Poseidon

https://docs.google.com/presentation/d/1mdp9VZPHhr8T0UGlnCwqqQwuXHfB2G7rYOJ8Ly6Tzpg/edit?usp=drivesdk

https://github.com/nishant42491/Nasa_Poseidon_Marine_Debris

Hypothesis-

To find a co-relation between oceanographic features such as Pressure, humidity, temperature and wind speed with Marine Debris. We aim to identify, observe and monitor the marine debris in ocean with the help of deep learning model

Reference: Research Paper by the University of Michigan

The above paper derived the relation between GDAS(Global Data Assimilation System) wind speed vectors and MSS (Mean Square Slope) values.

Conclusion showed a direct proportionality of MSS with the amount of plastic Marine Debris found.

Background:

Because of their relatively low concentrations in the environment sampling of microplastic particles generally requires large sample volumes. Thus, samples from the open water were usually taken with plankton nets of different mesh sizes. The sea surface is sampled for floating microplastics by manta trawls or neuston nets. While neuston catamarans can be operated even in higher waves, a manta trawl is best used in calm waters to prevent hopping on waves and damage to the device. Floating Bouys too contribute to the list of physical devices used to measure levels of plastic debris in our oceans

Methodology:

Data Exploration:

• One of the most important segments of creating deep learning models is data exploration and refining. We used 1D and Geo2D NetCDF files obtained from NASA CYGNSS datasets to get the latitude, longitude, mss and wind speed
• Furthermore, we removed the NAN and empty values to refine the datasets and merge the necessary attributes together.
• We used techniques like binary search and matrix Computation to co-locate records in our database
• We then extracted important features like Pressure, humidity, sea surface temperature, and wind speed from other PODAAC -JPL datasets
• To calibrate the deep learning model with the wind speed vectors, we made use of GDAS datasets and extracted values of wind speed vectors like 'u' and 'v' using a sequential methodology and algorithms
• These datasets are used to provide MSS Anomaly which is used to predict the mss score ( directly related to the quantity of marine debris)

Machine Learning:

Our method implements a deep learning algorithm known as multi level perception and a ML model called as Random-Forest Regressor to compute the relation between MSS and the physical features/

• Created a multi-level perceptron model consisting of 4 layers and 401 training parameters.
• Model consisted of sequential linear units, followed by relu activation functions.
• It was a supervised training method where MSS values were used as targets.
• The predicted MSS values, we received after feeding features to our model are used to calculate a matric called ''mss anomaly''
• MSS Anomaly is then used to quantify the amount of plastic pollutants in our oceans.
• MSS scaled values closer to 0 indicate higher presence of marine debris while the values closer to 1 indicate lower presence of marine debris
• Furthermore, trained a Random-Forest regressor using the scikit-learn library
• Helped us find the co-relation between different features and its effect on the model

A ) The multi level Perceptron Model

B ) The Random Forest Regressor :

The Random Forest Regressor model identifies and correlates physical features like pressure, temperature, wind speed and humidity. Below you will find the heatmap result.

Our Findings Show that MSS is heavily co-related with wind speed and mildly co-related with other factors such as temperature pressure and humidity.

Data Visualizations :

Here we plot some of our findings of the quantity of plastic marine debris polluting our oceans on a world map.

Application Programming Interface:

• Created a Flask API which handles GET and POST requests
• Latitude and Longitude as parameters and returns corresponding MSS anomaly, physical features
• Using pandas and numpy, it calculates the shortest length between two geo points.

Android Application:

• Created an android application using Android Studio and Java Programming Language
• Integrated Google Maps API to point markers at the users location and the location of nearest marine debris with scientific attributes added to it
• Called a JSON request to the above Flask API, to obtain mss_anomaly, pressure, humidity, temperature, and wind speed

What our next step would be if given more time to work on this project:

As a further extension to the applications, we can include people's contributions with the help of crowdsourcing. Application intakes users' current latitude and longitude along with the other attributes like temperature, surface pressure, humidity, etc using google Earth data APIs and PODAAC API

The further UI can be simplified so that people who aren't pretty comfortable with scientific applications will find easy to contribute

Github Link : nishant42491/Nasa_Poseidon_Marine_Debris (github.com)

CYGNSS datasets: CYGNSS Level 2 Science Data Record Version 3.0 | PO.DAAC / JPL / NASA

Podaac datasets for oceanraphiocal physical factors :

1 : Wind speed : MetOp-A ASCAT Level 2 25.0 km Ocean Surface Wind Vectors | PO.DAAC / JPL / NASA

2 : Sea surface temperature : GHRSST Level 4 MUR Global Foundation Sea Surface Temperature Analysis (v4.1) | PO.DAAC / JPL / NASA

 3 : Humidity,Pressure, Temperature : ECCO Atmosphere Surface Temperature, Humidity, Wind, and Pressure - Daily Mean 0.5 Degree (Version 4 Release 4) | PO.DAAC / JPL / NASA

GDAS : Global Ocean Currents Database | National Centers for Environmental Information (NCEI) (noaa.gov)

Nasa in collaboration with the SpaceApps India did a splendid job in hosting a wonderful 36hrs hackathon. It's great to see so much participation from all members and we wish them all the very best. We would like to thank all the co-ordinators who made this event possible without which we could never have imagined creating such a project for ourselves.

We chose this challenge as Plastic Marine Debris has been the number one cause for all manmade marine life deaths occurring in the oceans.

What better solution to make if not for Mother Nature itself.

This apps aims to create awareness of the amount of Debris present in our oceans and how as citizens of the planet we should come together in solving this crisis

Our approach was a very simple one, we analyzed multiple research papers from various universities and decided on our hypothesis.

Data collection and cleaning was a difficult task but nonetheless after that process we went on to run ML models and created an API for the dataset that we got from the model. Creating the app was simple as we didnt complicate with the UI/UX

An extended thank you to all the people who have come together to create such valuable datasets for us common people to use.

Resources:

University of Michigan:

Resources: Toward the Detection and Imaging of Ocean Microplastics With a Spaceborne Radar

https://ieeexplore.ieee.org/document/9449485

Softwares and Libraries used:

Google Co-lab

Jupyter Notebook

Android Studio

Visual Studio Code

Atom

Rainforest

Sci-Kit Learn

Matplotlib

pandas

numpy

Flask

#poseidon #water #ml #ai #saynotoplastic #SpaceAppIndia #SpaceAppIndia #Mumbai #SUMVN #Goalstreet

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