River bioassessment using diatoms: a classificatory analysis approach

This study explores the use of classificatory analysis, specifically stochastic gradient boosting trees, for river bioassessment using diatoms. We compared this approach to traditional diatom-based indices, such as the Specific Pollution sensitivity Index (SPI), across five major river basins in Ecuador. Our findings demonstrate that presence/absence data of diatom species can provide more accurate water quality assessments than quantitative matrices. A simplified protocol involving the identification of only 200 diatom valves at the species level yielded reliable results, with an 80% success rate in classifying water quality, significantly outperforming the SPI (43% accuracy) under similar conditions. The study also reveals that genus-level identification can be sufficient for effective biomonitoring, potentially reducing the expertise required for taxonomic analysis. Key indicator species, such as Gomphonema rigidum and Craticula vixnegligenda, were identified as crucial for the classification process. These results suggest that classificatory analysis using diatoms can offer a cost-effective and efficient alternative to traditional biomonitoring methods, with implications for streamlining water quality assessment protocols.