Effects of sand extraction, a socio-economically important industry in the Belgian Part of the North Sea, are monitored through macrobenthos-based indices, an expensive and time-consuming process. While environmental DNA metabarcoding and supervised machine learning can circumvent these issues, area and disturbance-specific protocols are needed. This study aimed to firstly explore if 18S eDNA metabarcoding data detects sand extraction effects on nematode diversity, a promising but underutilized indicator taxa; secondly, if this data can predict ecosystem status with machine learning. Sediment samples were used for DNA sequencing and morphologically identification of macrobenthos, after which alpha and beta diversity were analyzed. LASSO regression was used to predict BEQI ecosystem status. Analyses were performed at ASV and genus level to explore if taxon specificity affects predictive accuracy.
Extraction intensity had no effect on ASV alpha diversity but genus level data showed significant differences. Beta diversity was significantly different between high and low/no extraction in Thorntonbank. No significant difference was found for Oostdyck or Hinderbank. Predicted ecological status did not have enough agreement with macrobenthos-based assessment, either at ASV or genus level. We highlight future steps to increase predictive accuracy of a machine learning approach for impact assessments of sand extraction activity.
promotor/supervisor feedback
nothing yet
