Management of Natural and Environmental Resources for Sustainable Agricultural Development
Use of the Object Modeling System for Operational Water Supply Forecasting
ByTom Perkins (NRCS) & Tom Pagano (NRCS)
14 February 2006 Portland, OR
World Meteorological Organization (WMO)
The path we’ve trod…
•Early Years
•State-based Operations
•Centralized in Portland
•New Technologies
The first forecast of the Lake Tahoe “rise” was in 1911. This is the earliest known water supply forecast in the United States.
The snow water content on Mt. Rose, Nevada on April of 1910 was 597 mm. In 1911, it was 1128 mm. Based on those snow course measurements, Dr. James Church, predicted that the Lake Tahoe “rise” would be 189 percent of the previous year.
A bit short of water with such a short calculator!
• Modernize water supply forecasting environment
• Desire for new hydrograph- based products
• We’ve got all of this daily SNOTEL data
• Modeling dream
• 1983 reorganization
Simple linear regression
Stepwise linear regression
Principal components statistical techniques
Statistical analysis with jackknifing
Non-linear statistical analysis
Z-score statistical techniques
110 baud communications to Fort Collins, Colorado
Water resource managers and users want/need forecasts of any parameter you can derive from a hydrograph:
Flow on a dateDate of a threshold
Magnitude and date of peak
Daily scenarios for use in reservoir optimization programs
Weekly, monthly volumes
Some forecast methodologies are decades to centuries old
MODELS
Leavesley 1985Marron 1986(PRMS)
Jones 1986Perkins 1988(SSARR)
Cooley 1986 (NWSRFS)
Shafer/Marron 1987 (SRM)
Garen/Marks 1996-98(spatial snow model)
NRCS Operational Simulation Modeling History
Competition and increasing demands on our finite water resources are causing water managers to modify their management strategies
Water resource managers want to know more than just seasonal volumes
1991 NRCS Simulation Modeling Plan
• Compared options: SRM, SSARR, PRMS, NWSRFS
• Calibrate 200 basins in 5 years with 3 staff hydrologists
Several recent technological advances make simulation modeling do-able with fewer human resources
So…what’s changed?
Current Modeling Activities
•Models can be tailored to forecaster’s situation/need•There is no need to reinvent modeling infrastructure•System is flexible and up-gradable•Partnership and technology sharing with other agencies•Widely used and documented
Precipitation-Runoff Modeling SystemPRMS
Modular Modeling SystemMMS
Current modeling components• ArcGis to reproject downloaded basin digital elevation model data
• GIS Weasel to distribute hydrologic response unit (HRU) parameters
• Precipitation-Runoff Modeling System (PRMS) model - (includes HRU parameter distribution and ensemble streamflow prediction (ESP) modules)
• Microsoft Office Excel-based automatic data downloading routines and output viewer
Components missing on NWCC system• In-house application to develop relationships to distribute precipitation and temperature
over the basin hydrologic response units
• Automatic, step-wise, multiple-objective calibration procedure.
USGSModularModelingSystem
Currently using“off the shelf”PRMS package developed by
USGS
Calibration of PRMS Spatial Parameters
The GIS “Weasel”
Uses slope, aspect, elev, soils, veg type, veg density to define “Hydrologic Response Units” andassociated model parameters
“1-button” Weaselrecently developed and is being tested
Estimating non-spatial parameters
Traditional Approach
Manually tweak parameters to minimize bias,visually fit hydrograph to observed flow.
Not without its problems
Estimating non-spatial parameters
Traditional Approach
Manually tweak parameters to minimize bias,visually fit hydrograph to observed flow.
Not without its problems
Multi-step Automatic Calibration Scheme (Hay/USGS)
Iteratively and automatically calibratemodel internal states.
Not without its problems
Solar Radiation
PotentialEvapotrans
Water Balance
Peak Flows
Adjust radiation-related parameters. Check if model seasonal cycle matches “observed” radiation data.
Solar Radiation
PotentialEvapotrans
Water Balance
Peak Flows
Adjust radiation-related parameters. Check if model seasonal cycle matches “observed” radiation data.
Adjust evaporation parameters. Check seasonal cycle against “observed” potential ET data.
Solar Radiation
PotentialEvapotrans
Water Balance
Peak Flows
Adjust radiation-related parameters. Check if model seasonal cycle matches “observed” radiation data.
Adjust evaporation parameters. Check seasonal cycle against “observed” potential ET data.
Adjust water balance parameters.Check annual flow volume vs obs.
Solar Radiation
PotentialEvapotrans
Water Balance
Peak Flows
Adjust radiation-related parameters. Check if model seasonal cycle matches “observed” radiation data.
Adjust evaporation parameters. Check seasonal cycle against “observed” potential ET data.
Adjust water balance parameters.Check annual flow volume vs obs.
Adjust flow timing parameters.Evaluate flow on peak flow days.
Solar Radiation
PotentialEvapotrans
Water Balance
Peak Flows
Adjust radiation-related parameters. Check if model seasonal cycle matches “observed” radiation data.
Adjust evaporation parameters. Check seasonal cycle against “observed” potential ET data.
Adjust water balance parameters.Check annual flow volume vs obs.
Adjust flow timing parameters.Evaluate flow on peak flow days.
Rinse and repeat 4-8 times.
Gathering Data
Gathering DataWanted:A cheap, clean, reliable supply
Applied climate information system (ACIS)National Water Information System (NWIS)SNOw TELemetry System (SNOTEL)HYDROlogic and METeorologic Monitoring System (Hydromet)Others….
Automated data networks…
Snotel (NRCS) Network ACIS (NWS) Network
Mar 31, 2005
Blue: Reporting
Red: Missing
Data gathering and screening for MMSReal-time data automatically downloaded and
reformatted daily
• Real-time data quality “a concern”• Martyn Clark (University of Colorado) has created a
real-time temperature and precipitation quality control module that can be used stand-alone or as part of Modular Modeling System (MMS).
• Martyn also provided an initial cleaned up historical NWS/NRCS dataset
• United States Geological Survey (USGS)• National Weather Service (NWS)• Regional Climate Centers (RCC-ACIS)• Natural Resources Conservation Service (NRCS)
Results to date……..
PRMS-MMS Calibration and Operations
Missouri BasinColumbia BasinColorado/Rio Grande
16 headwater basins in diverse climates
NWCC personnel calibratedspatial parameters (Oct 2004)
USGS has automatic procedure to calibrate remaining parameters (Nov 2004)
USGS/USBR also running model in Gunnison, San Juan, Rio Grande, Carson, Yakima, Klamath, etc., so additional basins could be provided to us.
PRMS-MMS Calibration and Operations
Missouri BasinColumbia BasinColorado/Rio GrandePlanned
16 headwater basins in diverse climates
NWCC personnel calibratedspatial parameters (Oct 2004)
USGS has automatic procedure to calibrate remaining parameters (Nov 2004)
USGS/USBR also running model in Gunnison, San Juan, Rio Grande, Carson, Yakima, Klamath, etc., so additional basins could be provided to us.
Specific plans to increase roster to ~35
Animas River at Durango, Colorado
East River at Almont, Colorado
Yampa River at Maybell, Colorado
Black = observed
Red = simulated
NRCS Spreadsheet-based output interface
Little Wood Reservoir, Idaho – May 2005
Slide for fatheadPRMS
NOHRSC-SatelliteHistoricalSimulated
Snow Covered Area SimulationFr
actio
n co
vere
d
ESP conditional forecast using calibration #2Salmon Falls Creek nr San Jacinto, Nevada (USGS gage 13105000)
Initialized January 6, 2005
Mar-Jun Mar-Jul Mar-Sept Peak Peak Date(106m3) (106m3) (106m3) (m3/s)
Minimum 9.87 14.80 19.74 3.34 11-Mar90% exc 22.20 29.60 34.54 6.91 20-Apr70% exc 37.00 48.11 53.04 11.33 30-Apr50% exc 60.44 76.48 83.88 16.68 4-May30% exc 74.01 93.77 99.91 26.22 15-May10% exc 118.41 140.62 148.02 33.05 1-JunMaximum 151.72 185.02 194.89 44.32 10-Jun
Observed 109.16 115.08 120.63 39.64 18-May
Basin Area = 3755 km2
Probability distribution based on 23 historical years
PRMS Ensemble Streamflow Prediction (ESP) results:
Future Modeling Activities
OMS-PRMS
OMSObject-oriented Modeling System
•Library of science and database components
•Facilitates assembly of modeling packages
•Supported by graphical user interface modules
•Data retrieval, statistical, visualization utilities
•EXtensible Markup Language mechanisms
•Web based sharing of modeling resources
Pre-ProcessorsAccess & prepare data
Pre-ProcessorsAccess & prepare data
ModelsSimulate hydrologic & ecosystem processes
Pre-ProcessorsAccess & prepare data
ModelsSimulate hydrologic & ecosystem processes
Post-ProcessorsDisplay & analyze model
results
• OMS will be the modeling platform for NRCS• TR20, TR55, WEPP, PRMS will be included
initially• Other models will be considered in the future• The NWCC Water Supply Forecasting models
will be the initial program prototypes
Plots of individual parameters vs time
Plots of combined parameters vs time
Zoom feature
Data Plots
Probability Distributions
XY Plots
Flow Durations
Observed/Predicted Statistics
Etc.
Current OMS Modeling Components•PRMS model with all MMS modules
•Graphical User Interface
•Output module
Soon to be completed components•Hydrologic response unit parameter distribution modules•Hydrologic response unit delineation module•Automatic calibration module•Conditioned ensemble streamflow prediction scenarios•10-day quantitative climatic forecast interface•Data acquisition modules•Data quality control routines•Report analysis