Published In

Ecological Indicators

Document Type


Publication Date



Analysis, Bioassessment, Electric properties, Environmental quality, Fine sediment, Periphyton, Random forest, Runoff, Sediments (Geology), Traits-based


The diatom motility trait is widely used in bioassessment studies, however, there is no strong consensus as to what environmental conditions diatom motility reflects. We used random forest models to explore the behavior of several diatom-based motility indices and examined whether stronger diatom motility-environmental relationships could be developed by controlling for environmental factors that influence natural sediment loads and factors that co-occur with anthropogenic sediment loading. Across the study area, median values of most stressors were low; streambed fine sediments ranged from 0 to 100% (median: 6.7%) and total phosphorus concentrations ranged from 0 to 2587 µg L−1 (median: 17 µg L−1). Depending on the index, median abundance of motile taxa ranged between 8 and 47%, and was highest in the plains ecoregion. Few sites were dominated by motile taxa, with only 9 sites having a relative abundance of highly motile taxa >50%. Refinement of the motility trait did result in improved model performance; the RF model developed for highly motile taxa had the lowest RMSE (0.11) with environmental variables explaining 34.9% of the variability in the index. Variables indicative of streambed and water column sediments were not important predictors in the RF models for any motility index and motility trait refinement did not improve the relationship between motility index and streambed fine sediments. Variables related to water quality (e.g., total phosphorus, conductivity), flow (e.g., runoff, % of channel as fast moving water), and watershed characteristics (e.g., watershed slope) were the most important environmental predictors for the RF models for all motility indices, while variables indicative of streambed and water column sediments had lower importance values. Stratifying sites by ecoregion resulted in slight improvement of model performance for the mountain ecoregion (RMSE = 0.09) but did not result in stronger relationships between diatom-motility indices and instream sediments. Stratifying sites by instream total phosphorus concentrations also resulted in slight improvements in RF model performance (RMSE = 0.10), but did not result in stronger relationships between diatom motility indices and instream sediment conditions. Stratifying sites by geomorphic features had no effect on how diatom motility indices responded to instream sediments. Our results show that the relationship between the diatom motility trait and fine sediment is complex. Strong relationships between fine sediments and diatom motility may be difficult to detect in large-scale observational studies covering complex landscapes with multiple interacting factors.


Copyright (c) 2022 The Authors

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.



Persistent Identifier