Re: Estimating Home Ranges Of Individual Animals In ArcGIS

[kateb84]

Colin,

 

I realize this is an older post, but I have searched all over, and it is the only one that really seems to have any resemblence to an actual answer for me.

I am newer to GIS and especially ArcGIS10.  We're trying to establish a home range using kernel densities.  We'd like to incorporate the 50 and 90 (read it's less biased than 95%) PCV but are aware this isn't something we can do in the current version.  Is there something others have used in the past to do this? We also are curious as to how to incorporate boundries such as a river (could we draw in our boundries and use that as our environmental extent?).  I've played around a lot with the kernel density tool, but can't figure out these few things.  I also tried your suggestion above for boundries, but it didn't work at all for me.  What I'm looking for is a step by step to get me through this.  School just doesn't prepare you for these things. I appreciate the help

 

Thanks,

Kate

On Thursday, September 20, 2012 4:12:48 AM UTC-7, GIS in Ecology wrote:

Estimating Home Ranges Of Individual Animals In ArcGIS

 

It is often useful to be able to identify and compare the areas used by individual animals in a population. These areas are known as home ranges. There are two widely accepted ways to estimate an animal’s home range.  These are through a minimum convex polygon (MCP) or through a kernel density estimate (KDE).  An MCP is the smallest polygon that can be drawn that will encompass all the locations where an individual is recorded that contains no internal angles greater than 180 degrees.  It, therefore, provides a measure of every area where an individual has been recorded, but it provides no information about whether some areas are used more frequently than others. A KDE measures the density of records within each grid cell that covers a study area, and uses this to estimate the probability that an individual will use neighbouring cells.  It, therefore, provides an estimate of which areas an individual uses most frequently.  Typically, KDE are converted to percentage volume contours (PVCs) that identify areas where an individual is likely to occur 50% of the time (often referred to as the core range) and 95% of the time (often taken as a measure of the total home range).

 

There are two options for estimating home ranges within ArcGIS.  The first is to use ArcGIS tools directly. However, the tools that can be used to estimate home ranges are not easy to identify because of the names used for them.  To create an MCP, the tool that you would use is called MINIMUM BOUNDING GEOMETRY.  To use this to create an MCP, go to DATA MANAGEMENT TOOLS> FEATURES> MINIMUM BOUNDING GEOMETRY.  In the MINIMUM BOUNDING GEOMETRY window that will open, select CONVEX HULL for GEOMETRY TYPE (OPTIONAL), and this will allow you to create and MCP.

 

To create a KDE, you can use the KERNEL DENSITY tool.  This can be found in SPATIAL ANALYST TOOLS> DENSITY> KERNEL DENSITY.  The main limitation of this tool is that it assumes that there are no barriers to an individuals movements. These means that it is not well suited to creating KDEs in areas where such barriers exist.  For example, it would not be appropriate to use this tool to estimate the home range of individual bottlenose dolphins in a coastal population (as they cannot move across land). Instead, in such instances, you can use two other tools. First you can create a point density raster data layer using the tool POINT DENSITY (SPATIAL ANALYST TOOLS> DENSITY> POINT DENSITY. You can then convert this data layer to a point data layer and use a tool called KERNEL INTERPOLATION WITH BARRIERS (GEOSTATISTICAL ANALYSIS TOOLS> INTERPOLATION> KERNEL INTERPOLATION WITH BARRIERS) to create your KDE.  

 

In all cases, when creating a KDE, you need to make sure that you select appropriate grid cell sizes, extents and search radius for your individual animals and species.  

 

Alternatively, if you don’t wish to use these tools in ArcGIS directly, you can use one of several specialist extensions to create your home range estimates. These include Hawth’s tools and Home Range Tools for ArcGIS.  Unfortunately, at the moment most of these are designed to work with ArcGIS 9.3 or earlier and not the more recent ArcGIS 10. For ArcGIS 10 or later users, the only option that seems to be currently available is Geospatial Modelling Environment (GME).  It has tools that allow you to create MCPs and KDEs (although it does not allow you to create a KDE with barriers).

[GIS in Ecology]

Hi Kate,

 

In the latest versions of ArcGIS (10.0 and 10.1), this is not as easy as it used to be in earlier versions (ArcGIS 9.3, and indeed ArcView 3.3). However, there are ways round it.

 

If you wish to create kernels with boundaries, you can use the KERNEL INTERPOLATION WITH BARRIER tool (as outlined above). However, you need to be very careful about exactly you create the data layer you use as the INPUT ABSLUTE BARRIER FEATURES (OPTIONAL). You can't use a line data layer for this, only a polygon. In partiucular, this polygon needs to be a bounding polygon, that is one that defines the exact limits as to which cells will have kernel density values calculated and which will not. If you were wanting to use a river as a barrier, this means you would need to create a polygon where the river formed one edge while the rest of the polygon defined the other edges of your study area (this assumes the river acts as a barrier across your whole study area - things get more complicated if you have a river that only acts as a barrier for part of your study area - but it's not impossible to deal with so let me know if this is the case.

 

For this specific tool, there is also an issue with exactly how you create your INPUT FEATURES data layer, Specifically, this cannot just be a data layer of all the locations where an individual was recorded. Instead, it needs to have the number of times the individual was recorded in each grid cell. This can be a a bit tricky to do, but you can do this by assigning each locational record for a given individual a value of one and then using the POINT TO RASTER tool (CONVERSION TOOLS> TO RASTER> POINT TO RASTER) to create a raster of the number of locational records in each grid cell. Once you have this you can convert this back into a point data layer (using the RASTER TO POINT tool) and use it as the input.

 

The issue of how to create 50% and 90% PVCs is also complicated but can be done in ArcGIS.

 

The way to do it is to extract the kernel estimate values for your kernel density estimate for all the locations where your inndividual was recorded (the ones used to make the kernel density estimate in the first place) using the EXTRACT VALUES TO POINTS tool (SPATIAL ANALYST TOOLS> EXTRACTION> EXTRACT VALUES TO POINTS). Next, open the attribute table of point data layer which is created and sort the attribute table descending based on the grid code (this is the extracted kernel density estimate). Now work out how many records is 50% of your original sample (i.e. if you had 100 records, this would the 50th record) and scroll down to you reach this poin and read off the kernel denisty value. You then reclassify your kernel density estimate raster (using the RECLASSIFY tool - SPATIAL ANALYST TOOLS> RECLASS> RECLASSIFY) based on this value with values above this having a value of 50 (because it will form the 50% PVC), and values below this being classified as NO DATA. This reclassified raster is them converted into a polygon data layer (using the RASTER TO POLYGON tool in the CONVERSION toolbox) to give you your 50% PVC. This is the repeated for the records with the top 90% highest kernel density estimates to create the 90% PVC.

 

This is not a simple process and may take a few goes to get it right. In addition, it is not an exact measure of the PVCs, but rather is an approximation of them. However, in most cases they should be very similar to the actual PCVs calculated with tools such as the old HAWTHS TOOLS for ArcGIS 9.3.

 

Anyway, I hope this helps, and if anything isn't clear, or if you run into any problems, just post again on this thread.

 

All the best,

 

Colin

 

PS We should have a supplementary workbook on calculating home ranges in ArcGIS 10.1 available by the end of the year (which might be a bit late for you unfortunately, but I thought I'd mention it).

[Brenner, Kathryn]

Colin,

Like I said I'm new to this.  I'm trying to work through this piece by piece.  So far I created the point-raster then raster-point and used this as the input.  I used an output cell size of 10, and the barrier as a boundry polygon I had drew and saved as a shape file.  I tried each kernel function (the one pictured is the polynomial 5) but don't really understand the difference between them.  The bandwidth I made 640 and left everything else blank except used prediction for the surface type.  Every time I've tried it (adjusting numbers and functions along the way) it comes out looking like what I have pictured below.  The other screen shot is of the boundry polygon I created and the raster-point points.    Any help is appreciated!

Thanks,

Kate

--
-- ======================================================================
You received this message because you are subscribed to the "GIS In Ecology Forum" discussion group (http://www.gisinecology.com/GIS_in_Ecology_forum.htm).
 
To control how often you get emails from this group, go to http://groups.google.com/group/gis-in-ecology-forum/subscribe (you will need to log on to get to this page).
 
To post to this group, either log onto the group's home page or send an email to
gis-in-eco...@googlegroups.com.
 
The rules for posting to this group can be found here: http://groups.google.com/group/gis-in-ecology-forum/browse_thread/thread/df31a0822742203f#.
 
To unsubscribe from this group, email:
gis-in-ecology-f...@googlegroups.com
 
All information on this forum is provided on an 'as is' basis. GIS In Ecology is not responsible for checking the accuracy or suitability of any posting or response.
======================================================================
---
You received this message because you are subscribed to the Google Groups "GIS In Ecology Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gis-in-ecology-f...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[GIS in Ecology]

Hi Kate,

 

Sorry for the slow reply (my computer died on me at the start of the week and I'm only fully up and running again today). The screenshots you sent through do look a little odd. I think the best thing would be for you to email me a copy of your boundary polygon and your point data layer to cdmacleod (@) GISinEcology.com. I can then play around with it and give you some more specific instructions. Would this work for you?

 

All the best,

 

Colin

[rachael...@hotmail.co.uk]

Hello,

I've got a question regarding using the KDE function in ArcGIS 10 for estimating animal home ranges and the repeatably of this in other programs.  I have used the Kernel density function in ArcGIS10, and then the isopleth tool in GME to give me 95 % PVC's.  I have figured out the best cell size and search radius to use in the Kernel Density tool through trail and error, but if wanted to repeat this analysis in another program, how do these values relate to bandwidth, resolution and buffer values that are used in other programs? And what type of kernel is being used in ArcGIS (as I'll need to report this in my methods)? 

 I'm trying to investigate the asymptote of my home ranges with increasing numbers of fixes, so I can determine a minimum number of fixes required to estimate home range for an individual of my species. To do this in ArcGIS would take an age as I'm not great at writing code, and to do each KDE and then PVC by hand for each individual multiple times with an increase of 3 fixes each step, from 10 fixes through to hundreds for some individuals, is a ridiculous task that would probably take months to achieve! 

I've done this with MCPs using the asymptote function in program ABODE, and this suggests that as few as 15 fixes adequately describe the MCPs, but I need to do the same for KDEs and PVCs.   To use an asymptote function in a different program, (either ABODE, in ArcGIS, or in the rhr package for R) I need to specify bandwidth, and buffer values/types and a resolution.  Since I know what works in ArcGIS to give sensible home range estimates when I use all of my fixes, I'd like to just do the equivalent analysis in those other programs.  Do you know how I can convert a cell size of 10, and a 370m square search radius, into the equivalent bandwidth type or value, resolution and buffer?  Are they the very same things with different names, or do the different programs calculate KDEs and PVCs in such a different way that they are incomparable? I did try using the same values with buffer as my previous search radiun (370m) and bandwidth as my previous cell size (10m) but came out with very different 95% areas from my ArcGIS produced PVCs.

Any help you can give with this so that I don't have to start the whole trail and error procedure again in R and rhr, would be very greatly appreciated.

Kind regards,

Rachael.

[GIS in Ecology]

Hi Rachael,

 

>>I've got a question regarding using the KDE function in ArcGIS 10 for estimating animal home ranges and the repeatably of this in other programs.  I have used the Kernel density function in ArcGIS10, and then the isopleth tool in GME to give me 95 % PVC's.  I have figured out the best cell size and search radius to use in the Kernel Density tool through trail and error, but if wanted to repeat this analysis in another program, how do these values relate to bandwidth, resolution and buffer values that are used in other programs?

 

The resolution will be the same as cell size, and the search radius should be equivilant to the bandwidth used by other software. However, I'm not too sure what the equivilant would be for a buffer value. If you can let me know what the other software is, I can do some digging and find out for you.

 

 

>>And what type of kernel is being used in ArcGIS (as I'll need to report this in my methods)? 

 

I'm presuming you are using the KERNEL DENSITY tool here and not the KERNEL WITH BARRIER tool. If so, according to the ArcGIS help file for this tool: 'The kernel function is based on the quadratic kernel function described in Silverman (1986, p. 76, equation 4.5).'  You can find ther original source for this here: http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z00000011000000.htm.

 

 I'm trying to investigate the asymptote of my home ranges with increasing numbers of fixes, so I can determine a minimum number of fixes required to estimate home range for an individual of my species. To do this in ArcGIS would take an age as I'm not great at writing code, and to do each KDE and then PVC by hand for each individual multiple times with an increase of 3 fixes each step, from 10 fixes through to hundreds for some individuals, is a ridiculous task that would probably take months to achieve! 

>>I've done this with MCPs using the asymptote function in program ABODE, and this suggests that as few as 15 fixes adequately describe the MCPs, but I need to do the same for KDEs and PVCs.    

To use an asymptote function in a different program, (either ABODE, in ArcGIS, or in the rhr package for R) I need to specify bandwidth, and buffer values/types and a resolution.  Since I know what works in ArcGIS to give sensible home range estimates when I use all of my fixes, I'd like to just do the equivalent analysis in those other programs.  Do you know how I can convert a cell size of 10, and a 370m square search radius, into the equivalent bandwidth type or value, resolution and buffer?  Are they the very same things with different names, or do the different programs calculate KDEs and PVCs in such a different way that they are incomparable?

 

The bandwith and the search radius should be the same thing, and the resolution and the cell size should be the same thing. This is certainly the case in ABODE according to it's user manual. However, I'm not quite too sure what the buffer value is as I couldn't find a reference to it in the ABODE user manual. Is buffer a term that is specific to a certain software package? If so, let me know which and I can look into it. .

 

I did try using the same values with buffer as my previous search radiun (370m) and bandwidth as my previous cell size (10m) but came out with very different 95% areas from my ArcGIS produced PVCs.

 

Try it with the your previous search radius as your bandwidth and your cell size as your resolution and see if that gives you a similar 95% area.

 

I hope this helps and if it doesn't, just let me know and I can look into it further.

 

All the best,

 

Colin

[GIS in Ecology]

Hi Rachael,

 

I've been doing some digging into rhr.

I can now confirm that the search radius in ArcGIS is equivilent to the bandwidth in rhr. Also, the cell size is the same as the resolution. It turns out that the buffer option in rhr is somewha equivilant to the extent in ArcGIS, in that it adds a buffer around the extent of all your indivdual locations to allow the KDE to cover an area larger than the area where all the locations were recorded.

 

However, it seems that rhr and ADOBE/ArcGIS calculate the KDE in different ways. rhr uses a binned bivariate kernel density estimation based on a Gaussian distribution while ADBE/ArcGIS uses a biweight kernel for kernel density estimation based on a quadratic distribution.

 

This means that you will not get an identical KDE using the two different tool sets. At this stage, I'm not too sure how different these will be or what this means for transfering optimal search radii identified from asymptote from one to the other. This means you might need to do some further data exploration before you can know whether you can do this or not. Alternatively, you could choose to do all your analysis including the asymptote estimation) within a single software package as this would enure that everything is consistent throughout your project.

 

This is probably not the answer you were hoping for, but does seem to be the current situation with these tools. 

 

If there is anyone else out there with experience of looking at how the KDEs generated by ADOBE/ArcGIS and rhr, feel free to jump in on this discussion.

 

I hope this helps.

 

All the best,

 

Colin

 

 

[Sarah Woodward]

Dear Colin,

I'd like to start off by saying that ArcGIS is very new to me.  I am working my way through some bird telemetry data and would like your assistance with the Kernel Density Estimates.  I am using ArcMap 10.1. 

I feel like I should start at the beginning.  I tried unsuccessfully to plot my lat/long data in ArcMap using the Add XY data tool.  I'm not sure what the trick is, but I haven't gotten it to work.  I thought maybe I had found a way around this, when I remembered using the Minnesota DNR program a few years ago for a different project.  All of our telemetry data has been entered already into Google Earth (easier and simpler for techs to access data from their homes)  I downloaded the new version of DNR GPS and was happy to discover that I could load points from Google Earth into it and from there create a shapefile that could be easily displayed in ArcMap. 

So next I attempted to run the Kernel Density Estimate with no barriers as you describe above.  I ran the tool by choosing my telemetry shapefile from the drop down box for the input.  I allowed the rest of the boxes to auto-fill and pressed okay.  The KDE runs successfully and appears in my table of contents but is not visible on the map.  I have tried adjusting the output cell size and search radius with no change visible.  I read the previous comments about adjusting bandwidth as well but have been unable to find where this particular number is adjusted.

I should note I have also attempted to download the GME program to use instead of doing the calculations in ArcMap but have had no success getting it to actually run on my computer.

I would really appreciate any assistance you could provide on this issue and I hope I have posted in the correct place.

Thanks,
Sarah

[GIS in Ecology]

Hi Sarah,

 

>>I'd like to start off by saying that ArcGIS is very new to me.  I am working my way through some bird telemetry data and would like your assistance with the Kernel Density Estimates.  I am using ArcMap 10.1. 

>>I feel like I should start at the beginning.  I tried unsuccessfully to plot my lat/long data in ArcMap using the Add XY data tool.  I'm not sure what the trick is, but I haven't gotten it to work. 

 

This sounds like a problem with how you are handling the projection/coordinate system issue, and this is something which you need to deal with if you are going to be doing anything in ArcGIS. Basically, if you are adding data based on their latitude an longitude, these need to be in decimal degrees and you need to set the projection/coordinte system to the geographic projection based on the appropriate datum presumably WGS 1984 - but you need to double-check this). If you don't do this they won't plot in the right place.

 

>>I thought maybe I had found a way around this, when I remembered using the Minnesota DNR program a few years ago for a different project.  All of our telemetry data has been entered already into Google Earth (easier and simpler for techs to access data from their homes)  I downloaded the new version of DNR GPS and was happy to discover that I could load points from Google Earth into it and from there create a shapefile that could be easily displayed in ArcMap. 

While this might work for simply displaying information in a GIS, you may still have projection/coordinate system issues which you need to deal with before you.use the data layers with any of the tools in ArcGIS.

>>So next I attempted to run the Kernel Density Estimate with no barriers as you describe above.  I ran the tool by choosing my telemetry shapefile from the drop down box for the input.  I allowed the rest of the boxes to auto-fill and pressed okay.  The KDE runs successfully and appears in my table of contents but is not visible on the map. 

 

Again, this is almost certainly down to issues with the projection/coordinate system being used. Specifically, you need to be using the same projection/coordinate system for your data layer AND your data frame. If you don't (which I think will be the case here), a lot of the tools in ArcGIS will not work properly. This means that once you've successfully added your XY data to your GIS project (by whatever means), you will need to transform your data layer into the projection/coordinate system which you are using for your GIS project using the PROJECT tool (DATA MANAGEMENT TOOLS> PROJECTIONS AND TRANSFORMATIONS> FEATURES> PROJECT) before you can run the KDE tool. This projection/coordinate system should be a projected one which uses real world units (i.e. metres) as its unit of measurement rather than decimal degrees (which is the case for geographic projections). 

 

>>I have tried adjusting the output cell size and search radius with no change visible.  I read the previous comments about adjusting bandwidth as well but have been unable to find where this particular number is adjusted.

The bandwith and the search radius are the same thing, and you might sometimes also see this referred to as h. Selecting the most appropriate bandwith value and cell size value for your specific data set is rather complicated, and is something which is very hard to advise on. This is because it will be different for different data sets, and it can take a lot of work to get these right. There are a number of additional free software packages which can help you do this, including GME. However, GME has a number of limitations, in particular not being able to set the extent for your KDE, meaning it can be difficult to produce KDEs with identical extents if you wish to directly compare the KDEs of different individuals. In addition, you will still need to have correctly dealt with all projection/coordinate system issues before you can successfully run the KDE tools in GME.

>>I should note I have also attempted to download the GME program to use instead of doing the calculations in ArcMap but have had no success getting it to actually run on my computer.

GME can be difficult to get successfully installed on your computer. My advice would be to make sure that you definitely have the right version downloaded (this includes making sure you have the one for the service pack for your version of ArcGIS), and then try re-installing it. If this doesn't work, check the following (taken from the GME website): ArcGIS 10: it is very important to ensure you have the Microsoft .NET Framework 3.5 installed BEFORE you install ArcGIS 10, otherwise the ESRI .NET libraries will not be installed!

 

In addition, you need to install R before you install GME. More information on trouble-shooting the installation of GME can be found at http://www.spatialecology.com/gme/gmedownload.htm.

 

If this doesn't sort out the issues with installing GME, I would suggest you contact Spatial Ecology directly to see if they can help.

 

>>I would really appreciate any assistance you could provide on this issue and I hope I have posted in the correct place.

This is definitely the right place, and I hope the above helps. However, correctly creating KDEs is not something which is easy to do, especially if you don't already have much experience with ArcGIS. In addition, even when you can get ArcGIS to generate a KDE, this doesn't mean that it will necessarily be biologically correct. This will only be the case if you've managed to select an approrpiate search radius/bandwith/h and cell size, and these can be tricky to work out. This means that you need to read a lot of the background literature on creating KDEs and on the ecology of your species before you can start working with KDEs. This is not intended to try to put you off making them, but rather to try to make sure that you don't waste time doing them incorrectly the first time round.

 

In addition, since you are new to ArcGIS, I would suggest getting to know it, and especially how it deals with projections/coorinate systems, before you start trying to use the tools in it. While there are a number of places you can find information about this (including other postings on this forum), you can also find out more about this in my book 'An Introduction To Using GIS In Marine Biology'. While this is about using GIS in marine biology, it also provides information about projections and coordinate systems which is useful for all ecologists. This includes information on how to transform data layers between projection/coordinate systems, and also how to use the ADD XY DATA tool.

 

Anyway, I hope this helps. If you have any further questions about creating KDEs, feel free to post them on this thread. If you have questions specifically about projection/coordinate systems, these should be posted on a new thread.

 

All the best,

 

Colin

 

[Sarah Woodward]

Thank you for your help!

I believe I have made some progress.  I was able to successfully transform my shapefile from a geographic coordinate system to the projected WGS 84 coordinate system.  I also ran the KDE again and this time was at least able to view it on my map.  I think that it is still not "biologically correct" but I feel encouraged that I have made it this far.

I was just curious if the book you mentioned on your website, An Introduction to Using GIS In Marine Biology: Supplementary Workbook Four- Investigating Home Ranges of Individual Animals will be available soon?

Do you have any other suggestions for background reading to learn more about creating KDEs? Additionally, do you still recommend using GME and the Isopleth tool to convert KDE to percentage volume contours?

Sarah

--

-- ======================================================================
You received this message because you are subscribed to the "GIS In Ecology Forum" discussion group (http://www.gisinecology.com/GIS_in_Ecology_forum.htm).
 
To control how often you get emails from this group, go to http://groups.google.com/group/gis-in-ecology-forum/subscribe (you will need to log on to get to this page).
 
To post to this group, either log onto the group's home page or send an email to
gis-in-eco...@googlegroups.com.
 
The rules for posting to this group can be found here: http://groups.google.com/group/gis-in-ecology-forum/browse_thread/thread/df31a0822742203f#.
 
To unsubscribe from this group, email:
gis-in-ecology-f...@googlegroups.com
 
All information on this forum is provided on an 'as is' basis. GIS In Ecology is not responsible for checking the accuracy or suitability of any posting or response.
======================================================================
---

You received this message because you are subscribed to a topic in the Google Groups "GIS In Ecology Forum" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gis-in-ecology-forum/k13R3rNpXts/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gis-in-ecology-f...@googlegroups.com.

[GIS in Ecology]

Hi Sarah,

 

>> I believe I have made some progress.  I was able to successfully transform my shapefile from a geographic coordinate system to the projected WGS 84 coordinate system.  I also ran the KDE again and this time was at least able to view it on my map. 

 

Glad to hear that you're making progress, and that sorting out the projection/coordinate system seemed to help you move forward. If you have problems in GIS, this is always the first thing you should check as projection/coordinate system issues are at the route of most GIS problems.

 

>> I think that it is still not "biologically correct" but I feel encouraged that I have made it this far.

 

Often you need to focus on getting the tools working in GIS, and then sorting out the issues around making something biologically correct. As I tell people on the courses I run, assume you'll have to do everything in GIS at least 3 times before you get it right (once to work out where you're going wrong, once to work out how to do it properly from a technical point of view, and once to make sure it's biologically correct - anything less than this should be viewed as a bonus!)

 

>> I was just curious if the book you mentioned on your website, An Introduction to Using GIS In Marine Biology: Supplementary Workbook Four- Investigating Home Ranges of Individual Animals will be available soon?

 

I'm just about to send the proof off to the printer on this book, and I'm hoping it will be out by the end of January. However, this only covers the technical issues around creating home ranges in a GIS project. It doesn't cover the 'biological' aspects, such as how to select the best bandwidth, cell size etc as these are too complicated to cover in a simple GIS workbook.

 

>> Do you have any other suggestions for background reading to learn more about creating KDEs? 

 

The use of KDEs (as opposed to minimum convex polygons) for home ranges is still relatively new and it seems that no standard method has yet been established (e.g. different software will calculate the KDEs in different ways meaning the exact home range you'll get will depend on the software you use!). This means that the best way to learn more about creating KDEs is to read the academic papers which have used them. Unfortunately, these are often rather light on their coverage of the exact processes used to create them so you might have to dig around to find the ones that are really useful. I know this isn't a great answer, but currently it's the only one available. However, I'm hoping that this thread might eventually grow into a 'methods' paper covering all these issues as this would make life easier for everyone wanting to use KDEs to examine home ranges.

 

>> Additionally, do you still recommend using GME and the Isopleth tool to convert KDE to percentage volume contours?

 

I'm not too sure I would say I'd recommend using GME for KDEs. It's a great bit of software, but there are a few limitations when it comes to KDEs. The main one is that there doesn't seem t o be a way of setting the extent of the final KDE layer, meaning it's difficult to get the layers for different individuals to match up exactly. This isn't essential for all home range calculations, but it is needed in some cases. So, if you need all your home ranges to match up exactly in terms of their extents, you'd be better off just using ArcGIS. If not, go ahead and use GME to create your KDEs. 

 

In terms of using the isopleth tool to convert the KDE to percentage volume contours, as you will see from other posts on this thread, there is some concern that PVCs generated are covering too large or too small an area in some cases. I've yet to work out what's causing this, and this is something you need to watch out for. Personally, I would tend to work these out manually and directly in ArcGIS (the method should be in one of the earlier posts). This manual approach is also the one I've used in the supplementary workbook 4. However, if the isopleth tool in GME works for your data, and gives sensible resuls, then there's no reason not to use it as it makes life much simpler.

 

Anyway,  I hope this helps.

 

All the best,

 

Colin

 

[Ben Hewlett]

Hi there!!!

Im a 3rd biology student from oxford brookes currently trying to produce some kernel densities for some rhino data, and it seems like this is the place where people share there pains of kernel!!!!!

I have successfully plotted my XY data, exported them to .shp files etc. But when i run the actually kernel analysis I am always running into the same issue. 'Unable to create raster' and 'error in executing grid expressions'

Any ideas or solutions? Been at it for weeks and its driving me mad!!!!! 

It has worked once but randomly deleted

Thanks

Ben 

[GIS in Ecology]

Hi Ben,

 

Sorry to hear about the problems which you are having.

 

>>I have successfully plotted my XY data, exported them to .shp files etc. But when i run the actually kernel analysis I am always running into the same issue. 'Unable to create raster' and 'error in executing grid expressions'

>>Any ideas or solutions? Been at it for weeks and its driving me mad!!!!! 

 

This is most likely down to one of three possibilities. The first is that you don't have enough locational data, but I'm going to presume this is not the case.

 

The second is that there is a conflict between the projection/coordinate system of your data frame and point data layer/shapefile.

 

To check this:

 

1. Right click on the name of your data frame in the TABLE OF CONTENTS window (if you are using the default options this is most likely LAYERS) and select PROPERTIES. Click on the COORDINATE SYSTEM tab in the DATA FRAME PROPERTIES window and check what projection/coordinate system is being used. Write this down and then close the DATA FRAME PROPERTIES window. NOTE: This will need to be a projected coordinate system and not a geographic coordinate system. If it is a geographic coordinate system you will need to change it to a projected coordinate system suitable for your study area.

 

2. Next, in the TABLE OF CONTENTS window, right click on the name of your point data layer and select PROPERTIES. In the LAYER PROPERTIES window, click on the SOURCE tab and scroll down until you can see the projection/coordinate system of the data layer. For the kernel density estimate tool in ArcGIS to work, this needs to be the same as the projection/coordinate system of the data frame. If there is no projection/coordinate system listed, then you will need to assign one to your data layer using the DEFINE PROJECTION tool (DATA MANAGEMENT TOOLS> PROJECTIONS AND TRANSFORMATIONS> DEFINE PROJECTION). Once you have done this, repeat step 2.

 

3.If the projection/coordinate system of your data layer does not match that of your data frame, use the PROJECT tool (DATA MANAGEMENT TOOLS> PROJECTIONS AND TRANSFORMATIONS> FEATURES> PROJECT - unless you are using ArcGIS 10.2.1 in which case there is no FEATURES folder) to transform your data layer into the same projection/coordinate system of your data frame.

 

4. Try running the kernel density estimate tool again.

 

 

If it turns out that sorting out any projection/coordinate system does not solve this problem, the final possible solution is to look at the folder name and the file name which you are using. The file name needs to be less than 13 characters in length, and the folder name needs to be relatively simple (e.g. C:/GIS_PROJECT) and the folder needs to exist before you try to create your KDE. If changing this file name/folder doesn't work, try just letting ArcGIS save it in its default location and with its default file name to see if that works.

 

You might also want to check the individual settings you are using in the KERNEL DENSITY tool window. Make sure your POPULATION FIELD is a numeric field and not a text one. Then make sure that you sent a sensible cell size (in the units of your chosen projected coordinate system), and that to start off with you use the default SEARCH RADIUS (once you have the process working, you can think about changing this).

 

If these simple solutions don't work (they should in about 90% of cases), let me know and I will see if I can suggest something else which might solve your problems.

 

I hope this helps, and I'm sure we can get to the bottom of your problems relatively quickly.

 

All the best,

 

Colin

[Ben Hewlett]

Hi again

Sorry its taken me so long to get back it didnt give me a notification for some reason. 

Projected coordination system matches, all folders 13 characters or less, 

It wont let me have a population field as numerical, i can either have NONE, X or Y?

Thank you so much for the guidance!

Thanks, Ben 

[GIS in Ecology]

Hi Ben,

 

The easiest way to sort this out might be for you to email me a copy of your XY data for one individual (if that is possible - I'm presuming this is in a spreadsheet of some kind) along with information about what coordinate system the X and Y coordinates are in and the projection/coordinate system you're trying to create your kernels in. This will allow me to play around with it and see what the problem might be. You can email me here: cdmacleod[at]GISInEcology.com.

 

I hope this helps.

 

All the best,

 

Colin 

 

[Ben Hewlett]

Hi there Colin

Im afraid im not allowed to as when collected this data i signed a contract not to distribute it (protected species in high poaching area) which i have to honour. However since my last email some progress has been made. I am using GIS on my universities pooled network and it appears that the pooled version on our uni network it is not complete and full of glitches. I talked to the tech people here and this has apparently happened before. Which is extremely annoying but at least the problem is identified. Also when talking to the tech guys I ran my kernel on their system and it worked instantly!!! So I am going to use it on my friends laptop and see if it works there instead, which im thinking it will. Oh the troubles of GIS!

Hope this problem can be identified for others to avoid too much wasted time. 

Thanks for the support, Ben 

  

--

-- ======================================================================
You received this message because you are subscribed to the "GIS In Ecology Forum" discussion group (http://www.gisinecology.com/GIS_in_Ecology_forum.htm).
 
To control how often you get emails from this group, go to http://groups.google.com/group/gis-in-ecology-forum/subscribe (you will need to log on to get to this page).
 
To post to this group, either log onto the group's home page or send an email to
gis-in-eco...@googlegroups.com.
 
The rules for posting to this group can be found here: http://groups.google.com/group/gis-in-ecology-forum/browse_thread/thread/df31a0822742203f#.
 
To unsubscribe from this group, email:
gis-in-ecology-f...@googlegroups.com
 
All information on this forum is provided on an 'as is' basis. GIS In Ecology is not responsible for checking the accuracy or suitability of any posting or response.
======================================================================
---

You received this message because you are subscribed to the Google Groups "GIS In Ecology Forum" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gis-in-ecology-f...@googlegroups.com.

[GIS in Ecology]

Hi Ben,

 

No worries about not being able to share the data. However, as a general rule, it is always worth trying anything you just can't get to work in GIS on a different system. This is because it allows you to test whether there is a problem with your data, or if there is a problem with the way a specific computer is set up. 

 

In your case, it may well be that the problem is due to working on a network computer rather than on a personal computer. This used to cause major problems when using earlier versions of ArcGIS, especially when trying to do anything with raster data layers (and it was an absolute nightmare when trying to teach a class of 30 students and half of the computers would work for certain tools!), but I wasn't aware that it was still an issue with the more recent versions, although this could be because I no longer teach on networked computers rather than ESRI having solved all the issues.

 

The main problem here is that some of the tools in ArcGIS require access to specific file locations on computers, such as the temp folder on the C-drive, but on many networked computers, the C-drive is blocked from access by the user. This means that you can have major problems working on network computers, and it will never necessarily be clear that this is what is causing the issue (ArcGIS has always had a problem with no flagging up errors associated with file names and addresses which aren't compatible with the requirements of specific tools).

 

One way which might work to get round this is to try using the default file name and address which the tool suggests when you open it and see if that works (this might be why it worked when talking to the tech guy and why you had it working once before, because you didn't change this default name and address on those occasions but you have been when you were trying it for yourself). If you can make your kernels that way, you can then right click on the name of the data layer in the TABLE OF CONTENTS window and select EXPORT to save it to a different location.

 

Anyway, please report back if you find that it does work on your friend's laptop as this will be important for other people to know as well if they run into similar problems (and if you have, others will too). In particular, I wasn't aware that there was this potential issue of trying use the home range analyses tools on networked computers so this would be good to know (so no time wasted).

 

Good luck with your project, and hopefully this will solve this particular issue. If you come up with any more, or if this doesn't solve your problem, feel free to post on this thread again.

 

All the best,

 

Colin

[skyec...@boisestate.edu]

Instructions for Kernel Home Range using Arc10.1, GME, and R are posted at GIS4Geomorphology.com. The lessons cover the basics.

Kernel Home Ranges
http://gis4geomorphology.com/home-range-kernel/

MCP
http://gis4geomorphology.com/home-range-mpc/

Population Density
http://gis4geomorphology.com/home-ranges-density/

[GIS in Ecology]

When I originally created this thread, it was to provide some basic tips about how to analyse home ranges using the tools in ArcGIS. However, it has since proved to be one of the most popular threads on the forum, suggesting this is a subject which a lot of people out there are struggling with. As a result, I have created a 'Supplementary Workbook' in my 'An Introduction To Using GIS In Marine Biology' series specifically devoted to analysing home ranges in a GIS-based environment, and using ArcGIS 10.2 software to illustrate the processes.

 

This 'Supplementary Workbook' builds on the information which can be found for free in this thread, and contains five exercises based around investigating home ranges in individual animals. These include: 1. Creating a minimum convex polygon (MCP); 2. Analysing home ranges in habitats without barriers using kernel density estimates; 3. Analysing home ranges in habitats with barriers using kernel density estimates; 4. calculating the overlap and habitat preferences of the home ranges of individual animals; 5. Working out how the home ranges of individual animals in a population overlap spatially. These exercises provide fully worked examples of each process starting with the raw data on an individual's distribution, and so will be of most use to those struggling to apply the free  basic outline instructions provided on this thread to their own projects.

 

One subject which it does not cover is how to select the most appropriate values for H/bandwidth, cell sizes and spatial extents for kernel density estimates (KDEs). This is because this is still a subject of some discussion amongst those who use KDEs, and it requires some detailed reading of the available literature before deciding on what values are best for any individual project.

 

While it is primarily aimed at marine biologists, exactly the same processes are used in terrestrial environments, and when using kernel density estimates to examine the distribution of species, populations etc.

 

The workbook is called: An Introduction To Using GIS In Marine Biology Supplementary Workbook Four - Analysing Home Ranges Of Individual Animals - RRP 19.99. ISBN: 978-0-9568974-5-9.

It can be purchased from Amazon (http://www.amazon.com/Introduction-Using-Marine-Biology-Supplementary/dp/0956897452/) or from GIS In Ecology (http://www.gisinecology.com/Book_Shop.htm).

More information about this book (including extracts) can be found here: http://www.gisinecology.com/marine_supplementary_book4.htm

All the best,

 

Colin 

[John Vanek]

Wow, this is great thank you!

[Jake Wall]

Hi Colin,

You might be interested in an extension I released recently for ArcGIS 10.1 that currently calculates MCP, KDE, a-LoCoH, LTD and BBMM home-range methods among other things. It's called ArcMET - Movement Ecology Tools for ArcGIS and is freely available at www.movementecology.net. 

all the best,

Jake

[GIS in Ecology]

Hi Jake,

 

Thanks for posting about the extension you've released. It looks like it will be very useful. Does it work with ArcGIS 10.2 as well or only 10.1?

 

All the best,

 

Colin

[Jake Wall]

Hi Colin,

I hope to have a version for 10.2 posted soon (~ 1 month) but at present ArcMET only works with ArcMap 10.1. 

all the best,

Jake

--

-- ======================================================================
You received this message because you are subscribed to the "GIS In Ecology Forum" discussion group (http://www.gisinecology.com/GIS_in_Ecology_forum.htm).
 
To control how often you get emails from this group, go to http://groups.google.com/group/gis-in-ecology-forum/subscribe (you will need to log on to get to this page).
 
To post to this group, either log onto the group's home page or send an email to
gis-in-eco...@googlegroups.com.
 
The rules for posting to this group can be found here: http://groups.google.com/group/gis-in-ecology-forum/browse_thread/thread/df31a0822742203f#.
 
To unsubscribe from this group, email:
gis-in-ecology-f...@googlegroups.com
 
All information on this forum is provided on an 'as is' basis. GIS In Ecology is not responsible for checking the accuracy or suitability of any posting or response.
======================================================================
---

You received this message because you are subscribed to a topic in the Google Groups "GIS In Ecology Forum" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gis-in-ecology-forum/k13R3rNpXts/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gis-in-ecology-f...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
PhD Candidate
University of British Columbia
CAN: +16047231224
KEN: +254700504553
SAT: +881632556840
www.savetheelephants.org

[Jefferson Ferreira Ferreira]

Great Jake! 

Congratulations for the awesome work. I'll wait for the compliant version for 10.2.1 too. Actually I'm using R (adehabitatHR) for this task and would like to compare the results.

Best regards

-- 

Jefferson Ferreira-Ferreira

Geógrafo – GEOPROCESSAMENTO IDSM | Coordenadoria de TI

Jefferson...@mamiraua.org.br

Instituto de Desenvolvimento Sustentável Mamirauá

Ministério da Ciência, Tecnologia e Inovação

Telefone: +55 97 3343-9710

Google Maps - Mapas deste e-mail:

Exibir mapa ampliado

Contatos particulares:

(55) 9615-0100

You received this message because you are subscribed to the Google Groups "GIS In Ecology Forum" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gis-in-ecology-f...@googlegroups.com.

[David Robinson]

Hi Colin,

I am following your advice and think I am almost there with regards to creating a 50 and 95% home range using Kernel Interpolation with barrier. I just have a few issues with my results and I think it is due to assigning each location a value.

Can you explain how to do this please? Do we assign the value to the location shape file before converting to a raster and if so how do we do that? I know how to add the value after converting to a raster but not before? Can you also explain (in detail please) how in the point to raster tool you can 'create a raster of the number of locational records in each grid cell' please. The first time I did this, it looked like it worked but as there were less points when converted to a raster, but each point still had a value of 1 in the table even though it incorporated several points?

How do we access the information about grid cells?

Many thanks for your help, this thread and advice have been a great help!

Regards

David



On Wednesday, July 31, 2013 7:46:10 PM UTC+4, GIS in Ecology wrote:

Hi Kate,

 

In the latest versions of ArcGIS (10.0 and 10.1), this is not as easy as it used to be in earlier versions (ArcGIS 9.3, and indeed ArcView 3.3). However, there are ways round it.

 

If you wish to create kernels with boundaries, you can use the KERNEL INTERPOLATION WITH BARRIER tool (as outlined above). However, you need to be very careful about exactly you create the data layer you use as the INPUT ABSLUTE BARRIER FEATURES (OPTIONAL). You can't use a line data layer for this, only a polygon. In partiucular, this polygon needs to be a bounding polygon, that is one that defines the exact limits as to which cells will have kernel density values calculated and which will not. If you were wanting to use a river as a barrier, this means you would need to create a polygon where the river formed one edge while the rest of the polygon defined the other edges of your study area (this assumes the river acts as a barrier across your whole study area - things get more complicated if you have a river that only acts as a barrier for part of your study area - but it's not impossible to deal with so let me know if this is the case.

 

For this specific tool, there is also an issue with exactly how you create your INPUT FEATURES data layer, Specifically, this cannot just be a data layer of all the locations where an individual was recorded. Instead, it needs to have the number of times the individual was recorded in each grid cell. This can be a a bit tricky to do, but you can do this by assigning each locational record for a given individual a value of one and then using the POINT TO RASTER tool (CONVERSION TOOLS> TO RASTER> POINT TO RASTER) to create a raster of the number of locational records in each grid cell. Once you have this you can convert this back into a point data layer (using the RASTER TO POINT tool) and use it as the input.

 

The issue of how to create 50% and 90% PVCs is also complicated but can be done in ArcGIS.

 

The way to do it is to extract the kernel estimate values for your kernel density estimate for all the locations where your inndividual was recorded (the ones used to make the kernel density estimate in the first place) using the EXTRACT VALUES TO POINTS tool (SPATIAL ANALYST TOOLS> EXTRACTION> EXTRACT VALUES TO POINTS). Next, open the attribute table of point data layer which is created and sort the attribute table descending based on the grid code (this is the extracted kernel density estimate). Now work out how many records is 50% of your original sample (i.e. if you had 100 records, this would the 50th record) and scroll down to you reach this poin and read off the kernel denisty value. You then reclassify your kernel density estimate raster (using the RECLASSIFY tool - SPATIAL ANALYST TOOLS> RECLASS> RECLASSIFY) based on this value with values above this having a value of 50 (because it will form the 50% PVC), and values below this being classified as NO DATA. This reclassified raster is them converted into a polygon data layer (using the RASTER TO POLYGON tool in the CONVERSION toolbox) to give you your 50% PVC. This is the repeated for the records with the top 90% highest kernel density estimates to create the 90% PVC.

 

This is not a simple process and may take a few goes to get it right. In addition, it is not an exact measure of the PVCs, but rather is an approximation of them. However, in most cases they should be very similar to the actual PCVs calculated with tools such as the old HAWTHS TOOLS for ArcGIS 9.3.

 

Anyway, I hope this helps, and if anything isn't clear, or if you run into any problems, just post again on this thread.

 

All the best,

 

Colin

 

PS We should have a supplementary workbook on calculating home ranges in ArcGIS 10.1 available by the end of the year (which might be a bit late for you unfortunately, but I thought I'd mention it).

 

 

On Tuesday, July 30, 2013 11:33:59 PM UTC+1, kateb84 wrote:

Colin,

 

I realize this is an older post, but I have searched all over, and it is the only one that really seems to have any resemblence to an actual answer for me.

I am newer to GIS and especially ArcGIS10.  We're trying to establish a home range using kernel densities.  We'd like to incorporate the 50 and 90 (read it's less biased than 95%) PCV but are aware this isn't something we can do in the current version.  Is there something others have used in the past to do this? We also are curious as to how to incorporate boundries such as a river (could we draw in our boundries and use that as our environmental extent?).  I've played around a lot with the kernel density tool, but can't figure out these few things.  I also tried your suggestion above for boundries, but it didn't work at all for me.  What I'm looking for is a step by step to get me through this.  School just doesn't prepare you for these things. I appreciate the help

 

Thanks,

Kate

[jimmy....@gmail.com]

Dear Colin.

My name is Jimmy and I'm a marine biology student currently working on a behavioural study of localized Atlantic Cod. I've followed cod tagged with ultrasonic transmitters and am now trying to describe their spatial distribution. I'm doing most of my visual work in ArcGIS 10.1, and it is in this program that I want to do the kernel density estimation. I have no knowledge of how to use GME or R so I have not been able to work through any of those programs.

Seeing as this is a coastal population with several natural barriers (such as land) I thought it best to use the Kernel Interpolation with Barriers. However, I have not been able to get any effects of the barriers I use in the tool. I have read and followed your instructions in the thread but the output still seems to disregard whatever barrier I use. I believe the problem might lie in the barrier I use and that it is not boundary.

I should note that my schooling in ArcGIS consists of a basic introductory coarse and a lot of trial and error, sometimes leading to solutions and uses not completely textbook. But that is one of the things I like about the program, that it is very flexible and allows for many different approaches to a problem. As for the barrier problem I have tried every possible approach I can think of from using a simple clipped polygon layer representing the landmass, adding self made polylines or to converting the polygon to raster and mosaic layers for the Build Boundary function. I have pretty much given up, but thought that I would try and contact you in a last effort.

I would very much appreciate any help or tips you could offer. I could send you the required layer-files if that would help. I don't know if you would need it as a layer-package, just the layer or any other format. 

The tracking method does not give a exact point for each received signal but a general area of possible residence. The point data I use is derived from this general area and that has caused some points to appear on land. Would that be a problem that some points are on the barrier itself?

Thank you for your help and attention.

Best regards, Jimmy

[GIS in Ecology]

Dear Jimmy,

Thanks for your email and sorry to hear you’re having problems.

Here’s a first thing you can try. You mentioned that:

>>The point data you see is derived from this general area and that has caused some points to appear on land. Would that be a problem that some points are on the barrier itself?

This may well be what is causing the problem. Effectively, if you have points both inside and outside the barrier, it might cause the barrier to be ignored (this has cause problems with a few other similar processes in GIS I’ve used in the past). The first thing I would probably try is to either remove the points which fall on land, or move them to the closest point which is in the sea. Do this, and then try making the kernel with barriers again and let me know if it makes a difference. If you need instructions on how to do this, just let me know.

If this doesn’t work, I can take a look at your data layers and see if I can figure out what’s going wrong. The chances are that it is something quite simple, and once it has been identified and fixed, you’ll be able to create the desired kernels quite quickly (it might just be a matter of how you’re making the barrier polygons as this can be quite tricky).

 

All the best,

 

Colin 

[GIS in Ecology]

Hi David,

 

If I'm understanding correctly, the point you're having problems with is how you assign the values from your kernel density raster data layer to your original locations so that you can work out the thresholds for the 50% and 95% percentaage volume contours.  This is done after you've made the kernel density estimate and uses the EXTRACTION tool to link your original point locations to the KDE values. Once you have this, you can then open the attribute table and examine it to get the relevatn threshold values. Once you have done this, you can go back to the KDE raster data layer and use the RECLASSIFY tool to convert it into rasters of the 50% and 95% percentage volume contours. 

 

Does this seem like it would help you get over the part you are stuck with? If not, let me know.

 

By the way, if you want to find more detailed  instructions on how to create the KDE with barriers (including an example you can work through), you can find this in MacLeod, C.D. (2014) An Introduction To Using GIS In Marine Biology: Supplementary Workbook Four: Investigating The Home Ranges Of Individual Animals. Pictsh Beast Publications.(RRP: £19.99). This is available fro Amazon.

 

All the best,

 

Colin

[jimmy....@gmail.com]

Dear Colin.

That seemed to have been the problem. I removed the ones that was on land and it looks much better now when it takes the barrier into account. I might try and move the points from land but the number of points on land is negligible compared to the total dataset, so I think this will be fine.

Thank you very much for your help.

Best regards, Jimmy.

[David Robinson]

Hi Colin,

Many thanks for the reply.

My issue was with 'assigning a location a value' not with creating the PVC. The animal I am tracking spends lots of time in one 'location' and so I have multiple XY points within that area. However, following the above method it seems to give all locations (or grid cells) a value of 1 which isn't true, some should have higher values according to the amount of time (or XY points) within that location. So I think I need to assign each grid cell a value in accordance with the number of XY points within that grid cell, at the moment they all read 1 and so the program includes everything in the analysis instead of weighting the more frequently used grid cells.

Is there a way that ARC MAP can weight the grid cells according to the amount of 'time' (XY points) an animal has spent in that cell?

Many thanks

David

[GIS in Ecology]

Hi David,

Okay, I think now I see where the problem might be. When you applied the KERNEL INTERPOLATION WITH BARRIERS tools, did you use the raw animal locations?

 

If so, this will indeed give you a value of 1 for each cell where it was recorded, regardless of the number of records in a cell. This is because the KERNEL INTERPOLATION WITH BARRIERS tool operates in a different way form the KERNEL DENSITY ESTIMATE tool, and it doesn't calculate a density (its an interpolation tool rather than a density tool).

 

Instead, as mentioned in the original post on this thread, you need to calculate the density of records in each grid cell first using the POINT DENSITY tool (SPATIAL ANALYST TOOLS> DENSITY> POINT DENSITY). This will weight each grid cell based on the number of records within it (as you wanted to be able to do), and give you a density value. This is then converted into a point data layer, and it is this density point data layer (rather than the raw sightings locations) which is then used as the basis of the KERNEL INTERPOLATION WITH BARRIERS tool. The interpolation it produces effectively smooths the density from each cell into the surrounding ones to fill in the gaps between recorded locationsd and give the home range. This process is outlined in detail (with instructions for every step) in the Supplermentary Workbook mentioned below.

 

If you have been caclulating the density of records first and are still getting a value of 1 for each record regardless of the differences in density, let me know and maybe I can take a look at your data set to see what might be going on with it as this process should work.

 

All the best,

 

Colin

[David Robinson]

Hi Colin,

Many thanks for your reply (I have ordered your book and it is on its way to me)...

OK so I am still having issues getting this to work (Have got KDE to work with 50% and 90% PVCs). The animals I am tracking are sea turtles, so I am assuming the normal KDE is not appropriate and I have to use the Kernel Interpolation with Barriers?

I have created a COST RASTER file of the world's oceans with the ocean having a value 1 but the land having no value as I can't seem to define a value for it (is this correct?) Not sure how to incorporate this raster layer into the analysis to let the program know where my barriers are?

So my procedure was:

1.) Import XY location data from excel and export data to create a layer and shapefile.
2.) Use POINT DENSITY tool to weight each grid cell (stuck here)...
... This gives me the raster which ranges from 0-484, so basically there is a blue rectangle over my data from the 0-whatever range layer (I would usually make the first layer 'no colour' to make this rectangle disappear but this doesn't translate when converted).
3.) Without changing anything, convert this back with the RASTER to POINT TOOL which gives me an attribute table with thousands of points (and a black rectangle where the blue one used to be)... most of the points are 0 but some have a figure of 87.456.
4.) Use the created POINT layer to create the KERNEL INTERPOLATION and I get a rectangular rainbow over my points (which is very pretty but I am sure not correct and I think it is due to me not excluding values of 0? maybe?? :o). I think at the KERNEL INTERPOLATION point I should also be using my COST RASTER of the world's oceans (value of 1) to define my barriers, is this correct?

I stopped at this point as I am pretty sure I have gone horribly wrong, maybe at step 1 :o)?

Any help would be appreciated until your book arrives!

Kind regards

David

[GIS in Ecology]

Hi David,

 

>>OK so I am still having issues getting this to work (Have got KDE to work with 50% and 90% PVCs). The animals I am tracking are sea turtles, so I am assuming the normal KDE is not appropriate and I have to use the Kernel Interpolation with Barriers?

If your study area contains any any areas of land, then yes, you'd be better off using the Kernel Interpolation With Barriers approach. However, if thereis no land inn your study area, then you can just use the normal KDE approach.

>>I have created a COST RASTER file of the world's oceans with the ocean having a value 1 but the land having no value as I can't seem to define a value for it (is this correct?) Not sure how to incorporate this raster layer into the analysis to let the program know where my barriers are?

The barrier is incorporated when you run the Kernel Interpolation With Barriers tool. It is entered into the section of the tool window called INPUT ABSOLUTE BARRIER FEATURES (OPTIONAL). If I remember correctly, this needs to be a feature data layer and not a raster data layer. Therefore, you would need to convert your COST RASTER file into a polygon data layer using the RASTER TO POLYGON tool (CONVERSION TOOLS> FROM RASTER> RASTER TO POLYGON) before you can use it.  At this stage, check that you have no positions for your animals which fall on land (i.e. outside your barrier polygon). This can be done using the SELECT BY LOCATION tool (accessed through the SELECTION menu on the main menu bar). If you have any points outside your barrier polygon, the kernel interpolation with barriers tool will not work properly and you need to sort this out before you proceed either by removing points which fall in land, or adjusting their positions so they fall within the barrier. Which is appropriate will depend on your specific data set..

 

>>So my procedure was:

>>1.) Import XY location data from excel and export data to create a layer and shapefile.

 

Are you remembering to transform these data into an appropriate projection/coordinate system? The point data need to be in  a projected coordinate system rather than a geographic coordinate system otherwise the density estimates will not be calculated properly.

>>2.) Use POINT DENSITY tool to weight each grid cell (stuck here)...

This isn't so much providing a weighting as creating a measure of the density of points per grid cell.

 

>>... This gives me the raster which ranges from 0-484, so basically there is a blue rectangle over my data from the 0-whatever range layer (I would usually make the first layer 'no colour' to make this rectangle disappear but this doesn't translate when converted).

 

This is suggesting that you have between zero and 484 points per grid cell. This upper limit sounds like a lot, and it's a lot more than I'd expect. In addition, I would expect this upper limit to the density to have decimal places in it.

>>3.) Without changing anything, convert this back with the RASTER to POINT TOOL which gives me an attribute table with thousands of points (and a black rectangle where the blue one used to be)... most of the points are 0 but some have a figure of 87.456.

 

This is right in principle. However, I'd expect you to have the same maximum density in the point data layer as you had inthe raster data layer. This is suggesting that the maximum value for density you have is 87.456 (this is closer to what I'd have expected - given your data, does this seem like a realistic density.

 

>>4.) Use the created POINT layer to create the KERNEL INTERPOLATION and I get a rectangular rainbow over my points (which is very pretty but I am sure not correct and I think it is due to me not excluding values of 0? maybe?? :o).

 

The way I've generally done it, you don't remove the 0 values and you just use the point data layer. This assumes that the zero values represent real absences (i.e areas which were not used). This seems to give much better and smoother kernel outputs which better match the raw point data for the individual than when you exclude the zeros (otherwise you get areas on the perifery of where the animal occurs being given a high weighting in the basis of a single point where the animal was recorded nearby). Thus, this seems the appropriate option to use when applyign this tool to estimate kernel densities in environments with barriers.

 

In terms of getting a rectangular rainbow over yoru points. This should be right. However, plot your original point data layer on top and examine it to see whether the highest densities are matching up with the highest densities of the original points. In addition, you will need to change the legend to get it to look right (and have no values displayed on land).

 

>> I think at the KERNEL INTERPOLATION point I should also be using my COST RASTER of the world's oceans (value of 1) to define my barriers, is this correct?

Yes, but see above for this.

>>I stopped at this point as I am pretty sure I have gone horribly wrong, maybe at step 1 :o)?

>>Any help would be appreciated until your book arrives!

Hopefully the above can provide you with some help. However, without seeing your specific data set, it is difficult to see exactly where you might be going wrong or whether the results you are getting are correct or not. Once the book arrives, you will be able to work through an example from start to finish, and check it against the results you should be getting. This way, you'll know if you're doing things correctly or not. Once you have done this, you can adapt the instructions for your own specific data set.

 

As always I hope this helps, and as alwasy feel free to post again if you're still stuck.

 

All the best,

 

Colin

 

[Jake Wall]

Hi Colin,

I've just released a version of ArcMET that works with ArcMap v 10.2 at www.movementecology.net. 

all the best,

Jake

[GIS in Ecology]

Hi Jake,

 

Thanks for letting us all know that ArcMET has been updated for ArcMap 10.2. This, I'm sure, will be a great help to many people who are interested in looking at individual home ranges using GIS.

 

All the best,

 

Colin 

[cha...@wcs.org]

Hi Jake,

I'm keen to try your new tool out and have attempted to install it on my machine that has ArcGIS 10.2 installed and I'm getting the message "ArcGIS Desktop 10.2 is required before you can complete this setup". Do you have any idea what might be causing this problem? I have only just upgraded from 10.1, could there be an issue associated that?

Thanks
Chris

[gab_s...@hotmail.com]

Hi Colin,
I am very glad I found this thread as it is the only one related to my problem.
I am currently trying to do KDE using the latest version of GME on a 150 points shapefile on ArcGIS 10.1. It kept giving me the same error message but now the software keeps crashing. When it was working, I tested it with other tools (MCP and isopleth on a kde raster done with ArcGIS) and it all works. The only reason I really need to use GME to do kde is I need bandwidth selection method to be SCV (squared cross validation).

So my first question is:
Is it possible to select bandwidth with SCV in ArcMap 10.1? If yes, problem solved!

If no, I don't really know what to do to get GME to work.I read the entire thread and tried everything suggested:
-Uninstall and reinstall GME twice
-Shorten the path and file name
-Check for any projection issue
-Email GME creator (no answer)

Here is the command and error for GME:

kde(in="C:\Users\Utilisateur\Desktop\22781_GIS\kde2011.shp", out="C:\Users\Utilisateur\Desktop\22781_GIS\kern2011", bandwidth="scv", cellsize=200);

Error: The command text could not be interpreted. Please check the syntax of the command.

Error: An important error has occurred. Please include the information below if you submit a query about this error. La référence d'objet n'est pas définie à une instance d'un objet.

Any input on this would be appreciated.
Thank you,

Gabrielle

[GIS in Ecology]

Dear Gabrielle,

 

Unfortunately, there's no way to do SCV for kernels in ArcGIS directly.

 

In terms of GME, I'm not familiar with doing kernels in it (I mostly do kernels with barriers which GME can't do).

 

However, you could try changing the SCV in your command from small letters to capitals so it would read:

kde(in="C:\Users\Utilisateur\Desktop\22781_GIS\kde2011.shp", out="C:\Users\Utilisateur\Desktop\22781_GIS\kern2011", bandwidth="SCV", cellsize=200)

This is because it might be case sensitive. Also, there appears to be a semi-colon at the end which might be causing issues (or it may just be the way you've pasted in into the post).

As a by the way, SCV in GME doesn't stand for Squared Cross Validation, it stands for Smoothef Cross Validation - which is potentially something quite differeent. I've not heard of Squared Cross Validation before, but there is Least Squares Cross Validation, which might be what you're meaning. If you're wanting to use Least Squares Cross Validation, you would put  bandwidth="LSCV" into your command.

If you are meaning LSCV, and this is what you want to use, you could also try using the ADEHABITAT package in R (http://cran.r-project.org/web/packages/adehabitat/index.html). This allows you to do LSCV, but requires that you know howo to use R (not the easiest software to use!).

An alternative might be to use the ArcMET for ArcGIS (http://www.movementecology.net/). This doesn't have LSCV available yet, but this option is marked as 'Coming Soon@, so watch that space for more information.

Hopefully one of these options will help.

However, there's another possible issue here. I notice that in the error message you pasted into your post, part of it was in French. If you are working on a computer which is set to use a comma (,) as the decimal separator rather than a decimal point (.), this can cause all sorts of problems with GIS, and ones which can be very hard to track down and solve. In at least one instance when trying to create kernels in GME/ArcGIS, the problems were solved by changing the decimal separator from a comma (standard for many continental European computers) to a decimal point (standard in English, and in many software packages). However, this isn't something which you should do lightly as it may affect other aspects of functioning on your computer.

In addition, when working across a range of software packages (for the kernel analysis in GME, this involves GME interfacing with ArcGIS and R, as well as its own software code), making sure that any changes to the decimal separator are recognised across all the software packages can be difficult (the previous person who had this problem ended up resetting their entire computer to English and then re-installing all the software packages - this was only worth it because they were at the start of a 3 year Ph.D. in the UK).

Hopefully, this is not the reason you are having problems. If it is, you could try going down the ADEHABITAT route rather than using GME and see if that works (you'll be working within a single software package and which will reduce the risk of problems caused by decimal separators.

Sorry not to be able to provide you with an easier solution for this. If any of these does help, please post and let me (and others who might run into the same problem) know what works.

All the best,

Colin

On Saturday, June 14, 2014 2:31:14 PM UTC+1, gab_s...@hotmail.com wrote:

[Manuel Parejo Nieto]

Dear Gabrielle I have the same problem:

My name is Manuel Parejo Ph.D in Facultad de Ciencias (Universidad de Extremadura)
Spain, and I research animals fitted with GPS-GSM tags
I read your reply in the Arc GIS forum about the  Estimating Kernels Using Geospatial Modeling Environment

I have this problem with GME:




setwd( all="C:\Users\MPN\Desktop\Positions_table\shapes");

Current input working directory:
C:\Users\MPN\Desktop\Positions_table\shapes

Current output working directory:
C:\Users\MPN\Desktop\Positions_table\shapes

kde(in="pint01utm.shp", out="kdep_01_1dic.img", bandwidth="SCV", cellsize=10);

Error: The command text could not be interpreted. Please check the syntax of the
command.

Error: An important error has occurred. Please include the information below if you
submit a query about this error.
Excepción de HRESULT: 0x80040358




And I don´t Know if the problem is in the syntax of the command or in my data
because I have reviewed exhaustively

PD. My data were in coordinate system GCS-WGS 1984 and I transformer this data in
Universal Transverse Mercator (ETRS 1989 UTM Zone 30),

thank you very much,
Yours sincerely

Do you Know the solution?

[GIS in Ecology]

Hi Manuel,

 

I'm not too sure what's going on here, but the things I'd try is:

 

1. Try using a different method for calculating the best bandwidth. For example, try LSCV instead  of SCV. If it works with a different bandwidth, then it is something to do with the SCV bandwidth calculation option.

 

2. Try changing the output file type and see if it makes a difference.

 

3. Try using a different data set and see if you can get the commands to work with it.

 

4. Try calculating the kernel density estimate directly in ArcGIS using the kernel density estimate tool. If it works wtih your data, this would suggest if is something to do with how  GME is processing the data. If is doesn't work, it suggests it is something to do with the data set itself (most likely to do with some sort of incompatibility with the projection/coordinate system or with the choice of cell size - 10m is a very small cell size).

 

If 1, 3, and 3 don't work, but 4 does, the chances are there is a problem with the coding which you've used. Is the semi-colon at the nd of your coding part of your command? If so, this might be causing as problem. If not, check that your files have the right addresses (inccluidng y appropriate capital letters), and then double check that you haven't missed anything out in the syntax of your coding.

 

As I said to Gabrielle,  I don't generally use GME for home ranges, and do these directly in ArcGIS, but these would be my suggested starting points.

 

I hope this helps, and if this solves things for you, or if you find a different solution, please post here so others can read about how this problem was solved.

 

All the best,

 

Colin

cG

[Holley Kline]

Hi,

I am using ArcMap 9.3 and having trouble creating fixed KDEs using Home Range Tools. I have tried Hawth's tools, but need to use least squares cross validation. Home Range Tools will not allow me to create KDEs; it gives me errors such as "KDE:104, subscript out of range" and if I try to calculate the independence of observations, another error pops up - "Error 1001 Overflow". I have no idea what these mean or how to fix them. Does anyone know a solution?

Thanks,
Holley

[GIS in Ecology]

Hi Holley,

 

Thanks for the post. I haven't used the Home Range Tools myself, but this sounds like it might be an issue with either the projection/coodinate sysem of your data layers/data frame or with the extent of the kernel density estimate you're trying to create. Try checking that all your data layers and yourr data frame are in the same projection/coordinate system, and that this projection/coordinate system is a projected one which has a map units of metres ratehr than a geographic one where the map units are in decimal degrees.

 

Next, check that you don't have any points with missing latitudes and/or longitude values which are plotting in odd places. To do this, right click on your point data layer and select ZOOM TO LAYER. This will allow you to look ffor possible outliers that you willl need to deal with before running th KDE tools.

 

Finally, check that the extent you are using to create the KDEs is suitable and that it's been specified using the coordinates which are in the same units as your projection/coordinate system. In addition, it may need to enclose the entire distribution of your points which you are using to make the KDEs.

 

You might also want to check that you have enough points to make your KDEs in the first place.

 

Beyond that, I'm not too sure what might be causing these errors. Does anyone else have any ideas?

 

All the best,

 

Colin

[Holley Kline]

Colin,

Thanks for the reply! It doesn't seem to be the projection - I did have to re-project the locations, but there are still errors. There are definitely enough locations and none of them are missing long/lats, so those possibilities are out. However, I noticed you suggested checking to make sure the extent is appropriate. How might I determine what extent size will cover the distribution of my locations? I am thinking this is a possible solution since one of the errors that pops up says KDE:104 subscript out of range. I could be totally wrong about this though, as I am a newbie with ArcGIS. If anyone else has any suggestions, I'd really appreciate them!

Thanks for all of the help,
Holley

[GIS in Ecology]

Hi Holley,

 

I've been looking into this some more. The 'out of range' error means that there is some number somewhere which is much bigger or smaller than the software is expecting and so it is being flagged up as an error. Are you by any chance using a field with the number of individuals at each point in it when you'ree making your home ranges (for example a measure of group size or something similar? If so, it might be that some of these are missing, are zero, are decimal fractions or are non-numeric values.

 

Similarly, are you putting in cell sizes as metres instead of kilometres, or kilometres instead of metres? Or are you using a cell size that is to big or small in relation to the extent of the point data layer? There might also be isues with the bandwidth/h you are selecting (although if you're using LSCV, you're probably not setting this youself)?

 

Alternatively, if it's to do with the extent, you can select ZOOM TO LAYER so that you can see all your point locations, and then use the cursor to work out what the top, left, right and bottom values would be needed to cover this entire area by hovering your cursor over the appropriate position and reading the coordinates off bottom left corner the ArcMAP window.

 

One option to try to help you work ou what's going wrong would be to try doing a kernel analysis on your data with either Hawth's tools, or with the KDE tool in ArcMAP (If I remember correctly for the 9.3 version, it should be in SPATIAL ANALYST TOOLS> DENSITY> KDE in the TOOLBOX window. If you can successfully create your kernels using these other tools, then the problem almost cetainly lies with your Home Range Tools toolbox or one of the settings you are using. If, however, you still have problems, this would suggest that it's a problem with the data itself. Any additional error messages you get might help you track down what this is. 

 

I hope this helps.

 

All the best,

 

Colin 

[Eric Brossman]

This sounds familiar with what I am trying to do.  My question is basic, I know I need to use GME but I have no clue how to get started.

On Sunday, April 7, 2013 5:08:00 AM UTC-5, GIS in Ecology wrote:

Hi Sophie,

 

To start out with, I'll say that I tend to calculate kernels directly in ArcGIS rather than GME so I'm not that familiar with calculating it through GME. However, it uses the same code so it should give the same results. This having been said, I should be able to work out what might be going wrong here. However, I'd probably need some more information.

 

Firstly, what settings are you using when you run the KDE tool for things like cell size, bandwidth and weightfield?

 

Secondly, you say that you're calculating the 95% and 50% kernels, are you doing this with the KDE tool or are you using one of the other tools in GME to convert the KDE raster into 95% and 50% kernels?

 

Finally, what species are you working with (this is just for my own curiosity!)

 

I should also say at this point that your sample sizes are close to the minimum you'd need to be able to accurately estimate the home ranges (which is around 30 points). This isn't to say that it won't work, but it does mean you may need to be careful when it comes to interpreting the results.  

 

Anyway, if you can get me this additional information, I can see if I can help out. It might also be useful if you can send a screenshot of your KDE raster and then your 95% and 50% kernels.

 

All the best,

 

Colin

 

 

On Sunday, April 7, 2013 10:36:53 AM UTC+1, Sophie Rasmussen wrote:

Dear Colin. 

I have been wondering about the 50% kernels I have created in GME. The 95% kernels look decent, but the 50% kernels only contain around 5-8 location points out of 28-35, which of course is a lot less than 50%. Why might that be? I have tried to re-do them several times, and they always turn out the same way. 

Kind regards, Sophie 

 

[GIS in Ecology]

Hi Eric,

 

If you are interested in generating home range information, such as 50% percentage volume contours etc, these can be done through GME. However, if you are starting from scratch, I would suggest looking into using the ArcMET extension for ArcGIS. This will allow you to generate kernel density estimates and then from that percentage contours. As this package has been specifically created to look at animal movements and home range, it is probably easier to use than GME for this purpose.

 

However, you need to be careful when generating any sort of home range information using kernel density estimates. In particular, you need to ensure that you use an appropriate bandwirth/h value and this can be difficult to determine. ArcMET has some options which will help you do this though.

 

I hope this helps.

 

All the best,

 

Colin

[Eric Brossman]

Thanks Colin:

I suppose I have the same issue with ArcMET.  I got the extension but not sure how to generate a KDE. I set things up to run the KDE.  I set the input as the location points for the animal and I get a warning saying, "The Tracks feature class must be in  projected coordinates for this tool to work."  So, I changed the animal's location points layer projection to a projected coordinate system.  It ran the KDE and in the table of contents, the output was a geodatabase file and a summary table.  I'm not sure what to do or what needs to be done next.  Thanks for your help.

Eric

[GIS in Ecology]

Hi Eric,

 

Once you have created your KDE layer and added it to your GIS project, you can create the percentage volume contours. In ArcMET this is done using the PERCENT VOLUME CONTOUR tool (under the UD menu). Once you have that you can do things like calculate the area and the spatial overlap between different individuals.

 

However, while this is a fairly basic introduction to how to create KDEs/percentage volume contours, actually using the tools properly can be difficult. In particular, you need to select an appropriate cell size, extent, and most importantly, the bandwidth values. Selecting the most appropriate bandwith value can be done in a number of different ways, and there is no hard and fast rule on how to do this. You'll probably need to do quite a bit of background reading before you choose an approach so you can see what others who work in your particular field generally use. Also, you will probably want to do a bit of testing to see how big a difference using different bandwidth selection options makes (and using different cell sizes and extents).

 

I hope this helps, and if you get stuck with anything else, feel free to post again.

 

All the best,

 

Colin

[Hiro]

Hi, I had used older version of GME with arcMap10 until upgrading arcMap, and that was fine. 

Just today, I installed a new ver. of GME 0.7.3 with arc map 10.2.1 and R 3.1.1.

However, GME does not worked well. Maybe link with R is Okay, but it will be forced to shutdown when i push the folder button to select dataset in KDE, Isopleth and etc. commands.

I dont know why I cannot call dataset files. If you have any ideas towards this bug, I'd like you to help me.

Thanks,

Hiro

[GIS in Ecology]

Hi Hiro,

 

I would try uninstalling GME and R, and then make sure that any older versions of them have been uninstalled from your computer. Then, try re-installing R again.  When you do this, make sure that you use the default folders which it wants to install everything in. Once you have re-installed it, open R and make sure you download and install the required package for doing kernel analyses. To do this, use the code:

 

install.packages(pkgs=c("CircStats", "deSolve", "coda", "deldir", "igraph", "RandomFields", "ks"))

 

Once you have these installed, re-install GME, again using the default folders suggested.

 

Finally, try opening GME again and see if it will now work properly.

 

If you're still having problems, check that McAfee anti-virus software has never been installed on you computer.If you are using it as your antivirus software to protect your computer, that is fine. However, if it has been uninstalled at some point in the past, it may not have been uninstalled properly. If this is the case, go to: http://service.mcafee.com/faqdocument.aspx?id=TS101331 and follow the instructions to completely uninstall it.

 

Once you have done this, try removing and then re-installing GME again and see if that solves the problem.

 

Beyond this, I'm not really too sure what else to suggest. GME can be tricky to install sometimes, but usually the problem is solved by removing and re-installing it.  

 

All the best,

 

Colin

[Hiro]

Hi Colin,

Thanks for your soon reply. I tried uninstalling and re-installing R and GME, and updating microsoft NET framework, but GME cannot work.

GME appears to connect to "C:\Program Files\R\R-3.1.1\bin", 

and "CircStats", "deSolve", "coda", "deldir", "igraph", "RandomFields", "ks" pakages are in "C:\Program Files\R\R-3.1.1\library" with other pakages. 

Is it no probrem? I never downloaded Mcafee also.

Anyway, i have no idea, ill try in the other computer.

I really appreciate your help.

Regards,

Hiro

2014年9月8日月曜日 18時24分25秒 UTC+9 GIS in Ecology:

[Hiro]

Hi Colin,

Today I could install and use GME.

The cause of the problem may be the remained files and registries of older edition, arcGIS 10 Japanese pack.

Because of the installer bugs, I could not uninstall it until today.

After uninstall all arcGIS softs, delete the registry, and re-install 10.2.2 again, it worked well. 

Thank you so much for your thoughtful advises.

Regards,

Hiro

2014年9月8日月曜日 20時52分42秒 UTC+9 Hiro:

[GIS in Ecology]

Hi Hiro,

 

Sorry to take so long to reply to you. Good to hear that you finally got this sorted, and thanks also for posting how you solved this problem. This will hopefully help anyone else who runs into the same problems that you had.

 

All the best,

 

Colin

[Lianne]

I am trying to create 95% MCPs by year and individual, for several individuals.  For some it works fine, but for others it does not. It technically creates the polygon but the polygon does not show up on the map, and in the attribute table it says area = 0.  If I change it to a 99% MCP for the same animal/same locations, it works fine.  All MCPs have ~100 points, so it shouldn't be a small sample size issue.  I have no idea why this is happening, any insight will be much appreciated!

[GIS in Ecology]

Hi Lianne,

 

Thanks for the posting.  Can you tell me what software/extensions you're using, and also whether you're following my instructions from this thread or from somewhere else? That will hopefully help me work out what's going on.

 

In the meantime, my guess is that that you have a lot of points which plot in exactly the same locations (i.e. directly on top of each other because they have the exactly the same coordinates), and this is known to be a problem with many types of home range analyses. This would certainly give the problem that you are having and it would be something that would be worth checking into just to make sure that this isn't the issue.

 

Anyway, if you can let me know a little bit more about the software you're using and tthe approach you're taking, then  should be able to shed some more light on this.

 

All the best,

 

Colin 

[Lianne]

Colin,

I was using the Home Range Tools extension in ArcGIS 10.1.  There are identical locations, so that may be the issue.  I will look into that and let you know. 

I am able to use the Minimum Bounding Geometry tool as you suggested, however I want to create the 95% PVC (in Home Range Tools I was able to set the MCP to 95%).  The plan is to calculate the percent overlap of polygons for each individual for each year.  I am relatively new to GIS and absolutely new to R and GME, so I am trying to learn as I go!  This forum has been really helpful! Thank you!

[GIS in Ecology]

Hi Lianne,

 

Sorry to take so long to get back to you on this. Good to hear that the forum has ben useful to you. I think the fact that you have identical locations may well be at the heart of the problems you'e having as many of the home range tools really don't like this. One way to deal with this is to introduce a small amount of random error to your positions so that points don't stack directly on top of each other. This can be done quite easily in Excel.

 

Anyway, keep me posted on how you go with this, and I'm sure together we can get it working for you one waay or another.

 

All the best,

 

Colin

[Matt Danzl]

Hi Colin,

Trying to find home ranges of grouse. I used the kernel density in arc 10.1 and it produced this:

Is there some way that it doesn't cut off the edges? I let Arc choose its own radius and it looked different but I messed around with it in the symbology with the classify button to produce this. Is that an acceptable way to make 50 and 95% levels?

thanks,

Matt

[GIS in Ecology]

Hi Matt,

 

The reason that your kernel density is cut off around the edges is because you have used the default setting for the output extent of the kernel density tool. To avoid this, first, move the cursor around to the position where you would want the top left hand corner of your kernel density raster to be positioned and then read the coordinates for this position off the bottom righ corner of the ArcMAP user interface. Now repeat this to get the coordinates of the desired bottom right corner.

 

Once you have these coordinates, then open the kernel density tool, and once you have filled in all the required settings, click on the ENVIRONMENTS button. In the ENVIRONMENT SETTINGS window, click on PROCESSING EXTENT, select AS SPECIFIED BELOW and fill in your desired coordinates. Finally, click OK to close the ENVIRONMENT SETTINGS window, and then click OK to run the KERNEL DENSITY tool.  As long as you've got the coordinates right, this will mean that you no longer have your kernel density estimate cut off around the edges.

 

In terms of working out the 50% and 95% percentage volume contours (PVCs), you can do this by adjusting the symbology, but this will only change how it is visually displayed. This is okay if that is all you want to do, but if you want to do things like calculate the area of the PVCs, you would need to used the RECLASSIFY tool to generate data layers which only represent this information.

 

Finally, in terms of the search radius (also known as bandwidth or h), in general you shouldn't use the default option in ArcGIS (or any other GIS software). Instead, you should use some sort of objective selection process. Personally, I like the ad hoc method for doing this as it is one of teh easiest to apply and has a good logical consistency. You can find a paper detailing this approach here: http://www.animalbiotelemetry.com/content/pdf/2050-3385-1-13.pdf.

 

I hope this helps, and any further questions, just post back on this thread.

 

All the best,

 

Colin

[Benoit Laliberté]

Hi Colin,

Was is the meaning of the probability cut-off value in ArcMET?

By default, the value 1E-15 will return full polygons (ie. 100% contour) whether I input 0.50 or 0.95 as percentile value. Changing the cut-off values outputs different size polygons, but which cut-off value to use?

Thanks,

Benoit (Québec City)

[Benoit Laliberté]

To everybody's benefit, here the answer from the developper of ArcMET, Jake Wall:

To answer your question, the probability cut-off value defines the cell value where a pixel will be assigned a NULL value. (i.e., if a given pixel value is below the cut-off then it will be made NULL).

The density threshold distance will assign a NULL value to any pixel that is beyond the specified number of standard deviations from the fix point. This prevents the code from having to calculate every grid cell since the contribution to grids far away from a point is negligible.

[Benoit Laliberté]

I've been working on those Kernels for a while now and I can't seem to be able to go around a few basic problems. May be you can help. I've tried ArcGIS Kernel Density Tool, ArcMET extension, GME and Animal Space Use so far. The main problem I have is that I can't get any of these to work properly.

I have an ArcGIS 10.2.2 license and the best (only working) path so far as been to create Kernel rasters in ArcGIS and draw 95% contours using GME tool. I've read quite alot on the subject, but there are still a few things I can't grasp :

1- In ArcGIS Kernel Density Tool, the main parameter I can control is "search radius". Is that "search radius" equivalent the the "h-value" mentioned in other tools and literature?

2- If that is the case, I understand that I can basically choose the h-value that fits the best (I've read about h-ref, LSCV, etc.  and the ad hoc method paper by Kie, method comparison by Mitchell (2007) and others)?

3- Just to confirm, I assume that I pick different (ie: the best) h-values for each individual?

4- What is the meaning of the output values of the raster kernel? For example, if I don't input any specific h values/search radius, I get values from 0 to 3048.9729... And about 65,000 pixels, 20,000 of which have a value of 0.

5- And finally, how do these values relate to the 95% contours I'm trying to get (I'm thinking I should exclude all pixels with a value = 0, since they basically fill the extent of my raster, and then pick the pixels that sum up 95% of the values of the raster?)

Thanks

Benoit Laliberté 

[GIS in Ecology]

Hi Benoit,

 

>>I've been working on those Kernels for a while now and I can't seem to be able to go around a few basic problems. May be you can help. I've tried ArcGIS Kernel Density Tool, ArcMET extension, GME and Animal Space Use so far. The main problem I have is that I can't get any of these to work properly.

 

If you are consistently having problems across a lot of software packages, before you do anything else, I would double-check the projection/coordinte system of your point data layer and ensure that the one you are using is appropriate.

 

>>I have an ArcGIS 10.2.2 license and the best (only working) path so far as been to create Kernel rasters in ArcGIS and draw 95% contours using GME tool. I've read quite alot on the subject, but there are still a few things I can't grasp :

 

>>1- In ArcGIS Kernel Density Tool, the main parameter I can control is "search radius". Is that "search radius" equivalent the the "h-value" mentioned in other tools and literature?

 

Yes, these are the same thing (just with different names).

 

>>2- If that is the case, I understand that I can basically choose the h-value that fits the best (I've read about h-ref, LSCV, etc.  and the ad hoc method paper by Kie, method comparison by Mitchell (2007) and others)?

 

As long as you use an objective method to select the optmal h-value, yes. Personally, I'd go with Kie's ad hoc method myself.

 

>>3.- Just to confirm, I assume that I pick different (ie: the best) h-values for each individual?

 

That  depends on your exact method. Some times people will use different h-values for each individual, others will use the same one across all individuals.

 

>>4- What is the meaning of the output values of the raster kernel? For example, if I don't input any specific h values/search radius, I get values from 0 to 3048.9729... And about 65,000 pixels, 20,000 of which have a value of 0.

 

The values in the kernel raster are an estimate of the expected density of records in each cell. The higher the value, the greater the estimted usage.

 

>>5- And finally, how do these values relate to the 95% contours I'm trying to get (I'm thinking I should exclude all pixels with a value = 0, since they basically fill the extent of my raster, and then pick the pixels that sum up 95% of the values of the raster?)

 

The 95% Percentage Volume Contour (PVC) is generated by working out what pixel value will allow you to create a polygon that encloses 95% of the expected usage. This can be done using the tool in GME, or you can do this manually in ArcGIS by extacting values rom the kDE raster layer to you original point data layer. Sort these in descending order by the extracted KDE value, and then read off the DE value for the 95% record (to find this, multiply your total number of records by 0.95 and scroll down to the this number  from the top). Once you have this value, you acn ue the RECASSIFY tool to convert your KDE into a data layer that identifies the cells which fall in the 95% PVC, and the convert this into a polygon which will for the 95% KDE itself. The full instructions for doing this can be found 'An Introduction To Using GIS In Marine Biology: Supplementary Workbook Four' (Pictish Beast Publications).

[Anna Meissner]

Hi Colin, Hi all,

I am currently examining common dolphin distribution in NZ waters, using ArcGIS 10.2. For that, I divided my area into grid cells (4*4km). As I have information of numbers of groups and group sizes, I was able to calculate sighting frequencies and encounter rates for the area. I would like to investigate the species home range as the next step.

From all the very interesting posts and answers (thanks all for making this thread so useful), I will look at using kernel interpolation with barriers (to account for the coastline and few islands on the periphery of the area).

It has been suggested to first create a point density layer, instead of using raw points. Instead of only accounting for the number of groups, my idea is to use the group size (as the value for the population field).  When dealing with gregarious species, I feel like this parameter will best reflect density estimations. However, maybe I should look at the mean group size or median to take into account the number of groups as well? Also shouldn’t this value be somehow weighted by the effort of the grid cells?

 

Thanks a lot for your advice :-)

Best, Anna

-------------------------------------------------
Anna M. Meissner
PhD student
Coastal-Marine Research Group
Institute of Natural and Mathematical Sciences
Massey University
Private Bag 102 904
North Shore City, 0745
Auckland, New Zealand

Tel: +64 9 414 0800 ext 41520
Cell: +64 22 126 18 19
Fax: +64 9 443 9790

Email: a.m.me...@massey.ac.nz
Web: http://cmrg.massey.ac.nz

Recent paper:

Meissner AM, Christiansen F, Martinez E, Pawley MDM, Orams MB, Stockin KA (2015) Behavioural effects of tourism on oceanic common dolphins, Delphinus sp., in New Zealand: The effects of Markov analysis variations and current tour operator compliance with regulations. Plos One 10: e0116962.
-------------------------------------------------

On Thursday, September 20, 2012 at 11:12:48 PM UTC+12, GIS in Ecology wrote:

Estimating Home Ranges Of Individual Animals In ArcGIS

 

It is often useful to be able to identify and compare the areas used by individual animals in a population. These areas are known as home ranges. There are two widely accepted ways to estimate an animal’s home range.  These are through a minimum convex polygon (MCP) or through a kernel density estimate (KDE).  An MCP is the smallest polygon that can be drawn that will encompass all the locations where an individual is recorded that contains no internal angles greater than 180 degrees.  It, therefore, provides a measure of every area where an individual has been recorded, but it provides no information about whether some areas are used more frequently than others. A KDE measures the density of records within each grid cell that covers a study area, and uses this to estimate the probability that an individual will use neighbouring cells.  It, therefore, provides an estimate of which areas an individual uses most frequently.  Typically, KDE are converted to percentage volume contours (PVCs) that identify areas where an individual is likely to occur 50% of the time (often referred to as the core range) and 95% of the time (often taken as a measure of the total home range).

 

There are two options for estimating home ranges within ArcGIS.  The first is to use ArcGIS tools directly. However, the tools that can be used to estimate home ranges are not easy to identify because of the names used for them.  To create an MCP, the tool that you would use is called MINIMUM BOUNDING GEOMETRY.  To use this to create an MCP, go to DATA MANAGEMENT TOOLS> FEATURES> MINIMUM BOUNDING GEOMETRY.  In the MINIMUM BOUNDING GEOMETRY window that will open, select CONVEX HULL for GEOMETRY TYPE (OPTIONAL), and this will allow you to create and MCP.

 

To create a KDE, you can use the KERNEL DENSITY tool.  This can be found in SPATIAL ANALYST TOOLS> DENSITY> KERNEL DENSITY.  The main limitation of this tool is that it assumes that there are no barriers to an individuals movements. These means that it is not well suited to creating KDEs in areas where such barriers exist.  For example, it would not be appropriate to use this tool to estimate the home range of individual bottlenose dolphins in a coastal population (as they cannot move across land). Instead, in such instances, you can use two other tools. First you can create a point density raster data layer using the tool POINT DENSITY (SPATIAL ANALYST TOOLS> DENSITY> POINT DENSITY. You can then convert this data layer to a point data layer and use a tool called KERNEL INTERPOLATION WITH BARRIERS (GEOSTATISTICAL ANALYSIS TOOLS> INTERPOLATION> KERNEL INTERPOLATION WITH BARRIERS) to create your KDE.  

 

In all cases, when creating a KDE, you need to make sure that you select appropriate grid cell sizes, extents and search radius for your individual animals and species.  

 

Alternatively, if you don’t wish to use these tools in ArcGIS directly, you can use one of several specialist extensions to create your home range estimates. These include Hawth’s tools and Home Range Tools for ArcGIS.  Unfortunately, at the moment most of these are designed to work with ArcGIS 9.3 or earlier and not the more recent ArcGIS 10. For ArcGIS 10 or later users, the only option that seems to be currently available is Geospatial Modelling Environment (GME).  It has tools that allow you to create MCPs and KDEs (although it does not allow you to create a KDE with barriers).

[GIS in Ecology]

Hi Anna,

 

This is an interesting question,. Really, the techniques for home range analyses, such as creating kernel density estimates, are designed to be comparative rather than absolute. This means that they are primarily designed for comparing home range estimates between different individuals (or species, or any other grouping) which have the same level of survey effort for all of them. Therefore, the estimated home ranges are always implicitly preceeded by the phrase, 'given the survey effort for this study ...' and should be restricted to making comparisons in the home range between individuals (overlap, size, habitat preferences etc).

 

What you are wanting to do is to estimate the absolute home range of an individual species, rather than to make any comparions between species in your study area for a given level and distribution of survey effort. In addition, home ranges are, by definition, a single polygon that encloses all the areas is used by an individual/species, including corridors between areas of high usage (this goes back to the original home range definition from Burt 1943). Now, we know that common dolphin also occur outside your study area, so that the 'home range' which you would estimate will always be an under-estimate of the 'home range' of the species, and so should always be expanded to include most, if not all, of your study area, if the single polygon definition is to hold.

 

This having been said, you can use kernel density estimates to give you an idea of the locations in your study areas where your species preferentially occurs or uses, and as you said, it would be best to take your levels of survey effort into account in some way. With both the the kernel desnity with barriers approach (which you are right would be the best one for you to use) and the kernel density without barriers tools available in ArcGIS, there are no options to include a weighting to take survey effort into account (this is because of the underlying assumptions of equal survey effort). However, you could get round this by calculating the number of animals per unit survey effort recorded in each surveyed grid cell (see exercise four in my Supplementary Workbook One for full details of one way to do this - but I think you already know how to do this from your masters thesis work), and then use these values as the INPUT POINT FEATURE data layer when creating the DENSITY POINT RASTER DATA LAYER (if creating a kernel density estimate with barriers) which you would then use to create your kernel density estimate following the above instructions.

 

However, while this is possible, I would argue that it is not appropriate for you to do with common dolphin. This is because of the way that kernel density estimate work. Basically, they assume that the density of usage for an unsurveyed cell is proportionate to its distance from the value found in the nearest surveyed cell. This means it is effectively smoothing the values for surveyed cells into those that surround it. Now, this works really well for homogenous environments, or homogenously distributed species, or even environments with relatively smooth and shallow gradients in occurrence. However, my experience suggests that cetacean habitat use is extremely heterogeneous and that there are sharp demarkations based on environmental variaitons between where certain species occur or don't occur (although this might be less so for pelagic species, such as common dolphin, than the benthopelagic species I'm more used to working with). This means that the underlying assumption of kernel density estimates that the values for unsurveyed cells will be most similar to those closest to it may well be violated. As a result of this, simple smoothing approaches, such as kernel density estimates, kriging, and I would also say density surfaces which use latitude and longitude as explanatory variables, shouldn't be applied to analysing the distribution of single cetacean species (but that's very much mu individual opinion).

 

Instead, I would argue that you would be better off creating a species distribution model (SDM), based on the relationships with environmental variables, which takes the varying levels of survey effort in each grid cell into account (my Supplementary Workbook One shows a way to calculate survye effort per grid cell using a polygon grid data layer). Once you have this SDM, you can then make a prediction of the expected distribution across your entire study area (see Supplementary Workbook Three), using the same environmental variables. Once you have this, you can then select a threshold value and convert this into the equivilant of a 'home range' that would identify the cells whihc wouldbe considered part of the the species normal usage (i.e. its core habitat). There's a number of ways of identifying this threshold, but I always like the method I explored with Laura Mandleberg for her masters thesis, where you look at the relationship between size of the predicted area of occurrence vs the number of false absenses in a model testing data set, and use this to identify the point where lowering the threshold any further increases the area of predicted occurrence without proportionately reducing the number of false absences.

 

Anyway, that is just my thoughts and I'm sure others would differ in their opinions. Hopefully, this will allow you to go down the KDE route, if you so choose, while taking the levels of survey effort into account, but will also lewt you know how to go down the SDM route if, given what I've said, you feel that this would be better.

 

I hope this helps, and if you get stuck with any of it, or if you have any further questions about the above, just post back on this thread.

 

All the best,

 

Colin

 

 

 Wednesday, February 4, 2015 at 11:41:25 PM UTC, Anna Meissner wrote:

Hi Colin, Hi all,

I am currently examining common dolphin distribution in NZ waters, using ArcGIS 10.2. For that, I divided my area into grid cells (4*4km). As I have information of numbers of groups and group sizes, I was able to calculate sighting frequencies and encounter rates for the area. I would like to investigate the species home range as the next step.

From all the very interesting posts and answers (thanks all for making this thread so useful), I will look at using kernel interpolation with barriers (to account for the coastline and few islands on the periphery of the area).

It has been suggested to first create a point density layer, instead of using raw points. Instead of only accounting for the number of groups, my idea is to use the group size (as the value for the population field).  When dealing with gregarious species, I feel like this parameter will best reflect density estimations. However, maybe I should look at the mean group size or median to take into account the number of groups as well? Also shouldn’t this value be somehow weighted by the effort of the grid cells?

 

Thanks a lot for your advice :-)

Best, Anna

-------------------------------------------------
Anna M. Meissner
PhD student
Coastal-Marine Research Group
Institute of Natural and Mathematical Sciences
Massey University
Private Bag 102 904
North Shore City, 0745
Auckland, New Zealand

Recent paper:

[Anna Meissner]

Hi Colin,

I am comparing home ranges of fur seals and common dolphins in NZ. Im working on ArcGIS 10.2.
Is it possible to get negative values while mapping kernel estimates? I have no problems with Delphinus. Mapping fur seal home ranges causes me a problem. If I map a kernel without barriers the minimum value equals 0, however, when I map it with the barrier, the minimum reaches -0.004 (please see attachment).

Thanks for your feedback,

Best anna

[GIS in Ecology]

As a result of discussions during a recent course, I finally sat down and put together a custom toolbox for ArcGIS users for automatically running home range analyses. These tools are based on the processes first outlined on this thread, and later fleshed out in An Introduction To Using GIS in Marine Biology: Supplementary Workbook Four - Investigating The Home Ranges Of Individual Animals. This includes the ability to create kernel density estimates (KDE) in environments with barriers to movements. Currently, there are few other ways of doing this, and none that I know of which allow you to take into account complex barriers to movements (such as coastlines). These tools can be used  with both marine and terrestrial data sets.

 

In addition, it provides tools for extracting percentage volume contours (PVCs) from KDEs using the approach I developed for Supplementary Workbook Four, as well as tools to allow you to generate multiple 95% PVCs from KDEs from the same data set each with a different h value (also known as the smoothing parameter, bandwidth or search radius) to help with selecting the optimal value for this using the ad hoc approach from Kie 2013 (Animal Telemetry 1: 13), which you can download from here http://www.animalbiotelemetry.com/content/pdf/2050-3385-1-13.pdf, which is the approach I recommend for doing this.

 

If you are interested in using this toolbox, you can download it from: http://www.gisinecology.com/Home_Range_Tools.zip. There's a Readme file in the compressed folder which provides information about how to install it on your computer and how to import it into your version of ArcGIS. This toolbox is free to use, customise and distribute (although you will need the usual ArcGIS licences to use it), and is provided on an 'as is' basis. That means while it works for me, I can't guarantee that it'll work for you, or in every possible circumstance. Nor can I guarantee that it's completely error free.

 

Any feed back you have is welcome, while any questions about using it can be ask on this forum.

 

All the best,

 

Colin

 

[Anil Shrestha]

Dear Colin,

First of all many thx for providing custom toolbox for ArcGIS. I have added your tool box in ArcGIS 10.1. However, when I tried to run home range option (I have add xy data, projected in UTM, it gives me error message. Will you please suggest me whats might go wrong? Thank you so much in advance. Apart from this, I also wanted to know about your book An Introduction To Using GIS in Marine Biology: Supplementary Workbook Four - Investigating The Home Ranges Of Individual Animals. Is it also available in ebook format. Many thx.

Best

Anil

[GIS in Ecology]

Hi Anil,

 

Sorry to hear you are having problems with this. I should be able to help, but I'll need a bit of clarificaion first.

 

Can you confirm that you have installed the tool properly in the correct folder on your C drive (as instructed in the README file)? It should have the address C:/Home_Range_Tools/Home Range Tools.tbx?

 

Can you also confirm that you have gone into the GEOPROCESSING menu, selected GEOPROCESSING OPTIONS and that you have a tick beside OVERWRITE THE OUTPUT OF GEOPROCESSSING OPERATIONS?

 

If you have checked these and you are still having a problem, then you should get an error message in a pop up window. Select this entire error message (you may have to scroll up to select it all), copy it (by pressing the CTRL and C leys at the same time) and paste it into a Word document. Next, save this document as Home_Range_Tool_Error_Anil.Doc, and then when you reply to this message on this thread, attach the Word document with the error message to it. This will allow me to see exactly where things are going wrong, and so advise how to sort it out.

 

It will also help if I have an idea of which UTM you are using, what part of the world you are working in, and what species you're working with (okay, this last one isn't actually important, and is just for my own curiosity!).

 

In terms of Supplementary Workbook Four, unfortunately at the moment, it is only available in paperback and not as an ebook. This may change in the future, but probably not for at least 12 months. The best place to order it through is Amazon or (depending on where in the world you are based) Creatspace.com.

 

All the best,

 

Colin

[Cesar Munoz]

Hi Collin

I'm an undergraduate Biology student of Chile (at the Pontificia Universidad Católica), and i'm working measuring the home range of the Austral trush (Turdus falcklandii) at the urban matrix of Santiago (the capital of Chile).

Im usin the trial 60 days free versión of Arcgis, but it turns that because is the last version (10.3), there's no program that can work on it to make a Kernel (animal movement, etc..), except for your toolbox!.

I've tested your toolbox wihth the data date are includes as excersise (or examples), and the kernel wokrs perfect. But when i try to doit wiht mi own data, it pops up a error mesage and the kernel fails.

I instelled the program exactly like you said , and , like i told you, it work with the data that are included.

i'm attaching a word with the error mesage, and also a screenshot wiht the parameters of the kernel that i'm trying to do, and also the properties of the data and the atributte table.

I hope that you can helpme, i'm kind of stuk with this "simple" analysis for at least 2 weeks, and i need these data as soon as posible.

PS: sorry for mi bad english writting, i hope that you cand understand me .

Thanks

[Sarah Elmeligi]

Hi Colin,

Thanks for this post. It's actually the first I've found that is actually getting me close to understanding what I need to do... or at least what I'm trying to do.

I'm working in ArcGIS 10.2.1 and am having some issues creating Kernel Density Estimates for a grizzly bear population.

Using the Kernel Density tool in spatial analyst, I was able to create some kde raster layers for each individual grizzly bear's locations in a season. I read your post above about using the Isopleth tool in GME to then plot the 95% contour lines, but that's where I'm having trouble.

First of all, I couldn't actually use GME to develop my kde rasters and I don't really know why. The error message I was getting when I was trying that just said that the program couldn't understand my commands even though I've been using the boxes in the command screen.

So I created the kde's in ArcGIS, which is no big deal. When I try to use to the Isopleth tool in GME, I get this error message:

Error: The specified workspace does not exist. Cannot open raster dataset: C:\GPS\Input\GBLoc2013.gdb!GB128_F13kde2

Error: The raster data source has not been specified.

Error: The command text could not be interpreted. Please check the syntax of the command.

Error: An important error has occurred. Please include the information below if you submit a query about this error. Object reference not set to an instance of an object. I'm not sure what that's about. I've exported the raster data layer in ArcGIS and named it something else and I still get the same error message. I'm wondering if for some reason GME is not working with my computer, but I'm not sure why. I've got the most recent versions of all programs. So I'm not sure what to do.... Thoughts? Thanks so much Sarah

[GIS in Ecology]

Hi Sarah,

 

I'm glad you're finding this thread useful, and that it is helping you understand what you need to be doing. I'll do my best to answer your questions below.

>>Using the Kernel Density tool in spatial analyst, I was able to create some kde raster layers for each individual grizzly bear's locations in a season. I read your post above about using the Isopleth tool in GME to then plot the 95% contour lines, but that's where I'm having trouble.

 

A few people have reported having problems with this tool, so it's not just you that this is going wrong for

 

>>First of all, I couldn't actually use GME to develop my kde rasters and I don't really know why. The error message I was getting when I was trying that just said that the program couldn't understand my commands even though I've been using the boxes in the command screen.

 

My guess here would be that this is a projection/coordinate system issue. Before you can use GME to generate a kernel density estimate, you need to make sure that it has been transformed into the projection/coordinate system which you are going to use for your study. This projection/coordinate system has to be a projected one, and not a geographic one, because the map units have to be in metres (and not decimal degrees). So, the first thing I would do would be to plot your data in ArcGIS and then look at the projection/coordinate system of the data layer, and for your data frame. For the data layer, this can be done by right clicking on its name in the TABLE OF CONTENTS window and selecting PROPERTIES. In the LAYER PROPERTIES window, click on the SOURCE tab, and you should be able to find the projection/coordinate system information. For your data frame, right click on its name in the TABLE OF CONTENTS window and select PROPERTIES. In the DATA FRAME PROPERTIES window which opens, click on the COORDINATE SYSTEM tab and you can check it projection. If these are not the same, then you will need to the one for the data frame and/or transform the data layer into a different projection using the PROJECT tool. Once you have your data layer in an appropriate projection/coordinate system, it should hopefully work in GME.

 

>>So I created the kde's in ArcGIS, which is no big deal. When I try to use to the Isopleth tool in GME, I get this error message:

>>Error: The specified workspace does not exist. Cannot open raster dataset: C:\GPS\Input\GBLoc2013.gdb!GB128_F13kde2

>>Error: The raster data source has not been specified.

>>Error: The command text could not be interpreted. Please check the syntax of the command.

I think this error message is because you'v e saved your raster data layer in a Geodatabase rather than just in a normal folder on your computer. When you make the raster data layer using the KDE tool, you need to specify a specific folder and file name to save it to, and not save it in a Geodatabase (you can it something is being stored in a geodatabase because it as the extension .gdb somewhere in its file name/address). Alternatively, did you still have the raster data layer open in ArcGIS (or still have ArcGIS open)? This might have created a lock issue, meaning that you couldn't access the raster data layer in GME.

 

>>I'm wondering if for some reason GME is not working with my computer, but I'm not sure why. I've got the most recent versions of all programs. So I'm not sure what to do....

 

The quickest way to get you back on track would probably be for you to try using the Home Range Tools toolbox which I have out together for ArcGIS (http://www.gisinecology.com/Home_Range_Tools.zip). This allows you to do everything in ArcGIS itself and not have to use GME (which is a great bit of software when it is running, but it can sometimes be tricky to get installed and working properly).  The Home Range Tools toolbox has tools for creating KDEs, and for creating percentage volume contours (which is effectively what the ISOPLETH tool in GME does) based on the KDE estimate values for your original point data set.  However, you will still need to make sure that all your data and your data frame are using the same projection/coordinate system, that this is a projected and not a geographic coordinate system, and you will need to make sure you use appropriate values for the cell size, the search radius and the extent for your specific study area and species (i.e. it probably won't work with the default values).

 

I hope this helps, and good luck with your research project. Any further problems, or if the above doesn't help (or isn't clearly explained), just post back on this thread.

 

All the best,

 

Colin

 

[GIS in Ecology]

Hi Cesar,

 

I know that we have worked this problem out over email, but I wanted to add a reply here so that others can try the solution we worked out if they encounter a similar problem.

 

The issue here is the choice of paramters for the cell size, the extent and the search radius (also known as bandwidth or H value). The default values for the tool are based on the example data set for looking at the home range of an individual dolphin in the waters off northeast Scotland. When you are applying them to a different study system, you will need to change these (i.e. they will not work with the default values). For a fine scale study covering a small area, you might want to use a cell size of 10m or 20m, rather than the default, which is 1,000m. Similarly for the search radius cfor such studies, you might use a value of 30m or 50m rather than the default of 4,1667m. Similarly, if you were applying this tool to a much larger study area (e.g. a whole ocean or continent), you might want to select a cell size of 50,000m or 100,000m, and a search radius of 100,000m or 200,000m. You will also need to make sure that you use values for the extent which are suitable for the size and area covered by your study area. These will be given by the coordinates or the top left and bottom right corners of your study area in the map units of the projection/coordinate system you are using to calculate your honme range estimates.

 

So, in  summary, to get the kernel density estimate tool from the home range tools toolbox to work properly, you need to make sure that you have selected appropriate values for all the paramters, and that you don't try running it with the default values for your own data set.

 

As always, I hope this helps.

 

All the best,

 

Colin

[Sergio A. Balaguera-Reina]

Hi Jake

I am trying to download ArcMET but the webpage is not working, do you know other way to I can get it?

Thanks!

Sergio

On Sunday, May 11, 2014 at 2:03:41 PM UTC-5, Jake Wall wrote:

Hi Colin,

I've just released a version of ArcMET that works with ArcMap v 10.2 at www.movementecology.net. 

all the best,

Jake

On Tuesday, 1 April 2014 02:54:12 UTC-7, GIS in Ecology wrote:

Hi Jake,

 

Thanks for posting about the extension you've released. It looks like it will be very useful. Does it work with ArcGIS 10.2 as well or only 10.1?

 

All the best,

 

Colin

On Tuesday, April 1, 2014 2:03:00 AM UTC+1, Jake Wall wrote:

Hi Colin,

You might be interested in an extension I released recently for ArcGIS 10.1 that currently calculates MCP, KDE, a-LoCoH, LTD and BBMM home-range methods among other things. It's called ArcMET - Movement Ecology Tools for ArcGIS and is freely available at www.movementecology.net. 

all the best,

Jake

On Tuesday, 18 February 2014 02:47:13 UTC-8, GIS in Ecology wrote:

When I originally created this thread, it was to provide some basic tips about how to analyse home ranges using the tools in ArcGIS. However, it has since proved to be one of the most popular threads on the forum, suggesting this is a subject which a lot of people out there are struggling with. As a result, I have created a 'Supplementary Workbook' in my 'An Introduction To Using GIS In Marine Biology' series specifically devoted to analysing home ranges in a GIS-based environment, and using ArcGIS 10.2 software to illustrate the processes.

 

This 'Supplementary Workbook' builds on the information which can be found for free in this thread, and contains five exercises based around investigating home ranges in individual animals. These include: 1. Creating a minimum convex polygon (MCP); 2. Analysing home ranges in habitats without barriers using kernel density estimates; 3. Analysing home ranges in habitats with barriers using kernel density estimates; 4. calculating the overlap and habitat preferences of the home ranges of individual animals; 5. Working out how the home ranges of individual animals in a population overlap spatially. These exercises provide fully worked examples of each process starting with the raw data on an individual's distribution, and so will be of most use to those struggling to apply the free  basic outline instructions provided on this thread to their own projects.

 

One subject which it does not cover is how to select the most appropriate values for H/bandwidth, cell sizes and spatial extents for kernel density estimates (KDEs). This is because this is still a subject of some discussion amongst those who use KDEs, and it requires some detailed reading of the available literature before deciding on what values are best for any individual project.

 

While it is primarily aimed at marine biologists, exactly the same processes are used in terrestrial environments, and when using kernel density estimates to examine the distribution of species, populations etc.

 

The workbook is called: An Introduction To Using GIS In Marine Biology Supplementary Workbook Four - Analysing Home Ranges Of Individual Animals - RRP 19.99. ISBN: 978-0-9568974-5-9.

It can be purchased from Amazon (http://www.amazon.com/Introduction-Using-Marine-Biology-Supplementary/dp/0956897452/) or from GIS In Ecology (http://www.gisinecology.com/Book_Shop.htm).

More information about this book (including extracts) can be found here: http://www.gisinecology.com/marine_supplementary_book4.htm

All the best,

 

Colin 

[Jake Wall]

Hi Sergio,

Sorry about that! Looks like there was a glitch with the website but it's back up and running now. My apologies. 

all the best,

Jake

--
-- ======================================================================
You received this message because you are subscribed to the "GIS In Ecology Forum" discussion group (http://www.gisinecology.com/GIS_in_Ecology_forum.htm).
 
To control how often you get emails from this group, go to http://groups.google.com/group/gis-in-ecology-forum/subscribe (you will need to log on to get to this page).
 
To post to this group, either log onto the group's home page or send an email to
gis-in-eco...@googlegroups.com.
 
The rules for posting to this group can be found here: http://groups.google.com/group/gis-in-ecology-forum/browse_thread/thread/df31a0822742203f#.
 
To unsubscribe from this group, email:
gis-in-ecology-f...@googlegroups.com
 
All information on this forum is provided on an 'as is' basis. GIS In Ecology is not responsible for checking the accuracy or suitability of any posting or response.
======================================================================
---
You received this message because you are subscribed to a topic in the Google Groups "GIS In Ecology Forum" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gis-in-ecology-forum/k13R3rNpXts/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gis-in-ecology-f...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Dr. Jake Wall

Post-Doctoral Fellow

Fish, Wildlife and Conservation Biology

Colorado State University
CAN: +16047231224
KEN: +254700504553

www.movementecology.net

[Lachlan]

Hi Colin. Really useful forum and this thread is particularly interesting. Its taken me the whole day to read through it and there is a lot of interesting data.

Ive just ordered your books...volumes in the photo, might take a little while to reach Australia though. Hopefully I can ask a few questions before they arrive.

Will those volumes cover how to select an appropriate smoothing parameters, cell size and search radius? or is there another one I should order.

This sentence was of particular interest (as I gather i will be able to use the books to test out how to do a KUD with barrier etc)

"In all cases, when creating a KDE, you need to make sure that you select appropriate grid cell sizes, extents and search radius for your individual animals and species."

Could you please explain more about what these terms mean and how to do this? or point me in the direction of where to look. I see a lot of packages that explain everything in depth then say ' then choose appropriate parameters'... 

Im not sure I understand what the  h value is exactly (other than its in metres) and then how  grid size and extent relate to that value exactly (I had thought that extent and search radius were the same thing but the quote above seems to contradict that. 

In any case I have been playing around with KUDs to try and get a feel for what those parameters to.

To test those smoothing parameters I've used some data on 1 fish to do 95% KUDs using the href and lscv methods (and various different manually selected h values I dont include here) and there doesn't seem to be a huge difference in KUDs for href and lscv. Ive attached the example output which also includes MCP and detected positions , its a bit messy but hopefully gets the point across( href is the cleaner contour, MCP the largerst)...lscv and href appear fairly similar so what makes one better than the other? I mean the href looks 'nicer' but how does one objectively decide?

Thanks for your time and hopefully those questions arent too tedious.

Lachlan

ps: do you ever do courses in the southern hemisphere and also when do you release you course dates for the first half of 2016?

[GIS in Ecology]

HI Lachlan,

 

Thanks for your post, and hopefully you'll find the books useful once they arrive.

 

In terms of selecting a appropriate cell size and smoothing parameter (which is the same thing as a search radius), this isn't covered in the books. This is because they focus on the practical elements of doing home range analyses in GIS. In addition, there are, unfortunately, no hard and fast rules for selecting the most appropriate values. Similarly, there's no general agreement on the best approach for selecting a smoothing parameter.

 

However, here are the general bits of advice I give on my home range training course. Firstly, for cell size. You need to select a cell size that is small enough to capture all the features which influence your individual's distribution. For example, if you have a narrow channel linking two different areas of water that your individual can move between, then you need to select a cell size that is small enough to capture such channels. The basic idea here is that the cell size will be determined by the resolution of features in your study area rather than necessarily by any features of the individual's movement patterns. It may take several attempts to get the most appropriate cell size, but it is worth the effort to get it right. Interestingly, though, the cell size has relatively small impact on the estimated home range size (as long as all other parameters are kept constant).

 

The issue of selecting the most appropriate smoothing parameter is much more complex. The smoothing parameter is also known as h, the search radius or the bandwidth. What they do is set the distance over which the recorded presence of an individual influences the estimated occurrence in surrounding areas. There are many different suggested ways for selecting the most appropriate smoothing parameter, each of which may have their own benefits and limitations. The one I generally recommend is the ad hoc approach outlined in by Kie (2013), which can be found here: http://link.springer.com/article/10.1186/2050-3385-1-13. Really, the choice of approach you select is up to you, and will depend on things like the amount of data that you have, the tools you have to help you create home ranges, the approaches used by any previous studies you want to compare your results to, and so on, but the key thing is to make sure that you can justify your choice, that you understand any biases it might have, and that you use consistent approach across all of your data.

 

I'm sorry not to be able to provide a single clear answer on this, but it is not a straight-forward subject, and there is no general agreement on the best approach to use.

 

In terms of running courses in the Southern Hemisphere, we do run these occasionally, and hope to run more in the near future. If we do, we'll make sure to post the announcement here on the forum.

 

All the best,

 

Colin

[Lachlan]

Thanks Colin...I guess I was hoping you would have some magic answer to the what appears (at least from my reading and confirmed by what you say here) to be a common issue. 

Since I asked that question I have already done quite a bit more and I have already used the ad hoc method in the paper you cited (I found a reference to it on this forum) on  the 5 fish I am looking at currently and it seems to work fairly well, one thing that I found though is that increasing the href by 0.10 increments was required on two fish otherwise their homerange was fragmented and unrealistic. I notice that the paper decided to take href as the smoothing parameter if this occurred rather than increase the value. I suppose if I can justifty it based on the animals im using then it is ok. 

I think I am getting the hang of it, I know how to make fixed KUDs at any size and change the h estimator, grid size and extent. I don't think I really have  proper grasp of how the kernel is actually made though. This line from the paper you linked is like trying to decipher latin... "The starting point in kernel analyses is to construct a bivariate kernel estimate of a probability density function around each data point (animal location)" . Do you know any examples or tutorials where I can determine this on a very simple example by hand? I like to understand the whole process before relying on a program to do it for me.

Thanks again for your time, I look forward to using the books when they arrive and looking over more of the forum.

Lachlan

ps: I am still interested in courses next year over in Europe as I have a small amount funding to travel  overseas next year for a conference or research collaboration so could possible add in an extra week for a course like yours. I have recently atteneded an R course here and there is a spatial version coming up but its not aimed at movement studies and definitely not marine based so I might skip it for now.

[GIS in Ecology]

Hi Lachlan,

 

Yes, unfortunately there isn't a magic answer to this very common issue,

>>Since I asked that question I have already done quite a bit more and I have already used the ad hoc method in the paper you cited (I found a reference to it on this forum) on  the 5 fish I am looking at currently and it seems to work fairly well, one thing that I found though is that increasing the href by 0.10 increments was required on two fish otherwise their homerange was fragmented and unrealistic. I notice that the paper decided to take href as the smoothing parameter if this occurred rather than increase the value. I suppose if I can justifty it based on the animals im using then it is ok. 

 

Yes, really the critical thing is to be able to justify the values and the approach that you end up using. 

>>I think I am getting the hang of it, I know how to make fixed KUDs at any size and change the h estimator, grid size and extent. I don't think I really have  proper grasp of how the kernel is actually made though. This line from the paper you linked is like trying to decipher latin... "The starting point in kernel analyses is to construct a bivariate kernel estimate of a probability density function around each data point (animal location)" . Do you know any examples or tutorials where I can determine this on a very simple example by hand? I like to understand the whole process before relying on a program to do it for me.

 

I don't know of any tutorials, but if you think of a normal distribution centred on an individual grid cell, the height of this along the Z axis (the height above the XY surface) will be the recorded value for that central cell based on an actual record. The probability of animals occurring in surrounding cells is then estimated by the curve of the normal distribution flowing out in all direction. The bandwidth/search radius defines how quickly this distribution approaches a value of zero. The greater the bandwidth, the greater the distance this distribution takes to reach zero. This is done for every cell where an individual or species has been recorded, and the modelled values for each grid cell are summed to estimate the most likely usage of each grid cell. Thus, the greater the bandwidths, the more that the recorded occurrence in each grid cell is 'smeared' into those that surround it. I'm not too sure how clear this will be as an explanation, but it's how I generally try to explain the basic principle behind KDE to people.  

 

>>ps: I am still interested in courses next year over in Europe as I have a small amount funding to travel  overseas next year for a conference or research collaboration so could possible add in an extra week for a course like yours. I have recently atteneded an R course here and there is a spatial version coming up but its not aimed at movement studies and definitely not marine based so I might skip it for now.

 

I'll add your email address to the mailing list for future courses.

 

All the best,

 

Colin

[Lachlan]

I think that explanation of the KUD makes sense, hopefully am I understanding it correctly. When you say a recorded value what does this look like? In the case of animals isn't what value do you have other than it was there 1 and it wasn't 0? 

On a side note I might be getting a big confused by the different terminology in the  various programs I use. Do you know if I am I correct in saying that 

- Smoothing parameter/h/bandwidth/search radius are all the same thing?

- Grid size (zoatrack and adehabitatHR) is the same thing as processing extent in arcmap?

"Grid size info: The utilisation distribution is estimated in each pixel of this grid superimposed on the location fixes (detections) for each animal. The Extent (see below) is split into a specified number of intervals (= grid size). For example, setting this value to 100 will return a 100 x 100 grid. Default for this value = 60. Increasing this parameter can help create a smoother kernel surface and identify relatively compact areas of usage, however it will also increase processing time. See the adehabitatHR package notes for more information."

- Extent (zoatrack and adehabitatHR) is the same as cell size in arcmap?

"Extent info: This parameter controls the coverage of the grid (i.e. the polygon’s bounding box), where the minimum X coordinate of the bounding box is Xmin – Extent * difference between the minimum and maximum X coordinates of the relocations. Conversely the maximum Y coordinate of the grid is Ymax + Extent * difference between the minimum and maximum Y coordinates of the relocations. The default for this value = 1. If your kernel polygon has a flat edge, try increasing this parameter value. See the adehabitatHR package notes for more information".

thanks again for your time

Lachlan

[GIS in Ecology]

Hi Lachlan,

 

Apologies for taking so long to answer your post.

>>When you say a recorded value what does this look like? In the case of animals isn't what value do you have other than it was there 1 and it wasn't 0? 

 

By the recorded value of a cell when making kernel density estimate, I mean the number of occurrences in a specific cell. This might be 1 if you only recorded an individual or a species once in it, but if you have recorded more than one individual in a cell, this will be the total number of occurrences. Thus, it would be the total number of records you'd have in a cell based on your occurrence data (i.e. if you just converted your point data into a raster grid with the same extent and cell size as your KDE, this would be the values they would have). Does this make it any clearer?

 

>>On a side note I might be getting a big confused by the different terminology in the  various programs I use. Do you know if I am I correct in saying that  Smoothing parameter/h/bandwidth/search radius are all the same thing?

 

Yes, rather confusingly these are all terms of the same thing. Broadly these should be known as a smoothing parameter, but they are just what different people refer to it as. 

 

>>- Grid size (zoatrack and adehabitatHR) is the same thing as processing extent in arcmap?

 

These don't appear to be the same thing. The extent is the area covered by your KDE, and it should always be larger than the spread of all your data points. In addition, if you are wanting to compare KDEs calculated from different subsets of data (e.g. between individuals), these should all have exactly the same extent, so it's worth exploring all your data for all your data before you decide on what extent to use.

 

In contrast, given the information you provided, the grid size appears to be analogous to cell size in ArcMap (although they are setting the resolution of the resulting KDE in slightly different ways). However, I haven't used adehabitatHR, so I can't be certain of this.

 

>>- Extent (zoatrack and adehabitatHR) is the same as cell size in arcmap?

 

Again, given the information you provided, the extent info in adehabitatHR seems to be analogous to the processing extent in ArcMap, and sets hot the cell size, but total area that your grid will cover. However, be careful, because it is set in a different way between the two software packages, and this may cause problems if you try to compare KDEs made in different packages. In addition, it may be much harder to create KDEs with exactly the same extents in adehabitatHR  than in ArcMap due to the variations in the way they are set.

 

>>thanks again for your time.

 

No worries, always glad to be able to help when and where I can.

 

All the best,

 

Colin

 

[Greg]

Hello all, I am new to the group and have a qucik question about the Geospatial Modeling Environment, if anyone is familiar with it. I am trying to determine if the GME uses an adaptive or fixed kernel. I know that it uses a Gaussian kernel, but I do not know if that is adaptive or fixed. Any help or literature to help is appreciated!!

Greg
On Tuesday, July 30, 2013 at 6:33:59 PM UTC-4, kateb84 wrote:

Colin,

 

I realize this is an older post, but I have searched all over, and it is the only one that really seems to have any resemblence to an actual answer for me.

I am newer to GIS and especially ArcGIS10.  We're trying to establish a home range using kernel densities.  We'd like to incorporate the 50 and 90 (read it's less biased than 95%) PCV but are aware this isn't something we can do in the current version.  Is there something others have used in the past to do this? We also are curious as to how to incorporate boundries such as a river (could we draw in our boundries and use that as our environmental extent?).  I've played around a lot with the kernel density tool, but can't figure out these few things.  I also tried your suggestion above for boundries, but it didn't work at all for me.  What I'm looking for is a step by step to get me through this.  School just doesn't prepare you for these things. I appreciate the help

 

Thanks,

Kate

On Thursday, September 20, 2012 4:12:48 AM UTC-7, GIS in Ecology wrote:

Estimating Home Ranges Of Individual Animals In ArcGIS

 

It is often useful to be able to identify and compare the areas used by individual animals in a population. These areas are known as home ranges. There are two widely accepted ways to estimate an animal’s home range.  These are through a minimum convex polygon (MCP) or through a kernel density estimate (KDE).  An MCP is the smallest polygon that can be drawn that will encompass all the locations where an individual is recorded that contains no internal angles greater than 180 degrees.  It, therefore, provides a measure of every area where an individual has been recorded, but it provides no information about whether some areas are used more frequently than others. A KDE measures the density of records within each grid cell that covers a study area, and uses this to estimate the probability that an individual will use neighbouring cells.  It, therefore, provides an estimate of which areas an individual uses most frequently.  Typically, KDE are converted to percentage volume contours (PVCs) that identify areas where an individual is likely to occur 50% of the time (often referred to as the core range) and 95% of the time (often taken as a measure of the total home range).

 

There are two options for estimating home ranges within ArcGIS.  The first is to use ArcGIS tools directly. However, the tools that can be used to estimate home ranges are not easy to identify because of the names used for them.  To create an MCP, the tool that you would use is called MINIMUM BOUNDING GEOMETRY.  To use this to create an MCP, go to DATA MANAGEMENT TOOLS> FEATURES> MINIMUM BOUNDING GEOMETRY.  In the MINIMUM BOUNDING GEOMETRY window that will open, select CONVEX HULL for GEOMETRY TYPE (OPTIONAL), and this will allow you to create and MCP.

 

To create a KDE, you can use the KERNEL DENSITY tool.  This can be found in SPATIAL ANALYST TOOLS> DENSITY> KERNEL DENSITY.  The main limitation of this tool is that it assumes that there are no barriers to an individuals movements. These means that it is not well suited to creating KDEs in areas where such barriers exist.  For example, it would not be appropriate to use this tool to estimate the home range of individual bottlenose dolphins in a coastal population (as they cannot move across land). Instead, in such instances, you can use two other tools. First you can create a point density raster data layer using the tool POINT DENSITY (SPATIAL ANALYST TOOLS> DENSITY> POINT DENSITY. You can then convert this data layer to a point data layer and use a tool called KERNEL INTERPOLATION WITH BARRIERS (GEOSTATISTICAL ANALYSIS TOOLS> INTERPOLATION> KERNEL INTERPOLATION WITH BARRIERS) to create your KDE.  

 

In all cases, when creating a KDE, you need to make sure that you select appropriate grid cell sizes, extents and search radius for your individual animals and species.  

 

Alternatively, if you don’t wish to use these tools in ArcGIS directly, you can use one of several specialist extensions to create your home range estimates. These include Hawth’s tools and Home Range Tools for ArcGIS.  Unfortunately, at the moment most of these are designed to work with ArcGIS 9.3 or earlier and not the more recent ArcGIS 10. For ArcGIS 10 or later users, the only option that seems to be currently available is Geospatial Modelling Environment (GME).  It has tools that allow you to create MCPs and KDEs (although it does not allow you to create a KDE with barriers).

[GIS in Ecology]

Hi Greg,

 

As far as I know, both Geospatial Modelling Environment and ArcGIS use fixed kernels when creating kernel density estimates. In a fixed kernel, the same bandwidth width (or smoother parameter) is used across the whole study area, while with an adaptive kernel, the value of the bandwidth/smoothing parameter varies depending on some aspect of the distribution of the points used to create the kernels from (usually some measure of the density). You can find out a little more about this here: https://en.wikipedia.org/wiki/Variable_kernel_density_estimation.

 

I hope this helps. Any further questions, feel free to post again.

 

All the best,

 

Colin

[Greg]

Thanks Colin,

I assumed that based off the previous GME (Hawth's tools, which was Fixed) that the GME is also fixed, but wanted some verification. Reading the procedures on the GME website does not state either way. I appreciate your time.

Greg

[GIS in Ecology]

Hi Greg,

 

No worries, and glad to be able to help.

 

All the best,

 

Colin

[jls.sc...@gmail.com]

Hi,

I am trying to do a 95% fixed kernel estimate to get home home range size. I have a points layer in arc (one points later, with multiple animals). Can someone offer guidance on how to do that? When using the kernel density tool in arc, what is values should be used for "population field" , "output cell size" and/or "search radius" ? Thanks for the help!! 

[orlafro...@gmail.com]

Hi Colin,

I was wondering what your views on using a mask in the environments settings under Raster analysis within the KDE tool in ArcGIS 10.3? I looked into the KERNEL INTERPOLATION WITH BARRIERS tool  but the z value field threw me off. I was wondering if you knew does the mask in the basic KDE basically act like a barrier? Or how does one deal with the z-value filed (not an optional parameter), which is the height or magnitude value for unweighted points?

My data are seabird movement tracks. From reading these posts it makes sense to exclude all land masses as the birds do not feed on land. I have a vector (polygon) layer that is just water (i.e it excludes land), which I was using as my mask layer in the KDE tool. My output looks good but I was just wondering about whether it was more meaningful than if I hadn't done the mask in the first place?

You're feedback is much appreciated,

Orla

[GIS in Ecology]

Hi Orla,

Thanks for the post. You can simply mask a basic KDE to exclude areas of land, but in general it doesn't solve the problem of barriers to movements (but see below for thoughts on your specific case). With the basic KDE tool, the values are interpolated equally in all directions based on the setting used. If you have a very simple, straight coastline, then this might not be a problem because you could simply cut off any areas that flow onto the land. However, with more complex coastlines you run into the problem that the biological distance between two points (i.e. the distance the animal has to travel to get from one to the next) is not a straight line. 

For example, if you have animals feeding around a peninsula that is one km wide and 20 km long, and they don't like to, or can't, travel over land, then the straight line distance from a point on one side to the other will be 1km, and this is the distance that the basic KDE tool will use when interpolating values on either side of the peninsula. However, the biological distance between two points on either side of the peninsula will be up to 40km (as the animal has to move along one side it, around the end, and then back down the other. Since the basic KDE tool will not use this distance, it will potentially, substantially over-estimate the usage of the far side of the peninsula. Masking the output of the basic KDE tool will remove the values that overlap with the land but it won't alter the over-estimated values for the sea areas on the far side of he peninsula which have been created during the interpolation process to create the KDE in the first place.

Using the KDE with barriers approach, the KDE function fits its interpolation along the edges of the barrier, and so takes the biological distance into account. As a result, it doesn't create the same levels of over-estimations for areas that are in close proximity spatially, but that are separated by much larger biological distances. However, using the KDE with barriers tool is more complicated because you need to first create a density raster data layer of your data, and then use these density values as the input for the Z values in the KDE with barriers tool. That's the step you're currently missing, ad you can do it using the DENSITY tool, and then convert the resulting density raster into a point data layer with a suitable Z value for using in the KDE With Barriers tool. If you get stuck, you can find the whole work flow, and an example of how to do it, in my book 'An Introduction To Using GIS In Marine Biology: Supplementary Workbook Four'.

For your particular example, things might be a bit more tricky, because with birds, barriers such as the coast are potentially more fluid than for truly obligate marine species, such as cetaceans. Some seabird species may well be quite happy to fly across areas of land to get between foraging hot spots, in which case the biological distances will be straight line distances. This means that you can simply use the basic KDE tool, and mask out the areas of land to help you identify foraging hot spots. However, other species will be much less willing to cross areas of land, meaning that you'd have to use the 'with barriers' approach. 

Just to make things even more complicated, some species may well be okay with crossing shorter distances over land, but not longer ones, meaning that you would need to process your land shapefiles to make this into account in some way (probably by using the buffer tool to create a new shapefile which would incorporate this information into the 'with barriers' work flow. I haven't done this last part before, but if you'd like a basic outline of the steps you'd need to follow to create this type of barrier shapefile, just let me know.

So, the bottom line here is that you will need to know how willing your particular species is to cross over land areas in your particular study area to get between foraging hot spots before you can decide whether you can use the basic KDE tool and mask for areas of land, or whether you will have to use the KDE With Barriers tool instead.

As always, I hope this helps, and feel free to post back on this thread if you need any clarification on any of the above.

All the best,

Colin

[orlafro...@gmail.com]

Colin, thank you so much, your dedication to insightful and informative answers are very impressive and appreciated. This has given me a new perspective on my methodology which as you suggest is highly related to the species and location of the site.

It does help indeed!

Orla