Model-assisted estimation for national forest inventory via triangular tessellation

Abstract

National forest inventories (NFIs) play an important role in the monitoring and management of natural resources and are typically implemented as sample-based surveys that provide a probabilistic basis for making inferences about the population being sampled. Some estimators incorporate data collected from the population that are auxiliary to the sample, especially remote sensing imagery. The study presented here examined the efficiency of some of these estimators, including the post-stratified (PS) and model-assisted estimators. It introduces an estimator that combines the efficiency of model-assisted regression (MAR) with the disclosure risk mitigation benefits of aggregation, called the tessellated model-assisted regression (TMAR) estimator. The results of the study, using NFI data and a Landsat-based model of live tree aboveground biomass density from the USDA Forest Service Forest Inventory and Analysis program, along with a triangular tessellation derived from 648 km² US EPA Environmental Monitoring and Assessment Program hexagons, demonstrate the relative efficiency of the MAR and TMAR estimators compared to the PS estimator for a sample of counties across the contiguous United States.

Citation

​Wilson, Barry Tyler. 2025. Model-assisted estimation for national forest inventory via triangular tessellation. Forest Ecology and Management. 596: 123080. 5 p. https://doi.org/10.1016/j.foreco.2025.123080.