EFFECT OF SPATIAL AND TEMPORAL RESOLUTION UNCERTAINTY ON PREDICTED BEST MANAGEMENT

1
EFFECT OF SPATIAL AND TEMPORAL RESOLUTION
UNCERTAINTY ON PREDICTED BEST MANAGEMENT
PRACTICE (BMP) EFFICIENCY Robert Miskewitz,
Josef Kardos, Christopher Obropta, and Katie
Giacalone October 25, 2007
2
Cohansey River Watershed Restoration Plan
The Upper Cohansey River was placed on the 303d
list as an impaired water body. As a result a
Total Maximum Daily Load (TMDL) created for it.
Using Areal Loading calculations the New Jersey
Department of Environmental Protection (NJDEP)
determined that the amount of fecal coliform and
phosphorus inputs to the river should be reduced
by 91.
Upper Cohansey Watershed
3
The Upper Cohansey River Watershed (UCRW) is a 65
square kilometers agriculturally-based watershed
that drains to Delaware Bay.  The land use is
predominantly vegetable, field and container
nursery, and sod production. 
4
The Watershed Restoration Plan included
identification potential sources and mitigation
strategies. A comprehensive field sampling
program which involved collecting water samples
at ten locations throughout the watershed for a
period of six months in the summer of 2006 was
completed, and a hydrologic model was
constructed. The model was used to pinpoint
sources, and predict the results from various
treatment scenarios.
5
(No Transcript)
6
Hydrologic Modeling

• Selected the Soil Water Assessment Tool (SWAT)
  for phosphorus modeling in the watershed. SWAT is
  a river basin scale hydrologic model developed to
  quantify the impact of land management practices
  in large, complex watersheds.
• This effort represent one of the first
  applications of SWAT to a watershed with a large
  amount nursery agriculture.
• Modified NJDEP 2000/2002 land use/ land cover GIS
  data to include field nurseries, container
  nurseries, and sod farms each land use modeled
  with a different curve number.

7
Data used for construction of this SWAT model was
obtained from

• Topography (New Jersey Department of
  Environmental Protection, NJDEP)
• Land use (NJDEP 2002, adapted from site
  investigations)
• Soils (SURGO database)
• Weather (South Jersey Resource Conservation
  Development Council RISE system)
• Agricultural Practices (For this study farmers
  were surveyed by county agricultural agents)
• Ground Water (From USGS)
• Nutrient Parameters (USGS)

8
Initial Calibration
In its initial form the model was run for a
period of 5 years and was calibrated for flow on
a monthly basis. This model was delineated into
50 sub-basins. The data used to calibrate it was
collected from the USGS gauge at the lowest point
of the watershed, and phosphorus concentration
measurements at only six locations.
9
(No Transcript)
10
Average Annual Total Phosphorus Concentration
11
(No Transcript)
12
The model was re-built in order to address the
inaccuracies due to monthly flow calibration. It
was also found that by re-delineating the model
to the 10 sampling locations increased the speed
of the model 5xs and no impact on the
accuracy of the model was seen because
calibration was only possible where data exists.
The new model was them calibrated for daily flow
at the outlet of the watershed for three years,
and biweekly at each of the 10 sampling locations
over the six month sampling campaign.
13
Daily Flow Calibration at Seely Lake USGS Gauge
The model was calibrated for daily flow at the
base of the watershed using measurements from the
USGS gauge at Seely Lake. The Nash-Sutcliffe
Efficiency Coefficient for the calibration period
of 2004-2006 was 0.57.
14
Flow Calibration
NSE values range from 0.04 to 0.8.
15
Phosphorus Concentrations at Sampling Location C1
16
What is are the sources of phosphorus in this
watershed?

• The sources in this watershed are predominately
  Non-Point Sources.
• The two greatest suspected offenders are

17
But the problem with Geese is
18
They fly away
19
The modeling effort must concentration on
stationary sources such as farms, urban runoff,
and stream bank erosion. Is this
fair? Transient sources often get viewed as one
time events, but their impact on sediment pools
of phosphorus may be significant.
20
Annual Phosphorus Loadings from a Model
Calibrated using Two Different Time Scales
Daily Calibration
Monthly Calibration
21
Siting of BMPs using SWAT

• SWAT can be used to predict hotspots in terms of
  landuse or sub-basin or both
• Sub-basins lack interior routing routines (i.e.
  All HRUs are connected)
• SWAT cannot explicitly place a BMP into the model
  (except filter strips)
• SWAT cannot account for transient nutrient loads
  which means even non-point sources must be
  stationary

22
Load determination by land use
The largest loadings of phosphorus and sediment
in the watershed come from agricultural land.
Although this is partly due to its predominance
in the watershed, 41.78 of the total area, it is
also the largest normalized source as well. As a
result the filter strip option in the mgt. files
was used to install a 15 meter filter strip
around all agricultural land uses in the
watershed.
23
(No Transcript)
24
(No Transcript)
25
Results

• Predicted phosphorus loads for the first
  iteration of the model are higher by a factor of
  4 7.
• Sediment loads are allocated differently. Higher
  sediment in headwaters, less at downstream
  locations.
• Filter strip efficiency is predicted to be the
  similar for phosphorus, however load reduction in
  terms of mass is vastly greater in Model I.
• Sediment loads predicted by Model I are far less
  than Model II and the filter strips have a much
  greater effect.

26
Load determination by sub-basin
This type of analysis can be used to determine
the location of the sub-basins that might be
ideal for installing BMPs. However, it doesnt
tell us where in the sub-basin that the loads are
coming from. This could be broken down further
by looking at individual HRUs. In this analysis
it was determined that subbasins C4 and FR1 were
ideal for installing BMPs. As a result
bio-retention ponds were simulated in C4 and
constructed wetlands in FR1.
27
Annual Phosphorus Loadings from a Model
Calibrated using Two Different Time Scales
Daily Calibration
Monthly Calibration
28
Phosphorus Loads from sub-basin C4 (kg/year)
The sub-basin that represents the largest source
is the sub-basin that drains to C4. Due to the
high percentage of agricultural landuse in this
area it was determined that bio-retention ponds
would be installed to receive the runoff from 80
of this sub-basin.
29
Phosphorus Loads from sub-basins FR1 and HR1
(kg/year)
Model I predicted that HR1 was the second largest
source in the watershed, while when the model was
calibrated daily HR1 was no longer one of the
largest sources while FR1 was. By installing
wetlands in each of these sub-basins an
evaluation of the potential misallocation of
resources was completed.
30
Results

• The additional data requirements and time to
  complete the daily time step calibration of this
  model has resulted in the revision of annual
  estimates of TP loads from 6,065 kg to 1,570 kg
  for water year 2005 and from 10,071 kg to 1878 kg
  for water year 2006 at the outlet of the modeled
  watershed.
• These are differences of 4,495 and 8,193 kg of
  TP, respectively.
• BMPs designed to treat loads of 6,000 to 10,000
  kg per year will be vastly different from those
  that are designed to treat 1,500 to 1,800 kg per
  year.

31
Conclusions

• The extra time spent refining a model will be
  cheaper than building inappropriate BMPs for the
  conditions in a watershed.
• Inaccuracies due to temporal scale can result in
  inappropriate management decisions
• Spatial resolution of calibration is extremely
  important for accurate load allocations and thus
  siting of BMPs