Sounds like a great idea. Observer is now a variable in your model. There is an art to survey sampling (line transect sampling). Some people are very accurate. Others may tend to miss individuals or count individuals multiple times. If you are very unlucky they do both with equal abandon. You can now correct some of the data for this effect.

Alternatively, you mean that you have two people writing counts on the same data sheet. As they walk together they may correct each other and end up with fewer errors overall. Or they partition the field and because they ignore the other person you get more double counts. The bird flew from your side to my side and we both counted it. You might get more errors because one person misses an animal because their view is blocked by the other person, or they stop to chat about something. In this case you have a binary variable to add to your statistical model. Number=0 means that only one person collected data, Number=1 means that two people gathered data. Is number significant in your model?

Timothy, in the monitoring program we are working on, for safety reasons, one observer and one assistant conducted the sight sampling of mammals in protected areas in Brazil. We had stablished that only the observer data would be collected, however its like impossible to blind the assistant, and in some sites, sights of both were collected in the same data sheet.

if we control the number of observers (covariates), as you said, the final abundance and density of one species is comparable between our sites and years, and with data from other studies that used only one observer?

regards

I guess the main influence, if you do not control for observer difference, is higher variation in the encounter rate, as well as in the detection probability at different distances (some people can't see far from the transect line, for example, when they are not too tall) and then in the final variance of the estimates (wich is composed by the enconter rate, detection and group size). If the estimates are adequate (small SE and CV), you would disregard the differences. However, if the contrary is your reality, the only way to reduce variation is controlling for observer differences.

Hi Waldrido, I'm not sure if we have the control of number of observers in the same site. What did happened in fact is that in some sites two observers collected the same data and other only one. In this configuration we not able to understand if the SE between sites is caused by the N of observers or environmental variation. We are working with 1-3 transects of 5 km in 23 federal protected areas.

We are thinking that two observers (local people), in theory, have higher probability of detection, and if we are improving the prob. of detection, are we estimating the density ou reative abundance with fewer effort? since the real population density in one site is dependent of local conditions,  independent of the number of observers. Are we right?

regards

Hi Ricardo,

If you don´t know who collected the data, the two observers just end up being considered a single very good observer. It probably doesn´t matter much, except that it could give you a bimodal detection curve that is harder to model. It would also probably make it difficult to detect the effects of covariates. You will probably just end up ignoring it and treating the compound observer as you would a single one. In future, you should avoid this or have each observer survey only one side of the transect.

All the Best

Bill

Hi Bill, I'd checked if we have information about the number of observers for sightings  and, unfortunately, we don't have it. We can accept that few sites (protected ares) this could happened and considering it in the results discussion. All the samplers were trained to one observer only, but we can't control it on field and samplers and managers are asking us to change the protocol for two observers recording sightings. For them it may be easier and more logical...

We are not sure if it will give a good trade-off in the final results.

What do you think abut it? Anyone have an idea?

regards

Ricardo