Misidentification Bias in Citizen Science: the case of White-crested and Small-billed Elaenias

GORLERI F; ARETA JI

Congreso; XI Neotropical Ornithological Congress; 2019

Citizen Science (CS) permit the study of bird populations at large spatial and temporal scales byengaging volunteers in recollection of data. However, information is lacking on the consistencyof volunteer and professional data in groups of morphologically similar species. Systematicconfusions between species may produce an error cascade where false positives and falsenegatives are generated simultaneously. This can lead to erroneous results and interpretationsabout the distribution and phenology of the species involved.Most species in the genus Elaenia (Aves; Tyrannidae) are morphologically similar and standamong the hardest birds to identify in the Neotropical Region. Here we focus on the White-crested and Small-billed Elaenias, two endemic breeders of southern South America.The White-crested Elaenia breeds mostly along a thin stripe in the Andes and is narrowlyparapatric or widely allopatric to the lowland breeding Small-billed Elaenia. However,geolocator studies and meager specimen data indicate that White-crested Elaenia migratethrough the lowlands of Argentina and Uruguay towards NE Brazil, leading to a momentarysympatry with Small-billed Elaenia.Few records of White-crested on migration are available in CS platforms, and they fail toprovide a solid characterization of the migratory passage of the species in Argentina andUruguay. This could stem from a systematic large-scale misidentification of White-crested asSmall-billed Elaenia. To test this, we reviewed all photographic records of both species up toMarch 2019 in two CS projects (eBird and EcoRegistros) in Argentina and Uruguay, andcorrected those that were misidentified. We then compared phenology using raw data andreviewed data, and also performed species distribution models (SDMs) using these datasets toassess their differences.Reviewed data agreed in phenology and distribution with reference data, but differed stronglyfrom raw data. We found a strong unidirectional bias during autumn migration in raw data.White-crested Elaenia was misidentified as Small-billed Elaenia in 35% of the cases duringMarch to May (n= 138), while only one Small-billed was submitted as White-crested Elaenia.This generated systematic false positives of Small-billed Elaenia and false negatives of White-crested Elaenia. Raw data was unable to detect a peak migratory passage of White-crestedElaenia and indicated a late departure date for Small-billed Elaenia (late April). Reviewed datadetected a peak passage of White-crested Elaenia at the end of March, and showed a three-weeks earlier departure date for Small-billed Elaenia. Only SDMs using reviewed data detectedlarge rivers to be an important predictor during White-crested Elaenia migration coincidingwith geolocator studies.The error rate we found represents a lower bound for these species in CS databases, becausephotos can be verified by reviewers or users, but unvouchered records are elusive to data-quality filtering. We suggest to explore ways and develop tools to detect and control forsystematic large-scale misidentification biases. This should lead to more robust inferentialresults from analyses of CS data.Citizen Science (CS) permit the study of bird populations at large spatiotemporal scales byengaging volunteers in recollection of data. However, information is lacking on the consistencyof volunteer data in groups of morphologically similar species. Systematic misidentificationsbetween species can lead to erroneous results and interpretations about the distribution andphenology of the species involved.Most species in the genus Elaenia (Aves; Tyrannidae) stand among the hardest birds to identifyin the Neotropical Region. Here we focus on the White-crested and Small-billed Elaenias, twoendemic breeders of southern South America. Both are mostly allopatric during breedingseason, but they can be found together during migration times. Despite both are abundant,few records of White-crested on migration are available in CS platforms. This could stem froma systematic large-scale misidentification of White-crested as Small-billed Elaenia.To test this, we reviewed all photographic records of the species submitted to eBird andEcoRegistros up to March 2019 in Argentina and Uruguay. We then compared phenology anddistribution using raw data and reviewed data to assess their differences.White-crested Elaenia was misidentified as Small-billed Elaenia in 35% of the cases duringMarch to May (n= 138). Only reviewed data detected a peak migratory passage of White-crested Elaenia (end of March). Also, reviewed data indicated a three-weeks earlier departuredate (early April) for Small-billed Elaenia than raw data. Only SDMs using reviewed datadetected large rivers to be an important predictor during White-crested Elaenia migrationcoinciding with geolocator studies.The error rate we found represents a lower bound for these species in CS databases, becausephotos can be verified, but unvouchered records are elusive to data-quality filtering. Wesuggest to explore ways and develop tools to detect and control for systematic large-scalemisidentification biases. This should improve inferential results from analyses of CS data.