Over the last decade, advancements in instrumentation have produced a deluge of water column sonar data from ships, moorings and autonomous vehicles. These data provide unprecedented opportunities to study marine ecosystems at diverse spatial and temporal scales. However, to date, these data remain significantly under‑utilized. The root cause of this problem is the lack of interoperable data format and scalable analysis workflows that adapt well with increasing data volume.
echopype aims at addressing the above issues by creating an interoperable netCDF sonar data format, and supporting scalable computations through xarray and dask based on these files. This approach makes it possible to build sonar data analysis pipelines that are scalable and agnostic of the instrument origin of the data. In addition to using analysis modules provided by echopype, the converted data in netCDF format can be easily combined and analyzed flexibly with other climate and oceanographic data sets, facilitating the integration of ocean sonar data in interdisciplinary oceanographic research.
echopype contains tools for converting data from various manufacturer-specific formats into a standardized netCDF files. This is useful, because netCDF is the current defacto standard in climate research and is supported by many powerful libraries for efficient computation in many programming languages. This means that once data are converted, researchers can then pick and choose what they need directly from the data and are not restricted by specialized and often proprietary data analysis software.
echopype is written to use as much as functionality of xarray as possible. xarray is a powerful Python package that lets you work with multi-dimensional labeled data set the same way as in pandas. Under the hood, xarray uses dask to support parallel computations and streaming computation on data sets that don’t fit into memory. Building on xarray and dask, the goal of echopype to make sonar data analysis efficient and scalable through distributed computing through either local clusters or on the cloud.