Intro to lidar forest mapping

Hello, my name is Chris Hopkinson and I am a professor of geography and environment at the University of Lethbridge. I am also director of the artemis laboratory, where we own and operate a variety of

lidar

sensors to monitor ecosystem properties and the change of note we fly. Missions across Canada using a Teledyne multispectral optical

lidar

system that allows us to map terrain and vegetation cover in high three-dimensional resolution using three laser wavelengths. Over the next few minutes, I will briefly

intro

duce lidar and how we use it to map the landscape. Well, first lidar is the acronym for light detection and ranging whereby laser pulses are emitted towards a target and then, using our knowledge of the speed of light, the round trip time of the pulse and its reflection to calculate the distance at which a scanner is located.

They are used to redirect these pulses across the target surface to generate a three-dimensional point cloud. Most commercial airborne LIDAR

mapping

systems operate just outside the visible range in the shortwave near-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 contact area of ​​the surface geometry and its reflectance on structured targets such as buildings or trees the emitted laser pulse will produce an elongated waveform corresponding to the time during which the pulse interacts with those surfaces in most commercial sensors this response is divided into discrete returns to produce a point charge in the case of vegetation the density structure of the foliage and height strongly influence the geometry of the cloud Point Raw point clouds can be displayed in three dimensions and colored using a variety of attributes such as elevation intensity or return number;

However, the main derivative of most LIDAR acquisitions is the terrain or digital elevation model in

forest

ry and most ecosystem research, of course the structure of the vegetation is equally important and this is where LIDAR excels at providing detailed canopy models and With the latest multispectral lidar systems we can produce 3-channel laser intensity-based composite images using one of our recent multispectral lidar data collections over the Old Man River floodplain. Here in the south of Alberta we can visualize some of these point clouds. attributes at the landscape scale on the left we see the laser return number in the center we have true RGB coloring and on the right we have elevation and intensity in combination, such LIDAR datasets provide a rich three-dimensional and thematic data environment for classification characteristics and modeling approach.

In the laser pulse intensity here we see the grayscale response over the same

forest

ed and bare soil landscape for the three laser channels of the Titan multispectral scanner for applications requiring vegetation structure such as biomass, carbon habitat, marketable volume or fuel for forest fires. A common approach is to extract statistical data. point load descriptors such as height percentiles, here for example we see how the 75th or p75th percentile can be extracted from a hypothetical LIDAR sampling plot using field plot data, we can develop regression models against these cloud metrics of points to predict a series of forest attributes. such as biomass, as can be seen for an area of ​​the Tiger Plains in the Northwest Territories and if we have clear data from our captured time series on naturally growing forest stands, then we can develop biomass growth curves as we see here for berm sites in saskatchewan.

Combining percentile metrics with three-channel laser intensity creates new data structures for

mapping

and modeling vegetation attributes, for example, on the 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 them. Green and shortwave infrared channels, one way to capitalize on this rich information content is to generate vertical profiles of normalized intensity ratios, plotting these as voxels, illustrates both the vertical structure of the canopy and the changing reflectance properties of foliage in the entire canopy as examples. Here we see a study from brenduza that used the combined structure of point cloud and intensity data to map species at the individual canopy level, while here we see a data fusion model from the mcdermott lab that adds additional data metrics to modeling coarse woody debris at ground level to conclude my

intro

duction to air. lidar forest attribute mapping i would like to summarize some of my lab's work at 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 we have an archive spanning 20 years and every generation of lidar sensors, as here we see the time series of the canopy height model illustrating the influence of treatment of growth stands, such as seasonal thinning phenology and differences in lidar hardware, some of these influences are easier to detect; However, if we examine the changes in the cross section of the point cloud over a mature conifer plantation, we see the growth of the canopy, but we also see the gradual densification of the point cloud as the instruments become faster and more capable if we focus attention on the small, fast-growing standard we have left. see how the point cloud captures the change in canopy height over time;

However, it is obvious from this progression that focusing on the upper surface of the canopy misses much of the internal structure of the forest, illustrating the value of extracting so much information from the point cloud. Finally, as is possible, even without point cloud structure information, we can visualize differences in land cover and forest type using only multispectral lidar intensity. The image on the right is a false color composite of the three channel intensity bands and on the left we see intensity signatures for dominant land covers within the study area, applying these signatures to a simple maximum likelihood classifier allows us mapping these land covers and separating the dominant forest types, so this concludes my brief introduction to aerial lidar for forest vegetation attribute extraction and I hope it has been informative if you want to delve deeper into the concepts or case studies mentioned.

Here is a list of related sources or studies and finally I would like to thank you for looking and I must also thank the various funding agencies. and partners who have supported our research over the years, in particular, we owe a special debt of gratitude to all the students and fellows who have kept us at the forefront, as well as to my two most supportive lidar research partners , Dr. Laura Chasma and Teledyne Optics. thank you