Irina Benson - NOAA Federal
Resource Ecology and Fisheries Management Division, Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA
Research Fisheries Biologist
Seattle, WA
Fourier transform near infrared spectroscopy of otoliths coupled with machine learning can predict fish age more efficiently and with comparable precision compared to traditional ageing methods.
We explore advanced technologies using Fourier transform near infrared (FT-NIR) spectroscopy coupled with machine learning to estimate fish age more rapidly and with greater efficiency than traditional approaches. This technology has recently been applied to the otoliths from walleye pollock (Gadus chalcogrammus) and northern rockfish (Sebastes polyspinis) from the Bering Sea and red snapper (Lutjanus campechanus) from the northern Gulf of Mexico. In order to train and evaluate the performance of machine learning models for each species, data were split into a train set (80%) and a test set (20%). We employed TensorFlow to train neural networks to explore the underlying relationships between an otolith FT-NIR spectrum and fish age along with other biological and geospatial data that have an effect on fish growth. Results from the models between the FT-NIR predicted age and traditional age estimates for each of three species indicate that otolith spectra in the 4,000-7,000 cm-1 wavenumber region, which is associated with C-H and N-H functional groups, has the highest impact predicting fish age. Otolith weight, for which both fish age and somatic growth are determinants, has greater impact than fish length. The geographic coordinates or substock designation have an impact due to fish ontogenetic habitat shifts. Coefficients of determination (R2) are 0.91 to 0.93 for train data and 0.89 to 0.92 for test data. Root mean square errors (RMSE) are 0.83 to 3.37 for train data and 0.91 to 3.36 for test data set. Age estimation outcomes between the traditional and deep machine learning methods are largely indistinguishable based on Bland-Altman plots. Since FT-NIR spectroscopy method is about ten times faster than traditional age estimation methods, these results suggest that FT-NIR spectroscopy of otoliths coupled with deep machine learning can predict fish age more rapidly, with greater efficiency, and with comparable precision.
