e-mail from Bob Peet on 2013-1-24:

-----

One concern I have is that Aaron needs to recognize what constitutes a single plot in the input stream as there is a risk that our BIEN2 database included replicates (both spatial and temporal) as separate plots. There should be a total of around 125,000 plots prior to dropping unwanted plots, so this could be one test.

My second concern is the identification of planted versus natural stems. This is probably well documented, but I have not checked on this.

A third concern is the validity of the geo-coordinates.  Most plots have been randomly offset by up to 1.6 km, but this we can handle with an error term.  More critical is the assertion that "data for a proportion of the private locations have been switched between pairs of plot locations having the same forest type and owner class (but not the same owner) within a county." I do not know what this proportion is, but we may need to assume county-level precision for all private land, assuming we can identify the private holdings.

My fourth concern is the relatively high number of taxonomic problems in the database.  Some identifications are obvious errors (we have several western US trees that are never planted in the FIA inventory for NC).  Some taxa are obviously routinely composites of species but without acknowledgement of the problem. We may need to assess the meaning of the names in each state list as a separate project, which could be a painful undertaking.

A final issue is that we should probably try to capture the complete floristics data that are available for the 1/16th of the plots that are designated as "forest health" plots.  This is a rather recent innovation and not fully implemented yet, but where these data are available, we should try to capture them.  However, the taxonomic challenges for full floristics are enormous relative to trees and we need some assessment as to whether these data are of sufficiently quality to want to bother with.

-----