Improving the efficiency of sediment charcoal image analysis

Abstract

Imaging of charcoal particles extracted from lake sediments provides an important way to understand past fire regimes. Imaging of large numbers of particles can be time consuming. In this note we explore the effects of subsampling and extrapolation of area on estimates of sum charcoal area, using resampling of real and simulated data sets and propose a protocol in which all particles are counted with only the first 100 encountered being imaged. Extrapolated estimates of sum total area of charcoal for 40 real samples were nearly identical to actual values, and error introduced due to subsampling was low (Coefficient of variation <0.2) for all but samples originally containing fewer than 50 particles. Similarly, error was low for simulated data (CV <0.02). Extrapolation provided better estimates of charcoal area than did a regression-based approach. Our results suggest that imaging a fixed number of pieces of charcoal (n\u2009=\u2009100) and counting any additional pieces represents a time efficient way to estimate charcoal area while at the same time retaining useful information on particle size and shape