Variance estimation for systematic designs in spatial surveys

Area of Science:

• Ecology
• Spatial Statistics
• Wildlife Biology

Background:

• Systematic designs in spatial surveys offer lower variance for density estimation compared to random designs.
• Estimating variance for systematic designs is challenging, often leading to overestimation and loss of efficiency gains.
• Current methods approximate systematic designs with random or stratified designs, but these can be biased.

Purpose of the Study:

• To develop a novel, unbiased estimator for variance in systematic spatial surveys.
• To improve the precision of density estimates in ecological studies.
• To address limitations of existing variance estimation methods.

Main Methods:

• Developed a new "striplet" estimator based on modeling the spatial encounter process.
• Simulated various survey scenarios including strip-sampling, distance-sampling, and quadrat-sampling.
• Applied the estimator to spotted hyena density data in Serengeti National Park.

Main Results:

• The striplet estimator demonstrated negligible bias and excellent precision across diverse simulation scenarios.
• Compared to existing methods, the striplet estimator significantly reduced the reported coefficient of variation for spotted hyena density (11% vs. 20% and 17%).
• Simulations verified the substantial reduction in reported variance achieved by the new estimator.

Conclusions:

• The striplet estimator provides a significant advancement for variance estimation in systematic spatial surveys.
• This method enables more accurate reporting of precision for density estimates, particularly in complex ecological populations.
• The findings support the use of the striplet estimator for more reliable wildlife density assessments.