A signal-to-noise index to quantify the potential for peak detection in sediment–charcoal records

a r t i c l e i n f o Keywords: Signal-to-noise index (SNI) Charcoal analysis Fire history Lake sediment Paleoecology Charcoal peaks in lake-sediment records are commonly used to reconstruct fire histories spanning thousands of years, but quantitative methods for evaluating the suitability of records for peak detection are largely lacking. We present a signal-to-noise index (SNI) that quantifies the separation of charcoal peaks (signal) from other variability in a record (noise). We validate the SNI with simulated and empirical charcoal records and show that an SNI N 3 consistently identifies records appropriate for peak detection. The SNI thus offers a means to evaluate the suitability of sediment–charcoal records for reconstructing local fires. MATLAB and R functions for calculating SNI are provided. Introduction Numerous fire-history studies have been published on the basis of the identification of peaks in lake-sediment records of macroscopic (typically N100 μm in diameter) charcoal (e.g., see reviews by Whitlock and Larsen (2001) and Gavin et al. (2007)). Analytical methods are typically based on a decomposition approach by which a time series of charcoal accumulation rates (CHAR) is detrended to isolate background and peak series (e. The background series represents long-term shifts in fire regimes (e.g., area burned, fuel characteristics) and taphonomic processes unrelated to fire occurrence (e.g., slope wash, sediment focusing), and it may be modeled with a variety of moving-average methods. Background CHAR is removed from the raw CHAR series by either subtraction or division to obtain a peak series, which is assumed to represent charcoal from local fires, plus random variability (i.e., noise). To separate the signal of local fires from noise, a threshold function is determined by one of several methods, and samples exceeding the threshold are interpreted as local fire episodes. support suggesting that in records from small lakes (b ~20 ha), identified peaks represent fires within ca. 500–1000 m of the sampling location. However, sediment–charcoal records are highly variable, both within and between sites (Power et al., 2008), and some records are clearly more appropriate for peak analysis than others. While a record with large peaks distinct from background values fits well within the conceptual framework outlined above, processes including sediment mixing or reduced sedimentation rates can create more ambiguous records (Higuera et al., 2007). Despite broad adoption of the decomposition approach, methods for quantifying the suitability of a charcoal record for peak analysis were lacking prior to the …