Time as a dimension of the sample design in national-scale forest inventories
Historically, the goal of forest inventories has been to determine the extent of the timber resource. Predictions of how the resource was changing were made by comparing differences between successive inventories. The general view of the associated sample design was with selection probabilities based on land area observed at a discrete point in time. Time was not considered to be part of the sample design because it was not considered to be an element of the sampled population. Over the last few decades, the general goal of many national-scale forest inventories has been changing to monitoring the dynamic forest ecosystem. Here, we explore the inferential advantages of replacing the two-dimensional areal probability paradigm with a three-dimensional spatiotemporal probability paradigm. Our general discussion is augmented with a simulated example for estimating annual growth by diameter classes. Two assumptions of temporal indifference and remeasurement interval length indifference, which arose because of the two-dimensional view, are investigated through a simulation. The simulation compares and contrasts five estimators that differ in their reliance on those assumptions to make annual estimates. The results of the simulations often show those assumptions to be bias inducing.