A STATISTICAL COMPARISON OF PRE-DISTURBANCE VERSUS RECLAMATION WATER QUALITY AT A LIGNITE SURFACE MINE IN NORTHEAST TEXAS: A LUMPED PAIRS APPROACH

Near-surface lignite deposits in Texas are primarily associated with the Eocene-age Wilcox Group and lie within the recharge zone of the prolific Carrizo-Wilcox aquifer. Surface mining of these deposits can adversely affect groundwater quality by altering the redox conditions of the sediments, which can initiate or increase the oxidative dissolution of reduced iron sulfide minerals, releasing acidity as well as dissolved iron and sulfate ions. The presence of groundwater-fed, low pH seeps has delayed the release from bond of some reclaimed lands in the Sabine Uplift area of northeast Texas as federal regulations require that surface waters draining reclaimed mines have a stable pH between 6 and 9. However, this region has a history of low pH surface water and shallow groundwater which may be due to the natural weathering of iron sulfide-rich geologic units.

Using the Wilcoxon Rank Sum Test, this study tests for a lack of statistically significant changes of dissolved sulfate and of the dissolved sulfate to iron ratio (DSIR; a proxy for O₂ availability) in paired groups of water quality data collected at the Oak Hill Mine in northeast Texas. The Rank Sum Test ignores the ordinality of time series observations, instead treating two time series as a paired group of pre-disturbance and post-disturbance (reclamation) observations. In a previous study, 24 pairs were analyzed for statistically significant changes in groundwater pH using the Rank Sum Test. Inability to reject the null hypothesis was interpreted as evidence that groundwater pH is stable for that temporal pair. For this analysis I hypothesize that temporal pairs with an apparently stable pH did not experience significantly altered redox conditions after mining, and consequently that these pairs will show evidence of stable DSIRs and sulfate concentrations.

However, the results do not support this hypothesis. Additional investigation using simple linear regression on pH as a function of sulfate concentration indicates that, while the two variables are correlated, sulfate can explain no more than 20% of the variation in pH (p = 8.4E-41). Therefore, I conclude that paired before-and-after group analyses are not well-suited to this field site, perhaps because the expected correlations are being partially obscured by independent and secondary chemical reactions.