Reads the long-format BDsolos CSV (one row per horizon, with a
profile-id key) and returns a list of PedonRecord
objects. Auto-detects the column-name convention via
inspect_bdsolos_csv and maps to the soilKey horizon
schema. Texture (argila / silte / areia) is converted from g/kg to
percent (BDsolos canonical unit).
Value
A list of PedonRecord objects. Each pedon
has site$id from the profile-id column, the
taxonomic reference (when present) at
site$reference_sibcs, and one horizon row per CSV
row matching the profile id.
Details
Iron (v0.9.221): "CDB - Ferro (g/kg)" (dithionite-citrate-
bicarbonate, free Fe as the element) goes to fe_dcb_pct, which
the WRB Fe-dith criteria read (nitic horizon, Ferric, Ferritic);
"Ataque sulfurico - Fe2O3" (total Fe of the sulfuric attack, as
the oxide) goes to fe2o3_sulfuric_pct, which the SiBCS
ferrico / perferrico classes read; "Oxalato de Amonio - Ferro"
goes to fe_ox_pct. All three are converted from g/kg to %.
Until v0.9.220 the sulfuric Fe2O3 filled fe_dcb_pct, unconverted.
Silica and alumina (v0.9.243): "Ataque sulfurico - SiO2" and
"- Al2O3" go to sio2_sulfuric_pct and
al2o3_sulfuric_pct, converted from g/kg to %. The Ki index
(molecular SiO2 / Al2O3), which SiBCS asks of the B latossolico, is
computed from them; it reproduces the Ki column of the export.
Text of the horizon description (v0.9.245): the column
"Descricao Original" holds the field description of each horizon
as the survey wrote it. Two things are read from it where the structured
columns are empty, and the text itself is not kept: the amount of
slickensides written beside the term ("''slickenside'' comum e
moderado" fills slickensides with "Comum"; a mention with
no amount fills nothing), and a wedge-shaped or parallelepipedic
structure ("cuneiforme", "paralelepipedica", added to
structure_type; a structure "com tendencia a paralelepipedica" is
not one). SiBCS knows the vertic horizon by these two features, and the
surveys that predate the structured sheet describe them only there.
Since v0.9.251 the amount and the grade of the cerosidade are read the
same way ("cerosidade comum e moderada" fills
clay_films_amount with "Comum" and
clay_films_strength with "Moderada"), each where its own
cell is empty.
Reading (v0.9.222): the file is read with utils::read.csv2. The
export quotes every field, and its free-text fields hold line breaks and,
in some states, quotation marks of their own; read.csv2 returns
the export's records for all 27 state files. data.table::fread,
used until v0.9.221, split such records at the line breaks (in BA, GO, PI
and RS, where the lines of text became profiles of their own) or stopped
with an error that left the next fread() of the R session waiting
forever (after DF, MT, PA, PB or SP). An empty text field is NA in
every file; a quotation mark inside a text field is dropped.
Kind of sampling point (v0.9.242): the column "Tipo" of the export
goes to site$sampling_type, as written: "Perfil completo",
"Perfil parcial" or "Amostra extra". An extra sample is, in
most cases, two sampled layers of a point (a topsoil and a subsoil sample,
often with a gap between them), not a described profile: the thickness of
its layers is that of the samples. About a third of the records of the
national export are extra samples; benchmark_bdsolos
reports the kinds apart.
Profile-id columns are auto-detected: looks for any column whose
normalised name matches
"id_perfil|profile_id|cod_perfil|^perfil$|sample_id|^id$";
falls back to the first column when none match.