Estimating forestland area change from inventory data
Simple methods for estimating the proportion of land changing from forest to nonforest are developed. Variance estimators are derived to facilitate significance tests. A power analysis indicates that 400 inventory plots are required to reliably detect small changes in net or gross forest loss. This is an important result because forest certification programs may require additional precautions when wood from areas where forestland area loss is occurring is harvested or purchased. Net and gross forest area loss must be clearly differentiated to avoid confusion. Estimates of gross forest cover loss from satellite data should not be equated with net forest area loss, which can be better determined from remeasured forest inventory plots. Simultaneous tests of net and gross forest area loss should use multiple comparison procedures to ensure that overall error rates are correct. Examples of applications demonstrate how to properly perform these tests. A simulated example is used to verify that the variance estimators are reliable. An application to USDA Forest Service inventory data indicates that neither net nor gross forest loss at the state level was statistically significant for states that had sufficient remeasured plot data publicly available when this analysis was done.