Hi Johana,

The design is what matters the most, as I see it in distance sampling. The issue of pseudo replicates is a big one, most of the time we are not able to tease apart pseudo replicates from actual replicates. This can easily be done by following a robust sampling and data collection design and sticking to it honestly.

It depends on the species of group of species you are focusing on. I believe stratifying your study area into equal or near equal sizes (grids are often robust) and placing the transects in random manner so that you cover most of the area of your grids. Once you design it well, say with two to three or n number of transects which are well placed inside a grid will give you spatial replicates. And when you do it again in another season or time, you can have temporal replicates as well. Following the same design will make your data robust and represent the actual field state of the animal in focus.

Hope this helps!

Hi Johana,

Your question was a bit vague, since that pseudoreplicates (i.e. lack of independence) can have multiple interpretations in a distance sampling context.

Strictly, the detection process should be independent for each observation, as the methods are based on maximum likelihood estimators. The methods are also known to be extremely robust to this independence assumption (see e.g. Buckland 2006. The Auk).

Regarding the design itself, we recommend something along the lines of what @Shah explained: place a systematic grid over your study area and put one transect in each grid square, or then just put parallel lines systematically placed over the entire area. This will give you good coverage of the study area. You should aim to have at least 12 transects, possibly best to have 20+. That will ensure robust estimation of the variance in encounter rate, which will typically be a large proportion of the overall variance.

An additional point regarding design is that lines should be put at random with respect to the animal population. Further, if you expect strong animal density gradients, place your lines perpendicular to that gradient. This will have as consequence that all lines will go through the high density and the low density areas, hence leading to lower variances compared to a situation where some lines would be over the low density area only and some areas over the high density area only.

Hope this helps

Tiago