1. LMRFC March, 2009 Calibration at Finer Time and Space Scales

2. LMRFC March, 2009 Hydrologic Modeling Challenges We cannot directly apply physical laws to some components of the hydrologic cycle because boundary conditions and system properties are unknown at all locations, e.g. –Exact soil depth or plant rooting depth is unknown –Soil matrix hydraulic properties are unknown (e.g. hydraulic conductivity) –Underground flow paths are unknown Errors and uncertainty in data an models Model and data errors tend to increase at higher resolutions Modeling ungauged locations –Difficult to verify models –Difficult to determining warning thresholds

3. LMRFC March, 2009 Calibration with NEXRAD at smaller spatial and temporal scales Anytime a model is calibrated at one spatial and temporal scale it should be recalibrated if the time/space scale changes

4. LMRFC March, 2009 surface supplemental TCI direct interflow primary Sub-basin scale in HRAP bins 1X1 2X2 4X4 8X8 16X16 32X32 64X64 Relative sensitivity of SAC runoff components to sub-basin scale. Runoff values have been normalize by the value at the 8X8 scale. 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 Scaled runoff value

5. LMRFC March, 2009 (a) Lumped Basin (c) Basin disaggregated Into 16 cells (d) Basin disaggregated into 100 cells (b) Basin disaggregated into 4 cells “Truth Scale” and “Truth Simulation” Expectations: Effect of Data Errors and Modeling Scale

6. LMRFC March, 2009 Expectations: Effect of Data Errors and Modeling Scale Relative Sub-basin Scale A/A k 110 100 10 15 20 25 30 0 5 Relative error, Ek, % (lumped) (distributed) Noise 0% 25% 50% 75% Data errors (noise) may mask the benefits of fine scale modeling. In some cases, they may make the results worse than lumped simulations. Simulation error compared to fully distributed ‘Truth’ is simulation from 100 sub- basin model clean data

7. LMRFC March, 2009 Model Errors as a Function of Scale Flash floods 260 Distributed model (uncalibrated). Each point is an average peak flow error from approximately 25 events over an eight year study period Oct 1996- Sept.2004. Scaling relationship for an uncertainty index (Rq) from Carpenter and Georgakakos (2004) (secondary axis) Log-linear regression for distributed model data

8. LMRFC March, 2009 Overall Rmod vs Basin Size Calibrated Models Sprin Wsilo Caves Dutch KNSO2 Elm Powel Connr Savoy Lanag ELDO2 BLUO2 SLOA4 TIFM7 TALO2 DMIP 2 Results 37 sq km2484 sq km

9. LMRFC March, 2009 Tools XDMS spatial display (ABRFC) ICP PLOT-TS time series display Stat-Q statistics program Calibration Assistance Program (CAP) –Soils –Parameters Calb MAPX Calb MAP (1 hour)

10. LMRFC March, 2009 July 1, 1999 event. Rain fall on 6/30/99 hours 10,11,12, and 13 TALO2 T=10 T=12 T=11 T=13 Basin Shape: Case 2 - XDMS Plots of Radar Rainfall You can use XDMS now!

11. LMRFC March, 2009 Distributed Model Implementation Use with, not instead of, lumped model at same time step Part of natural progression to finer scales Lumped 6-hr Lumped 1-hour Distributed 1-hour Calibration is good training process for forecasting Current: –DHM: operation in NWS for headwaters, locals –HL-RDHM: Large area, soil moisture, FFG, etc Feedback to OHD

12. LMRFC March, 2009 Hydrograph at Location A Hydrograph at Location B Hydrographs at Basin Outlet 0 40 80 120 160 200 4/3/99 0:004/3/99 12:004/4/99 0:004/4/99 12:004/5/99 0:004/5/99 12:004/6/99 0:00 Flow (CMS) 0 40 80 120 160 200 4/3/99 0:004/3/99 12:004/4/99 0:004/4/99 12:004/5/99 0:004/5/99 12:004/6/99 0:00 0 40 80 120 160 200 4/3/99 0:004/3/99 12:004/4/99 0:004/4/99 12:004/5/99 0:004/5/99 12:004/6/99 0:00 Flow (CMS) B A Distributed Lumped Observed Flow (CMS) Use with, not instead of lumped model Distributed and Lumped Operations

13. LMRFC March, 2009 Case 1: October 23, 2007 24-hour Rainfall Black Creek near Brooklyn, Miss. Distributed Modeling for Operational River Forecasts 5 inches in 24 hours Basin Location

14. LMRFC March, 2009 Actual River Forecast: Black Cr. At Brooklyn, Miss. Oct. 23, 2007 Distributed Modeling for Operational River Forecasts Lumped model Observed flow Distributed model