Modeling land-based nitrogen loads from groundwater-dominated agricultural watersheds to estuaries to inform nutrient reduction planning

Excessive nitrate loads from intensive potato production have been linked to the reoccurring anoxic events in many estuaries in Prince Edward Island (PEI), Canada. Community-led watershed-based nutrient reduction planning has been promoted as a strategy for water quality restoration and initial nitrate load criteria have been proposed for the impacted estuaries. An integrated modeling approach was developed to predict base flow nitrate loads to inform the planning activities in the groundwater-dominated agricultural watersheds. Nitrate load is calculated as base flow multiplied by the average of nitrate concentration at the receiving watershed outlet. The average of nitrate concentration is estimated as the integration of nitrate leaching concentration over the watershed area minus a nitrate loss coefficient that accounts for long-term nitrate storage in the aquifer and losses from the recharge to the discharge zones. Nitrate leaching concentrations from potato rotation systems were estimated with a LEACHN model and the land use areas were determined from satellite image data (2006-2009) using GIS. The simulated average nitrate concentrations are compared with the arithmetic average of nitrate concentration measurements in each of the 27 watersheds for model calibration and in 138 watersheds for model verifications during 2006-2009. Sensitivity of the model to the variations of land use mapping errors, nitrate leaching concentrations from key sources, and nitrate loss coefficient was tested. The calibration and verification statistics and sensitivity analysis show that the model can provide accurate nitrate concentration predictions for watersheds with drainage areas more than 5 km(2) and nitrate concentration over 2 mg N L-1, while the model resolution for watersheds with drainage areas below 5 km(2) and/or nitrate concentration below 2 mg N L-1 may not be sufficient for nitrate load management purposes. Comparisons of normalized daily stream discharges among the active hydrometric stations indicated that stream base flow could be prorated for nitrate load calculation from the nearest gauging station in the absence of discharge measurements. Annual nitrate losses, including aquifer long-term storage, denitriflcation, and riparian plant uptake were estimated to be 0.8 mg N L-1, corresponding to 3.4 kg N ha(-1). The maximum and median base flow nitrate loads to the estuaries from among the 27 calibration watersheds were predicted to be 28.4 and 8.7 kg N ha(-1) respectively with a root mean square error (RMSE) as 2.3 kg N ha(-1). From among the 75 watersheds selected for model verification, the maximum and median base flow nitrate loads to the estuaries were estimated to be 29 and 5.5 kg N ha(-1) respectively with RMSE as 2.6 kg N ha(-1). At the estuaries with nitrate loads above the medians, the predominant nitrate sources (75-98%) were derived from the potato rotation lands, highlighting the importance of N use management with potato production for water quality restoration; nitrate load derived from atmospheric N deposits was estimated to account for 3.6-13% of annual nitrate loads in watersheds with nitrate loads exceeding the median values. The application of the model to nutrient reduction planning in the Southwest River watershed implies that a significant change on cropping practices has to be made in order to mitigate the anoxic events in this highly impacted agricultural watershed. Crown Copyright (C) 2015 Published by Elsevier B.V. All rights reserved.