Long-term deployments of autonomous echosounders deliver rich data embedded with information about mid-trophic level organisms of the marine food web. As a remote sensing tool, moored echosounders have played an important role in observing temporal changes of animal distributions in the water column over large temporal scales, ranging from months, seasons to years. Taking advantage of the power of matrix decomposition techniques in exploiting regularity in the data to automatically discover low-dimensional structures in large data sets, we apply Principal Component Pursuit (PCP) and temporally smooth Nonnegative Matrix Factorization (tsNMF) in a time-windowed fashion to analyze data from a network of moored, upward-looking echosounders deployed by the U.S. Ocean Observatories Initiative (OOI) from 2015 to 2018. We show that these echosounder time series are low-rank in nature and can be reproduced by a time-varying linear combination (activation) of a small number of distinct daily movement patterns (components). The components and the associated activation jointly provide a compact representation that is suitable for visualization and systematic analysis of mid-trophic animal activities in response to environmental changes, such as those from seasonal changes and extreme weather events.