Read monthly cropland soil cover from an LPJmL run's crop calendar.

Per-cell, per-month share of a cell's cropped area that is under a growing crop, derived from LPJmL's sowing (sdate.nc) and harvest (hdate.nc) dates and area-weighted by cftfrac.nc. Supply the result as data$cropland_cover to build_carbon_balance() to place the cropland soil-cover season where the crops actually are.

Usage

read_lpjml_crop_cover(
  run_dir = NULL,
  years = NULL,
  first_year = NULL,
  by = c("cropland", "regime"),
  example = FALSE
)

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) uses WHEP_LPJML_RUN_DIR.
years: Optional integer vector of calendar years to keep. NULL (default) keeps every year the files cover.
first_year: Calendar year of the files' first time step. NULL (default) reads it from the file's own time axis.
by: "cropland" (default) pools every calendar band into one cover per cell and month. "regime" keeps the rainfed and the irrigated bands apart – a band is irrigated when its name starts with irrigated – and adds regime and cropped_frac, the regime's share of the cell, so a consumer can pool the two back by area. Rainfed and irrigated stands of one crop sow and harvest on different dates; the per-regime layer is what the herbaceous rainfed and irrigated crop groups of build_carbon_balance() (crop_groups = list(method = "rotation_groups")) read, each from its own bands, while plain cropland reads the pool.
example: If TRUE, return a small fixture instead of reading a run. Defaults to FALSE.

Value

A tibble with lon, lat, year, month and cropland_cover (a fraction in 0-1); with by = "regime", also regime ("rainfed" or "irrigated") and cropped_frac, and up to two rows per cell-month.

Why the temperature proxy is not good enough

Without a calendar, soc_soil_cover_curve is anchored to each cell-year's warmest month as a stand-in for peak canopy. Measured against this run at 2010 over 18,548 cropland cells, the real area-weighted crop mid-season falls in the warmest month in only 5.2% of them, within one month in 22.6%, and three or more months away in 51.0% (median absolute offset three months). Winter cereals, Mediterranean systems and irrigated dry-season crops all grow away from the temperature peak, so the proxy puts modelled full canopy over real fallow and modelled bare soil over the real crop. It is a timing error rather than a level one: the curve's annual mean cover is 0.254 against the calendar's 0.343.

What the calendar does not cover

sdate/hdate carry 24 bands - the twelve named crops, rainfed and irrigated - while cftfrac carries 32. The others and managed-grassland bands have no calendar, which is 1.2% of cropped area; cells are returned with the cover of the area that does have one, and a cell whose cropped area is entirely others yields no row rather than a guess.

Band mapping is by NAME, never by index: only 12 of the 24 line up, because sdate band 13 is "irrigated temperate cereals" where cftfrac band 13 is "rainfed others".

Examples

read_lpjml_crop_cover(example = TRUE)
#> # A tibble: 12 × 5
#>      lon   lat  year month cropland_cover
#>    <dbl> <dbl> <int> <int>          <dbl>
#>  1  0.25  45.2  2010     1              0
#>  2  0.25  45.2  2010     2              0
#>  3  0.25  45.2  2010     3              0
#>  4  0.25  45.2  2010     4              0
#>  5  0.25  45.2  2010     5              1
#>  6  0.25  45.2  2010     6              1
#>  7  0.25  45.2  2010     7              1
#>  8  0.25  45.2  2010     8              1
#>  9  0.25  45.2  2010     9              1
#> 10  0.25  45.2  2010    10              0
#> 11  0.25  45.2  2010    11              0
#> 12  0.25  45.2  2010    12              0