WG - Lidar - Bathymetry - Public

As systems and processes evolve and advance, we continue to see new approaches to identifying the bottom. Traditional peak detection methods continue to be used where we have the data but we are now seeing image approaches, AI strategies, turbid water, shallow water, waveform stacking, area averaging, and more. 

USACE utilizing the CZMIl sensor  must enable the shallow water algorithms in less than 0.5m and we do enable a turbid water algorithm in some environments. The turbid water approach does extend coverage but also increases editing and QC time leaving the fun to USACE to validate and pass to class 40. There are "fuzzy" areas, I think Dewberry referred to as manatee poop from our last call, where this type data may or may not be the bottom or a less confident bottom than the more confident interest point methods but this data is not passed in our standard product as we are uncertain about it's certainty. 

Would the community support the wiki with a simple description of the algorithms and should we start storing this information within the las files?

06/09/2025 meeting notes - Please feel free to add comments and support clarifying our needs.

I shared CZMIL bathy lidar data and some of what I noted as "fuzzy" turbid water algorithm data. I also showed bathymetry class 40 data which used class 41 detected surfaces and ones that used class 42 modeled surfaces as well as shallow water algorithm returns which use a modeled class 42 surface. All of these bathy types are delivered as class 40 bathymetry. This system also has seven shallow receivers and a deep receiver which have unique calibration values on land and in water for some ranging functions. I was proposing modifying/extending the Bathy Domain Profile to include a field to capture the system/user algorithm applied to produce the class 40 bathy return. Even if proprietary algorithms can't be shared, the classifying or flagging of this data is often shared to the data producer  and can be useful for the end user. Example might be wanting to use traditional peak detection returns with high confidence for  charting purposes and feature detection but also wanting to share in the las files data extracted from waveform averaging or specialized turbid extractions that might be applicable for general bathymetry but may not fit all of the system specified accuracies or feature detection capabilities.

I didn't intend to confuse the conversation with our prior months call regarding classification confidence which as we begin to use more advanced extraction/classification techniques and AI are worth considering adding a classification confidence item to the las files. Example from that call was from ICESAT bottom classifications where rather than splitting the returns below the surface into class 40 bathy bottom and proposed class 49 low confidence bathy bottom, the confidence values of the algorithms themselves could be stored with the data for other users to determine thresholds.

Many notes on past bathy domain profiles to las,see below, and comments on this call remind us we have had figure of merit/SNR fields before but they are rarely requested or used outside of during data production. 

There is a little room for circular logic/room for confusion between class uncertainty and positional uncertainty when we discuss different algorithms used, even within a system, as we to discuss how signals and algorithms for extraction factor into positional uncertainty. System A with a per pulse central peak detection in clear waters in one area of a flight versus using a 3m local averaging of neighbors to extract depths in turbid areas should have differences in the horizontal uncertainty  estimated. Riegl showed at the 2024 JALBTCX Workshop in AK the factoring in of waveform metrics into the vertical uncertainty calculated so we are moving in that direction already. 

This conversation needs to be organized and maybe we can outline this in our Wiki. Please share your thoughts, ideas, and concerns here and we can continue this conversation on future calls.
 
  

LAS Working Group domain profile issue thread

https://github.com/ASPRSorg/LAS/issues/117

LAS Domain Profile (LDP) Description:Topobathy Lidar Version 2.0

https://github.com/ASPRSorg/LAS/issues/117

------------------------------
Nicholas Johnson
Physical Scientist
USACE
Kiln MS
------------------------------