Methodology for calculating behavior

[peterdesmet]

@ammouton, how do you want to calculate behavior? According to your flow chart this is based on:

• Number of receivers visited
• Direction
• Duration

  New behavior method

We would like to see if we can calculate the behavior for each record, without geospatial information. As a first step, we would like to figure out moving vs immobile, then type of movement.

This could be done based on:

• Distance travelled
• Number of receivers visited within a timeframe
• Number of receivers visited in previous/next occurrences
• A combination of those.

What do you think is the best approach? You can elaborate in a comment.

[ammouton]

Hi Peter, good idea: if we could avoid the additional spatial input, this would be interesting. but I should check more thoroughly, looks like you need more user-defined parameters for that approach? Some examples could be useful; e.g. what about a junction with two branches; I would define "consecutive receivers" as consecutive with respect to the flow direction; which will be hard if you only consider number of receivers(nor) /nor visited in prev/next occurs?

[peterdesmet]

1. So to recognize migration we need to have some knowledge of the network of receivers, how these are connected (consecutive receivers). We could either get that information automatically from their geospatial position (with some caveats) or by ordering them manually (more user-input).
2. Is there a time-component required to distinguish between homing and searching? E.g. if an eel is moving between 3 receivers, is there a duration at which you would say: this is homing, not searching?

I think we need an updated flowchart.

[bartaelterman]

Let's start with homing first. If I remember correct, this is only dependent on the number of receivers that detected the eel. Let's explore some example data.

Consider the following table:

record number	time	receiver
1	10:00	A
2	10:02	A
3	10: 10	A
4	11:45	A
5	12:01	B
6	12:05	B
7	12:07	B
8	12:10	A
9	12:13	A
10	12:15	C
11	12:17	A

Which records would you consider homing?

[ammouton]

For homing, I think it could be easy: there are several algorithms for calculating home ranges. We could calculate these home ranges, and see whether a given detection is at a receiver within the home range or not.

One problem is: how to decide which calculation of the home range is the best. This could be selected by the user after calculation of the home range (= number of receivers within the home range)?

[bartaelterman]

So you cannot say anything about the behavior (homing or not) without information on the distances between the receivers?

What algorithm do you currently use then? And what input parameters does it need?