L2A — Ground Elevation & Canopy Height¶

Version 2 — User Guide · ATBD Version 3 — User Guide · Data dictionary

Overview¶

GEDI L2A provides ground elevation, canopy height metrics, and relative height (RH) metrics for each GEDI footprint. Data is organized by 8 beams.

Quick Start¶

using SpaceAltimetry, DataFrames

g = granule("GEDI02_A_2020257225350_O09946_02_T08506_02_004_02_V003.h5")
t = table(g)
df = DataFrame(t)

Default Columns¶

Column	HDF5 Path	Type
`longitude`	`lon_lowestmode`	Float64
`latitude`	`lat_lowestmode`	Float64
`height`	`elev_lowestmode`	Float32
`height_error`	`elevation_bin0_error`	Float32
`datetime`	`delta_time`	DateTime
`intensity`	`energy_total`	Float32
`sensitivity`	`sensitivity`	Float32
`surface`	`surface_flag`	Bool
`quality`	`l2a_quality_flag_rel3`	Bool
`nmodes`	`num_detectedmodes`	UInt8
`sun_angle`	`solar_elevation`	Float32
`height_reference`	`digital_elevation_model`	Float32
`strong_beam`	attribute	—

Default Tracks¶

SpaceAltimetry.default_tracks(g)

("BEAM0000", "BEAM0001", "BEAM0010", "BEAM0011", "BEAM0101", "BEAM0110", "BEAM1000", "BEAM1011")

Canopy Heights¶

Use gedi_l2a_canopy_variables(g) to read highest return instead of lowest mode:

t = table(g; variables=gedi_l2a_canopy_variables(g))

Passing the granule picks the quality flag of its product version; without it the version 3 flag is used.

This reads elev_highestreturn / lon_highestreturn / lat_highestreturn.

Quality Filtering¶

filtered = t |>
    GEDI.Quality() |>
    GEDI.Sensitivity(gt = 0.9) |>
    collect

The GEDI.Quality filter auto-loads the receive assessment, stale/degrade, amplitude, and selected algorithm fields required by the full points(g; filtered=true) predicate. The optional GEDI.Sensitivity filter keeps gt < sensitivity <= 1.0; gt defaults to 0.9.

Version 3¶

Version 3 (PGE 004, filenames ending in _V003.h5) is the default since it was released in 2026. It is read with the same schema as version 2, with these differences (GEDI Level 2 User Guide V3.0, section 7.1):

	version 2	version 3
`quality` column	`quality_flag`	`l2a_quality_flag_rel3`: the version 2 conditions plus `sensitivity_a2 > 0.95` over land (`> 0.98` in tropical forest) and `elev_highestreturn − digital_elevation_model < 150` m. The version 2 flag is still in the file as `l2a_quality_flag_rel2`.
algorithm setting groups	1–6	1, 2, 5, 10
`GEDI.Quality` algorithm checks	`rx_processing_aN/{rx_algrunflag, zcross, toploc}` of the selected group	`l2_algrunflag`, which combines the same checks and adds `0 < sensitivity < 1`

gs = search(:GEDI, :GEDI02_A; extent = extent)                # version 3 (default)
gs2 = search(:GEDI, :GEDI02_A; version = 2, extent = extent)  # version 2

selected_algorithm in version 3

In version 3 files, selected_algorithm keeps the version 2 valid_range of 1–6, but shots also use group 10. Because nodata masking honours valid_range, those shots read as missing when you add selected_algorithm as a column. GEDI.Quality uses l2_algrunflag instead, so it is not affected.