Fish from space: Predicting mid-trophic levels biogeography via remote sensing and in-situ acoustic data fusion
Jerome Guiet, University of California, Los Angeles, jguiet@atmos.ucla.edu (Presenter)
Kaushik Srinivasan, University of California, Los Angeles, kaushiks@atmos.ucla.edu
Carrie Wall, NOAA, carrie.wall@noaa.gov
Daniele Bianchi, University of California, Los Angeles, dbianchi@atmos.ucla.edu
Mid-trophic level (MTL) organisms provide a link between lower trophic levels and large predators, and impact the ocean’s biological pump. However, their distribution and variability are poorly resolved, particularly in highly dynamic regions such as the California Current Ecosystem (CCE). This lack of understanding arises from the complex interaction of MTLs with their environment, and the difficulty of sampling these organisms in situ. To close this gap, we build a machine learning model of MTL that enables reconstruction of their distribution from well-sampled environmental predictors. The model is trained on 10 years (2007-2016) of coastwise multi-frequency fisheries acoustic observations. Environmental predictors, co-located with the acoustic data, include remote sensing (MEaSUREs, MODIS), reanalysis (ERA, SODA), and climatology (WOA) products. We use these data to train a feed-forward neural network to learn the distribution of acoustic backscatter, a proxy of MTL abundance, in the upper ocean (15 to 215m depth), reconstructing the variability of MTLs with high accuracy (correlation with observations of ~R2=0.80, out of bag). Our reconstructions show a consistent seasonal dynamics, with coastal pulses of backscatter that begin near the coast, spread offshore, and reach a minimum in winter, and a seasonal northward shifts of backscatter that follows upwelling. Comparison of day to night backscatter reveal regional variations in the strength of diel vertical migrations of mesopelagic MTL. Analysis of predictor importance indicates regional differences in the dominant environmental drivers of seasonal to interannual variability in MTLs. Our reconstructions offer a powerful new approach to study the dynamic of MTLs and their role in the global carbon cycle and fisheries.
Poster: Poster_Guiet_2-57_51_35.pdf
Associated Project(s):
Poster Location ID: 2-57
Presentation Type: Poster
Session: Poster Session 2
Session Date: Wed (May 10) 5:15-7:15 PM
CCE Program: OBB