Evaluation of SWAT streamflow components for

the Araxisi catchment(Sardinia, Italy)

Maria Grazia Badas1*, Mauro Sulis2, Roberto Deidda1, Enrico Piga1,

Marino Marrocu2 and Claudio Paniconi3

1. Dipartimento di Ingegneria del Territorio, Università di Cagliari, Italy.2. CRS4 (Center for Advanced Studies, Research and Development in

Sardinia), Cagliari, Italy3. INRS-ETE (Institut National de la Recherche Scientifique – Eau, Terre et

Environnement), Université du Québec, Canada

Outline

• Aims of the study• Description of the watershed and

dataset• Baseflow separation• Yearly calibration• Daily calibration• Conclusions

Aim of the study

• A classification made by the Sardinian Hydrological Survey, designated the basin as almost impermeable

• Pre-processing of measured data has suggested that this proposed classification has to be interpreted as absence of water loss to deep aquifer recharge

• Comparison between separation methods (related to response times), and streamflow components provided by the model (obtained from physically based equations)A reliable simulation of streamflow components is an important step in the adoption of a correct schematization of watershed characteristics

Case study: the Araxisi catchment

Overview of the watershed• Sardinian mountain

basin• Area=125km2

• Averageelevation=804 m

• Average steepness=30%

• DEM 100 m• 41 sub-basins (tr.

area=200 ha)

%[

%[%[

%[%[

$Z

$Z

']

&\

D e m

2 6 5 - 4 9 8

4 9 8 - 7 3 2

7 3 2 - 9 6 6

9 6 6 - 1 2 0 0

1 2 0 0 - 1 4 3 4

1 4 3 4 - 1 6 6 8

W a t e r s h e d

S t r e a m s

%[ R a i n g a g e s

$Z T e m p g a g e s

'] S o l a r g a g e s

&\ W e a g a g e s

Model Dataset1946-1975

• daily values of precipitation (5 gages), and maximum and minimum temperatures (2 gages ) supplied by the Sardinian Hydrological Survey

• solar radiation daily values from NCEP-NCAR analyses (National Centers for Environmental Prediction and for Atmospheric Research).

• daily discharge values supplied by the Sardinian Hydrological Survey

W a t e r s h e d

S w a t L a n d U s e C l a s s

A G R L

A P P L

F R S D

F R S E

F R S T

P A S T

U I D U

U R L D

U T R N

W A T R

S t r e a m s

Land Use Classification• obtained by a

satellite image with a resolutionof 400 m

• Prevalent land use classes:– 36% evergreen forest

(FRSE)– 26% mixed forest

(FRST) – 28% pasture (PAST)

Soil Classification• lacking of a detailed soil

map…

• main soil characteristicsfor the whole basin: – Clay 5%– Silt 25%– Sand 70%

•sandy loam soil (SL) according to the USDA soil texture triangle classification•soil stratification: two different configurations(single and multiple layers)

STREAMFLOW ASSESSMENT

• hydrograph components: surface runoff, subsurface flow and baseflow

• practically only two streamflow components are recognized: quick flow and recession flow, on the basis of response times, without any reference tothe underlying physical processes

Application of two separation techniques to the daily records for

baseflow estimation

1. A classical separation technique:• Qb=Qtot during interstorm periods;• logQb=linear trend during storm periods

2. A digital filter technique• proposed by the SWAT developers based on

the filtering of high and low frequency signals(surface runoff and baseflow)

• three passes: forward, backward, forward, with a decreasing baseflow rate at each pass[Arnold et al. (1995) and Arnold & Allen 1999)]

115 120 125 130 135 1400

2

4

6

8

10

12

observed rainfall observed streamflow manual baseflow filtered baseflow (forward) filtered baseflow (backward)filtered baseflow (forward)

Comparison of “classical” separation and digital filter technique

the first pass provides results comparable to those of the classical technique at an annual time step

1 2 3 4 5 6 7 8 9 10100

200

300

400

500

600

700

800

900

t (years)

observed streamflow manual baseflow filtered baseflow (forward) filtered baseflow (backward)filtered baseflow (forward)

Daily step Annual step

Baseflow recession factor (α)

• Manual technique: α = the average slope of linear trend of logarithmic streamflowswithin interstorm periods = 0.06 on a sample of 33 manual selected recession periods

• Digital filter method: it considers streamflow separation given by the filter, baseflow recession slopes computed only in low ET months and combined with the Master Recession Curve (MRC) method;α = 0.0143 calculated on 18 events

results are quite different!!

tt eQQ α−= 0

1 2 3 4 5 6 7 8 9 100

200

400

600

800

1000

1200

1400

1600

t(years)

observed rainfall observed streamflow simulated streamflowmanual baseflow filtered baseflow simulated baseflow lateral flow

Yearly calibration (1)(on the first decade of streamflow data)

Many likely sets of

parameters were

hypothesized,

but none of them led

to a realistic

separation.

EXAMPLE(RESULTS FOR ONE SIMULATION)

baseflow component practically absent in all the simulations!!!

Simulated baseflow

Observedrainfall

Simulated streamflow

Simulated lateral flow

Observed streamflow

Classical tech. baseflow

Filtered baseflow (I pass)

Yearly calibration (2)• incorrect calculation of the slope length

parameter (Lhill) in the AVSWAT interface

• lateral flow (kinematic wave approximation) is in inverse proportion to Lhill

• a reliable value is Lhill=50 m

• but the AVSWAT interface fixes Lhill= 0.05 m

overestimation of the lateral flowshortage of available soil water for groundwater recharge

1 2 3 4 5 6 7 8 9 100

200

400

600

800

1000

1200

1400

1600

t(years)

observed rainfall observed streamflow simulated streamflowmanual baseflow filtered baseflow simulated baseflow lateral flow

Yearly calibration (3)Correcting this parameter to50 m led to a considerable improvementin the streamflow separation

Lhill=50 m

Simulated baseflow

Observedrainfall

Simulated streamflow

Simulated lateral flow

Observed streamflow

Classical tech. baseflow

Filtered baseflow (I pass)

EXAMPLE(RESULTS FOR ONE SIMULATION)

1 2 3 4 5 6 7 8 9 10200

400

600

800

1000

1200

1400

1600

1800

2000

t (years)

Sensitivity to rainfall data

spatial distributionof the input rainfall:an important influence on the model response fora mountain basin

five gages could be unable to correctly reproduce the actual rainfall patterns

Observedrainfall (5 gages)

Simulated streamflow (5 gages)

Observed streamflow

Observedrainfall (1 gage)

Simulated streamflow (1 gage)

Daily calibration

superimposition of

– spikes during

storms

– smooth behaviour

during interstorm

periods

3200 3220 3240 3260 3280 3300 3320 3340 3360 3380 34000

2

4

6

8

10

12

14

16

18

20

Observed streamflow

Observed rainfall

Filtered baseflow (I pass)

Simulated streamflow

Simulated baseflow

Classical tech.baseflow

Conclusions and future work• Comparison between “classical” and digital

filter technique– Baseflow estimation– Baseflow recession factor (α)

• Incorrect calculation of Lhill in AVSWAT interface

• Work in progress…– Daily calibration– Validation of the model– Application of other separation techniques