Our scientific goal is to develop and operationally apply innovative theoretical and methodological approaches to better understand, map, monitor and model the multiscale dynamics of ecosystem patterns and processes. We conduct pure and applied research in remote sensing, wildlife ecology, urban studies, and natural resource management.
I am a Geospatial Scientist, developing innovative geospatial solutions for understanding and monitoring the physical environment at different scales, whether within multi-faced and diverse urban settings or broader natural settings. Consequently, I am involved in a variety of multidisciplinary research projects in the field of ecological modeling, wetland and vegetation mapping, grassland mapping and plant phenology analysis, and Urban Planning and Urban Energy Budget. The ultimate goal of my research works is to continue working towards a flexible and holistic approach to the use of geospatial tools for large-area resource management. My current role in BERA as a lead remote sensing scientists involve supervising geospatial research activities within the scope of the project.
Shannon Blackadder is a geospatial scientist who served as research technician for BERA in 2016-17. She is an expert in the application remote sensing and other geospatial tools to solving practical problems, and has served in this role in a variety of research projects. In her other life, Shannon is completing her MSc on species-occupancy modelling in the Crown of the Continent Ecosystem
Use of Unmanned Aerial Vehicles for Monitoring Recovery of Forest Vegetation on Petroleum Well Sites
I am a Research Technician for the BERA Project, providing support to the Remote Sensing Team and its various members, as well as undertaking my own research during the BERA research program. Since receiving my MSc in Geography (University of Calgary, 2008), during which I specialized in the satellite remote sensing of vegetation phenology, I have worked with the Alberta Biodiversity Monitoring Institute as a member of their Geospatial Centre, and have participated in a wide variety of remote sensing-focused, geospatial research activities.
These include land resource inventory protocols, land cover and land use mapping, vegetation mapping, change and spatial analyses, image time series transformation, the use of unmanned aerial vehicles (UAVs) in vegetation recovery assessment on human disturbance features in Alberta’s forested landscapes, and more recently, the application of cloud-computing, machine learning, and open-access satellite data archives to comprehensive wetland mapping in northern Alberta.
Jennifer’s research within the BERA project was completed in 2016 with a publication on “Use of Unmanned Aerial Vehicles for Monitoring Recovery of Forest Vegetation on Petroleum Well Sites”
Use of LiDAR to characterize the understory vegetation, for the enhancement of bird habitat modelling in the boreal forest
Remote Sensing of Coarse Woody Debris for Caribou Habitat Restoration in Alberta’s Boreal Forest
Quantifying human impact on the boreal forest of northern Alberta using LiDAR and photogrammetry
I am a Graduate Student at the University of Munich. The research for my thesis focuses on the evaluation of different techniques for quantifying forest disturbances in the boreal forest affected by natural resource extraction in northern Alberta. This past summer I spent two months at the University of Alberta to conduct my in field data collection. This data will be used to validate the three techniques involving Light Detection and Radar and Stereophotogrammetry for delineating the location and dimension of linear as well as non-linear disturbances. The goal of my research is to identify the best method to sustainably, reliably and accurately monitor forest disturbance and regeneration by comparing the point clouds derived from LiDAR, photogrammetry and a combination of the two, with regards to their accuracy as well as individual advantages and drawbacks. This will support provincial environmental protection plans as well as collect important information on the structure and inventory of the Canadian boreal forest.
Annette’s project was completed in November 2018 with a Masters thesis and a journal publication on the “Comparison of LiDAR and Digital Aerial Photogrammetry for Characterizing Canopy Openings in the Boreal Forest of Northern Alberta“.
Mapping Groundwater Table in Alberta’s Boreal Region Using Remote Sensing Techniques
The objective of my research is to develop an operational method to map groundwater table in the boreal region of Alberta using airborne optical and LiDAR data and photogrammetric techniques. My research work will primarily be built on the work of Rahman et. al. (2017), who used stable surface water (assumed to be saturated to the groundwater) as a guide to identify and map groundwater in a low lying peatland in the boreal region. Their method was observed to work well in low lying areas where there was an abundance of stable surface water. However, in areas where surface water was not visible (dense canopy and/or upland) their method did not perform very well. My research work will consist of two components; (i) apply the technique of Rahman et al. (2017) to a large and diverse study site in the boreal context, and (ii) locate the areas where this technique fails to perform well and thereby propose solutions for improvement.
Rahman, M. M., McDermid, G. J., Strack, M., & Lovitt, J. (2017). A new method to map groundwater table in peatlands using unmanned aerial vehicles. Remote Sensing, 9(10), 1057.
This project was completed in spring 2019 with a MGIS thesis on Mapping Depth to Water (DTW) in Alberta’s Boreal Region Using Remote Sensing Techniques
Semisupervized Object Detection in UAV Images
Michael is a Computer Scientist at the Ludwig-Maximilian University of Munich. His master thesis investigated the effectiveness of machine learning algorithms for automatic detection of coniferous seedling data along Boreal seismic lines. Since the seismic lines cover a length of more than 10,000 km, an automated solution is necessary. He used convolutional neural networks as a feature extractor on the images. Subsequently, they trained an object detector to learn the model to detect seedlings. After the training, the model can be used on unseen data to predict the location of the trees. In this work he also evaluated the accuracy of modern object detectors such as Faster R-CNN with regard to remote sensing capacity of conifer seedlings. Michael further investigated the problem by doing several experiments which focused on the special environmental variables in nature, including seasons and flight height of the drone. They also conducted experiments on the amount of data needed to achieve high accuracy and we also studied the influence of pre-trained networks on the object detector.
Michael’s project was completed in summer 2018 with a Masters thesis on Semisupervized Object Detection in UAV Images.
Man Fai Wu is an MSc student whose research involves developing remote-sensing tools for identifying conifer seedlings on human-disturbance features (seismic lines, well sites, etc). Conifer regeneration is one of the key elements of Alberta’s new seismic-line restoration guidelines, which are designed to encourage a return to forest cover and limit human and predator movement across the landscape.
Characterizing Vegetation Structure on Anthropogenic Features in Alberta’s Boreal Forest with UAV (unmanned aerial vehicle)
Characterizing vegetation structure is an important component for understanding ecological recovery on seismic lines and other non-permanent human footprint features (NPHF). Structural metrics provide an important baseline upon which to build a monitoring program, and a mechanism for comparing NPHF sites to un-disturbed reference locations. Accurate estimation of vegetation structural parameters provides a quantitative assessment of the vegetation status on and besides seismic lines, which is a prerequisite for studying vegetation recovery. However, current approaches to measuring vegetation structure rely on detailed field protocols that are costly and difficult to scale. UAVs (unmanned aerial vehicles) have shown great promise in characterizing vegetation structure in more cost-saving and effective way, compared to traditional field protocols. This project will evaluate the abilities of photogrammetric data from UAVs for characterizing vegetation structure on seismic lines. This study will also give conclusions and suggestions of optimal conditions, processing procedure and analysis method for obtaining the most accurate estimations of vegetation structural parameters. The project will contribute to establish repeatable, cost-effective, and final scale vegetation and ecological monitoring strategies on human disturbed features.
Shijuan’s project was completed in summer 2017 with a Masters thesis and a journal publication on “Measuring Vegetation Height in Linear Disturbances in the Boreal Forest with UAV Photogrammetry”.
The Development of Remote Sensing Tools for Mapping Linear Disturbance Features
The primary goal of my research is to develop remote sensing tools and protocols for accurately delineating the location and dimensions of linear disturbance features within a northern boreal environment in a more cost-efficient manner. To achieve this, airborne light detection and ranging data was used to extract linear features using a least-cost path analysis based on the relative height of vegetation canopies derived from digital elevation models and digital surface models. The results were compared to reference data collected in the field, which included unmanned aerial vehicle-collected imagery that was used to create point clouds for the extraction of comparable metrics (i.e., width of the line). The tools developed here will enhance our capacity to map human linear disturbances, and support ongoing efforts to understand the environmental effects of resource extraction in Canada’s boreal regions.
UAV Workflows for Assessment of Vegetation Structure
I am an MGIS student and Geospatial Technician for the BERA project and responsible for operating our fleet of unmanned aerial vehicles (UAVs) for the Remote Sensing Team at the University of Calgary. This summer I will be collecting aerial data for BERA and managing sensor integration, platform modifications and contributing to best practices for in field data collection. I have been providing support for several projects focused on vegetation analysis, but my personal research focus is to develop an automated workflow for rapid assessment of small replanted coniferous seedlings (<10 cm). UAVs are likely to decrease the need for onsite manual surveys from personnel for a variety of resource industries in Canada. The objective of my research is to create a scale invariant, sample-based algorithm that is stable across different study areas. Results of classified pine density estimates are accurate enough to suggest that UAVs can soon be used in an operational role for monitoring silviculture activities. My research also aims to prove that sample based flight plans are an effective method of extending UAV flight range for vegetation assessment tasks.
Photogrammetric point clouds from terrestrial and aerial imagery for forest mensuration
Traditional forest mensuration or measurement relies on detailed field sampling procedures that involve both direct tree measurements, such as diameter and height, and visual estimations of stand characteristics, such as canopy cover or crown closure. The latter can be subject to strong user bias, experience or error and detailed field protocols are time-consuming. Recent advances in computer technology and computer vision-based software, however, present an opportunity for a new approach to forest mensuration. Series of overlapping digital photographs taken of an object or surface from a variety of angles can be used in a computerized Structure-from-Motion workflow to generate dense, three-dimensional models of that object or surface, in the form of a point cloud. This point cloud is similar to what is produced by a LiDAR instrument, and such photogrammetric point clouds, built from overlapping aerial photographs, have already proven useful in forest characterization. Terrestrial digital photographs, however, present yet another source of imagery for point cloud generation and could be merged with aerially-based point clouds. This work examines best practices and optimal procedures for collecting digital imagery of a forest vegetation plot, both terrestrially and aerially, in support of point cloud-based forest mensuration analysis.
This project was completed in 2016.