Lidar Analysis of Vegetation Structure - Revision history

⚠️Wenzeslaus: add see also section

2018-09-23T21:51:17Z

add see also section

← Older revision		Revision as of 21:51, 23 September 2018
Line 603:		Line 603:

	[[File:Grasslidar_71.png\| R.out.gdal 2 ]]<br />		[[File:Grasslidar_71.png\| R.out.gdal 2 ]]<br />

			== See also ==

			* [[Processing lidar and UAV point clouds in GRASS GIS (workshop at FOSS4G Boston 2017)]]
			* [[Introduction to GRASS GIS with terrain analysis examples]]
			* [[From GRASS GIS novice to power user (workshop at FOSS4G Boston 2017)]]

	[[Category: Documentation]]		[[Category: Documentation]]
	[[Category: Tutorial]]		[[Category: Tutorial]]

Neteler: +categories

2016-11-09T08:28:43Z

+categories

← Older revision		Revision as of 08:28, 9 November 2016
Line 604:		Line 604:
	[[File:Grasslidar_71.png\| R.out.gdal 2 ]]<br />		[[File:Grasslidar_71.png\| R.out.gdal 2 ]]<br />

	~~Wikitext~~		[[Category: Documentation]]
	~~Preview~~		[[Category: Tutorial]]
	~~Changes~~
	~~Bold~~
	~~Italic~~
	~~Link~~
	~~Embedded file~~
	~~Reference~~
	~~Advanced~~
	~~Special characters~~
	~~Help~~

Neteler: +categories

2016-11-09T08:28:07Z

+categories

@@ Line 397: / Line 397: @@
 == Step 5:  Terrain analysis with geomorphons ==
-What are geomorphons?  Geomorphon is a new technique for classifying terrain based on a line of sight analysis, see http://geomorphometry.org/system/files/StepinskiJasiewicz2011geomorphometry.pdf
+What are geomorphons?  Geomorphon is a new technique for classifying terrain based on a line of sight analysis, see http://geomorphometry.org/system/files/StepinskiJasiewicz2011geomorphometry.pdf , and http://www.sciencedirect.com/science/article/pii/S0169555X12005028
 The technique uses a line of sight metric for terrain form calculation.  There are options for a simple 10 class representation, a 498 class numeric coding, or positive and negative encoding per cell on line of sight.
@@ Line 604: / Line 603: @@
 [[File:Grasslidar_71.png| R.out.gdal 2 ]]<br />

⚠️Dnewcomb at 02:11, 9 November 2016

2016-11-09T02:11:02Z

← Older revision		Revision as of 02:11, 9 November 2016
Line 543:		Line 543:

	[[File:Grasslidar_63.png\| Mapcalc treemask]]<br />		[[File:Grasslidar_63.png\| Mapcalc treemask]]<br />

			With
			int(bladen_test_5ft_ch@PERMANENT /20)
			as the equation in the Expression box
			Your mask raster should look like this:

			[[File:Grasslidar_64.png\| Height mask raster]]<br />

			All we care about for tree tops are the areas in class 1 or higher, so start r.mask.

			* Select Raster-->Mask{r.mask]
			* In the Raster tab , Select bladen_5ft_treemask_20ft
			* In the box for Raster values to use for mask type: 1 thru 8
			* Click Run

			[[File:Grasslidar_65.png\| Mask Menu]]<br />

			The same area now looks like this:

			[[File:Grasslidar_66.png\| Mask Applied]]<br />

			* Turn off the mask layer

			You should now see the treetops and canopy openings. Of the two classes here, the treetops are probably valid, but the canopy openings that show up with this mask are 20ft or more above the ground and are probably just low spots in the canopy rather than true canopy openings.

			[[File:Grasslidar_67.png\| Mask Applied]]<br />

			The raster mask will remain in place and affect all raster layers and raster operations until it is removed. To remove a raster mask:

			* Click on the Remove tab
			* Check the box for Remove existing mask
			* Click Run

			[[File:Grasslidar_68.png\| Remove Mask ]]<br />

			== Step 8: Exporting to GeoTiff ==

			While GRASS is a rich environment for analysis, the GRASS native datasets are not easily accessible by other applications. Exporting to a common file format such as geotiff would make the data more generally available.

			To export the a raster layer from GRASS :

			* From the Layer Manager window select File-->Export Raster map-->Common export formats

			[[File:Grasslidar_69.png\| Select Raster Export ]]<br />

			This will open the r.out.gdal menu, which uses the gdal library give you the option to write to many different raster formats.

			* Select bladen_per_shrub from the Name of raster map drop down
			* In the Name for output raster file option, Browse to the directory ( c:\grasslidar) and enter the name of the output file as bladen_per_shrub_20ft..tif
			* Keep the Raster data format as GTiff

			[[File:Grasslidar_70.png\| R.out.gdal 1 ]]<br />

			* Click on the Creation tab
			* Set the data type to byte ( 0-255 integer range)
			* In the Creation options window, type COMPRESS=DEFLATE,PREDICTOR=2

			The compression options in the creation options will cause the output geotiff to be losslessly compressed using the deflate method. Predictor= 2 makes the compression more efficient.
			The export process finishes in about 2 seconds.

			[[File:Grasslidar_71.png\| R.out.gdal 2 ]]<br />

⚠️Dnewcomb: Added geomorphon section

2016-11-09T01:38:46Z

Added geomorphon section

← Older revision		Revision as of 01:38, 9 November 2016
Line 497:		Line 497:

	You should then see the following legend with the integer values, classes and histogram by class.		You should then see the following legend with the integer values, classes and histogram by class.

			[[File:Grasslidar_61.png\| Geomorphon Legend]]<br />

			Pretty handy way of describing the landscape. What does this have to do with vegetation?

			== Step 7: Extracting canopy features using geomorphons ==

			* Add the bladen_test_5ft_ch ( add raster) .
			* Set the region to match bladen_test_5ft_ch(SettingsRegionSet regionSet region to match raster map)
			* Run a 3 cell common geomorphon analysis on the 5ft canopy height

			We want to look at the top of the canopy to pick out treetops and canopy openings. The treetops should be summits and the canopy openings should be depressions. We don’t care about any other class, so let’s create a new layer with just those two classes.




			* Go to Raster-->Change category values and classes-->Reclassify (r.reclass)
			* Select common_5ft_ch_3cell for the Name of raster to be reclassified.
			* Set the output raster to treetops_openings_5ft

			* Go down to the enter values directly box and type in:
			1 = Null
			2 = 1 Treetops
			3 thru 9 = NULL
			10 = 2 Openings

			* Turn off all other layers
			* Set your legend to treetops_openings_5ft

			Your display should look like this ( zoomed into the carolina bay)

			[[File:Grasslidar_62.png\| Treetops and Openings]]<br />

			First, a note about r.reclass. It does not create a new layer, it just creates a virtual layer with a filter pointing to the orginal raster. This makes it very fast and saves on hard drive space. GRASS will treat it like a real layer.

			Second, notice the line of “canopy openings” where the transmission right of way going across the pocosin vegetation and the line in the upper left and the tree tops in the field beyond the ditch. We need to filter out by canopy height.

			The best way to do that is to create a canopy height-based mask. In GRASS, the r.mask command, https://grass.osgeo.org/grass71/manuals/r.mask.html, is used. One issue is that our canopy height is a floating point raster and r.mask needs an integer raster. To fix this, we go back to r.mapcalc:

			* Start r.mapcalc
			* Select bladen_test_5ft_ch from the insert existing raster map layer dropdown
			* Click on the / button and type 20
			* Type int( at the front of the equation and ) at the end
			Your mapcalc window will look like this:

			[[File:Grasslidar_63.png\| Mapcalc treemask]]<br />

⚠️Dnewcomb at 01:23, 9 November 2016

2016-11-09T01:23:39Z

← Older revision		Revision as of 01:23, 9 November 2016
Line 460:		Line 460:

	* At the top of the Layer Manager window, Click Add raster map layer button		* At the top of the Layer Manager window, Click Add raster map layer button

			[[File:Grasslidar_56.png\| Select Raster Layer]]<br />

			* In the d.rast menu that pops up, select common_20ft_30cell

			[[File:Grasslidar_57.png\| Select Raster Layer d.rast]]<br />

			* Click OK



			Your results should look like this ( once you select the legend to show the current raster):

			[[File:Grasslidar_58.png\| Geomorphon Layer]]<br />

			A check of the Windows Task Manager shows that the r.geomorphon command is using less than 100 MB of RAM to process the area. This process will take 7.5 minutes on a 2010 vintage i7 laptop.

			Larger areas and larger cell radius extents will require more memory. This process scales fairly well. It has been used on the Statewide 20ft elevation grid for North Carolina ( 6.6 billion cells) with a radius of 30 cells, but it required more than 4GB of RAM and took overnight to run on a Linux workstation. GRASS 7.2 on Windows is 64 bit and will address more than 4 GB of RAM on that platform.


			You can see a lot of different patterns, but what do they mean? Click on the Query raster / vector maps tool and you can click individual cells to extract the value.

			[[File:Grasslidar_59.png\| Information Icon]]<br />

			For example, a blue pixel represents integer value 9 or valley.

			[[File:Grasslidar_60.png\| Query Results]]<br />

			Alternatively, you can
			* Click on the Add map elements Button
			* Select Show/Hide legend
			* Click on the Gradient tab
			* Check the boxes for draw smooth Gradient and ad histogram
			* Put 10 in Number of text labels
			* Click Apply

			You should then see the following legend with the integer values, classes and histogram by class.

⚠️Dnewcomb: Add geomorphon section

2016-11-08T03:15:05Z

Add geomorphon section

← Older revision		Revision as of 03:15, 8 November 2016
Line 403:		Line 403:

	[[File:Grasslidar_49.png\| Geomorphon Classes]]<br />		[[File:Grasslidar_49.png\| Geomorphon Classes]]<br />

			For this activity, we will use the simple 10 class calculation. Outputs will be cells with the following classes: Flat, peak, ridge, shoulder, spur, slope hollow, footslope, valley, and pit.

			To do our terrain classification, we will need to load the r.geomorphon addon to GRASS.

			GRASS has an online repository of Addons much like QGIS has for plugins. Most addons do not require administrative access to install.

			* In the Layer Manager window, click on Settings-->Addons extentions-->Install extention from addons

			[[File:Grasslidar_50.png\| Addon Menu]]<br />

			This will open the Fetch and install extension window. GRASS commands are named by their function: r.* commands are raster commands, v.* commands are vector commands, i.* commands are image analysis commands.

			* Click on the + next to raster to expand the view on all raster commands

			[[File:Grasslidar_51.png\| Addon Selection]]<br />

			* Select r.geomorphon
			* Click install

			In the Layer Manager window you can follow the progress of the installation.

			[[File:Grasslidar_52.png\| Addon Installation]]<br />

			In this case, downloading and installing the addon took 4 seconds on a 2010 vintage Windows 7 laptop.


			== Step 6: Running r.geomorphon ==

			The r.geomorphon command documentation is at http://grass.osgeo.org/grass70/manuals/addons/r.geomorphon.html and the publications for the technique can be found at the end of the command description.

			To start r.geomorphon,

			In the GRASS Command window, type r.geomorphon and press Enter.

			[[File:Grasslidar_53.png\| Start r.geomorphon]]<br />

			Entering the command without arguments brings up the r.geomorphon menu. You will notice that there are tabs across the top. The first tab is the Required Tab, with required data inputs to run the command.

			* From the dropdown for Input DEM, dem05ft@PERMANENT ( remember, the current resolution for the region is 20ft. GRASS will take the center value of the 5ft DEM in each 20ft cell and use that for the 20ft cell value, your output raster will be 20ft resolution)
			* In the Outer search radius, enter 30 ( number of cells around each cell to analyze) This is a very flat area, so the number may need to be higher to extract some features.

			[[File:Grasslidar_54.png\| R.geomorphon setup]]<br />

			You will note that as the command arguments are entered, they are echoed across the bottom.

			* Click on the patterns tab.
			* Type common_20ft_30cell in the box under Most common geomorphic forms.

			[[File:Grasslidar_55.png\| R.geomorphon setup 2]]<br />

			* Click Run

			The tab will immediately switch to the Command Output tab, the command options will be echoed , and a progress bar will show the progress of the command.
			You will need to manually add the result to the Display.

			* At the top of the Layer Manager window, Click Add raster map layer button

⚠️Dnewcomb: More of the exercise

2016-11-08T02:57:19Z

More of the exercise

← Older revision		Revision as of 02:57, 8 November 2016
Line 303:		Line 303:
	* for the zrange enter -10,250.		* for the zrange enter -10,250.
	* Click Run		* Click Run

			This gives a -10, 250 range around the high and low ground in the area (93-10 =83 and 128+250 = 378). This should exclude most of the extraneous points and leave a 5 ft resolution Digital Surface Model(DSM). Change the legend to display the new raster layer. What differences do you see?

			For now we are going to leave this dataset, but we will come back to it later.

			* Go back to Settings-->Region--> Set region to open the g.region dialog.
			* Select the Resolution tab.
			* Enter 20 in the 2D grid resolution box
			* Click Run

			[[File:Grasslidar_43.png\| g.gregion menu ]]<br />

			If you go to Settings-->Region-->Display region
			You will see :
			nsres: 20
			ewres: 20

			The Region resolution setting is now for 20ft cells.

			* Go back to the r.in.lidar menu and:
			* Set the output raster to bladen_test_20ft_ch and
			* Set the statistic to max.
			* Set the transform layer back to dem05ft
			* Check the box for use base raster resolution
			* Click Run
			* Set the legend to bladen_test_20ft_ch


			The Z values were converted to height above ground using the 5ft DEM and then statistics were calculated for 20ft cells. Notice how you see the outlines of the ditches.
			* Make sure that the bladen_test_20ft_ch is highlighted
			* Click on the Query selected raster/vector button ( 4th from the left in the display window)

			[[File:Grasslidar_44.png\| Query button ]]<br />

			* Place the crosshairs on one of the ditches
			* Left Click

			[[File:Grasslidar_45.png\| Query result ]]<br />

			A window will pop up giving the location coordinates, the layer being queried, the color of the feature, and the value of the feature. In this case, the canopy height of the feature. The Z conversion to height above ground is not perfect and may depend on the accuracy of the DEM and the steepness of the slope represented by the raster.


			Remember that a slope is a smooth surface that is being interpolated to a grid format.

			Choosing a cell size can be important. At a 20ft resolution, you will be getting parts of larger trees and 60ft may be more appropriate to capture a more completepicture of the landscape vegetaion. Converting to height above ground allows for the use of cells of arbitrary size. Your analysis can match up with the cell size and footprint to exactly match the cell alignment of existing data.

			There are several other statistical measures that are available in r.in lidar: n, mean, range, sum, standard deviation, variance, coefficeint of variance, median,percentile, skewness, trimmean. Mean height, variance and skewness are metrics used in vegetation analysis.

			At the 20ft resolution, repeat r.in.lidar for variance and mean methods. What differences do you see in the resulting rasters?

			Repeat r.in.lidar at 20ft resolution using n as the statistic to create a point count layer.
			* Go to the Selection tab and
			* Set the z range to 3,10
			* Set the output raster to bladen_20ft_shrub_count
			* Click Run
			* Go to the Selection tab
			* Set the zrange to 10,20
			* Set the output raster to bladen_20ft_midstory_count
			* Click Run

			As stated above, using the raw point count is not suitable for vegetation analysis in areas where there are overlaps in collection. However, if horizontal sections of the point counts per cell are divided by the total point count per cell you can compare the percentage of the points that occur in horizontal sections. In this case, we will be using raster algebra to divide the shrub and midstory point counts by the total point counts using r.mapcalc, https://grass.osgeo.org/grass71/manuals/r.mapcalc.html .

			* Go to Raster-->Raster map calculator (r.mapcalc)

			[[File:Grasslidar_46.png\| Raster Calculator Selection]]<br />

			This will open the r.mapcalc menu:

			* Enter bladen_per_shrub in the Name for new raster map to create
			* Use the dropdown in the Insert existing raster map section to select bladen_20ft_shrub_count
			* Click on the / button for division
			* Use the dropdown to select bladen_20ft_count
			Now things get interesting. The point count rasters are integers. Mathmatical operations between integers results in an integer. Dividing an integer by a larger integer number results in a number less than one. In an integer grid, this truncates to 0.

			To get around this, you convert each integer grid before the operation to a floating point grid, so bladen_20ft_shrub becomes float(bladen_20ft_shrub) and bladen_20ft_count becomes float(bladen_20ft_count) and you divide one floating point grid by another floating point grid and you get a floating point grid.

			Multiply this floating point grid by 100 to put the values back in the range of 100 to 0. Finally, convert results of the entire operation back to an integer grid by enclosing it with int() .
			The final command in the Expression box is:

			int((float(bladen_20ft_shrub_count@PERMANENT)/float(bladen_20ft_count@PERMANENT))*100)

			[[File:Grasslidar_47.png\| Raster Calculator Menu]]<br />

			The resulting raster should look like this:
			[[File:Grasslidar_48.png\| Raster Calculator Result]]<br />


			Repeat the exercise with the midstory layer and compare the two layers. Do you see differences between the shrub and midstory percentage of points layers?


			We will continue the vegetation analysis, but first we need to take a short side trip into terrain analysis that will relate back to our vegetation analysis.. Terrain analysis will be done using geomrphons.

			== Step 5: Terrain analysis with geomorphons ==

			What are geomorphons? Geomorphon is a new technique for classifying terrain based on a line of sight analysis, see http://geomorphometry.org/system/files/StepinskiJasiewicz2011geomorphometry.pdf
			and , http://www.sciencedirect.com/science/article/pii/S0169555X12005028

			The technique uses a line of sight metric for terrain form calculation. There are options for a simple 10 class representation, a 498 class numeric coding, or positive and negative encoding per cell on line of sight.

			[[File:Grasslidar_49.png\| Geomorphon Classes]]<br />

⚠️Dnewcomb at 02:26, 8 November 2016

2016-11-08T02:26:56Z

← Older revision		Revision as of 02:26, 8 November 2016
Line 276:		Line 276:

	[[File:Grasslidar_38.png\|Color editing ]]<br />		[[File:Grasslidar_38.png\|Color editing ]]<br />

			[[File:Grasslidar_39.png\|Updated Colors ]]<br />

			You will notice that in the light green and yellowish areas are significantly above the DEM surface. This is an indication that the laser did not penetrate the vegetation to hit the ground. While the DEM is not perfect, it gives us a better ground surface than just using the min value. There are ways to work around those differences, but that is not the topic for today.

			Look at the other end

			* Set the region to match the 5ft dem
			* Remove the base raster in the Transform tab and in the Optional tab
			* Remove the zrange on the Selection tab
			* Set the output raster to top_of_canopy_5ft
			* Run r.in.lidar with the max method

			Change your legend to match the top of canopy layer. What is the max height that you see?

			[[File:Grasslidar_40.png\|Updated Colors Max Height ]]<br />

			* Click Raster-->Reports and statistics-->Basic raster metadata (r.info)
			* Select dem05ft
			* Click Run

			[[File:Grasslidar_42.png\| r.info output ]]<br />

			R.info gives basic information about the raster layer. You see the rows, columns, number of cells, projection, the command that created the raster is listed as well. This is all useful information for metadata construction. The min and max values of the raster is what you need to look at now. The maximum value in the DEM is 2264.559. Since we have no trees in NC over 200ft, it seems we have some invalid high points. To remedy this:*
			* Go back to the Selection tab and and
			* for the zrange enter -10,250.
			* Click Run

⚠️Dnewcomb at 02:05, 8 November 2016

2016-11-08T02:05:33Z

← Older revision		Revision as of 02:05, 8 November 2016
Line 269:		Line 269:

	[[File:Grasslidar_36.png\|DEM ]]<br />		[[File:Grasslidar_36.png\|DEM ]]<br />

			This feature is called a Carolina bay. The northern half has been ditched and drained, and the southern half remains in the native thick vegetation. Turn the min values back on.

			* Go to raster-->Manage colors --> Color tables. This will open the r.colors menu.
			* Select the Define tab and in the Name of color table dropdown select elevation.

			[[File:Grasslidar_38.png\|Color editing ]]<br />