River nutrient loads and catchment size

We have used a total of 496 sample sites to calibrate a simple regression model for calculating dissolved inorganic nutrient fluxes via runoff to the ocean. The regression uses the logarithms of runoff and human population as the independent variables and estimates the logarithms of dissolved inorganic nitrogen and phosphorus loading with R-2 values near 0.8. This predictive capability is about the same as has been derived for total nutrient loading with process-based models requiring more detailed information on independent variables. We conclude that population and runoff are robust proxies for the more detailed application, landscape modification, and in-stream processing estimated by more process-based models. The regression model has then been applied to a demonstration data set of 1353 river catchments draining to the sea from the North American continent south of the Canadian border. The geographic extents of these basins were extracted from a 1-km digital elevation model for North America, and both runoff and population were estimated for each basin. Most of the basins (72% of the total) are smaller than 10(3) km(2), and both runoff and population density are higher and more variable among small basins than among larger ones. While total load to the ocean can probably be adequately estimated from large systems only, analysis of the geographic distribution of nutrient loading requires consideration of the small basins, which can exhibit significant hydrologic and demographic heterogeneity between systems over their range even within the same geographic region. High-resolution regional and local analysis is necessary for environmental assessment and management.