distance as a proxy for time

[ELR]

I have not come across any information in "the book" or on this forum on distance as a proxy for time so I'll go ahead and ask.
My study is on out-migrating salmon smolts in the Columbia River which takes place over a period of several months, as opposed to several years. The smolts are detected by a line of stationary acoustic receivers as they migrate down river and into the ocean. The receiver lines are not at equal distances within the river. Detection rate is high (near 100%), but varies from site to site.
The latest version of the manual was updated in January 2007 and on page 4-13 is a new sidebar on "Specifying and modeling uneven time-intervals between sampling occasion". Would it be wrong for me to use distance (scaled to one of the distance intervals) instead of time?

If anyone else is working on uni-directional migration past stationary receivers I would really like to compare notes.

Thanks for your time,
Erin

[Eric Janney]

Erin,

I have been involed in a study estimating and modeling coho smolt survival during outmigration in the Klamath River that sounds very similar to your study. Since you are measuring survival over distance intervals instead of time intervals it is appropriate to adjust the time intervals in MARK as a proxy to the distance between your acoustic receivers. This is important if you are interested in comparing survival of outmigrants in different sections of river.

[Eric Janney]

Depending on how close to 1.0 your detection probability is, you may be able to use a known fate model instead of Cormack-Jolly-Seber. The known fate has the advantage of not having to estimate p's. This will improve surival estimate precision.

[cooch]

Erin,

I have been involed in a study estimating and modeling coho smolt survival during outmigration in the Klamath River that sounds very similar to your study. Since you are measuring survival over distance intervals instead of time intervals it is appropriate to adjust the time intervals in MARK as a proxy to the distance between your acoustic receivers. This is important if you are interested in comparing survival of outmigrants in different sections of river.

If you're going to use time as a proxy for distance (with its inherent assumptions), it might be worth having a look at the sections of 'the book' which talk about how MARK handles time intervals. While its pretty simple, there are a few places where you can get into trouble if you're not careful. See especially the -sidebar- starting on p. 13 of Chapter 4.

[ELR]

Thanks for your replies.
Eric, I have very high detection efficiencies at a couple of lines, however some are closer to 70% so I still need to estimate p.
Evan, yes, these inherent assumptions are my concern...

I received an e-mail in response to my post and thought I would share with the forum in case it might be helpful to others:

Hi Erin, I've done this exact thing analyzing similar data in the
Connecticut River with Atlantic salmon smolts. If you don't standardize by
distance, you simply have suvival over the distance btw the receivers. If
you do, you get survival/distance. If there are big differences in the
distances btw the receivers, you'll get very different answers. This can
make a huge difference in interpretation if there are dams that decrease
survival (you can test for dam effects by using a dam/no-dam dummy
covariate). I'd be interested to see what you learn.

Cheers,
Erin