Size matters: Perspective and angle-correction improves accuracy in noninvasive image-based body size measurements

Abstract

River management should secure conservation of biodiversity and sustainable use of aquatic resources. Conservation of fish populations requires time-series data on the number of fish present and the size-structure. The number of fish and species composition can be resolved by video-surveillance, but detailed measurements of body size often come from more intrusive methods such as fish traps and electrofishing that impose additive stress or mortality on individuals. We have developed and tested a nonintrusive method for video-surveillance which enables estimation of fish length, of anadromous Arctic char (Salvelinus alpinus) in a subarctic river. We use pixel counts in images of fish that swim through a tunnel, to measure the size of the fish, and calibrate our measurements with two parallel laser lines (100 mm apart) that are visible on the fish, both from the side and from above, facilitated by a 45° mirror. We demonstrate how the accuracy in body length measurements depends on camera perspective, fish angle, body curvature and swimming speed, and evaluate this with three independent observers. Our results show that the typically used side-view camera (lateral view) underestimated the fish on average by 10.7%, but that accuracy could be significantly improved by including: (1) angle-correction (for non-perpendicular fish positioning), (2) by measuring the fish from above (dorsal view), and (3) by including the body curvature of actively swimming fish. Our method represents a cost-efficient approach for monitoring size-structure in vulnerable populations that is of management concern and where intrusive monitoring should be avoided.