Accuracy Assessment Reference Data

The evaluation sites against which the vegetation maps are compared are forest inventory plots established as part of the Forest Service's Forest Inventory and Analysis (FIA) program. Often, collecting enough field observations is so prohibitively expensive that valid map evaluation cannot be conducted. The FIA data provide a large quantitative dataset and help to avoid the expense of collecting additional field observations for accuracy assessment. Until recently, the plots collected by the inventory programs of the Forest Service in California were established on a permanent 3.4 mile grid, although the grid was intensified in some areas to increase sampling in National Forest plantations and significant vegetation communities. More recently, the basic sampling system has changed from a grid to a hexagonal sampling framework. Both are systematically random sampling systems independent of the maps they are used to assess.

The FIA plot data are further categorized into data collected on National Forest lands (NF) and data collected on ownerships other than National Forest. Historically NF plot data collection was administered by Regional personal while the Pacific Northwest Research Station collected data on all other lands. For the purposes of this discussion the two types of FIA data are referred to as NF or PNW. Differences in the amount, type and mode of data collection occur between NF and PNW plot data prior to 2002. The most salient differences between the two sets are:

PNW ground plots were installed only in specific forest conditions. Most other samples were photo-based with no quantitative measurements. All NFS plots are quantitative, ground-based measurements. See table 1.
NFS plots sample all National Forest lands including wilderness whereas PNW plots are subject to exclusion from parks and reserve areas.
NFS plots are from the 3.4 mile base grid plus additional plots that result from sampling intensification. See intensified.shtml
Field observations of vegetation types for shrub communties and rare forest types.

The most direct effect these differences have on map accuracy assessment are; 1) the frequency of plots within a given map condition, and 2) the ability to assess map attributes containing elements that cannot be precisely determined from aerial photography. In a systematic random sampling system, the number of plots is directly related to the probability of sampling a specific map condition. Therefore, NF plot data sample more map conditions and sample within conditions more frequently, resulting in a more robust statistical assessment. The PNW plot data that are photo-based are reliably used only for life form assessment. For areas outside National Forest containing a large amount of photo-sampled conditions, there are generally no data available for assessing vegetation type, tree size, and tree canopy closure. Table 1 illustrates the sample intensity and conditions associated with the PNW plots across California.

While sampling intensity varies between NF and PNW plot data, the plot design itself is generally the same. A discussion of plot design for FIA data is located at: http://www.fs.fed.us/r5/rsl/projects/inventory/sampling.shtml

Top of page

Updated page content 03/07/05
Updated html code 02/24/04

USDA Forest Service · Pacific Southwest Region