Intro to lidar forest mapping

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hello my name is chris hopkinson and i'm a professor of geography and environment at the university of lethbridge i'm also director of the artemis lab where we own and operate a range of lidar sensors for monitoring ecosystem properties and change of note we fly missions across canada using a teledyne optic multi-spectral lidar system that allows us to map terrain and vegetation cover in three-dimensional high resolution using three laser wavelengths over the next few minutes i'll briefly introduce lidar and how we use it to map landscape scale biomass attributes well first lidar is the acronym light detection and ranging whereby laser pulses are emitted towards a target then using our knowledge of the speed of light the round-trip travel time of the pulse and its reflection is recorded to calculate distance a scanner is employed to redirect those pulses across the target's surface to generate a three-dimensional point cloud most commercial airborne lidar mapping systems operate just outside the visible range in the near-infrared shortwave infrared and in addition to a 3d map of the earth's surface we also generate a laser-based intensity image the light energy returned to the sensor will be a function of the surface geometry contact area and its reflectance over structured targets like buildings or trees the emitted laser pulse will produce an elongated waveform corresponding to the time over which the pulse interacts with those surfaces in most commercial sensors this response is broken down into discrete returns to produce a point load in the case of vegetation the foliage density structure and height strongly influence the point cloud geometry raw point clouds can be visualized in three dimensions and colorized by a range of attributes like elevation intensity or return number however the primary derivative from most lidar acquisitions is the terrain or digital elevation model in forestry and most ecosystem research of course vegetation structure is equally important and this is where lidar excels by providing detailed canopy models and with the latest multi-spectral lidar systems we can produce composite images based on the 3-channel laser intensity using one of our recent multi-spectral lidar data collections over the old man river floodplain here in southern alberta we can visualize some of these point cloud attributes at the landscape scale at left we see the laser return number in the center we have true rgb colorization and at right we have elevation and intensity in combination such lidar datasets provide a rich three-dimensional and thematic data environment for feature classification and modelling focusing on laser pulse intensity here we see the grayscale response over the same wooded and bare ground landscape for all three laser channels of the titan multispectral scanner for applications that require vegetation structure like biomass carbon habitat merchantable volume or wildfire fuel a common approach is to extract statistical descriptors of the point load such as height percentiles here for example we see how the 75th percentile or p75 can be extracted from a hypothetical lidar sampling plot using field plot data we can develop regression models against these point cloud metrics to predict a number of forest attributes such as biomass as you can see for an area of the tiger plains in the northwest territories and if we have light our time series data captured over naturally growing forest stands then we can develop biomass growth curves as we see here for the berm sites in saskatchewan combining percentile metrics with three channel laser intensity creates new data structures for vegetation attribute mapping and modeling for example at top left we see height profiles for different canopy types in the near-infrared channel and in the two lower graphs we see the deviations within each of the green and shortwave infrared channels one way to capitalize on this rich information content is to generate vertical profiles of the normalized intensity ratios plotting these as voxels illustrates both vertical canopy structure and changing foliage reflectance properties throughout the canopy as examples here we see a study by brenduza that used the combined point cloud structure and intensity data to map species at the individual crown level while here we see a data fusion model from the mcdermott lab that adds in additional data metrics to model ground level coarse woody debris to conclude my introduction to airborne lidar forest attribute mapping i would like to summarize some of my lab's work over a vivian forest study site within the york regional forest in ontario we started collecting lidar here in 2000 using an op tech altm 1210 sensor and now have an archive spanning 20 years and every generation of lidar sensors since here we see the canopy height model time series that illustrates the influence of growth stand treatment such as thinning seasonal phenology and differences in lidar hardware some of these influences are more easy to pick out however if we examine the changes in the point-cloud cross-section over a mature conifer plantation we do see the growth of the canopy but we also see the gradual densification of the point cloud as instruments become faster and more capable if we focus attention on the small rapidly growing standard left we see how the point cloud captures the change in canopy height through time however it is obvious from this progression that by focusing on the upper canopy surface a lot of the internal forest structure is missed hence illustrating the value of extracting as much information from the point cloud as is possible finally even without point cloud structure information we are able to visualize differences in land cover and forest type using multispectral lidar intensity alone the image at right is a false colour composite of the three channel intensity bands and at left we see intensity signatures for dominant land covers within the study area applying these signatures to a simple maximum likelihood of classifier allows us to map out these land covers and separate the dominant forest types so this concludes my very brief introduction to airborne lidar for vegetation forest attribute extraction and i do hope it has been informative if you wish to dig deeper into the concepts or case studies touched upon here here is a list of the sources or related studies and finally i'd like to thank you for watching and as well i must thank the various funding agencies and partners that have supported our research over the years in particular a special debt of gratitude goes out to all the students and fellows who've kept us at the cutting edge as well as my two most supportive lidar research partners dr laura chasma and teledyne optic thank you
Info
Channel: Chris Hopkinson
Views: 9,704
Rating: undefined out of 5
Keywords: lidar, forestry, point cloud, biomass, vivian forest
Id: kTRqnB0usO8
Channel Id: undefined
Length: 6min 26sec (386 seconds)
Published: Sun Feb 07 2021
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.