Condense a ground-plot cloud into an age-binned series

Bins the observations on age and takes one quantile per bin, so every age band contributes once no matter how many plots landed in it. This is the non-parametric alternative to fitting a growth equation through the cloud: it assumes nothing about curve shape, which matters because the shape being tested is ForCS's own.

Usage

growth_bin_observations(
  obs,
  bin = 20L,
  probs = 0.5,
  site = NULL,
  weight = NULL
)

Arguments

obs

A tibble with age and aboveground_c_mg_ha.

bin

Numeric. Bin width in years.

probs

Numeric. Quantile to take within each bin.

site

Optional column name identifying the sampling location. When given, repeated visits to one location are averaged within a bin before the quantile is taken, so n counts locations rather than visits. Errors if the named column is absent, rather than silently skipping the correction.

weight

Optional column name holding a per-observation weight, typically from growth_climatic_weight(). When given, the within-bin quantile is weighted, so plots resembling the modelled landscape carry more of it. n still counts observations; weight reports the weight behind each bin, so a bin resting on many barely-relevant plots is visible as such.

Value

A tibble with age (bin mean), value, n, and weight.

Details

probs = 0.5 (the median) tracks the central tendency of realized stands. Raising it moves the series toward the upper envelope, which is arguably where a fully stocked, single-cohort simulation belongs: the plots span every stocking level, site quality and partial-disturbance history, and whole-plot volume is attributed to the leading species, which holds a median of 69% of the stand here.

Where observations come from a permanent-plot network, pass site so that each location contributes one value per bin. Permanent plots are remeasured on a schedule that reflects program history rather than anything ecological – in the network this was built against, 78% of locations carry more than one visit and some carry thirteen – so treating every visit as an independent observation silently weights each bin toward whichever locations happen to have been revisited most. That is pseudo-replication, and it biases the quantile rather than merely tightening it.

See also

Other growth calibration helpers: extract_landscape_cohort_structures(), growth_add_objective(), growth_auto_window(), growth_best_candidates(), growth_calibration_design(), growth_calibration_partition(), growth_calibration_work_root(), growth_climatic_distance(), growth_climatic_weight(), growth_expand_over_pseudo_species(), growth_factorial_ratio_grid(), growth_fitting_windows(), growth_identifiability(), growth_inflation_factor(), growth_mortality_onset_frac(), growth_plot_palette(), growth_pseudo_species_name(), growth_reference_curves(), growth_reference_inflection(), growth_score_fit(), growth_scoring_for(), growth_smooth_observations(), growth_structure_cell_curves(), growth_structure_cohort_table(), growth_structure_design(), growth_structure_summary(), growth_window_for(), plot_growth_calibration(), plot_growth_candidate(), plot_growth_factorial_sensitivity(), plot_growth_structures(), read_growth_scoring(), read_landscape_cohort_structures(), scale_linetype_growth_reference(), write_growth_review_bundle()