Introduction
The Hanford Site
What does in, must come out….
recharge Other losses..including external water exchanges
K-U1 Hanford (Unit 1) Hydraulic Conductivity Multiplier K-U5 Ringold (Unit 5) Hydraulic Conductivity Multiplier K-U7 Ringold (Unit 7) Hydraulic Conductivity Multiplier K-U9 Ringold (Unit 9) Hydraulic Conductivity Multiplier SY-H Hanford (Unit 1) Specific Yield Multiplier SY-RU5 Ringold (Unit 5) Specific Yield Multiplier F-CC Cold Creek Valley Flux Multiplier F-DC Dry Creek Valley Flux Multiplier F-NR Natural Recharge Multiplier F-RH Rattlesnake Hills Flux Multiplier. Model parameters Application of “principal of parsimony”
Hillside and Piezometers
System Properties Transmissivity = 100 m 2 /day Creek conductance is very high Recharge = 30 mm/yr
Groundwater levels
Transmissivity distribution - I 100 m 2 /day
12 m 2 /day 360 m 2 /day Transmissivity distribution - II
Chesapeake Bay
Watershed Model by Major Basin
P4 Potomac Segmentation
Simulated with HSPF
Loading Sources in Watershed Model PastureHay Imp Urb Cons. Till Perv UrbForest Conv. TillManure RIVER REACH Mixed Point Source Septic
Crop Simulation MeteorologyPrecipitation Runoff and Groundwater Land Morphology Nitrogen Cycle Sediment Export Phosphorus Cycle Nutrient Inputs
Crop Simulation Water Sediment AGCHEM PQUAL Water Sediment
Water simulation Ground Water Surface Interflow Lower Zone
Sediment Simulation Detached Sediment Soil Matrix (unlimited) Wash off Detachment Attachment f(rain intensity) f(time)
Trees Roots Leaves Particulate Refractory Organic N Particulate Labile Organic N Solution Ammonia Nitrate Solution Labile Organic N Adsorbed Ammonia Solution Refractory Organic N Nitrogen Cycle in Watershed Model Forest Atmospheric Deposition Denitrification
River Sediment Simulation Suspended Sediment Bed Storage (unlimited) Outflow Scour Deposition in Phase IV Watershed Model Inflow
N River Simulation Algae ORGN NO3 } Sediment NH3
SNOW section of the PERLND module of HSPF
PWATER section of the PERLND module of HSPF
(continued)
P4 Potomac Segmentation Gauging station
Daily Flow
Monthly Volume
Exceedence fraction
lzsn infilt agwrc deepfr E E-03 basetp E agwetp E E-03 uzsn intfw irc lzetp Parameter values
lzsn infilt agwrc deepfr E E e-2 basetp E e-3 agwetp E E e-3 uzsn intfw irc lzetp More parameter values
The common thread A model parameterised on the basis of “outside measurements” alone will probably fit field data poorly. Hence it must be “calibrated”.
The common thread No matter how many parameters a model has (ie. no matter how complex it is) normally only a handful of parameters will be adjusted during the calibration process. The model’s complexity then becomes a veneer.
The common thread Even after a model has been calibrated, it is still very rare for model to have the ability to simulate every nuance of a system’s behaviour.
The common thread Even if it could, a multiplicity of parameters would normally provide an identically good fit between model outcomes and field measurements. The more parameters that a model has (ie. the more complex that it is, the greater the extent of parameter nonuniqueness).
Soooooooo…. When we make a prediction using our model, especially one that involves fine detail, to what extent can we believe the model?
Soooooooo…. And which of the many possible parameter sets that calibrate the model do we use in making this prediction?