
Read natural-land soil cover from an LPJmL run.
Source:R/lpjml_natural_cover.R
read_lpjml_natural_cover.RdPer-cell, per-year vegetated cover of the natural stand, from LPJmL's
foliar projective cover (fpc.nc). Band 1 of that file is the natural
stand FRACTION – how much of the cell is natural – and bands 2 to 15 are
the fourteen natural PFTs' projective cover WITHIN that stand. The cover of
natural land is therefore the sum of bands 2 to 15, capped at 1, and never
band 1, which is an area share and a different quantity.
Supply the result as data$natural_cover to build_carbon_balance() to
replace soc_soil_cover_curve's constant 0.85 for the natural class.
Source
Schaphoff, S. et al. (2018). LPJmL4 - a dynamic global vegetation model with managed land - Part 1: Model description. Geoscientific Model Development, 11, 1343-1375. doi:10.5194/gmd-11-1343-2018 .
Arguments
- run_dir
Path to the LPJmL run output directory.
NULL(default) usesWHEP_LPJML_RUN_DIR.- years
Optional integer vector of calendar years to keep.
NULL(default) keeps every year the file covers.- first_year
Calendar year of the file's first time step.
NULL(default) reads it from the file's owntimeaxis.- example
If
TRUE, return a small fixture instead of reading a run. Defaults toFALSE.
Why it matters where it does
The constant is close on the global mean – measured cover runs 0.858 in
1901 to 0.884 in 2023 – and wrong in the distribution: the median natural
cell is fully covered at 1.000 while the 5th percentile is bare at roughly
zero. The RothC plant-retainment term is 0.6 + 0.4 * (1 - cover), so a
near-bare cell moves from 0.66 to 1.00, decomposes half again as fast, and
loses about a third of its equilibrium carbon. Those arid cells are exactly
where the model sits furthest from observation. The effect on the global
mean is about 1.8%; per cell it spans 0.66 to 1.10 times.
Managed grassland has no FPC band – fpc.nc covers the natural stand
only – so it necessarily stays on the curve.