Estimation Details
Estimation of forest variables follows the procedures documented in Bechtold and Patterson (2005) and Stanke et al 2020. Procedures for computing diversity indices are outlined in Hill (1973) and Shannon (1948).
Alpha-level indices are computed as the mean diversity of a stand. Specifically, alpha diversity is estimated using a sample-based ratio-of-means estimator of stand diversity (e.g. Richness) * land area of stand / total land area within the domain of interest. Thus estimates of alpha diversity within a stand are weighted by the area that stand represents. Gamma-level diversity is computed as a regional index, pooling all plot data together. Beta diversity is computed as gamma diversity - alpha diversity, and thus represents the excess of regional diversity with respect to local diversity.
Users may specify alternatives to the 'Temporally Indifferent' estimator using the method argument. Alternative design-based estimators include the annual estimator ("ANNUAL"; annual panels, or estimates from plots measured in the same year), simple moving average ("SMA"; combines annual panels with equal weight), linear moving average ("LMA"; combine annual panels with weights that decay linearly with time since measurement), and exponential moving average ("EMA"; combine annual panels with weights that decay exponentially with time since measurement). The "best" estimator depends entirely on user-objectives, see Stanke et al 2020 for a complete description of these estimators and tradeoffs between precision and temporal specificity.
When byPlot = FALSE (i.e., population estimates are returned), the "YEAR" column in the resulting dataframe indicates the final year of the inventory cycle that estimates are produced for. For example, an estimate of current forest area (e.g., 2018) may draw on data collected from 2008-2018, and "YEAR" will be listed as 2018 (consistent with EVALIDator). However, when byPlot = TRUE (i.e., plot-level estimates returned), the "YEAR" column denotes the year that each plot was measured (MEASYEAR), which may differ slightly from its associated inventory year (INVYR).
Stratified random sampling techniques are most often employed to compute estimates in recent inventories, although double sampling and simple random sampling may be employed for early inventories. Estimates are adjusted for non-response bias by assuming attributes of non-response plot locations to be equal to the mean of other plots included within thier respective stratum or population.
Working with "Big Data"
If FIA data are too large to hold in memory (e.g., R throws the "cannot allocate vector of size ..." error), use larger-than-RAM options. See documentation of link{readFIA} for examples of how to set up a Remote.FIA.Database. As a reference, we have used rFIA's larger-than-RAM methods to estimate forest variables using the entire FIA Database (~50GB) on a standard desktop computer with 16GB of RAM. Check out our website for more details and examples.
Easy, efficient parallelization is implemented with the parallel package. Users must only specify the nCores argument with a value greater than 1 in order to implement parallel processing on their machines. Parallel implementation is achieved using a snow type cluster on any Windows OS, and with multicore forking on any Unix OS (Linux, Mac). Implementing parallel processing may substantially decrease free memory during processing, particularly on Windows OS. Thus, users should be cautious when running in parallel, and consider implementing serial processing for this task if computational resources are limited (nCores = 1).
Definition of forestland
Forest land must have at least 10-percent canopy cover by live tally trees of any size, including land that formerly had such tree cover and that will be naturally or artificially regenerated. Forest land includes transition zones, such as areas between heavily forest and non-forested lands that meet the mimium tree canopy cover and forest areas adjacent to urban and built-up lands. The minimum area for classification of forest land is 1 acre in size and 120 feet wide measured stem-to-stem from the outer-most edge. Roadside, streamside, and shelterbelt strips of trees must have a width of at least 120 feet and continuous length of at least 363 feet to qualify as forest land. Tree-covered areas in agricultural production settings, such as fruit orchards, or tree-covered areas in urban settings, such as city parks, are not considered forest land.
Timber land is a subset of forest land that is producing or is capable of producing crops of industrial wood and not withdrawn from timber utilization by statute or administrative regulation. (Note: Areas qualifying as timberland are capable of producing at least 20 cubic feet per acre per year of industrial wood in natural stands. Currently inaccessible and inoperable areas are NOT included).