Background/Question/Methods:  
Mechanisms for resilience must exist for ecosystems to persist across space and time, often in the face of substantial disturbance, climatic, and environmental fluctuation.  Furthermore, resilience mechanisms must occur at spatio-temporal scales that can reinforce the patterns, processes, and interactions that operate in ecosystems.  Here, we examine the power law behavior of multi-scale vegetation features, topographies, and fire severity patches of eastern Cascade Mountain forest landscapes.  Our objectives were to: (1) use maximum likelihood estimation (MLE) techniques to fit statistical distributions to vegetation, topography, and fire severity patch-size distributions (PSDs) of pre-settlement era landscapes of four ecoregions, (2) evaluate concordance between vegetation, topography, and fire severity PSDs, (3) evaluate scale-invariance and power-law behavior of the best fitting distributions using MLE, broken-stick regression, left truncation, and neutral-modeling techniques, and (4) evaluate the quantitative evidence for exogenous  and endogenous controls on the distribution of vegetation (physiognomies, land cover types, structural classes, canopy cover classes) and fire severity patch sizes.
Results/Conclusions:   From physiognomies to canopy cover classes of all ecoregions, heavy-tailed PSDs were evident. Out of 25 statistical distributions tested, the Pareto and Generalized Beta II (GBII) distributions consistently fit the empirical inverse cumulative distribution functions (CDFs) of vegetation, topography, and fire severity patches. Top-down controls--When PSDs of any vegetation,  topography, or fire severity class were pooled across ecoregions, all tested statistical distributions failed to fit the resultant empirical CDFs.  However, K-S boot strap and log-likelihood ratio tests revealed that models fit to individual subregions provided good fits.  Bottom-up controls--The frequency-size distribution of N and S aspect patches followed a nearly pure power-law distribution for all ecoregions but one. Other topographic features such as slope, curvature, slope*aspect, slope*curvature also followed a power-law relation for some ecoregions, but inconsistencies were pronounced. Neutral models developed for vegetation, topography, and fire severity classes showed conclusive evidence that PSDs were not the result of random influences.  We conclude that (1) vegetation and aspect PSDs partially drove fire severity PSDs via mosaics of structural and compositional “fences and corridors” that either resisted or allowed penetration of otherwise contagious fires; (2) topography partially drove vegetation and fire severity PSDs of all ecoregions; (3) ecoregions partially drove vegetation and fire severity PSDs from above; and (4) fine-, meso, and broad-scale process domains may be quantified for vegetation, topography, and fire severity PSDs. This information should be useful to creating more resilient landscapes in an uncertain climatic future.