A guide to screening charcoal peaks in macrocharcoal-area records for fire-episode reconstructions

Abstract

Macroscopic charcoal records can be used to infer spatially explicit reconstructions of past fire history. However, a current deficiency in the charcoal-analysis toolbox has been the lack of a method to consider sampling variability and charcoal-particle area distributions for peak detection with charcoal-area records. We present a screening procedure specific for datasets comprising charcoal numbers and areas to screen the charcoal-area estimates with respect to the count sums. The rationale for screening charcoal-area peaks stems from the observation that although charcoal-area records can be more suitable in a statistical sense for peak detection (e.g. as established by the signal-to-noise index), charcoal-area peaks can be questionable if they are determined by just one or a few larger charcoal particles. Our method begins with a charcoal-area time series analysed by existing methods to identify peaks representing fire episodes. To screen these peaks, the method uses bootstrap resampling of charcoal-particle areas observed in a user-defined subsection of the record around each peak to obtain the range of likely charcoal areas for different counts. Peaks with total area within the likely range of bootstrapped samples (e.g. p > 0.05) are flagged as potentially unreliable, whereas samples with total area significantly greater than expected by chance are deemed robust indicators of past fire events. In an example application of the method to a charcoal record from Lake Brazi, Romania, several peaks failed to pass the screening suggesting that, as for count-based records, unscreened charcoal-area records may include spurious fire episodes and thus potentially underestimate past fire-return intervals.