Minimum count sums for charcoal concentration estimates in pollen slides: accuracy and potential errors

• Published in: Holocene (Sevenoaks) Vol. 15; no. 2; pp. 293 - 297
• Main Authors: Finsinger, Walter, Tinner, Willy
• Format: Journal Article
• Language: English
• Published: Thousand Oaks, CA SAGE Publications 01.02.2005; Turpin; London: Sage
• Subjects: Agricultural sciences; Biodiversity and Ecology; Botanics; Continental interfaces, environment; Environment and Society; Environmental Sciences; Environmental studies; Excavation and methods; Fauna; Flora; Global Changes; Interpretation, statistical and computer analysis; Laboratory methods; Life Sciences; Methodology and general studies; Prehistory and protohistory; Quantitative Methods; Sciences of the Universe; Silviculture, forestry; Vegetal Biology; New; Middle; error estimation; random sampling; Charcoal analysis; sample size; palaeoecology; Sample; Statistics; Decoration; Processing; Numeration; Palaeoecology; Simulation; Comparison; Analysis; Fire; Sampling; Charcoal
• ISSN: 0959-6836; 1477-0911
• DOI: 10.1191/0959683605hl808rr

Charcoal particles in pollen slides are often abundant, and thus analysts are faced with the problem of setting the minimum counting sum as small as possible in order to save time. We analysed the reliability of charcoal-concentration estimates based on different counting sums, using simulated low-to high-count samples. Bootstrap simulations indicate that the variability of inferred charcoal concentrations increases progressively with decreasing sums. Below 200 items (i.e., the sum of charcoal particles and exotic marker grains), reconstructed fire incidence is either too high or too low. Statistical comparisons show that the means of bootstrap simulations stabilize after 200 counts. Moreover, a count of 200-300 items is sufficient to produce a charcoal-concentration estimate with less than+5% error if compared with high-count samples of 1000 items for charcoal/marker grain ratios 0.1-0.91. If, however, this ratio is extremely high or low (> 0.91 or < 0.1) and if such samples are frequent, we suggest that marker grains are reduced or added prior to new sample processing.