Generating nitrate response functions for large regional watersheds

Nitrogen fertilizers applied to agricultural fields mostly enter groundwater in the form of nitrates. Once in an aquifer, nitrates may have a transit time to a stream ranging from nearly zero to, in theory, infinity when entering adjacent to a water divide. The actual distribution of transit times between these two extremes is dependent on a variety of hydrologic factors. This transit time distribution for a watershed plays an important role in the timing and concentration of nitrates in the baseflow of the outlet stream; it controls the nitrate response function of a watershed. In this thesis, two techniques for modeling transit time distributions and nitrate response functions are considered: 1) the finite difference groundwater model MODFLOW with its companion particle tracking code MODPATH and 2) a lumped parameter model (LPM). MODFLOW-MODPATH accounts for spatial variations in input parameters but may be cost and data prohibitive on a large regional scale. Alternatively, the lumped parameter model is more easily applied to large scale watersheds, but cannot account for spatial variations in hydrology or nitrate inputs.;We compared the MODFLOW-MODPATH and LPM approaches in the Maurice Watershed in New Jersey, and found that they generate very similar transit time distributions. The discrepancy between the two distributions is mostly caused by the inability of the LPM to account for weak sinks, which are sinks that do not draw water from the entire aquifer thickness (e.g. low order tributaries and small capacity pumping wells). On the scale of the Maurice Watershed, this discrepancy is minor because the local effects of weak sinks tend to average out. However, the LPM may become a lesser approximation at smaller (subwatershed) scales where weak sinks can heavily influence the local transit time distribution. MODFLOW-MODPATH and the LPM also generate very similar nitrate response functions for the Maurice Watershed, attributable in part to the similar transit time distributions but also to the scattered land usage throughout the watershed. Using hypothetical simulations, we demonstrate that the LPM may become a lesser approximation when a particular land use is only found either adjacent to a stream or a water divide. Overall, this thesis provides supporting evidence for the applicability of the LPM to large regional scale watersheds with non-point source nitrate inputs.