WLF 448: Fish & Wildlife Population Ecology

Fall 2005

LAB 4: AERIAL SURVEYS

I. Conceptual Framework (basic problems with estimating population size)

A. Observability or Sightability (B)

• Most methods of surveying animals do not result in counts or captures of all animals present on an area, i.e., the probability of seeing or catching an animal will generally be less than 1.

• To translate a count resulting from any survey method into an estimate of population size, we must estimate the proportion of animals counted (B) and then divide our count by B :

=  C / B

   where C = value of the count
         = estimate of true population size
         B = proportion of animals counted

For example, if we counted 20 birds during a survey and we know that we only see 25% of the total number of birds actually present, then = 20/0.25 = 80 birds.

• The bulk of the effort in developing population-estimation methods has involved ways of estimating B.

B. Sampling fraction (a)

• The second basic problem is that time and money are nearly always limited, so a particular survey method often cannot be applied to the entire area of interest. Therefore, sample area(s) must be selected that represent a fraction (a) of the total area of interest. The sampling fraction, a (alpha), is usually known with reasonable accuracy. If C represents the total number of animals counted on a representative sample area (assuming B =1), then the population size of the entire ^{area, , can be estimated as:}

^{= C / a}

For example, if 50 elk were observed on sample plots representing 10% of the total census zone, then C = 50, a = 0.10, and = 50/0.10 = 500, which is our estimate of the number of elk in the entire census zone.

• Sampling designs can range from simple-random sampling (with or without replacement) to more complex designs such as stratified-random sampling (see Ratti and Garton 1994:13).

C. Observability (B) + Sampling fraction (a)

• Combining equations for observability and sampling yields the following general population-size estimator:

^{= C / ( * a)}

	where 	 = estimate of population size.
	      	C  = the count of animals.
		 = estimate of the proportion of the animals counted.
		a  = fraction of area sampled.

Note: if we estimate population size (N) at some specified location of total area (A), we can estimate density as: = / A.

• Virtually all estimators of population size can be expressed in the form of the above equation. However, we also need to account for the variance associated with our estimate. For example, if a (sampling fraction) is known and there is no ^{correlation between C and B, then variance of is:}

var()= (N2)[(var(C)/C2)(1-a)+ (var()/B2)]

Total variance = sampling var + sightability var + other sources of variation

            sampling variance: originates from taking a sample of an area

            sightability variance: originates from different probabilities of sighting individuals, the bigger the correction factor the greater the variance

            other sources of variation: originates from the model and the accuracy of its parameters.

II. Aerial Surveys as a Population Estimation Technique

A. Application

• Common technique for estimating size and structure (age ratios, sex ratios) of big game and waterfowl populations.

• Population-census techniques, such as aerial surveys, are often used as a measure of relative abundance, recognizing that the technique is biased and that the bias cannot be removed or estimated, but only held constant.

B. Definitions/Concepts

1. Observability (B ) or Sightability - probability of seeing or catching an animal. In most cases, < 1 (i.e., not all individuals are seen). Note:  In lecture, detection functions were discussed, which is the same as a model for sightability.

2. Census zone - the whole area in which the number of animals is to be estimated. The census zone is usually divided up into a number of discrete units known as sample units (of equal or unequal size), and a specific number of these are chosen to sample.

3. Sample zone or unit - that part of the census zone that is searched and counted.

C. Types of Aerial Surveys (i.e., sampling strategies)

1. Transect Sampling.
2. Quadrat Sampling
3. Block Sampling

D. Factors to Consider in Selecting a Survey Method:

• cost
• navigation
• boundary effects
• counting
• sample error
• fatigue

Note: Based on the above factors, transect sampling is usually best, when it can be applied.

E. Types and Sources of Error in Aerial Surveys

1. Types of error

• Counting error--misclassifying animals, missing some animals in some habitats but not others, spending unequal amounts of search time all lead to increased variability and decreased precision.

• Counting bias - most biologists tend to undercount, and this is called counting bias.

undercounting is the rule in aerial censuses.

2. Sources of Error (i.e., variables affecting sightability)

• Transect width
• Altitude
• Speed
• Observers (training, experience, etc.)
• Other: topography, vegetative cover, snow cover, etc.

3. Errors resulting from poor survey design and inappropriate analysis:

"Appropriate sampling techniques coupled with standardized methods of survey ensure that even if our estimates are inaccurate at least they are repeatable. At the worst they will be useable as indices of density, if not as estimates of absolute density" (Caughley 1977:614).

III. Correcting for Sightability Bias in Aerial Surveys

A. Two basic concerns: Sightability (B) and Sampling fraction (a)

Note: a (alpha) is frequently known with reasonable accuracy. We assume the proportion of area sampled is equal to the proportion of animals sighted.  The bulk of the effort in developing population-estimation methods for animal populations has involved ways of estimating B.

B. Methods to Estimate Sightability Bias (B)

• Total Counts (e.g., aerial photographs, complete ground counts on sample plots)
• Incomplete Ground Counts (based on 2-sample Peterson estimator)
• Marked Subpopulation (radio telemetry can be used with this method).
• Multiple Observers (see Pollock and Kendall 1987, Lancia et al. 1994:229-230).
• Sighting Probability Models (Caughley et al. 1976, Samuel et al. 1987). Note: As discussed in lecture, these models are a detection function.

C. Sighting Probability Models

• Experimental surveys are conducted to identify variables that influence sighting probability (e.g., % vegetative cover, % snow cover, group size). Once identified, these variables are used in regression models to predict sighting probability (B). The raw-count data are then adjusted for sighting probability and extrapolated using survey-sampling techniques. Parameters estimated include population size and, where applicable, population structure (e.g., age ratios, sex ratios).

• Samuel et al. (1987) used data from radio-marked elk to develop a sighting probability model for use with aerial surveys in Idaho.

• Advantages - the costly process of estimating sighting probabilities is done only once during the initial experimental period of model development. After model(s) have been developed and tested (validated), survey efforts require only recording information on the model variables.

• Limitations - sightability model(s) are developed under specific conditions and may not perform equally well if applied using different aircraft or if applied to different species, geographic locations, habitat types, or times of the year.

"There is no one universal method for correcting biases in visibility from aerial counts. In some cases, the biases may remain of unknown magnitude, and aerial counts should then not be used as absolute population estimates" (Krebs 1989:103).

IV. Program Aerial Survey (Unsworth et al. 1994, 1998)

• Includes sightability models for elk, mule deer, bighorn sheep, and moose.

• Provides estimates of population size and structure (where applicable), and can be used to calculate sample-size requirements and test for differences among surveys.

• See In-class Exercise for instructions on how to run and interpret output from program Aerial Survey.

V. Problem Set

VI. Selected References

• Caughley, G. 1974. Bias in aerial survey. J. Wildl. Manage. 38:921-933.

• Caughley, G. 1977. Sampling in aerial survey. J. Wildl. Manage. 41:605-615.

• Krebs, C. J. 1989. Ecological methodology. Harper and Row, Publ., New York. 654pp.

• Lancia, R. A., J. D. Nichols, and K. H. Pollock. 1994. Estimating the number of animals in wildlife populations. Pages 215-253 in T. A. Bookhout, ed. Research and management techniques for wildlife and habitats. Fifth ed. The Wildlife Society, Bethesda, Md.

• Norton-Griffiths, M. 1978. Counting animals, 2nd ed. African Wildlife Leadership Foundation, Nairobi, 110pp.

• Pollock, K. H., and W. L. Kendall. 1987. Visibility bias in aerial surveys: A review of estimation procedures. J. Wildl. Manage. 51:502-510.

• Samuel, M. D., E. O. Garton, M. W. Schlegel, and R. G. Carson. 1987. Visibility bias during aerial surveys of elk in northcentral Idaho. J. Wildl. Manage. 51:622-630.

• Unsworth, J. W., F. A. Leban, D. J. Leptich, E. O. Garton, and P. Zager. 1994. Aerial Survey: User's Manual, 2nd ed. Idaho Dep. Fish and Game, Boise, Id. 84pp.

• Unsworth, J. W., F. A. Leban, D. J. Leptich, E. O. Garton, and P. Zager. 1998. Aerial Survey: User's Manual, 3rd (Electornic) ed. Idaho Dep. Fish and Game, Boise, Id.

Revised: September 09, 2005