Anomalous diameter distribution shifts estimated from FIA inventories through time

Forestry, Vol. 83, No. 3, 2010. doi:10.1093/forestry/cpq009 For Permissions, please email: [email protected] Advance Access publication date 7 April 2010 Since the 1930s, the individual, regional Forest Inventory and Analysis Units concentrated on increasing their efficiency of operation, often by improving the efficiency of the sample design for the specific forest conditions and the attributes that were considered the most important within their region while meeting national objectives. Despite a rich diversity of objectives between regions, all of the regions converted to probability proportional to size (pps) sampling for the selection of sample trees, using clusters of horizontal point samples, beginning in the late 1950s. This was done because, in these designs, trees are selected in proportion to their basal areas and basal area is highly correlated with important variables such as wood volume, age and stand size. More recently, growing environmental awareness has led to increased interest in forest biomass and carbon, which are highly correlated with volume and basal area. The diversity of methods between regions presented challenges for those wishing to compare results across regions. This, in combination with other considerations, such as an expanding scope and set of objectives for the program, helped to give rise to a clarion call to standardize methods across regions (McRoberts, 2005). The resulting national design uses nested, fixed-area plot clusters to select trees, within size categories, with equal probability (or probability proportional to size category, ppsc), rather than the previously favoured continuous-pps (horizontal point) sample clusters. The general field plot cluster for the new design is depicted in Figure 1. As seen in the figure, and discussed in Roesch (2007a), the cluster design allows for an optional macroplot surrounding each sub-plot for an increased sample of larger diameter trees. Only one of the four Forest Inventory and Analysis Program (FIA) regions (the Pacific Northwest Research Station) uses the macroplot option extensively, while two of the FIA regions (the Southern Research Station and the Northern Research Station) do not use it at all. Given the potentially significant increase in sampling effort required by the use of the macroplot clusters, we might wonder what differences in efficiency for meeting program goals are being experienced through their use. Figure 2 shows the number of sample trees per plot in 5-cm-diameter classes averaged over (1) all previous periodic inventories (mostly using pps designs) and (2) over the annual inventory since its inception, for each of the four FIA regions, included in FIA’s publicly available database, FIADB 4.0 (Anonymous, 2009). This figure shows that these two general sampling approaches have led to quite different allocations of effort in observing the diameter distributions (notably for the smallest and largest diameters) in each of the four regions. Note that observation of the upper end of the diameter distribution was Anomalous diameter distribution shifts estimated from FIA inventories through time