Problems with home range

[Former user]

Hi Shawn,

I have been trying to do home range analysis and have successfully got to the point of creating the residence time raster and converting it to a point file. I know my river raster is working, because the residence time raster follows the river between points rather than directly across land. However when I run the home range analysis the 50th pcentile and 90th pcentile rasters are perfectly circular - even though they do not display in the land areas. How do I force this stage to recognise my cost raster?

Thanks in advance Kate

[Shawn Laffan]

Hello Kate,

Could you please check that the path length raster is being used in the home range tool?

One way of producing circular patterns with a kernel density analysis, from which the percentiles are calculated, is to give it widely separated points. If it is passed the points along the path then it should roughly follow the path.

One other possibility is that the path length points nearly all have a value of zero. This should not happen, but that does not mean it cannot happen. Could you also check the attribute table of the path points?

Regards, Shawn.

[Kate Buckley]

I am definitely using the path points raster for the tool. The grid_code values in the attributes table from the path points raster are all small, eg. 0.003395, 0.004801. I have been manually tracking fish in a river with 15minute intervals, for 3 days; so the distances and time intervals are small. Is this a problem?

[Shawn Laffan]

Thanks Kate,

That information helps considerably.

The issue is in the percentile calculations. It converts the input raster to integer, using a default multiplier of 100. This is too small for your input data as anything less than 0.01 will be converted to zero.

The work-around for now is to multiply the path length raster by some constant factor such as 100 before converting it to point and feeding it into the home range tool. The one issue is that the resultant KDE raster will then need to be divided by 100 to get it back to the correct units (if you need it).

Please let me know if this works or not.

Regards, Shawn.

[Kate Buckley]

Wonderful, I will try this, thankyou!

[Kate Buckley]

Shawn, my KDE raster is still circular! Sorry.

[Shawn Laffan]

Hmmm. Would you be able to send me a copy of the data?

[Kate Buckley]

Shawn, this is a tiny subset of my data and the rough test polygon I was working with to ensure that everything was operational! I am using ArcMap10.2.2, and have the dataframe cooordinate system set to GCS_WGS1984. Thanks very much for looking at this Cheers Kate

[Shawn Laffan]

Thanks Kate. I'll have a look and get back to you.

Regards, Shawn.

[Kate Buckley]

Sorry Shawn, I sent the wrong polygon!!

[Shawn Laffan]

No worries. The boundaries aren't the source of this problem so a square works just as well.

What cell size are you using?

[Shawn Laffan]

Another question now I'm running through it, what radius are you using for the KDE analysis? The system does not correct for coordinate systems. For data in decimal degrees the default value of 150 will very nearly wrap around the planet, and also produces a circular pattern when I run it with your data.

One approach is to use a KDE radius in decimal degrees. The other approach is to project all your data into a projected coordinate system, possibly UTM (MGA) if they fall within one zone. Then all your distances will be in metres and the original problem with the percentiles should also disappear.

Regards, Shawn.

[Kate Buckley]

I have tried heaps of things, and the best results to date have been with a cell size of 0.00005 and the KDE radius of 5.....

[Shawn Laffan]

Thanks. 5 is still to large, as it is a radius of 5 degrees. If decimal degrees are easier to use than a projected coordinate system then try 15 times the cell size as a starting point. You would still need to work out how large that is in metres. Of course a value of 150 m is arbitrary, so has no more basis than 146.7, so the best approach is to use something sensible relative to the water body.

Decimal degrees do have the issue that the lines of meridian converge towards the poles, but that effect will be quite small for a study area of less than one degree in latitude.

Regards, Shawn.

[Kate Buckley]

Hi Shawn, I finally got it happening with a KDE radius of 0.00075 and by multiplying the path length raster by 100... I tried to project the data into UTM (MGA) Zone 52 but couldn't figure it out (the data and polygon wouldn't match up)... Anyway, I think I have it figured out enough to use- thankyou very much for your help! Cheers Kate

[Shawn Laffan]

Good to hear you got it working.

I've listed issue #17 to work on the percentile precision multiplier and will close this issue soon.

Regards, Shawn.

[Shawn Laffan]

Resolving issue.