Predicting Nitrogen in Streams: A Comparison of Two Estimates of Fertilizer Application

Decision makers frequently rely on water and air quality models to develop nutrient management strategies. Obviously, the results of these models (e.g., SWAT, SPARROW, CMAQ) are only as good as the nutrient source input data and recently the Nutrient Innovations Task Group has called for a better accounting of nonpoint nutrient sources. Currently, modelers frequently rely on county level fertilizer sales records combined with acreage of crops to estimate nitrogen sources from fertilizer for counties or watersheds. However, since fertilizer sales data are based on reported amounts they do not necessarily reflect actual use on the fields. In addition the reported sales data quality varies by state resulting in differing accuracy between states. In this study we examine an alternative method potentially providing a more uniform, spatially explicit, estimate of fertilizer use. Our nitrogen application data is estimated at a 30m pixel resolution which allows for scalable inputs for use in water and air quality models. To develop this dataset we combined raster data from the National Cropland data layer (CDL) data with the National Land Cover Data (NLCD). This process expanded the NLCD's 'cultivated crops' classes to included major grains, cover crops, and vegetable and fruits. The Agriculture Resource Management Survey chemical fertilizer application rate data were summarized by crop type and year for each state, encompassing the corn, soybean, spring wheat, and winter wheat crop types (ARMS, 2002-2005). The chemical fertilizer application rate data were then used to estimate annual application parameters for nitrogen, phosphate, potash, herbicide, pesticide, and total pesticide, all expressed on a mass-per-unit-crop-area basis for each state for each crop type. By linking crop types to nitrogen application rates, we can better estimate where applied fertilizer would likely be in excess of the amounts used by crops or where conservation practices may improve retention and uptake helping offset the impacts to water. To test the accuracy of our finer resolution nitrogen application data, we compare its ability to predict nitrogen concentrations in streams with the ability of the county sales data to do the same.