Spatial distribution and interannual trends of δ~(18)O, δ~2H, and deuterium excess in precipitation across North-Eastern Italy

摘要

NE Italy presents high-quantity, high-quality and easily exploitable groundwater resources that are seriously threatened by anthropogenic pressures. This study analyzes the oxygen and hydrogen isotopic composition of 2250 precipitation samples collected in 36 sites across the north-easternmost region of Italy, Friuli Venezia Giulia, between 1984 and 2015. This is an unprecedented dataset for North Italy with a high density of sampling sites and a decadal temporal extension. A series of both routine and original chemometric approaches were applied to investigate the temporal and spatial variability of the isotopic composition through relationships with geographical and weather variables. New statistical approaches were presented to model the seasonal and spatial patterns of isotopic composition as well as to summarize the large amount of isotopic data. Significant gradients of δ~(18)O and δ~2H were detected in the area due to the peculiar orography and climate of the region; the amplitude of the monthly patterns also presented similar gradients. The deuterium excess did not present a clear seasonality, but higher values were found in autumn. The deuterium excess-to-δ~(18)O ratio exhibited typical patterns throughout the region when grouping the sites for altitude and continentality; in winter, large differences of δ~(18)O were detected among groups, but deuterium excess remained almost unchanged. The inter-site correlations were moderately high across all the territory even for deseasonalized data. The local meteoric water line (using all the single samples δ~2H = 7.8?δ~(18)O + 8.9) was also estimated at annual and seasonal basis, evidencing the presence of spatial gradients according to the orographic and weather characteristics of the region. Statistically significant increasing interannual trends (0.23-0.87‰/y) were detected in 11 sites for deuterium excess; the presence of these trends was linked to local processes. Two multiple linear regression models we