Spatial and temporal validation of fire-scar fire histories.

摘要

Accurate information about historical fire regimes is needed to understand the long-term effects of fire and climate on ecosystem dynamics and guide ecosystem restoration. Fire scars are used widely to reconstruct historical fire regimes around the world but few empirical validation studies have been conducted. This dissertation consists of three integrated studies aimed at addressing the following questions: (1) how accurate are fire-scar fire histories compared to known patterns of fire occurrence; (2) how do these relationships vary spatially and temporally; (3) how representative statistically are search-based ("targeted") fire-scar sampling techniques? I utilized an empirical corroboration approach to validate fire-scar reconstructions against documentary fire perimeters for a 2,780 hectare ponderosa pine landscape in Saguaro National Park, Arizona (USA). Resampling statistics and spatial modeling were used to quantify interactions between spatial scale, sample size, and fire size. Statistical properties of targeted sampling were assessed by analyzing three case studies containing paired examples of targeted and non-targeted sampling (i.e., systematic and census). I found strong linear relationships between fire-scar synchrony (samples scarred in a given year) and annual area burned. Fire-scar derived estimates of fire frequency metrics, such as Mean Fire Return Interval and Natural Fire Rotation, did not differ significantly from the documentary record, and there was strong spatial coherence between fire frequency maps interpolated from fire-scar data and documentary maps. Scale and sample size dependence of fire-scar detection probabilities were variable for small fire years but relatively weak for widespread fires. This resulted in consistent and predictable influences on fire frequency reconstructions: statistical measures dependent on area burned were relatively stable and robust across a range of scale, sample size, and fire size. Targeted sampling did not differ statistically from non-targeted datasets, but targeted fire-scar data contained proportionately greater sample depth and longer temporal records with fewer samples. These results demonstrate collectively that key temporal and spatial fire frequency parameters can be reconstructed accurately from point-based fire-scar data. They also reaffirm general interpretations and management implications from past fire history research indicating that frequent, widespread burning was an important component of pre-settlement fire regimes in Southwestern ponderosa pine.