Explain one of the following distance sampling concepts in your own way:
Detectability
One of the four assumptions of distance sampling, and why it matters
Random placement
Detection is certain
No movement
Accurate measurements
Detection functions
Effective strip width
You can explain the concept however you wish, for example:
Draw a diagram
Tell a story from your own experience
Write your own definition
Add an audio recording
Whatever way works best for you, and allows you to express your ideas!
Detectability is basically about the fact that we donât see every animal that is actually there. I Imagined walking on a transect in a dense forest, a chimpanzee standing right next to the transect line is easy to be spotted, but one that is far away, hidden behind thick vegetation or large trunk of a tree, or moving quietly in fear of human presence is much harder to be detected. But distance sampling accepts this reality and models it by assuming that the probability of detecting a chimpanzee decreases as distance from the transect or camera increases.
By estimating detectability, distance sampling allows us to correct for the chimpanzees we missed and avoid underestimating population density. Without accounting for detectability, we would wrongly assume that the number of chimpanzees observed equals the number present, which is especially problematic in forests where visibility is low and some species are shy.
Random Placement is one of the key assumptions of distance sampling. The species or objects that you are surveying may or may not be randomly distributed, but itâs crucial that the transects are randomly placed to account for variation in distribution and detectability across the sampling area. Alignment of transects with spatial features like roads or certain types of terrain can influence sampling results, as animals could be attracted to or away from certain habitat or features within the sampling area.
This ensures that the detection function derived from the observed distance histogram is based on the detectability of the the objects rather than the habitat selection or other biases. This gives us a more accurate estimation of detectability across the sampling area. So, while we want to maximize detectability, selecting transects where we expect to see more animals, as opposed to random placement, incorrectly correlates detectability to density and doesnât give a true measure of detectability and density across the entire area.
Detectability â> The capacity that we have to visualize a specific group of individuals/ species/ populations when surveying them.
**I think of it this way: me with glasses and binoculars in a grassland== higher detectability // me with no glasses and equipment and a hangover in a dense canopy== lower detectability**
Assumption of random placement â> When doing distance sampling, the condition that we must take into account that consists of setting the location of our surveying transect/ points without being biased towards where we think we can see the study species more (or less).
Assumption of Detection to be certain â> When doing distance sampling, knowing that you can potentially see the study species where you are going to survey it.
Assumption of No movement â> One has to keep in mind that when you see an individual of your study species/ population, you have to catch its precise (or at least as much as possible) location before it runs away.
Assumption of Accurate measures â> The necessity to be accurate of the location of the surveyed individuals when doing distance sampling.
** Set my transect points randomly just knowing that I will not go surveying into a river searching for grasshoppers and keeping in mind that I have to be particularly silent to not scare what I encounter away. At the time of encounter, like Pokemon, I got to be sharp and catch its specific location.**
Detection functions ââ> the mathematical and statistical explanation behind the fact that I will be sharper in detecting things that are closer.
**Just a fancy term and complex maths that explain how well (or not) I get to see an individual at different distances away from me **
Effective strip width â> When surveying, the distance from you at which you have a 50:50 chance to detect the study individuals.
** I see (almost) everything that happens at 0.2m away from me/ I see (barely) nothing that happens at 500m away/ ESW = the distance in the middle at which I will catch the 50% of things (but then apply it to your study individuals)**
The detection function is a model we can use to understand that the further away an animal, the less likely it is that detection will take place. Or, in other words, the closer I am to a species, the more likely I will be to detect it. This is important to understand because the function states that just because my detection decreases with distance, doesnât mean that density does as well.
For example, I often see fields with prairie dog burrows here in Colorado. I see many prairie dogs up close running around, popping their heads out of their burrows, sun bathing, and even hear their chirps when they are tucked away in their dens. I have a harder time hearing and seeing them the further they are away from me; their chirps become more quiet with distance and they are small with tan coats that blend in with the dry land and grasses. Although I cannot see every burrow and prairie dog, that does not mean that they do not exist. I can prove this by walking around the area and picking a new point to monitor. I will then see many more burrows and prairie dogs.
Firstly, before random placement can be enacted, it must be ensured that there has been past observations proving that the specific species in question has occupied the area. Evidence such as scat, calls or any biological matter from the specific species is all that is needed.
After this has been confirmed, random placements is the method of placing random transects within the research diameter. The variability in the locations that may appear ensures the avoidance of bias. However, it is highly encouraged to avoid locations that may have certain amounts of terrain that may influence animal behavior, such as roads.
Hello, I have chosen âDetectabilityâ: The capability to visualise an individual or a population during the survey. Animals (or objects) closer to the observer can be easier to detect. Distance sampling takes this into consideration, as detectability drops as distance increases. This, therefore, allows researchers to correct counts and produce more accurate population estimates.
Thank you for your post, Byron. I found your example with the chimpanzees very useful to quickly visualise and therefore better understand the concept.
Random placement in distance sampling means that survey lines or points are selected randomly (within a study area), this is to ensure that they are independent to the survey specie distribution, this has some advantages:
it is fundamental to have an unbiased calculation of the detection function
avoid over or under sampling of some areas and habitats
avoid human bias (even if we try not to, we will always want to survey areas where we think we will find the specie we are looking for!)
Completely random placement of point of transect might not cover the study area well, so they are often placed within a grid, one transect/point for each square to ensure the area is covered evenly.
A common problem I encountred during fieldwork and survey design :
Depending on the kind of survey and the study area geographical features, land cover, and specie, complete randomness is often not feasible, is therefore important taking this into account when designing a survey.
for example, in evergreen forests with dense shrub layer, where is almost impossible to walk without creating a huge disturbance (having to cut through vines and bushes, with consequent sound) therefore impacting the detectability, chosing an existing path to conduct the survey is a good solution?
would it be better to âcutâ paths in the forest following the randomly selected transect lines a week or so before the survey?
Hi Morgan,
Thank you for your clear explanation. I am glad you mention the chirping of the praire dogs, because I have a question about call/sound surveys:
When doing a transect, would you consider these two data (visual vs Audio) separately when calculating the Effective Strip Width ? what I mean is, shall we take in consideration that sound waves and light travel in different way and therefore the detection range might change?
so when I come back from the filedwork I would analyse the data from the praire dogs I saw and the one I heard separately ?
Detection functions describe how probable it is to detect an individual or object at a certain distance. If you assume that at 0 distance from your line or point you will detect all individuals, then at 5 meters you will miss some but still see most, at 20 meters distance you will see some but miss most and you donât see any at 50 meters, for example. This then can be described by a detection function. Detection functions will differ for e.g. different habitats: a dense forest vs an open prairie will have different detectability at different distances.
I know this can be a tricky assignment with only a few concepts to choose from, so it feels like youâre repeating what has already been said. Hopefully the variety of explanations helps if you are confused or unsure about a concept
Adding examples from your own experience (if you have field time) is a great way to reduce the repetition. Iâm loving the @Byro_ns shy chimpanzees and @Morgan 's small and camouflaged prairie dogs ! These real-world examples really bring the ideas to life
@JManas your example of wearing your glasses to better detect the species highlights the possibility of variation between observers or teams, based on skill, experience, or physical capabilities. Training is important to minimise variation amongst observers, but there are other ways to tackle this too, which we explore during the Survey design and Analysis courses
@matthron ,thanks for emphasising the problems caused by selecting survey locations where you expect to see animals. Itâs worth noting that a selective approach would bias estimates upwards regardless of which method you use estimate abundance
@majellin - good point that only areas that contain the species will provide information on detectability! The benefit of using distance sampling to survey areas where you are unsure the species is present is that you gain more certainty about absence, based on your survey effort. Occupancy modelling is similar in this respect
The Effective Strip Width doesnât necessarily fall at the distance where you detect 50% of the animals, or where your probability of detection is 50%
ESW is the distance where the number of animals you miss closer than the ESW matches the number you see beyond it
Imagine that you are surveying for elephants in an undulating savannah, dotted with acacia. Your detection function will have a broad shoulder because elephants are easy to detect for quite a long distance. Further from the transect, detection may fall away rapidly because of terrain, a change in vegetation, or the accumulation of shrubs obscuring your view. In this situation, you might spot 90% of the elephants:
Alternatively, in another environment your detection function may have a narrow shoulder and decline smoothly (or dramatically!), and you might only detect 40% of the animals or less:
Great question! Provided you record the method of detection (visual or audio) with each observation, one approach would be to include âdetection typeâ as a detectability covariate during analysis. This allows you to model differently-shaped detection functions for auditory and visual detection, so all your data still feeds into a single density estimate
Thanks for mentioning this. These are the kind of topics we cover in the Survey design and Analysis with R courses
Detectability means Probability of detecting an animal given that it is present.
No Movement Before Detection:
Animal must not move toward or away from line before distance is measured.If animal moves, True perpendicular distance & Detection function will distort.
Detectability refers to the chance of detecting an animal given that it is present in the survey area. Note that: Detectability decreases with increasing distance from the transect line, making animals close to the line more likely to be seen than those far away.
Concept of Random Placement ensures representative sampling: Transects must be placed randomly with respect to animal distribution, if transects are placed near prefered habitats of the target animals such as waterholes or near roads. Some animals will be oversampled while others may be undersampled, making the density estimates become biased
In UAV wildlife surveys, detectability is a cornerstone of accurate population estimation, as it directly determines which animals are observed and which are missed. Detection probability typically declines with increasing distance from the UAV, but it is also strongly influenced by species-specific behaviorâcryptic, nocturnal, or burrowing animals are inherently harder to detect. Habitat structure plays a major role: dense vegetation, canopy cover, or rugged terrain can obscure animals from aerial sensors, while open habitats generally allow higher detectability. Temporal factors, such as time of day or season, also matter, as many species alter their activity patterns in response to temperature, light, or human disturbance.
Technical aspects of the UAV survey further shape detectability. Sensor typeâwhether RGB cameras, thermal imagers, or otherwiseâaffects which animals can be detected under different environmental conditions. Flight parameters, including altitude, speed, and transect overlap, influence both image resolution and coverage, with higher altitudes or faster flights reducing the chance of spotting smaller or camouflaged animals. Environmental conditions such as lighting, cloud cover, wind, and precipitation can further degrade visibility. Human factors also play a role: whether images are manually collected in real time or processed later using automated detection algorithms can affect both the likelihood and consistency of detecting animals.
Addressing these variables through thoughtful survey design and statistical analysisâsuch as incorporating detection functions, distance sampling, or covariate adjustmentsâensures that UAV-based monitoring accounts not only for observed individuals but also for those likely missed.
and 
! These real-world examples really bring the ideas to life
