A signal processing approach could be one way to solving this problem. If you are working with Full-waveform data then automatic layer extraction algorithms (RJMCMC for peak/layer extraction) can give you an estimate of tree height. If your foot print is small enough to cover an individual tree one at a time this approach may work. We have used this in the past (See link below). Even a simple 'peak-extraction' algorithm could work here.

Also, recently I have advocated that we can use large foot-prints in order to learn the waveforms itself. This could reduce the time on scanning individual trees.

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4359291&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4359291

Conference Paper Anomaly detection in clutter using spectrally enhanced Ladar

I am new to the indices world... used by Geo-scientists and the remote sensing community. So I cannot comment on the approach mentioned in the paper you attached. I am reading through them though. Just a thought.. have you tried segmenting your lidar signal or point cloud based on the percentile (in Lastools) how does it look? I'm guessing finding tree tops shouldn't be difficult.

Hello,

I send you a paper that might help you.

Hi Minerva, 

I'm focusing here on the canopy height profile part of the question and not the diversity part: I think your general thought is correct but there are a number of caveats. With the percentiles you mention, you have a good starting point but you're just measuring the density of lidar returns by height which cannot be directly translated to canopy density.

The relationship between vertical profiles of lidar returns and vertical canopy profiles is complex and governed by factors like lidar scan angle, beam footprint, leaf angle distribution, presence of non-photosynthetic vegetation, foliar clumping, and stand structure (i.e. isolated trees or not).

If your study area is dominated by one species (or several species with LAD's that can be modeled similarly) then you have a better chance of discerning the relationship (at least in a relative sense) between lidar height and canopy height. I'd also recommend limiting the scan angles you include to those that are close to nadir (within +/- 10 degrees if possible but sometimes that's not feasible). 

This paper has some useful info on the physical equations that could be considered: http://www.tandfonline.com/doi/pdf/10.5589/m03-026

For more applied and simple corrections (e.g., cosine correction for LAI) see: http://bit.ly/1dlqx5h (my application in urban environment) or http://bit.ly/1BHkkMu (Zhao and Popescu 2009). These are for LAI and not for profiles but some of the same considerations apply.

Hope this helps. For the record, I am not an advocate of making things overly complex. I think you can make rough estimates under less-than-ideal conditions but should just be aware of the potential sources of error.

Mike

Hi,

From math point of view percentiles are not proportions. But if you use LASTools you can actually calculate the proportions of lidar returns from height bins using height bins density option from lascanopy tool with the param "-d 0 2 4 8 30" if you are interested in the height bins 0 -2, 2-4, 4-8, 8-30 m. To establish the relationship between these densities and the foliage density probably you will need some ground truth data because this relationship will depend on the structure of the vegetation you have in your AOI, and the structure will depend on vegetation species. A predominantly fir forest will look very different structurally than a broad lief forest. Lidar responds to structure, and basically you need to establish the relationship between structure and vegetation species. 

The Shannon index is a diversity index and probably you can relate abundance to biomass. From lidar data you can calculate the height of median energy (HOME) that is the 50th percentile and this is correlated with biomass. HOME is sensitive to vegetation structure characteristics such as density, canopy openness, and biomass: Drake et al., 2002. Sensitivity of large-footprint lidar to canopy structure and biomass in a neotropical rainforest. remote Sensing of Environment, 81(20, 378-392. 

I hope this will help as a starting point. Good luck with your research,

Monica

For me research, I used two metrics related to FHD for airborne LiDAR data:

a)Foliage Height Diversity (FHD)- All points-- Foliage arrangement in the vertical direction.

b)Foliage Height Diversity (FHD)-Points above ground threshold---FHD calculated only from the points above the ground threshold.

Following is the code in MATLAB I used to calculate both:

%% For FHD

FHD=0;

for nb=1:20 %number of bins

bht=0.1*nb; %bin height

np(nb)=length(vegZ(vegZ(nb-1)*0.1 & vegZ> CT));

pnp(nb)=(np(nb)/nV);

if pnp(nb)~=0

FHD=FHD+(pnp(nb)*log(pnp(nb)));

end

end

FHD=-(FHD);

%% For FHD-Points above ground threshold

FHDGT=0;

for nb=1:(max(vegZ)/0.1) %alternate way to find the number of bins

bht=0.1*nb; %bin height

np(nb)=length(vegZ(vegZ(nb-1)*0.1 & vegZ> GT));

pnp(nb)=(np(nb)/nG);

%if pnp(nb)~=0

FHDGT=FHDGT+(pnp(nb)*log(pnp(nb)));%FHDGT-Points above ground threshold

end

FHDGT=-(FHDGT);

Or you can very easily calculate them using new version of BCAL Lidar Tools (http://bcal.boisestate.edu/tools/lidar). Go to BCAL LIDAR>>Create Raster Products>>Vegetataion Product.  You will find your two FHD metrics at the end of the list. 

Hope this helps.

MOre reading:

MacArthur, Robert H., and John W. MacArthur. 1961. “On Bird Species Diversity.” Ecology 42 (3): 594–98. doi:10.2307/1932254.

Hi

@Shital: Bical is a great tool but it gives errors

@Monica where in las canopy can I find the option for generating the height bins. Do you have a screenshot I could use? 

Hi,

You can find the help for lascanopy in LAStools/bin/lascanopy_README.txt. the readme file has explanations and some examples for different metrics, including height bin density, but also canopy cover, different statistics like max, min, mean, skew, etc, point count in height bins, canopy density and whatnot. It is a very helpful little tool. Make sure you work only with the points from the point cloud above ground and the z values are replaced by the height of the vegetation itself (you can think of it as the thickness of the vegetation ....). Also do a search for your area and see how tall is the tallest tree you can have. For example E of Mississippi the tallest tree is a little over 58 meters while in California you have trees a little over 114 meters, but you will not find them in other parts of US for example. Unfortunately point clouds have errors that were not necessarily taken out or classified by the vendor. Also check for flight overlap, and if there are points classified as overlap in the point cloud. These areas will have more points than normal and sometimes they show in some of the metrics, for example point density and point count. Lascanopy_README.txt file has some explanation how to keep only some of the classified points and not others, how to have a threshold height limit above which you can ignore the points and so on. In the same bin folder there are other readme files that may be of interest like lasheight and lasgrid. Lasheight will help with replacing the z values to vegetation height values, and lasgrid will be helpful to generate bare earth and first surface of your point cloud.

A free tool to quickly visualize points clouds and do some other few interesting things is called CloudCompare http://www.danielgm.net/cc/ but it is not geared towards vegetation analysis but mostly towards point cloud change analysis and maybe classification. Also helps with doing transects through point clouds, so you don't need an expensive license for LP360 for example. It is a very neat little tool.

I hope this helps,

Monica 

I do not understand your question well. What you want is the total height? is the FCC, or forest volume calculation? You can use any value if the error is statistically small. To work with LiDAR data I use "Fusion" and "R" for statistical computations. Note points per square meter, the type of vegetation and season of the year.

I'd highly recommend the Clawges reference that someone previously mentioned.  We've got a paper that compares the canopy height model to vertical metrics such as FHD and a digitized vegetation map http://www.cabnr.unr.edu/weisberg/publications/Weisberg2014Acta.pdf .  We found that for most of our study species we got the most bang for our buck using mean and standard deviation of canopy height, but some species had additional variance not explained by CHM but by vertical metrics such as the FHD.

Dear Dr Minerva Singh, 

We created maps of MVH and FHD using aerial LiDAR acquired in a mountainous forest area (about 5400 ha) in southern Italy.

You can have more detail about our analysis here:

https://www.researchgate.net/project/Analysis-of-Mediterranean-forests-using-aerial-LiDAR.

Kind regards,

Maurizio

Article Detection of Diversity and Stand Parameters in Mediterranean...