This document discusses linking water quality at Great Lakes beaches to land use in the surrounding watershed. It presents initial models that relate levels of E. coli bacteria at a beach in Racine, WI to rainfall, river discharge levels, and estimated bacterial loads from the nearby Root River watershed. The models explain around 40% of the variation in bacteria levels and include factors like wave height, water clarity, and seasonal effects. Future work is aimed at refining the watershed loading estimates and validating the linked beach-watershed modeling framework.

1. Linking Great Lakes Beach Water-
Quality to Land Use
Adam C. Mednick
Wisconsin Dept. of Natural Resources
WAFSCM Conference, 11/4/10, Wisconsin Dells

2. Outline
I. Problem/ Background
II. Beach Water-Quality “Nowcasts”
III. From Beach to Watershed
IV. Modeling/ Decision-Support Framework

3. I. The Problem
 Elevated fecal indicator bacteria
in nearshore recreational waters
 Public health impacts
Escherichia coli
 Economic impacts
 In WI 2003-2009:
 3,737 Swim Advisories
 912 Beach Closures

4. Difficult to ID/ Mitigate Sources
 Sources, pathways, and contributing
factors vary from beach to beach
Source: NOAA GLERL

5. Monitoring Challenges
 Standard methods take 18-24 hours
 Collect/ transport water quality samples
 Lab analysis
Quanti-tray enumeration

6. Monitoring Errors (Type I) .
 Wisconsin 2003-2009:
 63% of closings reflected false exceedances
of state closure guideline (1,000 CFU/100 mL)
 42% of advisories reflected false
exceedances of the EPA freshwater standard
(235 CFU/ 100 mL)

7. Monitoring Errors (Type II)
 Wisconsin 2003-2009:
 3% of non-advisory days with water quality
samples exceeded the state guideline
(1,000 CFU); i.e. should have been closed
 9%… exceeded the EPA standard (235 CFU);
i.e., should have been posted

8. Beach Water-Quality “Nowcasts”
Y = ϐ0 + (ϐ1*X 1) + (ϐ2*X 2) + … (ϐk *X k ) + ε
FIB Concentration

9. X1 = Rainfall
Source: NWS

10. X2 = Turbidity (NTU)

11. X3 = Wave Height
Source: USGS Ohio Water Science Center
Source: Racine Health Department

12. X4 = Sky Conditions
Source: NOAA

13. X5 = Wind Speed*Direction
Source: NOAA

14. X6 = Nearshore Current
Grand River Plume (4/20/10)
Grand Haven State Park, Michigan
Source: NOAA GLERL

15. X6 = Nearshore Current
Grand River Plume (4/20/10)
Grand Haven State Park, Michigan
Source: NOAA GLERL

16. X6 = Nearshore Current
Grand River Plume (4/20/10)
Grand Haven State Park, Michigan
Source: NOAA GLERL

17. X7 = FIB loading
Grand River Plume (4/20/10)
Grand Haven State Park, Michigan
Source: NOAA GLERL

18. First Operational “Nowcast” in WI
Port Washington, 2009-2010
 Explanatory Variables:
 48-hour Rainfall
 Turbidity (NTU)
 24-hr Stream Flow (total discharge)
 Wave Height
 Water Temperature
 Air Temperature
 “Third Qtr” of Beach Season (Y/N)

19. E. Coli (CFU/100 mL)
20
07
20 .05
1
10
100
1000
10000
07 .29
20 .06
07 .17
20 .07
07 .04
20 .07
07 .21
20 .08
07 .09
20 .08
08 .26
20 .06
08 .12
20 .06
08 .29
Date
20 .07
08 .15
20 .08
08 .02
"Combined Model"
20 .08
09 .18
20 .07
09 .02
20 .07
09 .30
.0
Predicted vs. Observed 2007-'09
8.
27
- Nowcast -
* R-sqr = 64%
First Operational “Nowcast” in WI
* MAE = 19 CFU
Standard
Predicted
Observed

20. First Operational “Nowcast” in WI
Sample date/time E. Coli E. Coli Units
(Model) (Lab)
Tues. 7/13/2010 8:30 4 23 MPN/ 100ml
Wed. 7/14/2010 8:30 24 24 MPN/ 100ml
Thurs. 7/15/2010 9:00 3,432 2,419 MPN/ 100ml
Friday 7/16/2010 8:50 166 127 MPN/ 100ml

21. Expanded “Nowcast” Modeling
 Federal Great Lakes
Restoration Initiative
 From 1 operational
nowcast to 20 by 2013
 59 candidate beaches Pt. Washington
 Adequate data
Study Area
 High-priority and/or
impaired

22. III. From Beach to Watershed

23. Root River, Racine
 Beach/Waterfront economically important
 Veryhigh number of closures and swim
advisories prior to 2006
 Inter-agency effort to address problem:
 City Public Health Dept., Engineering Dept.,
Planning Dept.

24. Multi-Year Sanitary Survey
 2003-present
 Intensive
sampling of nearshore waters,
stormwater outfalls, and Root River outlet
 Concurrent surveys of:
 Beach conditions (e.g., # bathers, gulls,
algae, litter…)
 Nearshore currents, wave height, turbidity
 Meteorological conditions

25. Site-Level Mitigation
 Redesigned stormwater outfall
 Constructed wetland cells
 Re-graded lower beach
 Modified grooming to increase aeration
 Ordinance against feeding gulls

26. Watershed Conditions
 Post-mitigation,
contamination events
primarily associated with major storm
events/ Root River discharge
 Riverplumes often
associated with
elevated FIB at Great
Lakes beaches, as well
as Marine beaches
(e.g. He & He 2008) Grand River at Grand Haven, MI
Source: NOAA

27. Watershed Conditions
 Startingin 2007, weekly samples collected
at 34 sites along lower Root River

28. 25 Open Water Sites

29. 9 Outfalls

30. Samples vs. Flow Conditions
Flow Duration Curve (daily data 1963-2009)
Wet Conditions Dry Conditions
10000
All Dates 1962-2009
1000 Sample Dates (N~75)
Log of flow (cfs)
100
10
1
0 10 20 30 40 50 60 70 80 90 100
0.1
0.01
Percent of time flow rate exceeded

31. Applied-Research Questions
 Can we develop a modeling framework
that links nearshore FIB concentrations to
watershed conditions?
 Land Use
 Development Practices (LID)
 Ifso, can we translate this into a usable
Planning-Support Tool?
 Nowcasts for Public Health officials
 Impact Assessment for Planners

32. Limited Research to date
 Kay and colleagues (2005) modeled real-
time FIB fluxes for a coastal watershed
discharging near recreational waters on
the Irish Sea.
 Export coefficients based on percentages
of different land uses
 Did
not link outputs to nearshore FIB
concentrations

33. IV. Modeling/ Decision-Support
Framework
Watershed modeling component
Rainfall
Beach “nowcast” component
Land Use Soils
Rainfall Waves Wind Turbidity Current
Runoff
NPS Nearshore
E. coli E. coli

34. Watershed Component
 Based on Long-Term Hydrologic Impact
Assessment (“L-THIA”)
 Spatially-distributed automation of rainfall-runoff
Curve Number (CN) method + event mean
concentration (EMC) coefficients
 Simple model
 Similar to PLOAD
 Basis for runoff/NPS in the PSS software “INDEX”
 Enables Web application

35. L-THIA (Online Version)
https://engineering.purdue.edu/~lthia/

36. L-THIA (“Real-Time”)

37. Beach Component
 U.S.
EPA’s “Virtual Beach”:
Model-building/ decision-support tool
 Walks beach managers through the
process of building
and operating
statistical models to
predict real-time FIB
concentrations

38. Initial Watershed Models
 Distributed
CN’s based on 30m land use
(NLCD 2001) and soils (SSURGO)
 Daily rainfall (2007-10) from Racine Airport
 CN’sadjusted for Antecedent Soil
Moisture based on previous 5 days
 Daily direct runoff estimates per 30m cell

39. Root River
Watershed
2010 Land Use (NLCD)
Open Water
Developed, Open Space
Developed, Low Intensity
Developed, Medium Intensit
Developed, High Intensity
Barren Land
Deciduous Forest
Evergreen Forest
Mixed Forest
Shrub/Scrub
Grassland/Herbaceous
Pasture/Hay
Cultivated Crops
Woody Wetlands
Emergent Wetlands
0 5 Kilometers

40. Curve
Numbers
< 65
65 - 70
70 - 75
75 - 80
80 - 85
85 - 90
90+
0 5 Kilometers

43. E. coli Coefficients
For each land use, 4 different sets of
EMC coefficients for total coliforms,
adjusted (EC = 0.625 * TC):
1. L-THIA default EMCs
2. WMM EMCs (Rouge River Natl. Wet
Weather Demonstration Project 1998)
3. PLOAD EMCs (U.S. EPA 2001)
4. NSQD median values (Pitt el al. 2009)

44. Initial Beach Model #1
 Rainfall + Clarity + Sky Conditions + Southerly Current
 R-square = 40.66%
Variable Coefficient CFU/100 mL t-Statistic p-Value
Constant + 1.97 14.870 0.000
(24-hr Rainfall [in]) -0.5 + 1.06 + 2.90 4.442 0.000
Wave Height (ft.) + 0.33 + 1.39 2.823 0.005
Water Clarity: "Somewhat Turbid" + 0.90 + 2.46 2.814 0.005
Water Clarity: "Turbid" + 0.69 + 2.00 2.586 0.010
Water Clarity: "Opaque" + 0.83 + 2.29 2.482 0.014
Sky Conditions: "Overcast" + 0.49 + 1.63 2.978 0.003
Stage-of-season: 3rd Qtr + 0.51 + 1.66 2.909 0.004
Stage-of-season: 4th Qtr + 0.47 + 1.60 2.569 0.011
(Southerly Surface Current [m/sec])-0.5 + 0.54 + 1.72 1.317 0.190

45. Initial Beach Model #2
 24-hour Avg. Root River Discharge in place of rainfall
 R-square = 40.24%
Variable Coefficient CFU/100 mL t-Statistic p-Value
Constant + 1.63 9.741 0.000
(Root River Discharge) -0.5 + 0.03 + 1.03 4.281 0.000
Wave Height (ft.) + 0.46 + 1.58 4.025 0.000
Water Clarity: "Somewhat Turbid" + 1.02 + 2.77 3.206 0.002
Water Clarity: "Turbid" + 0.55 + 1.73 2.011 0.046
Water Clarity: "Opaque" + 0.76 + 2.14 2.273 0.024
Sky Conditions: "Overcast" + 0.52 + 1.69 3.171 0.002
Stage-of-season: 3rd Qtr + 0.71 + 2.04 3.886 0.000
Stage-of-season: 4th Qtr + 0.67 + 1.96 3.553 0.000
(Southerly Surface Current [m/sec])-0.5 + 0.69 + 1.99 1.639 0.103

46. Initial Beach Model #3
 Estimated E. coli load in place of rainfall
 R-square = 38.71%
Variable Coefficient CFU/100 mL t-Statistic p-Value
Constant + 2.04 15.298 0.000
(Estimated E. coli load, WMM)-0.5 + 0.00 + 1.00 3.596 0.000
Wave Height (ft.) + 0.38 + 1.47 3.308 0.001
Water Clarity: "Somewhat Turbid" + 0.85 + 2.35 2.596 0.010
Water Clarity: "Turbid" + 0.64 + 1.89 2.329 0.021
Water Clarity: "Opaque" + 0.81 + 2.25 2.392 0.018
Sky Conditions: "Overcast" + 0.50 + 1.65 2.982 0.003
Stage-of-season: 3rd Qtr + 0.49 + 1.63 2.745 0.007
Stage-of-season: 4th Qtr + 0.48 + 1.61 2.561 0.011
(Southerly Surface Current [m/sec])-0.5 + 0.60 + 1.82 1.421 0.157

47. Forthcoming
 Watershed-specificEMC coefficients
estimated Root River data
 Data partitioned by flow regime and season
 Substitute lot-level land use for NLCD
 Calibrate curve numbers
 Validate E. coli estimates against sampling
 LID-adjusted curve numbers

48. Linking Great Lakes Beach Water-
Quality to Land Use
Adam C. Mednick
Wisconsin Dept. of Natural Resources
WAFSCM Conference, 11/4/10, Wisconsin Dells