Evaluation of soil, unsaturated, and saturated zone parameter uncertainty using GSFlow and PEST in an agricultural watershed

A coupled surface-ground water hydrological model of the Burley-DeMerritt Organic Dairy Research Farm in southeastern New Hampshire is under continued development in support of a long-term mission to understand nutrient dynamics and water use in sustainable New England dairy operations. To build on previous simulations of ground water recharge and nitrogen transport, an estimate of net recharge under an array of climate scenarios is required to facilitate modeling of nutrient dynamics for the projected life span of the dairy farm. The model must therefore incorporate spatially distributed surface and soil zone processes that influence the shallow ground water system. GSFlow couples the USGS Precipitation Runoff Modeling System and MODFLOW codes and is used to simulate surface, soil, and subsurface hydrological processes using a suite of empirical and process-based algorithms and parameters. Topography of the 83 hectare model domain was derived from a 1-meter horizontal resolution LiDAR DEM with centimeter-scale accuracy. Zonation at the soil surface was derived from detailed soils mapping, aerial land cover assessment, and drainage boundaries derived from the LiDAR DEM. Meteorological forcing data are taken from nearby (5 km) NCDC and AIRMAP meteorological towers. The farm's catchment consists of regionally common land covers including pasture, forest, and forested wetland. An array of surface model structures and spatial discretizations are evaluated ranging from fine scale, incorporating areas of consistent land cover or soil types, to the catchment scale. Parameter identification and uncertainty for both PRMS and MODFLOW components is conducted using PEST software, where distributed measurements of hydraulic head, soil moisture, and streamflow are weighted by measurement uncertainty and the relative measurement abundance or redundancy to define a model to measurement misfit objective function. Model validation will be conducted against data collected in the summer and fall of 2011, including improved resolution streamflow data from a recently installed Parshall flume on the small stream draining the majority of the farm. The relative contribution of the three measurement types to the overall model objective function, and parameter uncertainty from the PRMS soil zone, and MODFLOW saturated and unsaturated zones are used to assess data needed to improve the validity of water balance estimates, and suggest which model components may be least determined in similar coupled models.