Regression Analysis to Predict Selenium Levels at Two Surface Coal Mines in the Powder Basin, Wyoming
Journal American Society of Mining and Reclamation
Selenium (Se) is an element of interest in plant and animal nutrition because of the narrow range between essential and toxic levels. Many areas being mined in the Powder River Basin of northeastern Wyoming contain pockets ofhigb Se concentrations within the overburden material. Therefore, it has become increasingly necessary for industcy and state regulatory personnel to ti)' to quantify relationships between pre-existing soil Se levels to postmining levels in both soils and plants. A study
... plants. A study was initiated in 1991 to investigate the relationship of plant Se uptake · and soil/backfill Se levels at two active coal mines in the Powder River Basin, Wyoming. Soil and vegetation samples were collected in 1991 and 1992. Soil/backfill Se levels were determined by five methods: total Se and hot water, AB-DTP A, saturated paste, and dihydrogen phosphate extractable Se. Total plant Se was also determined. Plant Se levels of four vegetation lifeforms were regressed on eleven soil variables to construct appropriate models for assessing.plant-soil Se relationships. These regression analyses were conducted with soil depth, vegetation type (native versus reclaimed), and mine (large mine (Mine L) and (small mine (Mine S)) as important subcategories. Depth and type were significant in determining statistical relationships. Simple linear regression models were developed, but the majority of the slopes were not significantly different at the 0.05 probability level. Multiple linear regression models revealed that soil Se and pH were the most important predictors of plant Se levels for native areas; no specific parameter was dominant in reclaimed area analysis. The R 2 's for native areas were improved over the multiple linear models by deriving polynomial regression models. Polynomial regression models for the reclaimed areas resulted in marginal improvement ofR 2 values over the multiple linear regression models. Whereas hot water soluble Se appears to be a better predictor of plant Se concentrations, both AB-DTP A and phosphate extractable Se were also good predictors. The best statistical relationships were also determined with depths 2 and 3 of native areas. Inclusion of age of reclamation, however, improved the polynomial models for reclaimed areas. Additional Key Words: Soil-plant selenium relationships, regression models, predictive models.