A tribute to tributaries: River studies elucidate links between human activity and nutrient export across a broad range of watersheds

Abstract

In my early scientific career in environmental science I was exposed to a great deal of research on N cycling and transport in forests, wetlands, and coastal waters. Nitrogen was a hot topic because between the 1930s and 1980s, human alteration of the N cycle had increased dramatically. Nitrogen inputs increased not only to agricultural systems (from fertilizers) but also in precipitation (from fossil fuel combustion), to forested and coastal systems (Vitousek et al. 1997). In addition coastal systems receive nitrogen inputs from groundwater and rivers and the cumulative increase in N loads increased primary production and associated problems of eutrophication in these systems (Vitousek et al. 1997). Much of the research on nitrogen at this time was carried out at small scales (small watersheds, experimental plots, etc.) where detailed and accurate budgets could be made and experimental manipulations were feasible (Likens et al. 1970). In comparison there were few studies at the regional scale despite the fact that N pollution is a regional to global-scale problem. Large rivers offer the opportunity to study the impact of increasing N loads at the regional scale. Further when watershed characteristics are simultaneously collected the controls of N export can be examined by correlation or simple models. This crosssystem approach was used in the featured Ambio paper (Caraco and Cole 1999a). The Ambio work was an extension of an earlier crosssystem correlative study that I worked on with colleagues from The Institute of Ecosystem Studies. This study showed that nitrate export from large rivers was well correlated to human population density in the watersheds of those rivers (Peierls et al 1991). In the Ambio paper, we expanded on the population correlation to build a simple model of nitrate export that could do three things: 1. improve the relationship to human activity (e.g., increase the predictive power); 2. attribute the N exports to different human activities that are the sources of N; and 3. calculate the fraction of N load to the watershed that is exported as nitrate. This model was similar to one that I developed earlier for phosphate export from rivers (Caraco 1995). The model suggested that N sources to rivers varied substantially between rivers but on average fertilizer was the most important source (ca. 50%) followed by precipitation and point sources (both near 25%). The model also showed that fraction of N load exported from the watershed as nitrate was explained relatively well by water runoff alone and did not appear to be related strongly to N load. For example, for the heavily impacted Rhine River, less than 20% of the N loaded to the watershed as fertilizer and precipitation is exported to the river as nitrate. Nitrogen saturation is defined by the idea that if N loading of the watershed goes high enough, the retention of N in the watershed could decrease substantially (Aber 1992). Our study found little evidence for this saturation but we suggested that future saturation could have devastating effects, increasing dramatically N export even while N inputs to watersheds remained constant. While the Ambio paper presented a highly predictive model, it was deficient in some ways. One obvious problem was that this paper modeled nitrate and not total N export by rivers. This was done in part as there was a much larger dataset available for nitrate than for total N, but we felt this was justified in that nitrate export appeared to be the most reactive to human impacts examined here. Another problem was that the inputs to the watershed were incomplete. We did not consider N exchange in crop transport or fixation in forested or agricultural systems. We hypothesized that perhaps the N originating from atmospheric N fixation was more likely to be exported as organic N forms and this would be consistent with organic N dominating N export in many pristine systems. Further, while the modeled