Reconstruction of continental palaeoclimate and palaeohydrology is currently hampered by limited information about isotopic patterns in the modern hydrologic cycle. To remedy this situation and to provide baseline data for other isotope hydrology studies, more than 4800, depth- and width-integrated, stream samples from 391 selected sites within the USGS National Stream Quality Accounting Network (NASQAN) and Hydrologic Benchmark Network (HBN) were analysed for δ¹⁸O and δ²H (http://water.usgs.gov/pubs/ofr/ofr00-160/pdf/ofr00-160.pdf). Each site was sampled bimonthly or quarterly for 2·5 to 3 years between 1984 and 1987. The ability of this dataset to serve as a proxy for the isotopic composition of modern precipitation in the USA is supported by the excellent agreement between the river dataset and the isotopic compositions of adjacent precipitation monitoring sites, the strong spatial coherence of the distributions of δ¹⁸O and δ²H, the good correlations of the isotopic compositions with climatic parameters, and the good agreement between the ‘national’ meteoric water line (MWL) generated from unweighted analyses of samples from the 48 contiguous states of δ²H=8·11δ¹⁸O+8·99 (r²=0·98) and the unweighted global MWL of sites from the Global Network for Isotopes in Precipitation (GNIP) of the International Atomic Energy Agency and the World Meteorological Organization (WMO) of δ²H=8·17δ¹⁸O+10·35.

The national MWL is composed of water samples that arise in diverse local conditions where the local meteoric water lines (LMWLs) usually have much lower slopes. Adjacent sites often have similar LMWLs, allowing the datasets to be combined into regional MWLs. The slopes of regional MWLs probably reflect the humidity of the local air mass, which imparts a distinctive evaporative isotopic signature to rainfall and hence to stream samples. Deuterium excess values range from 6 to 15‰ in the eastern half of the USA, along the northwest coast and on the Colorado Plateau. In the rest of the USA, these values range from −2 to 6‰, with strong spatial correlations with regional aridity. The river samples have successfully integrated the spatial variability in the meteorological cycle and provide the best available dataset on the spatial distributions of δ¹⁸O and δ²H values of meteoric waters in the USA.