% bgcolor=ccccff> « back to results for ""
Below is a cache of http://www.cnr.uidaho.edu/wlf448/AERIALSU.pdf. It's a snapshot of the page taken as our search engine crawled the Web.
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content.
Aerial Survey Aerial Survey
Aerial Surveys for Large Mammals
What is our objective?
Gut feeling for herd condition?
Index of abundance?
Valid, consistent, repeatable estimates of
abundance and condition?
Absolute Abundance
Total count over whole area
Total counts on sample plots
Total counts on sample plots
Plots can be quadrats, strip
transects, or irregular shaped
areas
Critical assumption is that every
animal is observed and counted
once
Sample Units
Total counts on sample plots - Application
Delineate area
Apply survey sampling methods to
design an appropriate sample
Test assumptions and remove
bias if necessary

Common Sample Designs
Simple Random
Stratified Random
Cluster
Systematic Sample Design
High
High
High
High
Sample Design
Elk in Med Timber
Elk in Timber w/ Copter
Elk in Timber w/ Copter

Removing Bias from Sample Counts
Correction Factor (Ratio)
Mark Recapture
Sightability Model

Removing Bias: Correction Factor
Use depends upon assumption of
a constant factor under highly
variable conditions.
Correction Factor
One drainage to next has different
cover
Use of cover varies from flight to
flight
Size of groups vary continuously
Removing Bias: Mark-Recapture
Well established statistical basis
Questionable assumptions in most
aerial survey conditions
Extremely costly in time and
resources because must capture
and mark animals each time

Removing Bias: Sightability Model
Adaptable to a variety of
conditions
Cost efficient
Not applicable if visibility is very
low
Sightability Model
Mark elk (deer, sheep, etc.)
groups with radio-collars
Fly aerial survey
Determine which groups seen and
which groups missed
Depends on group size, tree &
shrub cover, snow cover, weather,
observers, type of helicopter, etc.
Sightability Model
Keep some factors constant
Develop a sightability model for
other factors
Use sightability model to correct
for factors which we cannot
control
Sightability Model
Apply same model in other areas
too if same type of helicopter and
approach used and original model
covered appropriate conditions in
terms of veg cover, snow, animal
behavior, etc.
Weve had original crews train
new crews to insure that model
still applies.
Simple Application
Suppose we determine that 1/3 of
groups are detected (p=0.33)
Then, if see 50, actually 150
present
How? Correction Factor (CF)= 1/p
CF= 1/0.33 = 3.0
N = N
obs
* CF = 50 * 3.0 = 150
Application to a Sample
Unit
Correct each group detected for
its probability of detection
(visibility)
Sum all corrected groups in a
sample unit for an unbiased
estimate of actual number of
animals present
Application to a Herd
Unit
Calculate means, ratios,
proportions, etc. according to
survey design
Calculate variances and
confidence intervals Sightability Model
Build a sightability model using
logistic regression
p = e
µ
/ 1+e
µ
where
µ
= a+b
1
X
1
-
b
2
X
2

e.g. X
1
= group size, X
2
= veg. cover
Sightability Model
p
Chance of
Seeing
Group
Size of Group
0
1.0
1
5
10
15
20
0.5
Sightability Model
p
Chance of
Seeing
Group
Cover
0
1.0
0.5
0%
50%
100%
Lochsa River - Unit 12
Traditional
Sightability
58

65

71

75

79

81

83

85

87
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
5,000
Lochsa River - Unit 12


1985
1986
1987
0
1,000
2,000
3,000
4,000
5,000
Traditional Counts
Sightability Model
Estimate of Total Numbers
m
ik
= number of animals in group i
in land unit k
CF
ik
= correction factor for group i
in land unit k
M
k
= m
ik

CF
ik
Average = M
k

/ No. land units
sampled
Total = Average * Total no. land
units
How good is estimate?
Variance of Total =
Sampling Variance +
Sightability Variance +
Model Variance
How good is estimate?
Sampling Variance = Variation
from one geographic unit sampled
to another
= S
2 How good is estimate?
Sightability Variance = Variation
(or error) from not seeing all the
animals
= proportional to CF
2
How good is estimate?
Model Variance = variance of
parameters in model (betas)
= proportional to e
(Variance-Covariance)
Lochsa River - Unit 12
1985
1986
1987
0
1,000
2,000
3,000
4,000
5,000
6,000
Sources of Variation
1985 Estimate of total elk = 4775
Sampling Variance = 59733
Sightability Variance = 16868
Model Variance = 825
Total Variance =77426
90% Bound =458
Herd Composition
Cows 2852 (269)
Bulls 968 (166)
Calves 857 (105)
Bulls per 100 Cows 34 (6.4)
Calves per 100 Cows 30 (5.4)
Northern Yellowstone
Elk
Surveyed in 1 day with 3 Super
Cubs
Survey of whole range (all units)
Developed sightability model in
80s
µ
= 0.969 +
+ 0.0369 Group Size
- 0.540 Vegetation Cover
+ 1.701 Activity
Northern Yellowstone
Elk
1987
1988
1989
1989
1989
1990
1991
0
5,000
10,000
15,000
20,000
25,000
AERIAL SURVEY PROGRAM
All calculations easily performed
Variety of sightability models
See Unsworth et al 1994