Canopy height map creation from LiDAR data

General discussions of measurement techniques and the results of testing of techniques and equipment.

Moderators: edfrank, dbhguru

Post Reply
User avatar
addy
Posts: 28
Joined: Fri Feb 24, 2017 9:16 pm

Canopy height map creation from LiDAR data

Post by addy » Tue Jul 17, 2018 12:22 pm

I've been making canopy height maps from LiDAR datasets (using ArcGIS for the rasters & LAStools for preprocessing) and I am wondering what settings do people use when you make the first returns raster? And what is the reasoning behind the choice of settings?

I use binning and maximum for the cell assignment method, that way each cell is given the value of the highest return in that cell. While I think this may slightly exaggerate the size of the areas of the highest returns it makes them large enough to be distinguished and stand out from surrounding lower returns and ensures the highest returns don't get muted and/or left out of the raster by being averaged out or otherwise processed into something lower. I use binning for the same reason, I'm afraid other cell assignment methods may mute, reduce or exclude the highest returns resulting in misleading results & underestimated canopy heights. I also use simple average to fill voids, although I don't think this is a very important setting its more for aesthetics so the raster doesn't have ugly little holes everywhere.

What cell size do you use in relation to the point spacing? I'm currently looking at a 0.7m point spacing LiDAR set and I am thinking ok well if I make the cell size too large it may grossly exaggerate the area of any particular high return but if I make the cell size too small it may reduce the high return cells to being so few & small that they become difficult to distinguish from surrounding lower return cells.

I have dome some limited ground truthing with canopy height maps I've made using binning, maximum and a cell size twice the point spacing and found the canopy height map to match hypsometer measured heights to between 2 to 7 feet.

User avatar
dbhguru
Posts: 4528
Joined: Mon Mar 08, 2010 9:34 pm

Re: Canopy height map creation from LiDAR data

Post by dbhguru » Tue Jul 17, 2018 4:08 pm

Addy,

I hope Michael Taylor has read your post. He's the most knowledgeable, I think, in terms of the kinds of settings you are describing.

Bob
Robert T. Leverett
Co-founder, Native Native Tree Society
Co-founder and President
Friends of Mohawk Trail State Forest
Co-founder, National Cadre

MarkGraham
Posts: 78
Joined: Mon Feb 15, 2016 10:19 pm

Re: Canopy height map creation from LiDAR data

Post by MarkGraham » Tue Jul 17, 2018 4:25 pm

Hello.

Sure you have reviewed this

http://desktop.arcgis.com/en/arcmap/10. ... ctions.htm


So when creating dsm file from las points filtered to first return it is mentioned to use binning as you suggest with cell assignment type MAXIMUM and void fill method NATURAL NEIGHBOR.
Then when creating dem file from las points filtered to ground return it is also binning but cell assignment type is AVERAGE and void fill method is still NATURAL NEIGHBOR.
Then there are further parameters for output data type, sampling type, sampling value, and Z factor. I would go as suggested except up the Z factor. Increasing the Z Factor will give you more accurate crown shapes but may also introduce first returns that are things such as birds or the wheel or the wing tip of the plane so be careful with that. There is a certain redwood I will call "Soaring Bird" that a couple people are familiar with based on my bad use of first return outliers.

After completing the dsm and dem files use the Minus function to subtract dem from dsm. Then the layer created from the Minus function can have Symbology applied to create various numbers of height breaks and color coding.

I have tried this method on redwoods in the west and white pines in the east and it is generally good within 3 feet, based on known height of certain trees. This is true for trees on relatively level ground as well as growing on moderate slopes.

User avatar
M.W.Taylor
Site Admin
Posts: 530
Joined: Sun May 29, 2011 11:45 am

Re: Canopy height map creation from LiDAR data

Post by M.W.Taylor » Sat Aug 18, 2018 3:03 pm

After cleaning up the parasite pixels and outliers, I rasterize in CloudCompare and can find the max and min z value of the entire point cloud (or selected snippet) in the toolbar. To make the highest points jump out visually, I colorize the highest pixels to white and red for the second highest band with a value of 90% of the white max value. Use the pick a point tool to pick the brightest white points to get the z value, use the x and y values to get the GPS coordinates of the point cloud in UTM. The colors blend so you can see the finer height banding. For instance if white is the top band at 100m and red the second band at 90m, the pixels at 95m will be pink.

Other parameters I use for rasterizing include:

cell height - maximum height
scalar fields - average value
empty cells - leave empty

With trees that lean over a hillside, the axis of the trunk relative to the perpendicular offset from base can be estimated using the ruler tool so the lean over the hill inflation is taken into account. The height banding on leaning trees is offset and not centered. This is an easy and quick way to tell if the highest pixel value is going to be over-estimated in ArcGIS. If you get those big holes, try making the pixel size bigger and make the pixels round, not square. The round pixels look better when enlarged and they dither into a more complete and easier to see point cloud. The square pixels when enlarged look wonky.
addy wrote:I've been making canopy height maps from LiDAR datasets (using ArcGIS for the rasters & LAStools for preprocessing) and I am wondering what settings do people use when you make the first returns raster? And what is the reasoning behind the choice of settings?

I use binning and maximum for the cell assignment method, that way each cell is given the value of the highest return in that cell. While I think this may slightly exaggerate the size of the areas of the highest returns it makes them large enough to be distinguished and stand out from surrounding lower returns and ensures the highest returns don't get muted and/or left out of the raster by being averaged out or otherwise processed into something lower. I use binning for the same reason, I'm afraid other cell assignment methods may mute, reduce or exclude the highest returns resulting in misleading results & underestimated canopy heights. I also use simple average to fill voids, although I don't think this is a very important setting its more for aesthetics so the raster doesn't have ugly little holes everywhere.

What cell size do you use in relation to the point spacing? I'm currently looking at a 0.7m point spacing LiDAR set and I am thinking ok well if I make the cell size too large it may grossly exaggerate the area of any particular high return but if I make the cell size too small it may reduce the high return cells to being so few & small that they become difficult to distinguish from surrounding lower return cells.

I have dome some limited ground truthing with canopy height maps I've made using binning, maximum and a cell size twice the point spacing and found the canopy height map to match hypsometer measured heights to between 2 to 7 feet.

Post Reply

Return to “Measurement and Dendromorphometry”