You don't have to have random counts. What is random anyway? What if there are no birds at your random stations? If there is no bird present, you won't count one lol. Point counts, transects (strip), or just walking paths are good techniques. Or you can use Distance sampling techniques where you measure perpendicular distances from the transect to the object and use these in the analysis. Anything, so long as you actually count (or hear) a bird lol :)

Trails, walking paths or transects, along with fixed points are standard techniques. Obviously, in a tropical montane rainforest, you can’t expect to open a number of new paths, for nature conservation reasons first, but also for time and energy waste. If the study area is characterized by already present trails, use them. In such environments, small rivers and streams can be other useful walking paths, added to the extant trails. The only situation where it can be recommended to open new trails is when you realize (especially after some local people tell you) that you don’t have walking path crossing specific microhabitats. In this case, I believe you should add a trail to investigate these environments that can present potentially different species. Referring to montane forests, it is important to cover most of the altitudinal range of the study area.

The length of each transect or walking path can be reduced to 200 m, or either brought to more than 250 m, and the time spent at each fixed point can be of 10 minutes. After a first trial, if you realize that these standards do not allow you to contact so many birds, you can change the parameters until you find the conditions that fit the area that you are going to investigate, and then start the study.

I recommend to have a look to similar studies, e.g., those lead in the Tilaran range and Monteverde rainforest (Costa Rica), by Jankowski & Rabenold (2007) and by Gasner et al. (2010), along with the literature quoted.

Depending of the analyses you want to process, you do not always need to refer to such standardized transects. If you have a look to the study by Barrantes et al. (2016), you can have an idea of other potential methods that can be adopted.

Eventually I report a sentence, because I believe it might serve to relax before starting the study of birds in these environments. I encourage to read also the paper where the sentence is from (Boyle & Sigel, 2015), and literature quoted, even if it is related to another typology of tropical forest (La Selva, Costa Rica):

“Long-term data sets from tropical regions collected under more rigorous protocols are neither available from either La Selva nor from the vast majority of Neotropical study sites.”

Thanks Andrew and Matteo for your comments on my question.

Matteo, I looked for and got all the papers you recommended. I think they would be very helpful. THANKS  A LOT!!! Let me give you a little more detail. The study is focused on a specific bird, the Sumichrast´s Wren (https://goo.gl/8ya40C), an endemic and rare bird, whose geographical range is restricted to a small area in central-southeast Mexico.  Now, the goal is mainly to assess abundance (density) using point counts (with distance sampling) [by the way, this is another challenge because of the dense vegetation, but that´s another part of the story!]. However, a recurrent critic I’ve seen in the literature regarding this topic, is that most bird studies (regardless if they use standardized methods) is that they can suffer from bias due to non-random selection of localities.  In fact, I hardly have seen this specific topic being addressed in most papers.  Nevertheless, I found Green et al. (2010) in a study of kaka (a parrot) in the tropical forest of New Zealand used “131 point count stations [variable circular plots] distributed systematically, with random start point”; their study area was “relatively flat” tho.   In our study, the area is very mountainous, with a karstic landscape, which makes impossible to emulate Greene's sample design. Our area was formerly tropical rain forest, now it has been replace by shade coffee plantations (with some forest patches), which fortunately retains a lot of bird diversity. Anyway, as I originally posted, it seems that the only way to move around the area is using the trails that people use to do so, but then, this bring bias to our study because we are not locating the point counts randomly. I was wondering if somebody has experience with such a situation. At the end, I guess logistics and safety (fer-de-lance snakes are not uncommon) are above statistics orthodoxy. Regards!!!

Hi.

In my opinion, point counts must be situated proportionally in relation to the different types of vegetation and the land surface of each, as we have made in some islands of the Canaries in different occasions. This implies, naturally, to have first a detailed vegetation map of the study area.

Best regards, and good luck in your work. 

Thanks to Rubén and Emerciano for your comments; feedback is always helpful.

Hi again.

A further comment: 250 m of distance between points is too short, I think. It's better to place the points at minimum distances of about 500 m, because this may avoid possible double counts. In the Canary Islands, when we have worked with forest bird communities the minimun distance between points has been of 500 m, and sometimes even 700 m or more. It's true that distance usually depends on the specific characteristics of the terrain in which one works, as conditions are not the same in open, flat landscape compared with mountains or dense forest habitats. 

Most of the answers will not help you. You don't say why you need density estimates. It's nearly impossible to achieve good density estimates for a rare species in forested terrain. Indeed the number of detections required for good distance estimates is unlikely to be achieved. So, you need to ask yourself what that information would give you that cannot be provided by a good index method (Dawson, Studies in Avian Biology 6:12 & 554; Johnson, D.H. 2008. In defense of indices: the case of bird surveys. Journal of Wildlife Management 72:857-868). Point counts are vastly superior in forested terrain, except where there is such an easy walking path that undivided attention can be devoted to counting. Distance sampling assumptions will not be met in tropical forests, so all you'll get is a raw count adjusted by a very poor correction factor. This is the elephant in the room and concern over random sampling is dwarfed by it. A compromise is to use binomial distance sampling to collect the simplest distance information (near/far with a divide at say 25 m and a cut-off distance of 200 m to reduce uncertainty over remote birds). The best reference for this is Buckland, S.T. 1987 Biometrics 43:363, although it gets a brief mention in "Distance Sampling" (Buckland et al 1993). You would then be able to check for great differences in detectability between places whilst retaining an efficient counting method and probably employing an index. I also recommend five-minute counts, as this is a compromise between declining return with increasing time spent and the risk of double counting at a given point and the need to spend as much time counting and counting as many points as possible. Ten minute counts are too long for efficiency. This question is rarely addressed, and then not well, but see Dawson & Bull Notornis 22:101 for a comparison of line transects, 5-minute and 10-minute counts.  200 m spacing is about right for such habitats, as few species are so conspicuous as to be counted in significant numbers beyond about 150 m, and the only problem with double counting that the variance is underestimated somewhat. Given the enormous variation from time to time and place to place, such problems are insignificant. With a rare species, it's vital to have as much time as possible counting, rather than moving between counting points. This suggests using existing walking routes and the most efficient method possible. There's much work showing that sampling bias because of non-random placement of points matters little for index methodology (this was a hot topic the the UK Common Birds Census and studied in detail there), but surely collect information on habitat co-variates to allow this to be checked.

Hi again.

I don't agree with some of the suggestions made by David Dawson (especially in relation to the distance between points, as I suggest more space between points), but it's true that 5-minutes for point counts may be OK. In any case, in my own experience, if the study is centred in a single (or very few) endangered, rare or elusive species, then 5 min. may be not sufficient to detect them. When we studied the forest bird community in one of the islands of the Canaries we confirmed that an increase of time up to 10-minutes produced a significant increase of the number of species, and the problem to make double counts could be avoided, as the densities there were in general low, and the total number of species was limited also, both in laurel and Canary Pine forests. Naturally, in a continental, tropical environment, the number of potential species to be detected is high in comparison to an oceanic island's environment, like occurs in the case of the Macaronesian archipelagos. 

Hi!  Sorry to thank the answers a little late. Hectic times lately, down here.

David Dawson: Thanks a lot for all your comments and advices; they’ve been extremely helpful, and make me think twice regarding some issues about the study design.  I’ve reading the papers you mention (including yours) and I’m learning a lot about the topic. Quickly, why we want to assess density? Well, basically, the species has not really protection but the area we want to survey is a (rather small) privately protected area. So, we want to know the population size encompassed by this protected area for conservation purposes; on the other hand, most likely, the bird is a habitat specialist linked to a limestone outcrop landscape (because it uses caves and crevices as nesting places) so we want to modeling habitat as well.

Rubén Barone:  Also thanks a lot for your advice.  Thank you for your suggestion regarding the spacing of the sampling sites. However, I think we will go for 250 m for practical reasons. On the one hand, the study area is relatively small (450 ha), difficult to access and we do not believe there are many trails, so increasing spacing would reduce the number of plots we can do. On the other hand, there is a report of territory size for the species of 0.5 ha, so most likely these birds do not move that much.

Finally, I highly recommend, for all those interested in these issues, a very interesting and insightful paper regarding bird surveys: Thompson (2002) Toward reliable bird surveys: accounting for individuals present but not detected. The Auk 119(1):18-25.

Beware of Thompson (2002), as he comes from a school of thought that regards all index methods as useless. The methods he advocates are very inefficient and often the assumptions cannot be tested, so you end up with an index anyhow! That said, there's no perfect solution and the use of index methods requires care. Have a look at Douglas Johnson's "In defense of indices: the case of bird surveys" Journal of Wildlife Management 72 (4): 857. For your purposes you need as many counting points as possible: 250m spacing is too great.

Hi again.

It may be useful to read this classic book:

BIBBY, C.J., N.D. BURGESS & D.A. HILL (1992). Bird Census Techniques. Academic Press. London. 257 pp.

(There is a more recent edition of it).

We have used the book in question when designing methodology in the study of bird communities in the Canary Islands...

Best regards. 

There is a second edition of Bibby et al. dated 2000. Like most such works, it's more a list of recipies than a full guide to study design, but good for what it's worth.

its very difficult to obtain density estimate for many species in dense forest with haigh biodiversity. I think that is better to concentrate in good abundance index. Mist nettin my be useful for many understory birds. You can have also some surprise with species very difficult to observe

Thanks Giovanni:

You are quite right! During the counts, we've hardly seen the bird, tho the sing is clear and easily identified, and to "locate". We are trying to assess distances to our best but most likely we will end reporting with an index, as D. Dwason also adviced us. Athuogh no focus on the species, we are netting too but so far none has been trapped. The species is all the time around caves, cracks and holes that are common in the karstic environment, but where it is difficult to put the nets, not to say walking or exploring.

regards

Dear Joshe, under existing forest conditions, your option appears appropriate. Go ahead with intense field observations.

Best wishes

Thanks Hemant!!