Classifies NPDB pedons (load_npdb_pedons) with
classify_cssc3 and compares the result with the survey's
classification at the level asked.
Usage
benchmark_npdb(
pedons,
level = c("order", "great_group", "subgroup"),
set_aside = "Chernozemic",
cores = 1L,
verbose = TRUE
)Arguments
- pedons
A list of
PedonRecordfromload_npdb_pedons.- level
"order","great_group"or"subgroup".- set_aside
Reference orders reported apart (
character(0)for none).- cores
Number of processes (
parallel::mclapply; always 1 on Windows).- verbose
Print a summary.
Value
A list: level; accuracy (share of agreement outside
set_aside); accuracy_decided; n_compared;
n_decided; accuracy_all (all orders); set_aside (a
table of where those pedons land); confusion;
per_class_recall; predictions (one row per pedon;
scenario marks pedons classified at the ends of their ranges and
scenario_split those whose versions disagree).
Details
The records hold no soil climate and no dry colour, which the key reads
for the Chernozemic order and its great groups (Ch 3, p. 34; Ch 5, p. 63)
and for the Gray Brown Luvisols; the orders named in set_aside
(by default the Chernozemic) are therefore reported apart, with where they
land, and left out of accuracy. A result is provisional when an
undecided class comes before the one assigned ($provisional_because
of classify_cssc3); accuracy_decided counts only the
results that are not provisional at the level asked or above.
Pedons from npdb_scenario carry estimates as ranges; each
is classified at its central estimate and at every combination of the
ends of its ranges (and of the moisture subclasses between its two
estimates), and a class counts as decided only where every
version gives it and none is provisional (otherwise the central class is
reported as provisional). Run such pedons with
set_aside = character(0) to include the Chernozems.