LAS Dataset to Raster: How do I determine the sampling size?

[DonaldPirie-Hay1]

I am attempting to create a DEM and DSM from a LAS Dataset. I know how to filter the appropriate returns for either the DEM or the DSM, but I am unsure what the sampling size should be (along with Sampling type: Cellsize).

From the metadata I've been provided, the lidar (LAS) files have a horizontal resolution of 0.31m. However, the average point spacing is about 1.3 (for each LAS file, there are only two). Should my sampling size be 0.31 or 1.3?

Any help is greatly appreciated.

Thanks

[JakeSkinner]

Hi Donald,

If you are adding the LAS data to a mosaic dataset, take a look at steps 6 & 7 in the below help document:

Adding lidar data to a mosaic dataset—Help | ArcGIS for Desktop

[DonaldPirie-Hay1]

Thanks Jake,

Quick question on how that relates to actual units and cell size.

I will take the average point spacing of 1.3, and multiply that by 4, resulting in 5.2. That will then be put in to the field for Sampling Size.

These units are in meters. Does that mean that a cell will be 5.2 meters square, or each side is 5.2 m long, thus resulting in a cell that is 27.04 meters square?

Thanks again,

Donald

[JakeSkinner]

These units are in meters. Does that mean that a cell will be 5.2 meters square, or each side is 5.2 m long, thus resulting in a cell that is 27.04 meters square?

Yes, that is correct.

[DonaldPirie-Hay1]

Sorry but it was more of an A or B question.

A) The cell is 5.2 meters square

B) The cell is 5.2 meters by 5.2 meters = 27.04 meters square.

[JakeSkinner]

That was my mistake, I didn't mean copy all of that statement.  The answer is B, the cell would be 27.04 square meters.

[IanMurray]

From most of the scientific papers I have read, generally it is acceptable to generate cells of the same dimensions as the average point spacing of the Las file.  That is if your average point spacing is 1.3, then it would okay for your cell size to be 1.3m by 1.3m(1.69m^2).  Otherwise I would say you are very much under utilizing your data, since you will be having ~20 points being used per pixel instead of ~1.

Donald, have a look at page 5 of this ISPRS paper, it should explain why the smaller cell size is acceptable.

[CodyBenkelman]

Hi Don
sorry for the very slow response - you may have experimented and moved on by now, but I wanted to answer.

We've just published a rather detailed document of best practices for Lidar at this link  Image Management..  It includes a discussion of deciding sample size.

For a quick answer, I typically recommend you make your bare earth DTM no finer than 4 points per cell (or 2x point spacing), e.g. if point SPACING is 1.3 meters, that would be cell size ~ 2.6 meters (thus 4 samples in your DTM cell looking at 2 dimensions).  However, your *average* point spacing will not account for areas with higher and lower point density - so in heavy canopy your spacing will be more coarse, and in truly bare areas your spacing will be better than the average value.  I recommend using the GP Tool "Point File Statistics as Raster" and summarize using PULSE_COUNT to get a raster map of your outgoing pulse density, and then using POINT_COUNT to get a raster showing point density - then you can see areas that might have thin coverage.  (one other detail - I said "density" above but note if you summarize this raster at 2.6 meter cell size, you'll get pulses or points per [2.6]^2 square meter - so if you want an actual numeric measure of density per square meter you have to divide each value by 6.76 to normalize)

I hope that helps!

Cody B

...oops, I didn't scroll down so I didn't realize Jake Skinner had already answered...  Thanks Jake.  I also didn't answer re: DSM; that's up to you but many users go as fine as 1 LAS point per cell, just to see the maximum structure - but it'll be noisy, so I usually make my DSM and DTM the same resolution just for consistency.