Search

Abstract: Studies of forest biomass dynamics typically use long-term forest inventory data, available in only a few places around the world. We present a method that uses photogrammetric measurements from [...]

Wildfire ignition distribution models are powerful tools for predicting the probability of ignitions across broad areas, and identifying their drivers. Several approaches have been used for ignition-distribution modelling, yet the performance of different model types has not been compared. This is unfortunate, given that conceptually similar species-distribution models exhibit pronounced differences among model types. Therefore, our goal was to compare the predictive performance, variable importance and the spatial patterns of predicted ignition-probabilities of three ignition-distribution model types: one parametric, statistical model (Generalised Linear Models, GLM) and two machine-learning algorithms (Random Forests and Maximum Entropy, Maxent). We parameterised the models using 16 years of ignitions data and environmental data for the Huron-Manistee National Forest in Michigan, USA. Random Forests and Maxent had slightly better prediction accuracies than did GLM, but model fit was similar for all three. Variables related to human population and development were the best predictors of wildfire ignition locations in all models (although variable rankings differed slightly), along with elevation. However, despite similar model performance and variables, the map of ignition probabilities generated by Maxent was markedly different from those of the two other models. We thus suggest that when accurate predictions are desired, the outcomes of different model types should be compared, or alternatively combined, to produce ensemble predictions. [ABSTRACT FROM AUTHOR]

The majority of landscape pattern studies are based on the patch-mosaic paradigm, in which habitat patches are the basic unit of the analysis. While many patch-based landscape indices successfully relate spatial patterns to ecological processes, it is also desirable to use finer grained analyses for understanding species presence, abundance, and movement patterns across the landscape and to describe spatial context by measuring habitat change across scales. Here, we introduce two multi-scale pixel-based approaches for spatial pattern analysis, which quantify the spatial context of each pixel in the landscape. Both approaches summarize the proportion of habitat at increasing window sizes around each pixel in a scalogram. In the first regressionbased approach, a third-order polynomial is fitted to the scalogram of each pixel, and the four polynomial coefficients are used as descriptors of spatial context of each pixel within the landscape mosaic. In the second shape-based approach, the scalogram mean and standard deviation, and the mean slope between forest cover at the smallest window size and each of the larger window sizes are calculated. The values emerging from these two approaches are assigned to each focal pixel and can be used as predictive variables, for example, in species presence and abundance studies. We tested the performance of these approaches on 18 random landscapes and nine actual landscapes with varying forest habitat cover. Results show that both methods were able to differentiate between several spatial contexts. We thus suggest that these approaches could serve as a complement or an alternative to existing methods for landscape pattern analysis and possibly add further insight into pattern-species relations. [ABSTRACT FROM AUTHOR]

Disturbances such as grazing, invading species, and clear-cutting, often act at small spatial scales, and means for quantifying their impact on fine scale vegetation patterns are generally lacking. Here we adopt a set of landscape metrics, commonly used for quantifying coarse scale fragmentation, to quantify fine scale fragmentation, namely the fine scale vegetation structure. At this scale, patches often consist of individual plants smaller than 1 m2, requiring the grain of the analysis to be much smaller. We used balloon aerial photographs to map fine details of Mediterranean vegetation (pixel size <0.04 m) in experimental plots subjected to grazing and clear-cutting and in undisturbed plots. Landscape metrics are sensitive to scale. Therefore, we aggregated the vegetation map into four coarser scales, up to a resolution of 1 m, and analyzed the effect of scale on the metrics and their ability to distinguish between different disturbances. At the finest scale, six of the seven landscape metrics we evaluated revealed significant differences between treated and undisturbed plots. Four metrics revealed differences between grazed and control plots, and six metrics revealed differences between cleared and control plots. The majority of metrics exhibited scaling relations. Aggregation had mixed effects on the differences between metric values for different disturbances. The control plots were the most sensitive to scale, followed by grazing and clearing. We conclude that landscape metrics are useful for quantifying the very fine scale impact of disturbance on woody vegetation, assuming that the analysis is based on sufficiently high spatial resolution data. [ABSTRACT FROM AUTHOR]