UKH

How do peeps.

As a part of my dissertation, I've the GPS records of many swift nest in buildings. I'm thinking that finding the average distance between nest sites, as well as the closest and the further distances might be a helpful addition.

Does anybody know if there is any kind of programme available with which one could enter what is essential a 'spiders web' or cluster of GPS coordinates, and find the above kinds of answers?

I'm currently in a lecture so may be slow at answering any 'Do you mean X?' questions.

Currently I have a student online access subscription to ARC GIS provided by ESRI, so could upgrade that if they provide that facility.

Many thanks. 

Excel?

Depending on what version of Arc you are using the near tool might work (combined with avregeing the resultant distance feild).

I assume you mean the average nearest distance between points.

Alternatively you could mean:

Nest A, average of distances between Nest B, C, D, E, F, G.....

and then 

Nest B, average of distances between Nest A, C, D, E, F, G.....

Average all of above. I assume this isn't what you are looking for.

If just looking for average nearest distances between points, then I would start by creating a Delaunay triangulation.

Arc can do this. Usually used for creating surfaces.

Then just export the triangles and average the triangle side lengths.

No idea what this means. closest and the further distances from what? A point you specify. If so excel can do this.

As the crow flies?

Seriously though is what you want a bit like of the pictures here? https://gis.stackexchange.com/questions/245524/measuring-distance-between-p...

QGIS is free but I reckon most things in QGIS will be doable in ArcGIS maybe with similar terminology.

I think you need to clarify what you're thinking of there.

A nest will have many neighbours (the entire set of mapped nests, at the limit case). Which nearby nests will you consider as neighbours?

Or is your intent to find the closest nest for each nest, and find the average of those values?

Converting GPS lat/long to linear coordinates isn't too hard, if you use a geoid model, and make some trivial approximations, and accept minor errors that are probably less than the recording errors. If your GPS logger included OSGB coordinates, it would remove the need to do that.

So, ypu first need to find linear positions of each nest. Then find the distances between each nest and the rest (n^2/2 calculations required), then find minimum distance for each nest. Then do whatever stats you like.

It might also be instructive to look at nest density, and factors affecting that; presumably resource-bound (food, building materials).

I can't speak to ARC or other GIS. I've just used AWK for parsing GPX files and mucking about with tracks, etc... For instance, a plot of a day's skiing, with track width scaled to velocity. Spot the GNSS anomoly...

It's the average of them collective when seen as a group, the average of all the distance measurements considered when they interrelate.

What is 'them'? Nearest neighbour?

I can imagine a decent GIS should have all sorts of tools for analysing relationships between things in a geographical context. About time I started to learn to use one...

Nearest nests.

If you speak 'R' which is free then this code (Googled) does it

library(geosphere)

distm (c(lon1, lat1), c(lon2, lat2), fun = distHaversine)

Easy enough to create an array of all possible combinations then apply this function over that array.

Nest, or nests? For each nest, that is? I might have to think a bit to come up with a way of automatically determine which of the surrounding nests are 'nearest'. Do they naturally come up with a hexagonal distribution?

If for example there were 30 nests, it'd be the average of each distance measured in a straight line between each pair of nests, with nests 1 2 and 3 making up two pairs. I hope that makes sense.

Don't take too much brain power away from your day job however.   

it sounds to me like you want a Delaunay triangulation then.

Thanks, that looks like it fits exactly. 

If there are 30 nests, then there are 29 pairing nests for each nest. Are you intending to measure the average distance for those 29 pairs? What it you had 1000 nests? The average starts to be meaningless. I think you are more concerned about nearest neighbours, not those nests on the other side of the 1000 sample area, because they will have little effect on the siting of a nest; no competing for resources.

Finding the single closest neighbour is easy. Deciding which of the sample are the 'closest neighbours' is much harder, I think.

Naturally, given uniform resources, they ought to fall into a hexagonal array.

I think that looks something like what I was envisioning as a means of determining nearest neighbours; the triangulation acts as a 'fence' to determine if another point is a closer neighbour or not.

Glad someone knows about this...

I was thinking exactly the same re the average of all the nests being meaningless, also I presume there is a study are boundary where nests outside aren't included.  Therefore the average for all the nests will be controlled by the size of your study area as much as any 'real' factors regarding the behaviour of swifts. 

Fortunately, assuming that Delaunay triangulation finds the nearest neighbours, as it appears to be doing, then the analysis can proceed to look at average distance to nearest neighbours, using the network 'connectivity' identified.

Which version of ArcGis do you need for being able to do that please, or which one is used in your case?

I used Civil3D. I would expect any version of ArcGIS to be able to create a TIN as it's a pretty basic function in GIS.

https://desktop.arcgis.com/en/arcmap/latest/manage-data/tin/fundamentals-of...

I'm sure whoever wrote that help page thought it was helpful, but...

I guess there's a whole bunch of precursor stuff that you need to understand first.

The description of the circumscribed points is pretty much the same as the wiki entry. Neither explains why this circumscription is useful/significant; maybe it's obvious to the more (geometrical) mathematically minded.

I only provided the link to show ArcGIS could do it.

As for the content, I think it's just trying to say that the triangles are laid out to avoid skinny triangles, and overlapping triangles.

Really easy to do manually, but mathematically quite complicated.

Hopefully my little sketch below shows how it can be done right (on the left) or not so right (on the right)

The previous page gives a bit more context

https://desktop.arcgis.com/en/arcmap/latest/manage-data/tin/fundamentals-of...

In ArcMap just use the Near Tool. If you have lets say 10 nest sites then run it 10 times. If you want I can run it for you as I use ArcGIS everyday. You can use a plug-in called Spider Diagram Geoprocessing Tool to make lines from each point.

was just going to post this which also mentions Near Tool..

https://gis.stackexchange.com/questions/259265/calculating-average-distance...

also wondering how useful this data would be..

Yeah either GIS will do that long post for you. Others have covered it in this thread. 

It's been found that even where swifts are being predated upon by other birds, the can still nest in holes which are close together, which suggests there's a benefit in them doing so, potentially in them passing on where to find food to one another, which may mean that a reduction in  building spaces for them to nest in is detrimental in this sense, less site density being unhelpful, as well as in the 'face value' sense of there being fewer places to nest.

If this combines with a reduction in available insects to eat as well, that's all kinds of bad news for swifts.

Thanks, I'll email you the nest grid references today.

ahh, didn't realise you were measuring nest sites within a nest site?

Nest sites within an area, the study I read about them being predated upon looked into the proximity of holes to one another, and compared where they were being predated upon with where they weren't. It's less of a central focus, and more something extra to talk about, and something to mention as having potential for further research.

how close do swifts nest to each other? maybe nest density would be a better measure?

'All swifts are different' seems to be a theme, but from half a metre to one and a half away from one another. 

have you heard of correspondence analysis?

youtube.com/watch?v=raPTZtjyNos&

good gps

Indeed. Way beyond the resolving capability of a consumer handheld receiver. You'd need a centimetric receiver and/or reference station for that sort of accuracy.

I assumed the nests were tens of metres apart.

This is from other research. Given that the buildings/nest I'm taking records (secondary data) from are a street's width apart, or more, it's another 'element to ponder'.