High-Level Project Summary
Landslides occur in all 50 states. Causing between 25 and 50 deaths every year in the US alone and thousands more in vulnerable areas around the globe. Landslides include debris avalanches, debris flows down carrying large rocks, block movements causing huge damage and loss to people, their homes, farms and cattle, industrial and other structures. Our project detects landslides' whereabouts and predicts their level of risk, potential casualties, and the cost of their impacts. Using a user-friendly website ,machine learning model trained on data from NASA satellites and ground-based sources. providing affected communities and governments with the information needed for timely decision-making.
Link to Project "Demo"
Link to Final Project
Detailed Project Description
How We Addressed This Challenge
Our team has developed a machine learning model that was trained on NASA's satellite's data of wind, rainfalls, and earthquakes from The Integrated Multi-satellite Retrievals (IMERG) algorithm that combines information from the early precipitation estimates collected during the operation of the Tropical Rainfall Measuring Mission (TRMM) satellite (2000 - 2015) and the recent precipitation estimates collected from the Global Precipitation Measurement (GPM) satellite (2014 - present).
The final project is an interactive, user-friendly website that will include our training datasets as well as understandable analysis, conclusions, and visualizations to help researchers and scientists to integrate them into their processes, and help to provide individuals with the information they need to easily find out whether or not that they are at risk of a landslide.
Importance Of Our Project:
- Providing forecasts to help governments take timely decision-making.
- Allowing rural communities to know whether or not they are at risk of landslides.
- Giving governments and communities enough time to prepare themselves before the occurrences of a landslide.
- Contributing to saving the natural environment from landslides' hazards.
- Authority have a scientific and statistically data for decision making.
Methodology Of Our Project:
Whenever the user specifies a location and time in our website heat map, the map shows if there have been previous landslides in that place and the model makes a forecast of the upcoming landslide date and the extent of damage caused in that particular structure.
Tools / Coding Languages/ Software Used:
- Pandas Library
- Numpy Library
- Scikit learn Library
- LGB machine learning model
- Python
- HTML
- CSS
Space Agency Data
Hackathon Journey
Problems Faced:
One of the biggest challenges that our team faced was gathering and collecting the data in a way that was usable while identifying the most useful data points to keep the size of our data set manageable. We spent 15+ hours figuring out how to efficiently pull satellite imagery from NASA’s worldview database.
Due to the lack of time, we weren't able to implement all of our ideas starting from collecting variety of landslides factors' datasets to making a front-end and back-end website that meets our expectations.
References
Resources
https://gpm.nasa.gov/data/imerg
https://gpm.nasa.gov/landslides/resources.html
https://gpm.nasa.gov/applications/landslides#modelingandreportinglandslides
https://asf.alaska.edu/data-sets/sar-data-sets/alos-palsar/
https://sentinel.esa.int/web/sentinel/thematic-areas/emergency-management/landslides
https://coast.noaa.gov/digitalcoast/data/home.html
Project Demo
https://bit.ly/3uzHGlk
Code on Github
Global Judging
This project has been submitted for consideration during the Judging process.