line transects in areas with lakes

[Kevin Hawkshaw]

Hi there, 

If you were conducting line transects for terrestrial birds in an area in with many small lakes and ponds, it seems possible for the distribution of birds to be non uniform with respect to the transect line, since on average the lake shores are going to be on average some distance away from the transects (assuming the observer is not walking through the lakes) and your terrestrial species cannot occupy the water. It's a bit more complicated and may not be as much an issue cause you may detect birds flying over the lake too, but it's possible that on average your bird locations could be biased towards the transect line since distance to a lake would be an upper cap on the distance you could detect a bird at. 

Could you diagnose this potential problem by fitting the detection function with a covariate for distance to the nearest lake? Expecting that for surveys closer to lakes that the detection function would be more left-shifted (detection distances smaller)? I dont think you could use this to correct counts, but just as an exploratory measure?

I hope I explained the potential problem well enough.

Thanks for your help,

Kevin

[Mark Wilson]

Hi Kevin,

I think it should be possible, at least in theory, to make a correction for the availability of land in different distance bands. Obviously, this is done routinely for point transects, for which availability increases linearly with distance (rather than being constant with respect to distance as it is for line transects), but making a similar correction for non-linear variations in availability might be trickier. You might be able to do this by adjusting sampling effort within different distance bands - I think this would mean categorising each distance band in each transect as a separate sample. I am not certain whether this would work. A more brute force correction would be to correct the number of observations in each band (per transect) according to the proportion of land 'missing' from that band. This would be a bit messy, particularly if needing to stick with an integer response, but could work. But other people are likely to know what the best way to handle this would be.

Potentially a trickier issue is that both occurrence and detectability of terrestrial birds may vary non-randomly with respect to proximity to waterbodies. If your lakes and ponds are surrounded by dense scrub, for example, this strip of lakeside vegetation may hold much denser concentrations (at least of scrub-loving species) than the rest of your land area, but birds in these strips may also be harder to detect. This becomes particularly problematic, from a distance analysis perspective, if the distribution of this strip is non-random with respect to distance from your transect (e.g. if your transects tend to run parallel to linear water features). If this isn't the case, then it might not be a problem for you (i.e. might be a similar situation to a transect through savanna dotted with scrub).

Cheers,

Mark

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To post to this group, send email to distance...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/89870e17-f92d-45b1-8755-94e273189f1f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Dr Mark Wilson

Research Ecologist

BTO Scotland

Unit 15

Beta Centre
Innovation Park

University of Stirling

FK9 4NF

Tel: 01786 458024

Registered Charity No SC039193 (Scotland) and No 216652 (England & Wales)
Company Limited by Guarantee No 357284 (England & Wales)

[Eric Rexstad]

Kevin

You are correct that it is difficult to detect non-uniform distribution of animals with respect to transects from distance sampling data alone.  Collection of ancillary data, as described by Marques et al. (2010, 2013).

I suspect your transects are a) not terribly numerous and/or b) not randomly placed (because you discard transects that cross water bodies, correct?).  You might overcome some of the non-uniform distribution difficulty by having perhaps shorter transects that intersect water bodies, like this

That creates more off-effort for field workers to get around water bodies, but may reduce some non-uniformity.  Increased number of transects, by sheer replication, should enhance the chances of conforming to the uniformity of animals w.r.t. transects design assumption.

Marques, T. A., S. T. Buckland, D. L. Borchers, D. Tosh, and R. A. McDonald. 2010. Point Transect Sampling Along Linear Features. Biometrics 66:1247–1255.

Marques, T. A., Buckland, S. T., Bispo, R., & Howland, B. (2013). Accounting for animal density gradients using independent information in distance sampling surveys. Statistical Methods & Applications, 22(1), 67-80. https://doi.org/10.1007/s10260-012-0223-2

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To post to this group, send email to distance...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/89870e17-f92d-45b1-8755-94e273189f1f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 
Eric Rexstad
Research Unit for Wildlife Population Assessment
Centre for Research into Ecological and Environmental Modelling
University of St. Andrews
St. Andrews Scotland KY16 9LZ
+44 (0)1334 461833
The University of St Andrews is a charity registered in Scotland : No SC013532

[Stephen Buckland]

Just adding to Eric’s comments, you would then exclude the water bodies from your study area.  And one more comment:  for walked line transects, you should not include birds in flight in your analysis.  The method assumes that the observer’s speed is appreciably greater than the average speed of the animals.  For flying birds, this is clearly not the case, and including them will give you density estimates that are biased high.

 

Steve Buckland

To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/f3f8370e-c9e7-127d-8e06-beee74e92397%40st-andrews.ac.uk.

[Kevin Hawkshaw]

Thanks to all three of you for your responses!

I'll try to flesh out my study a little more. My study area is quite large (~ 1500 km.sq land area), and i have ~ 450 1 km line transects. Habitat is very open - arctic tundra - so very little in the way of dense scrub that might lead to issues of detection that Mark mentions. There are no trees and only occasional small patches of shrubs that might get to 1 m in height if they're lucky. The issue is the water - the study area is full of it and in all different shapes and sizes. Ponds from ~30 m.sq or smaller, to lakes that are many km long. It looks something like Eric's diagram, except there are also some very large water bodies. I've attached a pic of the study design, note that each of those transects was repeated at least once within season, so there are more than double the number that appear in the picture. The start points for each transect are random, but in selecting ending points i made effort to pick a direction that would not have me walk through a lake (though starting or ending at a lake was fine), or walk a very convoluted path to avoid them. Nonetheless there are many transects that have a substantial amount of water on them (let's call that > 10% of the survey area as water). Again, like Eric's diagram there is a wide range of transect and water arrangements - some transects have no water at all, some start and/or end at a body of water, and others are adjacent to water at varying distances. I never traversed water. These are also pretty much comprehensive surveys of avian life, so there are species that cannot occupy the water (say, a songbird or a wader) and some that can (eg. a swan, goose or duck), intended to be modelled separately of course. 

I have run detection functions with proportion of water in the survey area as a covariate, and in no case did water influence detection, but I'm just not sure how valid that is for diagnosing this issue. perhaps i did have a large enough sample such that uniformity was more or less maintained. the only case that gave me some pause was ducks, as the detection function appears to have a bit lower than expected number of individuals detected close to the transect line (see attached), not unlike in Marquez et al 2013, when the kangaroos avoided the tracks the surveys were conducted from. Since i never walked through water, and the ducks were often found on the water, then it makes sense why there could be a lower number of birds detected near the line. 

To Stephen's comment - i did record whether the birds were in flight or what substrate they were on (including if they were on water), so they could easily be removed. Is this recommended to avoid double counting? Bird density is so low and habitat is so open that I can't imagine double counting is a serious issue in my case. 

On Tuesday, June 4, 2019 at 3:51:03 AM UTC-6, Stephen Buckland wrote:

Just adding to Eric’s comments, you would then exclude the water bodies from your study area.  And one more comment:  for walked line transects, you should not include birds in flight in your analysis.  The method assumes that the observer’s speed is appreciably greater than the average speed of the animals.  For flying birds, this is clearly not the case, and including them will give you density estimates that are biased high.

 

Steve Buckland

 

From: distance...@googlegroups.com <distance...@googlegroups.com> On Behalf Of Eric Rexstad
Sent: 04 June 2019 10:40
To: Kevin Hawkshaw <hawk...@ualberta.ca>; distance-sampling <distance...@googlegroups.com>
Subject: Re: [distance-sampling] line transects in areas with lakes

 

Kevin

You are correct that it is difficult to detect non-uniform distribution of animals with respect to transects from distance sampling data alone.  Collection of ancillary data, as described by Marques et al. (2010, 2013).

I suspect your transects are a) not terribly numerous and/or b) not randomly placed (because you discard transects that cross water bodies, correct?).  You might overcome some of the non-uniform distribution difficulty by having perhaps shorter transects that intersect water bodies, like this

That creates more off-effort for field workers to get around water bodies, but may reduce some non-uniformity.  Increased number of transects, by sheer replication, should enhance the chances of conforming to the uniformity of animals w.r.t. transects design assumption.

Marques, T. A., S. T. Buckland, D. L. Borchers, D. Tosh, and R. A. McDonald. 2010. Point Transect Sampling Along Linear Features. Biometrics 66:1247–1255.

Marques, T. A., Buckland, S. T., Bispo, R., & Howland, B. (2013). Accounting for animal density gradients using independent information in distance sampling surveys. Statistical Methods & Applications, 22(1), 67-80. https://doi.org/10.1007/s10260-012-0223-2

On 04/06/2019 04:33, Kevin Hawkshaw wrote:

Hi there, 

 

If you were conducting line transects for terrestrial birds in an area in with many small lakes and ponds, it seems possible for the distribution of birds to be non uniform with respect to the transect line, since on average the lake shores are going to be on average some distance away from the transects (assuming the observer is not walking through the lakes) and your terrestrial species cannot occupy the water. It's a bit more complicated and may not be as much an issue cause you may detect birds flying over the lake too, but it's possible that on average your bird locations could be biased towards the transect line since distance to a lake would be an upper cap on the distance you could detect a bird at. 

 

Could you diagnose this potential problem by fitting the detection function with a covariate for distance to the nearest lake? Expecting that for surveys closer to lakes that the detection function would be more left-shifted (detection distances smaller)? I dont think you could use this to correct counts, but just as an exploratory measure?

 

I hope I explained the potential problem well enough.

 

Thanks for your help,

Kevin

 

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.

To unsubscribe from this group and stop receiving emails from it, send an email to distance...@googlegroups.com.

To post to this group, send email to distance...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/89870e17-f92d-45b1-8755-94e273189f1f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 

Eric Rexstad

Research Unit for Wildlife Population Assessment

Centre for Research into Ecological and Environmental Modelling

University of St. Andrews

St. Andrews Scotland KY16 9LZ

+44 (0)1334 461833

The University of St Andrews is a charity registered in Scotland : No SC013532

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.

To unsubscribe from this group and stop receiving emails from it, send an email to distance...@googlegroups.com.