alsdb.processing.change.compute_change

alsdb.processing.change.compute_change(store: ALSZarrStore, variable: str, year_from: int, year_to: int, resolution: float, *, min_delta: float = 0.0, pct_min_abs: float = 0.0, overwrite: bool = False) → None

Compute per-pixel change between two survey years for variable.

Reads both time slices from store and writes three derived variables:

• {variable}_delta : absolute change year_to − year_from, same units as the source variable. NaN where either year has no data.

• {variable}_delta_pct : relative change in %. NaN where |year_from| < pct_min_abs (avoids division by near-zero and suppresses meaningless percentages for very small base values).

• {variable}_change_flag : +1 gain, −1 loss, 0 no significant change. Pixels are flagged only when |delta| > min_delta.

Parameters:

• store – ALSZarrStore containing the source variable.

• variable – Variable name, e.g. "chm", "lai", "biomass".

• year_from – Survey years to compare. Both must already be present in store.

• year_to – Survey years to compare. Both must already be present in store.

• resolution – Resolution group in metres (e.g. 1.0 or 10.0).

• min_delta – Minimum absolute change flagged as significant (default 0). Set to e.g. 0.5 m for CHM to suppress sub-pixel noise.

• pct_min_abs –

  Minimum absolute value of year_from required for a valid relative- change estimate. Pixels where |year_from| < pct_min_abs receive NaN in delta_pct. Choose a value appropriate for the variable:

  • CHM (m): pct_min_abs=0.5 suppresses extremes in open areas.

  • LAI / gap fraction (0–1): pct_min_abs=0.05 avoids blow-up near bare-ground cells.

  • AGB (Mg ha⁻¹): pct_min_abs=1.0 avoids extreme % in sparse scrub.

  Default 0.0 reproduces the legacy behaviour (threshold = 1e-6).

• overwrite – Re-compute even if the output variables already exist for year_to.