Estimating Metrics of Forest Spatial Pattern from Large Area Forest Inventory Cluster Samples

Abstract

Abstract The interest in ecologically meaningful information from forest inventories is increasing. Forest area and its characteristics of spatial distribution are among this information. This paper describes a technique for deriving some metrics of forest spatial pattern from nonmapped forest inventory samples. The technique is developed for clusters of subplots, though applicable also for other plot types. It evaluates the area of the three categories: forest, nonforest, and buffer, estimated by the percentage of cluster plots where all, none, or some subplot centers fall into forest. Of particular interest is the buffer area, which is an imagined strip along the forest boundary: the larger this area, the more forest boundaries there are, and the more fragmented the forest pattern is. The estimates of forest and buffer area percentage are used to derive metrics that are related to perimeter length and mean patch size. Variance estimators for these metrics are given. Two examples are presented to illustrate the characteristics of the method, one with a schematic map, one with real inventory data. The most meaningful results are obtained when the size of the cluster plots used is smaller than the forest patches and smaller than the distance between them. The examples suggest that, in order to obtain reasonably precise estimates, sample size should be n = 500 or more. These conditions commonly hold in large area forest inventories. The technique processes information that is readily available from the field measurements of large area forest inventories. It does not require extra measurements and adds an ecologically meaningful aspect to the data analysis. It is independent of the availability of complete maps of the inventory region or of mapped plots, and therefore also allows the retrospective analysis of old forest inventory sample data. FOR. SCI. 46(4):548–557.