News Contact Information

Dpt of Forest Sciences
P.O. Box 27
(Latokartanonkaari 7)
FI-00014 University of Helsinki
FINLAND

Student Services:
viikki-student@helsinki.fi
Registration of credits:
viikki-register@helsinki.fi
Communications and Community Relations:
comms-viikki@helsinki.fi
PhD Studies:
viikki-phd@helsinki.fi
Administration:
mmtdk-hallinto@helsinki.fi

Doctoral dissertation: ”Towards an enhanced understanding of airborne LiDAR measurements of forest vegetation”

Airborne laser scanning relies on LiDAR (Light Detection And Ranging) measurement principle. Airborne LiDARs perform 3D measurements by sending short laser pulses to the target and measuring the travel time of the pulse and the orientation and position of the sensor. Airborne LiDAR can be used in wall-to-wall forest management planning inventories as well as in large scale sampling surveys. Commercial sensors available on the market are normally used and in most cases the 3D information only, i.e. the point cloud, is utilized in interpretation. Aerial images are often used as a complementary data source to improve tree species classification. The studies focused upon use of LiDAR for forest inventory are usually empirical in nature, seeking for correlations between the characteristics of LiDAR point cloud and the forest attributes of interest. In this thesis a somewhat different approach was taken. With the help of careful field measurements and theoretical simulation, I examined LiDAR measurements and the information content in them at the level of individual pulses and trees. The aim was to gain better understanding of the measurement process and of the interplay between laser pulses, vegetation and sensor characteristics. Understanding of the LiDAR pulse interaction with vegetation and of the functioning of LiDAR sensors is needed in developing new data interpretation methods and designing sensors better suited for forest mapping.


The studies made in this thesis reflect the overall technical development in LiDAR sensors. Breakthrough in the use of LiDAR for forestry came in mid 90’s after the emergence of scanning LiDAR sensors that enabled efficient wall-to-wall mapping. Since those days the steady increase in pulse repetition rates has enabled more dense measurements to be performed at given cost, or alternatively reduction of costs by increasing flying altitude and speed. The measurement principles have however remained the same and improvements in the inventory results have been searched by developing new data interpretation methods and utilizing extra information such as laser intensity in addition to 3D points. Waveform-recording LiDAR sensors emerged on the market in mid 2000’s. These digitize the temporal shape of the LiDAR pulse reflected from the target, and provide thus more detailed description of the target characteristics in comparison to discrete return sensors that perform discrete distance measurements, called echoes. The waveform data is still rarely utilized in practice, because of larger data storage capacity and more difficult interpretation. The waveform data are well suited for studying the measurement process and sensor functioning, because fewer details are hidden by the manufacturer’s algorithms. Three of the four sub-studies in my thesis utilized waveform LiDAR data. In addition to increasing our knowledge on LiDAR measurement process, these studies showed the potential of waveform LiDAR for improving forest inventory results.


The first sub-study in this thesis focused upon discrete return LiDAR measurements of understory trees that are often ignored in forest inventories. Detailed investigation of LiDAR measurements at pulse level with the help of accurately positioned field reference trees was undertaken. Generally, the results showed that the intensity information of LiDAR echoes is of low value in characterizing species of understory trees. This is mainly because all species appear similar in the intensity signal, and because the transmission losses to upper canopy further complicate the interpretation. The abundance of understory trees could however be predicted based on echo height distributions. This shows the potential for coarse characterization of understory tree layer by means of LiDAR. This information could be utilized for example in planning the manual clearing activities needed before harvesting operations.


The second sub-study continued in the way of detailed pulse level analyses. This time the field reference data was acquired by means of terrestrial digital images that were looking the trees from below the canopy. The images were accurately oriented to be able to visualize the LiDAR pulses by projecting them onto the images, and to analyze the targets that the LiDAR pulses had intersected. This new method proved useful also in evaluating the geometric accuracy of the LiDAR data. The strength of laser backscattering was shown to correlate with the projected area extracted from the images.


In the third sub-study, a LiDAR simulation model was developed and validated against real measurements (link to video: http://www.helsinki.fi/~ajhovi/Programs/two_birches_animation.gif). The model was able to be used for sensitivity analyses to illustrate how plant structure or different sensor properties influence the LiDAR waveforms. Both simulated and real data showed that LiDAR waveforms are able to capture small-scale variations in the geometry and optical properties of juvenile forest vegetation. Waveform data have thus potential for enhanced mapping of seedling and sapling stands.


The fourth sub-study illustrated the potential of waveform LiDAR data in tree species classification. The results were compared against those obtained with intensity data from discrete return LiDAR. Improvements were seen particularly when the pulse density of the LiDAR data was the lowest. The waveform features that separated tree species were also dependent on other variables such as tree size and phenological status. Early summer acquisitions proved to be beneficial in separating tree species.


Overall, the thesis provided basic findings on how LiDAR pulses interact with forest vegetation, and served to link theory with real observations. The results contribute to an improved understanding of LiDAR measurements and their limitations, and thus provide support for further improvements in both data interpretation methods and specific sensor design.