Hacking Ecology

 Notes on a first project
 
 Land use change from remote sensing data
 Outline: Get Remote Sensing Data (mostely Satellite Images) automatically from NASA/ESA etc. databases, classify different land use types (forest, city...) using different methods and then present/analyse the data as time series
 Possible outcome: A nice web tool where you can input a GPS position or an area and then get a nice figure with the changes in land use as well as the data so that others can do more detailed analyses

•  Another possibility: Create a large dataset with a global map of land use change and then compare this data to other datasets ("ground data", ecosystem quality metrices) that we will have gahtered until then
• Might be interesting as part of the relevant background
• (though in German)
• https://www.boell.de/de/bodenatlas
• An example how spatial data can be visualised and linked with eco-societal-political arguments:

 https://correctiv.org/top-stories/2017/06/13/irrsinn-der-agrarpolitik/
 Correctiv won UmweltMedienPreis 2018 with this research: https://www.duh.de/ump_preistraeger/
 
 
 
 
 How would this project look like in smaller steps? (PLEASE fill in details and complete this!)
 
 1. Write something that automatically fetches the correct images from NASA/ESA sources.

•  Sources:
• "Correct" in the sense that the user can define the region and a range in time

2. Some Code that automatically classifies parts of the images in terms of land use

• This might need to be optimized to be fast
• Algorithms/tools that do this:
  • https://github.com/tropicalmentat/land-cover-to-land-use-classification
  • https://github.com/shakasom/Deep-Learning-for-Satellite-Imagery/blob/master/LULC_Final.ipynb
  • This blog post might be useful: https://towardsdatascience.com/land-use-land-cover-classification-with-deep-learning-9a5041095ddb

3. Create a nice GUI and command line interface