LIDAR Format Letter

Open Letter for the need for Open Standards in LiDAR

March 2015.

We, the undersigned, are concerned that the current interoperability between LiDAR applications, through use of the open "LAS" format, is being threatened by ESRI's introduction and promotion of an alternative "Optimised LAS" closed format. This is of concern as it reduces interoperability between applications and organisations, and introduces vendor lock-in.

We request that:

1. The OGC initiate the formalisation of an open standard for storing LIDAR data, and OGC sponsors prioritise the development of this LIDAR standard.
2. ESRI support the OGC in their mission "to advance the development and use of international standards and supporting services that promote geospatial interoperability." In particular, join the OGC in developing of an Open Standard for use of LIDAR data. This might include proposing ESRI's "Optimised LAS" as an Open Standard, and removing technical and legal hurdles to use of "Optimised LAS" as an Open Standard. A simple test to determine if "Optimised LAS" can be used as an Open Standard would be if "Optimised LAS" can legally be implemented by Open Source software such as LibLAS or LASzip.
3. Users and sponsors of LIDAR data, and the LAS Working Group (LWG) which is part of ASPRS, publicly state their preference for the use of an open format over closed when selecting software and services.

Signed

Name, Affiliation(s), Optional comment on interest in Open LIDAR format

1. Suchith Anand , Geo for All , committed to Open Principles in Geo Education and Policy.
2. Cameron Shorter, GeoSpatial Director at LISAsoft, Core contributor and coordinator of OSGeo-Live, Contributor to numerous OGC testbeds, technical lead on a range of previous Australian and New Zealand Open Government initiatives.
3. Stefan Keller, founder and director of Geometa Lab; researcher in GIS, databases, open (government) data and interoperability; maintainer of GeoConverter; contributor to open source software (GDAL/OGR, QGIS).

Background

About LiDAR

LiDAR (Light Detection And Ranging) is a form of high precision range measurement unit much like a radar system that uses laser light instead of electromagnetic waves. The standard LiDAR product is a 3D point cloud that can be conceptualized as a series of point measurements representing distance between the sensor to a returned emission. A common storage format for these point cloud data is the LAS format.

To date, [April 2015] there has been a common format for storing LiDAR data, the “LAS” format, and two open source libraries are available, libLAS and LASlib, that can read and write the format. Both these libraries have been incorporated within many LiDAR applications, allowing read/write access to a common exchange format, and resulting in full interoperability between these applications.

LAS limitations

Quoting Paul Ramsey:

LAS format is not without its drawbacks:

While it is a binary format and does not waste any space unnecessarily, neither does it apply any compression to the data it stores. That’s not good for archival use.

Also, LAS stores points in scan order, so accessing any particular chunk of points involves reading the whole file. That’s not good for random access.

Clearly there is a little more work to be done. Can LAS be improved? In fact, it already has been:

• An open source compression library LASzip can apply 20:1 lossless compression to LAS files, making them great for archival purposes.
• Other LAS users have experimented with re-ordering points in a LAS or LASzip file to allow random access to internal chunks of the LIDAR point cloud.

Basically, making LAS smaller and faster is not rocket science, and if the work were incorporated into libLAS then the whole LIDAR community could leverage it together, and the user community would only have one file type to interchange.

Closed "Optimized LAS" format from ESRI

ESRI has announced the release of an "Optimized LAS" format which is claimed to provide faster access and smaller file sizes and released a free tool which is claimed to convert LAS files to and from "Optimized LAS". This announcement created a outburst of vocal protest in the LiDAR community [1 2 3 4 5 6]. One year later ESRI released a free Windows DLL that can be used to read LiDAR data from their proprietary format as well as generate it.

However, the "Optimized LAS" format is neither published, nor available under any open license, which provides both technical as well as legal barriers to other applications reading and/or writing to this proprietary format. This creates a vendor lock-in scenario which is contrary to the principles of the Open Geospatial Consortium, the OSGeo Foundation, and many government IT procurement policies.

OGCs efforts so far to enable Open standards in LiDAR

Carl Reed (formerly on OGC staff) provided the following information on OGC's previous efforts to work towards enabling Open Standards in LiDAR

"Over the last 8 or so years, the OGC approached ASPRS at least twice regarding LAS and working with ASPRS to bring LAS into the OGC for consideration as both a Best Practice and an OGC standard. OGC member Rick Pearsall worked hard to try to make this work. Rick worked for NGA, was active in the OGC, and was also the Standards Committee chair at ASPRS. Rick and Carl tried and failed. These attempts go back to at least 2007.http://www.asprs.org/a/society/divisions/ppd/ppd_meetings/2007springppdreport.pdf. More recently, Carl Reed had an email dialogue with Lewis Graham about bringing LAS into the OGC as an OGC Best Practice. Carl thought progress was being made, but then for some reason all communication stopped."

There is current interest by the OGC in pursuing point cloud encoding standards, including a member-initiated mechanism to extend LAS data with OGC-standard XML content. The OGC invites interested members who wish to work on this effort to please contact Scott Simmons (Executive Director, Standards Program E-mail : ssimmons@opengeospatial.org ) to register their interest and discuss details. OGC will also be holding an ad hoc session at our next Technical Committee meeting in Boulder, CO, USA in early June to bring together all interested from all sectors (government, industry, academia) for this and plan next steps.

Some History on LAS and "Optimised LAS"

The original development of the LAS format started in 1998 according to Lewis Graham. The effort was at first led by pioneers of the LiDAR industry until the format was donated to the ASPRS. Since then the LAS Working Group (LWG) has been maintaining the LAS format, guiding it from the initial LAS 1.0 version until today's LAS 1.4 version. This effort created an incredible successful open data exchange standard for discrete LiDAR points that is nowadays supported by practically every LiDAR software.

ESRI did not join the LWG of the ASPRS until rather late, the 24th of August in 2011 to be exact. At that time ESRI was not really a "player" in the LiDAR market as they did not have much support for LiDAR in any of their products. That was to change soon as they were planning to add LAS as a native type in ArcGIS 10.1.

Shortly before that, in June 2011, Martin Isenburg received a personal message from the ESRI development team that made him quite happy: "I have a question about LAS compression. I’m evaluating some potential enhancements and support for LAZ is one of them. Compression time and amount is impressive." When inquiring two month later about the status of LAZ integration Martin was told: "We’re in beta and working to finish the current release. I very much doubt LAZ will make it in because it’s too big a change at this time. So, we’ll be considering, for the following release, what we want to do regarding compression and spatial indexing."

In June 2012 ESRI released ArcGIS 10.1 and introduced the "LAS Dataset file (*.lasd)" that would group collections of files into one logical unit. That was a useful addition but unfortunately ESRI was not inclined to share this improvement with others despite several private and eventually public requests.

However, there were encouraging follow-ups about compression in January 2012: "I would be interested in having a more in depth meeting with you on better understand the great work you are doing on with LAS and how we can possibly better partner on this." and again in December 2012: "If you have time, I'd like to set some time aside Tuesday afternoon to meet with <an imporant person> at the Esri booth. Does 2pm sound okay?" The meeting went well and it looked as if ESRI was going to embrace the de-facto open standard because shortly after the meeting Martin got word that: "I hear from <an imporant person> that it was a success and that he had a good meeting with you. I wanted to see if we can have a telephone discussion on Wed or Thu this week related to the potential of incorporating LAZ into ArcGIS."

Then the first signs of hesitation showed. First there were legal issues raised in February 2013: "Currently the legal aspects are being reviewed. They were having some issues separating the LPGL aspects from LASzip from the remainder of LAStools etc." and then in April 2013 concerns about the code were made: "We have started to look into the integration of LAZ, but come across some issues. We don’t want to copy files from the other package, mix and match or hack around. We were hoping for an API to stream points out of a LAZ file as well as write LAZ files. There should be a simple code sample for that?" and - following up on that - in June 2013 Martin was told "I see value in LASzip becoming a de-facto standard that provides compression to the LAS format. If this to happen then it needs to be bundled together with appropriate reference implementations, examples and documentation as a standard and so ensure that issues as defined earlier in mail to not occur."

At that time LASzip was an open source project without sponsorship. The original funding from USACE that had turned LASzip from an academic prototype into an industry strength compression engine had long run out. So right then Martin asked ESRI to become a sponsor to create the kind of clean, well-documented, and easy-to-use API for LASzip that ESRI was after. The answer was: "Esri often helps in the financing of Open Source projects and I could foresee Esri possibly helping in LASzip. [...] If you are interested in promoting LASzip as such a standard and doing the required work then I would recommend you put together a proposal and I can look to get Esri as a sponsor." Immediately Martin proposed the following to ESRI: "I hereby propose to write an easy to use open source DLL wrapper for LASzip that will make it easier to integrate LASzip in a standardized manner into other software products such as ArcGIS, LP 360, or Terrasolid. This will come with example code on how to use the DLL for reading and writing LAZ files and inlude a few compressed example files. I hereby ask ESRI to provide funding for this effort without imposing any limitations on the produced DLL API." All this happened in June 2013.

In July 2013 ESRI added another big improvement to their LAS file handling with the release of ArcGIS 10.2, namely spatial indexing to speed up area-of-interest queries. Just like the LAS Dataset file (*.lasd) this was done with a closed format. Their new proprietary (*.lasx) files have a practically identical look and feel as the open source spatial indexing (*.lax) files introduced a year earlier and published at the European LiDAR Mapping Forum (ELMF) in December 2012.

After more requests (in particular for USDA's FUSION) for an easier interface to LASzip Martin decided not to wait for funding from ESRI but "code it forward". On July 30th of 2013 Martin announced the release of a clean well-documented, and easy-to-use DLL with the expectation that developers at ESRI would use it to add read and write support for LAZ to their next release of ArcGIS.

All this changed suddenly in December of 2013 when several LAStools users contacted Martin with suspicions that ESRI may be working on their own proprietary LAS compression. Martin broke the news as soon as it became evident that ESRI had used the time it needed to resolve "legal issues" and "code problems" to put together their own proprietary compressed format with almost near-identical performance and functionality. Theit LAS compression scheme (that has since been dubbed the "LAZ clone") comes with one additional feature that LASzip is obviously lacking: it is ESRI-controlled technology.

Caught by an instant "shit-storm" (see comments) by the LiDAR community, an official FAQ was released by ESRI a few days later confirming the rumors. From day one Martin has taken a very outspoken stance, informing the community about ESRI's actions and lobbying against the "LAZ clone" using fun-filled imagery of "laser battles" and "clone wars". Martin tried to keep his critical stance light-hearted in order not to burn bridges with ESRI. At the same time Martin - again and again - engaged in direct personal communication with decision makers and product managers of ESRI pleading to work with the LiDAR community and avoid format fragmentation.

Martin's core argument for resolving this to everybody's benefit was that - coincidentally - a natural break was just happening in the LAS format with the introduction of the new LAS 1.4 version. Martin repeatedly outlined a detailed plan of how a joint development of LASzip for LAS 1.4 by rapidlasso and ESRI could exploit this natural break in the format to accomplish two things at once: add whatever additional features ESRI needed the compressed format to have (a vaguely uttered justification for their "LAZ clone") and also extend the LASzip compression scheme to handle the new point types introduced with the LAS 1.4 specification. There was no pressing need at the time to handle LAS 1.4 so Martin delayed the extension of LASzip to the new LAS 1.4 point types to make sure a cooperation with ESRI would remain a viable option.

Unfortunately this fell on deaf ears. In a final attempt to convince ESRI management and show them how positively the community would react to news of them embracing the open de-facto compression standard we released a prank press release on April Fools' Day. The press release was almost true except for the fact that ESRI had not agreed to the deal yet. Large parts of the community fell for the prank and the overwhelming positive reactions (see comments) leave no doubt about the sentiment on the issue in the LiDAR communitiy. The other objective of this prank press release was to get coverage about this controversy in the geospatial media.

In contrast to the entire "LAZ clone" controversy before, this press release was carried by practically every geospatial news outlet. Despite being one of the most talked about LiDAR topics of 2014 the geospatial media did not cover this story once. The extreme bias in sponsored news reporting is especially evident for LiDAR news who dutiful reported whatever positive bit they were told about "Optimized LAS" [1 2 3 4 5 6 7] but none of the raging controversy. The self-censorship of the geospatial news media who are obviously hesitant to "bite the hand that feeds them" limits the exposure that ESRI's lock-in tactics have been getting.

Martin was able to win Digital Coast, NOAA Coastal Services Center as a gold sponsor for LASzip and announced in October 2014 to develop the LAS 1.4 compatibility mode that would allow to add immediate support for the new point types to LASzip without closing the door on a potential cooperation with ESRI for a joint LAS 1.4 compressor by utilizing a clever recoding of new point types into old one.

The final signal from ESRI that they are absolutely not willing to cooperate on an open compression standard came in November 2014 with the announcement that ESRI had added their own extension for the new LAS 1.4 point types to the "LAZ clone". This was ending all hopes to devise a joint compressor and avoid format fragmentation by exploiting this "natural break" in the LAS format as Martin had suggested to ESRI both seriously in private messages and also funnily in public.

Value of Standards

The importance of Open Standards is described is most government IT policies. For instance, the United Kingdom policy states:

… Government assets should be interoperable and open for re-use in order to maximise return on investment, avoid technological or supplier lock-in, reduce operational risk in ICT projects and provide responsive services for citizens and business. This should also lower barriers to entry for more diverse sources of IT services, including citizens and SMEs.

The value of Open Standards have been described in numerous national studies on the effects of standards on economic growth.

... the national studies demonstrate that standards have a positive influence on economic growth due to the resulting improved diffusion of knowledge. The contribution of standards to the growth rate in each country is equivalent to 0.9% in Germany, 0.8% in 0.3% in the UK and 0.2% in Canada.

See Also

1. Paul Ramsey provides background to LAS vs Optimised LAS, http://boundlessgeo.com/2014/01/lidar-format-wars/
2. Running commentary by Martin Isenburg, author of LASlib, http://rapidlasso.com/2015/02/22/lidar-las-asprs-esri-and-the-laz-clone/
3. Earlier comment from Martin Isenburg, http://rapidlasso.com/2014/11/06/keeping-esri-honest/
4. First call-to-action by Martin Isenburg, http://rapidlasso.com/2013/12/30/new-compressed-las-format-by-esri/
5. ESRI Announces "Optimised LAS", http://blog.lidarnews.com/esri-announces-las-compression/
6. ESRI description of "Optimised LAS", http://www.lidarnews.com/content/view/10214
7. Discussion background on this topic at Geo for All list , http://lists.osgeo.org/pipermail/ica-osgeo-labs/2015-March/001225.html