ArcGIS 10 (ESRI, Inc.) geographic information system software was used to spatially analyze the data. Water sampling locations were classified according to their selleck compound distance to the closest existing natural gas well, as well as their topographic position (valley vs. upslope). The samples were also classified by the geohydrologic units in which the water well was finished (bedrock formations vs. unconsolidated sand and gravel).

Locations of existing natural gas wells in Chenango County were obtained from the NYSDEC (NYSDEC, 2012), and a threshold of 1000 m was used to group water wells into ‘close’ or ‘far’ from a gas well (Osborn et al., 2011). Topographic position was determined using two methods. Following Molofsky et al. (2013), one method determined location in a valley according to distance to the nearest stream. Locations within 305 m (1000 feet) of a stream were considered to be valleys, where streams were defined using the USGS National Hydrography Dataset (NHD). A second approach focused on the geohydrologic setting and used surficial geology maps (Cadwell, 1991) and georeferenced USGS maps of valley-fill aquifers in Chenango County (McPherson, 1993) to classify ‘valley’ wells as those located in mapped valley-fill aquifers. LY294002 These approaches were similar to the methodology used by a recent USGS study in south-central New York; however, their valley

delineation factored in additional parameters including stream slope and elevation change between streams and adjacent uplands (Heisig and Scott, 2013). Well finishing geology in this study was determined as a specific bedrock formation

or unconsolidated sand and gravel fill by using information on well depth (as reported by the homeowner) along with depth to bedrock estimated from USGS survey maps (McPherson, 1993) and bedrock geology maps (Fisher C225 et al., 1970). Finishing geology was only determined for locations where well depth was reported by the homeowner. R (The R Project for Statistical Computing) was used for statistical analysis of the data. For statistical analysis of all analytes, values below the method detection limit were treated as being equal to their analyzed values ( Gilliom et al., 1984). The Mann–Whitney non-parametric test was used to analyze the dissolved gas data, as grouped according to proximity to gas wells and topographic position (valleys vs. upland). A non-parametric test was chosen due to the skewed distribution of the methane dataset and since log transformation of the data was not sufficient to normalize the distribution. For any analysis of δ13C-CH4 data, values were excluded for samples where the methane concentration was below the method detection limit of 0.01 mg L−1. The Kruskal–Wallis non-parametric test combined with a pairwise comparison (‘kruskalmc’ in R package ‘pgirmess’) was used where there were more than two groupings for methane data.