Wildfire is a natural disturbance common to many forested ecosystems of North America. In the fall of 2011, a lightning-ignited wildfire, the Pagami Creek Fire (PCF), burned over 38,000 hectares of forest in northeast Minnesota, most of which occurred in the protected Boundary Waters Canoe Area Wilderness (BWCAW).

Satellite remote sensing data has been used within the upper Midwest to detect and quantify forest vegetation extent, composition, and associated structure. However, very few studies have quantified the composition and structure of early seral forest regeneration.

We were interested in two objectives:

1) Use a combination of remote sensing imagery (multi-temporal Landsat data and National Agriculture Imagery Program (NAIP) color-infrared aerial imagery) and derived variables to develop models estimating quaking aspen (Populus tremuloides Michx.) regeneration abundance and structure as well as vegetation species richness derived from field plots. These empirical models were used to spatially extrapolate field data to produce continuous spatially explicit biophysical estimates for the entire PCF burned area.

2) Examine hypothesized relationships between pre-fire aspen basal area (BA) and resulting fire severity, and to evaluate whether burn severity and pre-burn aspen BA may explain observed variability in aspen regeneration (size, abundance, and distribution) and vegetation species richness within the PCF burned area.

Remote sensing-based models for vegetation species richness (RMSE=2.47 species, Adj. R²=0.60) was the strongest observed, followed by average aspen stem diameter (RMSE=2.21 mm, Adj. R²=0.53). We also generated statistically significant (p < 0.01) empirical models for average aspen basal area, height, density, and percent cover.

We used simple linear regression to examine empirical relationships between field measurements of regenerating aspen abundance and vegetation species richness measurements as functions of satellite derived fire severity and pre-fire aspen BA estimates. Aside from one of our aspen measurements (percent foliar cover), there was a significant relationship (p < 0.01) and meaningful coefficient of determination with pre-fire aspen BA and post-fire aspen measurements (Adj. R² range = 0.33 - 0.12). There was not a significant relationship (p = 0.46) between pre-fire BA and vegetation species richness (Adj. R² = -0.01). There was not a significant relationship (p-value range = 0.14 - 0.57) between fire severity and all other vegetation recovery response variables.

We expect both forest managers and ecologists to benefit from the ability to detect, model, and quantify early-seral stage forest regeneration over large, often remote, areas. Such remote sensing methods may further facilitate understanding of landscape scale processes (pattern, distribution, and abundance) with respect to fire and other stand-replacing forest disturbances in this and similar forested environments.