Skip

Blanchard Watershed ANNAGNPS Modeling Project- Final Report 

Prepared for:  U.S. Army Corps of Engineers- Buffalo District
Prepared by: LimnoTech
AnnAGNPS Project Team
  • Dr. Ron Bingner, Dr. Fred Theurer, USDA Agricultural Research Service
  • Dr. Kevin Czajkowski, David Dean, University of Toledo
  • Steve Davis, Jim Stafford, USDA Natural Resources Conservation Service
  • Greg Koltun, U.S. Geological Survey
  • Dr. Pete Richards, Heidelberg College
  • Byron Rupp, Billy Johnson, U.S. Army Corps of Engineers
  • Joe DePinto, Greg Peterson, Laura Weintraub, Amanda Flynn, Pranesh Selvendiran, LimnoTech
  • Libby Dayton, OSU

Download the Full Report (requires Adobe Acrobat).

Blanchard Watershed ANNAGNPS Modeling Project- Final Report (PDF; 2.43 MB)

Executive Summary

This report describes an interagency effort, funded under the authority of Section 516(e) of the Water Resources Development Act (WRDA) of 1996, to apply a watershed model, AnnAGNPS, to the Blanchard River Watershed, Ohio. The goal of the modeling effort was to predict direct runoff as well as sediment and nutrient loading from the highly agricultural watershed. A set of potential land management alternatives were evaluated to estimate the potential benefits in terms of reduced sediment and nutrient loading.

The watershed was represented in the model using data from several sources. A 30 m DEM was used to delineate the watershed into 3,830 subwatershed cells with an average area of 52 ha. Spatial information and attribute data from SSURGO and NASIS databases were used to define soil conditions. Stream channel geometry was based on a collection of surveyed cross-sections in the Blanchard and neighboring watersheds. A four-year crop and tillage rotation data layer was developed based on remote sensing data. Crop land management practices and fertilizer/manure application rates were defined in the model based on local knowledge. Point source loads from 13 permitted discharges were also included.

The model was calibrated against observed stream flow and water quality data for the period from 2002-2009. A model confirmation was also conducted using best available data from 1995-2001. For the calibration period, the model prediction of direct runoff was good, yielding R2 and Nash-Sutcliffe model efficiencies (NSE) greater than 0.75 on an annual basis, and ranging from 0.63 to 0.69 on a monthly basis. Percent error and percent different calculations were both less than 20% and met the calibration target. Visual comparison of model results indicated an underprediction of runoff in the late winter/early spring period, potentially attributed to the model’s limitations in modeling a change in infiltration under frozen soil conditions.

The simulation of suspended sediment yield and loading was good, with NSE and R2 values greater than 0.86 on an annual basis and near 0.40 on a monthly basis. Similar to the direct runoff predictions, the model under-predicted sediment during the late winter/early spring period. AnnAGNPS estimated that ephemeral gully erosion accounted for approximately 85% of the total landscape erosion in the watershed, while sheet and rill erosion amounted to the remaining 15%. The model simulated total phosphorus and total nitrogen loading in the watershed with less accuracy than direct runoff or suspended sediment.

A set of land management alternatives were run including tile drain management, conservation tillage, cover crops, conversion of crops to grassland, and improved nutrient management. A pre-settlement “all natural” watershed scenario was also developed. In general, all scenario runs showed reasonable reductions in suspended sediment. For example, the model estimated a suspended sediment loading reduction of 54% with a conversion of 10% of highest eroding cropland to grassland, and a 60% reduction for a combined management scenario involving conservation tillage, conversion of crop to grassland, and improved nutrient management. The model Blanchard Watershed ANNAGPS Modeling Final Report October 19, 2010 LimnoTech Page ES-2 predicted a sediment loading reduction of 99.8% under an all-natural watershed condition.

Simulation of phosphorus and nutrient loading reductions under proposed land management was reasonable for most scenarios. A cover crop scenario resulted in an estimated 25% reduction of total phosphorus and 39% reduction of total nitrogen. The model predicted that a 60% reduction of fertilizer application could result in a 21% decrease in total phosphorus and 60% decrease in total nitrogen loading. The model produced unexpected total phosphorus results for scenarios involving the conversion of cropland to grassland or forest. Model diagnostic runs suggest that phosphorus in non-crop land uses are represented almost entirely in a dissolved form which continually leaches out of cells during the simulation period. These results suggest that the phosphorus cycling algorithms within AnnAGNPS warrant further investigation.

This modeling exercise was a successful attempt at quantifying direct runoff and suspended sediment loading contributions from the Blanchard River watershed under baseline and potential management scenarios. The simulation of nutrient loading from the watershed under most management scenarios was informational; however, model nutrient calculations related to conversion of cropland to non-cropland land uses were problematic.

The application of AnnAGNPS to the Blanchard River watershed was a detailed analysis for a complicated problem over a larger watershed system. Because of the number of watershed cells and the complexity of supporting databases (e.g., crop and tillage rotations), a high level of resources was expended for model set-up, execution, and interpretation of model results. A simplified model configuration (e.g., smaller watershed or coarser spatial scale) which did not involve calibration may have required fewer resources.

The simulation of ephemeral gullies for delivery of sediments and associated nutrients is an important process captured in AnnAGNPS which is not an element of many other watershed models. However, additional empirical observations of ephemeral gully formation and erosion may help support the improvement of model process formulation. Simulation of nutrients within AnnAGNPS is less mature than algorithms, which model direct runoff and suspended sediment. Further investigation and testing of these processes would help to improve future applications of this model.