IconLoCoH: A k-NNCH Implementation

Update, December 2013

The LoCoH webapp and R script on this website have been updated to correct a discrepancy between the code and the LoCoH algorithm as described in Getz et al (2007). The discrepancy concerns how ties are handled when sorting hulls in the adaptive and fixed-r methods. See below for details.

A newer version of LoCoH, called T-LoCoH, is available as a R package at http://tlocoh.r-forge.r-project.org. The new version does not yet have a webapp or ArcToolbox, but we recommend R users transition to the newer version which has more features and better performance.


LoCoH is a powerful family of algorithms for finding homeranges or utilization distributions. The LoCoH algorithms work by creating convex hulls around each point in the data set and then iteratively joining these hulls together from smallest to largest into isopleths. The 10% isopleth contains 10% of the points, the 100% isopleth 100% encompasses all the points. The smaller the hull, the more heavily used the region. Therefore, these isopleths can be used to determine how frequently a region is used.

There are current three LoCoH algorithms:

  • Fixed k LoCoH Fixed k LoCoH works by creating hulls from the (k-1) nearest neighbors to the root point. Also known as k-NNCH, Fixed k LoCoH was originally described in Getz and Wilmers (2004).
  • Fixed r LoCoH Fixed r LoCoH operates by generating local hulls from all the points within r distance from the root point. This method is very similar to Fixed Kernel Methods.
  • Adaptive LoCoH Adaptive LoCoH creates hulls from the maximum number of nearest neighbors to a point such that the sum of their distances from the root point is less than or equal to a distance measure a.

Compared to other methods for constructing homeranges (such as Minimum Convex Polygon and kernel methods), the LoCoH methods have several advantages that make them particularly well-suited for landscapes with 'sharp' features such as lakes, fences, or steep terrain. These type of landscape features often result in spatial distributions containing holes, sharp boundaries, corners, or corridors. In these cases, the LoCoH density isopleths have been shown to better approximate the true area represented by the data than kernel or alpha-hull methods. LoCoH isopleths also have the property of converging to the true area represented by the data as the number of data points increases, thus the method is particularly well suited when there is a lot of observational data (e.g., from a GPS collar).

Read the LoCoH Web Application TutorialForward

This web-based version of LoCoH allows you to explore the k-NNCH method with your own point data. The number of points that can be analyzed via this web-based version of LoCoH is limited due to memory limits on the server. However if you get an 'out of memory' error message you can download the R script to run it on your own PC. LoCoH is also available as toolbox for ArcGIS 9.x.

An in review journal article comparing and contrasting the three LoCoH algorithms in addition to kernel algorithms is Getz Et al, 2007.

Step 1: Enter Your Data

LoCoH currently supports tab delimited textfiles and ESRI's shapefile format. You can either upload your own data, or experiment with a demo dataset

Use the demo dataset. Upload my own tab delimited textfile. Upload my own point shapefile.

Select a tab deliminated text file. The first row will be used as labels.
What is this?
Select .shp file: (2MB limit).
What is this?
Locate matching .shx file: (2MB limit).
Locate matching .dbf file: (2MB limit).
This dataset contains 429 sightings of wolves in the Yellowstone. You can use it if you just want to run LoCoH through its paces.

Step 2: Set Options

Mode: Fixed k Fixed r Adaptive
What is this?
Minimum value of k:
What is this?
Value of k:
What is this?
When encountering duplicated points: Displace duplicates by units.
Include duplicates in nearest neighbor searches.
Delete them.
What is this?
Isopleth Levels:
What is this?

Step 3: Analyze

Please check over your settings and then click...

Hull sorting code correction details

When using the adaptive or fixed-radius method, hulls should be sorted by the number of enclosed points (largest to smallest), and then progressively unioned to form isopleths. When two hulls enclose the same number of points, the one with the smallest area (most dense) should be unioned first (Getz et al 2007). The webapp and R code on this website did not specify area as a second sort field as it should have, so ties were broken in an arbitrary manner. This was corrected in December 2013. The discrepancy does not apply to the k-method, nor does it exist in the LoCoH implementations in the adehabitatHR or T-LoCoH packages for R.

In most cases, the second sort field is not needed and/or does not make much difference in the isopleths. The difference will only be noticeable when an isopleth is capped at a tie-point. Random noise in the data, such as the location of a point within the accuracy of the recording device, are far more likely to occur and have an impact on the shape and size of isopleths. Any one wishing to check whether the hull sorting correction makes a significant difference in the isopleths is encouraged to re-run their data with the correction and compare the isopleth maps and area curves.

If you like this program, why don't you download the LoCoH R script that powers it! The script allows you to use all three LoCoH methods on your own computer without limits (and gives our servers a break!). The script contains much more flexiblity and power than this basic web interface gives you access to.

Learn More About LoCoH for RForward

We also have an ArcGis Toolbox script that will integrate the LoCoH R script right into ArcMap.

Learn About the ToolboxForward

Email scottfr@gmail.com | All contents copyright 2005 Wayne Getz lab. | Programmed by Scott Fortmann-Roe.