Towards a whole-system framework for wildfire monitoring using Earth observations.

Fire seasons have become increasingly variable and extreme due to changing climatological, ecological, and social conditions. Earth observation data are critical for monitoring fires and their impacts. Herein, we present a whole-system framework for identifying and synthesizing fire monitoring objectives and data needs throughout the life cycle of a fire event. The four stages of fire monitoring using Earth observation data include the following: (1) pre-fire vegetation inventories, (2) active-fire monitoring, (3) post-fire assessment, and (4) multi-scale synthesis. We identify the challenges and opportunities associated with current approaches to fire monitoring, highlighting four case studies from North American boreal, montane, and grassland ecosystems. While the case studies are localized to these ecosystems and regional contexts, they provide insights for others experiencing similar monitoring challenges worldwide. The field of remote sensing is experiencing a rapid proliferation of new data sources, providing observations that can inform all aspects of our fire monitoring framework; however, significant challenges for meeting fire monitoring objectives remain. We identify future opportunities for data sharing and rapid co-development of information products using cloud computing that benefits from open-access Earth observation and other geospatial data layers.

Uncovering spatial and ecological variability in gap size frequency distributions in the Canadian boreal forest.

Analyses characterizing canopy gaps are required to improve our understanding of spatial and structural variations in forest canopies and provide insight into ecosystem-level successional processes. Gap size frequency distributions (GSFD) are indicative of ecological processes and disturbance patterns. To date, GSFD in boreal forest ecosystems have not been systematically quantified over large areas using a single consistent data source. Herein we characterized GSFDs across the entirety of the Canadian boreal forest using transects of airborne laser scanning (ALS) data. ALS transects were representatively sampled within eight distinct Canadian boreal ecozones. Gaps were detected and delineated from the ALS-derived canopy height model as contiguous canopy openings ≥8 m2 with canopy heights ≤3 m. Gaps were then stratified by ecozone and forest type (i.e. coniferous, broadleaf, mixedwood, wetland-treed), and combinations thereof, and GSFDs were calculated for each stratum. GSFDs were characterized by the scaling parameter of the power-law probability distribution, lambda (λ) and Kolmogorov-Smirnov tests confirmed that GSFDs for each stratum followed a power-law distribution. Pairwise comparisons between ecozones, forest types, and combinations thereof indicated significant differences between estimates of λ. Scaling parameters were found to be more variable by ecozone (1.96-2.31) than by forest type (2.15-2.21). These results contrast those of similar studies done in tropical forest environments, whereby λ was found to be relatively consistent across a range of site types, geological substrates, and forest types. The geographic range considered herein is much larger than that of previous studies, and broad-scale patterns in climate, landforms, and soils that are reflected in the definition of unique ecozones, likely also influence gap characteristics.

Near Real-Time Wildfire Progression Monitoring with Sentinel-1 SAR Time Series and Deep Learning.

In recent years, the world witnessed many devastating wildfires that resulted in destructive human and environmental impacts across the globe. Emergency response and rapid response for mitigation calls for effective approaches for near real-time wildfire monitoring. Capable of penetrating clouds and smoke, and imaging day and night, Synthetic Aperture Radar (SAR) can play a critical role in wildfire monitoring. In this communication, we investigated and demonstrated the potential of Sentinel-1 SAR time series with a deep learning framework for near real-time wildfire progression monitoring. The deep learning framework, based on a Convolutional Neural Network (CNN), is developed to detect burnt areas automatically using every new SAR image acquired during the wildfires and by exploiting all available pre-fire SAR time series to characterize the temporal backscatter variations. The results show that Sentinel-1 SAR backscatter can detect wildfires and capture their temporal progression as demonstrated for three large and impactful wildfires: the 2017 Elephant Hill Fire in British Columbia, Canada, the 2018 Camp Fire in California, USA, and the 2019 Chuckegg Creek Fire in northern Alberta, Canada. Compared to the traditional log-ratio operator, CNN-based deep learning framework can better distinguish burnt areas with higher accuracy. These findings demonstrate that spaceborne SAR time series with deep learning can play a significant role for near real-time wildfire monitoring when the data becomes available at daily and hourly intervals with the launches of RADARSAT Constellation Missions in 2019, and SAR CubeSat constellations.

Breaking the Habit(at).

Earth Observation has revolutionised the mapping of species; however, habitat maps are often categorical, static representations of reality that result in issues relating to accurate change estimation and application to multiple species. We must break the habit of simplistic discrete classes of habitat and derive continuous, interval value, change-sensitive, habitat descriptions with an ability to drive hypotheses.

Uncovering regional variability in disturbance trends between parks and greater park ecosystems across Canada (1985-2015).

We assess the protective function of Canada's parks and protected areas (PPAs) by analyzing three decades of stand-replacing disturbance derived from Landsat time series data (1985-2015). Specifically, we compared rates of wildfire and harvest within 1,415 PPAs against rates of disturbance in surrounding greater park ecosystems (GPEs). We found that disturbance rates in GPEs were significantly higher (p < 0.05) than in corresponding PPAs in southern managed forests (six of Canada's 12 forested ecozones). Higher disturbance rates in GPEs were attributed to harvesting activities, as the area impacted by wildfire was not significantly different between GPEs and PPAs in any ecozone. The area burned within PPAs and corresponding GPEs was highly correlated (r = 0.90), whereas the area harvested was weakly correlated (r = 0.19). The average area burned in PPAs/GPEs below 55° N was low (0.05% yr-1) largely due to fire suppression aimed at protecting communities, timber, and recreational values, while the average burn rate was higher in northern PPAs/GPEs where fire suppression is uncommon (0.40% yr-1 in PPAs/GPEs above 55° N). Assessing regional variability in disturbance patterns and the pressures faced by PPAs can better inform policy and protection goals across Canada and the globe.

A thirty year, fine-scale, characterization of area burned in Canadian forests shows evidence of regionally increasing trends in the last decade.

Fire as a dominant disturbance has profound implications on the terrestrial carbon cycle. We present the first ever multi-decadal, spatially-explicit, 30 meter assessment of fire regimes across the forested ecoregions of Canada at an annual time-step. From 1985 to 2015, 51 Mha burned, impacting over 6.5% of forested ecosystems. Mean annual area burned was 1,651,818 ha and varied markedly (σ = 1,116,119), with 25% of the total area burned occurring in three years: 1989, 1995, and 2015. Boreal forest types contained 98% of the total area burned, with the conifer-dominated Boreal Shield containing one-third of all burned area. While results confirm no significant national trend in burned area for the period of 1985 to 2015, a significant national increasing trend (α = 0.05) of 11% per year was evident for the past decade (2006 to 2015). Regionally, a significant increasing trend in total burned area from 1985 to 2015 was observed in the Montane Cordillera (2.4% increase per year), while the Taiga Plains and Taiga Shield West displayed significant increasing trends from 2006 to 2015 (26.1% and 12.7% increases per year, respectively). The Atlantic Maritime, which had the lowest burned area of all ecozones (0.01% burned per year), was the only ecozone to display a significant negative trend (2.4% decrease per year) from 1985 to 2015. Given the century-long fire return intervals in many of these ecozones, and large annual variability in burned area, short-term trends need to be interpreted with caution. Additional interpretive cautions are related to year used for trend initiation and the nature and extents of spatial regionalizations used for summarizing findings. The results of our analysis provide a baseline for monitoring future national and regional trends in burned area and offer spatially and temporally detailed insights to inform science, policy, and management.

Estimating changes in lichen mat volume through time and related effects on barren ground caribou (Rangifer tarandus groenlandicus) movement.

Lichens form a critical portion of barren ground caribou (Rangifer tarandus groenlandicus) diets, especially during winter months. Here, we assess lichen mat volume across five herd ranges in the Northwest Territories and Nunavut, Canada, using newly developed composite Landsat imagery. The lichen volume estimator (LVE) was adapted for use across 700 000 km2 of barren ground caribou habitat annually from 1984-2012. We subsequently assessed how LVE changed temporally throughout the time series for each pixel using Theil-Sen's slopes, and spatially by assessing whether slope values were centered in local clusters of similar values. Additionally, we assessed how LVE estimates resulted in changes in barren ground caribou movement rates using an extensive telemetry data set from 2006-2011. The Ahiak/Beverly herd had the largest overall increase in LVE (median = 0.033), while the more western herds had the least (median slopes below zero in all cases). LVE slope pixels were arranged in significant clusters across the study area, with the Cape Bathurst, Bathurst, and Bluenose East herds having the most significant clusters of negative slopes (more than 20% of vegetated land in each case). The Ahiak/Beverly and Bluenose West had the most significant positive clusters (16.3% and 18.5% of vegetated land respectively). Barren ground caribou displayed complex reactions to changing lichen conditions depending on season; the majority of detected associations with movement data agreed with current understanding of barren ground caribou foraging behavior (the exception was an increase in movement velocity at high lichen volume estimates in Fall). The temporal assessment of LVE identified areas where shifts in ecological conditions may have resulted in changing lichen mat conditions, while assessing the slope estimates for clustering identified zones beyond the pixel scale where forage conditions may be changing. Lichen volume estimates associated with barren ground caribou movement metrics in an expected manner and, as such, show value for future habitat assessments.

Forest Connectivity Regions of Canada Using Circuit Theory and Image Analysis.

Ecological processes are increasingly well understood over smaller areas, yet information regarding interconnections and the hierarchical nature of ecosystems remains less studied and understood. Information on connectivity over large areas with high resolution source information provides for both local detail and regional context. The emerging capacity to apply circuit theory to create maps of omnidirectional connectivity provides an opportunity for improved and quantitative depictions of forest connectivity, supporting the formation and testing of hypotheses about the density of animal movement, ecosystem structure, and related links to natural and anthropogenic forces. In this research, our goal was to delineate regions where connectivity regimes are similar across the boreal region of Canada using new quantitative analyses for characterizing connectivity over large areas (e.g., millions of hectares). Utilizing the Earth Observation for Sustainable Development of forests (EOSD) circa 2000 Landsat-derived land-cover map, we created and analyzed a national-scale map of omnidirectional forest connectivity at 25m resolution over 10000 tiles of 625 km2 each, spanning the forested regions of Canada. Using image recognition software to detect corridors, pinch points, and barriers to movements at multiple spatial scales in each tile, we developed a simple measure of the structural complexity of connectivity patterns in omnidirectional connectivity maps. We then mapped the Circuitscape resistance distance measure and used it in conjunction with the complexity data to study connectivity characteristics in each forested ecozone. Ecozone boundaries masked substantial systematic patterns in connectivity characteristics that are uncovered using a new classification of connectivity patterns that revealed six clear groups of forest connectivity patterns found in Canada. The resulting maps allow exploration of omnidirectional forest connectivity patterns at full resolution while permitting quantitative analyses of connectivity over broad areas, informing modeling, planning and monitoring efforts.

Barren-ground caribou (Rangifer tarandus groenlandicus) behaviour after recent fire events; integrating caribou telemetry data with Landsat fire detection techniques.

Fire regimes are changing throughout the North American boreal forest in complex ways. Fire is also a major factor governing access to high-quality forage such as terricholous lichens for barren-ground caribou (Rangifer tarandus groenlandicus). Additionally, fire alters forest structure which can affect barren-ground caribou's ability to navigate in a landscape. Here, we characterize how the size and severity of fires are changing across five barren-ground caribou herd ranges in the Northwest Territories and Nunavut, Canada. Additionally, we demonstrate how time since fire, fire severity, and season result in complex changes in caribou behavioural metrics estimated using telemetry data. Fire disturbances were identified using novel gap-free Landsat surface reflectance composites from 1985 to 2011 across all herd ranges. Burn severity was estimated using the differenced normalized burn ratio. Annual area burned and burn severity were assessed through time for each herd and related to two behavioural metrics: velocity and relative turning angle. Neither annual area burned nor burn severity displayed any temporal trend within the study period. However, certain herds, such as the Ahiak/Beverly, have more exposure to fire than other herds (i.e. Cape Bathurst had a maximum forested area burned of less than 4 km2 ). Time since fire and burn severity both significantly affected velocity and relative turning angles. During fall, winter, and spring, fire virtually eliminated foraging-focused behaviour for all 26 years of analysis while more severe fires resulted in a marked increase in movement-focused behaviour compared to unburnt patches. Between seasons, caribou used burned areas as early as 1-year postfire, demonstrating complex, nonlinear reactions to time since fire, fire severity, and season. In all cases, increases in movement-focused behaviour were detected postfire. We conclude that changes in caribou behaviour immediately postfire are primarily driven by changes in forest structure rather than changes in terricholous lichen availability.

Improving carbon monitoring and reporting in forests using spatially-explicit information.

BACKGROUND: Understanding and quantifying carbon (C) exchanges between the biosphere and the atmosphere-specifically the process of C removal from the atmosphere, and how this process is changing-is the basis for developing appropriate adaptation and mitigation strategies for climate change. Monitoring forest systems and reporting on greenhouse gas (GHG) emissions and removals are now required components of international efforts aimed at mitigating rising atmospheric GHG. Spatially-explicit information about forests can improve the estimates of GHG emissions and removals. However, at present, remotely-sensed information on forest change is not commonly integrated into GHG reporting systems. New, detailed (30-m spatial resolution) forest change products derived from satellite time series informing on location, magnitude, and type of change, at an annual time step, have recently become available. Here we estimate the forest GHG balance using these new Landsat-based change data, a spatial forest inventory, and develop yield curves as inputs to the Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3) to estimate GHG emissions and removals at a 30 m resolution for a 13 Mha pilot area in Saskatchewan, Canada. RESULTS: Our results depict the forests as cumulative C sink (17.98 Tg C or 0.64 Tg C year-1) between 1984 and 2012 with an average C density of 206.5 (±0.6) Mg C ha-1. Comparisons between our estimates and estimates from Canada's National Forest Carbon Monitoring, Accounting and Reporting System (NFCMARS) were possible only on a subset of our study area. In our simulations the area was a C sink, while the official reporting simulations, it was a C source. Forest area and overall C stock estimates also differ between the two simulated estimates. CONCLUSIONS: Both estimates have similar uncertainties, but the spatially-explicit results we present here better quantify the potential improvement brought on by spatially-explicit modelling. We discuss the source of the differences between these estimates. This study represents an important first step towards the integration of spatially-explicit information into Canada's NFCMARS.

Applying circuit theory for corridor expansion and management at regional scales: tiling, pinch points, and omnidirectional connectivity.

Connectivity models are useful tools that improve the ability of researchers and managers to plan land use for conservation and preservation. Most connectivity models function in a point-to-point or patch-to-patch fashion, limiting their use for assessing connectivity over very large areas. In large or highly fragmented systems, there may be so many habitat patches of interest that assessing connectivity among all possible combinations is prohibitive. To overcome these conceptual and practical limitations, we hypothesized that minor adaptation of the Circuitscape model can allow the creation of omnidirectional connectivity maps illustrating flow paths and variations in the ease of travel across a large study area. We tested this hypothesis in a 24,300 km(2) study area centered on the Montérégie region near Montréal, Québec. We executed the circuit model in overlapping tiles covering the study region. Current was passed across the surface of each tile in orthogonal directions, and then the tiles were reassembled to create directional and omnidirectional maps of connectivity. The resulting mosaics provide a continuous view of connectivity in the entire study area at the full original resolution. We quantified differences between mosaics created using different tile and buffer sizes and developed a measure of the prominence of seams in mosaics formed with this approach. The mosaics clearly show variations in current flow driven by subtle aspects of landscape composition and configuration. Shown prominently in mosaics are pinch points, narrow corridors where organisms appear to be required to traverse when moving through the landscape. Using modest computational resources, these continuous, fine-scale maps of nearly unlimited size allow the identification of movement paths and barriers that affect connectivity. This effort develops a powerful new application of circuit models by pinpointing areas of importance for conservation, broadening the potential for addressing intriguing questions about resource use, animal distribution, and movement.

Exploring the ecological processes driving geographical patterns of breeding bird richness in British Columbia, Canada.

British Columbia (BC), Canada, has a diverse landscape that provides breeding habitat for > 300 avian species, and the recent development of the BC Breeding Bird Atlas data set presents key information for exploring the landscape conditions which lead to biological richness. We used the volunteer-collected raw breeding bird evidence data set to analyze the effects of sampling biases on spatial distribution of observed breeding bird species and implemented regression tree analysis (Random Forests) to examine the influence of productivity, ambient energy, and habitat heterogeneity on independently measured breeding bird richness. Results indicated that total breeding species richness is correlated with total survey effort (alpha < 0.001). By stratifying species richness by survey effort, we observed that ambient energy is the top-ranked environmental predictor of breeding bird richness across BC, which, when used in combination with a number of other environmental variables, explains -40% of the variation in richness. Using our modeled relationships, we predicted breeding bird species richness in the areas of BC not presently surveyed between three and six hours. The majority of the productive Boreal Plains, the southern portion of the Taiga Plains region, the lowlands of the Southern and Central Interior, along the Rocky Mountain Trench, and the coastal areas of the Georgia Depression are predicted to have the highest categories of breeding richness (35-57 unique species). Our results support ongoing species diversity gradient research, which identifies ambient energy as an important factor influencing species diversity distributions in the Northern Hemisphere. By linking breeding bird richness to environmental data derived from remotely sensed data and systematically collected climate data, we demonstrate the potential to monitor shifts in ambient energy as a surrogate for vertebrate habitat condition affecting population levels. By analyzing the influence of survey effort on species richness metrics, we also highlight the need to consider adding attributes to the raw breeding bird data set to describe observer experience, such as hours or seasons spent surveying, and provide survey dates to allow greater flexibility for removing survey bias. These additions can increase the utility of atlas data for species richness studies useful for conservation planning.

Measuring forest structure along productivity gradients in the Canadian boreal with small-footprint Lidar.

The structure and productivity of boreal forests are key components of the global carbon cycle and impact the resources and habitats available for species. With this research, we characterized the relationship between measurements of forest structure and satellite-derived estimates of gross primary production (GPP) over the Canadian boreal. We acquired stand level indicators of canopy cover, canopy height, and structural complexity from nearly 25,000 km of small-footprint discrete return Light Detection and Ranging (Lidar) data and compared these attributes to GPP estimates derived from the MODerate resolution Imaging Spectroradiometer (MODIS). While limited in our capacity to control for stand age, we removed recently disturbed and managed forests using information on fire history, roads, and anthropogenic change. We found that MODIS GPP was strongly linked to Lidar-derived canopy cover (r = 0.74, p < 0.01), however was only weakly related to Lidar-derived canopy height and structural complexity as these attributes are largely a function of stand age. A relationship was apparent between MODIS GPP and the maximum sampled heights derived from Lidar as growth rates and resource availability likely limit tree height in the prolonged absence of disturbance. The most structurally complex stands, as measured by the coefficient of variation of Lidar return heights, occurred where MODIS GPP was highest as productive boreal stands are expected to contain a wider range of tree heights and transition to uneven-aged structures faster than less productive stands. While MODIS GPP related near-linearly to Lidar-derived canopy cover, the weaker relationships to Lidar-derived canopy height and structural complexity highlight the importance of stand age in determining the structure of boreal forests. We conclude that an improved quantification of how both productivity and disturbance shape stand structure is needed to better understand the current state of boreal forests in Canada and how these forests are changing in response to changing climate and disturbance regimes.

Ecosystem classifications based on summer and winter conditions.

Ecosystem classifications map an area into relatively homogenous units for environmental research, monitoring, and management. However, their effectiveness is rarely tested. Here, three classifications are (1) defined and characterized for Canada along summertime productivity (moderate-resolution imaging spectrometer fraction of absorbed photosynthetically active radiation) and wintertime snow conditions (special sensor microwave/imager snow water equivalent), independently and in combination, and (2) comparatively evaluated to determine the ability of each classification to represent the spatial and environmental patterns of alternative schemes, including the Canadian ecozone framework. All classifications depicted similar patterns across Canada, but detailed class distributions differed. Class spatial characteristics varied with environmental conditions within classifications, but were comparable between classifications. There was moderate correspondence between classifications. The strongest association was between productivity classes and ecozones. The classification along both productivity and snow balanced these two sets of variables, yielding intermediate levels of association in all pairwise comparisons. Despite relatively low spatial agreement between classifications, they successfully captured patterns of the environmental conditions underlying alternate schemes (e.g., snow classes explained variation in productivity and vice versa). The performance of ecosystem classifications and the relevance of their input variables depend on the environmental patterns and processes used for applications and evaluation. Productivity or snow regimes, as constructed here, may be desirable when summarizing patterns controlled by summer- or wintertime conditions, respectively, or of climate change responses. General purpose ecosystem classifications should include both sets of drivers. Classifications should be carefully, quantitatively, and comparatively evaluated relative to a particular application prior to their implementation as monitoring and assessment frameworks.

Comparing the impacts of mitigation and non-mitigation on mountain pine beetle populations.

Mountain pine beetles, Dendroctonus ponderosae (Hopkins) attack and can ultimately kill individuals and groups of pine trees, specifically lodgepole pine (Pinus contorta Dougl. ex. Loud var. latifolia Engl.). In British Columbia, beetle attack has increased from 164 000 ha in 1999 to over 13 million ha in 2008. Mitigation efforts can play a key role in addressing the impact beetle infestations can have on the forested landscape. In this research, the impact of mitigation on a mountain pine beetle infestation is examined within a network of 28 research plots where sanitation harvesting was completed (10 mitigated plots) and not completed (18 unmitigated plots). Three forest stand level modelling scenarios which predict the number of attacked trees, based on current infestation within the plots, were utilized to compare the differences between mitigated and non-mitigated plots. In the first scenario in the non-mitigated plots, 125 trees were infested after 10 years, while in the mitigated plots no trees were infested in the same time period. The second scenario indicates the level of mitigation required to suppress beetle infestations where the proportion of mitigated trees was calculated for each plot by counting the residual attack and the number of mitigated trees. The average mitigation rate over all plots of 43% (range 0-100%) is not sufficient to provide control. In the non-mitigated plots, the average population expansion rate was 5 (range of 0-18) which requires a detection accuracy of 74% to reliably detect infestation. The third scenario estimated the length of time required for ongoing detection, monitoring, and mitigation to bring an infestation under control. If mitigation efforts were maintained at the current rate of 43%, the beetle population would not be adequately controlled. However, when aided by continued detection and monitoring of attacked trees, mitigation rates greater than 50% are sufficient to control infestations, especially with persistent implementation, aided by continued detection and monitoring of infested trees.

Using digital time-lapse cameras to monitor species-specific understorey and overstorey phenology in support of wildlife habitat assessment.

Critical to habitat management is the understanding of not only the location of animal food resources, but also the timing of their availability. Grizzly bear (Ursus arctos) diets, for example, shift seasonally as different vegetation species enter key phenological phases. In this paper, we describe the use of a network of seven ground-based digital camera systems to monitor understorey and overstorey vegetation within species-specific regions of interest. Established across an elevation gradient in western Alberta, Canada, the cameras collected true-colour (RGB) images daily from 13 April 2009 to 27 October 2009. Fourth-order polynomials were fit to an RGB-derived index, which was then compared to field-based observations of phenological phases. Using linear regression to statistically relate the camera and field data, results indicated that 61% (r (2) = 0.61, df = 1, F = 14.3, p = 0.0043) of the variance observed in the field phenological phase data is captured by the cameras for the start of the growing season and 72% (r (2) = 0.72, df = 1, F = 23.09, p = 0.0009) of the variance in length of growing season. Based on the linear regression models, the mean absolute differences in residuals between predicted and observed start of growing season and length of growing season were 4 and 6 days, respectively. This work extends upon previous research by demonstrating that specific understorey and overstorey species can be targeted for phenological monitoring in a forested environment, using readily available digital camera technology and RGB-based vegetation indices.

Multiscale satellite and spatial information and analysis framework in support of a large-area forest monitoring and inventory update.

Many countries undertake a national forest inventory to enable statistically valid monitoring in support of national and international reporting of forest conditions and change. Canada's National Forest Inventory (NFI) program is designed to operate on a 10-year remeasurement cycle, with an interim report produced at the 5-year mid-point. The NFI is a sample-based inventory, with approximately 18,850 2×2-km photo plots across the country, distributed on a 20×20-km grid of sample points; these photo plots are the primary data source for the NFI. Capacity to provide annual monitoring information is required to keep policy and decision makers apprised of current forest conditions. In this study, we implemented a multistage monitoring framework and used a Moderate Resolution Imaging Spectroradiometer (MODIS) change product to successfully identify 78% of the changes in forest cover area that were captured with a Landsat change detection approach. Of the NFI photo plots that were identified by both the Landsat and MODIS approaches as having changes in forest cover, the proportion of change area within the plots was similar (R2=0.78). Approximately 70% of the Landsat-derived change events occupied less than 40% of a single MODIS pixel, and more than 90% of the change events of this size were successfully detected with the MODIS product. Finally, MODIS estimates of the proportion of forest cover change at the NFI photo plot level were comparable to change estimates for the ecoregions as a whole (R2=0.95). High-temporal, low-spatial resolution imagery such as MODIS, in combination with other remotely sensed data sources, can provide information on disturbance events within a national forest inventory remeasurement cycle, thereby satisfying the interim information needs of policy and decision makers as well as the requirements of national and international reporting commitments.

Characterizing the forest fragmentation of Canada's national parks.

Characterizing the amount and configuration of forests can provide insights into habitat quality, biodiversity, and land use. The establishment of protected areas can be a mechanism for maintaining large, contiguous areas of forests, and the loss and fragmentation of forest habitat is a potential threat to Canada's national park system. Using the Earth Observation for Sustainable Development of Forests (EOSD) land cover product (EOSD LC 2000), we characterize the circa 2000 forest patterns in 26 of Canada's national parks and compare these to forest patterns in the ecological units surrounding these parks, referred to as the greater park ecosystem (GPE). Five landscape pattern metrics were analyzed: number of forest patches, mean forest patch size (hectare), standard deviation of forest patch size (hectare), mean forest patch perimeter-to-area ratio (meters per hectare), and edge density of forest patches (meters per hectare). An assumption is often made that forests within park boundaries are less fragmented than the surrounding GPE, as indicated by fewer forest patches, a larger mean forest patch size, less variability in forest patch size, a lower perimeter-to-area ratio, and lower forest edge density. Of the 26 national parks we analyzed, 58% had significantly fewer patches, 46% had a significantly larger mean forest patch size (23% were not significantly different), and 46% had a significantly smaller standard deviation of forest patch size (31% were not significantly different), relative to their GPEs. For forest patch perimeter-to-area ratio and forest edge density, equal proportions of parks had values that were significantly larger or smaller than their respective GPEs and no clear trend emerged. In summary, all the national parks we analyzed, with the exception of the Georgian Bay Islands, were found to be significantly different from their corresponding GPE for at least one of the five metrics assessed, and 50% of the 26 parks were significantly different from their respective GPEs for all of the metrics assessed. The EOSD LC 2000 provides a heretofore unavailable dataset for characterizing broad trends in forest fragmentation in Canada's national parks and in their surrounding GPEs. The interpretation of forest fragmentation metrics must be guided by the underlying land cover context, as many forested ecosystems in Canada are naturally fragmented due to wetlands and topography. Furthermore, interpretation must also consider the management context, as some parks are designed to preserve fragmented habitats. An analysis of forest pattern such as that described herein provides a baseline, from which changes in fragmentation patterns over time could be monitored, enabled by earth observation data.