Diatoms are known to be efficient bioindicators for water quality assessment because of their rapid response to environmental pressures and their omnipresence in water bodies. The identification of benthic diatoms communities in the biofilm, coupled with quality indices such as the Indice de polluosensibilité spécifique (IPS) can be used for biomonitoring purposes in freshwater. However, the morphological identification and counting of diatoms species under the microscope is time-consuming and requires extensive expertise to deal with a constantly evolving taxonomy. In response, a molecular-based and potentially more cost-effective method has been developed, coupling high-throughput sequencing and DNA metabarcoding. The method has already been tested for water quality assessment with diatoms in Central Europe. In this study, we applied both the traditional and molecular methods on 180 biofilms samples from Northern Europe (rivers and lakes of Fennoscandia and Iceland). The DNA metabarcoding data were obtained on two different DNA markers, the 18S-V4 and rbcL barcodes, with the NucleoSpin Soil kit for DNA extraction and sequenced on an Ion Torrent PGM platform. We assessed the ability of the molecular method to produce species inventories, IPS scores and ecological status class comparable to the ones generated by the traditional morphology-based approach. The two methods generated correlated but significantly different IPS scores and ecological status assessment. The observed deviations are explained by presence/absence and abundance discrepancies in the species inventories, mainly due to the incompleteness of the barcodes reference databases, primer bias and strictness of the bioinformatic pipeline. Abundance discrepancies are less common than presence/absence discrepancies but have a greater effect on the ecological assessment. Missing species in the reference databases are mostly acidophilic benthic diatoms species, typical of the low pH waters of Northern Europe. The two different DNA markers also generated significantly different ecological status assessments. The use of the 18S-V4 marker generates more species inventories discrepancies, but achieves an ecological assessment more similar to the traditional morphology-based method. Further development of the metabarcoding method is needed for its use in environmental assessment. For its application in Northern Europe, completion and curation of reference databases are necessary, as well as evaluation of the currently available bioinformatics pipelines. New indices, fitted for environmental biomonitoring, should also be developed directly from molecular data.
