Detection method for diel vertical migration pattern using 2D cross-correlation with ADCP backscatter time-series data

Abstract

1. Diel vertical migration (DVM ), which refers to the global daily migra t io n o f zooplankton and mic ro ne k to n (hereafter ‘zooplankton’) , serves an important function in oceanic ecosystems. DVM can be recognized as a variation in the sound scattering layer (SSL), but no standard exists for identifying SSLs, and no definition has been proposed to describe the basic features of DVM. Hence, a standardized DVM detection method is needed to report consistent results and efficiently compare the parameters over broad ocean areas and lo ng time scales. 2. We developed an automated, quantitative method for detecting DVM and identifying the characteristic parameters using 2D cross-correlation. We established the DVM trajectory model with line a r a nd s inus o id a l parts featuring a specific height and width. The 2D cross-correlation method was applied to acoustic echogram images of the volume backscattering strength (Sv, dB re 1 m-1) with a synthetic image made from the DVM trajectory model. From the cross-correlation coeffic ie nt s, we found the candidate lines exhibit i ng the strongest possibilit y of being DVM trajectories. 3. We tested the DVM detection method on the acoustic echograms for 273 days, of which Sv was ga thered by a bottom-moored, upward-looking acoustic Doppler current profiler (ADCP). For each identified DVM trajectory, 4 quantitative parameters describing their spatial and temporal structures (the maximum and minimum depths and ascent and descent time s ) were determined, and the existence of multi- layer was recognized. We confirmed that this method showed better performance than the prior method based on the weighted mean depth (WMD), and as a result of visua l scrutiny inspection, our method showed 88% DVM detection performance. 4. O ur goal was to suggest a robust, quantitative, automated DVM detection method using 2D cross-correlation. The method was successfully applied to detect DVM trajectories and define the characteristic parameters regardless of fluc tua t io n or the mult i- la yer structure due to its robustness and s imp le mo d e l. Analyzing DVM behaviors across environme nts with this method could help reveal the significance of these behaviors in ocean environments.