Horvitz-Thompson-like estimation with distance-based detection probabilities for circular plot sampling of forests.

Area of Science:

• Forestry and ecological surveying
• Statistical modeling and estimation

Background:

• Circular plot sampling is standard for forest characteristic estimation.
• Obstructed visibility (e.g., from terrestrial laser scanners) prevents observing all trees within a sample plot, biasing traditional methods.

Purpose of the Study:

• To develop an unbiased estimator for forest population totals (e.g., stem density, basal area) in the presence of hidden trees.
• To provide variance estimation and confidence intervals for the proposed method.

Main Methods:

• Proposed a Horvitz-Thompson-like estimator incorporating distance-based detection probabilities.
• Derived detection probabilities using stochastic geometry.
• Evaluated the estimator's performance via simulation studies using field data and point processes, comparing it against benchmark methods.

Main Results:

• The proposed estimator is unbiased for Poisson forests.
• Simulation results indicate lower or comparable error rates compared to existing methods.
• Observed small bias (0.3%-2.2% for stem density) and conservative or nominal confidence interval coverage in field data simulations.

Conclusions:

• The novel estimator effectively addresses tree occlusion in circular plot sampling.
• It offers a statistically sound approach for accurate forest inventory, including variance estimation and confidence intervals.
• The method demonstrates superior or competitive performance in simulated forest environments.