Dennis P. Lettenmaier Department of Civil and Environmental Engineering University of Washington

Dennis P. Lettenmaier

Department of Civil and Environmental EngineeringUniversity of Washington

GEOSS Workshop XXXIII: Using Earth

Observations for Water Management

San Francisco

December 18, 2009

Advances in seasonal hydrologic prediction

Talk Outline1) Background

2) The University of Washington west-wide seasonal hydrologic forecast system

3) Current and recent research -- assimilation of satellite data

4) Is there hydrologically useful skill in climate forecasts?

5) Concluding thoughts

1. Background: The importance of Seasonal Hydrologic Forecasting

water management hydropower

irrigationflood controlwater supply

fisheriesrecreationnavigation

water quality

Aug Dec Apr

Res

ervo

ir S

tora

ge

Aug

Sn

ow w

ater

con

ten

t on

Ap

ril 1

April to August runoff

McLean, D.A., 1948 Western Snow Conf.

SNOTEL Network

Application of statistical methods to seasonal hydrologic prediction in the western U.S.

PNW

Overview: ESP Hydrologic prediction strategy

ESP data flow

The ESP “spider web”

2. The University of Washington west-wide seasonal hydrologic forecast system

6-month ESP streamflow forecasts for western U.S. and Mexico effective 12/7/09

UW Seasonal HydrologicForecast System Website

Forecast System Initial State information

Soil MoistureSimulated Initial

Condition

SnowpackSimulated Initial Condition

Observed SWE

Streamflow Forecast Details

Flow location maps give access to monthly hydrograph plots, and also to raw forecast data.

Clicking the stream flow forecast map also accesses current basin-averaged conditions

Streamflow Forecast Results: Westwide at a Glance

3a: Current and recent research: Snow data assimilation

NCDC met. station obs. up to 2-4 months from current

local scale weather inputs

Initial Conditions: soil moisture,snowpack

Hydrologic model spin up

MODIS Update

Ensemble Forecast:streamflow, soil moisture, snowpack, runoff

25th Day of Month 01-2 years back

LDAS/other real-time met.

forcings for remaining

spin-up

Hydrologic simulation

End of Month 6 - 12

MODIS updating of snow covered area

Snowcover before MODIS update Snowcover after MODIS update

Change in Snowcover as a Result of MODIS Update for April 1, 2004 Forecast

Unadjusted vs adjusted forecast errors, 2001-2003, for reservoir inflow volumes (left plot) and

reservoir storage (right)

4: Is there hydrologically relevant skill in climate forcings

Wood et al 2005: Retrospective Assessment: Results using GSM

General finding is that NCEP GSM climate forecasts do not add to skill of ESP forecasts, except…

April GSM forecast with respect to climatology (left) and to ESP (right)

Wood et al 2005: Retrospective results for ENSO years

October GSM forecast w.r.t ESP: unconditional (left) and strong-ENSO (right)

Summary: During strong ENSO events, for some river basins (California, Pacific Northwest) runoff forecasts improved with strong-ENSO composite; but Colorado River, upper Rio Grande River basin RO forecasts worsened.

Reverse ESP vs ESP – typical results for the western U.S.

Columbia R. Basin

Rio Grande R. Basin

ICs more impt

fcst more impt

DEMETER forecast evaluation

• VIC model long-term (1960-99) simulations at ½ degree spatial resolution assumed to be truth

• DEMETER reforecasts with ECMWF seasonal forecast model for 6 month lead, forecasts made on Feb 1, May 1, Aug 1, Nov 1 1960-99

• 9 forecast ensembles on each date

• Forecast forcings (precipitation and temperature) downscaled and bias corrected using Wood et al approach (also incorporated in UW West-wide system)

• On each forecast date, 9 ensemble members also resampled at random from 1960-99 to form ESP ensemble

• Forecast skill evaluated using Cp for unrouted runoff

Test sites

Missouri River at Fort Benton

Snake River at Milne

Concluding thoughts

• Hydrologic prediction skill at S/I lead times comes mostly from initial conditions.

• Hence more focus on data assimilation, and its implications for hydrologic forecast skill, needs more attention.

• The role of model error in hydrologic predictions needs more focus – how do we best weight land models in multimodel ensemble?

• Do hydrologists (and the land data assimilation community) need to expend more effort on hydrologic forecasting?

Streamflow forecast skill, observed streamflow simulated (left panel) and forecasted (right two) using model soil moisture and SWE; MAMJ streamflow conditioned on January 1 model conditions

7) Multimodel approaches

UW Multi-model monitor

•Same approach as VIC-based SWM

•Models include VIC, Noah, CLM, Sac

The challenge: Different land schemes have different soil moisture dynamics

Model simulated soil moisture at cell(40.25N, 112.25W)

Areas for spatially averaged soil moisture percentiles

Box sizes are 5 x 5 degrees

NW

NENE

SWSWSE

NW

SE

Soil Moisture Percentiles w.r.t. 1920-20032008-07-01

CLM

SAC NOAH

ENSEMBLE

VIC

US Drought Monitor

Summary• West-wide forecast system and SW Monitor are

templates for exploration of new forecasting methods

• Methods perform well in the U.S., where surface obs are relatively abundant.

• However, ongoing work illustrates the potential for using similar methods in areas where in situ obs are sparse, using e.g. remotely sensed precipitation, and/or weather prediction model analysis fields.

• New remote sensing data sources (e.g. SWOT) offer tremendous opportunities for extension of these methods to the underdeveloped world.