Login / Signup

Assessing data bias in visual surveys from a cetacean monitoring programme.

Cláudia Oliveira-RodriguesAna Mafalda CorreiaRaul ValenteÁgatha GilMiguel GandraMarcos LiberalMassimiliano RossoGraham PierceIsabel Sousa-Pinto
Published in: Scientific data (2022)
Long-term monitoring datasets are fundamental to understand physical and ecological responses to environmental changes, supporting management and conservation. The data should be reliable, with the sources of bias identified and quantified. CETUS Project is a cetacean monitoring programme in the Eastern North Atlantic, based on visual methods of data collection. This study aims to assess data quality and bias in the CETUS dataset, by 1) applying validation methods, through photographic confirmation of species identification; 2) creating data quality criteria to evaluate the observer's experience; and 3) assessing bias to the number of sightings collected and to the success in species identification. Through photographic validation, the species identification of 10 sightings was corrected and a new species was added to the CETUS dataset. The number of sightings collected was biased by external factors, mostly by sampling effort but also by weather conditions. Ultimately, results highlight the importance of identifying and quantifying data bias, while also yielding guidelines for data collection and processing, relevant for species monitoring programmes based on visual methods.
Keyphrases
  • electronic health record
  • big data
  • randomized controlled trial
  • quality improvement
  • risk assessment
  • artificial intelligence
  • rna seq