Re: Help with OpenJUMP LoCoH Analysis (2)

[Stefan Steiniger]

Hey Jennifer, (I CC to the list, hope you don't mind)

it took much longer than expected to be able to revise the code and see
what actually happens.

However, first the final result with the option to calculate 10%, 30%,
50%, 70% and 90% isopleths is probably clear: so for the one bear
dataset that I have I get in the column "percent-cover-hull" actually
the real values (e.g. 12.08% instead of 10%, and 29.8% instead of 30%).
The 100% covers all points (usually only one polygon), while all 12.08%
polygons together (or all 29.8% polygons together) covers only a "core"
set of points, with the smaller value, i.e. the 12.08% or 30% instead of
100% referring to polygons that mark hotspots / cores ( = high density
of points).

So given that, the density for each hull polygon calculated (I for
instance used k=20 points) will set the number of points found in the
hull to the area of the hull polygon, or in the code:
density = numPts/(area / 1000000.0);

This means, the smaller the polygon area the higher the density if, as
in my example k=20 points (i.e. number of points is fixed). This also
means that the density must be larger than 0 and has no upper limit.

So, sorting the polygons by density will give you polygons where many
points are in a small space. That means, if the sampling rate stays
constant, then this is a hotspot or core area or much visited place.
However, if sampling rate is not constant (e.g. some devices sample
every 5 minutes and other every 30mins) then this means nothing.

In principle you can use LoCoH to achieve what you want, but I would
also ask: Why not using KDE, Line-KDE methods for that (or as comparison)?
I attach an image where I used LineKDE (bandwidth = 2000m , 200m
grid)and LoCoH-k with k=20 on grizzly bear data. You can see the LoCoH
polygons coloured; the more yellow the lower the "percent-cover-hull"
value (the more likely a core/hotspot area). In comparion the resulting
p=95% (brownish) and p=65% (blueish) percentil lines for the Line-KDE.
While the 95% percentil line (area) encompases most points, the 65%
encompases rather a core-set of points, or marks hotspot areas. The
LoCoH result is more obvious here, but if I set p=30% I should get more
pronounced hotspot areas. Well, of course, at the end there is probably
more parameters to set "correctly/meaningfull" than for LoCoH.

hope this helps,
stefan


Am 15.01.16 um 10:49 schrieb Jennifer Newton:
> Dear Dr. Steiniger,
>
> Thank you for the quick response. I appreciate you looking into the
> density value. It will be very helpful to me, as I am using this
> analysis to investigate where different groups of people go within a
> trail system for my dissertation. I hope to find the most
> visited/heavily used areas within the trail system for each group, and
> then compare them.
>
> I look forward to your response.
>
> Thank you!
>
> Jenn
>
> On Fri, Jan 15, 2016 at 6:45 AM, Stefan Steiniger <sst...@geo.uzh.ch
> <mailto:sst...@geo.uzh.ch>> wrote:
>
> Hi Jenn,
>
> thanks for your feedback. I shall get back to you in the next days,
> as i need to check what i programmed and don't have the sources with
> me at the moment. However, I don't think that there was/is an upper
> limit on the density value. As far as I recall the original paper
> does not talk about density and only calculates areas/ranges, but i
> implemented it to have something more comparable to the KDE approach.
>
> until later,
>
> stefan
>
> ------ Originalnachricht ------
> *Von: *Jennifer Newton__
> *Datum: *13.01.2016 11:08
> *An: *sst...@geo.uzh.ch <mailto:sst...@geo.uzh.ch>;
> *Betreff:*Help with OpenJUMP LoCoH Analysis
>
> Dear Dr. Steiniger,
>
> Thank you for creating the HoRAE toolbox for OpenJUMP. I have found
> it very helpful, however I currently need some help....
>
> I am doing a k-LoCoH analysis in OpenJump. The analysis is working,
> but I'm having trouble understanding the output. There is a
> "density" column in the output, and I'm not sure what units this is
> in, or what it indicates (the numbers range from a 3 digit number to
> 7 digit number).
>
> I know by highlighting some of the polygons that the higher the
> number in the density column the more used the area...but I still
> would like to know what this number actually means before I continue
> with my results. Any help you could provide is greatly appreciated!
> (I am conducting this analysis as part of my dissertation :) ).
>
> Thank you!
>
> Jenn
> --
> *Jennifer N. Newton*
> Doctoral Candidate
> Recreation, Park and Tourism Management and
> Human Dimensions of Natural Resources and the Environment
> The Pennsylvania State University
> 804 Ford Building
> Tel: (724)651-3069 <tel:%28724%29651-3069>
> Email: jnn...@psu.edu <mailto:jnn...@psu.edu>
> **
>
>
>
>
> --
> *Jennifer N. Newton*
> Doctoral Candidate
> Recreation, Park and Tourism Management and
> Human Dimensions of Natural Resources and the Environment
> The Pennsylvania State University
> 804 Ford Building
> Tel: (724)651-3069
> Email: jnn...@psu.edu <mailto:jnn...@psu.edu>
> **

[pimmelm...@gmail.com]

sometimes the Algorithm cant calculate these specific isopleths from the data. i guess thats all

[Stefan Steiniger]

Hi,

perhaps yes, perhaps no. It depends your type of data. If you can be a
bit more specific what you try to achieve and with what method?

But of course if there is no (sufficient) gradient between points, then
one can't calculate volumes or so.

best,
Stefan

El 7/18/17 a las 16:09, pimmelm...@gmail.com escribió:

[Marjorie Mattes]

Hi everyone,

Thank you for making this tool! I'm also currently working on making a LoCoH-k analysis on feeding areas and foraging areas of two types of bats.

First of all, I am a complete newbie in that regard. I started digging into the topic yesterday and read through the papers available concering this and tried to find a solution on how to create home ranges with the LoCoH method.

According to the instructions I received, the result should display the feeding areas (50%-LoCoH) and the foraging areas (95%-LoCoH), comparable to the screenshot below:

Right now, I have a shapefile containing the 147 locations (points) and I use the LoCoH tool with the k-method selected. From what I know, the parameter value can be calculated (heuristically) with the square root of the points (?) which would result in a parameter value of 12.12.

With this and the box to calculate the percentile isopleths checked, I get the following:

Unfortunately, it seems like the isopleths are not loaded. Is there a way to ensure that they are being displayed?

Also, how is there a way  to display the 50%- and 95% range comparable to the screenshot above?

I'm sorry that my question might not be precise enough. But if you need any further information please let me know!

Many thanks in advance for your help! It's greatly appreciated.

Best,

Marjorie

[Marjorie Mattes]

Hi there,

Thank you for creating the HORAE Toolbox for OpenJump.

I read through your manual and different papers concerning the LoCoh-Method and have to use it in order to determine core feeding areas (with  LoCoH 50%) and foraging areas (LoCoH 95%) of two kinds of bats.

I'm sorry that this is not precise, but I am a complete newbie concerning this topic and need to visualize this for a customer. 

I have copied an example screenshot from what it should look like:

Right now, I have a dataset of 147 point locations available and went on to use the LoCoH-K option for Home Ranges. Provided I got it right, the parameter value is (heuristically) calculated with the square root of these 147 points.

So, with the k-method, a parameter value of 12.12 and the checkbox for the percentile isopleths checked, I get somewhat the following:

Unfortunately, no layer with the percentage-isopleths was generated. Are my parameters wrong or what can I do to ensure that they are being generated?

Also, what is needed to get the 50%- and 95%-LoCoH values and to display the percentile range from 5-95 % ?

Help is greatly appreciated as this is the only tool I could find for this (usually using ArcGIS Pro or Arcmap 10.8, so couldn't use GME tool).

Many thanks in advance and let me know if you need any further information from my side in that regard.

Best,

Marjorie

Stefan - HoRAE Developer schrieb am Donnerstag, 20. Juli 2017 um 10:55:21 UTC+2:

[Stefan Steiniger]

Hi Marjorie,

I just want to confirm that I received your email - i was in vacation last week. But i also need to check back with my version as to my knowledge i did not implement ways to obtain isopleths (nor am i sure at the moment its possible) for LoCoH, given the approach. And instead tried to do something similar by counting densities. Its been a few (12?) years since i implemented it. I hope i can find time tomorrow or thursday to check it out.

However i assume results between methods should not be so different so as to (cross) validate them - and if looking at the results (pattern) makes sense from a behavioural point of view is another form to validate (but of course sometimes we always look for the exception that does not seem to make sense ;)

Best,

Stefan

PD. I don't mind if you send me one of the datasets to explore. If you have time stamps than other methods such as Brownian Bridges or Moving Kernels could be interesting too - for these there may be also R packages.

--
You received this message because you are subscribed to the Google Groups "OpenJUMP HoRAE Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ojhorae-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ojhorae-users/de9c5f17-fd99-478a-942b-88580db7bb16n%40googlegroups.com.

[Stefan Steiniger]

Hi Marjorie,

seems like I am wrong - there are isopleths of course (too long since I used LoCoH).

I have tried with another bat dataset using k-LoCoH, and it returned the isopleths in the layer that is produced first. I attach a screenshot. Note that there may be multiple polygons with the same value for the attribute "percent-cover-hull". That means for 50% you would have 3 core area polygons in my example. This attribute is also what you are looking for. However, as you see,  the software is not calculating the 95%, and also the exact percentages for the hulls, which means roughly 90%, 70%, 50% etc.

I hope this helps, 

Stefan