The consequences of dam passage for downstream-migrating American eel in the Penobscot River, Maine

American eel (Anguilla rostrata) often pass hydropower dams during adult spawning migrations. We conducted a 4-year acoustic telemetry study that characterized passage risks through two dams (West Enfield and Milford) in the Penobscot River, Maine, USA. We released tagged fish (n = 355) at two sites, estimated survival and delay under variable river conditions, and compared performance among dammed and free-flowing river sections. Survival rates (standardized per river kilometre, rkm) were lower at West Enfield (Φrkm = 0.984 ± 0.006 SE) and Milford (Φrkm = 0.966 ± 0.007 SE) compared with undammed River sections (Φrkm = 0.998 ± 0.0003 SE). Cumulative mortality was 8.7% (4.4 km) and 14.2% (5.5 km) through dammed sections and 8.7% throughout the rest of the river (58.1 km). Fish that already passed an upstream dam incurred higher downstream mortality compared with individuals without passage experience. Additionally, fish endured long delays at dams, and >10% of fish were delayed >24 h. Low flows exacerbated the risk of mortality and delay. These results offer evidence for direct, latent, and sublethal consequences of dam passage for migrating eels.

Deep Learning for Automated Detection and Identification of Migrating American Eel Anguilla rostrata from Imaging Sonar Data

Adult American eels (Anguilla rostrata) are vulnerable to hydropower turbine mortality during outmigration from growth habitat in inland waters to the ocean where they spawn. Imaging sonar is a reliable and proven technology for monitoring of fish passage and migration; however, there is no efficient automated method for eel detection. We designed a deep learning model for automated detection of adult American eels from sonar data. The method employs convolution neural network (CNN) to distinguish between 14 images of eels and non-eel objects. Prior to image classification with CNN, background subtraction and wavelet denoising were applied to enhance sonar images. The CNN model was first trained and tested on data obtained from a laboratory experiment, which yielded overall accuracies of >98% for image-based classification. Then, the model was trained and tested on field data that were obtained near the Iroquois Dam located on the St. Lawrence River; the accuracy achieved was commensurate with that of human experts.