Geochemical evolution of produced waters from hydraulic fracturing of the Marcellus Shale, northern Appalachian Basin: A multivariate compositional data analysis approach

Multivariate compositional data analysis methods were used to investigate geochemical data for water injected during hydraulic fracturing and for water produced from 19 Marcellus Shale gas wells in the northern Appalachian Basin. The data were originally published as part of an industry report. The analysis was adapted to consider the compositional nature of the data and avoid potentially spurious correlations present in raw concentration data through the application of log-ratio transformations. Techniques such as robust variation arrays, robust principal component analysis, and relative variation plots were applied to log-ratio transformed data. Results from this battery of multivariate tools indicate that two primary processes affect the chemical evolution of the water returned to the surface during the first 90 days of production: mixing of injected water with formation brines of evaporated paleoseawater origin and injection of sulfate-rich water during hydraulic fracturing may stimulate sulfate reduction at some sites. Spatial variability in sulfate/alkalinity ratios appears to influence variations in geochemical controls on strontium versus barium with elevated proportions of strontium being found in more bicarbonate-poor environments, while barium is a larger proportion in sulfate-poor areas. Comparison of results using a log-ratio approach versus the more common analysis of concentration data reveals both similarities and some marked differences in the resulting interpretations. Results from this work are important in terms of both demonstrating methods to avoid mathematical inconsistencies from using raw brine geochemical data and to further investigate the geochemical controls on produced waters generated from shale gas reservoirs.