The Challenge | Where's the Water?
Use satellite and other data to allow farmers, landowners, and land managers in your locale to identify and visualize water resources in their surroundings.
Check out the live demo at http://tinyurl.com/WaterTrackerApp
Water Tracker will help monitor water-related ecosystems including wetlands, rivers, aquifers, and lakes. Any sudden or large changes in water quantity can be identified and a machine learning algorithm can be applied to correlate it with environmental conditions and find the potential causes.
Overview:
- Water
Tracker is a neural network to locate and track water (in near-real-time) using
satellite imagery
- It can
be used by farmers, scientists, researchers, and citizens through an
easy-to-use web-app
- A possible use is finding water during a drought
- Another potential
use for this data is flood prediction by correlating water movement with
elevation data
- Also, it can be used to analyse water movement over time
- It
is trained on known water locations and
crowdsourced data
Features:
- A user can open the web-app and choose some options, then predict water locations using the neural network:
- Precision
–
More precise calculations take longer, but can detect smaller bodies of water
- Date
–
The system is able to use archived data to track changes in water over time
- It
will show the areas which are water and the areas which are (most likely) not
water
- There
is also an option to download the data for further analysis instead of viewing
it on the map
Future Direction:
- More training data for neural network – Using more training data (from existing topographical maps) would allow the neural network to be smarter and more accurate.
- More types of datasets – The project could also be improved by using more datasets along with satellite data (i.e. LIDAR from satellites and other NASA data which is not publicly available)
- Mobile App - A simple, easy-to-use mobile app for anybody to easily view water location data.
Data Used:
- United States Geographical Survey Earth Explorer & Datasets
- Sentinel-2 Satellite Imagery
- NASA Shuttle Radar Topography Mission Water Data (Only for training data for neural network)
Technology Used:
- Neural Network - Tensorflow Machine Learning library (with Python)
- Web App - Node.js with HTML5 & JavaScript
- Data Storage - GeoTIFF & CSV files