The NADP network collects precipitation weekly at ~215 sites across the U. S. for chemical analysis [http://nadp.sws.uiuc.edu]. The samples are archived for five years before being discarded. Welker began collecting these discarded samples in 1994 resulting in network samples that begun in 1989. To verify that these samples preserved their original isotope values, we measured the oxygen ( d 18O) and hydrogen ( dD) isotope ratios, and used the deuterium excess (d-excess), which shifts during evaporation and vapor exchange, as a quality control tool to remove bad samples. Only about 5% of the samples could be identified as probably altered; that is, outside the typical d-excess range of -5‰ to 25‰ [Welker, 2000; Harvey and Welker, 2000]. As a second, check on the quality of these samples, metoric water lines where generated for each site and evaluated as to whether they were within the range of MWL published for other stations across the globe. Here, we report on over 10,000 weekly isotopic analyses from ~80 sites for the period of July 1989 through December 1995. This data set represents by far the most spatially and temporally complete isotopic data set of precipitation ever compiled for the continental U. S. We chose to analyze samples from these sites due to mass spectrometer instrumentation limitations and spatial overlap between sites. Analysis sites were chosen that are part of existing networks such as the LTER [Long Term Ecological Research-NSF-[http:\\www.lternet.edu] where other ancillary data is available and sites where chosen to provide biome-wide representation so that the data sets and findings would be applicable to the broadest range of scientists and citizens. At present the web site only contains the δ 18O weighted annual averages over our 6 year sampling period.

In this presentation, we are focusing on attributes of the monthly, seasonal and annual averages. As expected, for any given site, 52 weeks of precipitation for the years, 1989, 1991, 1992, 1993, 1994 and 1995 do not exist, nor do they for other isotope network data sets and published findings [Rozanski et al. 1992; Rozanski et al. 1993]. We have taken the following steps to assure that the numerical presentation and the reliability of the results are robust and reliable, given that not all weeks had precipitation for d 18O and dD analysis. In order to conform to the monthly average framework of sample collection and analysis of GNIP (Rozanski et al. 1993), we averaged the weekly SIP measurements into amount-weighted monthly values. Unfortunately, not all weeks for which precipitation occurred during the 1989 to 1995 interval, have yet been analyzed for either d 18O or dD. Approximately 60% of the months for which any isotopic measurements were made were “complete” or all possible weekly samples were analyzed for SIPs. Using only complete months (i.e. 4 values per month) would greatly reduce the density of the isotopic measurements, thus we also calculated the monthly averages from the incomplete months.

To assure ourselves of the robustness of these incomplete months, we determined that that 14%, 33%, 27%, 17% and 10% were characterized by 1, 2, 3, 4 or 5 weeks of precipitation samples, respectively. We found that the accuracy of isotopic values, derived from incomplete months, in representing complete monthly averages decreased linearly with the fraction of samples used in the calculations. So as to limit the number of monthly averages with high uncertainties, we only included monthly averages for which at least 50% of the weekly samples per month were analyzed. With this criteria, over 90% of the average standard deviations for each site ranged from 0.4‰ and 0.6‰ d 18O. 2,056 monthly isotopic values were used in this presentation.

While the majority (70%) of our d 18O values are derived from explicit analysis, a portion of the values used in this presentation were calculated from dD values using site specific MWL. This was necessary as at times our analytical capacity to complete dD analyses exceed the d 18O analyses. ArcGIS (ESRI) was the program used to interpolate d 18O-values between the 61 sampling sites. Spherical semi-variogram model kriging was the interpolation scheme applied, based on the 12 sites closest to the location where d 18O-values are being calculated. The interpolation maps are meant to describe probable isotopic values for all points within a grid, based on spatial functions calculated from site-specific empirical data. Lambert’s conformal mapping projection was chosen to describe the spatial relationships because of its ability to best represent shapes at mid latitudes. NAD83 was the datum used.

The time period during which our samples were collected happens to be inclusive of divergent SOI phases consisting of two Neutral phases (1989 & 1995) and multiple years of El Ni ño (1990, 1991, 1992, 1993, 1994) based on the Western Regional Climate Center SOI indexes (http://www.wrcc.dri.edu/). These two phases are based on the NOAA depictions of the SOI and reflect the pressure differences between Tahiti and Darwin, June through November. While there exists a multitude of indices, this one is consistent with others presented by NOAA and other agencies.