TriHyb is a Triton-based Remora developed by the Marine Bioacoustics Research Collaborative designed to produce minutely Hybrid Millidecade (HMD) soundscape products (Martin et al. 2021), in alignment with the current standard of the NOAA National Centers for Environmental Information (NCEI). These products can be used for diverse applications, such as studying different environments through sound analysis, identifying sound sources within underwater soundscapes, and comparing soundscapes between different locations. By examining minutely HMD products, users can visualize ambient sounds across a user-defined frequency band. TriHyb collects data from user-specified sites to generate daily netCDF files. From this output, time series plots can be generated to display sound levels across deployments.

Wiki created by: Miguel Gonzalez (Scripps JT-SURF Intern 2025)

Learn how to download or clone the Triton repository with the TriHyb Remora repository in the quick setup section and how to add the Remora to Triton.

In the Control Window, use Remoras pull-down menu, hover over TriHyb, and select Compute HMD Products.

After opening, you will be shown a pop up that contains information on the HARP which is all on the tab Metadata Compiler.

With the Metadata Compiler, the user is able to input information on the HARPS they are looking at; the picture above gives an example of what some info should look like. The user can either manually type in information on their HARP or click on the Import from .xlsx file, which will grab information from any .xlsx file and import it into the corresponding section. It’s important that the headers on the file directly match the ones on the Remora so that it can distribute the information correctly.

The picture above shows what a correctly formatted .xlsx file should look like. I will give some background on what each of the heading means.

• Title: Title of the dataset with resolution and site.
• Summary: Summary of what dataset contains.
• History: Who created the product.
• Source: How data was conducted.
• Acknowledgment: Acknowledgement for funders / supports / collaborators who contributed to the data.
• Citation: Citation for the dataset.
• Conventions:
• Creator Role: Roles of the creators
• ID: Identification of the organization who created the products.
• Institution: Institution where data comes from.
• Keywords: Keywords about the data produces
• License: License number
• Product Version: Version of the product
• References: References that helped in computing the HMD products
• Comment: Comments to include in metadata (depth of site location, elevator of sensor above seafloor, any information about the data that is needed)
• Creator Name: Name of creator
• Creator URL: Creator of project URL link
• Publisher URL: URL link to publisher
• Instrument: Name of the instrument
• Keywords Vocabulary: GCMD Science Keywords
• Naming Authority: NOAA National Centers for Environmental Information
• Publisher Name: Name of Publisher

If you look next to the Metadata Compiler you will see another tab, Compute HMD. This is where you can input which data you are using as well as where your data output goes.

This is what the Compute HMD section looks like on the Remora. Just like the Metadata compiler I will give some background on what each heading means.

• Input Directory: The input directory is where you pick a file to pull data from. You can either pull data from a full deployment or examine individual time frames of data. Each deployment file has multiple “disks” that contain data from different timeframes. For example, if I wanted to look at a time frame from October to November 2023, I would pick CINMS_B_49_disk01 because it has data from September 27, 2023, until November 13, 2023. However, if you want to look at a full deployment of data, you can check the Search Subfolders box on the top right. This will have the program gather all the data from the deployment rather than just one disk.

• Filename Pattern: Since there may be multiple files in one deployment file (df20 and df100), this filename pattern allows you to pick a flag for the program to look for. If I just wanted to run df20 data, I would put "* df20 *" here with asterisks (no spaces).

• Output Directory: The output directory is where you can choose where you want to store the data computed through the Remora. You can make a new folder on your files and have the data stored there.

• Transfer Function File: Transfer function calibration file with .tf file extension

• Organization: Organization that collected and processed the data

• Project: Project name of dataset

• Site Name: Site name of dataset

• Site Location: Coordinates of the site in decimal degrees

• Deployment #: Deployment number

• Deployment Start/End Date: Deployment start and end data in YYMMDD format

• Start/End Frequency: Range of frequencies you want to look at

• Minimum Effort for Minute Bin(%):

As with all software and file formats, working with .netcdf files in MATLAB can sometimes present unexpected issues. Below is a list of quirks we've encountered during development. If you come across additional ones, please let us know—we’ll do our best to resolve them.

• Persistent File Locks : MATLAB occasionally struggles to fully release .netcdf files after opening or writing to them. Even after issuing a close, the file may remain locked for the remainder of the session. If you’re unable to access or overwrite a .netcdf file, try running close all or restarting MATLAB entirely before attempting to open or write to the file again.
• Overwrite Restrictions (CLOBBER Issues) : NetCDF-4 files do not support overwriting existing files (also known as clobbering) by default. If you attempt to write to a file that already exists, the operation may fail silently or return an error. Create a new output folder or manually delete the existing file before running your deployment or disk-writing script.
• Deck test/Recovery: When deploying the site, the crew conducts a test on the HARPs to ensure data is being recorded properly. During this time, loud noises from the crew and surface activities are also captured. To prevent these deck test noises from appearing in long-term analyses, it is recommended to delete the first and last days of data from the output folder after running TriHyb.

TriHyb was developed by the Marine Bioacoustics Research Collaborative to compute hybrid-millidecade bands in acoustic data collected by High-frequency Acoustic Recording Packages (HARPs). With the HMD products, users can analyze long-term time series to identify different sound sources, such as ships, whales, and fish, and observe the duration of these sound sources within the soundscape. Examples of source sources may include: ship noise with typical ship noise creating high amplitudes below 100 Hz; blue whales, which vocalize at low frequencies around 10–40 Hz; humpback whales make sounds ranging between 10 and 2 kHz. Users can define the target frequency band to identify sound sources of interest. A typical frequency range may be 10–4000 Hz to capture noise from both ships and whales.

All natural sounds in the ocean consist of multiple frequencies with varying intensities (spectrum levels). Sounds emitted by ships, however, produce persistent high-intensity spectrum levels, disrupting the ocean's natural soundscape. Researchers use visualization tools like spectrograms to analyze and display the amplitudes of different frequencies over time.

_{Figure 1: An example of a spectrogram that is created through the TriHyb remora. The legend that shows spectrum level (PSD) is in the box labeled A. Box B shows what a ship looks like on the graph when passing through the site. Box C is an example of what a humpback whale calls looks like on the spectrogram.}

This is an example of a spectrogram (Figure 1) generated using the TriHyb Remora. At low frequencies, ships generate intense noise. The color bar to the right of the graph (Box A) shows the spectrum levels, bright yellow displaying the high amplitudes and dark blue displaying the lowest. The high-energy signal (Box B) is a ship passing by. As you can see, the intensity of the ship noise is represented by a bright yellow, which is high amplitude and can significantly disrupt marine mammal communication that is within the same frequency band. To give an example, it is even louder than an airplane flying over you, which can be disruptive. In the smaller red box (Box C), you can see a lower energy sound signal, which indicates humpback whale calls!

_{Figure 2: Shows a day where a lot of ships passed by our site on March 19, 2024.}

This is another example graph (Figure 2), the data shows that on March 19, 2024 there were about 11 ships that passed the Channel Islands. Imagine if you heard that many airplanes pass by you each day. How can you be productive and relaxed when all of these ships are passing at almost every moment of the day?

_{Figure 3: Shows a line graph that was created using the PSD averages from the remora. It shows average psd levels from our first CINMS deployment.}

In Figure 3, we show an example of an average HMD spectra for a whole deployment in the Santa Barbara Channel. The graph is plotted with frequency (Hz) on the x-axis and HMD levels on the y-axis. As shown on the graph, the spectrum levels are higher in the lower frequency range (left side), indicating high noise intensity at low-frequencies, likely due to the intense ship traffic in the Santa Barbara Channel.

_{Figure 4: Shows a long-term spectrogram that was calculated using a year's worth of data, which was computed using TriHyb.}

Figure 4 provides another example of why TriHyb was developed. The graph is a spectrogram generated from data collected at our Channel Islands site, spanning from October 2023 to January 2025. It shows consistently high average PSD levels throughout the year, indicating frequent loud noise near the site.

_{Figure 5: Shows long term average PSD levels at different frequencies with data gathered through TriHyb.}

Lastly, Figure 5 shows the average PSD levels at different frequencies. This is useful for analyzing PSD at specific frequencies. For example, ships typically generate noise around 63 Hz, so we can use the graph to assess how often ships produce noise or to determine the average PSD level throughout the year.