Overview

Design Flood Estimation Using RORB • Calibration philosophy

10th-11th February 2020 • Model parameters
Warragul • Errors
• Baseflow
• Regional relationships for kc
Model Calibration

1 2

Calibration philosophy Calibration Issues

• Objective function for calibration:
• 2 basic types of applications
– what is being optimised?
– Gauged catchments:
• calibration to observed rainfall/flood events Rainfall
Input
• independent testing of calibrated model Match: Peak?
• reconciliation of design flood estimates with results of Volume?

Flow
flood frequency analysis “Blackbox”
• comparison with regional parameter values of
Streamflow
Catchment
– Ungauged catchments, parameter estimates from: Output
Time
Model Parameters
• adjacent gauged catchment(s)
• regional estimation methods
- calibration represents compromise between
all relevant pieces of information

3 4

Choice of objective function RORB Model Parameters

• Depends on how the model will be used
• Routing parameters:
– E.g. Peak flow or volume sensitive problem?
– Kc
Calc. Calc. – m (most likely 0.8)
Actual Actual
• Losses - three choices of loss model
Flow

Flow

– Initial loss/Continuing loss (most likely appropriate)

– Initial loss/Proportional loss
Time Time – SWMOD
– Either of these calibrations may be appropriate
depending on the selected metric

5 6
 Effect of model parameters - Kc Effect of model parameters - IL
• Increasing Kc attenuates the hydrograph • Reduces effective rainfall at the start of the
– Lower and later peak event (until the IL is completely satisfied)
– No change in hydrograph volume • Less impact on the end of the hydrograph

Rain
High Kc High IL
Low Kc Low IL
Flow

Flow
Time
Time

7 8

Effect of model parameters - CL Model Parameters - Issues

• Reduces effective rainfall by a flat rate in
• Data problems:
each time-step • errors in basic rainfall/streamflow data
• Lowers entire hydrograph • inadequate data resolution to define
temporal and spatial rainfall patterns
=> errors are forced into parameter estimates
Rain

High CL
Low CL
• Parameter interaction:
• between loss and routing parameters
=> requires sequential calibration approach:
Flow

1. loss parameters: hydrograph volume

2. routing parameters: hydrograph shape
• between IL and CL, or kc and m:
=> fix one parameter, optimise the other
Time

9 10

Model Parameters - Issues Baseflow

Quickflow (rainfall
• Subjective decisions by modeller: excess runoff,
• baseflow separation
quick response)
Total streamflow

• interpretation of rainfall data

• Inadequate model structure: Baseflow (delayed

• difficulties in calibration may indicate response, e.g.
problems with model structure groundwater)
=> only solution: change model structure
• Extrapolation to different conditions:
• needs to be guided by hydrological reasoning Time

11 12
 Baseflow Baseflow Separation
• RORB only models surface runoff – baseflow • Manually drawn (somewhat subjective)
processes aren’t included • At the start and end of the event quickflow equals zero
• Baseflow increases from the time of peak flow to the
• For calibration, can estimate a baseflow point of inflection on the falling limb
hydrograph from gauged data
Baseflow Peak
• Two choices under peak baseflow

• Remove the baseflow from the gauged data

Flow

Flow
• Input the baseflow estimate as an inflow hydrograph
• For design, see ARR2016 Book 5 Chapter 4 for
guidance on how to model baseflow Time Time

13 14

RORB settings – FIT vs DESIGN

• FIT mode best suited to analysis of observed
Calibration to hydrograph
Observed Event • Initial loss selected by user to match hydrograph
rise
• Continuing/proportional loss parameter selected
by program to match volume of observed
hydrograph
• Care needed in baseflow separation and rainfall
inputs
• DESIGN mode can be used to manually fit both
initial and continuing/proportional losses

15 16

Interactive parameter fitting Regional Relationships for Vic

kc = 2.57A0.45 Mean annual rainfall > 800 mm

kc = 0.49A0.65 Mean annual rainfall < 800 mm

A = catchment area (km2)

17 18
 Regional kc Relationships Average flow distance
• Average of the distances from each sub-area
centroid to the model outlet
• Computed by RORBwin

Greer (1992): Flood data for Queensland catchments including design flood estimates Water Resources Commissions

19 20

c0.8 = kc/dav v Area Parameter values for c0.8

Pearse, M., Jordan, P., Collins, Y., 2002. A simple method for estimating RORB model parameters for ungauged
rural catchments, in: Hydrology and Water Resources Symposium, Melbourne.

10

Group Victorian Yu CRCCH

c0.8 Expected 1.25 0.96 1.14

c 0.8

c0.8 Low
0.75 0.47 0.61
(-1 SD in log)
0.1 c0.8 High
10 100 1000 10000 2.07 1.94 2.13
Area (km²)
(+1 SD in log)

21 22

21 22

Summary of Normal Ranges for Parameters

RORB
Parameter
Normal Range
Mathematically
Possible Range
Notes
Summary
kc Varies with size Varies with size of Increasing kc causes peak flows to decrease
of catchment. catchment.
Expect to be Expect to be within • Calibration of routing and loss parameters to
within ±30% of -50%/+100% of the
the kc value kc value from historic events
from calibration. calibration.
m 0.8 0.65 to 1 Normally only has a secondary influence on the
results of verification. Avoid modifying from 0.8
• Use of regional information
unless you have strong evidence to the contrary.
Initial Loss 20 to 60 mm 0 to 150 mm Increasing IL causes peak flows to decrease.
(Median Increasing IL may increase the critical duration.
value of Regional estimates available from ARR data hub
distribution)
Continuing 2 to 5 mm/h 0 to 8 mm/h Increasing CL causes peak flows to decrease.
Loss Increasing CL may increase the critical duration.
Regional estimates available from ARR data hub

23 24