Estimators for forest attributes on a simulated population of medium-sized tropical forest in southern Mexico

Abstract

Throughout the data of the National Forest and Soil Inventory (INFyS) of Mexico, it is not uncommon to find clusters with less than four subplots (incomplete). The consequences of this condition on the forest parameters estimates are yet to be completely analyzed. The main objective of this work was to compare the behavior of different sampling estimators under such conditions of cluster completeness. Using an artificial population of 9,370,000 trees, created on a 10,000 ha surface, a total of 88 systematic sampling grids using four-plot circular clusters were set. Each grid had 81 clusters, separated by 1 km. On each sampling grid, three different completeness conditions were tested: a) full completeness (all clusters with four subplots) b) 88 % completeness and c) 63 % completeness. On each condition, timber volume (m³ ha^-1) and tree density (tree ha^-1) were estimated using the following estimators: 1) Forest Inventory and Analysis (FIA) 2) Van Deusen Estimators 3) Means of ratio and 4) Ratio of means. The estimators were evaluated using relative bias on the mean and the variance. For volume, on each of the three completeness conditions, the mean estimates were similar and unbiased using the proposed four estimators. Nevertheless, the FIA estimator produced biased variance estimates ranging from two to five times larger for 88% and 63% completeness respectively. Similar behavior was observed on tree density. The FIA estimators will produce biased results on the variance estimator when a high percentage of clusters is incomplete.