Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Interception
P
SnowP
Soil
moistu
reS
Soilmoist
ure
ETDirectruno↵Q
Evapo-transp.ET
InterflowS
Perc
ola
tion
S
Routing
Q
Geolog
yS
0
0.45
2
61014
1822
2630
3438
42
4650
Sobol indexSi ST i � Si
Inter-‐ cep*on
rou*
ng
perco-‐
la*on
soil moisture
storage
run-‐ off
inter-‐ cep*on
6. Calibra*on with OF3 using ESA-‐CCI soil moisture
2. Study domain and mesoscale Hydrologic Model (mHM) mHM [1] is a distributed model which treats grid cells as hydro-‐logic units. mHM’s key feature is the Mul*scale Parameter Regio-‐naliza*on technique (MPR).
3. Calibra*on objec*ves
Towards constraining hydrologic models using satellite retrieved soil moisture
1. Mo*va*on Hydrological models are usually calibrated against observed discharge at the catchment outlet and thus are condi*oned by an integral catchment informa*on. This procedure does not take into account any spa*o-‐temporal variability of fluxes or state variables and can lead to uncertain*es in model internal states as e.g. soil moisture (SM). Satellite data may help to beMer constrain model parameters. The first objec*ve of this study is to calibrate a hydrological model with synthe*c soil moisture data to inves*gate the skill of different objec*ve func*ons (OF) for recovering model parameters. The different approaches aim either for spa*al, for temporal or for spa*o-‐temporal matching of the two paMerns. A second objec*ve is to calibrate the hydrologic model with ESA-‐CCI satellite soil moisture using the superior OF.
MaIhias Zink, Juliane Mai, Oldrich Rakovec, Mar*n Schrön, Rohini Kumar, David Schäfer, and Luis Samaniego
Helmholtz Centre for Environmental Research – UFZ, Leipzig, Germany ([email protected])
7. Conclusions • SM-‐sensi*ve parameters should be considered if deciding for an appropriate OF. • Objec*ve func*ons aiming on temporal paMerns of SM almost recover SM-‐sensi*ve parameters. • SM is not sufficient to constrain model parameters for discharge predic*on.
References [1] Samaniego, L., Kumar, R., & Attinger, S. (2010). Multiscale parameter regionalization of a grid-based hydrologic model at the mesoscale. Water Resources Research, 46(5)
www.ufz.de/mhm
Two dis*nct European basins a – Neckar (humid climate) b – Guadalquivir (semi-‐arid climate)
Euclidian parameter distances
Discharge predic*on
b – Guadalquivir catchment a – Neckar catchment
2000 2001 2002 2003 20040200400600800
100012001400
Dis
char
geQ
[m3
s�1 ]
Qobs
Qsim
Qre f
NSE = 0.14
2000 2001 2002 2003 20040200400600800
100012001400
Dis
char
geQ
[m3
s�1 ]
NSE = 0.14
1955 1956 1957 1958 1959050
100150200250300350400
Dis
char
geQ
[m3
s�1 ]
NSE = 0.33
1.72.02.32.6
Distance for all parametersDistance for all parametersDistance for all parametersDistance for all parameters
0 5000 10000 15000No. of Iterations
0.00.30.60.9 Distance for SM-sensitive parametersDistance for SM-sensitive parametersDistance for SM-sensitive parametersDistance for SM-sensitive parameters
OF1 OF2 OF3 OF4
1.72.02.32.6
Distance for all parametersDistance for all parametersDistance for all parametersDistance for all parameters
0 5000 10000 15000No. of Iterations
0.00.30.60.9 Distance for SM-sensitive parametersDistance for SM-sensitive parametersDistance for SM-sensitive parametersDistance for SM-sensitive parameters
OF1 OF2 OF3 OF4
�4�W �3�W37�N
38�N
reference OF1
�4�W �3�W37�N
38�N
OF2
�4�W �3�W
OF3
�4�W �3�W
OF4
�4�W �3�W
0.5 0.6 0.7 [mm mm�1]
[%]�15�10 �5 0 5 10 15
48�N
49�N
OF1
9�E 10�E
referenceOF2
9�E 10�E48�N
49�N
OF3
9�E 10�E
OF4
9�E 10�E 9�E 10�E
0.6 0.7 0.8 0.9 [mm mm�1]
[%]�5 �3 �1 1 3 5
Rel. difference to reference
field
0.2 0.5 0.8
0.20.50.8
G(y
)
rs=0.99802
0.2 0.5 0.8
rs=0.96456
0.2 0.5 0.8
rs=0.99973
0.2 0.5 0.8F(x)
rs=0.99972
0.0 0.2 0.4 0.6 0.8 1.0
0.2 0.5 0.8F(x)
0.20.50.8
G(y
)
rs=0.99973
0.2 0.5 0.8
rs=0.99860
0.2 0.5 0.8
rs=0.99970
0.2 0.5 0.8
rs=0.99979
0.0 0.2 0.4 0.6 0.8 1.0
Copula between reference field
and OFs
OF1 – KGE of catchment average SM (spa*o-‐temporal paMern)
OF2 – paMern similarity (spa*al paMern)
OF3 – sum squared errors of SM anomaly
(temporal paMern)
OF4 – spa*al average of temporal corre-‐ la*on (temporal paMern)
-1 6 5
13 7 14
11 10 11 12
Pattern A
7 9
13 7 5
12 11 13 11
Pattern B
A·B+1= 12
-1 1
0
2
-1 -1 -1 -1
1 1
1
-1 1 1 1
1 -1 2 2
0 2
2 16 2
Æ 0.75
SM SON 1997
The model parameters which are sensi*ve to soil moisture are determined due to a sensi*vity analysis.
5. Determining the appropriate objec*ve func*on (synthe*c dataset)
4. SM-‐sensi*ve parameters
Sobol indexSi ST i � Si