1. Fritz Fiedler Calibration 2290 East Prospect Road, Suite 1 Fort Collins, Colorado 80525 National Weather Service River Forecast System Cooperative Program for Operational Meteorology, Education and Training Hydromet 00-1

2. C8 Calibration  Calibration process –Estimation of parameter values which will minimize differences between observed and simulated streamflows  Calibration problems –Parameter interaction –Non-unique solutions –Time-consuming –Inaccuracies –Non-linearities –Lack of understanding

3. C8 Calibration Continued...  Goals –Unbiased reproduction of historical conditions –Parameters that cause model components to mimic the hydrologic processes they were designed to represent –Ability to extrapolate beyond conditions encountered in historical record  To do this, must understand the watershed processes as they relate to model structure

4. C8 Calibration System Parameter estimation/optimization and watershed simulation  Input –Point or areal estimates of historical precipitation, temperature, and potential evaporation –Initial hydrologic conditions  Output –Basin areal averages for point value inputs –Simulated hydrographs for historical analysis or use in ESP –Parameter values for models in operational forecast and ESP systems

5. C8 Calibration System (continued)  Characteristics –Performs computations for few forecast points for many time steps –Uses operations table –Compatible with operational system and ESP –Produces graphical output for manual calibration –Includes algorithms for automatic optimization  Applications –Historical watershed simulation –Model calibration

6. C8 Model Calibration  Basin-Wide Strategy –Select river system –Prepare data  MAP - Mean Areal Precipitation  MAT - Mean Areal Temperature  PE - Potential Evaporation  QME - Mean Daily Discharge  QIN - Instantaneous Discharge

7. C8 Model Calibration (continued) –Calibrate least complicated headwater basins  Select calibration period  Estimate initial parameter - observed Qs  Trial and error using MCP3/ICP (Interactive)  Statistics, observed versus simulated plots  Hydrograph components  Proper approach to parameter adjustment  Automatic parameter optimization - OPT3  Fine tuning - MCP3 –Calibrate other headwater areas –Calibrate local areas

8. C8 Model Calibration (continued)  Important considerations –Model structure, simulation processes –Effects of parameter changes –Use of the forecast information

9. C8 Data Preparation MAP Algorithms - Mean Areal Precipitation Techniques for converting point precipitation measurements into areal measurements and distributing them properly in time Daily and hourly data Grid point algorithm Estimating precipitation at a point (1/D 2 ) Estimate: >least, <greatest 100-150 points within basin Normalize at each grid point, then renormalize Thiessen weights Grid point versus Thiessen Two-pass algorithm - distribute daily, then estimate missing Consistency plots MAT Algorithms - Mean Areal Temperature Max - min data Grid point algorithm (1/D) Elevation weighting factor Centroid (1/D P ) Conversion to mean temperatures Consistency plots MAPE - Mean Areal Potential Evaporation Evaporation pan data MAPE vs. Mean seasonal curve QME QIN

10. C8 Historical Data Analysis General Information Needed Station data in Calibration files Station history info - observation times, changes, location, moves Topog map of basin MAP Specific Information Non- Mountainous Mountains --basin boundary --isohyetal map --station weights MAT Specific Information --mean max/min temperatures Non-Mountainous Mountains --basin boundary --areal-elev curve MAPE Specific Information --Evaporation maps --mean monthly evap --station weights MAP3 (re)check consistency generate time series of MAP PXPP check consistency compute normals MAT3 generate time series of MAT MAT3 check consistency TAPLOT3 get mean max/min for mean zone elevation MAPE check consistency generate daily time series of MAPE Precipitation TemperatureEvaporation

11. C8 Sacramento Soil Moisture Accounting Model

12. C8 Sacramento Model Structure E T Demand Impervious Area E T Precipitation Input Px Pervious Area E T Impervious Area Tension Water UZTW Free Water UZFW Percolation Zperc. Rexp 1-PFREE PFREE Free Water Tension Water P S LZTW LZFP LZFS RSERV Primary Baseflow Direct Runoff Surface Runoff Interflow Supplemental Base flow SideSubsurface Discharge LZSK LZPK Upper Zone Lower Zone EXCESS UZK RIVA PCTIM ADIMP Total Channel Inflow Distribution Function Streamflow Total Baseflow

13. C8 Hydrograph Decomposition Supplemental Baseflow Primary Baseflow Interflow Surface Runoff Impervious and Direct Runoff Discharge Time

14. C8 Sacramento Soil Moisture Components Impervious and Direct Runoff Surface Runoff Interflow Supplemental Baseflow Primary Baseflow SAC-SMA Model Evaporation Precipitation Upper Zone Lower Zone PerviousImpervious

15. C8 Initial Parameter Estimates By Hydrograph Analysis

16. C8 Initial Parameter Estimates By Hydrograph Analysis (continued) LZSK - Supplemental baseflow recession (always > LZPK) Flow that typically persists anywhere from 15 days to 3 or 4 months

17. C8 Initial Parameters Estimates by Hydrograph Analysis (continued)

18. C8 Initial Soil Moisture Estimates by Hydrograph Analysis (continued)

19. C8 Interactive Calibration Strategy  Remove large biases that may mask other problems  Adjust baseflow parameters  Adjust tension water capacities  Adjust parameters to get proper storm runoff simulation  Final adjustments to improve seasonal and flow interval bias statistics

20. C8 Multiyear Statistical Output

21. C8 Multiyear Statistical Output (continued)

45. C8 Calibrated Models in Forecasting  Data Issues –Differences in Raingauge Networks  New/Removed Stations; Mountainous Areas  Examine Effect on MAP Seasonal Statistics –Differences in Measurement Sensors  Radar –Use of Forecast rather than Observed Precipitation  Adjust for Biases, Estimate Uncertainty –Differences in Spatial and Temporal Resolution

46. C8 INTERACTIVE vs. AUTOMATIC  Process representation  Requires knowledge of physical model basis  Labor intensive  Multiple performance criteria  Less affected by data quality  Good potential for reliable future simulations  Model fit to data  Treats model as non- linear regression  Not labor intensive  Small number of statistical criteria  Highly sensitive to data quality  Uncertain value of future simulations