andy wood, alan hamlet and dennis p. lettenmaier university of washington

1
Andy Wood, Alan Hamlet and Dennis P. Lettenmaier University of Washington A west-wide seasonal to interannual hydrologic forecast system We have implemented the Variable Infiltration Capacity (VIC) macroscale hydrology model over the western U.S. at 1/8 degree spatial resolution for experimental ensemble hydrologic prediction at lead times up to six months. We have implemented the Variable Infiltration Capacity (VIC) macroscale Climate forecast ensembles are presently taken from the NCEP Global Spectral Model (GSM) and the NASA NSIPP model, and will eventually be expanded to incorporate more models in a multi-model ensemble. As a benchmark, we also use the VIC model to produce parallel forecasts via the well-known Extended Streamflow Prediction (ESP) method. The ESP forecasts are further composited to provide ENSO and PDO conditioned ensembles, which past work has shown can considerably reduce seasonal forecast error variance. Retrospective skill assessment for forecasts of basin averages of hydrologic and climate variables GSM wrt CLIM GSM wrt ESP Strong ENSO Composite: GSM wrt ESP Shown below are the skill scores for GSM-based forecasts over 1979-1999, relative to two forecast baselines (CLIM and ESP), for all years (top 2 sets) and for a strong ENSO composite (abs(Nino 3.4) > 1). NOTE: Skill Score = 1 - RMSE(GSM) / RMSE(baseline) where baseline is either: CLIM = unconditional climatology ESP = ESP-derived forecast References Wood, A.W., E.P. Maurer, A. Kumar and D.P. Lettenmaier, 2002. Long Range Experimental Hydrologic Forecasting for the Eastern U.S., J. Geophys. Res., 107(D20). Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges, A Simple hydrologically Based Model of Land Surface Water and Energy Fluxes for GSMs, J. Geophys. Res., 99(D7), 14,415-14,428, 1994. VIC Hydrologic Model (Liang et al., 1994) Models *numbered locations were used for retrospective streamflow forecasting analysis (results not shown) 6-Month Ensemble Forecasts of System Storage for the Columbia River Basin Using VIC Streamflow Forecasts and the ColSim Reservoir Model Initialized by Observed Reservoir Elevations (~ Feb 1, 2001) Hydrologic Forecasting Simulations Forecast Products streamflow soil moisture runoff snowpack derived products model spin-up forecast ensemble(s) climate forecast informati on climatology ensemble 1-2 years back start of month 0 end of mon 6-12 NCDC met. station obs. up to 2-4 months from current LDAS/ other real- time met. forcings for remainin g spin- up data source s snow state informati on Step 2: Downscaling climate model spatial scale VIC (1/8-1/4 deg) simple inverse distance interpolation of precip & temp anomalies climate model temporal scale (monthly) VIC (daily) conditional resampling of historic record imposition of daily signal for precip & temp (same month) rescaling of precip & shift of Tmin, Tmax according to forecast anomaly Method described in Wood et al. (2002) Downscaling Climate Model Output At the climate model scale (1.8-2.5 deg), use a quantile to quantile mapping from climate model climatology to observed historical climatology, for precipitation & temperature separately, e.g., Step 1: Statistical Bias Correction Our initial model domain is the Pacific Northwest. Initial testing in real-time began with bi-monthly updates starting at the end of December, 2002, and ran through April 2003. Upgrades to the modeling system during the test period included: a) the development of a simple method for assimilating snow water equivalent observations at the start of the forecast, and b) a modification of the surface forcing estimation immediately prior to the forecast start using a set of real-time index stations in lieu of the Land Data Assimilation System (LDAS) real-time forcings. We also describe the development of a set of reservoir system models for the western U.S., and their implementation within the system to produce ensemble forecasts of reservoir system storages, operations and releases. NCEP GSM forecasts • T62 (~1.9 degree) resolution • 6 month forecast duration • each month, ensemble product has: 20 forecast members 210 rolling climatology members (derived from 10 initial atmos. condition perturbations for each year of a 21 year climatology period) • we use monthly total precip & average temperature NSIPP forecasts 2 x 2.5 degree (lat x lon) resolution 7 month forecast duration 9-member forecast ensembles fixed 50 year climatology based on 9 continous AMIP runs A retrospective forecast skill analysis for the NCEP seasonal forecasts over the entire western U.S. domain was undertaken to ascertain the value of the climate model forecasts, relative to the ESP forecast and climatological forecasts baselines. In general, the GSM retrospective forecasts did not improve upon the skill of the ESP streamflow forecasts; however, in years when strong ENSO anomalies were present in the forecast initiation month, the GSM-based forecasts yielded skill increases in California and the Columbia River basin, but lower forecast skill (relative to ESP) in the Colorado and the upper Rio Grande River basins. Real-time Hydrologic Forecasting for Columbia River Basin in Winter 2003 Dam Power Plant River/Canal Transfer Eastman, Hensley, & Millerton New Don Pedro & McClure Del ta New Hogan Pardee & Camanche Stanislaus River Tuolumne & Merced Rivers Delta Outflow Mokelumne River Calaveras River S a n J o a q u i n R i v e r New Melones San Luis Trin ity Whiskeyto wn Shast a Oroville (SWP) Folso m Clear Creek American River Feather River Trinity River Sacramento River Dam Power Plant River Transfer Del ta Colorado River San Joaquin River Columbia River Sacramento River computer disk failure halted UW forecasts Feb 1 Jan 15, 2003 Dec 28, 2002 Feb 1, 2003 Mar 1, 2003 Apr 1, 2003 Initial Snow Water Equivalent Simulated System Storage (acre-ft) Simulated System Storage (acre-ft) Apr-Sep Stream flow Forecasts Colum bia R iveratthe D alles,O R 50 60 70 80 90 1-Jan 1-Feb 1-M ar 1-Apr 1-M ay forecastdate percent of no UW NRCS BestEstim ate Apr-Sep Stream flow Forecasts S nake R iver"nearm outh" 50 60 70 80 90 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al UW NRCS BestEstim ate Apr-Sep S tream flow Forecasts Libby ReservoirInflow 50 60 70 80 90 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al UW NRCS BestEstim ate Apr-Sep Stream flow Forecasts Dworshak ReservoirInflow 50 60 70 80 90 100 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al UW NRCS BestEstim ate blue/red are storage boundaries green is ensemble mean thick red is historical average black: init. cond. with normal climate Use of real-time SWE observations (right) (from the 600+ station USDA/NRCS SNOTEL network and several ASP stations in BC, Canada, run by Environment Canada) to adjust snow state at the forecast start date (left) (spin-up met. data improvements method not shown) Streamflow volume forecast comparison with NRCS official forecasts Streamflow hydrograph forecasts (example from February 1) Initial hydrologic condition estimates Forecast Approach upgrades Reservoir system forecasts January Forecast April Forecast July Forecast October Forecast implementing remainder of western U.S. domain with real-time forecasts to recommence in working on alternative spin-up meteorology approaches expanding products to include spatial fields (snow, soil moisture), wider reservoir sys improving web site (http://www.ce.washington.edu/pub/HYDRO/aww/w_fcst/w_fcst.htm) developing a downscaling approach for official forecasts from NCEP and other centers pursuing linkages to NRCS and NWS streamflow forecasting operations groups (April and July forecasts not shown; also, streamflow forecasts not shown) January Forecast October Forecast Ongoing Work

Upload: rufus

Post on 06-Jan-2016

19 views

Category:

Documents


2 download

DESCRIPTION

Mokelumne River. Pardee & Camanche. Delta Outflow. Delta. Calaveras River. Shasta. Trinity. Whiskeytown. Trinity River. New Hogan. Clear Creek. Hydrologic Forecasting Simulations. Stanislaus River. San Joaquin River. San Luis. Oroville (SWP). Feb 1. Feather River. New Melones. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Andy Wood, Alan Hamlet and Dennis P. Lettenmaier University of Washington

Andy Wood, Alan Hamlet and Dennis P. LettenmaierUniversity of Washington

A west-wide seasonal to interannual hydrologic forecast system

We have implemented the Variable Infiltration Capacity (VIC) macroscale hydrology model over the western U.S. at 1/8 degree spatial resolution for experimental ensemble hydrologic prediction at lead times up to six months.

We have implemented the Variable Infiltration Capacity (VIC) macroscale Climate forecast ensembles are presently taken from the NCEP Global Spectral Model (GSM) and the NASA NSIPP model, and will eventually be expanded to incorporate more models in a multi-model ensemble.

As a benchmark, we also use the VIC model to produce parallel forecasts via the well-known Extended Streamflow Prediction (ESP) method. The ESP forecasts are further composited to provide ENSO and PDO conditioned ensembles, which past work has shown can considerably reduce seasonal forecast error variance.

Retrospective skill assessment for forecasts of basin averages of hydrologic and climate variables

GSM wrt CLIM GSM wrt ESPStrong ENSO Composite: GSM wrt ESP

Shown below are the skill scores for GSM-based forecasts over 1979-1999, relative to two forecast baselines (CLIM and ESP), for all years (top 2 sets) and for a strong ENSO composite (abs(Nino 3.4) > 1).

NOTE: Skill Score = 1 - RMSE(GSM) / RMSE(baseline) where baseline is either:

CLIM = unconditional climatologyESP = ESP-derived forecast

ReferencesWood, A.W., E.P. Maurer, A. Kumar and D.P. Lettenmaier, 2002. Long Range Experimental Hydrologic Forecasting for the Eastern U.S., J. Geophys.

Res., 107(D20).Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges, A Simple hydrologically Based Model of Land Surface Water and Energy Fluxes for GSMs, J.

Geophys. Res., 99(D7), 14,415-14,428, 1994.

VIC Hydrologic Model (Liang et al., 1994)

Models

*numbered locations were used forretrospective streamflow forecasting analysis (results not shown)

6-Month Ensemble Forecasts of System Storage for the Columbia River Basin

Using VIC Streamflow Forecasts and the ColSim Reservoir Model Initialized by Observed Reservoir

Elevations (~ Feb 1, 2001)

Hydrologic Forecasting Simulations

Forecast Productsstreamflow soil moisture

runoffsnowpack

derived products

model spin-upforecast ensemble(s)

climate forecast

information

climatology ensemble

1-2 years back start of month 0 end of mon 6-12

NCDC met. station obs. up to

2-4 months from

current

LDAS/other real-time

met. forcings for remaining

spin-up

data sources

snow state information

Step 2: Downscaling

• climate model spatial scale VIC (1/8-1/4 deg) • simple inverse distance interpolation of precip & temp anomalies

• climate model temporal scale (monthly) VIC (daily) • conditional resampling of historic record• imposition of daily signal for precip & temp (same month)• rescaling of precip & shift of Tmin, Tmax according to forecast anomaly

• Method described in Wood et al. (2002)

Downscaling Climate Model Output

At the climate model scale (1.8-2.5 deg), use a quantile to quantile mapping from climate model climatology to observed historical climatology, for precipitation & temperature separately, e.g.,

Step 1: Statistical Bias Correction

Our initial model domain is the Pacific Northwest. Initial testing in real-time began with bi-monthly updates starting at the end of December, 2002, and ran through April 2003.

Upgrades to the modeling system during the test period included: a) the development of a simple method for assimilating snow water equivalent observations at the start of the forecast, and b) a modification of the surface forcing estimation immediately prior to the forecast start using a set of real-time index stations in lieu of the Land Data Assimilation System (LDAS) real-time forcings.

We also describe the development of a set of reservoir system models for the western U.S., and their implementation within the system to produce ensemble forecasts of reservoir system storages, operations and releases.

NCEP GSM forecasts• T62 (~1.9 degree) resolution• 6 month forecast duration• each month, ensemble product has:

• 20 forecast members• 210 rolling climatology

members (derived from 10 initial atmos. condition perturbations for each year of a 21 year climatology period)

• we use monthly total precip & average temperature

NSIPP forecasts• 2 x 2.5 degree (lat x lon)

resolution• 7 month forecast duration• 9-member forecast ensembles• fixed 50 year climatology based

on 9 continous AMIP runs

A retrospective forecast skill analysis for the NCEP seasonal forecasts over the entire western U.S. domain was undertaken to ascertain the value of the climate model forecasts, relative to the ESP forecast and climatological forecasts baselines. In general, the GSM retrospective forecasts did not improve upon the skill of the ESP streamflow forecasts; however, in years when strong ENSO anomalies were present in the forecast initiation month, the GSM-based forecasts yielded skill increases in California and the Columbia River basin, but lower forecast skill (relative to ESP) in the Colorado and the upper Rio Grande River basins.

Real-time Hydrologic Forecasting for Columbia River Basin in Winter 2003

Dam

Power Plant

River/Canal

Transfer

Eastman, Hensley, & Millerton

New Don Pedro & McClure

Delta

New Hogan

Pardee & Camanche

Stanislaus River

Tuolumne & Merced Rivers

Delta Outflow

Mokelumne River

Calaveras River

San

Joaquin

R

iver

New Melones

San Luis

Trinity

Whiskeytown

Shasta

Oroville (SWP)

Folsom

Clear Creek

American River

Feather River

Trinity River

Sac

ram

ento

R

iver

Dam

Power Plant

River

TransferDelta

Colorado River San Joaquin River

Columbia River Sacramento River

computer disk failure halted UW forecasts

Feb 1

Jan 15, 2003Dec 28, 2002 Feb 1, 2003 Mar 1, 2003 Apr 1, 2003

Initial Snow Water Equivalent

Sim

ulat

ed S

yste

m S

tora

ge (

acre

-ft)

Sim

ulat

ed S

yste

m S

tora

ge (

acre

-ft)

Apr-Sep Streamflow Forecasts Columbia River at the Dalles, OR

50

60

70

80

90

1-Jan 1-Feb 1-Mar 1-Apr 1-Mayforecast date

perc

ent

of

norm

al

UW

NRCS

Best Estimate

Apr-Sep Streamflow Forecasts Snake River "near mouth"

50

60

70

80

90

1-Jan 1-Feb 1-Mar 1-Apr 1-Mayforecast date

pe

rce

nt o

f no

rma

l

UW

NRCS

Best Estimate

Apr-Sep Streamflow Forecasts Libby Reservoir Inflow

50

60

70

80

90

1-Jan 1-Feb 1-Mar 1-Apr 1-Mayforecast date

pe

rce

nt o

f no

rma

l

UW

NRCS

Best Estimate

Apr-Sep Streamflow Forecasts Dworshak Reservoir Inflow

50

60

70

80

90

100

1-Jan 1-Feb 1-Mar 1-Apr 1-Mayforecast date

pe

rce

nt o

f no

rma

l

UW

NRCS

Best Estimate

blue/red are storage boundariesgreen is ensemble mean

thick red is historical averageblack: init. cond. with normal climate

Use of real-time SWE observations (right) (from the 600+ station USDA/NRCS SNOTEL network and several ASP stations in BC, Canada, run by Environment Canada) to adjust snow state at the forecast start date (left)

(spin-up met. data improvements method not shown)

Streamflow volume forecast comparison with NRCS official forecasts

Streamflow hydrograph forecasts (example from February 1)

Initial hydrologic condition estimates

Forecast Approach upgrades

Reservoir system forecasts

January Forecast

April Forecast

July Forecast

October Forecast

implementing remainder of western U.S. domain with real-time forecasts to recommence in Sept. working on alternative spin-up meteorology approaches expanding products to include spatial fields (snow, soil moisture), wider reservoir system coverage improving web site (http://www.ce.washington.edu/pub/HYDRO/aww/w_fcst/w_fcst.htm) developing a downscaling approach for official forecasts from NCEP and other centers pursuing linkages to NRCS and NWS streamflow forecasting operations groups

(April and July forecasts not shown; also, streamflow forecasts not shown)

January Forecast

October Forecast

Ongoing Work