I think line transect surveys would help minimize double counting.  Basically, the observers walk along a predetermined route and record birds on either side of them and estimate the distance and direction of each bird.  The link below has a little more information.

http://www.ebcc.info/wpimages/other/birdsurvey.pdf

Thank you   Coby Klein, for your answer and additional references. Basically, i am following the method of line transect positioning, which may minimize double counting and provides unbiased estimate as  compared to point count method, but still there are gaps with use of  line transect survey, within similar habitat types and small study areas.

Depending on the spatial scale of your study, synchronized observations may minimize this kind of bias. If there are more observers, they should visit prefixed observation points at the same time, and they should record the time of the detections precisely.

Best regards

I think that the main bias is the lack of detectability rather than double count.Distance sampling will correct it. You can look this paper, it is mentionned that double-counting may have little impact : Landbird counting techniques: current practices and an alternative. SS Rosenstock, DR Anderson, KM Giesen, T Leukering… - The Auk, 2002 - BioOne

The authors refer to Buckland (1993)

BUCKLAND, S. T., D. R. ANDERSON, K. P. BURNHAM, AND J. L. LAAKE. 1993. Distance Sampling: Estimating Abundance of Biological Populations.

Chapman and Hall, New York.

But you might find interesting answer in its updated version of 2001 (Distance Sampling).

There's no survey method that is absolutely free of errors. You can hardly avoid some few cases of double counting in a line transect survey method. But using programs like DISTANCE and other stats methods can help explain how far you are from being correct. I would say you just have to keep sampling using the assumptions of line transect method. You can look at "Bird Census and survey techniques by Bibby et al., 1999"                                                                                                                            

Bibby is a good reference.

The  method you need depends on your research question.

There are problems with all methods, e.g. see

Barry, S.C. and Welsh, A.H. (2001) Distance sampling methodology. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 63, 31-53.

Welsh, A.H. (2002) Incomplete detection in enumeration surveys: whither distance sampling? The Knibbs lecture for 2000. Australian and New Zealand Journal of Statistics 44, 13-22.

Welsh, Alan H., Lindenmayer, David B. and Donnelly, Christine F. (2013) Fitting and Interpreting Occupancy Models. PLoS ONE 8, e52015

Birds are particularly difficult. Line transects for birds has the problem of behavioral differences between species affecting flushing distance. Flighty birds might flush before you get there and so you don't detect all on the line. 

If you have a randomised sampling design for point counts (which is just a modified Distance method) then counting the same bird twice at different points is less of a problem than violation of some of the other assumptions of Distance, e.g. flushing and inaccurate direction and distance measurement.

One might assume that the probability that a bird you counted at site a) moves to site b ) is equivalent to the probability that a bird at site b when you were at site a) moved to site a) before you got to site b), and they cancel out. If that assumption holds, then counting twice does not matter.  However, simultaneous counting overcomes that to a degree, but imposes observer bias, which can be measured using blind Tandem/ double counting.

The Distance estimator is limited by the number of observations. Depending on who you read,  a minimum of 60  observations is required, which is a problem for rare species. Saturation and inaccuracy of enumeration becomes a problem for abundant flocking species like European starlings. 

 Perhaps Occupancy modelling

MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.L. and Hines, J.E. (2006) Occupancy estimation and modelling: Inferring patterns and dynamics of species occurrence. (Academic Press (Elsevier): London)

might answer your dispersion question.

Really thank you Peter  Fleming, for your detail and clear explanations.

Beside the ideas expained above, what I often do to minimeze biases is to avoid always starting the transect on the same point. If I go from point A to B in one day, the next I go from B to A. If this is possible it makes sure that errors caused by bird movements will not always be in the same direction. This way, when you combine the data from multiple transects the biases should be minimal.

Hi Getachew,

Peter F covers most points nicely.

Final method selection will also depend on the nature of the landscape you are working in and spatial scale you wish to assign bird records to. If you are surveying a heterogeneous landscape and trying to assign bird records to different habitats/vegetation types within that landscape, it can often be easier to survey using point counts rather than line transects - bird records can be assigned to the specific vegetation surrounding a point. How you choose to describe or quantify the vegetation is a different matter! You may want to consider systematic random placement of points - you can stratify the sampling effort (points) based on different habitats (strata) to assign bird records to these habitats, while it can also help reduce the variance associated with your abundance estimates.

I would recommend employing Distance sampling point transects (also called

variable circular plot sampling) - as Peter F mentions, providing any (unknown) double counting of birds at different points is random then this should not be major a problem. To help reduce any bias associated with bird records not being independent at a given point, you should use a count period that is short enough to detect all birds on or near the point. I've attached a paper we did on distance sampling count periods for surveying forest birds (also at http://onlinelibrary.wiley.com/doi/10.1111/j.1474-919X.2007.00790.x/pdf).

You could also think about using a cue count method at points where you count the number of cues - bird songs - rather than birds themselves. You can use Distance sampling to calculate the density of cues, and then species density by dividing the cue density by an estimate of cue rate (song rate, which must be obtained separately to the point transect surveys).

Employing occupancy methods could be used to give you a measure of habitat/site occupancy, which could be monitored as an index of bird abundance.

I hope this is of some help. Good luck.

Article Adjusting count period strategies to improve the accuracy of...

In some European countries standardised method of research of birds – the common birds census – is based on the so called „point transect method“(Svensson 1975, Janda, Šťastný 1984, Šťastný et al., 1989, Kropil 1995, 1998). The bird count is made on 20 points, at each point for 5 minutes. Distance of each points is 300 m at minimum. When moving from 1 point to another point no birds are counted. This method can minimise multiple counts of the same birds, especially when we suppose that most songbirds are territorial and size of their territories do not exceed 300 m.

Thank you Samuel P. for your experience sharing.

Hi, I think double counts are inevitable in some cases. But I would suggest you to walk pseudo transects or point transects to minimize the errors as done for the cases of grassland birds and large mammals. When this data is inserted in Program Distance, it basically gives you a pretty gud result depending on detection probability rather than mere numbers only.

Thank you Zaara Kidwai, for your scientific information.

As already suggested by Gergely Schally, it is better to have synchronized survey involving multiple set of observers. In this technique, you must also take care of observer’s capability.

Also just an idea: if You are counting resident bird community of breeding species of birds, they have certain territories and within breeding season mostly do not move away from its territory...If You know size of their territories, it is easy to avoid double or multiple counts. This is also basis of the idea of skipping 300 m between individual points of the point transect.

In all relevant articles on bird survey. I have noticed that most authors/surveyors have maintained atleast a distance of 300 meters from plots/transects. So I think you should not place point count stations closer than 300 meters.