Abstract
Large multiobjective forest monitoring efforts such as the USDA’s Forest Inventory and Analysis Program and other National Forest Inventory (NFI) systems are usually described to the public in terms of the relationship of their sample designs to the land base of interest. Sometimes the third dimension of time is included in the description of the sample design. Additionally, there is a tendency to favor descriptions that arguably support design-based views of the used estimation systems. The accompanying justification usually relies on an assumption that design-based estimators are “objective,” whereas model-based estimators are “subjective.” This article posits that this argument misses the mark because design-based estimation begins with the assumption that a probability sample exists and that the sample observations have been obtained without error. In most large sampling efforts, it is known that violations of the sample design not only might, but will, occur. Additionally, some objectives in a multiobjective inventory require estimation of attributes of tangentially related populations. Here we unify estimation methods into an overall theoretical framework, or “wrapper” model, that allows for the recognition, acknowledgement, and accounting for nonsampling errors and imperfect sample frames into estimation systems capable of combining all observations available to analysts of these large inventory systems.
Study Implications
This study describes a theoretical framework, named a “wrapper” model, that gives developers of national forest inventory and monitoring systems a more complete way to describe the assumptions and information sources contributing to the outputs from those systems. Rather than the usual description of the outputs in terms of the sample design and corresponding estimation system, these elements are recognized simply as (albeit major) components within the wrapper model.
The Wrapper Model for Multiobjective Forest Monitoring Systems
