|
Abstract:
|
It has become increasingly common for data about individuals to be spread across multiple sources, making it necessary to identify which records refer to the same individual. The task of record linkage is to estimate this linkage structure in the absence of a unique identifiable attribute. We present a Bayesian hierarchical record linkage model for spatial location data motivated by estimation of population growth and mortality functions for forests. The location data comprises overlapping lidar scans taken at varied points in time. Each scan is post-processed to determine the location of the treetops and an estimate of canopy volume. Tree mortality and growth functions are then estimated, dependent upon correctly identifying unique individuals across scans without a unique identifying attribute. Application of the model identifies the unique individuals while quantifying the uncertainty in the linkage process and incorporating it into the inference on tree growth and mortality. We discuss the model formulation, including data and process models, and assess performance on simulated data sets and scans of the Upper Gunnison Watershed provided by the Rocky Mountain Biological Lab.
|
