Movement Corridor Detection

MoveApps

Github repository: https://github.com/chei90/EMAC_Corridors

Description

This app is about detecting corridors of animal movement. The current output of this app is a html file (containing a map, can be opened with browser) that highlights areas where more individuals have traversed through. The darker the color, the more individuals were recorded.

EuroDeer Roe deer in Italy 2005-2008

North Sea population tracks of greater white-fronted geese 2014-2017 (data from Klzsch et al. 2019)

LifeTrack White Stork SW Germany

Documentation

1. The input data is stripped from all unwanted data. Only the individual id, the timestamp and position are kept.
2. The trajectories are simplified by using Ramer-Douglas-Peucker-Algorithm. The magnitude of simplification can be controlled via the Rdp Epsilon parameter. As described in the Settings section.
3. The trajectories are traced against a grid
4. The number of different individuals per grid is counted
5. Adjacent cells (moore neighbourhood) are joined, the MCP is calculated.
6. The sum of mcps is simplified and returned.

Input data

The app expects a TrajectoryCollection whose GeoDataFrames contain the following columns:

• timestamps - column containing the time when the data point was recorded
• individual.local.identifier - column containing the distinct individual id
• geometry - point geometry (lon, lat), the gps/surface position where the data was recorded

Output data

This app produces multiple polygons that denote the different passage corridors. Since TrajectoryCollection doesn't support polygons, the data has to be provided as artifact (downloadable content). See below.

In consequence: the output data of this app is not intended to be processed further. The app outputs the input data.

Artifacts

• corridors_map.html - Can be opened with the browser. Contains an OpenStreetmap with the computation results as overlays.
• Shapefile - Verified compatible with QGis and Arcgis.

Settings

All settings values are positive integers, so negative values are not allowed.

• Rdp Epsilon - The Ramer-Douglas-Peucker-Algorithm is used to reduce the trajectories point count, hence improving the computation time. Greater values will reduce the point count per trajectory, smaller values will preserve more points. Think of it like, remove every point from the trajectory that is closer to a given start/end segment than N meters.
• Grid resolution - The resolution of the raster (in meters) that is used to bin different individuals traversing map segments. Smaller values will make the result more fine grained, but will probably reduce the avg amount of captured individuals per raster cell (as it's smaller now). Result of very small values (e.g. 50 meters) will look mostly like the input trajectories. Reducing the grid size will increase computation time.
• Capture threshold - The minimum number of captured individuals per grid cell to appear as polygon on the map. Grid cells with an individual count lower than this number will not appear.

Working Example

This section is intended to give you a good basis on where to start while checking out the app. Create a new workflow:

If you want to start by reproducing my results, choose the following movebank dataset as datasource:

Note Don't forget to use the translator app that converts R dataframes to pythons' Trajectorycollection (see in the workflow above: moveStack to MovingPandas).

Open the App Settings of Movement Corridor Detection:

Set the settings to:

• Rdp Epsilon: 350
• Grid resolution: 2000
• Capture threshold: 2

You will get the following result