SPARROW Modeling of Surface Water Quality: Applications to the Lake Michigan Basin

1
SPARROW Modeling of Surface Water Quality
Applications to the Lake Michigan Basin
By Dale M. Robertson and David A. Saad,
Wisconsin WSC Richard B. Alexander and Gregory
E. Schwarz, National Center, Reston, VA
dzrobert_at_usgs.gov [phone]
2
SPARROW Water-Quality Model - DescriptionSPAtiall
y Referenced Regression on Watershed Attributes
http//water.usgs.gov/nawqa/sparrow Smith et al.
1997

• Hybrid statistical and mechanistic process
  structure mass-balance constraints data-driven,
  nonlinear estimation of parameters
• Separates land and in-stream processes
• Once calibrated, the model has physically
  interpretable coefficients model supports
  hypothesis testing and uncertainty estimation
• Predictions of mean-annual flux reflect
  long-term, net effects of nutrient supply and
  loss processes in watersheds

• Hybrid statistical and mechanistic process
  structure mass-balance constraints data-driven,
  nonlinear estimation of parameters

3
SPARROW Predictions of Nitrogen Flux USEPA RF1 -
62,000 reaches nationally (3,200 Upper Miss.)
HUC12
SPARROW?SPAtially Referenced Regressions On
Watershed Attributes
4
Total Nitrogen Load
1992 Nitrogen SPARROW Model Output
Alexander and others, 2007
5
Total Nitrogen Delivered Incremental Yield
6
Total Nitrogen Delivered Incremental Yield
Top 150
2002 Nitrogen SPARROW Output
7
Ranked Incremental Nitrogen Yields From the HUCS,
with 90 CIs
8
90 Confidence Intervals for Yields and Ranks
9
HUCS In or Potentially In The Top 150 For TN
10
Take Advantage of Data from Other USGS and Other
Agency Programs
Sites used in National Models
Sites Planned to be used in Regional Models
11
U.S. Geological Survey SPARROW models
12
Mississippi River SPARROW Model
13
SPARROW Modeling Result for the Upper Midwest
14
Incremental Yield
Ranking by Incremental Yield
15
Future Improvements from Regional SPARROW Models
1. Better spatial resolution More sites and
especially more smaller sites, should lead to
more accurate predictions at smaller scales. 2.
Further reductions in biases. 3. Better
definition of source terms better point-source
data, more sites in unique areas, possible better
local GIS inputs. 4. Better able to address more
regional and local questions.