Geochemical Analysis and Data-Driven Modeling of Lithium-Rich Oil and Gas Produced Waters in the United States
Faculty Mentor Name
Mary Kay Camarillo
Research or Creativity Area
Engineering & Computer Science
Abstract
Lithium is an increasingly important resource due to its role in energy storage technologies. This study investigates lithium-rich oil and gas produced waters as a potential alternative source using the U.S. Geological Survey (USGS) Produced Water Database. A subset of samples with lithium concentrations greater than 100 mg/L was analyzed to focus on potentially viable extraction targets.
Statistical analysis was performed on major and trace elements, including Ca, Mg, FeTot, Mn, Zn, Si, As, Ba, Pb, Sr, and Ni. Key geochemical ratios such as Ca/Li, Mg/Li, and (Mn+Zn)/Li were calculated to assess processing complexity and scaling potential. Spatial mapping techniques were used to visualize the distribution of lithium and related variables across the United States, revealing regional patterns.
To explore underlying structure in the data, K-means clustering was applied to identify groups of samples with similar geochemical characteristics. The results suggest that lithium-rich waters can be categorized into distinct types, reflecting different formation conditions. Additionally, linear regression was used to evaluate whether lithium concentrations could be predicted from Ca and Mg. The low R² value (~0.024) indicates that these variables alone have limited explanatory power, suggesting that lithium enrichment is controlled by more complex processes.
A missing data analysis highlighted gaps in key trace elements, which may impact model accuracy. Overall, this study demonstrates that integrating statistical analysis, geospatial visualization, and machine learning provides valuable insights into lithium distribution and supports improved resource evaluation.
Purpose
Our goal is to better understand the distribution and geochemical characteristics of lithium-rich oil and gas produced waters in the United States. Lithium concentrations were analyzed using samples greater than 100 mg/L, and we aimed to examine variability across different regions and chemical conditions, including potential solid waste generation and toxic elements. The long-term goal is to identify factors that influence lithium concentration and to provide insights that can be used to support more efficient resource evaluation and extraction strategies.
Results
A total of 319 samples with lithium concentrations greater than 100 mg/L were identified from the USGS Produced Water Database. These samples were distributed across multiple states, with the highest counts observed in Arkansas, Pennsylvania, and Texas. Summary statistics revealed substantial variability in major and trace element concentrations, particularly for calcium and magnesium, which are relevant to scaling and processing conditions.
Geochemical ratio analysis (Ca/Li, Mg/Li, and (Mn+Zn)/Li) showed a wide range of values, indicating differences in chemical complexity among lithium-rich samples. Spatial mapping demonstrated regional clustering of high-lithium samples, suggesting associations with specific geological basins.
K-means clustering grouped samples into distinct clusters based on multi-element geochemical data, reflecting differences in brine composition. Linear regression analysis using calcium and magnesium as predictors of lithium concentration resulted in a low coefficient of determination (R² ≈ 0.024), indicating weak predictive capability.
Missing data analysis revealed incomplete measurements for several trace elements, including lead (Pb) and arsenic (As), which limits the completeness of statistical modeling and interpretation.
Significance
This study is important because it evaluates lithium-rich produced waters as a potential alternative resource for lithium extraction in the United States. By combining geochemical analysis, spatial visualization, and machine learning, the work provides a data-driven framework for understanding variability in lithium distribution and associated chemical conditions.
The results highlight that lithium enrichment cannot be explained by simple relationships with major elements such as calcium and magnesium, suggesting that more complex geological and geochemical processes control lithium occurrence. This has important implications for developing predictive models and improving extraction strategies.
Additionally, the presence of scaling elements and toxic trace metals such as lead (Pb) indicates that lithium extraction may generate solid waste and environmental challenges that must be considered in process design. Identifying these factors helps inform more sustainable and efficient extraction approaches.
Overall, this work contributes to improved resource evaluation, highlights limitations in existing datasets, and supports future research in lithium recovery, environmental assessment, and energy sustainability.
Geochemical Analysis and Data-Driven Modeling of Lithium-Rich Oil and Gas Produced Waters in the United States
Lithium is an increasingly important resource due to its role in energy storage technologies. This study investigates lithium-rich oil and gas produced waters as a potential alternative source using the U.S. Geological Survey (USGS) Produced Water Database. A subset of samples with lithium concentrations greater than 100 mg/L was analyzed to focus on potentially viable extraction targets.
Statistical analysis was performed on major and trace elements, including Ca, Mg, FeTot, Mn, Zn, Si, As, Ba, Pb, Sr, and Ni. Key geochemical ratios such as Ca/Li, Mg/Li, and (Mn+Zn)/Li were calculated to assess processing complexity and scaling potential. Spatial mapping techniques were used to visualize the distribution of lithium and related variables across the United States, revealing regional patterns.
To explore underlying structure in the data, K-means clustering was applied to identify groups of samples with similar geochemical characteristics. The results suggest that lithium-rich waters can be categorized into distinct types, reflecting different formation conditions. Additionally, linear regression was used to evaluate whether lithium concentrations could be predicted from Ca and Mg. The low R² value (~0.024) indicates that these variables alone have limited explanatory power, suggesting that lithium enrichment is controlled by more complex processes.
A missing data analysis highlighted gaps in key trace elements, which may impact model accuracy. Overall, this study demonstrates that integrating statistical analysis, geospatial visualization, and machine learning provides valuable insights into lithium distribution and supports improved resource evaluation.
