Use of principal component analysis to extract environmental information from lake water isotopic compositions

• Published in: Limnology and Oceanography
• Main Authors: Kopec, B. G., Feng, X., Posmentier, E. S., Chipman, J. W., Virginia, R. A.
• Other Authors: National Science Foundation
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2018
• Subjects: Greenland; North Atlantic; North Atlantic oscillation
• Online Access: http://dx.doi.org/10.1002/lno.10776; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002%2Flno.10776; https://aslopubs.onlinelibrary.wiley.com/doi/pdf/10.1002/lno.10776

Abstract Lake water hydrogen (δD) and oxygen (δ 18 O) isotopic ratios have been widely used to study the hydrologic cycle. In a given region, δD and δ 18 O values from multiple lakes are often correlated and define a local evaporation line (LEL). The slope of and extension of each lake along this line contain information of many variables with entangled effects of important lake fluxes (e.g., evaporation). We present a new method to extract regional and local (a particular lake) environmental information based on principal component analyses (PCAs) of the δD‐δ 18 O relationships. Water of up to 30 lakes in West Greenland was sampled every summer for 5 yr, and measured for δD and δ 18 O values. For a given year, PCA yields values of the first (Enrichment) and second (Deviance) principal components for each lake. In addition, we obtain the slope of the line along the first eigenvector, which is similar, but not identical, to the LEL. We refer to this line as the regional isotopic enrichment line (RIEL), and demonstrate that the RIEL slope changes interannually with the dominant atmospheric circulation patterns affecting regional humidity, specifically the North Atlantic Oscillation and frequency of sea breezes. The Deviance of a lake correlates strongly with its longitude and surface area, both of which influence local humidity and isotopic ratios of vapor. Conversely, Enrichment carries little extractable environmental information because it is controlled by many variables of competing effects. This method can be widely applied to both modern hydrological studies and paleoclimate reconstructions using lake sediments.