Nicholas Charles Coops

The Earth is facing a biodiversity crisis, with over one million species at risk of extinction. To inform recovery actions for threatened species, statistical models are used to predict and understand the habitat required for species survival. Airborne Laser Scanning (ALS) data (a lidar technology) is proving increasingly useful for modelling habitat, providing detailed 3D environmental information. In British Columbia, the prediction of marbled murrelet (Brachyramphus marmoratus, hereafter ‘murrelet’) nesting habitat is of considerable interest. Murrelets are threatened seabirds that nest in large trees along the Pacific coast. This habitat has severely declined due to industrial logging and fine-scale maps of remaining nesting areas are needed to inform recovery actions. This thesis is composed of two complementary research chapters. First, I investigate how ALS technology can be used to advance habitat modelling. I synthesize the available approaches to infer habitat characteristics with ALS and provide a non-technical overview of processing options and data considerations. I advocate for the use of ALS predictors that capture the specific resources, risks, and conditions that directly determine habitat suitability. In doing so, habitat models can be tailored to a species’ ecology and test a wider range of biologically realistic species-environment relationships. Second, I apply these approaches to model murrelet habitat. Two rare-species modelling approaches - ensembles of small models (ESMs) and maximum entropy modelling (MaxEnt) - were evaluated to link 58 nests with ALS and non-ALS predictors at a 100 m resolution in Desolation Sound (DS). An independent model transfer was conducted in Clayoquot Sound (CS), with 21 nests. A MaxEnt model using two ALS predictors (forest vertical complexity and the volume of internal forest gaps) was identified as the best model, achieving good performance in DS (AUC = 0.77, Boyce index = 0.99) and reasonable performance in CS (AUC = 0.63, Boyce index = 0.67). On average, the most suitable habitat was found in small clusters (
View record

Wildfires burn with a mixture of severities across the landscape and create complex forest structures. Quantifying the structural changes in post-fire forests is critical to evaluating the ecological impacts of wildfires. Advances in drone laser scanning (DLS) and mobile laser scanning (MLS) have enabled the acquisition of high-density point clouds to better resolve detailed tree structures. Yet, few studies have examined their combined capability to describe forest structures. To characterize post-fire tree attributes in interior dry forests in British Columbia, nine study sites in the area burned by the 2017 Elephant Hill wildfire were scanned in 2019 using DLS and MLS. First, I examined the utility of DLS and MLS both individually and combined to estimate tree attributes using quantitative structure models (QSMs) across varying canopy cover levels. Second, I investigated the QSM-derived tree attributes to interpret the effects of burn severities at individual tree and plot scales. The results showed that the fused laser scanning datasets outperformed single laser scanning datasets in estimating tree attributes across canopy cover levels. Specifically, with increasing canopy cover, diameter at breast height, crown diameter, and crown base height were best predicted using the fused data. Height was accurate regardless of canopy cover, which was independent of data collection platforms. Tree volumes could be best modelled by the fused data with increasing canopy cover. In terms of burn severities, the results suggested that smaller pre-fire trees tend to experience higher levels of crown scorch than larger pre-fire trees. Among trees with similar pre-fire sizes, those within mature stands experienced relatively lower levels of crown scorch than those within young stands. At the plot level, low-severity fires had minor effects, moderate-severity fires mostly decreased tree height, and high-severity fires significantly reduced diameter at breast height, height, and biomass. The results also revealed that stands dominated by trees with large crowns and relatively wide spacing could burn less severely than stands characterized by regenerating trees with high densities. The findings of this thesis facilitate forestry practitioners to select appropriate laser scanning tools in forest inventory assessment and develop site-specific management plans for fire-prone forests.

View record

Active forest management for timber production, through the harvesting of forest stands using cut blocks, frequently overlaps grizzly bear (Ursus arctos) habitat on multi-use landscapes in North America. Thus, it is critical to understand how forest harvest management can effectively support grizzly bear conservation efforts. While many localized studies have investigated relationships between forest harvest and grizzly bear habitat use, a synthesis of our current understanding of these complex interactions is warranted. This thesis reviewed publications that empirically assessed grizzly bear use of recently harvested forest stands (
View record

Grizzly bears utilize large home ranges and move through many stages of forest succession. Research to date has confirmed bears display preference for habitats along forest edges, as well as certain landcover classes. However, the effect of fine-scale patterns and changes in forest structure on bear movement and habitat selection is less well understood. To date, habitat selection studies of grizzly bears have thoroughly examined the effects of various habitat characteristics such as anthropogenic disturbance, food availability, topographic variables, and plant communities, though there is more opportunity to leverage advanced remote sensing, three-dimensional data to refine and enhance our understanding. To address this, I processed Airborne Laser Scanning data in the Yellowhead region of Alberta to characterize forest structure and used resource selection functions to determine whether ALS-derived descriptions of habitat could effectively model habitat selection by bears. I found: 1) ALS, when combined with a topographic index, provides enough structural information and context to reliably describe habitat selection. 2) Canopy cover and vegetation height both influence selection as a function of distance from the forest edge. 3) Outside a forest stand, cover > 2 m increases the probability of selection, while inside forested stands, higher canopy cover is negatively related to selection. My final model cross-validation demonstrated selective use of forest edge habitat. Selective use of forest edges implies that the shape and spatial arrangement of forest cut blocks and associated retention patches left during harvesting operations may be optimized to minimize human-bear encounters and associated human-caused mortality of grizzly bears, in accordance with the “emulating natural disturbance” forest management philosophy.

View record

Canada’s urban areas experienced extensive growth over the past-quarter century; however, there has been no reliable, spatially explicit approach for quantifying urban vegetation and land use patterns nationally. Satellite imagery, such as Landsat, provide an opportunity to measure large areas frequently, consistently, and accurately over long temporal periods. Using Landsat imagery, this thesis characterizes multi-decadal trends of vegetative greenness and land use transitions from 1984 to 2016 on the pixel and census dissemination area levels for 18 major Canadian urban areas. First, I developed an urban greenness score using spectrally unmixed Landsat imagery to categorize multi-decadal greenness change relative to its current greenness level across Canadian geographic strata and population density groups. Second, using unmixed fractions and land cover information, I derived a dynamic land use classification scheme and applied it across Canadian peri-urban areas to quantify multi-decadal land use transitions and their correlation with current socio-economic states. Most Canadian urban areas sustained a moderate (∼ 20–40%) or low (≲ 20%) level of greenness between 1984 and 2016, with denser urban areas experiencing the greatest losses (16% of DAs). Eastern urban areas maintained the most greenness over time, while urban areas in the Prairies had the greatest increase in greenness. During the same time, 47% of Canadian peri-urban areas experienced an increase in urban use, converting about 2,000 km² of natural and agriculture uses each. The region with the most area transitioned from natural to urban use occurred in the West (39%), whereas the Prairies observed more marked rates of urbanization on natural areas (median of 3.9% per annum per sq-km) and more area with agriculture conversions (41%). Regional differences in correlations between land use trends and current levels of population density, income, new residential construction, and single-detached housing highlight the potential influence of urban policies and/or socio-economic trajectories. The methods presented take advantage of sub-pixel information from the open and longstanding Landsat archive to provide a comprehensive and accessible framework to understand current and historic urban land dynamics, which supports long-term strategic planning and can be transferred to other regions across spatial and temporal scales.

View record

The diversity of forest birds has long been linked to forest structure. Forest structurecan be quantified by remote sensing with airborne laser scanning (ALS). To date, ALSbased wildlife monitoring studies have been limited in geographic scope, principallydue to limited aerial coverage of the ALS datasets. This thesis first assesses the abilityof ALS and other remote sensing and geospatial datasets to estimate bird occurrenceover a focus site of boreal forest in Newfoundland, Canada and then develops anapproach to extend predictions of bird occurrence in areas that have no ALS coverage.To do so, I developed models using best subset regression of avian habitat suitabilityusing ALS data acquired in Newfoundland, Canada. Model data included plot surveys,multispectral information, topographic, landscape, climatic and ALS datasets. Whilefield plot metrics outperformed remote sensing metrics from a single sensor; modelscombining ALS, multispectral and/or environmental data, were most significant. Giventhe models were developed in highly managed areas of boreal forest, the resultssuggest these approaches can help monitor the environmental impacts ofmanagement decisions and climate change on biodiversity.Second, I used discontinuous ALS and spectral time series from Landsat to predict wall-to-wall structural metrics across Newfoundland. Three forest structural metrics(kurtosis, height and understory) were selected as strong predictors of avian habitatquality. Using imputation approaches, I extended these structural metrics over theentirety of Newfoundland with results indicating kurtosis was not predicted accurately(r 2 =0.08) when compared to independent ALS data, however, height (r 2 =0.43) andunderstory (r 2 =0.37) showed strong correlations. Using a random forests modellingapproach, I produced habitat suitability models to predict the occurrence of nine birdivspecies using citizen science data from eBird and provincial permanent sample plots(PSPs). All species had higher area under the receiver operator curve values using eBirddata (min=0.73, median=0.85, max=0.91) compared to PSP data (min=0.52, median=0.60,max=0.71), demonstrating the strength of citizen science data. Overall, the resultsdemonstrate structural metrics from ALS, combined with avian observations fromforest plots and citizen science data can produce accurate predictions of avianoccurrence and abundance over large areas.

View record

Knowledge of forest structure can be used to guide sustainable forest management decisions. Currently, Airborne Laser Scanning (ALS) has been well established as an effective tool to delineate and characterize canopy structure of forested biomes. However, the use of ALS to characterize forest secondary structure is less well developed. Secondary structure consists of suppressed sub-canopy trees, short-stature vegetation and coarse-woody-debris (CWD). I utilized discrete return ALS to develop methodologies which characterize two secondary structural units, sub-canopy trees and CWD, within natural forest stands in central British Columbia. I first segmented the forest vertically into canopy versus sub-canopy and computed a suite of ALS metrics to develop predictive models of sub-canopy stand attributes. Calibrated against 28 ground plots, models were developed using stepwise regression resulting in the strongest predictors being a combination of height, structure and cover-based metrics. Two sets of models were developed, one with the canopy removed and another with it retained. The sub-canopy set of models resulted in stronger cross-validated R-squared values for volume and basal area and as a result the sub-canopy volume model was used to map sub-canopy volume over the entire study area. The second structural unit, CWD, is a meaningful contributor to forest carbon levels and biodiversity. In this work I detail a novel methodology that isolates CWD returns from large diameter logs (>30cm) using a refined grounding algorithm, a mixture of height and pulse-based filters and linear pattern recognition to transform returns into measurable vectorized shapes. Height values are extracted directly from the point cloud to calculate volume for detected shapes. This approach is then demonstrated by successfully mapping CWD and estimates of volume as well as providing an assessment of individual log and plot-level attributes that influence successful detection. I compared plot volume totals calculated from ALS-derived CWD against field measured CWD and found a strong correlation. Lastly this methodology was applied over a larger region to quantify CWD volume differentials between stands. These methodologies demonstrate the capability to generate a secondary structure inventory that can highlight locations for selective logging, model fire susceptibility and carbon sequestration, and quantify wildlife habitat.

View record

Understanding patterns in vegetation phenology under increasing anthropogenic pressures, and within a changing climate, is essential for determining the availability of key plant-food resources, which drive habitat selection of wildlife. In creating a fine scale phenology product to monitor daily phenology trends, this thesis determines how variations in availability of key plant-food species is affecting daily habitat selection of grizzly bears (Ursus arctos), across years, within the Yellowhead Bear Management Area, Alberta, Canada.Dynamic Time Warping is used to combine Landsat satellite data and Moderate Resolution Image Spectroradiometer (MODIS) imagery, to quantify daily changes in vegetation at a 30 m resolution, from 2000-2018. This approach, entitled DRIVE, was validated against the start and end of season dates (SOS and EOS respectively) derived from time-lapse imagery obtained from ground cameras. Results showed correlations of r = 0.73 at SOS and r = 0.85 at EOS with a mean absolute error of 7.17 and 10.76 days at SOS and EOS respectively. Analysis of the DRIVE product also indicated that SOS is advancing at a maximum rate of 0.78 days per year from 2000-2018.A set of new methods were then developed to create daily vegetative food species availability layers from 2000-2017. Annual species distribution models (SDMs) were created using maximum entropy modelling. SDMs were combined with DRIVE outputs to create daily plant-food availability layers for eight food species. Food availability layers were combined with environmental variables and grizzly bear GPS collar data to create resource selection functions modelling daily and seasonal selection. Results determined that in the dry spring, selection for roots was stronger and occurred earlier than in the average/wet years, in the wet summer the length of selection increased for forbs and in the dry fall, the period of selection for berries was longer than in the wet year.Through this research, I found that variations in phenology driven by climate and anthropogenic processes, has the potential to affect grizzly bear habitat selection into the future. The datasets and approaches developed here will provide resource managers with an important tool for use in grizzly bear habitat management and population recovery.

View record

Anthropogenic disturbances, including roads, are known to influence animal habitat selection and mortality. However, little is known about the role of sensory perception in animal responses to disturbance. The goal of this thesis was to investigate the effect of visual and auditory perception around roads on grizzly bears (Ursus arctos) in Alberta, Canada. As an apex predator, the greatest threat to grizzly bear populations in my study area is human-caused mortality near roads, yet grizzly bear behavioural responses to roads are not fully understood. In this thesis, detailed topographic and land cover data from airborne Light Detection and Ranging (lidar) and Landsat imagery were used to estimate visibility and audibility around roads. Using a modified semivariogram approach with data on step lengths from GPS-collared grizzly bears, I found that grizzly bears responded to roads at slightly further distances when roads were perceptible (80 m) than when roads were imperceptible (60 m). I extended the analysis of grizzly bear response by modelling habitat selection as a function of road perception and other environmental variables using integrated step selection analysis. I also assessed mortality risk in visible areas by comparing habitat selection between grizzly bears that died and grizzly bears that survived. Grizzly bears were less likely (p
View record

Determining the number of trees in a stand, known as stand density, is one of the key fundamental inventory metrics that inform managers of the current conditions of the forest. Stand density can be a challenging variable to measure using field-based surveys, and as a result, foresters have turned to remote sensing to help identify and count individual trees in a given area. Currently, the field of remote sensing is undergoing rapid changes through the utilization of remotely piloted aircraft systems (RPAS) technology. The increased utilization of this new technology has seen renewed interest in the application of individual tree detection (ITD) to digital aerial photogrammetric point clouds (DAP). To date, much of the previously developed ITD methodologies have been developed for airborne laser scanning (ALS) and traditional aerial imagery. DAP is unique as it provides point cloud-based data as well as retaining the spectral information from the original images. This thesis analyzes the application and subsequent modification of an established tree detection routine to RPAS acquired DAP. In particular, modifications are proposed and evaluated which aim to improve its applicability to DAP, with particular emphasis on the incorporation of spectral information. The new routine, which utilizes point-level spectral information in sub-crown level clustering, improved overall true positive detection by ~6.3%, with the most significant improvements in true positive detection found in lower canopy class stems. The ITD routine was also tested without the inclusion of spectral information, and was shown to produce poorer results by ~4.1%; indicating that the inclusion of these data is beneficial for ITD approaches. This new ITD routine and the use of point level spectral information, represent an advancement of understanding the role of DAP for ITD in the era of RPAS remote sensing in forested environments.

View record

Distributed systems of small satellites, called CubeSats, are generating a new type of remote sensing data: multispectral imagery with high spatial resolution and near-daily global coverage. Recent studies have shown that this data is valuable for monitoring land cover change. However, in order to achieve widespread application for this purpose, all satellites in the distributed system, or constellation, must acquire imagery with accurate geolocation and consistent radiometric properties. I developed a preprocessing method to automatically co-register and radiometrically normalize a temporally dense series of images using reference imagery. To demonstrate the effectiveness of this approach, I normalized a time series of CubeSat images at both the pixel and the polygon level. Images from the PlanetScope (PS) constellation were used, focusing on a forested area in British Columbia, Canada that was heavily affected by forest fires in 2017. By examining the normalized difference vegetation index (NDVI) before and after the fires, I found that this method allowed simple identification of burned and unburned areas, which was not readily possible without applying the normalization method. After establishing the validity of the preprocessing method, I developed a multi-temporal change detection method which integrates this technique. This methodological approach is resistant to cross-sensor radiometric inconsistencies. I illustrate the effectiveness of the approach using imagery from the PS constellation and the Harmonized Landsat Sentinel-2 (HLS) virtual constellation. The approach is two-fold; first, a bitemporal method is applied exclusively to PS data, and then a multi-temporal method is applied to radiometrically normalized PS and HLS data. I apply this method to generate a forest fire disturbance map at 3.0 m resolution with a sub-weekly time step. A comparison with a more conventional disturbance map from Landsat difference normalized burn ratio shows improved capture of fine-scale spatial heterogeneity in the burn patterns. This method allows for integrated radiometric normalization with high-resolution change mapping at a sub-weekly time step using CubeSat imagery, suitable for fine-scale land cover analysis. These approaches can help fully exploit remote sensing datasets that have high spatiotemporal resolution but contain radiometric inconsistencies in order to quickly identify land cover changes.

View record

Autonomous aerial systems-based digital aerial photogrammetry (AAS-DAP) is an emerging technology that has the capacity to generate dense three-dimensional (3D) point clouds similar to airborne laser scanning (ALS). Over forested stands, these point-clouds can be used to model forest attributes using an area-based approach however, model accuracy is dependent on digital elevation model quality used to gather vegetation heights above ground. It is known that canopy occlusion contributes to larger gaps in terrain registration from AAS-DAP compared to ALS point clouds. Due to the recent emergence of AAS-DAP as a cost-effective remote sensing platform, few studies have investigated the terrain modelling and forest inventory capacity of AAS-DAP over complex conifer forests. In Chapter 3, through the use of a sensitivity analysis, I established a set of optimal ground points from AAS-DAP by using commercially provided ALS ground points as reference. This optimal set of ground points was then used to test common terrain surface interpolation routines in Chapter 4. Interpolation routines include inverse-distance weighted, natural neighbour, triangulated irregular network, and spline with tension. Using field-measured tree height and stem diameter, allometric relationships were established for dependent variables: mean tree height (Hmean), Lorey’s height (HLorey) and stem volume per hectare (Vstem). Models were then fit among dependent variables and metrics calculated from the vertical distribution of the AAS-DAP point cloud normalized by the different AAS-DAP terrain surfaces in addition to a reference surface generated from commercially provided ALS ground points. A Kruskal-Wallis with Dunn’s posthoc test found no significant difference between predictions derived from different terrain surfaces for all three dependent variables; however, the inverse-distance-weighted method produced a distribution of predictions most similar to those from the ALS-DEM. The best performing forest attributes models for Hmean, HLorey and Vstem yielded mean root-mean-square errors (RMSE) of 1.19 m (7.29%), 0.92 m (5.04%) and 54.55 m³·ha⁻¹ (26.66%) respectively across the four AAS-DAP terrain surfaces generated. Model performance was higher yet comparable when using the ALS-DEM for point cloud height normalization with RMSE of 0.73 m (4.43%), 0.59 m (3.24%) and 37.31 m³·ha⁻¹ (18.24%).

View record

Many migratory terrestrial mammal species within North America rely on particular habitat characteristics to provide shelter from snow cover in order to assure inter-annual survivorship. Mule deer (Odocoileus hemionus) in particular, are reported to be in decline across South-central British Columbia, likely due to reductions in adequate winter range habitat, limiting the degree of shelter that can be utilized to avoid snow and low temperatures. This thesis sought to evaluate predictions from multiple step selection functions (SSF) by considering both mule deer responses to the timing and distribution of snow cover as well as forest stand attributes including canopy cover and forest edge. In order to generate such SSFs, increasing spatial and temporal information regarding snow timing and distribution across the landscape was required. Previously however, predictions of fine-scale snow dynamics across the landscape suitable for analysis with hourly telemetry data were limited. Therefore, the first component of this thesis was to utilize the strengths of both medium spatial resolution and high temporal resolution satellite imagery and develop a data fusion algorithm to predict snow cover dynamics at a 30m spatial resolution daily, since 2000 using Landsat data with MODIS (Moderate Resolution Imaging Spectroradiometer) snow map data as inputs. The final fused snow map product (MODSAT-NDSI) achieved an overall accuracy of 90% using 33 validation test sites, which included government snow pillow data and an installed camera network. Environmental covariates from MODSAT-NDSI snow maps and 77 deer’s GPS telemetry data in the mule deer SSFs were used to produce predictions of relative probability of use for population-level estimates of habitat selection patterns. The top-ranked SSF models (based on AIC) indicated that mule deer avoided areas with greater, and more persistent, snow cover, and selected areas closer to forest edge. Key thesis outcomes include generated snow cover maps that can be updated and utilized in further studies, a data fusion algorithm that can be replicated for other remote sensing metrics, and habitat selection models that may help to inform future mule deer habitat management.

View record

Snow dynamics influence seasonal behaviors of wildlife, such as denning patterns and habitat selection related to the availability of food resources. Under a changing climate, characteristics of the temporal and spatial patterns of snow are predicted to change, and as a result, there is a need to better understand how species interact with snow. Through the generation of fine-scale snow cover data, this thesis examines grizzly bear (Ursus arctos) spring habitat selection and use in the Yellowhead Bear Management Area, Alberta, Canada. First, a new approach was developed to create a daily time-series of 30-m resolution snow cover observations (called SNOWARP). SNOWARP was derived from daily Moderate Resolution Imaging Spectroradiometer (MODIS) data to capture the temporal dynamics of snow cover and Dynamic Time Warping to re-order historical Landsat observations to account for inter-annual variability. The SNOWARP product was produced for 2000-2018 and calibrated against a network of time-lapse cameras and snow pillows. Results indicate the root mean squared error of the fractional product ranges from 31.3% to 68.3%, while F score of the binary product ranges from 87.7% to 98.6%.Second, data from SNOWARP and other environmental variables were combined with GPS collar locations from grizzly bears to test the hypothesis that grizzly bears select for locations with less snow cover and areas where snow melts sooner during spring. Using Integrated Step Selection Analysis, a series of models were built to examine weather snow variables improved models constructed based on previous knowledge of grizzly bear selection during the spring. Comparing four different models fit to 62 individual bear-years, it was found that the inclusion of fractional snow covered area (fSCA) improved model accuracy 60% of the time based on Akaike Information Criterion tallies. Probability of use was then used to evaluate grizzly bear habitat use in response to snow and environmental attributes. The results of this thesis provide one example of the application of newly derived daily 30-m fSCA and indicate grizzly bears select for lower elevation, snow-free locations during spring, which has important implications for management of threatened grizzly bear populations in consideration of changing climatic conditions.

View record

Enhancing forest inventories using airborne laser scanning (ALS) and digital aerial photogrammetry (DAP) is a spatially extensive means of providing accurate and consistent measures of forest stand structure. While the cost of multi-temporal ALS is still sometimes prohibitive to its integration for growth assessment, DAP point cloud data have been proposed as a cost-effective alternative to those from ALS for inventory re-measurement. As such, the primary objective of this thesis was to examine the capacity of ALS and DAP technologies to assess height growth (HL) in a disturbed boreal forest near Slave Lake, Alberta.First, this thesis determined the variables to be used in modeling height growth, and investigated how the predictive model errors responded to stand condition. To evaluate appropriate variables for predictive modeling, a model using only height metrics (growth_single) was compared with one using height, canopy cover and height variability metrics (growth_multi). The growth_multi model estimated height growth with an RMSE of 1.42 m (%RMSE = 164.18%) and the growth_single model estimated height growth with an RMSE of 1.76 m (%RMSE = 203.03%). To evaluate error response to stand condition, an iterative process was used to measure the accuracy of optimized height models while incrementing the mortality in the dataset. %RMSE increased with increasing plot-level mortality as a parabolic asymptotic curve. When the maximum allowable mortality was approximately 25% the %RMSE was just below 100%.Second, this thesis determined growth patterns near Slave Lake with respect to eight ecological variables, ecosite type and ecosite phase. Analysis of variance (ANOVA) tests were conducted to test the significances of differences between the means of height growth (ΔH). Patterns demonstrated by the ecological variables were most apparent using nutrient regime, moisture regime, species dominance and the soils classification. Growth patterns among ecosites and ecosite phases followed the patterns of the ecological variables that describe them.This research finds that, prior to utilizing multi-temporal remote sensing methods to assess stand-level height growth, forest managers must first understand local forest growth rates and mortality rates to ensure that the growth magnitudes and forest condition permit accurate height growth estimation using predictive models.

View record

Near real-time mapping of anthropogenic linear networks (e.g. roads, seismic lines and fireguards) in forests has a range of applications including monitoring rapid management responses to disturbances such as fire. Synthetic aperture radar imagery is well suited for near real time monitoring because microwaves penetrate clouds and smoke, and satellite images are acquired weekly in many parts of the world, assuring regular coverage.In this study, we created maps of fireguard networks constructed during the 2017 wildfires in Alex Fraser Research Forest in interior British Columbia, Canada based on Sentinel-1 SAR image time series and Sentinel-2 image pairs.I developed two methods to summarize the Sentinel-1 backscatter time series in a single summary raster suitable for human interpretation in Google Earth Engine. The first method is to fit a trend line to the backscatter time series for each pixel, and display the value of this line at the start and end of the observation window in red and green. The second is to fit a single-step function and display the left and right tail values along with the R² value of the fit as red, green and blue values. I assessed the utility of these summary images for manual delineation of fireguard networks by simulating the accuracy and timeliness of fireguard detection based on acquisition in near real-time. For reference, I compared these methods with a straightforward before-after analysis of Sentinel-2 multispectral images and with ground truth maps.From the trend line and step function summary images, interpreters detected 22–41% and 24–55% of fireguard length respectively while delineation from multispectral imagery attained a detection rate of 84–86%. Delineation from Sentinel-2 images was most precise with an average deviation of 5–6 metres from the ground truth, followed by the trend image with 8–15 metres deviation and the step image with 11–16 metres. In the best case, a change was detected based on a step image within 6 days. The developed method can be used to monitor linear feature construction where more accurate methods are unavailable.

View record

Urban areas are where most people in the world directly benefit from ecosystem services (ES), yet there is evidence that ES are distributed inequitably with respect to socio-economic status which can lead to environmental injustices. There are a number of barriers to understanding UGS equity, some conceptual and some technical. One barrier is that few studies of environmental justice use similar quantitative methods, nor do they use concordant conceptual frameworks of UGS and ES equity. Another barrier is the fine scale information needed to accurately map greenspace can be difficult and expensive to obtain. In light of these barriers, this thesis seeks to contribute to UGS/ES equity studies in two fundamentally important ways.First, I explore the concepts of equity in ecosystem services as applied to urban settings. I undertake a review of trans-disciplinary literature on urban systems to answer the question “How has environmental justice been considered and incorporated into urban ES research?” I characterize types of urban ES and measure the breadth of justice issues addressed in each article using a new environmental justice index (EJI). I also highlight the methods and results of key quantitative and qualitative papers that can inform future urban ES justice frameworks. Second, I explore how new advances in remote sensing can better characterize UGS distributions via more accurate mapping of heterogeneous urban areas. I combine three-dimensional information from airborne Light Detection and Ranging (LiDAR) data with RapidEye high spatial resolution imagery in a Geographic Object-Based Image Analysis (GEOBIA) approach to classify urban landcover in a large metropolitan region. Though 5m RapidEye pixels were often mixed in urban areas, LiDAR data enabled accurate classification of fine spatial objects such as street trees and single-family dwellings. Ultimately, I propose that mapping ES distributions among urban socio-demographic groups and assessing potential ES tradeoffs is not enough to avoid injustices. Because ES are socio-political constructs, gaining a comprehensive understanding of urban ES injustices is not merely a process of mapping greenspace, but also understanding how the groups in question ascribe value to the ES supply sources around them.

View record

Observing landscape patterns at various temporal and spatial scales is central to mapping ecosystems. Traditionally, ecosystem mapping uses a combination of fieldwork and aerial photography interpretation. These methods, however, are time-consuming, prone to subjectivity, and difficult to update. Airborne Laser Scanning (ALS) is an advanced remote sensing technology that has increased in application in the past decade and has the potential to significantly increase and refine information content of ecosystem mapping, especially in the vertical dimension. ALS technology provides detailed information on topography and vegetation structure and has considerable potential to be used for terrestrial ecosystem classification and mapping. In this thesis, the potential to use ALS data to advance ecosystem mapping is examined. The current state of the science for using ALS data to classify and map key ecosystem attributes within an existing ecosystem mapping scheme is discussed by focusing on British Columbia’s Terrestrial Ecosystem Mapping (TEM) and its associated Predictive Ecosystem Mapping (PEM). Based on a detailed literature review, a site-specific case study was also developed with the goal of mapping TEM polygons for a forested landscape on Vancouver Island, British Columbia. To do so an object-based image analysis approach was used. The analysis examined which were the best suite of ALS-based terrain and vegetation metrics to define and distinguish individual site series. It established a workflow for the classification of site series within the study site and examined the capacity to map site series based on ALS derived values. Best segmentation parameters were first established and then the study area was classified for slope position-wetness and finally into the specific site series. In the classification of site series two approaches were used. One approach used only terrain metrics and the other incorporated vegetation metrics. Overall accuracies were 59% and 56% respectively. While this workflow requires refinement, it shows potential for improved accuracies by applying suggestions discussed.The thesis concludes with a discussion summarizing the findings of this research and highlighting future refinement to the methods applied in the case study, while also providing recommendations for the current application of ALS technology to TEM.

View record

Vegetation structure is an important biodiversity indicator providing biological and physical environment that supports and maintains forest biodiversity. The airborne lidar (Light Detection and Ranging) systems have the advantage of directly measuring three-dimensional vegetation structure, and have been widely used in wildlife habitat mapping and species distribution modeling at the local scales. As lidar data are increasingly compiled into broad spatial coverage, the development of structural inventory and indicators to categorize habitat types and identify important patches would be beneficial to regional-level conservation planning and biodiversity monitoring. However, this area of research has not been adequately explored. Large-area mapping of critical habitat patches is also a fundamental step towards modeling habitat connectivity. Quantification and dynamic modeling of habitat connectivity under long-term influence of land cover change events provide insights into forest management and conservation planning, and including climate change constraints into the modeling framework also helps maintain ecosystem integrity and improve conservation effectiveness.Therefore, the objectives of this thesis are to 1) characterize vegetation structure and identify important habitat patches with critical structural traits using regional lidar dataset, and 2) build habitat networks to model connectivity dynamics under land cover change events. To do this, first, a novel approach using cluster analysis to process large-area lidar data into categorical classes representing natural groupings of habitat structure was applied to derive eight unique structure classes in the managed forested area in Alberta, Canada. Second, the structure classes indicating high levels of structure complexity combined with Landsat-derived forest cover types were used to identify important habitat patches to develop habitat networks. Lastly, spatial prioritization schemes based on different aspects of connectivity and climate constraints were generated and implemented through scenario-based simulations of land cover change events. Connectivity dynamics through the simulations were assessed and compared between scenarios. The result showed that the conservation strategies considering both habitat area and habitat spatial configuration were best at maintaining habitat connectivity, and taking climate constraints into consideration didn’t affect overall connectivity. Overall, this research provides an integrated approach to characterize habitat structure using large-area lidar data for dynamic connectivity modeling following land cover change simulations.

View record

Invasive plants are increasingly present in ecosystems, producing both positive and negative effects. Proactive management of plant invasions is critical to curbing their spreads, especially in urban areas which often act as centres of invasions. Therefore, municipalities require new tools to map invasions for both management and information. Remote sensing technologies provide opportunities to detect plant invasions over large areas at fine spatial resolutions. In Surrey, British Columbia, Canada, Himalayan blackberry (HB; Rubus armeniacus) and English ivy (EI; Hedera helix) are two understory invasive plants that can negatively influence native ecosystems and harm users of urban natural areas. Two remote sensing technologies, hyperspectral imagery and light detection and ranging (LiDAR) data, were utilized to map these two species across the entire area of Surrey. Analysis of spectral characteristics of HB and EI were used with hyperspectral imagery to examine the feasibility of spectrally detecting these species. Spectra were obtained from a ground-based handheld spectrometer from the two species and other common species in Surrey and processed through a spectral channel selection algorithm to identify key wavelengths for distinguishing these species. Once identified, a spectral classification routine used these wavelengths and training plots to detect HB and EI across open areas in Surrey. Results showed accuracies of 76.4% for HB and 80.0% for EI. Mapping HB and EI across all land covers of Surrey required detecting the two species in forested areas. Field plots, LiDAR-derived topographic and forest structure variables, hyperspectral data, a land cover classification, and a LiDAR-derived irradiance model were all used as inputs into random forest models to detect the species across the entire land base. Model accuracies ranged from 77.8% to 87.8%. Open areas were classified better than forested areas. EI was found more across the city than HB. The research in the thesis has advanced detection of invasive plants by demonstrating the feasibility of mapping understory invasions of EI and HB in urban areas at fine spatial resolutions and can form the basis for a future monitoring system using data acquired at regular intervals. Future work is recommended to enhance data collection and increase map specificity.

View record

As of 2016, there were 57 community forestry organizations in British Columbia apart of various community forest agreements (CFA). Community forests allow for the development of multi-use management plans to reflect a diverse set of values. The availability of detailed information of the forested area is vital to maximizing a community’s benefits and profits. Airborne laser scanning (ALS) can provide estimates of conventional forest attributes, advance inventory attributes along with spatially describing ecosystem services (ES). This thesis combines ALS data, ground sampling data and vegetation resource inventory (VRI) data for the Sunshine Coast Community Forest (SCCF) located near Sechelt, British Columbia in a case study of the application of ALS data to benefit a community forest. Primary attributes (height, diameter at breast height, stem number, quadratic mean diameter, Lorey’s height, volume and biomass) were calculated using an area-based-approach. A secondary attribute (stem size distribution) was calculated using a two-parameter Weibull probability density function. Finally, a tertiary attribute - site indices - was calculated using maximum height from ALS. The reliability of primary attributes predictions varied, with stem number being the poorest (R²=0.51, p-value
View record

Detailed observations of natural and anthropogenic disturbances that alter the forest structure and the distribution of carbon are essential to estimate changes in forest carbon sinks and sources. Remote sensing is one of the primary sources to provide observations of land cover and land-cover change for carbon studies and other ecological applications due to its ability to monitor the Earth’s surface on a regular and continuous basis. However, observations of change are often not attributed directly to an underlying disturbance type and are not well validated, especially in tropical areas.The overall objectives of this thesis are to 1) assess forest disturbances (natural and anthropogenic) and derive activity data for carbon budget modeling, and 2) estimate the impact of different activity data on the terrestrial carbon balance for REDD+ in Mexican tropical forests. To do so, a novel Multi-Source, Multi-Scale Disturbance (MS-D) assessment method was developed to: 1) characterize natural and anthropogenic forest disturbances; 2) obtain land-cover change observations; and 3) attribute land-cover changes to their most likely disturbance driver. Spatially-explicit layers of major disturbance types were generated in annual time steps for carbon modeling across the Yucatan Peninsula from 2005 to 2010. Using geospatial techniques and regression-tree analysis the MS-D approach successfully attributed 86% of land-cover changes derived from the MODIS satellite imagery to their underlying disturbance cause, creating synergies between remote-sensing products, forest inventory and ancillary datasets. Four remote-sensing products derived from Landsat and MODIS satellites were then compiled, providing inputs of activity data for carbon modeling with the CBM-CFS3. Two map sequences were generated for each product, with and without attributing land-cover changes to disturbance type with the MS-D approach. Annual carbon fluxes were simulated to compare the impact of: 1) spatial resolution, 2) temporal resolution, and 3) attribution/non-attribution of land-cover changes by disturbance type on carbon flux estimates. The results clearly demonstrated that different choices of satellite imagery and attribution of changes to disturbance types change the estimated carbon balance. This study provides an integral cost-effective approach to derive activity data for carbon modeling, and support policy and decision-making for forest monitoring and REDD+.

View record

Improving our ability to track and monitor changes on Earth’s surface will inevitably enhance our ability to manage and monitor the biosphere. Remote Sensing technologies developed to monitor the Earth’s surface have already improved our understating of dynamic land cover change at a variety of scales. Fundamental to the identification of land cover change is the detection of abrupt disturbance events. These events constitute direct changes to the composition and structure of ecological systems and may have long lasting effects. In a forestry context it is important to identify disturbances in a timely manner in order to inform management decisions. The RapidEye constellation is a series of five identical Earth orbiting optical sensors capable of achieving five meter spatial resolution imagery with a daily return time. In this thesis we present two studies which assess the capacity of RapidEye to detect (1) stand replacing disturbances and (2) non-stand replacing disturbances in British Columbia. In the first study we develop a robust method to identify stand-replacing disturbances across seven regions in British Columbia. Overall accuracy for the classification of forest disturbance ranged from 83.65 ± 0.77% to 97.65 ± 0.25% for individual 25 X 25 km test locations.In the second study the utility of the RapidEye constellation to detect and characterize a low severity fire in a dry Western Canadian Forest was examined. Estimates of burn severity from field data were correlated with a selected suite of common spectral vegetation indices. All correlations between the ground estimates and vegetation indices produced significant results (p
View record

Previous work has shown that many populations of birds residing in British Columbia are declining in number. However, given the size and remoteness of many parts of the province, direct sampling of BC’s bird communities throughout the province is unlikely. Remote sensing has been shown to be an attractive option in these types of situations, providing environmental data in remote areas at spatial scales appropriate for a provincial level of analysis. Additionally, the spatial coverage of remotely sensed data allows for species occurrences to be estimated in un-sampled locations using models derived from areas where sampling has occurred.The overall objectives of this thesis are twofold, and were tested in two separate studies. First, the ability of remotely sensed environmental variables to predict the distribution of coastal bird species for the entire BC coast was investigated. Second, multiple environmental regionalization schemes were evaluated with regard to their ability to delineate avian Beta diversity across the province, and were compared to a regionalization built using species data directly.In Chapter 3, the distributions of 60 species of birds were estimated along the BC coast. Distribution models were built using species occupancy data linked to oceanic, terrestrial, and anthropogenic remotely sensed variables, as well as interpolated climate indices and spatial variables, in both single and ensemble models. The use of these four different types of environmental variables improved the distribution models’ ability to estimate species occurrence, as did the use of ensemble modeling and the inclusion of spatial variables. Significant changes in the amount of occupied habitat by year were detected in 16 species for the eight year study period. In Chapter 4, four environmental conservation regionalization schemes were compared using analysis of similarity (ANOSIM) tests to assess their ability to delineate Beta diversity. A new, species-based regionalization was then created to act as an ideal scenario, and was subsequently tested against each environmental regionalization scheme. These analyses demonstrated that all environmental regionalization schemes delineated significant patterns in Beta diversity, with the Bird Conservation Regions scoring highest in ANOSIM testing overall and being the most similar to the species-based regionalization.

View record

Forests are considered important reservoirs of organic carbon and have been identified as essential in moderating climate change. Measuring the amount of carbon stored in forests helps improve our understanding of the carbon budget and help with climate change adaptation strategies. Therefore, effective and accurate methods in characterizing changing forest cover and biomass densities are needed.Both LiDAR (light detection and ranging) and radar (radio detection and ranging) technologies can contribute towards the study of forest biomass but one sensor alone cannot provide all the information necessary to monitor forests. Understanding and investigating synergies between different remotely sensed data sets provides new and innovative opportunities to monitor forests.The overall objective reported in this thesis is to demonstrate novel methods to integrate two remotely sensed data sets (i.e., radar and LiDAR) for the application of biomass estimation. This research was divided into two main questions: (1) can shorter wavelength radar variables provide improved biomass estimates when combined with LiDAR data; and (2) can the use of space-borne radar extend aboveground biomass estimates over a larger area using spatial modeling methods.In the first study, relationships between biomass and biomass components with LiDAR and radar data were examined through regression analyses to determine the best combined parameters to estimate biomass. Results indicated that integrating radar variables to a LiDAR-derived model of aboveground biomass helped explain an additional 17.9% of the variability in crown biomass. This corresponded in an improvement in crown biomass estimates of 10% RMSE. Furthermore, InSAR coherence magnitudes from C-band and L-band radars provided the best estimate of aboveground biomass using radar alone.In the second study, aboveground biomass transects derived from plot-based field data and LiDAR, and wall-to-wall radar were spatially integrated using three kriging techniques. The results indicated the importance of correlation between primary and secondary variables when using these kriging approaches. Also a 1000 m distance between biomass transects, was found to provide reasonable compromise between ease of use, accuracy, and cost of obtaining LiDAR data for the study area. Insights into other opportunities for further development in spatial modeling techniques are discussed.

View record

Natural variability and disturbance events drive spatial and temporal variation in ecosystem processes and play key roles in ecosystem variety and the maintenance of species diversity. As a result, an improved understanding of the links between natural environmental variability and species diversity is needed to guide prioritisation of conservation and management actions. Ontario, the second largest province in Canada, covering approximately 1 million km², is environmentally diverse and is subject to a large amount of natural and anthropogenic disturbances. Remote sensing is uniquely capable of monitoring dynamic ecosystems over large areas in a repeatable and cost effective manner and has been shown to provide considerable benefit to assess species distribution and biodiversity.This thesis (1) examines an approach for detecting natural variability and disturbances of vegetation productivity from a remote sensing time-series and (2) demonstrates the use of satellite-derived indicators for the characterisation of moose habitat across Ontario. First, an approach was developed to assess temporal trends in vegetation productivity which utilised a Theil-Sen’s non-parametric statistical trend test over a 6-year period (2003-2008) of ten-day composites of Medium Resolution Imaging Spectroradiometer (MERIS) fraction of Photosynthetically Active Radiation (fPAR). Results indicated that this novel remote sensing approach can be used to characterise trends in landscape productivity patterns over large areas and can aid in provincial and national monitoring activities. Second, the research investigated the application of remotely sensed indicators such as vegetation productivity, land cover, topography, snow cover and natural and anthropogenic disturbances to predict moose occurrence and abundance. Results indicated that remotely sensed indicators were significantly correlated to moose habitat suitability with moose distribution being more accurately estimated than moose abundance. In addition to providing insights into the relative importance of the predictor covariates for moose occurrence and abundance, this study creates opportunities for further development of spatial models that closely examine the occurrence/abundance-habitat relationships which are highly valuable for habitat management decisions.

View record

National parks in western Canada experience wildland fire events at differing frequencies, intensities, and burn severities. These episodic disturbances have varying implications for various biotic and abiotic processes and patterns. To predict burn severity, the differenced Normalized Burn Ratio (dNBR) algorithm, derived from Landsat imagery, has been used extensively throughout the wildland fire community. Researchers have often employed this approach to study the effects of fire across multiple contrasting landscapes. Many remote sensing scientists have concluded that incorporating pre-fire information into the current remote sensing dNBR methodology may make such models more transferable. In the first study the main purpose was to investigate the accuracies of the absolute dNBR versus its relative form (RdNBR) to estimate burn severity, in which was hypothesized that RdNBR would outperform dNBR based on former research by Miller and Thode (2007). The secondary purpose was to examine and compare the accuracies of RdNBR and dNBR algorithms in pre-fire landscapes with low canopy closure and high heterogeneity. Results indicate that the RdNBR-derived model did not estimate burn severity more accurately than dNBR (65.2% versus 70.2% classification accuracy, respectively) nor indicate improved estimates in the more heterogeneous and low canopy cover landscapes. In addition, we concluded that RdNBR is no more effective than dNBR at the regional, individual, and fine-scale vegetation levels. The results herein support the continued use of both the dNBR and RdNBR methods and the pursuit of developing regional models.In the second study, we compare the transferability of an overall model and those stratified by land cover and ecozone. Our second objective was to test the statistical benefit of incorporating pre- and post-fire information into standard dNBR approaches. We determined that an overall dNBR derived model successfully estimated burn severity for the majority of our study fires, which supports its transferability across multiple western Canadian landscapes. Results indicate that both pre- and post-fire remote sensing information provides a means of further understanding the different post-fire responses as well as showing minimal statistical burn severity estimates across the majority of fires, however, significant improvement was evident for three of the ten study fires.

View record