Implementation of a baseline - Githubissues

ai4er-cdt / geograph

GeoGraph provides a tool for analysing habitat fragmentation and related problems in landscape ecology. GeoGraph builds a geospatially referenced graph from land cover or field survey data and enables graph-based landscape ecology analysis as well as interactive visualizations.

https://geograph.readthedocs.io

MIT License

39 stars 10 forks source link

Implementation of a baseline #9

Closed sdat2 closed 3 years ago

sdat2 commented 3 years ago

EDIT: Remaining things to be done:

[x] Get ESA CCI data as test data
[x] Define region of interest
[x] Clip raster to region of interest
[x] Separate zones
[x] Reproject to suitable CRS
[x] Perform pylandstats analysis on timeseries from 1992 to 2015
[x] Sanity checks
[ ] Transfer final pipeline code from jupyter notebooks to python script
[ ] Finalize plotting procedures and add to python script

Croydon-Brixton commented 3 years ago

Purpose: Create a baseline of the currently existing fragmentation algorithms (probably starting from land cover data)

Requirements:

[x] Landcover data (?) to compute the metrics on
[x] (Anything else?)

Baseline metrics/algorithms:

Percolation theory (Nature paper)
Papers by Fahrig (Fahrig 2003, Jaeger 2000)
R package from James Ball --> in Python (https://r-spatialecology.github.io/landscapemetrics/)

Croydon-Brixton commented 3 years ago

@arduinfindeis @KMacfarlaneGreen @sdat2 I opened a branch for us to work on this: feature/fragmentation-baseline

Croydon-Brixton commented 3 years ago

Update: I started to add a few convenience scripts for working with rasterio and plotting land-cover data, as I imagine we will use this a lot.

For an explanation of the first steps with Rasterio, see this notebook.

@KMacfarlaneGreen has already done a first evaluation of some fragmentation characteristics via pylandstat around the region of London in this notebook here - thanks Katie!

Next step will be to compute this efficiently without having to do to-disk memory operations, so that we can eventually do it on larger areas (pylandstats is also quite slow at the moment).

Note: We can discuss the contents of the notebooks nicely via the links posted above, on ReviewNB

Croydon-Brixton commented 3 years ago

Update: We now have the geometry of the exclusion zone, as well as buffer zones (30 km & 60 km) loaded from Open Street Map. Further, the ESA CCI Landcover data from 1992-2015, cut to our region of interest, is available in our joint group workspace directory.

This means we can now properly start the fragmentation analysis covering the years 1992-2015.

For more information, see this notebook

Croydon-Brixton commented 3 years ago

Update: Spatiotemporal analysis in zone is performed in notebook no. 5 and saved to disk in netCDF form. It can be loaded via xarray as shown here to plot.

The next step in this workstream is to produce some compelling visualizations of a few key metrics to guide our general analysis.

Once that is done, this workstream is complete (and code can be re-run with our own habitat data once it's ready).

Croydon-Brixton commented 3 years ago

Remaining things to be done:

[ ] Transfer final pipeline code from jupyter notebooks to python script
[ ] Finalize plotting procedures and add to python script