholder requests:1. Address diverse stakeholder planning needs (terrestrial and aquatic

ecosystems, water management, human health, energy, etc.)2. Provide comprehensive coverage over large geographic areas using

consistent methods3. Increase spatial resolution

-- address both large-scale and small-scale planning efforts in a consistent manner

4. Increase temporal resolution-- address changes at daily timescales and assess changes in hydrologic extremes

5. Quantify uncertainties in future projections

The Columbia Basin Climate Change Scenarios Project

This 2-year research project (finalizing Spring 2010) is designed to provide a comprehensive suite of 21st century hydroclimatological scenarios for the Columbia River basin and coastal drainages in OR and WA

Collaborative Partners:•WA State Dept. of Ecology (via HB 2860)•Bonneville Power Administration•Northwest Power and Conservation Council•Oregon Water Resources Department•BC Ministry of the Environment

297 297 streamflowstreamflow

sitessites

• Provide a wide range of products to address multiple stakeholder needs

• Increase spatial and temporal resolution

• Provide a large ensemble of climate scenarios to assess uncertainties

• Address hydrologic extremes (e.g. Q100 and 7Q10)

Project Goals and Objectives

http://www.hydro.washington.edu/2860/

Draft

Overview of

Methods

Primary Meteorological Stations Used to Create Spatial Dataset* (1915-2006)

*1/16 spatial resolution & daily timestep

Climate Impacts Group 2009, WA Assessment, Ch. 1http://cses.washington.edu/cig/res/ia/waccia.shtml

Climate Change Scenarios

Figure shows change compared with 1970 -1999 average

IPCC AR 4 Emissions Scenarios:

A1B Medium High

B1 Low

Projected change in mean annual T

Projected change in mean annual P

Downscaling Relates the “Large” to the “Small”

~200 km(~125 mi)resolution

~5 km(~3 mi)

resolution

Statistical Downscaling Approaches

• Simple, easy to explain “snapshot” of average future conditions• Incorporates realistic historical daily time series and spatial variability• Provides 91 years of historical variability combined with changes in T and P for

each future time frame and emissions scenario• Projections are easily relate to stakeholder knowledge of historical impacts

Composite Delta Method

Strengths:

Limitations:

• Incorporates only average changes in mean monthly T and P, not extremes• Changes are assumed to be the same throughout the region

Statistical Downscaling Approaches

• Incorporates more information from the monthly GCMs, including altered monthly T and P variability in space and time

• Produces a transient (i.e. continually varying) daily time series for 150 years (1950-2098+)

• Simulates rates of change in hydrologic variables• Offers flexible time period of analysis (one run gives all future time periods)

Bias Corrected and Spatially Downscaled (BCSD)

Strengths:

Limitations:• Quality of downscaled realizations is dependent on GCM performance – What

you get out is only as good as what you put in!• Ensemble analysis needed to account for decadal P variability due to relatively

small sample size in future time slices (~30 years)• Daily time series characteristics are not suitable for some kinds of analyses

(e.g. analysis of hydrologic extremes)

Statistical Downscaling Approaches

• Combines the strengths of the Composite Delta and BCSD methods, while avoiding most of the weaknesses of both

• Incorporates more information from the monthly GCMs, including altered spatial patterns of T and P changes

• Incorporates realistic historical daily time series and spatial variability• Provides 91 years of historical variability combined with changes in T and P for

each future time frame and emissions scenario• Arguably the best method for evaluating hydrologic extremes (floods and low

flows)

Hybrid Delta Method

Strengths:

Limitations:• Quality of downscaled realizations is dependent on GCM performance (spatial

patterns only)• Constrained by the time series behavior in the historic record

Available PNW Scenarios

2020s – mean 2010-2039; 2040s – mean 2030-2059; 2080s – mean 2070-2099

Downscaling Approach A1B Emissions Scenario

B1 Emissions Scenario

Hybrid Delta

hadcm cnrm_cm3 ccsm3 echam5 echo_g cgcm3.1_t47 pcm1 miroc_3.2 ipsl_cm4 hadgem1

2020s 10 10

2040s 10 10

2080s 10 10

Transient BCSD

hadcm cnrm_cm ccsm3 echam5 echo_g cgcm3.1_t47 pcm1

1950-2098+ 7 7

Delta Method

composite of 10

2020s 1 12040s 1 12080s 1 1

Snow Model

Schematic of VIC Hydrologic Model• Sophisticated, fully distributed,

physically based hydrologic model• Widely used globally in climate change

applications• 1/16 Degree Resolution

(~5km x 6km or ~ 3mi x 4mi)

General Model Schematic

Sample Results Using Different Statistical

Downscaling Approaches

http://www.hydro.washington.edu/2860/

HydrologicProducts

Draft

Changing Watershed Classifications:Transformation From Snow to Rain

• Based on Composite Delta Method scenarios (multimodel average change in T & P)

• Historical period includes 1916-2006 water years (Oct-Sep)Map: Rob Norheim

Trend (Days per Decade) * 50mm SWE Plotting Threshold

Linear trend for the ECHAM 5 A1B Scenario downscaled using the

transient BCSD Downscaling Method

Trends in Date of Peak SWE

http://www.hydro.washington.edu/2860/products/sites/?site=6092

Snow and Runoff Summary (Cispus River near Randle)

Scenario Ensembles Ensemble Mean Historical Mean

Draft study results (to be finalized Spring 2010) are already being used and evaluated by a wide range of stakeholders including:•USGS•Bonneville Power Administration•U.S. Bureau of Reclamation•U.S. Army Corps of Engineers•U.S. Forest Service•U.S. Fish and Wildlife Service•Boise Aquatic Research Laboratory•National Marine Fisheries Science Center

Who’s Using the Data?

Next steps…

Extending the approach to additional western U.S. watersheds in partnership with:

•US Forest Service•US Fish and Wildlife Service•Boise Aquatic Sciences Lab•Trout Unlimited

Eric Salathé

JISAO Climate Impacts GroupUniversity of Washington

Ruby Leung & Qian Fu PNNLYongxin Zhang CIG, NCARCliff Mass UW

Part 2

Regional Climate Modeling for Impacts Applications in the US Pacific Northwest

Downscaling and Regional Climate ModelingStatistical Downscaling

•Maps the climate change signal from a global model onto the observed patterns•Computationally efficient•Can tune to observed climate•Preserves uncertainty in Global Climate Models•Cannot represent fine-scale patterns of climate change

Regional Climate Models (“Dynamic Downscaling”)

•Extend the physical modeling of the climate system to finer spatial scales•Computationally demanding•Cannot correct bias in global model•Adds to uncertainty from Global Climate Models

12-50 km or~7-32 mi

Global ClimateModel

6-hourlyMonthly

100-200 km

6 km or~3.7 mi

150-km GCM

High resolution is needed for regional studies

Washington

Oregon

IdahoC

asca

de R

ange

Rocky M

ountains

Snake Plain

Olympics

Global models typically have 100-200 km (62-124 mi.) resolution

•Cannot distinguish Eastern WA from Western WA

•No Cascades

•No land cover differences

150-km GCM

Washington

Oregon

IdahoC

asca

de R

ange

Rocky M

ountains

Snake Plain

Olympics

Regional models typically have 12-50 km (7-32 mi) resolution

• 12 km WRF at UW/CIG

• Can represent major topographic features

• Can simulate small extreme weather systems

• Represent land surface effects at local scales

12-km WRF

High resolution is needed for regional studies, cont’d

Why do we want to simulate the regional climate?

Process studies Topographic effects on temperature and

precipitation Extreme weather Attribution of observed climate change Land-atmosphere interactions

Climate Impacts Applications Streamflow and flood statistics Water supply Ecosystems Human health Air Quality

Regional Climate Modeling at CIG

WRF Model (NOAH LSM) Resolution: 12 to 36 km

(~7- 32 mi)

ECHAM5 forcing

CCSM3 forcing

(A1B and A2 scenarios)

HadRM Resolution: 25 km

(~15 mi)

HadCM3 forcing

Statistical Downscaling CCSM3

Fall difference between 1990s and 2040s

Low spatial detail for climate change signal

°C %

Temperature (°C) Precipitation (% change)

WRF “Dynamic Downscaling” CCSM3

Fall difference between 1990s and 2040s

High spatial detail for climate change signal

%

Temperature (°C) Precipitation (% change)

°C

Regional Effects andApplications

Regional Effects andApplications

Land-Atmosphere Interactions

Snow Cover Change Temperature Change

Change in winter temperature (degrees C)Change in fraction of days with snow cover

Wintertime Change from 1990s to 2050s

Salathé et al. 2008

Land-Atmosphere Interactions

Wintertime Change from 1990s to 2050s

Salathé et al 2008

Solar Radiation

Example: Endangered Species

Pika

The American Pika depends on cold alpine climate

Loss of snow accelerates warming

Andrea J Ray, et al. Rapid-Response Climate Assessment to Support the FWS Status Review of the American Pika. U.S. Fish and Wildlife Service and the NOAA Earth System Research Laboratory

Projected Changes to 2040s in Spring Pacific Northwest Snowpack

Extreme Precipitation

Change from 1970-2000 to 2030-2060 in the percentage of total precipitation occurring when daily precipitation exceeds the 20th century 95th percentile

•Larger increase on windward slopes of Cascades, Columbia basin•Smaller increase or decrease along Cascade crest

Example: Stormwater Management

Rosenberg, E.A. et al, 2010. Climatic Change, in press.

After Bias Correction, WRF results were used to compute flow in Thornton Creek, Seattle, to project storm-water impacts

(David Hartley, Northwest Hydraulic Consultants)

Percent time exceedance of 50% of the peak 2-year flow for the period 1970–2000 using HSPF model with Bias-Corrected WRF simulations

Output Data for Applications Models

WRF:

Output provides the full hourly 3-D meteorological data needed as input for many applications models

•Air quality•Puget Sound circulation

Air Quality Modeling with WRF

WRF Meteorology

Air Quality Air Quality ModelModel

(CMAQ)(CMAQ)

AnthropogenicBiogenic

Emissions

Human Health Risk Assessment

Example: Summertime Ozone Mortality

King

1997-2006

King

2045-2055

Spokane

1997-2006

Spokane

2045-2055

Population 1,758,260 2,629,160 424,636 712,617

O3 20.7 26.5 35.5 41.6

Daily Mortality rate

0.026 0.033 0.058 0.068

Deaths 69 132 37 74

Jackson, JE. et al, 2010. Climatic Change, in press.

Ozone projections from WRF-CMAQ and population projections are used to assess the future risk of mortality due to ozone exposure

Ongoing and Future Directions

Ongoing and Future Directions

Model Validation and Improvement

1. RCMs add significant local gradients to climate projections

2. If the magnitude or placement of gradients is wrong, this can degrade the climate projections

• Model Validation‒ Both “forecast mode” and “climate mode”‒ Compare to station observations‒ Compare to satellite observations

• Model Improvement‒ Modified snow model‒ Optimizing model for accuracy and efficiency

WRF Ensemble Climate Projections

• To understand uncertainties in regional climate prediction

• Follows similar modeling and statistical analysis used in UW Ensemble Prediction System

• Goal: Use a large set of simulations to project probabilities for future changes

• WRF ensemble simulations underway now

– 4 members completed

– 12 km

– 100-year or 30-year duration

climateprediction.net Oxford University

climateprediction.net •Oxford University and Hadley Centre•Global Climate Experiments•Run by volunteers on personal computer

Regional climateprediction.net •Western US project at OCCRI and CIG•Super ensemble of regional simulations•Integrated global and regional modelling system•25 km grid•Monthly-mean and statistical results•Beta version soon•Results over the next year

The North American Regional Climate Change Assessment Program (NARCCAP)

• Exploration of multiple uncertainties in regional model and global climate model regional projections.

• Development of multiple 50-km regional climate scenarios for use in impacts assessments.

• Evaluation of regional model performance over North America.

www.narccap.ucar.edu

50-km Grid

GFDL CGCM3 HADCM3 CCSM

MM5 X X1

RegCM X1** X

CRCM X1** X

HADRM X X1

RSM X1 X

WRF X X1

Red = run completed

SummarySummary

Regional models can simulate unique local responses to climate change and improve on global models and statistical downscaling

High spatial resolution is necessary to provide added value

Many applications require information only available from a regional climate model

Some applications require additional bias correction or downscaling

Errors in regional models can amplify uncertaintyModel validation is essential

Large uncertainties remain due to Interannual variability Global climate projections Regional climate model differences

An ensemble regional modeling approach is required

More to Come

In the coming year, the CIG plans to host:

• Workshop(s) and/or webinars on the Columbia Basin Climate Change Scenarios Project

• A series of thematic webinars on global and regional climate variability and change, including webinars on specific sectors(e.g., water, forests) and planning for climate change

Announcements will be sent via CIG listserve: climateupdate@uw.edu (see CIG home page)

Thank You!

Climate Science in the Public Interest

Additional opportunities for Q&A:

• Post-webinar survey• Email CIG: cig@uw.edu

CIG Website: http://cses.uw.edu/cig/