Efficient and Unbiased Estimation of Population Size.

Area of Science:

• Ecology
• Social Sciences
• Image Analysis

Background:

• Accurate population sizing from aerial images is crucial but challenging.
• Existing automatic detection algorithms often fail, and manual counting is impractical for large populations.
• Current density estimation methods lack detailed sampling protocols and error assessment.

Purpose of the Study:

• To develop a novel, unbiased method for population size estimation from aerial images.
• To adapt geometric sampling principles for ecological and social science applications.
• To provide a reliable error prediction formula for population estimates.

Main Methods:

• Application of geometric sampling principles to planar population estimation.
• Implementation using random superimposition of coarse quadrat grids.
• Development and validation of a new theoretical error prediction formula using Monte Carlo resampling.

Main Results:

• The proposed geometric sampling design is unbiased regardless of population size, pattern, or perspective artifacts.
• Counting 50-100 individuals in 20 quadrats can achieve relative standard errors of 8%-5%.
• The new error prediction formula demonstrates improved accuracy over traditional assumptions.

Conclusions:

• Geometric sampling provides a robust and efficient solution for population sizing from aerial imagery.
• This method overcomes limitations of automatic detection and manual counting, making semi-automatic sampling a viable option.
• The approach is applicable to diverse populations including people, animals, and trees.