4.4 Limitations and Recommendations for Future Research
While the approach developed in this manuscript successfully describes spatiotemporal connectivity dynamics, there are several assumptions and limitations that future work should consider. Connectivity strength metrics used here enable the development of a continuous value between 0 and 1, however the meaning of the connectivity strength values are dependent on the internal variation within a system and interpretation may differ among systems. As such, we suggest that when identifying thresholds in different connectivity states, future research should consider a distribution of possible connectivity strength values (see Figure 9). An additional limitation is that the empirical connectivity functions we developed assume stationarity in the underlying floodplain structure and fitting a single model assumes a lack of hysteresis in functional connectivity between source and target sites between rising and falling limbs. Our observations of hysteresis at two sites in 2018, likely driven by beaver activity, suggests that such changes in connectivity thresholds are likely relatively common. However, such assumptions are equally present in all approaches reliant on static physical datasets such single date LIDAR acquisition or field surveys (Passalacqua et al., 2015). Therefore, we believe our approach is valuable but for longer-term studies, but that conditions should be monitored through time and relationships updated similar to how a rating curve used to estimate discharge needs to be updated if the underlying channel morphology is altered during the study period.
While our work demonstrates that aquatic microbiomes can be utilized for inference into hydrologic connectivity, our ability to determine the broad applicability of this technique is limited. This study was conducted in a relatively small river-floodplain system with relatively homogenous surrounding land cover. To apply the microbial connectivity metric more widely, future work will need to assess how residence time thresholds in different systems and at different scales interact with microbial membership. Further, as investigations move to larger, more heterogeneous landscape scales, it will be necessary to consider how to incorporate more diverse sources of water and microbiomes into the approach. To address this, we suggest that future work implement finer-scale OTU-level analysis rather than a coarse community level similarity metric, which we believe can increase the ability to detect weaker flow paths and potentially be used as a multi-tracer to simultaneously measure connectivity from multiple water sources.