Hello Tanvir,

The stratification of the study area is important. I believe your team is surveying different forest patches which are different in size from one another. As for replicates, the more the better but please keep an eye for lack of independence, i. e. they should not be pseudo-replicates (this could be temporal or spatial). The replicate surveys should be independent from each other. More replicates would help you to come up with variation and nestedness. The optimal number of replicates will depend on the sample site size you have.

Thanks a lot Shah Nawaz Jelil sir for your kind answer. Yes, we are surveying heterogeneous habitats of different sizes. Heterogeneous in such a way of different vegetation types (e.g. Bamboo plantation, semi-deciduous forests, different plantation), indigenous community habitations and agriculture practice into the forests.

I have 2 related question to know.

1. How can i figure out the optimal replication numbers for a forest?

2. I have confusion in one thing is that if local tourist's guide helps to find out groups based on their knowledge of occurrence any group in any location of the forest, doesn't that denote dependent?

With best regards

Tanvir

Hi Tanvir,

1. One way optimal replicate numbers can be achieved is through simple plots where you plot cumulative number of species in every replicate, until you reach a stable number of species, and species number don't fluctuate much. You may also look at the concept of 'power' in field surveys.

If you follow a grided survey, it's better or well distanced transects for spatial replicates , as for temporal replicates, you could do it in different seasons or different months.

2. I don't think that's a good approach, transects should be placed at random and not previous knowledge of the area. Randomisation would be compromised if you follow people's direction.

Lastly, do you intend to compare these different patches or are they just different sites for a wholesome study area. Please mention the size.

Hi Tanvir,

In distance sampling, what matters is the number of independent transects you cover (and then we usually say 12 is a minimum, but more than 20 is great). Variance in encounter rate comes from the variance across transects, that is why a reasonable number of transects is required. You actually do not need any "replication" (in this case I am assuming you are referring to covering the same transects several times).

But then you need that from these you get enough detections to estimate a detection function. That means about 60-80 detections (groups if you are doing a group based analysis) on a line transect survey. Note strictly that will be the number of detections per estimate required, so if you wanted estimates in two different habitats separately, you would want that minimum number of detections in each stratum. If you have a covariate that explains detectability then you can get away with less observation in each stratum say, if you make detectability a function of the covariates and then make your estimates conditional on the covariates observed in each stratum.

Repeating the same transect many times can be helpful if you do not have enough observations in a single passage to estimate the detection function, but then you need to care about independence issues. Imagine you have a transect with 10 km. Effort should be recorded as 10 km if you only cover it once. If you cover it K times, say 20, then you could record the effort as 200 km and that would account for the non-independence.

Thanks a lot Shah Nawaz Jelil and Tiago A. Marques Sir for yours helps. I got my answers.