4. DISCUSSION
We found that shellfish sanitation data collected routinely through a systematic random sampling strategy as defined by the National Shellfish Sanitation Program (NSSP) could cautiously be used for long-term water quality trend analysis. By comparing salinity measurements collected by the NCDWR, which maintains an unbiased monitoring program, and NCDMF, which only samples when shellfish waters are open for harvest, we were able to assess whether the sampling constraints imposed on the NCDMF measurements influenced the trend testing results. We found that the NCDMF and NCDWR salinity time series behaved similarly across all SGAs (Figure 5). However, the NCDWR data only spanned 10 years while the NCDMF data spanned 20, and the difference in time series length limits our ability to fully corroborate the NCDMF data using NCDWR observations. Additionally, though not strongly evident in the salinity data analyzed here, the risk for sampling bias to affect routine monitoring data collected by shellfish sanitation programs exists and should always be considered when analyzing their measurements.
We expect sampling bias risk to be greatest in conditionally approved waters with low rainfall thresholds (i.e., 1 to 2 inches), such as SGA E. In contrast, in areas with relatively high rainfall thresholds (e.g., 4 inches), routine FC samples can typically be collected at any time during the year since these waters remain open unless an exceptional event, such as a hurricane or major frontal storm, has occurred. Because waters with high rainfall thresholds largely remain open, the six annual samples are collected under a wider range of environmental conditions, and there is less risk of sampling bias potentially affecting FC trends quantified from the routine monitoring data. For example, SGA G represents an area with high rainfall thresholds (4 inches). These high rainfall thresholds create less restrictive conditions for routine sampling, effectively increasing the variety of conditions captured in the sampling. Accordingly, FC trends determined from shellfish sanitation data from these stations are likely representative of the true improvement or degradation in water quality observed in the system, which also helps to explain why the FC trend results we reported corroborate findings from other studies that have evaluated water quality in this region. In contrast, areas that are more restricted in the time and conditions that routine sampling is able to occur (i.e., areas that are conditionally managed with low rainfall thresholds), such as SGA E, are associated with routine observations that have higher risk of being biased, and there is increased complexity in terms of interpreting these data to infer general water quality trends. Low rainfall thresholds dictate higher rates of closures for even mild meteorological events, which effectively restricts the open times available for routine sampling. However, we demonstrated that the use of an external water quality dataset, in this case for salinity, can be used to assess how sampling bias may have affected measurements collected by shellfish sanitation programs.
Nonpoint source runoff is considered a major contributor to FC loads in estuaries located near developed landscapes (Mallin et al., 2000; Coulliette et al., 2009; Kirby-Smith et al., 2006; Campos et al., 2013). Therefore, the increasing trends we documented in FC concentrations in SGAs B, E, and H align with the known relationship between FC and development. Specifically, the positive correlative relationship between change in developed land cover across a watershed and increasing FC trends was seen in SGAs B, E, and H, while A, C, D, F, and G were associated with negative correlations. Relationships between developed land use change and FC trends could potentially be clarified further by using population density change over watersheds, stormwater management, or differentiating impervious surfaces (Mallin et al., 2000; Carle et al., 2005; Cahoon et al., 2016; Freeman et al., 2019).
The negative correlation between FC and salinity along all SGAs was consistent with established water quality relationships except for a few contradictory results. The inverse relationship between FC and salinity could be a result of the coupled effect of increased freshwater input that comes with increased precipitation (Campos et al., 2013; Coulliette et al., 2009). It is known that FC concentrations increase following runoff after rainfall events, especially in more developed areas (Mallin et al., 2000; Carle et al., 2005; Cahoon et al., 2016; Freeman et al., 2019). These same rainfall events that increase the FC concentrations also decrease salinity, which is illustrated in the inverse relationships reported in this study across each SGA, with the exception of SGA E (Table 2, Figure 6). However, the inverse relationship between FC concentration and salinity trends was often noisy (Figure 6b), with the correlation coefficient between FC concentration and salinity trends being in the range of [-0.147, 0.161] for 5 out of 8 SGAs (Table 2). In the case of SGA E, where a positive correlation between salinity and FC trends was observed, the correlation appears to have been influenced by outlying values (Figure 6b), particularly since most of the βSal values reflected increases in salinity (Figure 6c) while the βFC values showed there were FC concentrations decreases across most sampling locations (Figure 6a, 6b).
The noisy relationships between FC concentration and salinity trends in our results could be explained by our dataset not capturing short-term FC concentration increases following storm events and instead capturing FC during baseflow conditions. Because the data analyzed in this study was produced from routine systematic random sampling, which is collected when waters are open for harvest to capture baseline fecal coliform loading, the observations will not capture changes in storm-driven FC concentrations. Instead, the measurements may reveal if there is chronic loading in an area (e.g., due to continuously failing septic systems or poorly performing wastewater treatment plants). Therefore, the trends from this analysis are representative of baseflow conditions. Accordingly, had the routine sampling data captured post-storm conditions, we expect stronger correlations between FC and salinity trends would have been observed. Instead, we believe that factors such as increases in tidal flushing (e.g., due to inlet dredging) and changes in baseflow FC loads in these systems play a larger role in explaining the negative relationship between FC and salinity than changes in rain and runoff.
In addition to providing insights on long-term water quality trends, shellfish sanitation data can also be used to assess the efficacy of current management practices. For example, a conditionally managed waterbody with low rainfall thresholds that is still showing a trend towards increasing FC concentrations could indicate a decline in water quality that has not been met with intense enough action by the current management plan. As a result, trends in fecal coliform observations could be used as an “early warning system” to help pinpoint areas where more intense management measures need to be taken. For example, the way in which these data could be used as an “early warning system” is demonstrated by focal SGA B (Figure 7c), where the mouth of the Cape Fear River likely shows increasing FC concentrations due to degradation of water quality that may need to be met with new management actions.