All reasonable efforts have been made to remove data affected to any degree by questionable
observational conditions, however questionable data likely are still present. Here are the
currently-identified issues with the LVIS dataset:
Canopy height bias.
- Low-lying ground fog was present in some areas (swamps, river valleys etc). This will cause
canopy height parameters (e.g., zt, rh25, rh50, rh75) to be higher than in reality.
- Canopy return misselected during data processing. Weak canopy top returns or spurious
higher-amplitude background noise samples were misselected during data processing.
- Misselected ground return. Obviously, if the ground location has been misselected then canopy
height products will be biased. Data product zt is unaffected since its elevation is relative to
the eliposoid, not the ground.
Ground elevation bias.
- Ground return is not contained within the waveform and thus cannot be identified.
- Ground return misselected during data processing, caused by:
- A mode from a higher reflecting layer than the ground was misselected as the
ground return. This misselection usually implies the ground reflection is "weak"
(i.e., contains only a small proportion of the reflected energy from the footprint
as a whole) and thus has been "overlooked" by the interpretation algorithm.
- Spurious, higher ampitude background noise was misselected as the ground return.
- Ground return is indistinct, i.e., the lowest reflection has become convolved with reflections
from higher surfaces. This effect is compounded by within footprint slope and surface roughness.
Improvements in our methods of waveform interpretation will enable better data accuracy.
Digitally recording the shape of the return laser pulse means that these improvements can easily
be applied. The LVIS L1B data product has been structured to provide interested end users the infomation needed to apply
their own interpretation algorithms and regenerate elevation and height products.
