**4. Discussion**

It is rational to expect groundwater atrazine concentration in Nebraska counties with continuously high atrazine usage to be significantly elevated or at least remain constant over time. Instead, low-level groundwater atrazine is frequently observed even though no groundwater atrazine elimination process was identified in these counties. This raises some issues addressed by this study. Before delving into these critical issues, the longterm significance of this study will be reiterated. The toxic or carcinogenic effects of atrazine are common knowledge due to evidence from experimental [19–23] and ecological

studies [4,24,25]. However, due to sparse epidemiological evidence [26,27], atrazine is often absolved of the supposed toxicity observed in experimental and ecological studies.

Meanwhile, it must be noted that only the use of individual-level atrazine exposure data would credibly predict the health outcomes associated with atrazine exposure. However, only a handful of studies are available for such designs. Hence most studies utilize county-level groundwater atrazine as exposure for potential disease outcomes. This may be downplaying the toxic effects of atrazine, since groundwater atrazine measurement reveals low-level atrazine concentration, which would interfere with identifying existing correlations between atrazine exposure and suspected pathological conditions when county-wide atrazine use data is used instead of individual atrazine exposure data. To this end, this study provided evidence for why the inclusion of groundwater atrazine measurements as explanatory variables for most models in epidemiological studies may fail to predict proposed atrazine-induced pathological conditions accurately.

The inferences drawn from this study may have direct human implications, given that approximately 60% of the sampled wells were domestic. Moreover, the domestic wells in this study were among the deepest. This is interesting and reassuring, since previous studies have observed correlations between better water quality and deeper wells [28]. Furthermore, the average atrazine amount detected in all the wells for the entire study period is significantly less than the United States Environmental Protection Agency Maximum Contaminant Level (MCL). Additionally, two primary atrazine metabolites, D.E.A. and D.I.A., were detected, suggesting atrazine degradation during the study period. While D.E.A. and D.I.A. are not the only atrazine metabolites, they were the only metabolites sufficiently detected during the study period. Hydroxyatrazine, another atrazine metabolite, was not detected, indicating dealkylation as the predominant metabolic pathway for atrazine degradation in the sampled groundwater. While abiotic pathway was previously reported for atrazine dealkylation, most atrazine dealkylation processes are attributed to biotic pathways [29]. This may suggest microbial co-contamination of the sampled groundwater [30–32]. Although microbial contaminant is not the focus of this study, this needs to be verified by future studies.

Atrazine depletion corresponded to D.E.A. and D.I.A. depletion in this study. Moreover, atrazine, D.I.A., and D.E.A. depletions were more apparent in the year 2000 than in other years. It is difficult to conclude any relationship between atrazine and D.I.A. or D.E.A. depletion, given that no baseline data for any of the pesticides was captured in this study. Furthermore, atrazine depletion was observed with the formation of its metabolites as time progressed. Although atrazine half-life in the sampled water supply wells may be challenging to determine, atrazine degradation to D.E.A. or D.I.A. contributes to atrazine depletion. Atrazine half-life depends on environmental factors. For example, it may range between 2 weeks and 16 weeks in surface soils. Moreover, it could be four years, or degradation may not even occur [33]. Atrazine degradation was not observed in groundwater after 77 weeks [34], and another study reported atrazine's half-life in groundwater as 83 weeks [35]. However, this may be as short as 24 weeks in the presence of sunlight [36].

As this is an environmental observational study, environmental effects, including climatic changes, cannot be excluded from groundwater atrazine's fate. Climatic changes were in this study described in terms of hydrometeorological factors such as precipitation and annual mean air temperature. While daily mean air temperatures observed for most of the years were in the range of extreme heat or cold, there appears to be evidence of drought in 1995, 2000, and 2012. Drought during a growing season reduces the groundwater recharge rate. Consequently, the lower recharge also reduces the leaching of atrazine to wells. Hence, the sharp atrazine, D.I.A., and D.E.A. depletion in the year with the longest drought duration may be due to a drought-induced decrease in atrazine leaching [37]. Moreover, high precipitation was observed in cluster 7, characterized by high groundwater atrazine concentration. This suggests the involvement of precipitation in the deposition of atrazine in groundwater [38].

Data used in this study provided evidence of seasonal variation of atrazine, D.E.A., and D.I.A. While May and June are the peak season for atrazine application [39,40], December, January, February, and March, which are winter/early spring seasons, were observed in this study as the months with groundwater peak atrazine, D.E.A., and D.I.A. In contrast, another study reported peak atrazine concentration in late summer and early autumn. This was attributed to rainfall [41]. Nebraska's intense rainstorms in May and June may contribute to atrazine leaching after application. This, therefore, suggests that peak groundwater atrazine detected in the winter and early spring may result from a time lag of five to seven months required for atrazine transition from the application site to groundwater.

Moreover, other studies conducted in the Midwest have reported groundwater atrazine peaks in the winter and early spring [42], which is in congruence with the findings of this study. In addition, winter is known to slow down atrazine degradation [43]. This may partly contribute to the seasonal variation of atrazine in favor of winter and early spring.

Atrazine depletion with time was indeed observed in this study. Given the extreme climatic changes in Nebraska, one may easily attribute this to the time-dependent atrazine depletion. However, only a slight precipitation effect was observed. Instead, well depth highly predicted low-level groundwater atrazine. This finding is not novel because the associations between well depth and decreased atrazine level were previously reported [28]. This underscores the significance of the irrigation and domestic wells, the deepest wells in this study. To determine other factors beyond well depth which affects groundwater atrazine concentration, a cluster analysis was performed. The effect of well depth was excluded by comparing clusters with the same well depth but different groundwater atrazine concentrations. County clusters with low groundwater atrazine concentration had approximately three times the population supplied by domestic groundwater compared to county clusters with high-level atrazine.

Furthermore, counties with low-level atrazine are more metropolitan than counties with high-level atrazine. In contrast to the wells in high-level atrazine counties, wells in low-level atrazine counties were mostly utilization wells. This suggests that low-level groundwater atrazine in cluster 5 may be due to excessive groundwater usage [44].

In addition, cluster 5 with low-level atrazine counties (mainly in groundwater discharge areas) was characterized by slightly lower precipitation than cluster 7 with high-level atrazine counties. The fate of atrazine in these two clusters could be due to the interplay between degradation processes, leaching to groundwater wells, and groundwater abstraction. In general, the expectation is that the greater the well depth or the depth to water table, the more the groundwater wells should be protected from atrazine contamination. Since these two clusters have similar well depth and groundwater temperature (based on annual mean air temperature), the difference in their mean atrazine concentrations may be due to differences in their precipitation, their rates of groundwater abstraction (as a function of the population of potential groundwater users and density of wells, Table 2), their bedrock geology, or whether they are in recharge or discharge areas.

Compared to cluster 5, the higher average groundwater atrazine concentration in cluster 7 could be attributed to leaching, since its counties are in recharge areas, coupled with higher precipitation and higher irrigation well density (more abstraction for irrigation during growing seasons). In contrast, the low average groundwater atrazine concentration in cluster 5 counties could be attributed to the fact that they are primarily in discharge areas (and close to major streams and rivers) with lower precipitation (less leaching), low irrigation well density, and high population (more drinking water wells). This could also result from groundwater mixing since some drinking water wells in discharge areas could draw water from nearby streams, lakes, or rivers.

The reasons for groundwater atrazine depletion are highly convoluted. This current study demonstrated robustness for deciphering the factors associated with low-level groundwater atrazine in counties of high atrazine usage. However, it was limited by reliance on publicly available data representing only county-level groundwater atrazine estimation. Atrazine measurement of abstracted groundwater at usage sites may be more explicit. Another limitation is that not all wells sampled for atrazine had measurements for atrazine metabolites. Hence, atrazine metabolites were under-reported in this study.

Furthermore, this study failed to account for the transport process of atrazine through the vadose zone. In addition, no data was available regarding atrazine at the recharge areas. However, county-level data summarized all the potential defects that the aforementioned limitations would provide if we had used individual wells for this study.
