PlastiTracker

https://biteable.com/watch/3266008/be013fac31daa5a4eb274c1e1d76368e

https://drive.google.com/drive/folders/1MgWAgu5O7BdauLK-BZ8HfjOemgNv8aYh?usp=sharing

In the ocean, plastic waste has effects on the surface that can be detected via remote sensing. We exploited two of these effects in our analysis: 1- plastic causes surface temperatures to rise and 2- plastic blocks chlorophyll from phytoplankton at the surface. We used AQUA/MODIS chlorophyll detections and Global Temperature Anomaly detection data over time to find areas in the ocean that displayed both high surface temperatures and low levels of chlorophyll.

After downloading, cleaning, and transforming the data, we built and tested several AI/ML models in an attempt

to add predictive features to the amassed dataset. This dataset then feeds a geographic chart in Tableau which can be used to interactively analyze the data.

For this project, we added the data downloads to Google Drive and then did all Python coding in Google Colab. After the model runs in Colab, the final dataset is saved in Google Drive where it can be read by Tableau. The dashboard will be published to Tableau Public where it can be viewed publicly.

Through this project, we have found that (aside from time restrictions) there are several benefits and pitfalls to AI/ML of plastic debris tracking. We have found some of the benefits to be: the relative ease of creating a solution, the solution can be cost effective, there is abundant data to aid in the research. Some of the pitfalls of AI/ML implementation are: the reliance on assumptions, the need for field work to validate hypothesis tests (‘is there plastic where we think there is?’), and being beholden to limitations within the data.

We utilized NASA Earth Observations for the two main data sources in this project:

1- Chlorophyll Concentration (1 Month at 0.5

degrees)

2003-2007 & 2019-2021

https://neo.sci.gsfc.nasa.gov/view.php?datasetId=MY1DMM_CHLORA&year=2021

2- Global Temperature Anomaly (1 Month at

0.5 degrees)

2019-2021

https://neo.sci.gsfc.nasa.gov/view.php?datasetId=GISS_TA_M&year=2021

We used the chlorophyll data from 2003-2007

to get baseline average readings by latitude, longitude, and month. We then

used these baseline averages to detect differences in the more recent monthly

chlorophyll data. The Global Temperature Anomaly data already provides its

readings as changes in surface temperature for the given latitude, longitude,

and month.

As a team, we have learned a lot during this hackathon. We have come to appreciate the value of a team vs the individual as well as the importance of time management. We made good progress on developing a plan to answer the questions, collecting data, and transforming the data. However, we ran into technical snags (limited RAM on PC) which forced us to re-work the size of the dataset, the variables in use, and we needed to get creative with pickling and deleting dataframes. Given more time, we believe that would have been able to properly finish our code to work around our constraints, tune the models, and build an interactive geographical dashboard. We have taken away many lessons from this exercise and we feel we are now more prepared to tackle similar challenges in the future.

• Chlorophyll Data: NASA Earth Observations - https://neo.sci.gsfc.nasa.gov/view.php?datasetId=MY1DMM_CHLORA&year=2021
• Global Temperature Anomaly Data: NASA Earth Observations - https://neo.sci.gsfc.nasa.gov/view.php?datasetId=GISS_TA_M&year=2021
• https://www.onegreenplanet.org/environment/plankton-under-threat-tiny-life-in-major-need-of-your-help/
• https://india.mongabay.com/2019/08/what-would-happen-if-the-oceans-are-completely-covered-with-plastic/
• Tableau
• Google Drive
• Google Colab
• Python [numpy, pandas, os, sklearn]
• Biteable

This project has been submitted for consideration during the Judging process.