SkyFolks!| Automated Detection of Hazards

Automated Detection of Hazards

Countless phenomena such as floods, fires, and algae blooms routinely impact ecosystems, economies, and human safety. Your challenge is to use satellite data to create a machine learning model that detects a specific phenomenon and build an interface that not only displays the detected phenomenon, but also layers it alongside ancillary data to help researchers and decision-makers better understand its impacts and scope.

SkyFolks - Automatic Hazard Detection

Summary

Automated Huricane Detection by using Machine Learning models for prediction of sattelite data from MODIS. The final application consists from a web application with an interactive map that highlights Hazards world-wide and gives to the user the ability topick a region on the map and check if the probability to have a Hurricane in the area.Main functionalities:- Web application - Machine Learning model build with the transverse bands dataset- Real time data from MODIS- Predictions of possible hurricanes in areas of choice- Interactive world map

How We Addressed This Challenge

Final result:

https://drive.google.com/file/d/1Nd-fI9TDo2n7bYOvaog-tVOKNnhFgdOH/view?usp=sharing


The challenge is to automatically detect hazards, therefore we build a Machine Learning model using open source NASA satellite images dataset that predicts Transversal Cirrus Bands on an interactive world-map.






The application consists of 3 main services:


  • ui-service
  • hazard-prediction-service
  • hazard-analytics-service



ui-service

Interactive Web interface build with React and Redux, HTML5 and CSS. Consists of 3 main sections (Menu section, Map sections, Analytics Section). Interactive map using Google Maps API and the satellite view.



hazard-prediction-service

The role of this service is to use call the MODIS APi to get a satellite data in real time from the location selected by the user. After that to apply the Machine Learning model and predict if in the location selected there is an probability to be an Transverse Cirrus Band like hurricane or thunderstorms.


hazard-analytics-service

Expose different kind of statistics and events about hazards around the globe. This service is aiming to reuse existing Kaggle datasets and mine insightfull analytics so that are expose in the web application.

How We Developed This Project

We build a fully working web application wich contains a interactive world map and does predict automatically hazards (huricanes) in the region that the user selects.


It uses the lates Machine Learning techniques (Random Forest algorithm) to make prediction and categorise satellite images as isHurricane or isNotHurricane.


Real-time calls and retrieve of satellite images from MODIS which are after that passed as imputed to the ML Model mentioned above for calculating the probabilities.


Architecure wise:




We developed those services using:



  • ui-service: Javascript, HTML5, React Framework, Redux, CSS, Material Design, Google Maps API
  • prediction-service: Python, Pandas, Scikit-Learn, Matplotlib, Flask
  • hazard-analytics-service: Python, Pandas, Matplot, Numpy, Flask



Gode on Github and links are at the bottom of the page.

How We Used Space Agency Data in This Project

Used MODIS sattelite data



  • MODIS


Used the open dataset posted for this challenge with satellite data for Transverse Bands Data



  • Transverse Cirrus Bands are bands of clouds oriented perpendicular to the atmospheric flow in which they are embedded. TCBs are often an indicator of strong turbulence and often associated with severe weather such as hurricanes and thunderstorms or atmospheric jets.
  • Transverse Cirrus Bands Data: s3://impact-datashare/transverse_bands/
  • This dataset has been used in training the prediction model.


In addition we used some Kaggle open datasets for wildfires in (Brazil, Australia, California)



  • The purpose of this data is to analyse economic, social and geographic impact that arises from wildfires damages
  • Analysis will consist of building correlation matrixes, barcharts as well as graphs to better showcase the data relations in order to help people understanding more about the impacts


Project Demo
Project Code
Data & Resources
  1. Phenomena Detection challenge Resources | Transversal Cirrus Bands dataset that was specific for this challenge available on AWS S3. https://github.com/nasa/spaceapps-phenomena_detection/tree/dev/data/labeled )
  2. Euro Data Cube via MODIS Nasa API. Documentation: https://cmr.earthdata.nasa.gov/search/site/docs/search/api.html
  3. Example on how to query MODIS providede in the challenge resources | https://github.com/nasa/spaceapps-phenomena_detection/blob/dev/examples/cmr_query/cmr_search_example.ipynb
  4. World view tool from Earth Data to explore events datasets | https://worldview.earthdata.nasa.gov/
  5. Hurricane Damage Overvie Notebook | Kaggle | https://www.kaggle.com/kmader/hurricane-damage-overview
  6. Feature-based Damage Classification Notebook | Kaggle | https://www.kaggle.com/kmader/feature-based-damage-classification#Linear-Regression-Model
  7. California Wildfire Dataset | Kaggle | https://www.kaggle.com/ananthu017/california-wildfire-incidents-20132020
  8. Forest Fires in Brazil | Kaggle | https://www.kaggle.com/gustavomodelli/forest-fires-in-brazil
  9. Huricane Damages Dataset| Kaggle | https://www.kaggle.com/kmader/satellite-images-of-hurricane-damage?
  10. Huricane Dataset Dataset| Kaggle | https://www.kaggle.com/noaa/hurricane-database
  11. Australia Wildfire Dataset | kaggle | https://www.kaggle.com/carlosparadis/fires-from-space-australia-and-new-zeland
Tags
#artificial_inteligence #automatic_hazard_detection #machine_learning #world_map
Judging
This project was submitted for consideration during the Space Apps Judging process.