An outlier detection approach for water footprint assessments in shale formations: case Eagle Ford play (Texas)

The increasing trend on water use for hydraulic fracturing (HF) in multiple plays across the U.S. has raised the need to improve the HF water management model. Such approaches require good-quality datasets, particularly in water-stressed regions. In this work, we presented a QA/QC framework for HF data using an outlier detection methodology based on five univariate techniques: two interquartile ranges at 95 and 90% (PCTL95, PCTL90), the median absolute deviation (MAD) and Z score with thresholds of two and three times the standard deviation (2STD, 3STD). The cleaning techniques were tested using multiple variables from two data sources centered on the Eagle Ford play (EFP), Texas, for the period 2011–2017. Results suggest that the PCTL95 and MAD techniques are the best choices to remove long-tailed statistical distributions of different variables, classifying the minimum number of records as outliers. Overall, outliers represent 13–23% of the total HF water volume in the EFP. In addition, outliers highly impacted minimum and maximum HF water use values (min–max range of 0–47 m³/m and 5.3–24.6 m³/m of frac length, before and after the outlier removal process, respectively), that are frequently used as a proxy to develop future water–energy scenarios in early-stage plays. The data and framework presented here can be extended to other plays to improve water footprint estimates with similar conditions.