Effects of Ecological Complexity on Student Identification Accuracy in a School-Based Citizen Science Program

School-based citizen science is increasingly used to support biodiversity monitoring, but the reliability of student-generated ecological data remains uncertain. I evaluated a nationwide barn owl (Tyto alba) pellet project in Israel involving 1612 students and 107 teachers over two academic years. Teacher participation was moderate to high: 66.4% of teachers returned at least one complete kit, generating 3333 prey records, mostly from small mammals, thereby demonstrating that schools can contribute significant ecological sampling effort across broad geographic areas. Student identification accuracy was moderate: 50% of pellet analyses were perfectly identified, with an overall item-level accuracy of 62%, and 76% of pellet attempts included at least one correctly identified prey item. The probability of perfect identification declined sharply with pellet complexity. Identification accuracy also varied among prey taxa: even under minimal complexity (single-species pellets), certain taxa exhibited significantly higher misidentification rates than others. Each additional prey item reduced the odds of correctly identifying all prey items in a pellet by about 68%, indicating that identification reliability decreases as sample complexity increases. Although many students initially found pellets disgusting, engagement remained high, and willingness to repeat the activity was strongly linked to enjoyment. These findings show that school-based citizen science can support biodiversity monitoring, but reliable ecological inference requires validation frameworks, especially when sample complexity is high.