Use of MODIS Data in Dynamic SPARROW Analysis of Watershed Loading Reductions

Wednesday, 17 December 2014: 3:25 PM
Richard A Smith1, Gregory E Schwarz1, John W Brakebill2, Anne Hoos3, Richard B Moore4, Anne Walden Nolin5, Jhih-Shyang Shih6, Celeste Journey7 and Molly Macauley6, (1)USGS, Reston, VA, United States, (2)USGS Maryland/Delaware/District of Columbia Water Science Center, Baltimore, MD, United States, (3)USGS Tennessee Water Science Center, Nashville, TN, United States, (4)U.S. Geological Survey, Pembroke, NH, United States, (5)Oregon State University, Corvallis, OR, United States, (6)Resources for the Future, Washington, DC, United States, (7)USGS South Carolina Water Science Center, Columbia, SC, United States
Predicting the temporal response of stream water quality to a proposed reduction in contaminant loading is a major watershed management problem due to temporary storage of contaminants in groundwater, vegetation, snowpack, etc. We describe the response of dynamically calibrated SPARROW models of total nitrogen (TN) flux to hypothetical reductions in reactive nitrogen inputs in three sub-regional watersheds: Potomac River Basin (Chesapeake Bay drainage), Long Island Sound drainage, and South Carolina coastal drainage. The models are based on seasonal water quality and watershed input data from 170 monitoring stations for the period 2002 to 2008.

The spatial reference frames of the three models are stream networks containing an average 38,000 catchments and the time step is seasonal. We use MODIS Enhanced Vegetation Index (EVI) and snow/ice cover data to parameterize seasonal uptake and release of nitrogen from vegetation and snowpack. The model accounts for storage of total nitrogen inputs from fertilized cropland, pasture, urban land, and atmospheric deposition. Model calibration is by non-linear regression. Model source terms based on previous season export allow for recursive simulation of stream flux and can be used to estimate the approximate residence times of TN in the watersheds.

Catchment residence times in the Long Island Sound Basin are shorter (typically < 1 year) than in the Potomac or South Carolina Basins (typically > 1 year), in part, because a significant fraction of nitrogen flux derives from snowmelt and occurs within one season of snowfall. We use the calibrated models to examine the response of TN flux to hypothetical step reductions in source inputs at the beginning of the 2002-2008 period and the influence of observed fluctuations in precipitation, temperature, vegetation growth and snow melt over the period. Following non-point source reductions of up to 100%, stream flux was found to continue to vary greatly for several years as a function of seasonal conditions, with high values in both winter (January, February, March) and spring due to high precipitation and snow melt, but much lower summer yields due to low precipitation and nitrogen retention in growing vegetation (EVI). Temporal variations in stream flux are large enough to potentially mask water quality improvements for several years.