Remote cameras are a common method for surveying wildlife and recently have been promoted for implementing large-scale regional biodiversity monitoring programs. The use of camera-trap data depends on the correct identification of animals captured in the photographs, yet misidentification rates can be high, especially when morphologically similar species co-occur, and this can lead to faulty inferences and hinder conservation efforts. Correct identification is dependent on diagnosable taxonomic characters, photograph quality, and the experience and training of the observer. However, keys rooted in taxonomy are rarely used for the identification of camera-trap images and error rates are rarely assessed, even when morphologically similar species are present in the study area. We tested a method for ensuring high identification accuracy using two sympatric and morphologically similar chipmunk (Neotamias) species as a case study. We hypothesized that the identification accuracy would improve with use of the identification key, and with observer training, resulting in higher levels of observer confidence and higher levels of agreement among observers. We developed an identification key and tested identification accuracy based on photographs of verified museum specimens. Our results supported predictions for each of these hypotheses.  In addition, we validated the method in the field by comparing remote camera data with live-trapping data.  We recommend use of these methods to evaluate error rates and to exclude ambiguous records in camera-trap datasets. We urge that ensuring correct and scientifically defensible species identifications is incumbent on researchers and should be incorporated into the camera-trap workflow.

We tested the efficacy of our final identification key by comparing the accuracy of observers using identification resources from the literature (hereafter, “literature observers”; N=19) to that of observers using our identification key (hereafter, “key observers”; N=15). We provided all observers with Adobe PDF files that included instructions, identification resources, and a test. We provided the literature observers with identification resources that consisted of excerpts from Mammalian Species accounts for both species (Best et al., 1992; Verts and Carraway, 2001) and a popular field guide to North American mammals (Reid, 2006). These materials represented the best available identification information attainable without examining specimens. We highlighted sections pertaining to pelage traits to guide observers to the most relevant information for identifications from photographs. We provided the key observers with the identification key. For both groups of observers, the test consisted of 20 slides, each showing three views of a single chipmunk specimen (dorsal, lateral, and ventral). We used three views for testing because in our field applications cameras fire multiple times providing photographs of an animal from multiple angles – on average, we captured 10.6 photographs of a chipmunk with each visit to a camera and only 7.2% of chipmunk visits to a camera resulted in a single photograph. For each slide, observers recorded a species identification and the numeric confidence-rank. Numeric confidence-ranks ranged from 1 to 4 for each slide, based on the observer's confidence in the attribution of species, from 1: no confidence, 2: not very confident, 3: somewhat confident, 4: very confident. Observers could only view their own responses during the testing process. The observers were field technicians working on chipmunk field research or undergraduate students in wildlife biology, but they did not have any prior knowledge about chipmunk identification.

We tested if a training program would improve the accuracy of observers who used our identification key. All key observers (N=15) completed the training program. For the training program, observers practiced using the identification key to identify photographs of chipmunk specimens in two separate training sets. After each training set, we provided the trainees with the answer key, so that they could compare their answers to the correct answers and learn from mistakes. The first training set was the original 504 randomized slides showing nonants of specimens of chipmunks, used by the authors for the development of the identification key. The trainees coded each slide for each pelage trait, assigned a species identification based on their overall impression, and reported a numeric confidence-rank, following the procedure used for the development of the key. The second training set consisted of 168 randomized slides showing a single view (dorsal, lateral, or ventral) of a specimen. For each slide, the trainee assigned a species identification and reported a numeric confidence-rank. After completing both training sets and reviewing the correct identifications, we considered observers to be fully trained (hereafter “trained key observers”). We tested trained key observers using a post-training test, which consisted of a set of 56 slides, each showing three views of a single chipmunk specimen (dorsal, lateral, and ventral). For each slide, observers recorded a species identification and the numeric confidence-rank.