Assessing anthropogenic pressures on streams: A random forest approach based on benthic diatom communities

Area of Science:

• Ecology
• Environmental Science
• Freshwater Biology

Background:

• Benthic diatoms are key bioindicators for freshwater ecosystem health, particularly under international policies like the Water Framework Directive (WFD).
• Current diatom-based indices quickly reflect water quality changes but struggle to pinpoint specific co-occurring anthropogenic pressures.
• A need exists for tools that can identify diverse environmental stressors impacting aquatic ecosystems.

Purpose of the Study:

• To develop predictive models for estimating stream impairment risk from multiple anthropogenic pressures.
• To utilize both taxonomic and trait-based characteristics of diatom assemblages for risk assessment.
• To create a practical tool for stream managers to identify and prioritize environmental management actions.

Main Methods:

• Employed large-scale monitoring data to build Random Forest models.
• Developed individual models to assess impairment risk for six chemical and five hydromorphological/land-use pressure categories.
• Evaluated model performance using Area Under the Curve (AUC) metric, with a threshold of ≥0.70 for good assessment.

Main Results:

• Eight models achieved good impairment risk assessment (AUC ≥ 0.70).
• Models for chemical pressures showed higher accuracy than those for hydromorphological or land-use pressures in multi-pressure scenarios.
• The models successfully detected both ecological restoration and degradation trends over long-term surveys.

Conclusions:

• Developed robust models for assessing stream impairment risk from various anthropogenic pressures using diatom data.
• The R-based routine provides a user-friendly tool for stream managers to proactively identify ecological degradation.
• This approach enhances the ability to prioritize effective water quality management and conservation efforts.