How to create a DSM and a DEM from las dataset alone?

[TedRakel]

My client wants to keep lidar data in a las dataset and not use terrain datasets or mosaic datasets.  They also do not want to create multipoint features in the geodatabase from the las files.  I need to create a DSM and a DEM from the las dataset alone.  I have read the excellent esri paper available here : http://www.esri.com/library/whitepapers/pdfs/lidar-analysis-forestry-10.pdf   This paper gives great adivce on how to create a DSM and a DEM but it requires creating mulitpoint features in the geodatabase, which my client does not want.

Is there a way to create a DSM and DEM from a las dataset directly without creating mulitpoint features or using terrain or mosaic datasets?  I'm using arc gis desktop 10.2.1.

Thanks

[IanMurray · Accepted solution]

Once you import your .las files into a Las Dataset(.lasd), you can control what return values you want the las dataset to use and which point types in the Las Dataset properties, or you can make a Las Dataset Layer file, with only the return types you want.  For a DEM, you would want to use only returns classified as ground(I sincerely hope they are already broken down into ground/non ground) in your las dataset, then can use the Las Dataset to Raster tool, for a DSM, you would want only First returns in your Lasd Dataset, since it the DSM is showing the return of those surfaces that are hit first by the lidar pulse.  Choose your value field(elevation in this case) and interpolation type(see the help about selecting one).

Output cell size is a sticky topic, a fair number of published journals have recommended you can use a cell size that is equivalent or so to the nominal point spacing of your lidar data.  ESRI suggests using a cell size 2-3 times larger than the nominal point spacing, but from personal experience, using the same as the nominal point spacing works well.

Making Las Datasets ArcGIS Help 10.1

Las Dataset to Raster ArcGIS Help 10.1

[NeilAyres]

Just to underline what Ian has said, once you have imported the las files into a lasd dataset, you can check the various statistics in ArcCatalog.

Then in ArcMap, layer properties, filter, you can set which returns you want. Look at the data carefully for badly classified points. Once happy, save this as a layer file and use as the input into the LAS Dataset to Raster (in Conversion tools).

[TedRakel]

Thanks for you help on this.  I've been experimenting with the 'LAS Dataset to Raster' GP tool.  There are a lot of options.  The LiDAR data that I have is of a flight over a transmission line.  So it's long and thin and piecewise linear.  Here's a screen shot of the vegetation points in arc map (about 3.5 million points)

When creating a raster from this, I selected 'Binning' for the interpolation type.  For the void fill method, I've tried 'none', 'simple', 'linear', and 'nearest neighbor'.  'nearest neighbor' ran overnight without finishing so I don't know how well it works, it's just not practical with this data.  With 'linear' I get this:

With 'simple', I get this

I think 'linear' is the best option except it ends up void filling way outside the area of the actual lidar samples.  Is there any way to restrict the void filling when using 'linear' to within a polygon or to prevent the void filling from extending so far?  For cell size I used 3; the las dataset properties report the point spacing as 2.78.

[TedRakel]

Looks like this was addressed in the 2014 UC video here: http://video.esri.com/watch/3862/arcgis-3d-analyst-lidar-applications

I needed a surface constraint derived using the delineate data area model described in the video.  Now my DSM and DEM look much better.

[JoshuaBrengel]

The surface constraint discussion starts around 30:00 in the video.