Estimating the Sources and Transport of Nitrogen in the Mississippi River Basin Using Spatially Referenced Modeling Techniques R.B. Alexander, R.A. Smith,

Estimating the Sources and Transport of Nitrogen in the Mississippi River Basin

Using Spatially ReferencedModeling Techniques

R.B. Alexander, R.A. Smith, G.E. Schwarz,

and J. Nolan

NAWQA, Nutrient Synthesis Group

Upper Mississippi River Basin Nutrient WorkshopMarch 25-26, 2002

http://water.usgs.gov/nawqa/sparrow/

Topics Presented

• Spatially referenced modeling techniques– Background on SPARROW

• Model applications in the Mississippi Basin– N sources and transport

• Updates and enhancements to the models

• Near-term and future research

SPARROW (SPAtially Referenced Regression on Watershed Attributes)

Land Use & Sources

Drainage & Impoundments

Landscape Features

Monitoring Data

Integrates watershed data over multiple spatial scales

to predict origin & fate of contaminants

Features of SPARROW SPAtially Referenced Regression

on Watershed Attributes

• Spatially referenced source inputs & watershed attributes

• Mechanistic structure (mass balance; flow paths; non-conservative flux)

• Nested monitoring sites• Spatial nonlinear regression• Empirical estimates of flux to

streams & watershed outlets from point & diffuse sources

• Mean annual or seasonal flux• Constituents: nutrients, atrazine,

fecal bacteria, suspended sediment, streamflow

SPARROWEstimated Equation

StreamLoad

Sources

Land-to-watertransport

Aquatictransport

Error

Nutrient ModelsFertilizer

Animal WastesAtmosphere (TN)

Industrial & Municipal WastesNonagricultural Diffuse Sources

Soil PermeabilitySlopeRunoff

Stream DensityTemperature (TN)

StreamflowWater Velocity

Channel LengthReservoir Hydraulics

TN Yield NASQAN I Sites

R-square 0.88 to 0.96

• Stream measurements of nutrient flux (monitoring data)

• Literature rate coefficients Catchment yields by land use, per capita waste loads, in-stream decay, reservoir settling rates

Evaluations of SPARROW Models Independent Verification ofCoefficients & Predictions

• Stream measurements of nutrient flux (monitoring data)

• Literature rate coefficients Catchment yields by land use, per capita waste loads, in-stream decay, reservoir settling rates

• Inter-model comparisonsU.S., Chesapeake Bay, Neuse/Tar R, & New Zealand SPARROWs, SWAT, HSPF, RivR-N, GWLF, regression methods, watershed process models, N budgets (NRC, 2000; Valigura et al. 2001 AGU volume; Seitzinger et al. in press)

• Spatial analyses of prediction errors (test of model misspecification)e.g., SCOPE N project (Alexander et al. in press)

Evaluations of SPARROW Models Independent Verification ofCoefficients & Predictions

SPARROW Applications to the Mississippi River Basin

Source Characterization &Nitrogen Delivery to the

Gulf of Mexico

Total Nitrogen Yield Total Nitrogen Yield Delivered to the Gulf Delivered to the Gulf

from NASQAN from NASQAN Monitoring SitesMonitoring Sites

• Transport: model empirically estimates loss

as function of 1st-order decay in different sized channels & water travel time

• Sources: model accounts for inputs of major diffuse and point sources of N

Alexander et al. Nature, 2000

Nitrogen Flux from the Mississippi River to the Gulf of Mexico: Share from Major Sources (with

90 percent confidence intervals)

0

10

20

30

40

50

60

70

Point Sources Fertilizer Use Livestockwastes

Atmosphericdeposition

Nonagriculturalnonpointsources

Per

cent

Sources of Total Nitrogen at the Mississippi River Outlet to the Gulf

SPARROW Estimates of Delivered TN YieldSPARROW Estimates of Delivered TN Yield

Agriculture

Point Sources

Atmosphere

Estimates of In-StreamNitrogen Loss

• SPARROW 1st-order loss rates estimated separately for different sized streams

• Illustrates use of model to test hypotheses about N loss—

inverse relation theoretically expected; i.e., less contact btwn.

water column and stream bottomin large channels

• SPARROW rates compare favorably with each other and

literature rates of denitrification- induced losses; depth major limiting factor explaining N loss

• Other channel properties account for large variability – need to understand mechanisms

• SPARROW provides effective tool for estimating large-scale transport of N

Dendritic Pattern of N Delivery toGulf of Mexico from Watershed Outlets

• Percentage estimated as a function of stream attributes only (loss rate, velocity, length)

• Large differences in delivery from neighboring watersheds

• Higher proportions of N delivered from areas close to large rivers (small streams

remove N more effectively)

• Management strategies should consider location of sources relative to large streams – a

commensurate level of control near large streams will remove more N downstreamAlexander et al. Nature, 2000

SPARROW Modeling

Updates and Enhancements

SPARROW Constituents & Methods

New / updated constituents• Combine 1991 NAWQA with

historical NASQAN WQ data• Nutrients (TN, TP; 1992

update)• Suspended Sediment • Fecal coliform • Atrazine (updated methods)

• Flow / velocity

New methods• Stream load estimation• Calibration accommodates

more spatially and mathematically complex descriptions of transport or storage (e.g., reach-specific decay)

• More spatially detailed watershed infrastructure

New SPARROW InfrastructureRF1 Streams Linked to 1-km DEM Watersheds

and 30-m MRLC Land Use

RF1 reach

RF1 reachwatershedboundary

Sources of Total Nitrogenin U.S. Streams

(1992 model: R2=0.90; 370 sites)

• NLCD land-use data allows more detailed estimation of nonagricultural diffuse sources (urban, forests)

• Other sources insensitive to addition of land-use terms

• Model fitted land-use yields and per-capita waste loads compare favorably with literature rates

Alexander et al., Water Resour. Res., in press

Total Nitrogen Loss in 75 Reservoirs of the Waikato River Basin, New Zealand

• Empirically estimated loss with SPARROW as function of 1st-order settling velocity and areal water load (based on

Vollenweider-type models)

• TN loss inversely related to reservoir flushing rate—i.e., smaller losses occur in more rapidly flushed reservoirs

SPARROW Estimates of Nitrogen Lossin Reservoirs

• SPARROW settling rates compare favorably with literature

• Magnitude of SPARROW rates suggest denitrification (rather than algal uptake and particulate burial) may be a dominant long-term loss process in reservoirs

• Impoundments are prominent features of U.S. landscape (> 70,000)—their location and size may be important to

understanding N fate in watersheds

SPARROW Applications

Water-Quality Management

NATURAL BACKGROUND CONCENTRATIONS OF

NUTRIENTS (BY NUTRIENT ECOREGION)

STUDY DESIGN• Used study of reference sites by Clark et al, 2000• Hybrid SPARROW model • Compare with models of S. American & African R.• Make predictions for all reaches in 14 “ecoregions”• Summarize as frequency distributions (see figure)

CONCLUSIONS• TN concentrations exceed background by

larger factor than do TP concentrations

• Large variation in background in several regions due to runoff & stream size

• Background exceeds EPA-proposed criteria in many regions

Objective: Select the “optimal” set of monitoring locations that improves the precision of model estimates of the “delivered TN yield” to the Gulf of Mexico

Method: (a) stratify the distribution of delivered yield for sites (273 NASQAN & NAWQA) and reaches; (b) determine the sample size from each strata that satisfies the objective; (c) randomly select 100 locations from the four strata

Network Design Using SPARROW

Additional design scenarios possible: (a) alternate populations of streams

having different attributes; (b) effect of station sample size; (c) different objective functions (e.g., concentration)

SPARROW Near-Term / Future Research

• Temporally variable models:– Stream loads modeled explicitly as function of time

(mean-annual loads estimated for selected time periods such as 1987, 1992, and 1997)

– Account for multi-year terrestrial storage of nutrients

– Include ’91, ’94, and ’97 NAWQA data and 1996-2000 NASQAN data

• Simultaneous multi-contaminant models (e.g., N forms; pesticides)

• NAWQA Cycle II activities (HST, ACT and NEET topical teams)

SPARROW Near-Term / Future Research

• Linking deterministic models to SPARROW– Tests of process hypotheses – More detailed management simulations– examples: TOPMODEL, SWAT, GW models (regional

SPARROWs)

• Evaluation / validation of model source characterizations and in-stream decay rates– N, O isotopes– experimental measurements of denitrification /

mass balance studies

SPARROW Near-Term / Future Research

• Biological modeling– Microbiological (pathogens, indicators)– Chlorophyll and algae– Fish tissue– Benthic invertebrates

• “Emerging” contaminants (e.g., antibiotics)

Cyber Seminar Presentation on Regional Sparrow Models—April 18, 2002

(Chesapeake Bay, New Eng., Neuse/Tar R.)

Studies provide an infrastructure for integrating local monitoring data, research, and management activities

SPARROW Workshop (Fall, 2002)

Three day workshop in Reston, VA:• Introduction to SPARROW modeling for

initiating regional or national studies• Presentation of results from national and

regional studies• Description of new capabilities• Discussion of potential regions, constituents

and applications for future modeling

SPARROW Web Site: http://water.usgs.gov/nawqa/sparrow/

• Mean-annual streamflow, water velocity, drainage area• MRLC land use (1992)• Population, waste disposal type (1990 Census)

Internal URL: http://hqsun2.er.usgs.gov/~mierardi/hucmaster

SPARROW WEBWatershed Data and Model Predictions

for 62,000 Stream Reaches

• Mean-annual nutrient conditions (yield, concentration, sources, prediction uncertainties)

• Natural background nutrient conditions• Public release: Sept. 2002