Wildfire is a significant driver of forest and land cover change in the central interior of British Columbia, Canada. Fuel type maps are a primary input to fire behavior calculations and simulation studies that assess wildfire threat at the landscape level. However, these thematic maps are not easily produced at the scale and speed needed to assess and mitigate wildfire threat on an annual basis. The objective of this research was to explore how an artificial neural network could be used with remotely sensed satellite imagery to map and update fuel types on an annual basis. We applied the artificial neural network over a 40 000-km 2 landscape in central interior British Columbia that burned from a megafire in 2017. Fuel maps were generated for the years 2014-2018, assessed through an independent validation, and evaluated against an existing fuel type map. The highest cross-validation overall accuracy during training was 66.5% and overall accuracy from the independent validation was 63.1%. Generally, the maps had fair agreement with the existing fuel type map (circa 2016), with Cohen's Kappa ranging from 0.28 in 2018 to 0.35 in 2015. Several recommendations are provided for future research using artificial neural networks for fuel typing such as assuring quality of training samples through rigorous standards, designing the network architecture, choosing appropriate cost functions and regularization, incorporating learning of temporal features, and identifying novel fuel types from the output activations.

Abstract Most regulatory and certification agencies in Canada now require forest management plans to include some level of historical fire pattern approximation. As a result, sustainable forest management and enhancements to existing fire management policies and practices require a thorough understanding of the spatial fire patterns created and maintained by fire as well as the environmental conditions when they occur. To date, however, no boreal fire pattern study has examined relationships between the spatial arrangement of fire patterns, including vegetation remnants, and its main environmental top‐down and bottom‐up controls, based on a large number of fire events across large areas of the Canadian boreal forest. In this study, we leverage a recent, comprehensive, Landsat‐derived fire pattern dataset that includes information on fire vegetation remnants for the Canadian boreal plains ecozone, covering 507 fires and 2.5 Mha, to characterize the predictability of six fire pattern metrics. We then compare these metrics to multiple top‐down (monthly climate) and bottom‐up (topography, fuels, natural barriers, and disturbance history) controls on fire behavior. To do so, we first reduced the fire pattern metrics to three principal components and used a random forest modeling approach to better understand the main environmental explanatory controls. Across this large number of fires, we identified three dimensions of fire patterns: compactness, or the complexity of the perimeter; patchiness, or the spatial heterogeneity in the burned patches; and residualness, or the amount of fire vegetation remnants within the burned patches. We found that patchiness was mostly conditioned by the land cover through variables characterizing the type and connectedness of the fuels; however, summer and spring drought were locally important. Compactness responded to a combination of the disturbance history, land cover, and summer drought. The presence of water resulted in less compact fires. Residualness was a function of the disturbance history, topography, and land cover. Fires in lower elevations presented the most variable patterns, in response to changes in the amount and types of fuel. This research offers an enhanced understanding of the hierarchical interactions between resulting fire patterns and environmental conditions that are critical to supporting management decisions.

We compared three monthly adaptations of the daily Drought Code (DC) of Canada’s Fire Weather Index System and applied them to interpret drought conditions associated with historical fires in montane forests of south-eastern British Columbia. The three adaptations were compared with the monthly mean DC calculated from daily values for the Palliser fire-weather station. Two adaptations improved on the existing Monthly DC calculated from monthly climate data by (1) accounting for overwinter drying and an early start to the fire season, and (2) improving estimates of effective precipitation. Using a cross-dated fire-scar record from 20 sites in montane forests surrounding the Palliser station, we found significant fire–drought associations from June to August with all adaptations, and significant associations in April and May with the two new adaptations. Of the 17 fire years from 1901 to 2013, 6years had low initial drought conditions that increased late in the fire season, and 5 years had high drought conditions throughout the fire season. We conclude that variable drought within and among fire seasons influenced fire severity. Our findings provide a connection between modern drought indices used to rank fire danger and drought effects on the historical mixed-severity fire regime in montane forests of south-eastern British Columbia.

Measures of forest fragmentation, and how fragmentation is changing through time, offer required information for understanding the status and dynamics of forest ecosystems, habitat conditions, and ecosystem functions. In this research, we investigate the multi-temporal characterization of forest fragmentation across the forested ecosystems of Canada (> 650 million ha) and characterize the past three decades of forest fragmentation, providing useful context against which future analyses can be compared. Using 33 years of annual land cover maps produced from classified Landsat image best-available pixel composites (1984-2016), we describe and quantify the different forest patterns and dynamics that emerged in areas that were not disturbed in the analysis period, as well as following stand-replacing (i.e., wildfire, harvest) and non-stand-replacing (e.g., insects, water stress) disturbances. Baseline levels of fragmentation for each ecozone were determined by analyzing unchanged areas. Fragmentation dynamics by dominant forest disturbance showed that harvest activities generally lead to an increase in fragmentation related to the amount of forest cover (composition), while wildfires result in increasing fragmentation as a function of the spatial arrangement of the forest. The results presented herein also allow for characterization of the recovery of vegetation spatial patterns following various disturbance types, with areas dominated by fire presenting slower spatial recovery rates compared to harvest. By the end of the analysis period following disturbance events, forest fragmentation metrics in harvest-dominated landscapes were comparable to the pre-harvesting baseline, reaching 96% of mean pre-disturbance levels for mean forest patch size, and 83% for number of forest patches. In contrast, fire-dominated landscapes resulted in more event related fragmentation, with reduced forest cover (mean Thiel Sen slope = -0.13% year(-1)), mean forest patch size (-0.22 ha year(-1)), an increase in forest patches (0.11 year(-1)), and forest-non-forest join counts (0.83 year(-1)). By the end of the analysis period following disturbance events, mean forest patch size reached 96% of mean pre-disturbance levels and 68% the number of forest patches. Overall, non-stand replacing changes had no impact on the behaviour of the forest fragmentation metrics. The open access to Landsat's image archive combined with the analysis methods presented herein enable the systematic quantification and characterization of Canada-wide trends in forest fragmentation trends and post-disturbance spatial patterns over three decades. The results reported herein provide detailed information on the temporal evolution of spatial forest patterns, and illustrate that given an adequate time period, spatial patterns in areas where land use has not changed, recover to resemble pre-disturbance conditions.

Landsat time series (LTS) enable the characterization of forest recovery post-disturbance over large areas; however, there is a gap in our current knowledge concerning the linkage between spectral measures of recovery derived from LTS and actual manifestations of forest structure in regenerating stands. Airborne laser scanning (ALS) data provide useful measures of forest structure that can be used to corroborate spectral measures of forest recovery. The objective of this study was to evaluate the utility of a spectral index of recovery based on the Normalized Burn Ratio (NBR): the years to recovery, or Y2R metric, as an indicator of the return of forest vegetation following forest harvest (clearcutting). The Y2R metric has previously been defined as the number of years required for a pixel to return to 80% of its pre-disturbance NBR (NBRpre) value. In this study, the Composite2Change (C2C) algorithm was used to generate a time series of gap-free, cloud-free Landsat surface reflectance composites (1985–2012), associated change metrics, and a spatially-explicit dataset of detected changes for an actively managed forest area in southern Finland (5.3 Mha). The overall accuracy of change detection, determined using independent validation data, was 89%. Areas of forest harvesting in 1991 were then used to evaluate the Y2R metric. Four alternative recovery scenarios were evaluated, representing variations in the spectral threshold used to define Y2R: 60%, 80%, and 100% of NBRpre, and a critical value of z (i.e. the year in which the pixel's NBR value is no longer significantly different from NBRpre). The Y2R for each scenario were classified into five groups: recovery within <10 years, 10–13 years, 14–17 years, >17 years, and not recovered. Measures of forest structure (canopy height and cover) were obtained from ALS data. Benchmarks for height (>5 m) and canopy cover (>10%) were applied to each recovery scenario, and the percentage of pixels that attained both of these benchmarks for each recovery group, was determined for each Y2R scenario. Our results indicated that the Y2R metric using the 80% threshold provided the most realistic assessment of forest recovery: all pixels considered in our analysis were spectrally recovered within the analysis period, with 88.88% of recovered pixels attaining the benchmarks for both cover and height. Moreover, false positives (pixels that had recovered spectrally, but not structurally) and false negatives (pixels that had recovered structurally, but not spectrally) were minimized with the 80% threshold. This research demonstrates the efficacy of LTS-derived assessments of recovery, which can be spatially exhaustive and retrospective, providing important baseline data for forest monitoring.

Spring represents the peak of human-caused wildfire events in populated boreal forests, resulting in catastrophic loss of property and human life. Human-caused wildfire risk is anticipated to increase in northern forests as fuels become drier, on average, under warming climate scenarios and as population density increases within formerly remote regions. We investigated springtime human-caused wildfire risk derived from satellite-observed vegetation greenness in the early part of the growing season, a period of increased ignition and wildfire spread potential from snow melt to vegetation green-up with the aim of developing an early warning wildfire risk system. The initial system was developed for 392,856 km2 of forested lands with satellite observations available prior to the start of the official wildfire season and predicted peak human-caused wildfire activity with 10-day accuracy for 76% of wildfire-protected lands by March 22. The early warning system could have significant utility as a cost-effective solution for wildfire managers to prioritize the deployment of wildfire protection resources in wildfire-prone landscapes across boreal-dominated ecosystems of North America, Europe, and Russia using open access Earth observations.

Understanding the development of landscape patterns over broad spatial and temporal scales is a major contribution to ecological sciences and is a critical area of research for forested land management. Boreal forests represent an excellent case study for such research because these forests have undergone significant changes over recent decades. We analyzed the temporal trends of four widely-used landscape pattern indices for boreal forests of Canada: forest cover, largest forest patch index, forest edge density, and core (interior) forest cover. The indices were computed over landscape extents ranging from 5,000 ha (n = 18,185) to 50,000 ha (n = 1,662) and across nine major ecozones of Canada. We used 26 years of Landsat satellite imagery to derive annualized trends of the landscape pattern indices. The largest declines in forest cover, largest forest patch index, and core forest cover were observed in the Boreal Shield, Boreal Plain, and Boreal Cordillera ecozones. Forest edge density increased at all landscape extents for all ecozones. Rapidly changing landscapes, defined as the 90th percentile of forest cover change, were among the most forested initially and were characterized by four times greater decrease in largest forest patch index, three times greater increase in forest edge density, and four times greater decrease in core forest cover compared with all 50,000 ha landscapes. Moreover, approximately 18% of all 50,000 ha landscapes did not change due to a lack of disturbance. The pattern database results provide important context for forest management agencies committed to implementing ecosystem-based management strategies.

Site productivity, an important measure of the capacity of land to produce wood biomass, is traditionally estimated by applying species-specific, locally designed models that describe the relation between stand age and dominant height. In this paper we present an approach to derive chronosequences of stand age and height estimates from remotely sensed data to develop site productivity estimates. We first utilised an annual Landsat time series to identify areas of stand replacing disturbances and to estimate the time-since-disturbance, a proxy for stand age. Airborne Laser Scanning data were used to provide estimates of dominant height for these stands. Non-linear regression was used to fit a site productivity guide curve for stands aged 7 to 32 years. Existing and developed productivity models, together with remote sensing and inventory data as inputs, were used to validate the site productivity model in three different comparisons. Site productivity was overestimated by 0.70 m (RMSE = 5.55 m) relative to existing forest inventory estimates; further, 89% of remote sensing estimates were within ±1 derived site class of the forest inventory estimates. We conclude that the presented approach is suitable for estimating site productivity for young stands in areas that lack wall-to-wall forest inventory data.

ABSTRACT A critical component of landscape dynamics is the recovery of vegetation following disturbance. The objective of this research was to characterize the forest recovery trends associated with a range of spectral indicators and report their observed performance and identified limitations. Forest disturbances were mapped for a random sample of three major bioclimate zones of North American boreal forests. The mean number of years for forest to recover, defined as time required to for a pixel to attain 80% of the mean spectral value of the 2 years prior to disturbance, was estimated for each disturbed pixel. The majority of disturbed pixels recovered within the first 5 years regardless of the index ranging from approximately 78% with normalized burn ratio (NBR) to 95% with tasselled cap greenness (TCG) and after 10 years more than 93% of disturbed pixels had recovered. Recovery rates suggest that normalized differenced vegetation index (NDVI) and TCG saturate earlier than indices that emphasize longer wavelengths. Thus, indices such as NBR and the mid-infrared spectral band offer increased capacity to characterize different levels of forest recovery. The mean length of time for spectral indices to recover to 80% of the pre-disturbance value for pixels disturbed 10 or more years ago was highest for NBR, 5.6 years, and lowest for TCG, 1.7 years. The mid-infrared spectral band had the greatest difference in recovered pixels among bioclimate zones 1 year after disturbance, ranging from approximately 42% of disturbed pixels for the cold and mesic bioclimate zone to 60% for the extremely cold and mesic bioclimate zone. The cold and mesic bioclimate zone had the longest mean years to recover ranging from 1.9 years for TCG to 4.2 years for NBR, while the cool temperate and dry bioclimate zone had the shortest mean years to recover ranging from 1.6 years for TCG to 2.9 years for NBR suggesting differences in pre-disturbance conditions or successional processes. The results highlight the need for caution when selecting and interpreting a spectral index for recovery characterization, as spectral indices, based upon the constituent wavelengths, are sensitive to different vegetation conditions and will provide a variable representation of structural conditions of forests.

Resource development can have significant consequences for the distribution of vegetation cover and for species persistence. Modelling changes to anthropogenic disturbance regimes over time can provide profound insights into the mechanisms that drive land cover change. We analyzed the spatial patterns of anthropogenic disturbance before and after a period of significant oil and gas extraction in two boreal forest subregions in Alberta, Canada. A spatially explicit model was used to map levels of anthropogenic forest crown mortality across 700 000 ha of managed forest over a 60-year period. The anthropogenic disturbance regime varied both spatially and temporally and was outside the historical range of variability characterized by regional fire regimes. Levels of live forest crown within anthropogenic disturbances declined and edge density increased following oil and gas development, whereas patch size varied regionally. In some places, anthropogenic disturbance generated profoundly novel landscapes with spatial patterns that had no historical analogue in the boreal system. The results illustrate that a shift in one sector of the economy can have dramatic outcomes on landscape structure. The results also suggest that any efforts to better align cumulative anthropogenic disturbance patterns with the historic baseline will almost certainly require a concerted and collaborative effort from all of the major stakeholders.

Abstract. Site productivity, an important measure of the capacity of land to produce wood biomass, is traditionally estimated by applying species-specific, locally designed models that describe the relation between stand age and dominant height. In this article, we present an approach to derive chronosequences of stand age and height estimates from remotely sensed data to develop site productivity estimates. We first utilized an annual Landsat time series to identify areas of stand replacing disturbances and to estimate the time-since-disturbance, a proxy for stand age. Airborne laser scanning data were used to provide estimates of dominant height for these stands. Nonlinear regression was used to fit a site productivity guide curve for stands aged 7 to 32 years. Existing and developed productivity models, together with remote sensing and inventory data as inputs, were used to validate the site productivity model in three different comparisons. Site productivity was overestimated by 0.70 m (RMSE = 5.55 m) relative to existing forest inventory estimates; further, 89% of remote sensing estimates were within ±1 derived site class of the forest inventory estimates. We conclude that the presented approach is suitable for estimating site productivity for young stands in areas that lack wall-to-wall forest inventory data. Résumé. Le potentiel du site, une mesure importante de la capacité des terres à produire de la biomasse de bois, est traditionnellement estimé en appliquant des modèles spécifiques d’espèces, conçus localement, qui décrivent la relation entre l’âge du peuplement et la hauteur dominante. Dans cet article, nous présentons une approche pour dériver des séquences chronologiques d’estimations de l’âge du peuplement et de la hauteur à partir de données de télédétection pour établir des estimations du potentiel du site. Nous avons d’abord utilisé une série temporelle annuelle Landsat pour identifier les zones de perturbations menant au remplacement des peuplements et pour estimer le temps écoulé depuis la perturbation, un estimateur pour l’âge du peuplement. Des données laser aéroportées ont été utilisées pour fournir des estimations de la hauteur dominante de ces peuplements. Une régression non linéaire a été utilisée pour obtenir une courbe de potentiel du site pour les peuplements âgés de 7 à 32 ans. Les modèles de productivité existants ainsi que des données de télédétection et d’inventaire ont été utilisés comme entrées pour valider le modèle du potentiel du site à partir de trois comparaisons différentes. Le potentiel du site a été surestimé de 0,70 m (RMSE = 5,55 m) par rapport aux estimations existantes d’inventaires forestiers. De plus, 89% des estimations de télédétection étaient à ±1 classe dérivée du site des estimations de l’inventaire forestier. Nous concluons que l’approche présentée est appropriée pour estimer le potentiel du site pour les jeunes peuplements dans les zones avec des données incomplètes d’inventaire forestier.

The accelerated development of energy resources around the world has substantially increased forest change related to oil and gas activities. In some cases, oil and gas activities are the primary catalyst of land-use change in forested landscapes. We discuss the challenges associated with characterizing ecological change related to energy resource development using North America as an exemplar. We synthesize the major impacts of energy development to forested ecosystems and offer new perspectives on how to detect and monitor anthropogenic disturbance during the Anthropocene. The disturbance of North American forests for energy development has resulted in persistent linear corridors, suppression of historical disturbance regimes, novel ecosystems, and the eradication of ecological memory. Characterizing anthropogenic disturbances using conventional patch-based disturbance measures will tend to underestimate the ecological impacts of energy development. Suitable indicators of anthropogenic impacts in forests should be derived from the integration of multi-scalar Earth observations. Relating these indicators to ecosystem condition will be a capstone in the progress toward monitoring forest change in landscapes undergoing rapid energy development.

Human transformation of the terrestrial biosphere via resource utilization is a critical impetus for monitoring and characterizing anthropogenic change to vegetation condition. The primary objective of this research was to detect anthropogenic forest disturbance for a recent Landsat time series. A novel combination of an autonomous change detection procedure and spectral classification scheme was applied and tested in a landscape that has undergone significant resource development over the last 30 years. Anthropogenic disturbance was detected with greater than 93% accuracy. Most disturbances were correctly classified as within ±1 year. The signal of anthropogenic disturbance was significant in the landscape, accounting for more than 91% of all disturbances and 86% of total disturbed area during the 23-year study period. The study demonstrated a robust approach for examining historical disturbance trends related to human-modification of the environment.

Ecosystem-based management (EBM) has emerged as a dominant paradigm for the Canadian boreal forest. One of the principles of EBM is to maintain ecosystem function by means of management activities that approximate the historic patterns or processes responsible for maintaining a range of landscape conditions. This ideal has been manifested as planning schemes are shifting away from traditional sustained yield harvests toward designs based on historic wildfire disturbance patterns. Wildfire disturbance patterns represent a coarse-filter management strategy, and are well-suited to the boreal forests of Canada. Forest management professionals in the boreal have been leaders in adopting these strategies over the past decade. However, two key questions remain unanswered: (1) to what degree have these forest management efforts resulted in disturbance patterns that resemble wildfire burning patterns?; and (2) to what degree do the other sources of anthropogenic disturbance activities align with historic wildfire patterns? In this paper, an existing knowledge of historic range of variability (HRV) of wildfire patterns and the NEPTUNE (Novel Emulation Pattern Tool for Understanding Natural Events) decision support tool were used to test both questions.The results suggest that forest harvest disturbances better approximated historic disturbance patterns than did energy extraction disturbances, though in both cases some of the metrics were beyond the HRV. Significant differences were found between traditional dispersed patterns (e.g., multi-pass harvesting) and the more recent aggregate harvest (e.g., single-pass) designs. Aggregate harvests were characterized by low proportional area in matrix remnants, moderate levels of combined island and matrix remnants, and a high proportional area in the single largest disturbed patch (LDP). Dispersed harvests tended to have a higher proportional area in matrix remnants and better approximated the HRV in terms of proportional area in proportional island remnants area and largest island remnant. Aggregate harvest patterns did not perform as well for a few metrics such as proportional area in island remnants, mean island remnant size and largest island remnant. Overall, the results suggest that aggregate harvest designs better approximated key HRV patterns such as proportional matrix remnants and LDP, than did dispersed harvest designs on the landscape. The results also suggest that forest harvesting was significantly more effective at approximating historic disturbance patterns than the activities of the energy sector. Energy sector disturbances were smaller and had fewer island remnants than the HRV. The composition of surviving remnant trees within anthropogenic disturbance events (i.e., matrix and island remnants) remains a critical area of research for approximating HRV patterns. (C) 2013 Elsevier B.V. All rights reserved.

Report on study to characterise disturbance patterns of energy and forestry related disturbances, and compare them to natural wildfire patterns.