ABSTRACT
Invertebrate biodiversity remains poorly explored although it comprises much of the terrestrial animal biomass, more than 90% of the species-level diversity, supplies many ecosystem services. Increasing anthropogenic threads also require regular monitoring of invertebrate communities. The main obstacle is specimen- and species-rich samples consisting of thousands of small specimens. Traditional sorting techniques require manual handling based on morphology and are too slow and labor-intensive. Molecular techniques based on metabarcoding struggle with obtaining reliable abundance information. We here present a fully automated sorting robot for small specimens that are detected in the mixed sample using a convolutional neural network. Each specimen is then moved from the mixed sample to a well of a 96-well microplate in preparation for DNA barcoding. Prior to movement, the specimen is being photographed and assigned to 14 particularly common “classes” of insects in Malaise trap samples. The average assignment precision for the classes is 91.4 % (75-100 %) based on a preliminary neural network that is expected to improve further as more images are used for training. In order to obtain biomass information, the specimen images are also used to measure the specimen length and estimate the body volume. We outline how the “DiversityScanner” robot can be a key component for tackling and monitoring invertebrate diversity by generating large numbers of images that become training sets for species-, genus-, or family-level convolutional neural networks, once the imaged specimens are classified with DNA barcodes. The robot also allows for taxon-specific subsampling of large invertebrate samples. We conclude that the combination of automation, machine learning, and DNA barcoding has the potential to tackle invertebrate diversity at an unprecedented scale.
Competing Interest Statement
The authors have declared no competing interest.
