Abstract
Abstract. A regionalized cluster-based water isotope prediction (RCWIP) approach, based on the Global Network of Isotopes in Precipitation (GNIP), was demonstrated for the purposes of predicting point- and large-scale spatio-temporal patterns of the stable isotope composition (δ2H, δ18O) of precipitation around the world. Unlike earlier global domain and fixed regressor models, RCWIP predefined 36 climatic cluster domains and tested all model combinations from an array of climatic and spatial regressor variables to obtain the best predictive approach to each cluster domain, as indicated by root-mean-squared error (RMSE) and variogram analysis. Fuzzy membership fractions were thereafter used as the weights to seamlessly amalgamate results of the optimized climatic zone prediction models into a single predictive mapping product, such as global or regional amount-weighted mean annual, mean monthly, or growing-season δ18O/δ2H in precipitation. Comparative tests revealed the RCWIP approach outperformed classical global-fixed regression–interpolation-based models more than 67% of the time, and clearly improved upon predictive accuracy and precision. All RCWIP isotope mapping products are available as gridded GeoTIFF files from the IAEA website (www.iaea.org/water) and are for use in hydrology, climatology, food authenticity, ecology, and forensics.
Highlights
Spatial patterns in stable hydrogen and oxygen isotope ratios of precipitation were first observed in the 1950s (Epstein and Mayeda, 1953; Friedman, 1953; Dansgaard, 1954; Craig, 1961), and increasingly revealed as long-term δ2H and δ18O data sets from around the world accumulated in the International Atomic Energy Agency’s (IAEA) global network of isotopes in precipitation (GNIP – Fig. 1) (Dansgaard, 1964; Rozanski et al, 1993; Aggarwal et al, 2010; IAEA/WMO, 2013)
The past decade has experienced increasing demand for accurate spatio-temporal predictions of point, regional, and continental-scale δ2H and δ18O values in precipitation, especially for some regions where little or no Global Network of Isotopes in Precipitation (GNIP) data existed. This demand for isotopic predictive capacity arose from the ecological, wildlife, food source traceability, and forensic sciences after it was revealed that the δ2H values of some plant, animal, and human tissues closely mirrored isotopic patterns of precipitation (Hobson and Wassenaar, 1997; Bowen et al, 2005a)
regionalized cluster-based water isotope prediction (RCWIP) differed from this key earlier work in that it used the best performing model from a suite of regionalized domain multivariate regression equations to determine the isotopic composition of precipitation at a known location i as a function of the selected predictor variables available
Summary
Spatial patterns in stable hydrogen and oxygen isotope ratios of precipitation were first observed in the 1950s (Epstein and Mayeda, 1953; Friedman, 1953; Dansgaard, 1954; Craig, 1961), and increasingly revealed as long-term δ2H and δ18O data sets from around the world accumulated in the International Atomic Energy Agency’s (IAEA) global network of isotopes in precipitation (GNIP – Fig. 1) (Dansgaard, 1964; Rozanski et al, 1993; Aggarwal et al, 2010; IAEA/WMO, 2013). The past decade has experienced increasing demand for accurate spatio-temporal predictions of point, regional, and continental-scale δ2H and δ18O values in precipitation, especially for some regions where little or no GNIP data existed This demand for isotopic predictive capacity arose from the ecological, wildlife, food source traceability, and forensic sciences after it was revealed that the δ2H (and δ18O) values of some plant, animal, and human tissues closely mirrored isotopic patterns of precipitation (Hobson and Wassenaar, 1997; Bowen et al, 2005a).
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have