Bias and variation in salmonid redd counting using remotely piloted vehicles

Abstract

Redd surveys estimate spawning population size for many salmonid species. Studies of field‐based redd counting methods highlight observer bias caused by redd density, observer experience, and environmental factors. Researchers have begun using remotely piloted vehicles (RPVs, drones) to count redds; yet, no studies have quantified bias and variability in counts. This study aimed to quantify the influence of redd density, observer experience, and environmental factors (namely, water clarity) on redd counting bias and variability when using RPVs. We found that technological and procedural improvements from our previous study increased precision and reduced variability among observers (coefficient of variation, сυ = 11% compared to сυ = 42%). Redd density was the leading covariate causing differences between RPV and both “best counts” (p < 0.05) and field counts (p < 0.05). We found a reduction in variability with experience level (no experience сυ = 78%; semi‐experienced сυ = 33%; experienced сυ = 20%), with no directional bias in counting. Our paper is the first to quantify observer bias in RPV‐based redd counts. This study describes RPV methods and can help agencies decide how to use RPVs in redd counting and incorporate RPV methods into long‐term datasets.