IAHS 2013, 23th of July 2013
Enhancing model realism by incorporating
landscape characteristics: Topography driven
rainfall runoff modeling (FLEX-TOPO)
By:
Shervan Gharari
Markus Hrachowitz
Fabrizio Fenicia
Hongkai Gao
Tanja Euser
and Hubert Savenije
Hunting the Hydrological Snark
(Andreassian et al., 2009)
Questions…
How to see the hydrological
Snark…?
How far is the hydrological
Snark…?
How big is the hydrological
Snark…?
How does the hydrological
Snark look like…?
The Hydrological Snark &
Questions…
Are we in love with the
hydrological Snark?
Or…
Are we in love with the
instruments used to find the
hydrological Snark?
What are the objectives…
• To apply the lessons learned from case specific
studies and to be able to incorporate our expert
knowledge into modeling.
[Savenije (2009)]
• To use topography and HRUs in model
development.
• To get the most out of the least knowledge
available.
Wetland
Hillslope
Plateua
Gharari et al., (2011) HESS
• Landscape classes for this case study are:
-Wetland: water table close to the surface, saturation
overland flow
-Hillslope: steep slopes,
shallow rapid
subsurface flow
-Plateau: modest slope, deep percolation
• Assigning dominant runoff processes to each
landscape and develop the model structures
accordingly.
FLEXB
FLEXA
P
I
P
I
Lumped
T
P
Remainder
(85%)
T
Ru
Rf
RC
RP
Wetland
(15%)
T
Ru
Qf
I
Rf
Ru
Rf
Qfr
Rs
Rs
RP
Qfw
RC
RP
Qs
Qs
FLEXC
P
I Plateau
(40%)
T
P
I
P
Hillslope
(45%)
T
Ru
I Wetland
(15%)
T
Ru
Rf
Ru
Qfh
RP
Rs
RP
Rs
Rf
Qfw
RC
RP
Qs
Although complex …
Parameter constraints:
-Hillslope are mostly forested  Imaxh> Imaxw & Imaxh>Imaxp
-Hillslope has longer lag time compare to wetland  Lagw< Lagh
-Plateau has a deeper unsaturated zone compared to wetland and
hillslope  Sumaxp>Sumaxh & Sumaxp>Sumaxw
FLEXC
P
I Plateau
(40%)
T
P
I
P
Hillslope
(45%)
T
Ru
I Wetland
(15%)
T
Ru
Rf
Ru
Qfh
RP
Rs
RP
Rs
Rf
Qfw
RC
RP
Qs
Although complex …
Process-based constraints:
-Peak flows during a dry period are mostly generated by the
wetland Qfw,dry/Qfh,dry>1
-Peak flows during a wet period are mostly generated by the
hillslope  Qfw,wet/Qfh,wet<1
-Annual runoff coefficient  0.2<(Q/P) & (Q/P)<0.45 (Budyko)
FLEXC
P
I Plateau
(40%)
T
P
I
P
Hillslope
(45%)
T
Ru
I Wetland
(15%)
T
Ru
Rf
Ru
Qfh
RP
Rs
RP
Rs
Rf
Qfw
RC
RP
Qs
Then how to parameterize?!
Parameter
space
Satisfies 2
or more constraints
Satisfies N-1
or more constraints
Satisfies 1
or more constraints
Parameter space
which satisfies all
the constraints but
is not optimal
calibrated on
observed
hydrograph
Un-calibrated
but
constrained
Satisfies N
or more constraints! (All)
Un-calibrated but constrained
NS = -0.87
NS = 0.64
NS = 0.71
Then how to parameterize?!
Parameter
space
Satisfies 2
or more constraints
Satisfies N-1
or more constraints
Satisfies 1
or more constraints
Calibrated
and
constrained
Pareto
optimal
Un-calibrated
but
constrained
Satisfies N
or more constraints! (All)
Calibrated and constrained
NS = 0.91
NS = 0.87
NS = 0.88
Validated for calibrated and constrained
NS = 0.75
NS = 0.80
NS = 0.80
Comparison & conclusion &
discussion!
FLEX-TOPO case study in Heihe, China
But answering my research questions!
Questions…
Answers…
How to see the hydrological
Snark…?
How far is the hydrological
Snark…?
How big is the hydrological
Snark…?
How does the hydrological
Snark look like…?
To be honest! I don’t know ...
But the hydrological Snark
might not be that “far away” or
that “close”…
It might not be that “big” or
that “small”…
It might not be that “ugly” or
that “beautiful”…
These are just my feelings!
But answering my research questions!
Questions…
Answer…
Are we in love with the
hydrological Snark?
Or…
Are we in love with the
instruments we use to find the
hydrological Snark?
Seems dangerous to answer!
Thank you for Your Attention