libera_utils.footprint_matching.aggregation.weighted_log_mean

libera_utils.footprint_matching.aggregation.weighted_log_mean(values: ndarray, weights: ndarray, **_: Any) → dict[str, float]

PSF-weighted geometric mean exp(sum(w*ln x)/sum(w)).

Used for pixel-level cloud optical depth, which CERES averages in log space: “Cloud optical depth is averaged in logarithm form, since the logarithm of the cloud optical depth is proportional to the visible radiance” (CERES FluxByCldTyp Ed4A Data Quality Summary, p. 9, https://ceres.larc.nasa.gov/documents/DQ_summaries/CERES_FluxByCldTyp_Ed4A_DQS.pdf), and the SSF records the PSF-weighted mean of the natural logarithms of the pixels’ visible optical depth (SSF-85, CERES SSF Collection Guide, Release 2, p. 76, https://ceres.larc.nasa.gov/documents/collect_guide/pdf/SSF_CG_R2V1.pdf). This strategy returns the exponential of that mean, so the value stays in optical-depth units. Only strictly positive values contribute – ln is undefined at and below zero – so those samples are dropped in addition to the usual NaN exclusion. Its _Standard_Deviation companion is weighted_positive_std(), over the same positive samples.