application of a coupled swat-bathtub model to evaluate
TRANSCRIPT
1
Application of a coupled SWAT-BATHTUB model to
evaluate phosphorus critical source areas and land
management alternatives on the water quality of Lake
Prespa, Macedonia
24 June 2015
Michael Winchell, Naresh Pai Stone Environmental, Inc.
Dave Dilks LimnoTech
Nikola Zdraveski United Nations Development Programme (UNDP)
Presented at the 2015 international SWAT conference, Sardinia, Italy
2
Presentation Outline
Study area
Study objectives
SWAT model:
• Development
• Calibration and Validation
BATHTUB model
• Background and development
• Simulation results
Phosphorus critical source area (CSA) analysis
Alternative management practices (preliminary results)
Conclusions and next steps
3
Study Area: Lake Prespa
You
are
here!
International:
• Macedonia
• Albania
• Greece
Watershed area:
• Land: 1,054 km2
• Lake: 305 km2
Elevation:
• Min: 823 m
• Max: 2,420 m
4
Study Area: Lake Water Quality Status
Based on 2014 monitoring data,
Prespa lake can be classified as:
• eutrophic based on TP
• mesotrophic based on
secchi depth and Chl-A
concentrations
5
Study Objectives
Develop a coupled SWAT/BATHTUB model
Identify phosphorus CSAs
Develop and evaluate alternative management practices
Determine impact of alternatives on lake water quality
Compile knowledge gained into a management tool for UNDP to guide future
activities in the watershed
6
SWAT Development: Subbasin/HRU Delineation Data
UNDP (20 m) and
ASTER (30 m) stitched to
produce 20 m DEM
DEM (m)
825 - 1,041
1,042 - 1,301
1,302 - 1,573
1,574 - 1,865
1,866 - 2,442
AGRL
AGRR
APPL
BARR
FRSD
FRST
ORCD
PAST
RNGB
RNGE
UCOM
URLD
URML
WATR
WETL
CORINE 2000/2006 (100 m) and vector orchard boundaries stitched and resampled to produce 20 m land use
ESDB (1 km), FAO (10
km) and 32 UNDP field
samples integrated to
produce soils data
7
SWAT Development: Subbasin/HRU Delineation Results
Land Use:
• 80% undeveloped
• 19% agriculture
• 1% developed
Soils: 48 soils classes, hydro group C dominant
Slope: 5 classes
• 0 – 3%: 10%
• 3 – 15%: 21%
• 15 – 25%: 17%
• 25 – 50%: 38%
• > 50%: 13%
HRU Delineation: No thresholds applied
• 126 subbasins
• 5,061 HRUs total
8
SWAT Model Development: Weather and Agronomy
Daily precipitation and temperature :
• NCEP CFSR, 1979-2013, 4 stations
• UNDP 2006-2014, 2 stations
Lapse rates generated from long term
isohyetal and temperature maps
Agronomic management schedules:
• Apples, wheat, potato (Macedonia)
• Corn, lima bean (Greece, Albania)
• Timing and frequency of planting,
harvesting, tillage, irrigation, and fertilizer
applications derived by local agronomist.
9
SWAT Calibration: Streamflow data
Monthly streamflow: 1983 – 2010
• Brajcinska River at Brajcino
• Golema River near Resen
(estimated)
Model warmup: 1979-1982
Calibration: 1997-2010
Validation: 1983 – 1996
Model performance evaluated based on
guidelines by Moriasi et al. (2007)
Golema
River near
Resen
Brajcinska
River at
BrajcinoMoriasi, D. N., J. G. Arnold, M. W. Van Liew, R. L. Bingner, R. D. Harmel and T. L. Veith. 2007.
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE 50(3):885-900.
10
SWAT Calibration: Example Streamflow Results
0
2
4
6
8
10
121
/1/1
99
7
9/1
/19
97
5/1
/1998
1/1
/1999
9/1
/19
99
5/1
/20
00
1/1
/20
01
9/1
/20
01
5/1
/20
02
1/1
/2003
9/1
/2003
5/1
/20
04
1/1
/20
05
9/1
/20
05
5/1
/20
06
1/1
/20
07
9/1
/2007
5/1
/2008
1/1
/20
09
9/1
/20
09
5/1
/20
10
Mo
nth
ly S
trea
mfl
ow
(m
3/s
)
Observed Simulated
Brajcinska River at Brajcino
PBIAS -2%, NSE 0.65, RSR 0.59
Based on performance evaluation criteria, PBIAS is “very good”, NSE is
“satisfactory”, and RSR is “good”
11
SWAT Validation: Example Streamflow Results
Based on performance evaluation criteria, PBIAS is “very good”, NSE and RSR
are “satisfactory”
0
2
4
6
8
10
121
/1/1
983
9/1
/19
83
5/1
/19
84
1/1
/1985
9/1
/19
85
5/1
/1986
1/1
/19
87
9/1
/1987
5/1
/19
88
1/1
/19
89
9/1
/19
89
5/1
/19
90
1/1
/1991
9/1
/19
91
5/1
/1992
1/1
/19
93
9/1
/1993
5/1
/19
94
1/1
/19
95
9/1
/1995
5/1
/19
96
Mo
nth
ly S
trea
mfl
ow
(m
3/s
) Brajcinska River at Brajcino
PBIAS -8%, NSE 0.59, RSR 0.64
12
SWAT Calibration: Water Quality Data
Monitoring data: 8 sites, 12/2013-
12/2014 (continuing in 2015)
Between 6 and 11 samples per site
Majority of samples during low flows
Calibration Strategy:
• Qualitative comparison of
observed concentration
distributions with simulated
distributions (same flow rate
range)
• Watershed-level calibration,
uniform parameter adjustmentsWater
quality
stations
14
SWAT Model Calibration: Example Water Quality
Results, 3 of 8 Sites
Sediment and TP simulations very close to observed data. Over predictions in the
highest percentiles likely reflect lack of monitoring data during high flows
0.00
0.05
0.10
0.15
0 0.2 0.4 0.6 0.8 1
Sed
imen
t (g
/L) Observed
Simulated
0
0.05
0.1
0.15
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
0 0.2 0.4 0.6 0.8 1
0
200
400
600
800
0 0.2 0.4 0.6 0.8 1
Tota
l P
hosp
horu
s
(ug/L
)
Percentile
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1Percentile
0
200
400
600
800
0 0.2 0.4 0.6 0.8 1Percentile
SP1 SP2 SP3
15
SWAT Model Calibration: Landscape Analysis, Plant
Growth
0
5
10
15
20
25
Jan
Feb
Mar
Apr
May Jun
Jul
Aug
Sep
Oct
Nov
Dec
Bio
mass
(t/
ha) Apple
harvest
Sep 30pruning
Mar 3041
42
43
44
45
46
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Deciduous Forest
43
44
45
46
47
48
49
50
Jan
FebMar
Apr
MayJunJul
Aug
SepOct
Nov
Dec
Mixed Forest
Seasonal biomass growth pattern considered reasonable by a local agronomist
3.0
4.0
5.0
6.0
7.0
8.0
9.0
Jan
FebMar
Apr
MayJunJul
Aug
SepOct
Nov
Dec
Pasture
0
2
4
6
8
10
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Bio
mass
(t/
ha) Potato
harvest
Aug 15planting
Apr 15
0
2
4
6
8
10
12
14Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Winter Wheat
harvest
July 15
planting
Nov 1
16
SWAT Model Calibration: Landscape Analysis, Total P by
Soil Hydrologic Group
A B C D
Apple 1.24 1.33 1.53 1.48
Barren 0.34 NA NA NA
Urban 0.13 0.15 0.28 0.57
Corn NA NA 10.31 NA
Forest Deciduous 0.09 0.09 0.11 0.06
Forest Generic 0.09 0.09 0.13 NA
Lima NA NA 8.21 NA
Pasture 0.09 0.10 0.13 0.06
Potatoes 2.01 4.27 6.26 1.57
Range Brush 0.11 0.10 0.16 NA
Range Grass 0.11 0.12 0.18 NA
Wetlands 0.09 0.10 0.13 0.11
Winter Wheat 0.79 1.79 2.66 0.80
Soil Hydrologic Group
Forest and wetlands
generate lowest total P
Loads from urban areas
higher than undeveloped
Highest loads from heavily
cultivated agriculture (corn,
lima, potato)
P load generally increases
from A to C soils
D soils, a very small fraction
of the watershed (0.5%) did
not follow expected trend.
17
SWAT-BATHTUB Coupling: BATHTUB Background
Steady-state model
A combination of mechanistic and empirical
sub-models
Empirical equations based on 2.5 million
observations from 271 lakes
Model outputs include:
• Total phosphorus
• Total nitrogen
• Chl-A
• Transparency
• Hypolimnetic oxygen demand
0.001
0.01
0.1
1
10
100
1000
0.01 0.1 1 10 100 1000
Total Phosphorus
Ch
loro
ph
yll
a
18
SWAT-BATHTUB Coupling: BATHTUB Results
BATHTUB simulation for Lake Prespa run
with 2013 -2014 SWAT P load as input
Model calibrated to 2014 monitoring data
Results showed Total P, Chl-A, and Secchi
depth observations within range of simulation
20
CSA Analysis: Total P Load by Land Use
Corn and lima beans contribute highest total loads, followed by winter wheat potato,
and apples
267.3
212
.8
109.3
86.1
72
.8
45.7
41.1
19.6
3.1
2.6
2.1
1.9
0.4
0
50
100
150
200
250
300
350
Tota
l P
(kg/y
r) x
10
0
21
CSA Analysis: Cumulative Watershed P Load
22
Alternative Management Practices
Alternative management practices currently being considered include:
• Reduced irrigation in orchards
• Fertilizer best practices (placement and rate)
• Apple orchard waste management
• Wetland restoration (point source and non point source P)
• Erosion control from agricultural land
23
Alternative Management Practices: Irrigation Code
Modification
Current SWAT Code: Irrigation depth
capped to field capacity when using
outside source, resulting in lower than
intended irrigation amount
Revised SWAT Code: Set to use allow
user-specified depth (regardless of field
capacity), resulting in intended irrigation
amount
Code change impacted overall nutrient
flux
Surface runoff from irrigation events
currently a user-defined fraction of
irrigation, not based on a mechanistic or
empirical model
irrsub.f
revision
24
Alternative Management Practices: Orchard Irrigation
Reduction
Current Irrigation Practice: 864 mm/year,
simulated as manual irrigation in SWAT
Based on a UNDP soil sampling,
recommendation is to apply either 318 mm
(drip) to 454 mm (surface) of irrigation,
simulated as auto-irrigation in SWAT
Based on a 30-year simulation under
alternative practice:
• Average irrigation of 369 mm/yr
• Water use reduced by 57%
• Total annual P reduced by 26%
• Simulated biomass growth increased,
potentially due to greater nutrient
availability in the soil profile
25
Conclusions and Next Steps
A coupled SWAT-BATHTUB model for the Prespa Lake watershed was able to
simulate the hydrology and water quality of tributary rivers and the lake
Preliminary results suggest that 10% of the landscape contributes 64% of the
total P, indicating high potential for meaningful P reduction with targeted
alternative practices
Initial evaluation of improved irrigation practices showed benefits in terms of
water use savings and lower pollutant losses, with no effect on crop yield
Early estimates suggest that a reduction of total P load of ~40% would bring
lake conditions solidly into the mesotrophic range
Additional management alternatives will be assessed in the coming months
Through the support of the UNDP, water quality monitoring will continue
throughout the watershed, providing data for future model refinements
26
Thank you.
Acknowledgements:
Dimitar Sekovski, UNDP
Mary Watzin, North Carolina State University
Ordan Cukaliev and Cvetanka Popovska,
University of Ss. Cyril & Methodius
For more information, contact: