calibration and sensitivity analysis of swat for a small forested … · 2011. 6. 16. ·...
TRANSCRIPT
UNIÃO EUROPEIA Fundo social Europeu
Rial-Rivas, M.E.1; Santos, J. 1;Bernard-Jannin L. 1 ; Boulet, A.K. 1 ; Coelho, C.O.A. 1; Ferreira, A.J.D.4; Nunes, J.P. 1 ; Rodríguez-Suárez J.A. 1, 2; Rodríguez-Blanco M.L. 1,3;
Keizer, J.J. 1
Calibration and sensitivity analysis of SWAT for a small forested catchment, north-
central Portugal
2011 International SWAT Conference- June 15-17 Toledo, Spain
1CESAM and Dept. Environment & Planning. University of Aveiro. PORTUGAL E-mail: [email protected]
2Vegetal Biology and Soil Science Department. University of Vigo. SPAIN. 3 Faculty of Sciences. University of A Coruña. SPAIN.
4CERNAS, Coimbra Agrarian Technical School. PORTUGAL
HIDRIA project
UNIÃO EUROPEIA Fundo social Europeu
HIDRIA project
“A multi-stage approach for addressing input data uncertainties in
process-based rainfall-runoff modeling for small forested
catchments upstream of the Ria de Aveiro”
The project foresees the development of:
A stepwise approach to rainfall-runoff modelling
To assess the implications of
existing data constraints
To establish priorities for additional field
and laboratory data gathering
UNIÃO EUROPEIA Fundo social Europeu
SPECIFIC OBJECTIVE WITHIN THIS WORK
Assess the influence of:
Nash-Sutcliffe efficiency (NSE)
Percent bias (PBIAS)
Root Mean Square Error (RMSE)
METHODS
MODEL EVALUATION
SENSITIVITY ANALYSIS
(ArcSWAT 2009 interface)
AUTO-CALIBRATION
(ArcSWAT 2009 interface)
Latin Hypercube (LH) and One-factor-At-a-Time (OAT)
sampling.
Parameter Solution (ParaSol)
with uncertainty analysis
Ranges of variation of these
parameters
in the SWAT model auto-calibration results for the study catchment.
Number of parameters included in
the auto-calibration.
HIDRIA project
STUDY AREA
HIDRIA project
Serra de Cima
Rainfall:
1000-2500 mm/yr
4 micro-catchments
Area <1 km2
Caramulo mountain
range
STUDY AREA
HIDRIA project
HYDROLOGICAL MODELLING: Input Data
LAND-USE SOIL-TYPES
Data widely available in Portugal (e.g. to
hydrologists from a consultancy company developing a
Watershed Management Plan)
CLIMATE
HYDROLOGICAL MODELLING: Input Data
ArcSWAT 2009
HUMIC CAMBISOL
ELEVATIONS
•Rainfall
•Temperature
•Relative Humidity
•Wind velocity
•Solar radiation
DAILY
•Temperature
•Rainfall
IM: Coimbra_G
WEATHER
GENERATOR
Study area
HYDROLOGICAL MODELLING: CLIMATE Input Data
CLC06: Transitional woodland-shrub 6.58% total area
Catchment area: 0.53 km2
CORINE LAND-COVER 2006
Young Eucalypt
plantations
HYDROLOGICAL MODELLING: LAND-USE Input Data
CLC06: Broad-leaved forest: 93.42% total area
Eucalypt
Forest
HYDROLOGICAL MODELLING: Simulated period
13.2
13.4
13.6
13.8
14
14.2
14.4
14.6
14.8
15
15.2
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Tem
pera
ture
, ºC
Rain
fall, m
m/y
r
Rainfall, mm/yr Mean Temp, ºC
HYDROLOGICAL MODELLING: Simulated period Warm-up period Study period
Mean Rainfall = 1245 mm/yr
Calibration period Validation period
0
30
60
90
120
1/1/09 3/2/09 5/1/09 6/30/09 8/29/09 10/28/09 12/27/09
Daily
Str
eam
flow
, m
m
0
30
60
90
120
150
180
Daily
Rain
fall,
mm
0
30
60
90
120
01/01/10 03/02/10 05/01/10 06/30/10 08/29/10 10/28/10 12/27/10
Daily
Str
eam
flow
, m
m
0
30
60
90
120
150
180
Daily
Rain
fall,
mm
Calibration period Validation period
HYDROLOGICAL MODELLING: Simulated period
PBouça = 1585 mm
Qmm = 1004 mm
Q max obs = 47.9 mm
Q min obs = 0 mm
PBouça = 1367 mm
Qmm = 671 mm
Q max obs = 19.9 mm
Q min obs = 0 mm
HYDROLOGICAL MODELLING: Sensitivity Analysis
HYDROLOGICAL MODELLING: Sensitivity Analysis
26 flow-related parameters Ranking
Mean
PARAMETER LO BOUND UP BOUND iMet
Alpha_Bf 0 1 1
Biomix 0 1 1
Blai 0 1 1
Canmx 0 10 1
Ch_K2 0 150 1
Ch_N2 0 1 1
Cn2 -25 25 3
Epco 0 1 1
Esco 0 1 1
Gw_Delay 0.001 10 2
Gw_Revap 0.001 0.036 2
Gwqmn 0.001 1000 2
Revapmn 0.001 100 2
Sftmp 0 5 1
Slope -25 25 3
Slsubbsn -25 25 3
Smfmn 0 10 1
Smfmx 0 10 1
Smtmp -25 25 3
Sol_Alb -25 25 3
Sol_Awc -25 25 3
Sol_K -25 25 3
Sol_Z -25 25 3
Surlag 0 10 1
Timp 0 1 1
Tlaps 0 50 1
INPUT OUTPUT
0
5
10
15
20
25
Alp
ha
_B
f
Bio
mix
Bla
i
Ca
nm
x
Ch
_K
2
Ch
_N
2
Cn
2
Ep
co
Esco
Gw
_D
ela
y
Gw
_R
eva
p
Gw
qm
n
Re
va
pm
n
Sft
mp
Slo
pe
Sls
ub
bsn
Sm
fmn
Sm
fmx
Sm
tmp
So
l_A
lb
So
l_A
wc
So
l_K
So
l_Z
Su
rla
g
Tim
p
Tla
ps
With observed data
Without observed data
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
Alp
ha
_B
f
Bio
mix
Bla
i
Ca
nm
x
Ch
_K
2
Ch
_N
2
Cn
2
Ep
co
Esco
Gw
_D
ela
y
Gw
_R
eva
p
Gw
qm
n
Re
va
pm
n
Sft
mp
Slo
pe
Sls
ub
bsn
Sm
fmn
Sm
fmx
Sm
tmp
So
l_A
lb
So
l_A
wc
So
l_K
So
l_Z
Su
rla
g
Tim
p
Tla
ps
With observed data
Without observed data
HYDROLOGICAL MODELLING: Sensitivity Analysis 26 flow-related parameters
Ranking
Mean
OUTPUT
0
5
10
15
20
25
Alp
ha
_B
f
Bio
mix
Bla
i
Ca
nm
x
Ch
_K
2
Ch
_N
2
Cn
2
Ep
co
Esco
Gw
_D
ela
y
Gw
_R
eva
p
Gw
qm
n
Re
va
pm
n
Sft
mp
Slo
pe
Sls
ub
bsn
Sm
fmn
Sm
fmx
Sm
tmp
So
l_A
lb
So
l_A
wc
So
l_K
So
l_Z
Su
rla
g
Tim
p
Tla
ps
With observed data
Without observed data
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
Alp
ha
_B
f
Bio
mix
Bla
i
Ca
nm
x
Ch
_K
2
Ch
_N
2
Cn
2
Ep
co
Esco
Gw
_D
ela
y
Gw
_R
eva
p
Gw
qm
n
Re
va
pm
n
Sft
mp
Slo
pe
Sls
ub
bsn
Sm
fmn
Sm
fmx
Sm
tmp
So
l_A
lb
So
l_A
wc
So
l_K
So
l_Z
Su
rla
g
Tim
p
Tla
ps
With observed data
Without observed data
WITHOUT WITH
OBS DATA OBS DATA
Alpha_Bf Esco
Cn2 Gwqmn
Esco Sol_Awc
Canmx Canmx
Blai Cn2
Surlag Sol_Z
Ch_K2 Gw_Revap
Gw_Revap Blai
Gw_Delay Alpha_Bf
Sol_Z Sol_K
Ch_N2 Ch_K2
Sol_Awc Biomix
Sol_K Slope
Slope Epco
Biomix Revapmn
Slsubbsn Gw_Delay
Epco Surlag
Gwqmn Ch_N2
Sol_Alb Sol_Alb
Revapmn Slsubbsn
Sftmp Sftmp
Smfmn Smfmn
Smfmx Smfmx
Smtmp Smtmp
Timp Timp
Tlaps Tlaps
HYDROLOGICAL MODELLING: Auto-calibration
Three auto-calibrations were carried out using:
Auto-calibration A
26 flow-related
parameters and full
default ranges of
variation
HYDROLOGICAL MODELLING: Auto-calibration
Auto-calibration C Auto-calibration B
13 flow-related
parameters and full
default ranges of
variation
13 flow related
parameters and
narrow ranges of
variation
The 13 most sensitive parameters selected from
Sensitivity Analysis using observed data.
Upper and lower bounds for narrow ranges using
max and min values from the good parameter
sets from Auto-calibration A.
• ParaSol with uncertainty analysis
• Fixing the number of simulations runs in 20000 and the optimization settings as the default.
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Calibration period
0
30
60
90
1200
30
60
90
120
1/1/2009 2/1/2009 3/1/2009 4/1/2009 5/1/2009 6/1/2009 7/1/2009 8/1/2009 9/1/2009 10/1/2009 11/1/2009 12/1/2009
Daily
Rain
fall,
mm
Daily
Str
eam
flo
w, m
m
Daily Rainfall, mm
Daily Observed Flows, mm (Calibration period)
Auto-calibration A: Daily Simulated Flows, mm (Calibration period)
Auto-calibration B: Daily Simulated Flows, mm (Calibration period)
Auto-calibration C: Daily Simulated Flows, mm (Calibration period)
Auto-calibration A Auto-calibration B Auto-calibration C
PBIAS 9.84 16.05 9.75
RMSE 4.14 4.11 4.37
NSE 0.62 0.63 0.58
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Calibration period
SD=6.72
Satisfactory
Very good
0
30
60
90
1200
30
60
90
120
1/1/2009 2/1/2009 3/1/2009 4/1/2009 5/1/2009 6/1/2009 7/1/2009 8/1/2009 9/1/2009 10/1/2009 11/1/2009 12/1/2009
Daily
Rain
fall,
mm
Daily
Str
eam
flo
w, m
m
Daily Rainfall, mm
Daily Observed Flows, mm (Calibration period)
Auto-calibration A: Daily Simulated Flows, mm (Calibration period)
Auto-calibration B: Daily Simulated Flows, mm (Calibration period)
Auto-calibration C: Daily Simulated Flows, mm (Calibration period)
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Validation period
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Validation period
0
20
40
60
80
100
120
140
160
180
2000
20
40
60
01/01/10 01/31/10 03/02/10 04/01/10 05/01/10 05/31/10 06/30/10 07/30/10 08/29/10 09/28/10 10/28/10 11/27/10 12/27/10
Daily
Rain
fall,
mm
Daily
Str
eam
flo
w, m
m
Daily Rainfall, mm
Daily Observed Flows, mm (Validation period)
Auto-calibration A: Daily Simulated Flows, mm (Validation period)
Auto-calibration B: Daily Simulated Flows, mm (Validation period)
Auto-calibration C: Daily Simulated Flows, mm (Validation period)
Auto-calibration A Auto-calibration B Auto-calibration C
PBIAS -23.6 -19.81 -21.64
RMSE 1.5 1.92 1.69
NSE 0.74 0.55 0.66
0
20
40
60
80
100
120
140
160
180
2000
20
40
60
01/01/10 01/31/10 03/02/10 04/01/10 05/01/10 05/31/10 06/30/10 07/30/10 08/29/10 09/28/10 10/28/10 11/27/10 12/27/10
Daily
Rain
fall,
mm
Daily
Str
eam
flo
w, m
m
Daily Rainfall, mm
Daily Observed Flows, mm (Validation period)
Auto-calibration A: Daily Simulated Flows, mm (Validation period)
Auto-calibration B: Daily Simulated Flows, mm (Validation period)
Auto-calibration C: Daily Simulated Flows, mm (Validation period)
SD=2.83
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Comparison between best parameter sets for each auto-calibration
HYDROLOGICAL MODELLING: Auto-calibration RESULTS
Auto-calibration A Auto-calibration B Auto-calibration C
Alpha_Bf 0.269 0.562 0.195
Biomix 0.677 0.218 0.180
Blai 0.156 0.061 0.059
Canmx 4.412 7.327 7.421
Ch_K2 127.6 106.9 126.4
Cn2 50-95 56-100 51-97
Esco 0.131 0.198 0.139
Gw_Revap 0.050 2.009 0.060
Gwqmn 933 962 1858
Sol_Awc1 0.169 0.203 0.193
Sol_Awc2 0.130 0.157 0.149
Sol_K 3.46 4.28 4.15
Sol_K2 3.41 4.22 4.09
Sol_Z1 335 358 334
Sol_Z2 1342 1432 1337
Groundwater “revap” coefficient
Afects the amount of water that
recharges the capillary fringe after
evaporation during the dry periods.
Depth from soil surface to bottom layer
Threshold depth of water in the
shallow aquifer required for the
return flow to occur (The ground
water flow to the main channel is
allowed only when the depth of water in
the shallow aquifer is equal to or greater
than
Baseflow recesion coefficient is a direct
index of groundwater flow response to
changes in recharge
Comparison between best parameter sets for each auto-calibration
Canopy Interception and Max. Potencial
LAI
1.- Check the feasibility of the obtained parameters
2.- Check other auto-calibration methodologies and with different objective functions.
3.- Testing SWAT with data obtained from a meteorological station in the study area as well as
from fieldwork in the Serra de Cima catchment, aiming at improving model results and
decrease problems related with equifinality.
CONCLUSIONS
The best results were obtained for the set with the largest number of
parameters and the widest ranges of variation.
Sensitivity analysis was helpful in reducing the number of parameters
included in the auto-calibration and, auto-calibration time, without
seriously affecting model results.
The use of narrow ranges of variation for the parameters also reduced the
time needed for auto-calibration whilst still producing results that can be
regarded adequate, especially for general-purpose studies.
The fact that several parameter sets have given good results, indicating a
problem with equifinality of model parameterization.
ONGOING WORK
UNIÃO EUROPEIA Fundo social Europeu
Thanks for your attention