BMP Evaluations Using SWAT Model and Associated Uncertainties

1
BMP Evaluations Using SWAT Model and Associated
Uncertainties

• A. Shirmohammadi and T. W. Chu.
• Biological Resources Engineering Department
• University of Maryland, College PArk

2
Forests and wetlands trap sediments and help
slow the flow of pollutants into the Bay. Their
loss, coupled with the decline of grasses and
oysters in the 1970s and 1980s, caused the Bay to
lose much of its resilience.

• Chesapeake Quarterly
• MD Sea Grant College, Vol. 3, Num. 3
• October 2004

3

• Ecological Resilience Provides a measure of the
  amount of disturbance that an ecosystem can
  withstand without shifting into an alternate
  stable state

4

• For Bay The shift from a food web dynamic
  driven by benthic processes- such as underwater
  grasses and oysters- to one driven by
  phytoplankton in the water column is a classic
  example of regime shift, a shift between stable
  states.

5
(No Transcript)
6

• Identifying Thresholds Researchers must develop
  a systematic way to anticipate when a system is
  getting close to a threshold or tipping point and
  prevent it from going over the edge. They also
  need to develop methods to turn around the state
  of a system such as the Chesapeake Bay from
  undesirable to desirable! ---Load Reduction!

7
Background

• Hydrologic/ water quality models are the main
  tools used to tabulate total maximum daily loads
  (TMDLs)
• Procedure for tabulating TMDLs
• Use monitored data as input into model to
  represent base conditions
• Simulate alternate management scenarios
• Choose management scenario that meets water
  quality standards
• Determine total load and allocate load among
  sources

8
Background

• General expression for TMDL allocations
• TMDLSWLA SLA Future Growth MOS
• Waste Load Allocations (WLA)- point source
  contributions
• Load Allocations (LA)- non-point source
  contributions including background sources
• Margin of Safety (MOS)- accounts for
  uncertainties about the relationship between
  pollutant loads and receiving water quality
  (USEPA, 1999a)

9
Background

• Types of Uncertainty in Modeling
• Model Structure
• Parameter Values
• Natural Variability (Spatial and Temporal)
• Data Uncertainty
• Model Prediction

10
Watershed/Basin Scale Model SWAT (Arnold et al.,
1998)
11
Data Source

• Study Site
• Warner Creek Watershed

12
Data Collection

• Precipitation
• Stream flow
• Sediment
• NO3-N
• NH4-N
• TKN
• PO4-P
• TP

13
Location Frederick County, Maryland Area 346 ha
(856 acres) Soil Type 1/3 Area
Manor-Edgemont-Brandywine 2/3 Area
Penn-Readington-Croton Slope 95 Area Slope
lt15, 5 Area Slope between 15 and
25 Erodibility 65 Area Moderately erodible
12 Area Severely erodible 23 Area
classified not erodible Land Use Pasture, Dairy,
Beef, Cropland
14
(No Transcript)
15
Background on SWAT Model
16
Components of SWAT Model

• Hydrology
• Sedimentation (Erosion)
• Nutrients
• Pesticides
• Bacteria

17
(No Transcript)
18

• Calibration period (1994-1995).
• Validation period (1996-2002).

19
(No Transcript)
20
Table. Statistical results comparing measured and
simulated flow data at station 2A after
adjustment to the subsurface flow contribution
from outside the watershed.
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
-
NO3 - N Statistics
25
(No Transcript)
26
List of BMPs Simulated
27
Total areas for row crops planting within Warner
watershed
28
Results of BMP Simulation
29
Comparison of annual total streamflow at the
outlet of the watershed based on different BMP
implementations.
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
30
Comparison of annual surface runoff at the outlet
of the watershed based on different BMP
implementations.
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
31
Comparison of annual surface runoff at the outlet
of the watershed based on different BMP
implementations without winter crop planting .
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
32
Comparison of annual sediment loading at the
outlet of the watershed based on different BMP
implementations.
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
33
Comparison of annual nitrate nitrogen loading at
the outlet of the watershed based on different
BMP implementations.
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
34
Comparison of annual soluble phosphorus loading
at the outlet of the watershed based on different
BMP implementations.
Background up and downhill planting with
conventional tillage BMP1 contour planting with
conventional tillage BMP2 contour planting with
conservation tillage BMP3 contour planting
with no-till BMP4 contour stripcropping with
no-till
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
Comparison of average annual (1994-2002) model
prediction at the outlet of the watershed based
on different BMP implementations with and without
winter crop planting
41
Latin Hypercubic Sampling

• For Model Uncertainty

42
Latin Hypercube Sampling (LHS)
n5
Uniform distribution
Normal distribution

• Divided into n non-overlapping intervals on the
  basis of equal probability.
• One value from each interval is then selected
  randomly with respect to the probability density
  in the interval.

43
Model output distribution of Streamflow at the
watershed outlet
44
Model output distribution of Sediment loading at
the watershed outlet
45
Model output distribution of Nitrate loading at
the watershed outlet
46
Model output distribution of Streamflow at the
watershed outlet
1004
BMP4 BMP4 (1996)
47
Model output distribution of Nitrate loading at
the watershed outlet
BMP4 (1996) BMP4 (1996) without winter crop
48
Acknowledgement

• Our Cooperator,
• Dr. Linda Abbot of USDA/OCE/ORACBA

49
Model output distribution of Sediment loading at
the watershed outlet
BMP2 BMP4 (1996)