Double-observer line transect methods: levels of independence
read more
Citations
Unmanned aerial vehicles (UAVs) for surveying marine fauna: a dugong case study.
Estimation of population density by spatially explicit capture–recapture analysis of data from area searches
Aerial surveys of seabirds: the advent of digital methods
Using mark–recapture distance sampling methods on line transect surveys
A Unifying Model for Capture-Recapture and Distance Sampling Surveys of Wildlife Populations
References
Introduction to Distance Sampling: Estimating Abundance of Biological Populations
Model selection: An integral part of inference
Advanced distance sampling
Incorporating covariates into standard line transect analyses.
Related Papers (5)
Frequently Asked Questions (13)
Q2. What were the three forms of independence used in the SCANS II survey?
Models were fitted corresponding to full independence (α = β = 0), point independence (α = 0, β unconstrained), and limiting independence with α ≥ 0, β ≥ 0.
Q3. What is the main argument of Laake?
Laake (1999) argued that heterogeneity is less of a problem on the line, where probability of detection is relatively high, than away from the line, so that assuming independence only on the line should yield less biased estimates of abundance.
Q4. How many minke groups were detected by the primary platform?
Using a truncation distance of 700 m, the tracker detected 54 minke groups totaling 62 animals, while the primary platform detected 57 groups totaling 59 animals; 17 groups (19 animals) were detected by both tracker and primary platform.
Q5. Why were double-observer line transect survey methods used?
Double-observer line transect survey methods were used because for many species detection of animals on the trackline was expected to be less than unity.
Q6. What is the full likelihood for double-platform data?
The full likelihood for double-platform data may be expressed as L = LnLzLy |zLω where Ln is the component accounting for variation in total number of animals n detected by at least one observer, Lz corresponds to any observation-specific covariates z, Ly |z corresponds to the conditional distribution of distances y, given covariates z, and Lω corresponds to the mark–recapture data (Laake and Borchers, 2004).
Q7. What is the common Petersen estimator?
If the authors estimate pj by p̂j = n12/nj ′ for j = 1, 2, j ′ = 3 − j, and substitute in, the authors find that N̂c = n 1n 2n 12 , which is the familiar Petersen estimator.
Q8. What was the bias in the limiting independence model?
Model-averaged estimates had low bias, except for scenarios 3 and 4, for which around 40% of analyses under the limiting independence model were rejected due to high correlation between α̂ and λ̂0j .
Q9. What is the full likelihood for a double-platform data?
if the full independence assumption holds then it is only necessary to use Lω (Borchers et al., 1998) and with the point independence assumption, Lω and Ly can be maximized independently using models for pj |j ′(y) and pj (y), which separateinto the two respective likelihood components (Borchers et al., 2006).
Q10. What is the standard assumption for limiting independence?
1. If the authors consider the following logistic model for limiting independence:loge{ δ0(y)1 − δ0(y)} = α + βy + loge { 1 − L(y) U (y) − 1 } , (1)then α = 0 specifies point independence at y∗ = 0, and α = β = 0 specifies a full independence model.
Q11. How many models were fitted to the two continuous covariates?
In each case, three models were fitted: the first with observer as a factor and distance as a covariate, the second with the addition of an interaction term between the two, and the third with the squared distance as an additional covariate, together with interaction terms between observer and the two continuous covariates.
Q12. What is the way to ensure that pj ′(y) 1?
This requires a model for δ(y) with the following properties to ensure valid probabilities:(1) δ(y) ≤ U (y), where U (y) = min{1/p1(y), 1/p2(y)}, which ensures that pj |j ′(y) ≤
Q13. How do authors minimize the bias in the estimation of animal abundance?
Authors have attempted to minimize this bias, for example, by modeling the effects of covariates (Borchers, Zucchini, and Fewster, 1998; Borchers et al., 1998, 2006; Borchers, 1999; Schweder et al., 1999; Laake and Borchers, 2004), or by assuming independence in the detections of instantaneous cues (such as whale blows) rather than of animals (Schweder et al., 1999; Skaug and Schweder, 1999).