1 of 13

Predictive Foobar: new method for predicting Foobar Hackosa spawning events using Argos

math goes here

2 of 13

Project idea: brief description of project goal or focus

Pitch + Ideation: �Fill in these once you start discussion so others can chime in

  • Zoom breakout room #
  • Ideation board: URL
  • Slack channel: ohw23_proj_foobar

Build the project team:�Fill in these once project group is somewhat established

  • Github repo: URL
  • Team members: AAA (UTC-7), BBB (UTC+3), CCC (UTC+8), …

Other content to help people understand the project!

3 of 13

Inertial oscillations in the marginal ice zone

This project will satellite imagery to track inertial oscillations at the sea ice edge. These currents are driven by storms and contribute to ocean mixing.

  • Zoom breakout room 3
  • Ideation board
  • Github URL
  • Slack: owh23_proj_sea_ice_oscillations

1) Source and plot satellite imagery

2) Apply object tracking algorithm

3) Learn about ocean currents

Tools and skills: Imagery

Acquiring and plotting imagery from multiple platforms.

Satpy seems useful!

Tools and skills: Floe tracking

Image processing and computer �vision methods for tracking

OpenCV’s Optical Flow seems useful!

Team members: �Laura C. (Seattle in-person)�Dalton KS (virtual) … Aditya Sharma (Australia in-person)

4 of 13

The main idea of this project is to automate the process of identifying a possible oil slick in a satellite image, cropping the area of ​​interest to then start validating whether it is oil or a look-alike.

  1. Read the satellite image file

2) Histogram and statistical analysis to find the possible oiled area

3) Crop the area of interest and start validations

Project idea: Oil spill Monitoring: Segmentation of Satellite Imagery

  • Zoom breakout room # 3
  • Ideation board: link
  • Slack channel: ohw23_proj_oil
  • Github repo: https://github.com/oceanhackweek/ohw23-oil
  • Team members: Amanda D. (UTC-7), Maria G. (UTC-3), José G. (UTC-3), Jaelyn Bos (UTC -4), Anand Sekar (UTC-7)

5 of 13

Project idea: Mooring processing and data page 📜

Pitch + Ideation: We want to develop some data visualization tools for mooring data in Puget Sound and La Perouse Bank. Ideally, these could be interactive plots, hosted on Github Pages, where users could explore the data on their own. The learning goals are to improve skills in Github pages, data interactivity, and data visualizationZoom breakout room #❓🔮

Build the project team:�Github repo: https://github.com/oceanhackweek/ohw-fancymoorings

  • Team members: Andrea, Lu, Seth, Sam, Tobias, Johnathan, Hafeez (UTC-7), Hameed, Halley

6 of 13

Project idea: Marine Species Distribution Modeling tutorial: sea turtles

There are many Species Distribution Model (SDM) tutorials for land applications and the applications for marine SDMs from marine reserves or fishing zones, evaluating impacts of human use of the ocean, estimating fish densities. However there are almost no marine SDM tutorials available. A few basic tutorials with marine examples would be very helpful for people who want to learn how to create marine SDMs and could be used for future OceanHackweeks. Eli Holmes (who pitched idea) will use this in an upcoming Ocean hackweek in India in Sept 2023. She has a ‘analysis ready zarr file’ with environmental variables that we can use for the Arabian Sea and Bay of Bengal.

Proposed tutorial (Jupyter notebook or Rmarkdown file) with code:

  • Start to the Rmarkdown file: https://github.com/oceanhackweek/ohw23_proj_marinesdms/blob/main/tutorial/tutorial1.Rmd
  • Background on SDM models (4 paragraphs + references)
  • How to access and format marine species presence/absence data:
    • Where you can find marine spp occurrence data? how to get it in the format needed
  • What kind of environmental variables do you need
    • For the project we have a pre-prepared zarr file of env variables for N Indian Ocean but could add more variables
    • What sort of variables are good for marine species (we could interview the OHW participants to flesh out more variables that we might use)
    • Where you can find the environmental data
  • How to run a SDM models using the biological data and environmental variables
    • Example with dismo R package
  • Visualization of SDM model output

Project team pages

  • Ideation board: Slide 8
  • Slack channel: #ohw23_proj_marinesdms
  • Github repo: ohw23_proj_marinesdms
  • Team members: Eli H, Paulo F, Caitlin O, Mary S, Laura T, CCC (UTC+8), …

SDM example

7 of 13

Project idea: SST spatial distribution prediction using machine learning

Pitch + Ideation: Predict SST anomalies (upwelling, etc), generate SST spatial distribution, the model can also be used with other type of data as input! We plan to use satellite data (MUR)

STEPS:

  1. Data, boundary box (time [2000-2020], lat [-5,32], lon[45,90])
  2. Split data on training, validation and testing datasets
  3. Model Architecture: we plan to use TensorFlow
    1. ConvLSTM
    2. 3D CNN
    3. Transformers
    4. Hybrid: CNN + Transformer + LSTM
    5. * SHAPE CORRECT
  4. Complie and fit
    • Early-stop
    • * Loss function: MSE, MAE, SSIM
    • * metric
  5. Visualization of result and Interpretation!

  • Ideation board: https://jamboard.google.com/d/1lOgVwnqQLvNRPAOEVEGnWXm8FSTuPYQWbteptKrslTM/viewer?f=10
  • Slack channel: ohw23_proj_sst

Build the project team:�Github repo: URL: https://github.com/oceanhackweek/ohw23_proj_sst/

  • Team members: Jiurai Yu, Boris, Paula Birocchi (Seattle, US), Hao Tang, Danyang Li, Chandrama Sarker, Zhengxi Zhou? Does anyone want to join us?cc

8 of 13

Machine Learning for Argo Data QC

PROJECT: Apply a ML Approach for Argo Data QC to improve the quality of the data

HOW:

  • Develop python scripts to process Argo Floats data from GDAC. The focus will be on obtaining real-time and delayed mode core Argo (T, S and P)
  • Apply an unsupervised Machine Learning model (Gaussian Mixture Model - GMM), using a Profile Classification Model. With this technique, it is expected to be able to identify outliers in the data.
  • Apply IOOS QC on Argo data? Generate Spikes?
  • Apply a Dense Neural Network model on a training set of models and then test with the entire dataset

Build the project team:�Github repo: URL: https://github.com/oceanhackweek/ohw23_proj_argo_ml

Goal: Better determination of suspicious data in Argo Data

9 of 13

Project: Benthic habitat mapping (image processing/seabed classification)

Pitch + Ideation: Habitat Mapping using Irish national seabed mapping program INFOMAR multibeam echosounder data- bathymetry and backscatter data and utilise a range of machine learning techniques in SciKit-Image Python library.

Start off simple to increasingly complex models and utilise the confusion matrix for accuracy assessment.

Possible habitats include coral reefs, continental shelf or slope or more exciting environments from national seabed mapping data!!

  • Zoom breakout room # 7 Benthic Habitat Mapping
  • Ideation board: Slide18
  • Slack channel: ohw23_project_habitatmapping

Build the project team:

  • Github repo: URL
  • Team members: Siddhi Joshi (UTC+1), Minh Phan (UTC?), Colin Sauze (UTC?), Valentina (mentor)

Aim for Tuesday: Get data loaded to Jupyterhub shared folder

10 of 13

Direct Georeferencing of Drone Images

Using drones over the ocean makes it difficult to use structure Shallow water areas we can use structure from motion (SfM), in deeper water with waves SfM fails.

Is this the same turtle?

Where is this rock?

Where is this point in a 10m satellite pixel?

  • Slack channel: #ohw23_proj_drone_georef
  • oceanhackweek/ohw23_proj_drone_georef
  • Team members: Alson, Diana, Paul, Alessio, Nick

11 of 13

Investigating the variability of the suppression of the South Australian upwelling.

Context: Playing with the IMOS mooring data in the past Ocean Hack, I noticed a suppression of South Australian upwelling that lead to a question:

What is the variability of the South Australian upwelling and what is driving it?

12 of 13

Passive acoustics project: linking soundscape metrics with other ocean variables

Idea/pitch: use open access acoustics datasets and extract some acoustics metrics from the soundscape. Then fetch some other variables available for the same locations to explore links and correlations.

Steps:

  • Make a team: Lucille, Minh,
  • Build github repo: ohw23-proj-pamproject
  • Choose OA datasets
  • Fetch data
  • Extract acoustics metrics: LTSAs, SPLs, power spectra vs time, …
  • Plots/visualisations
  • Choose other variables to find: temp, pH, current/flow, tide, moon phase, animal presence/absence…
  • Fetch data and visualisations
  • Correlations and statistic analysis

13 of 13

Project Idea: Compare Different Bioinformatic Pipelines to Produce Standardized Output For Amplicon Sequence Data

Does The Pipeline Chosen Affect The Biological Interpretation?

Possible Pipelines that are available on GitHub using different tools:

Could possibly be incorporated into the Species Distribution Modelling Idea (US Proposed); - will the differences in bioinformatic pipeline chosen lead to difference in species distribution; eDNA data becomes the biological input data into the species distribution maps

Data integration: incorporate abiotic factors into analyses; ideally depth, habitat and other relevalent abiotic variables

Goal for Wednesday: demultiplex data, upload it to the cloud, create samplesheets for different pipelines, run pipelines, identify problems/bugs.

Clustered image map showing integration between 16S data and abiotic parameters. Source: http://mixomics.org/mixkernel/