Overview
This procedure describes how to access, download, and process killer whale calls through the Orcasound database. Using the following processing technique, you can visualize the audio data of several various killer whale calls in spectrogram form.
Dataset sourcing
Orcasound is a great organization focused on saving the endangered orcas that hunt salmon along the west coast of North America. Collaboration between my lab at UW and Orcasound has been rewarding as we work towards new and improved acoustic detection methods in the Puget Sound area.
Using this data with Python
Orcasound stores their data in publicly accessible Amazon Web Service buckets. My code walks through accessing these buckets via python, allowing you to download date-selected data onto your personal computer. Visit the Orcasound github linked above for more data access information.
Orcasound has several hydrophone stations spread across Puget Sound. This tutorial accesses the “Sunset Bay” hydrophone, however the process is identical with the others as well.
Getting Started:
This tutorial assumes that the user has a working python code editor such as VS code on their personal computer. If you do not, start off with a tutorial to get both python and a code editor on your system.
Downloading the necessary files to your computer:
Navigate to my github repository and download all of the files to what will become your working folder on your personal computer. Open this folder in VS code or your python code editor.
Accessing the data:
Start off with the aws_sunsetbay_datasort.ipynb. In running the first code block, check that your python environment has all of the required packages.
Once you have run the first four code blocks successfully, you have a choice. You can pull a singular audio file (step 4a) or you can pull multiple audio files spanning a larger time range (step 4b).
STEP 4A:
Input the specific timestamp that you would like to listen to. You can also change the duration of the audio sample you will download, though files larger than 300 seconds (5 minutes) will be challenging to work with and process. Use the following timestamp to look at a series of great SO1 killer whale calls:
year, month, day = "2025", "01", "12"
start_hour, start_min, start_sec = "21", "07", "38"
date = f'{year}-{month}-{day}_{start_hour}.{start_min}.{start_sec}'
duration = 60
STEP 4B:
Instead of manually entering a timestamp that you want to listen to, this code section will pull audio files as listed in the provided .xlxs file. The entries are filtered to select for detected orca calls in the last couple months of 2024. Follow the exact format of the file and its entries if you would like to alter it, otherwise the code will not read it properly.
At this point, go into the folder that holds all of these code files and you should see a new folder titled “”. Inside that folder you should see the audio files in .wav format.
Processing the data:
Congrats, you are ready to start processing the data now! Open the hydro_spectro.ipynb file, again checking that you have the necessary packages installed.
Process a singular file:
# some important lines here
Process all the files:
# some important lines here
What do the results mean? What am I looking at??
Understanding these plots is equally as important if not more so than being able to reproduce them.
Description of simulation parameters
| Parameter | Value | Language | Time period | Description |
|---|---|---|---|---|
| $\alpha$ | $1/2$ | NA | 1930–1954 | Tempor dolor in |