Welcome to the March edition of the Point Cloud. In this issue, Nicholas Coops gives us an update on AWARE and we feature the Piotr Tompalski (Q13) and the species identification research (Q21) being done by Jean-Francois Prieur.
We are also proud to announce that Shane Furze (Q10) has had his manuscript on “Assessing Soil-related Black Spruce and White Spruce Plantation Productivity” accepted by the Open Journal of Forestry. This will be published in June of this year.
Year End Message from Nicholas
I wanted to provide a quick update to you all re our overall AWARE objectives and our ongoing progress.
As you know, we are approaching the end of the second year of our five years. As part of this milestone, we have prepared our second annual report to NSERC which details our results thus far. Compiling all of the projects and the students, postdocs and faculty as well as the strong links to industry, and federal and provincial government researchers was exciting and humbling to see so many great projects underway.
All core sites have become key research areas with multi-spectral LIDAR being flown in Ontario, full waveform in Newfoundland and one of the largest digital photogrammetry datasets compiled in Alberta. In addition, the field work to support these campaigns has also been staggering. The data compiled at the Blackbrook site in New Brunswick is close to globally unique with tree level species, inventory, soil and site variables all measured and used in a wide range of models and applications.
We continue to build the AWARE community with online student forums, this newsletter, the website and of course, the annual general meeting, which this year is being held in May in New Brunswick. All of these activities allow us to build professional and personal relationships, share science as well as stay grounded to our industrial partner needs.
As we continue to report to NSERC on progress, please make sure you let me or Curtis know of talks, presentations, and papers that you are producing on your AWARE work. It is critically important we record these to demonstrate to NSERC the success of the project.
Thanks for all of your hard work thus far and I look forward to see you all again, as well as meeting new members, in May.
Piotr Tompalski was born and raised in Poland and studied forestry at the University of Agriculture in Krakow. He spent 6 months on exchange at Wageningen University in the Netherlands where he became very interested in GIS and remote sensing. From this experience, he decided to do a master’s thesis at UAKrakow on applications of terrestrial laser scanning in forestry.
After obtaining his M.Sc. in forestry in 2008, Piotr began his PhD studies on the applications of various geotechnologies for forestry and nature conservation with a focus on vegetation growth in greatly disturbed areas. His research employed both airborne and terrestrial laser scanning data in addition to satellite imagery. He used OBIA tools to process his data, skills he acquired during his internship at Trimble Geospatial Division in Munich, Germany (2010). His research methods were influenced not only by the scientists at his home university but also by the researchers at Vienna University of Technology, where he spent a short time as a visiting PhD candidate.
After obtaining his PhD degree in 2013, he moved to Vancouver, BC, Canada, and started to work as a postdoctoral research fellow at the Integrated Remote Sensing Studio, University of British Columbia, where he continues his research on airborne laser scanning in forestry.
In 2015 Piotr joined AWARE and is working on enhancing forest growth estimates with ALS and image-based point clouds. For his research on Q13 in AWARE, he focuses on extracting forest stand attributes like volume, basal area, and height at two different moments in time. He then uses those estimates to match a growth curve to a stand, at 20 x 20 m cell level. Once a growth curve is matched, stand attributes can be projected to a desired year, with increased level of detail.
Aside from his scientific work, Piotr enjoys traditional woodworking.
SPECIES IDENTIFICATION AT THE INDIVIDUAL TREE LEVEL
Jean-François Prieur is a PhD student under the supervision of Benoît St-Onge (Université du Québec à Montréal) and Richard Fournier (Université de Sherbrooke). His research project (question 21 of theme 4) centers on the identification of tree species (within conifer dominated (fir/spruce) and mixed hard- and softwood stands) at the individual tree level using a combination of lidar data and optical imagery. Better species information will improve the accuracy of forest inventories and lead to a greater understanding of forest structure and dynamics. His current study areas are Petawawa Research Forest (PRF) located in Petawawa, Ontario, and Black Brook Forest (BBF), near Edmunston, New-Brunswick. The data he is currently working with comes from a 2016 overflight of PRF using the Teledyne Optech Titan multispectral lidar, and single channel lidar data acquired at BBF in 2012.
The first objective in our tree species identification methodology is an analysis of within-species lidar variation feature values (height percentiles, shape, porosity, intensity statistics and normalized differences, etc.) to find features that are robust to changes in tree size within a given site. Furthermore, in order to develop an identification method applicable across different forest sites, we must here again understand which lidar features are most representative of each species (maximise inter-species variance), and vary the least for a given species (minimize within species variance), in order to ensure maximum separability and generalizability. We must therefore obtain a set of training crowns of different size, across different sites, for which species is known (from field work) in order to analyse variances and train our classifier. One of the technical challenges related to these objectives is the efficient extraction of the point cloud data for each training crown at the calibration stage, and for all crowns of a given site at the classification stage.
The initial identification of tree crowns from the lidar data is performed using in-house software called SEGMA. Approximately 1500 training crowns were extracted from the lidar data for PRF as well as 2000 training crowns for BBF. At PRF, the average size of each flight strip is 7-8 GB and contains between 50-60 million points. Each mature tree is typically composed of 10-100 points. Locating and extracting these points is a time-consuming operation since it involves a spatial join between the crown polygon and each individual point in the point cloud. In our laboratory we had been using R (an open-source statistics application) to perform these extractions that, while successful, ran into R’s limitations: memory management and iteration speed. Python solves these two shortcomings but until recently, nothing comparable to R functionalities (namely the use of spatialized dataframes) was available.
The release of the geopandas (0.2 is latest version) module for Python addresses these shortcomings and provides a spatial join function (sjoin) that has rtree spatial indexing built into it. Rtree indexing speeds up spatial joins by creating neighborhoods of like points so that when the spatial join is performed on a particular lidar point it does not have to search throughout the entire file but a subset of points that have been indexed by use of rtrees. This is all accomplished in just a single line of code using the geopandas.sjoin function.
Figure 1: Graphical representation of rtree index for 3d point clouds. Blue bounding boxes are spatial indices, axes represent x, y and z dimensions.
The results using the 2016 PRF lidar data and the training crowns show that there is a strong linear relationship between the number of points and processing time up until 60 million points. The PC used for these tests has 32GB of RAM so larger point clouds start using swap memory which is much slower (but still feasible). Results at BBF (1km*1km tiles and 100,000 crowns/tile) were processed in approximately 10 minutes, much faster than previous methods. The Black Brook study area is more representative of the volumes encountered in practice with over 2000 tiles of lidar data and crowns. Any improvement in speed during this extraction has repercussions when processing such large volumes of data, notably the acceleration of all the steps that precede the actual training a classifier.
AWARE Research Opportunities
TERRAIN BASED INDICATORS IN PREDICTIVE SPATIAL REPRESENTATIONS OF FOREST GROWTH AND WOOD FIBRE ATTRIBUTES
A MESc. Student at Nipissing University will investigate the potential to derive a suite of terrain based indicators from newly acquired LiDAR data and other geospatial information at the Newfoundland core site and integrate these with existing soil and forest structure information available from the provincial growth plot network. The student will then assess the capacity of these indicators to be used as inputs into land classification schemes which can be used to produce predictive spatial representations of forest growth and wood fibre attributes.
The ideal candidate will have some background in quantitative analysis at the undergraduate level, a Bachelor’s degree in forestry, biology or geography and some experience in growth and yield modelling or forest inventory.
Interested candidates are encouraged to submit a cover letter and curriculum vitae by Jan. 1, 2017 to:
Dr. Jeff Dech
Department of Biology and Chemistry
100 College Drive
North Bay, ON, Canada P1B 8L7
Tel: (705) 474-3450 x.4701
Fax: (705) 474-1947 email@example.com
AWARE Year 2 Annual General Meeting
AWARE’s 2nd AGM will be held from May 16-18th, 2017 at the Madawaska Historical Museum in Edmundston, New Brunswick. Please see our news page for registration and the preliminary agenda.
If you are making your own travel arrangements for the AGM, a block of rooms has been set aside for AGM attendees at the Best Western, Edmundston. Book Best Western
You must reserve at the front desk or online prior to April 16th, 2017.