A Systematic Review of Small Domain Estimation Research in Forestry During the Twenty-First Century From Outside the United States

Richard W Guldin

doi:10.3389/ffgc.2021.695929

Abstract

Small domain estimation (SDE) research outside of the United States has been centered in Canada and Europe—both in transnational organizations, such as the European Union, and in the national statistics offices of individual countries. Support for SDE research is driven by government policy-makers responsible for core national statistics across domains. Examples include demographic information about provision of health care or education (a social domain) or business data for a manufacturing sector (economic domain). Small area estimation (SAE) research on forest statistics has typically studied a subset of core environmental statistics for a limited geographic domain. The statistical design and sampling intensity of national forest inventories (NFIs) provide population estimates of acceptable precision at the national level and sometimes for broad sub-national regions. But forest managers responsible for smaller areas—states/provinces, districts, counties—are facing changing market conditions, such as emerging forest carbon markets, and budgetary pressures that limit local forest inventories. They need better estimates of conditions and trends for small sub-sets of a national-scale domain than can be provided at acceptable levels of precision from NFIs. Small area estimation research is how forest biometricians at the science-policy interface build bridges to inform decisions by forest managers, landowners, and investors.

Highlights

Defining Small Domain Estimation and Small Area EstimationA study domain is a major segment of some population for which separate statistics are needed
SAE research and applications are underway in many European countries to improve estimates— reduce the RMSE or confidence intervals—of forest attributes based on sample data collected on NFI field plots
Airborne LiDAR data are becoming increasingly popular as auxiliary data, especially where country-wide laser scanning has replaced country-wide aerial photography as the raw data for national topographical mapping, transportation, or other agencies

Summary

Introduction

Defining Small Domain Estimation and Small Area EstimationA study domain is a major segment of some population for which separate statistics are needed. Small Domain Estimation Outside USA based on data from complete enumerations of the population’s members, commonly called censuses. The local density of a target variable is defined at each point of the region as the ratio of the Horvitz-Thompson estimate of the tree population total of the variable, based on a probability sample of trees taken at the point, to the area of the region. A simple random sample from the first-phase points is used to locate field plots where data are collected on the target variables (i.e., their local densities). The Swiss NFI follows this design-based Monte Carlo paradigm with its two-phase simple random sampling for post-stratification estimation, where the auxiliary information is used to do the post-stratification

Objectives

Methods

Results

Discussion

Conclusion