Difference between revisions of "Movement Analysis"

Latest revision as of 10:35, 2 March 2021

We have create a toolbox for the free GIS software OpenJUMP that allows to analyse movement data, called OpenJUMP HoRAE - Home Range Analysis and Estimation (pronounced: hooray). The software is and can be freely distributed under the General Public License verison 2.0 (GPL) - so you can give it to your co-worker or even modify it.

Update: 28. Apr. 2016 - fixed bug that prevented correct execution of Asymptote Analysis HR method
Update: 7. Mar. 2015 - changed some download and document links, since our file hosting provider stopped offering private file download service.
Update: 19. Nov. 2014 - A version packaged with OpenJUMP 1.7.1 was released. No warranties though. Please report to stefan---geosteiniger.cl if you spot differences.
Update: 20. Feb. 2013 - A new version has been released including a new CSV file reader, adding the GeoEllipse method by JA Long and TA Nelson (2012), and a function to extract time attributes from date fields.

About the Home Range Analysis Toolbox

OpenJUMP HoRAE Screenshot

The toolbox contains functions for visualization of movement tracks and allows the creation and analysis of animal home ranges. Home range estimation methods that have been implemented are (see also the article below):

• Minimum Convex Polygon (MCP),
• Kernel Density Estimation (KDE), methods: (a) href, (b) LSCV, and (c) ad-hoc (region split)
• Line-based KDE, scaled and un-scaled
• Brownian Bridges
• Line Buffer
• Local Convex Hulls (LoCoH), methods: radius r, k-1 nearest neighbors, alpha region.
• Potential Path Area / Geo-Ellipse (in new version only, see above)

Additionally one can:

• Perform an asymptote analysis for MCP, Point-KDE, Line-Buffer
• Calculate core areas
• Classify home range parts
• Derive the skeleton for a home range region.
• Analyze the daily travel

The toolbox was originally developed to analyse location data from grizzly bears (ursus arctos) collected with GPS collars within the Grizzly Bear Programm of the Foothills Research Institute (FRI, Alberta, Canada). Funding was provided by Canadian Phase IV GEOIDE grant (#3).

Download

The software should run on Windows and Linux systems. MacOSX users can use the same version.

• download software - version from Apr. 2016 ca. 48 MB including documentation - this version is packaged with OpenJUMP 1.7.1 instead of 1.6 as for the March 2013 version; (Note, a backup copy on another server can be downloaded from here)

  older well-tested versions are still available: HoraeMar2013 HoraeNov2011

• install instructions - note, the software requires Java Runtime Environment (JRE) installed
• a short article (pdf) that describes the functions of the toolbox (this is the 1st version; the final is here Wildlife Society Bulletin)

Some preliminary results for a comparison of the estimators with grizzly bear data: GIScience 2010 extended abstract (pdf). (The link to the poster (pdf) seems to be dead).

A test dataset with 14 GPS points to check out the functions of the toolbox is shipped with the distribution and can be found in the folder "/testdata/".

MacOSX users please read: MacOSX Users Short Installation Guide - at the end we also say what keyboard-combinations are important for context menu access.

Documentation

A revised version of the user manual for OpenJUMP HoRAE is available since Feb 2013 ( download pdf). It is also included in the software distribution. (Important: In the 2011 Version was an error in the core area calculation description; but not in the calculation.)

The manual describes only use of the software. For more background information on the methods we recommend to have a look at the ABODE manual since most functions implemented were "inspired" by the ABODE toolbox (but software calculations are differently).

We have also created a wiki page describing Problems with the HoRAE Toolbox. We had a start with a wiki for the manual, Home Range Calculation Manual, but it needs to be revised.

Other documents describing new methods implemented:

• Steiniger, S. (2009, unpublished): A vector-based algorithm to detect core, edge, patch and corridor areas and comparison with its raster-based complement. (pdf) --- The document explains the different types of classes such as core, edge, patch and corridor that can be derived with the function MOVEAN>HRE Analysis>Extract Core, Patch, Edge; and: MOVEAN>HRE Analysis>Extract Corridors.
• Steiniger, S., Hunter, A.J.S. (2013): A scaled line-based kernel density estimator for the retrieval of utilization distributions and home ranges from GPS movement tracks. Ecological Informatics 13:1-8 (Elsevier link, manuscript pdf) --- The document explains how the line-based Kernel Density Estimator (Line-KDE) works and compares it with Point-based KDE and the Brownian Bridge (BB) estimator. The function is to be found under MOVEAN>HRE>Line KD>....

If you want to cite/reference OpenJUMP HoRAE, then we ask you to use our 2012 article in the Wildlife Society Bulletin (see also above, WSB link).

Loading Your Data

Currently we have no special GPS file loaders and our input files for testing have been Shapefiles (though there is a GPS extension for OpenJUMP). If you have problems to get your data into OpenJUMP, then please write us (to Stefan). Also, if you need to create the "unique days" attribute and you don't know how - let us know, so we can write a guide. There is no guide on that yet, because we had only pre-processed data and the unique days were generated with a Matlab script and not with an office application.

related:

• Converting GPX files to Shapefiles

Support and Contact

If you have questions on the use of OpenJUMP HoRAE you can post to our user list at Google Groups:

• http://groups.google.com/d/forum/ojhorae-users .

Please note, that you need to login or subscribe first to send a message.

However, you can contact me, Stefan Steiniger, writing an email to ssteiniger(=at=)uc.cl. Alternatively you may contact Andrew Hunter.

Source Code & Future Development & Bug Reporting

The source code for the HoRAE toolbox (an extension to OpenJUMP) can be found on SourceForge.net. OpenJUMP has there, as Jump Pilot Project, a code repository. In that SVN repository the HoRAE code can be found under:

• http://jump-pilot.svn.sourceforge.net/viewvc/jump-pilot/plug-ins/horae/

How to access the SVN is described here.

There are several ideas for future developments (e.g. to evaluate co-occurrence for two individuals). However, currently we don't have funding for that. If you wish a specific function in the toolbox, then it should be possible to implement that when you can provide funding. However, bug fixing is free of cost and depends on (my) free time. If you find a bug, then let me (Stefan) know. As the source code is freely available - others can do feature development and bug fixing too. Please let us know if you want to contribute, need help, or even write how you use HoRAE.

HoRAE TODO

Comparison of Results with HRT

Thanks to Christine Richter (Göttingen, GER), we got aware of the fact that results calculated with HRT (by A.R. Rodgers and J.G. Kie) and OpenJUMP cannot be directly compared. If you use the same bandwidth parameter h for HRT and OpenJUMP HoRAE, then results (contours) will be much more detailed than those from HRT, with the latter being much smoother. However, if you take the bandwidth parameter from HRT (e.g. 63m) and multiply it with about 2.78 (e.g. 2.78 * 63 = 175m) and use this value (i.e. 175m) for HoRAE, then the contours should be about the same. The reason for the differences and the need to multiply the values is, that in ArcGIS/HRT such multiplication of h_ref is performed. This is done, as the general h_ref formula is an estimate for a Gaussian function/kernel, but will be applied to a bi-weight Kernel (in HRT). As you can see in HoRAE and on Wikipedia the kernel functions have different maximum values and properties. Internally HoRAE does not perform a scaling so one can directly compare density results. The multiplication with the value of 2.78 for the Gaussian-based h_ref value to be used with a bi-weight kernel will however, result in comparable density estimation with a Gaussian-based density estimation (see Silverman 1986).