Date of Award


Degree Type


Degree Name

Master of Science (MS)


Agricultural and Biological Engineering

Committee Chair

Jane Frankenberger

Committee Member 1

Laura Bowling

Committee Member 2

Venkatesh Merwade


Estimation of nutrient load is critical for many applications in water quality management; however, infrequent data monitoring and measurement error could raise considerable uncertainty in the load estimations. The objectives of this study were to quantify the overall uncertainty in annual nitrate-N load estimates, and develop a statistical model to predict nitrate loads based on subsurface drainage characteristics in Indiana watersheds. Nitrate was selected as the study object because of the high loads common in Midwestern streams, and its important influence on hypoxia in the Gulf of Mexico.

A standard error propagation method was used to quantify the uncertainty from both measurement and load calculation processes to evaluate the accumulative effect. Results showed that the estimated measurement uncertainty, expressed as coefficient of variation (CV), ranged from 11.7 to 12.4%, and the load calculation uncertainty CV for a 30-day frequency ranged from 7% to 32%. The estimated overall uncertainty ranged from 14 to 34% in annual nitrate-N load estimates for a 68% confidence interval.

Load estimation uncertainty was found to be affected by watershed size and streamflow flashiness. Smaller watershed size often lead to greater uncertainty in load estimates; and the R-B index and the hydrologic reactivity index were found to be significantly positively related to the load estimates uncertainty, while the autocorrelation coefficient (Lag 1) indicated a negative linear relationship.

A statistical model was developed based on the linear relationship between the flowweighted nitrate-N concentration from nonpoint sources and tile drained area percentage in Indiana watersheds. The linear relation is strong for the annual model and for monthly models from December to July, and model was found to especially suitable for medium and highly drained watersheds. Therefore, this model can be used as a simple and effective tool in estimating nitrate loads for unmonitored Midwestern tile-drained watersheds, and the potential for nitrate reduction when various tile drainage management techniques are employed.