Bayesian SPARROW Model Song Qian Ibrahim Alameddine The University of Toledo American University of Beirut.

Slides:



Advertisements
Similar presentations
SPARROW Modeling in the Mississippi and Atchafalaya River Basins (MARB) Dale Robertson, Richard Alexander, and David Saad U.S. Geological Survey USGS/USEPA.
Advertisements

Delaware River Basin SPARROW Model Mary Chepiga Susan Colarullo Jeff Fischer
U.S. Department of the Interior U.S. Geological Survey Welcome to the USGS Webinar: New Science and Online Management Tools to Help Guide Action on Nutrients.
Developing Modeling Tools in Support of Nutrient Reduction Policies Randy Mentz Adam Freihoefer, Trip Hook, & Theresa Nelson Water Quality Modeling Technical.
The Wisconsin River TMDL: Linking Monitoring and Modeling Ann Hirekatur, Pat Oldenburg, & Adam Freihoefer March 7, 2013 Wisconsin River TMDL Project Team.
Estimating the Sources and Transport of Nitrogen in the Mississippi River Basin Using Spatially Referenced Modeling Techniques R.B. Alexander, R.A. Smith,
Land Use Change and Its Effect on Water Quality: A Watershed Level BASINS-SWAT Model in West Georgia Gandhi Raj Bhattarai Diane Hite Upton Hatch Prepared.
SPARROW Modeling in the Mississippi River Basin Iowa Science Assessment Davenport, IA Nov. 14, 2012 (608) By Dale M. Robertson*
0 The National Hydrography Dataset Plus a tool for SPARROW Watershed Modeling Richard Moore (presented by Alan Rea)
Computational statistics 2009 Sensitivity analysis (SA) - outline 1.Objectives and examples 2.Local and global SA 3.Importance measures involving derivatives.
Modeling Water Quality. Special reference of this work to….
Hydrologic Cycle and Watersheds. Hydrologic Cycle Components Precipitation Infiltration Percolation Runoff Evapotranspiration.
Water Quality Modeling in GIS Application of Schematic Network Processing Schema Links and Nodes have unique behaviors based on their type A framework.
Great Environmental Indicators (GLEI) Lakes.
Soil Water Assessment Tool (SWAT) Model Input
Introduction- Tax Value How are people using stacked ecosystem services? – “Bringing Ecosystem Services to Market” (**can we mention?) – Broad spatial.
SPARROW Water- Quality Modeling: Application of the National Hydrography Dataset What is SPARROW? Use of NHD SPARROW results By Craig Johnston and Richard.
Economic and Biophysical Models to Support Conservation Policy: Hypoxia and Water Quality in the Upper Mississippi River Basin CARD Resources and Environmental.
Transitory storage of N R in watersheds occurs in both surface and subsurface reservoirs. The factors pertaining to storage examined here are: precipitation,
Hydrologic Modeling: Verification, Validation, Calibration, and Sensitivity Analysis Fritz R. Fiedler, P.E., Ph.D.
Estimating Pollutant Loads Caroni River Bolivar, Venezuela Global Applications of GIS Technology Lee Sherman.
U.S. Department of the Interior U.S. Geological Survey Regional scale point source nutrient load estimation in support of SPARROW* modeling Gerard McMahon,
NHD Products and Applications InterMountain GIS 2014 NHD Workshop April 7, 2014.
Assessing Alternative Policies for the Control of Nutrients in the Upper Mississippi River Basin Catherine L. Kling, Silvia Secchi, Hongli Feng, Philip.
1 Evaluating and Estimating the Effect of Land use Changed on Water Quality at Selorejo Reservoir, Indonesia Mohammad Sholichin Faridah Othman Shatira.
Patapsco/Back River SWMM Model Part I - Hydrology Maryland Department of the Environment.
112.3 Jessica L. Feeser, M. Elise Lauterbur & Jennifer L. Soong Research Project for Systems Ecology (ENVS 316), Fall ’06 Oberlin College, Oberlin OH BackgroundFindings.
Gulf of Mexico Hypoxia and Mississippi River Basin Nutrient Losses Herb Buxton, USGSRob Magnien, NOAA Co-Chairs, Monitoring, Modeling, and Research Workgroup,
Effect of Spatial Variability on a Distributed Hydrologic Model May 6, 2015 Huidae Cho Water Resources Engineer, Dewberry Consultants Part-Time Assistant.
U.S. Department of the Interior U.S. Geological Survey Using Monitoring Data from Multiple Networks/Agencies to Calibrate Nutrient SPARROW* Models, Southeastern.
Delaware River Basin SPARROW Model Mary Chepiga, , Susan Colarullo, , Jeff Fischer, ,
Vision for the National Geospatial Framework for Surface Water Robert M. Hirsch Associate Director for Water U.S. Department of the Interior U.S. Geological.
Gulf of Mexico Hypoxia and Nutrient Management in the Mississippi River Basin Herb Buxton, U.S. Geological Survey.
Results and Discussion The above graph depicts FC colony plate averages for each sample site. Samples are ordered from upstream to downstream as indicated.
BASINS 2.0 and The Trinity River Basin By Jóna Finndís Jónsdóttir.
Identifying Changes to Stream Condition caused by Urbanization How understanding the responses can improve ecological risk characterization
1. The Study of Excess Nitrogen in the Neuse River Basin “A Landscape Level Analysis of Potential Excess Nitrogen in East-Central North Carolina, USA”
Relating Surface Water Nutrients in the Pacific Northwest to Watershed Attributes Using the USGS SPARROW Model Daniel Wise, Hydrologist US Geological Survey.
Opportunities for Collaboration on Water- Quality Issues in the Mississippi River Basin Herb Buxton, Office of Water Quality.
Integrating the NAWQA approach to assessments in rivers and streams By Donna Myers, Bill Wilber, Anne Hoos, and Charlie Crawford U.S. Geological Survey,
Exposure Assessment for Health Effect Studies: Insights from Air Pollution Epidemiology Lianne Sheppard University of Washington Special thanks to Sun-Young.
Modeling Stream Flow of Clear Creek Watershed-Emory River Basin Modeling Stream Flow of Clear Creek Watershed-Emory River Basin Presented by Divya Sharon.
SPARROW: A Model Designed for Use With Monitoring Networks Richard A. Smith, Gregory E. Schwarz, and Richard B. Alexander US Geological Survey, Reston,
U.S. Department of the Interior U.S. Geological Survey Automatic Generation of Parameter Inputs and Visualization of Model Outputs for AGNPS using GIS.
QUANTIFYING UNCERTAINTY IN ECOSYSTEM STUDIES Carrie Rose Levine, Ruth Yanai, John Campbell, Mark Green, Don Buso, Gene Likens Hubbard Brook Cooperators.
U.S. Department of the Interior U.S. Geological Survey Reston, Virginia (703) NHD Flow and Velocity Project Greg Schwarz, Reston,
Controls on Catchment-Scale Patterns of Phosphorous in Soil, Streambed Sediment, and Stream Water Marcel van der Perk, et al… Journal of Environmental.
The Future of NHDPlus 2009 NHD Stewardship Conference Denver, CO April 14-17, NHD Stewardship Conference Denver, CO April 14-17, 2009 Tommy Dewald.
NOPOLU System2 Large scale assessment of non-point nutrient sources.
1 A case-study of Continental Water dynamics in the Guadalupe and San Antonio River Basins Cédric David University of Texas at Austin.
Using Regional Models to Assess the Relative Effects of Stressors Lester L. Yuan National Center for Environmental Assessment U.S. Environmental Protection.
Using the NHDPlus for drainage area delineation and site matching Kirsten Cassingham, NC Water Science Center Silvia Terziotti, NC Water Science Center.
High Rock Lake Nutrient Modeling Update Pam Behm - NC Division of Water Resources Environmental Management Commission Water Quality Committee Information.
Graduate Students, CEE-6190
Puget Sound Watershed Characterization Project
Dave Clark and Michael Kasch
Estimation of Runoff & nonpoint source pollution using GIS techniques
Approaches to Continental Scale River Flow Routing
US Environmental Protection Agency
Optimising Predictions of Sediment and Nutrient Loads Using AnnAGNPS
Water Quality Acquisition Systems in Australia
Water Quality Analysis for Selected Streams in Utah and Idaho
Paper by: Bloniarz D. , M. Matteo, T
Digital Elevation Models and Hydrology
Public Meeting February 19, 2009
1. The Study of Excess Nitrogen in the Neuse River Basin
Regional Hydraulic Model for the City of Austin
Department of Bioscience
Soil Gross Nutrient Balances
Explaining the recent changes in agricultural nutrient load in Finland
Presentation transcript:

Bayesian SPARROW Model Song Qian Ibrahim Alameddine The University of Toledo American University of Beirut

SPARROW SPARRO WSPARROW: SPAtially Referenced Regressions On Watershed attributes SPARROW estimates the origin and fate of contaminants in river networks It is a semi-empirical non-linear model It is spatial in structure and takes into account the nested configuration of monitoring stations in a basin Can be used to link changes in the watershed to changes in water quality

SPARROW EQUATION Nutrient loading (L) at a downstream water quality monitoring station i: # of sources # of upstream reaches Contribution from Different sources (S) Losses/sinks Multiplicative error term

SPARROW Shortcomings Some of the shortcomings of SPARROW: Temporal and Spatial average Coarse spatial resolution  regional specifics often omitted autocorrelationSpatial autocorrelation in model residuals Model developed to run under

What Did We Do? temporally dynamicWe changed the model’s architecture to make it temporally dynamic new regionalizingWe developed a new regionalizing approach –Substitute space (# of stations) with time (# of years) nestedWe nested the model within a larger scale regional model changesWe assessed changes in loading over time for the Neuse subwatersheds open sourceWe moved the model to an open source platform

How it works for BSPARROW? USGS Flow and WQ data, GIS data (LULC, soil, …) Data preparation Bayesian computation Statistical data analysis GIS for mapping & display

Neuse SPARROW: Bayesian, Dynamic, & Regional Nested the model within the lager scale Nitrogen Southeast model (Hoos & McMahon, 2009) Updated the model over time (time step = 2 years) –Used 12 years of data  Regionalization over time –Data and model parameters change over time (dynamic) –Bayesian updating (posterior of t-1 = prior at t)

Main Tasks Develop a regionalized basin- specific model architecture Make it dynamic  update the model over time and regionalize catchments Pinpoint catchments of concerns and identify pollution types

Main Tasks Synthetic data  Assess parameter correlations  Check for model equifinality  Quantify value of extra monitoring on model fit Lake Erie data  Develop a regionalized basin- specific model  Update the model over time  Identify sub-basins of concerns

Bayesian SPARROW: Synthetic Data 1.Fixed model coefficients to literature reported values 2.Generate synthetic loads 2.Generate synthetic loads leaving each sub-catchment (+ accounted for land and aquatic losses) different arrangements 3.Randomly generated different arrangements of monitoring networks  simulate the effect of additional monitoring  4.Assigned vague priors on the model coefficients  Data driven inference Attempted to assess model convergence, parameter correlations, and model fit

Results: Parameter Correlations Strong correlations between some model parameters –Affects the mechanistic interpretation of the model coefficients –Hard to interpret the coefficients separately α: land-to-water delivery coefficient (a function of soil properties, e -α soil j ) β 1 : source coefficient on agricultural sources (Kg/Kg) β 2 : source coefficient on urban sources (MT/Km 2 ) β point : source coefficient on point sources (Kg/Kg)

Results: Model Fit (Increased Stations) Increasing the number of stations improves model fit > 1 stStations on > 1 st order streams improve the fit better # of stations Stream order R2R2R2R2RMSE13NA > > > > >

Model Fit: # of stations (1 st order)

Model Fit # of Stations  #of stations

Results: Convergence (Increased Stations) The common metrics (R 2 and RMSE) used to assess model fit can be deceiving Increased # of stations identifibalityIncreased # of stations help in the identifibality of model parameters β1β1 β point β2β2

Results: Multi-Dimensional Convergence (Increased Stations) 3-D Posterior Space 1348

Main Tasks Synthetic data  Assess parameter correlations  Check for model equifinality  Quantify value of extra monitoring on model fit Neuse River data  Develop a regionalized basin- specific model  Update the model over time  Identify sub-basins of concerns

& Do the Model Coefficients Vary Over Time? & How Do They Compare to Their SE Counterparts?

Neuse: Source Coefficients decreases Uncertainty around model coefficients decreases with time Some coefficients close to SE model, but other very different Some of the temporal variations we are seeing is due to: 1.Convergence 2.Changes occurring at the landscape α: land-to-water delivery coeffiecinet β 1 : source coefficient agricultural sources (Kg/Kg) β 2 : source coefficient urban sources (MT/Km 2 ) β point : source coefficient point sources (Kg/Kg)

Neuse: Aquatic Attenuation In-stream decay close to SE model Most of the nitrogen removal happens in small low flowing stream segments Lake attenuation different from SE model (less efficient) K s : In-stream decay K r : Lake removal Small segments Medium segments Large segments Lake

How Did the Neuse BSPARROW Model Perform Over Time?

Neuse SPARROW: Model Fit

How Do We Compare to the SE Model? (Hoos & McMahon, 2009)

Where Are the Areas of Concern? Have They Changed Over Time?

1990 Neuse Nitrogen Export by Basin 2001

Yield to Neuse Estuary by Basin Durham Cary Morrisville Raleigh Kinston Durham Cary Morrisville Raleigh Kinston

Conclusions Bayesian temporally dynamic nestedRegionalization of SPARROW to basin level possible: Bayesian temporally dynamic nested modeling framework changeLoads (and model coefficients) across the basin change over time and the model is capturing these changes Urban runoffUrban runoff seems to be a concern for TN loading in the Neuse Nitrogen loading to the Neuse Estuary have decreased  relative success in environmental management