Structural causal model over the exact covariate column names that
appear in the v1.3.1 case-study bundle (WoSIS 0-10 cm topsoil SOC +
SoilGrids + WorldClim + SRTM + WorldCover). Unlike
causal_cerrado_dag(), which uses short schematic labels
(elev, slope, twi, map_mm, ndvi, soc), this DAG is
wired against bio1, bio12, soilgrids_clay,
wc_landcover_trees etc. so that causal_estimate_effect() can
consume the real profiles data frame without renaming.
Value
A dagitty DAG whose nodes match the column names of the
profiles data frame in
inst/extdata/case_cerrado_results.rds.
Details
The edges encode six classes of Cerrado pedogenetic relations:
- Relief -> climate
Elevation modulates temperature (adiabatic lapse) and precipitation (orographic forcing) via
elev -> bio1,elev -> bio12,elev -> slope.- Climate -> vegetation / land cover
Bio1 (mean annual temperature) and bio12 (mean annual precipitation) drive the fraction of land covered by trees, grassland and cropland.
- Relief -> texture
Steep slopes export fine fractions (
slope -> soilgrids_clay) and accumulate coarse fractions downslope (slope -> soilgrids_sand).- Texture -> bulk density
Fine-textured soils compact differently (
soilgrids_clay -> soilgrids_bdod).- Climate + texture -> SOC (direct)
Both sides drive decomposition vs mineral protection.
- Land cover -> SOC
Native savanna vs. pasture vs. cropland produce 3-4x SOC differences in Cerrado topsoil; the land-cover fractions are the dominant single factor.
See also
causal_cerrado_dag() for the short-label schematic
version; causal_adjustment_set() and
causal_estimate_effect() for identification + estimation.