Steve G. Cumming

Statistical models of landscape pattern metrics, with applications to regional scale dynamic forest simulations

Forest managers in Canada need to model landscape pattern or spatial configurationoverlarge (100,000 km2) regions. This presents a scalingproblem, as landscape configuration is measured at a high spatial resolution,but a low spatial resolution is indicated for regional simulation. We present astatistical solution to this scaling problem by showing how a wide range oflandscape pattern metrics can be modelled from low resolution data. Our studyarea comprises about 75,000 km2 of boreal mixedwoodforest in northeast Alberta, Canada. Within this area we gridded a sample of 84digital forest cover maps, each about 9500 ha in size, to aresolution of 1 ha and used FRAGSTATS to compute a suite oflandscape pattern metrics for each map. We then used multivariate dimensionreduction techniques and canonical correlation analysis to model therelationship between landscape pattern metrics and simpler stand table metricsthat are easily obtained from non-spatial forest inventories. These analyseswere performed on four habitat types common in boreal mixedwood forests: youngdeciduous, old deciduous, white spruce, and mixedwood types. Using only threelandscape variables obtained directly from stand attribute tables (totalhabitatarea, and the mean and standard deviation of habitat patch size), ourstatistical models explained more than 73% of the joint variation in fivelandscape pattern metrics (representing patch shape, forest interior habitat,and patch isolation). By PCA, these five indices captured much of the totalvariability in the rich set of landscape pattern metrics that FRAGSTATS cangenerate. The predictor variables and strengths of association were highlyconsistent across habitat classes. We illustrate the potential use of suchstatistical relationships by simulating the regional, cumulative effects ofwildfire and forest management on the spatial arrangement of forest patches,using non-spatial stand attribute tables.

ACCOUNTING FOR SYSTEM DYNAMICS IN RESERVE DESIGN

Systematic conservation plans have only recently considered the dynamic nature of ecosystems. Methods have been developed to incorporate climate change, population dynamics, and uncertainty in reserve design, but few studies have examined how to account for natural disturbance. Considering natural disturbance in reserve design may be especially important for the worlds remaining intact areas, which still experience active natural disturbance regimes. We developed a spatially explicit, dynamic simulation model, CONSERV, which simulates patch dynamics and fire, and used it to evaluate the efficacy of hypothetical reserve networks in northern Canada. We designed six networks based on conventional reserve design methods, with different conservation targets for woodland caribou habitat, high-quality wetlands, vegetation, water bodies, and relative connectedness. We input the six reserve networks into CONSERV and tracked the ability of each to maintain initial conservation targets through time under an active natural disturbance regime. None of the reserve networks maintained all initial targets, and some over-represented certain features, suggesting that both effectiveness and efficiency of reserve design could be improved through use of spatially explicit dynamic simulation during the planning process. Spatial simulation models of landscape dynamics are commonly used in natural resource management, but we provide the first illustration of their potential use for reserve design. Spatial simulation models could be used iteratively to evaluate competing reserve designs and select targets that have a higher likelihood of being maintained through time. Such models could be combined with dynamic planning techniques to develop a general theory for reserve design in an uncertain world.

Calibrating indices of avian density from non-standardized survey data: making the most of a messy situation

Summary The analysis of large heterogeneous data sets of avian point-count surveys compiled across studies is hindered by a lack of analytical approaches that can deal with detectability and variation in survey protocols. We reformulated removal models of avian singing rates and distance sampling models of the effective detection radius (EDR) to control for the effects of survey protocol and temporal and environmental covariates on detection probabilities. We estimated singing rates and EDR for 75 boreal forest songbird species and found that survey protocol, especially point-count radius, explained most of the variation in detectability. However, environmental and temporal covariates (date, time, vegetation) affected singing rates and EDR for 73% and 59% of species, respectively. Unadjusted survey counts increased by an average of 201% from a 5-min, 50-m radius survey to a 10-min, 100-m radius survey (n = 75 species). This variability was decreased to 8·5% using detection probabilities estimated from a combination of removal and distance sampling models. Our modelling approach reduced computation when fitting complex models to large data sets and can be used with a wide range of statistical techniques for inference and prediction of avian densities.

Effects of biotic feedback and harvest management on boreal forest fire activity under climate change

Predictions of future fire activity over Canadas boreal forests have primarily been generated from climate data following assumptions that direct effects of weather will stand alone in contributing to changes in burning. However, this assumption needs explicit testing. First, areas recently burned can be less likely to burn again in the near term, and this endogenous regulation suggests the potential for self-limiting, negative biotic feedback to regional climate-driven increases in fire. Second, forest harvest is ongoing, and resulting changes in vegetation structure have been shown to affect fire activity. Consequently, we tested the assumption that fire activity will be driven by changes in fire weather without regulation by biotic feedback or regional harvest-driven changes in vegetation structure in the mixedwood boreal forest of Alberta, Canada, using a simulation experiment that includes the interaction of fire, stand dynamics, climate change, and clear cut harvest management. We found that climate change projected with fire weather indices calculated from the Canadian Regional Climate Model increased fire activity, as expected, and our simulations established evidence that the magnitude of regional increase in fire was sufficient to generate negative feedback to subsequent fire activity. We illustrate a 39% (1.39-fold) increase in fire initiation and 47% (1.47-fold) increase in area burned when climate and stand dynamics were included in simulations, yet 48% (1.48-fold) and 61% (1.61-fold) increases, respectively, when climate was considered alone. Thus, although biotic feedbacks reduced burned area estimates in important ways, they were secondary to the direct effect of climate on fire. We then show that ongoing harvest management in this region changed landscape composition in a way that led to reduced fire activity, even in the context of climate change. Although forest harvesting resulted in decreased regional fire activity when compared to unharvested conditions, forest composition and age structure was shifted substantially, illustrating a trade-off between management goals to minimize fire and conservation goals to emulate natural disturbance.

Land cover, more than monthly fire weather, drives fire-size distribution in Southern Québec forests: Implications for fire risk management

[This corrects the article DOI: 10.1371/journal.pone.0179294.].

Modeling distribution and abundance of multiple species: Different pooling strategies produce similar results

Quantifications of spatial distribution and abundance of animals are essential to identifying key landscape characteristics and targeting locations for conservation action. Since conservation decisions often focus on multiple species aggregated in groups, e.g., guild-level, rather than individual species, predictions of species group abundance are of central importance. However, areas chosen for conservation action may differ if results from various modeling strategies also differ. Therefore, we compared three different strategies for modeling species group distribution and abundance: predict first, assemble later (PA); assemble first, predict later (AP); and the combined assemble, predict, then assemble (APA). All strategies were performed using Boosted Regression Trees (BRTs), which were fit to individual species data and then grouped after modeling, or fit to datasets that were grouped before modeling. Modeling strategies produced very similar results in terms of statistical performance assessed throug...

Modeling configuration dynamics of harvested forest landscapes in the Canadian boreal plains

Habitat configuration has important implications for the persistence of faunal and floral populations at a variety of spatial scales. Forest harvesting alters habitat configurations. However, measuring and predicting such alterations remains challenging, in part because previously developed metrics of habitat configuration are often not statistically independent of habitat amount. Thus, their ability to measure independent effects of habitat configurations and habitat amount on ecosystem components such as wildlife populations has been limited. Here, we evaluate habitat configuration based on newly developed metrics that are independent of habitat amount but do not depend on regression residuals of abundance and configuration relationships on any population of landscapes. We use these new metrics to measure and predict changes in habitat configuration following forest harvesting in the boreal forest of Alberta, Canada. Our findings clearly demonstrate changes in habitat configuration resulting from forest harvesting can be predicted precisely with information about initial habitat patch structure and harvesting patterns. Because forest harvesting has significant implications for habitat configuration, accurately predicting these changes is critical for determining if forest harvesting strategies are sustainable for ecosystem components and processes. This study provides a set of novel, robust metrics for tracking landscape-scale changes in habitat configuration in harvested boreal forests.

Configuration dynamics of boreal forest landscapes under recent fire and harvesting regimes in western Canada

Harvesting and forest fire change the spatial configurations of forest habitat. We used multivariate statistical models to evaluate the individual and cumulative effects of these two disturbances on habitat configuration in managed boreal forest landscapes in western Canada. We evaluated three aspects of configuration (core area, inter-patch distance and shape) using indices normalized for total habitat abundance. The two disturbances types had different effects on the three configuration metrics in terms of both the magnitude and direction of change. We found that the magnitudes of harvesting effects were larger than for fire. The direction of change was the same for core area and shape, but opposite for inter-patch distance which decreased slightly after fire. The combined effects of the two disturbances are distinct from the effects of either disturbance alone, and the effects are not always additive or compensatory for all metrics. Pre-treatment configuration was a significant covariate in all models, and total habitat abundance was significant in 4/9 models, but these were often not the most important covariates. In the cumulative disturbance model, covariates for the number or size of cut-blocks were significant.

Exploiting Poisson additivity to predict fire frequency from maps of fire weather and land cover in boreal forests of Québec, Canada

&NA; Predictive models of fire frequency conditional on weather and land cover are essential to assess how future cover‐type distributions and weather conditions may influence fire regimes. We modelled the effects of bottom‐up variables (e.g. land cover) and top‐down variables (e.g. fire weather) simultaneously with data aggregated or interpolated to spatial and temporal units of 100 km2 and 1 yr in the boreal forest of Québec, Canada. For models of human‐caused fires, we used road density as a surrogate for human access and behaviour. We exploited the additive property of Poisson distributions to estimate cover‐type specific fire count rates, which would normally not be possible with data of this spatial resolution. We used piecewise linear functions to model nonlinear relations between fire weather and fire frequency for each cover‐type simultaneously. The estimated conditional rates may be considered as expected mean counts per unit area and time. It follows that these rates can be rescaled to arbitrary spatial and temporal extents. Our results showed fire frequency increased nonlinearly as aridity increased and more quickly in disturbed areas than other types. Road density exerted the strongest influence on the frequency of human‐caused fires, which were positively correlated with road density. The estimates may be used to parameterize the fire ignition component of spatial simulation models, which often have a resolution different from that at which the data were collected. This is an essential step in incorporating biotic and abiotic feedbacks, land‐cover dynamics, and climate projections into ecological forecasting. The insight into the power of Poisson additivity to reveal high‐resolution ecological processes from low‐resolution data could have applications in other areas of ecology.

Boreal bird abundance estimates within different energy sector disturbances vary with point count radius

ABSTRACT Responses of boreal birds to changes in forest structure and composition caused by construction of well pads, seismic lines, and pipelines are poorly understood. Bird species associated with older forests are predicted to experience larger population declines with increased disturbance compared with species associated with younger or open habitats; however, point count methods may influence apparent outcomes because the proportional area of disturbed vegetation and the magnitude, uncertainty, and detection of a disturbance response by birds vary as a function of sampling area. We analyzed point count data from 12 energy sector studies and measured how disturbance type and point count radius interacted to affect 531 impact ratios (mean abundance at point counts centered within disturbances relative to abundance at point counts within forest 150–400 m from the nearest edge bordering those disturbances [59 species*3 disturbance types*3 point count radii]). We observed larger disturbance effects (impact ratios) within larger-radius point counts at well pads (100-m and unlimited-distance) and pipelines (unlimited-distance) compared with 50-m point counts at seismic lines, and within 50-m point counts at well pads relative to 50-m point counts at seismic lines. Effect uncertainty was higher at well pads and pipelines than seismic lines, and lower within larger-radius point counts. The probability of detecting a disturbance response was greater for larger-radius point counts at pipelines than for 50-m point counts at seismic lines, and within 50-m point counts at well pads relative to 50-m point counts at seismic lines. On average, a species was more likely to increase in abundance near an energy sector disturbance if the species was not associated with older (>75 yr) forest stages. While the effects of disturbance varied by species and with disturbance type, the effects of pipelines and seismic lines were better detected by larger-radius point counts, while the effects of well pads were better detected by smaller-radius point counts.