Improving Seasonal Forecasting in the Snake River Basin, Idaho

1
Improving Seasonal Forecasting in the Snake River
Basin, Idaho
University of Washington University of British
Columbia Fall Hydrology Workshop Oct 3,
2003 Marketa McGuire With contributions
from Alan Hamlet, Andy Wood, Kostas Andreadis,
Dennis Lettenmaier
2
Objectives

• To evaluate the impact of remote sensing data for
  improved estimates of initial snow conditions in
  streamflow forecasting
• To evaluate the impact of streamflow forecast
  products for improved water resources operations

3
Motivation

• In regions like the Snake River basin, where
  spring and summer streamflows are dominated by
  snow-melt, it is important to know accurately the
  extent of snow in the initial condition of a
  streamflow forecast
• Previous work by Maurer et al (2003) suggests
  that MODIS remotely sensed snowcover has the
  potential to improve hydrological modeling and
  prediction in the Snake River basin
• Current meteorological station data are
  provisional
• Result estimates of forecast initial conditions
  have large uncertainty

4
Forecasting Approach using MODIS Updating
Initial Conditions soil moisture, snowpack

• Ensemble Forecast
• streamflow,
• soil moisture,
• snowpack,
• runoff

Hydrologic model spin up
Hydrologic simulation
local scale weather inputs
NCDC met. station obs. up to 2-4 months from
current
LDAS/other real-time met. forcings for remaining
spin-up
MODIS Update
25th Day of Month 0
1-2 years back
End of Month 6 - 12
5
Variable Infiltration Capacity (VIC) Model
Snake River 1/8 Resolution Routing Flow Network
6
MODIS Snowcover April 4, 2000(500 m)
VIC SWEApril 4, 2000(1/8 degree)
Snow
Land
Snow (SWE gt 5mm)
Clouds
No Data/No Decision/Saturated
Land (within Snake River Basin)
7
Updating VIC Snow State

• Current Version of VIC model
• Snow model runs on a 3 hours timestep
• Each grid cell has up to 5 elevation bands
• Each elevation band either has snow (coverage
  1) or does not (coverage 0)
• VIC model with updated snowcover (2 options)
• Apply MODIS snowcover uniformly over elevation
  bands based on some threshold fraction of
  snowcover
• Utilize VIC model version 4.1 that incorporates
  fractional snowcover (in testing phase Fall 03)

8
MODIS Fractional Snowcover

• Idaho National Environmental and Engineering
  Laboratory (INEEL) is processing snow cover
  fractions for VIC model grid cells
• Have obtained all Daily Snow cover tiles
  available for Snake Basin from February 2000 to
  present
• Have automatic subscription with the NSIDC to
  obtain all newly processed scenes, with lag time
  of 2-3 days
• Working toward fully automating the prototype
  algorithm to provide near real-time snow cover
  fractions for the Snake River Basin test
  application

9
Strategy for Evaluating the Impact of MODIS

• Conduct a sensitivity analysis to determine the
  importance of snow in the Snake River basin
• Analysis based on discrepency between MODIS
  snowcover and VIC snowcover
• Compare streamflow forecasts, with and without
  MODIS updating, beginning at various dates
  throughout the winter
• Hypothesis Updating will be more valuable for a
  streamflow forecast in early winter and spring
  when snow cover tends to change rapidly
• Compare streamflow forecasts, with and without
  MODIS updating, to streamflow forecasts produced
  by the NRCS for a subset of basins within the
  Snake River Basin

10
Objectives

• To evaluate the impact of remote sensing data for
  improved estimates of initial snow conditions in
  streamflow forecasting
• To evaluate the impact of streamflow forecast
  products for improved water resources operations

11
Modeling Approach to Evaluate Operations
Observed Meteorological Time Series
Precip, temp, wind, etc.
Streamflow Forecasts
Reservoir Forecasts
Water Resources Operations Model (SnakeSim)
Hydrology Model (VIC)
Storage, reliability, spill, energy
Streamflow at 21 locations
MODIS fraction of snowcover
12
Ensemble Streamflow Prediction
Ensembles
Streamflow Ensemble
13
Sample Streamflow Forecast
Jackson Lake Inflows
www.hydro.washington.edu/Lettenmaier/Projects/fcst
/index.htm
14
Jackson Lake
1997 WY
1992 WY
1977 WY
15
Jackson Lake
1997 WY
1962 WY
1961 WY
16
(No Transcript)
17
Linking Streamflow Forecasts to SnakeSim
SnakeSim Water Resources Operations Model
Ensemble Streamflow Forecast
Bias Correction
Demand Scenarios
Initial Reservoir Contents from USACE or USBR
Storage Ensemble
18
SnakeSim Operations Model Overview

• Developed by Nathan VanRheenen, UW
• Stella modeling environment
• Simulation for 1950 - 1992
• 21 Inflow Nodes, utilizing
• Historic naturalized flows
• Routed flows from VIC model
• 18 Reservoirs Modeled
• 13.3 MAF Total Storage (16.4 BCM)
• Simulation of Snake River Plain Aquifer
• Historic Demand Scenarios

19
SnakeSim Operations Model Assumptions

• Current levels of operation adhering to IDWR,
  BOR, COE rules for reservoir storage and releases
• Instream targets for fish, water quality, and
  hydropower production
• Surface water diversions grouped by river reach
• Groundwater response curves are linear and based
  on University of Idaho algorithms
• 1980 groundwater pumping curves and irrigation
  areas

20
Portion of Domain in Storage Forecast
21
System Storage Forecast from SnakeSim Jackson
Lake Palisades Island Park Ririe American
Falls Lake Walcott 11 ENSO neutral years Random
historic demand scenarios
Full Pool
Full Pool
22
Strategy for Evaluating the Use of Streamflow
Forecasts in Water Management

• Do similar comparisons as with streamflow
  forecasts
• Sensitivity analysis to determine importance of
  snow
• Compare operations, using streamflow forecasts
  with and without MODIS updated snow cover,
  beginning at various dates throughout the winter
• Other

23
Summary Current Status

• We produced first set of streamflow forecasts for
  Sept 1, 2003 for 21 locations in the Snake River
  basin
• Snake River basin streamflow forecasts, updated
  every month, will now be available on the web as
  part of the UW S/I Hydrologic Forecast System
• (www.hydro.washington.edu/Lettenmaier/Projects/fcs
  t/index.htm)
• Acquisition of near real-time MODIS fractional
  snowcover for the Snake Basin is in progress
• Testing of VIC 4.1 planned for Fall 2003
• (utilizes fractional snowcover for each elevation
  band)

24
Questions ?
25
Bias Correction Objectives
Raw
Bias Corrected
Result Bias corrected hydrologic simulations
are quite consistent with observed streamflows in
absolute value and climate change signals are
translated without significant distortion.
26
Quantile-Based Bias Correction (Wood et al. 2002)
VIC Input 19000
Bias Corrected Output 10000
27
(No Transcript)