Source identification of nitrate by means of isotopic tracers in the Baltic Sea catchments
- 1Baltic Sea Research Institute, Seestr. 15, 18119 Rostock, Germany
- 2Stockholm University, Department of Systems Ecology, 10691 Stockholm, Sweden
- 3Stockholm University, Department of Applied Environmental Science, 10691 Stockholm, Sweden
- 4Finnish Environment Institute, P.O. Box 140, 00251 Helsinki, Finland
- 5Sea Fisheries Institute, Kołłątaja 1, 81-332 Gdynia, Poland
- 6Ruhr Universität Bochum, Universitätsstr. 150, 44780 Bochum, Germany
Abstract. Nitrate input to a river is largely controlled by land use in its catchment. We compared the information carried by the isotopic signatures of nitrate in 12 Baltic rivers, in relation to the vegetation cover, land use, and fertilization of agricultural land of their catchments. We found isotope values in nitrate ranging from −2 to 14‰ for δ15N and 8 to 25‰ for δ18O. The annual variability of riverine nitrate isotope signatures is presented in detail for one Nordic, the Kemijoki, and two southern rivers, the Vistula and Oder. Nordic rivers with relatively pristine vegetation in their catchments show not only low δ15N values and high δ18O-NO3− but also lower annual variability than rivers draining densely populated land. Seasonal signals were found in all the rivers. We used load weighted nitrate isotope data and data from the three major N sources (farmland/sewage, atmospheric deposition and from runoff of pristine soils) to theoretically estimate the shares of nitrate from these sources. The results of an isotope mixing model (IMM-1) agree reasonably well with the same estimates for agricultural land derived from a Global Land Cover (GLC) data base, with a deviation varying from −16% to +26%. The comparison with an emission model (EM) reveals relatively good agreements for intensively used catchments (−18 to +18% deviation). Rather unsatisfactory agreement was found between the IMM-1 and GLC calculations for pristine catchments (−36 to +50% deviation). Advantages and limitations of the tested model are discussed.