A multi-model hydrologic ensemble for seasonal streamflow forecasting in the western U.S.

Ted Bohn, Andy Wood, Ali Akanda and Dennis P. Lettenmaier

Department of Civil and Environmental Engineering

for

HEPEX Workshop

Foothills Lab, NCAR, Boulder, Colorado

July 19-22, 2005

OUTLINE

experimental western U.S. forecasting system

expansion to multiple hydrologic model framework

research issues

Experimental W. US Hydrologic Forecast System

Experimental W. US Hydrologic Forecast System

NCDC met. station obs.

up to 2-4 months from

current

local scale (1/8 degree) weather inputs

soil moisturesnowpack

Hydrologic model spin up

SNOTEL

Update

streamflow, soil moisture, snow water equivalent, runoff

25th Day, Month 01-2 years back

LDAS/other real-time

met. forcings for spin-up

gap

Hydrologic forecast simulation

Month 6 - 12

INITIAL STATE

SNOTEL/ MODIS*Update

ensemble forecasts ESP traces (40) CPC-based outlook (13) NCEP CFS ensemble (N) NSIPP-1 ensemble (9)

* experimental, not yet in real-time product

Experimental W. US Hydrologic Forecast System

Soil MoistureInitial

Condition

SnowpackInitial Condition

targeted statisticse.g., runoff volumes

monthly hydrographs

spatial forecast maps

Seasonal Hydrologic Forecast Uncertainty in Western US

Importance of uncertainty in ICs vs. climate vary with lead time …

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

Forecast

Uncertainty actual

perfect data, model

streamflow volume forecast period

model + data uncertainty

low

high

IC error lowclimate fcst error high

IC error highclimate fcst error low

… hence importance of model & data errors also vary with lead time.

Expansion to multiple-model framework

ESP

ENSO/PDO

ENSO

CPC Official Outlooks

Coupled Forecast System

CAS

OCN

SMLR

CCA

CA

NSIPP/GMAO dynamical

model

VIC Hydrology Model

NOAA

NASA

UW

Seasonal Climate Forecast Data Sources

Expansion to multiple-model framework

ESP

ENSO/PDO

ENSO

CPC Official Outlooks

Coupled Forecast System

CAS

OCN

SMLR

CCA

CA

NSIPP/GMAO dynamical

model

Model 2

NOAA

NASA

UW

Multiple Hydrologic Models

Model 1

Model 3

weightings calibrated via retrospective analysis(regression, bayesian, …)

SIMMA: “Standard Interface Multi-Model Array”

...

Forcings(ALMA, NetCDF)

Model 1Wrapper

Model 2Wrapper

Model NWrapper

Post-Processing(ALMA, NetCDF)

•Compare•Combine•Route

Wrapper Script

Set up simulationparameters

Translate forcings(if necessary)

Run model

Translate output(if necessary)

(the intended design)

SACNOAHVIC

PET

SIMMA Features

• Common forcings (ALMA, NetCDF)• Common output format (ALMA, NetCDF)

– Easy to compare results• Wrapper Scripts

– Handle translation– Handle model-specific processing

• Minimal changes to model code– Easy to update a model

• Modular– Easy to add a new model– Re-use code

ALMA Standards - Forcings

Name Units DescriptionSWdown W/m2 Surface incident shortwave radiationLWdown W/m2 Surface incident longwave radiationTair K Near surface air temperatureQair kg/kg Near surface specific humidityPSurf Pa Surface air pressureRainf kg/m2 Rainfall rateSnowf kg/m2 Snowfall rateWind m/s Near surface wind speed

Energy Balance TermsSWnet W/m^2 Net shortwave radiationLWnet W/m^2 Net longwave radiationQle W/m^2 Latent heat fluxQh W/m^2 Sensible heat fluxQg W/m^2 Ground heat fluxQf W/m^2 Energy of fusionQv W/m^2 Energy of sublimationQa W/m^2 Advective energyDelSurfheat J/m^2 Change in surface heat storageDelColdCont J/m^2 Change in snow cold content

Water Balance TermsSnowf kg/m^2s Snowfall rateRainf kg/m^2s Rainfall rateEvap kg/m^2s Total EvapotranspirationQs kg/m^2s Surface runoffQsb kg/m^2s Subsurface runoffQsm kg/m^2s SnowmeltDelSoilMoist kg/m^2 Change in soil moistureDelSWE kg/m^2 Change in snow water equivalentDelSurfStor kg/m^2 Change in Surface Water StorageDelIntercept kg/m^2 Change in interception storage

Surface State VariablesSnowT K Snow Surface TemperatureVegT K Vegetation Canopy TemperatureBaresoilT K Temperature of bare soilAvgSurfT K Average surface temperatureRadT K Surface Radiative TemperatureAlbedo unitless Surface AlbedoSWE kg/m^2 Snow Water EquivalentSWEVeg kg/m^2 SWE intercepted by vegetationSurfStor kg/m^2 Surface Water Storage

ALMA Standards - OutputsSubsurface State VariablesHLayerDepth m Hydrological Soil Layer DepthSoilMoist kg/m^2 Average layer soil moistureSoilTemp K Average layer soil temperatureSMLiqFrac unitless Average layer fraction of liquid moistureSMFrozFrac unitless Average layer fraction of frozen moistureSoilWet unitless Total soil wetness

Evaporation VariablesPotEvap kg/m^2s Potential EvapotranspirationECanop kg/m^2s Interception evaporationTVeg kg/m^2s Vegetation transpirationESoil kg/m^2s Bare soil evaporationEWater kg/m^2s Open water evaporationRootMoist kg/m^2 Root zone soil moistureCanopInt kg/m^2 Total canopy water storageSubSnow kg/m^2s Snow sublimationSubSurf kg/m^2s Sublimation of the snow free areaACond m/s Aerodynamic conductance

Cold Season Process VariablesSnowFrac unitless Snow covered fractionIceFrac unitless Ice-covered fractionIceT m Sea-ice thicknessFdepth m Frozen soil depthTdepth m Depth to soil thawSAlbedo unitless Snow albedoSnowTProf K Temperature profile in the snowSnowDepth m Depth of snow layer

UW Multi-model Results

ourforecast-related multi-modelresults to date

+ gets you

Arctic Basin application• linearly combined CHASM, ECMWF, NOAH, and VIC based on snow cover simulation performance• combination reduced annual runoff error

West-wide forecast application

SIMMA implementation “hardships”Model Versions and Parameters:major experiment (PILPS, NLDAS) versions are a specialized branch off the main development tree. Reconciling desired combination of parameters, I/O, and model physics takes substantial effort.• PILPS-2e version of NOAH used NetCDF I/O, which we adopted.• NLDAS NOAH had desired CONUS domain, but we want physics from v. 2.7.1.

No standard format for parameters: - Different versions of the parameters (for the same model) had different formats (sequential binary, arcinfo, xmrg, grib, etc). - Different versions sometimes were on different grids, or even different individual parameters within the same set were on different grids.

Different model versions (for same model) use different parameters: - Some versions used different soil/veg classification schemes than others, requiring mapping them to the desired scheme.e.g. NLDAS parameters for an older version of NOAH than the desired one (2.7.1). We want parameters from NWS OHD for CONUS domain, not sure what version

Some models (e.g. SAC) use conceptual parameters that lack an obvious relation to standard soil/veg types that are easily available.

Efforts to standardize formats & datasets

(forcing/output/parameter), and/or provide

conversion tools to/from standards, will be useful.

UW forecast/nowcast application of SIMMA

Test CaseSalmon R. at Whitebird, ID

Future ApplicationsWestwide forecast domainCONUS nowcast

Starting PointNLDAS-grid implementationsof VIC/NOAH/SAC

UW Real-time Daily Nowcast

For more information:

Ted Bohn, tbohn@hydro.washington.edu

Research Issues Implementation Hurdles: grid specs, parameter inputs,

model physics, I/O formats, etc.

Combination methods in ungaged areas? (e.g., US nowcast) What are OBS?

What if combinations that produce best streamflow do not also produce best SWE or soil moisture? Is use as diagnostic physical tool compromised?

END

NEW

Expansion to multiple-model framework

Our current research with multi-model simulations is promising: 4 land surface models used to simulate arctic basin hydrology, 100 km resolution following linear combination approach of Krishnamurti et al., (2000)

weighting calibration based on simulation of snow-covered area results for streamflow and other hydrologic variables evaluated

multi-model errors are lower than single model errors, in most cases(work by Ted Bohn at U. of

Washington)

Expansion to multiple-model framework

annual discharge predictions