Transcript Document

The potential value of hydrologic predictability on Missouri River
main-stem reservoir systems
GEWEX Americas Prediction
Project 2003 PIs Meeting
1
Edwin P. Maurer and Dennis P. Lettenmaier
1. Department of Atmospheric Sciences, Box 351640, University of Washington, Seattle, WA 98195
2. Department of Civil Engineering, box 352700, University of Washington, Seattle, WA 98195
Understanding the links between remote conditions, such as tropical sea surface
temperatures, and regional climate has the potential to improve streamflow
predictions, with associated economic benefits for reservoir operation. Better
definition of land surface moisture states (soil moisture and snow water storage) at
the beginning of the forecast period provides an additional source of streamflow
predictability. We examine the value of long-lead predictive skill added by climate
forecast information and land surface moisture states in the Missouri River basin.
Forecasted flows were generated that represent predictability achievable through
knowledge of climate, snow and soil moisture states at the time of forecast. For the
current main stem reservoirs (90 billion m3 storage volume) only a 1.8%
improvement in hydropower benefits could be achieved with perfect forecasts for
lead times up to one year. This low value of prediction skill is due to the system’s
large storage capacity relative to annual inflow. To evaluate the effects of
hydrologic predictability on a smaller system, a hypothetical system was specified
with a reduced storage volume of 36 billion m3. For this smaller system there was a
7.1% increase in annual hydropower benefits for perfect forecasts, representing
$25.7 million. Using realistic streamflow predictability, $6.8 million of the $25.7
million are estimated to be realizable. The climate indices provide the greatest
portion of the $6.8 million, and initial soil moisture information provides the largest
incremental value above climate knowledge. An analysis of the seasonal variation
in the value of runoff predictability provides further insights. In general, the value of
predictability is greatest in the spring, when interannual variability is greatest;
whereas in winter and spring, the incremental benefits due to soil moisture
knowledge (beyond those realizable from knowledge of climate and snow water
equivalent state at the time of forecast), are greatest. This illustrates the potential
value of soil moisture knowledge in determining spring and summer inflows. The
results demonstrate that the use of climate forecast information, along with better
definition of the basin (snow and soil) moisture states, can provide modest
economic benefits, and that these benefits in general will increase as reservoir
storage decreases.
Science Questions
3
Varying Levels and Sources of Runoff Predictability
Season of
runoff being
predicted
Selection of Indices Characterizing
Sources of Predictability
SOI – An index identifying ENSO phase
AO – An index of phase of the Arctic
Oscillation
SM – Soil moisture
SWE – Snow water equivalent
Climate
Land
1. SOI/AO
2. SWE
3. SM
Incremental predictability assessed
for each tier
Multiple linear regression between selected predictors
(SOI/AO/SM/SWE) and runoff at different lead times
r2
Lead -2
Lead 1
Lead-0
Runoff
SOI/AO
SOI/AO
D
J
Dec 1
F
M
Mar 1
A
M
Jun 1
J
A
J
S
Sep 1
N
O
Forecast
Season
DJF
Dec 1
r2 of regression
is indicator
of predictability
Initialization Dates
for DJF Forecast
r2SWE
SWE
Missouri
River
Basin
Contributing Areas
3) What is the value of increased predictive skill to the management of
a water resources system?
Land Surface Data Used in this Study
90% of inflow
E P
Q


To derive W and E, use observations of P (and
T), which have better spatial representation to
drive a hydrologic model
Q
Illustrate that model reproduces observed Q
By water balance, E must be close over long
term
Using a physically-based land surface
representation gives confidence in seasonal
variation represented in model
dW
PE
dt
VIC model used to generate time
series of soil moisture, snow, and
runoff
Features:
•Developed over 10 years at Princeton
and UW
•Energy and water budget closes at
each time step
•Multiple vegetation classes in each cell
•Sub-grid elevation band definition (for
snow)
•Subgrid infiltration/runoff variability
Resulting Data Set used in this study:
• 50-year+ simulation using the VIC
hydrologic model
• 3-hour time step, aggregated to
monthly and seasonal values
• 1/8 degree (~12 km) resolution
• Variables include all water and energy
budget components
• Long term spatial data set allows
characterization of variability
• Described in Maurer et al., 2002

31%
90% of inflow and storage
capacity at upstream 3
reservoirs
Inflows dominated by spring
and early summer snowmelt
Variability greatest in spring
and early summer
Predictability of spring and
summer flows should provide
greatest benefits
47%
12%
Ft. Peck Garrison
4%
6%
Oahe
•Important runoff forecast skill at long lead times is limited, and due to
modest predictive skill in areas with high runoff
5000
4500
Flow
Std. Dev.
4000
3500
3000
Step month by month, for 99 years
(1898-1996) making new forecasts for
12 months ahead
Predictability
4 x 5 grid of
average
predictabilities for
each area
• Greatest value of predictability in DJF and MAM – affecting
large future inflows.
• Knowledge of soil moisture in winter and spring provides
the greatest incremental increase in benefits above that
already attainable with climate signals.
• Increased predictability in JJA with soil moisture lowers
annual value, due to variable monthly value of
hydropower.
1500
1000
500
0
Feb
Mar
Apr
May Jun
Jul
Aug Sep
Oct
Nov Dec
MOSIM system simulation model developed to simulate system operation
and hydropower generation at monthly time step
•Simulates operation of upstream 3 reservoirs – downstream are runof-river
•March 1 reservoir evacuation target for each dam: drain during fall
and winter to base of Multiple Use zone
•Uses model constraints from Corps of Engineers
•Physical limits of dams, penstocks
•Release constraints:
•Navigation
•Endangered species
•Spawning, water supply, irrigation
•Minimum hydropower generation
•Maximum release for flooding
•Maximum winter release for ice
Validation of MOSIM Storage and Energy Simulations
Predictability level set for chosen predictors in
designated season; zero predictability in all other
seasons.
Average Annual Hydropower Benefits
The 90th percentile flows (upper decile)
are the assumed level of risk (for
flooding) used for this study
2000
Missouri Main Stem hydropower:
•Constitutes the largest current system benefit
•Provides a metric for benefits of runoff predictability
Effect of differing levels of predictability on Missouri River
mainstem hydropower generation
Benefits with zero predictability: $530
million/year
Benefits with perfect forecast: $540
million/year – 1.8% gain
This is within trajectory of past studies
Benefits with MOSIM without any forecast
component: $510 million
•Smaller systems can see greater benefits of
improved determination of initial condition and
climate state
Existing System
Results with Reduced Volume System
Scenario/Forecast
Knowledge
Annual Hydropower
Benefits, $ million
Zero Predictability
359.8
Climate State
363.2
Climate + Snow
364.5
Climate + Snow +
Soil Moisture
366.6
Perfect Predictability
385.5
Multiple Use Zone
Carryover
Storage
Reductions to
Carryover Storage
and Dead Pool
Zones
Carryover
Storage
Dead
Pool
Summary
•Benefits of added predictability for a large system
are limited
Hypothetical Reduced system
To investigate the potential effect of predictability on a
smaller system in this geographical setting, a
reduced-volume scenario was developed:
Flood Control
5
Reduced-volume Missouri
River system applies
proportional reductions to 3
upstream dams
This shows:
1. Change to benefits due to modification of system
operation to incorporate forecast information
exceeds benefits added by predictability
2. System capacity is large (multi-year storage), so
seasonal predictability effect is small
Flood Control
Multiple Use Zone
Dead
Pool
Synthetic forecast inflows derived for
each reservoir by adding noise to
~
served inflows: X t  X t  t
2500
Jan
Domain coincides with LDAS-NA
For each season
and lead time:
•Establish average
predictability for
each contributing
area
•Weight each grid
cell by runoff
Seasonal Distribution of Predictability
Benefits with Reduced-Volume System
Big Bend Ft. Randall Gavins Pt.
Inflow and Variability
Inflow toReservoir
3 Upstream
Reservoirs
Flow, million cubic meters

W
•At lead times over 1 season, limited potential forecast skill due to land
surface in west and climate signal in east
Development of Predicted Inflow Sequences for each Reservoir
2) What are the relative contributions of climate conditions, snow and
soil moisture content to runoff predictability?
•Multi-decadal records needed to define
variability of soil moisture, snow water,
runoff on a seasonal time scale.
•Variability of these states and fluxes
cannot generally be determined with
observations (Roads et al., 2003)
•At a lead-0 (1.5 month), soil moisture is dominant for predictive
capability of runoff
Shaded areas are locally significant at 95%
confidence
Color indicates r2 of regression at each grid
cell
X indicates no basin-wide field significance
at 95% confidence level
Water Management Implications of Runoff Predictability in the Missouri River basin
4
1) Where is seasonal hydrologic predictability greatest, and through
what lead time is it significant?
2
Runoff predictability due to soil moisture
Variables introduced in order of
how well indices represent current
knowledge of state:
Only Use Indices in Persistence Mode
Lead-3
Runoff predictability due to snow
Increasing
Lead Time
Varying Lead Times between Initial Conditions (IC)
and Forecast Runoff
Lead-4
Runoff predictability due to climate
Benefits above zero predictability, $ million
ABSTRACT
1
2
•These benefits can be large amounts, but
represent small relative increases over current
technology
•These results are case-specific and depend on:
•Physical system for management
•Operating rules of system
•Natural variability (current vs. potential predictability)
•Time value of water
References:
•Maurer, E.P., A.W. Wood, J.C. Adam, D.P. Lettenmaier, and B.
Nijssen, 2002, A Long-Term Hydrologically-Based Data Set of Land
Surface Fluxes and States for the Conterminous United States, J.
Climate 15(22), 3237-3251.
•Roads, J., E. Bainto, M. Kanamitsu, T. Reichler, R. Lawford, D.
Lettenmaier, E. Maurer, D. Miller, K. Gallo, A. Robock, G. Srinivasan,
K. Vinnikov, D. Robinson, V. Lakshmi, H. Berbery, R. Pinker, Q. Li, J.
Smith, T. von der Haar, W. Higgins, E. Yarosh, J. Janowiak, K. Mitchell,
B. Fekete, C. Vorosmarty, T. Meyers, D. Salstein S. Williams, 2003,
GCIP Water and Energy Budget Synthesis, J. Geophys. Res. (in review).