TESTING ACCURACY AND PRECISION OF LOSS-ON-IGNITION ANALYSIS FOR SOILS AND PALEOSOLS

Loss on ignition (LOI) analysis is widely used to estimate volatile content in soils and paleosols. In this study, we sought to optimize precision and accuracy of LOI measurements in the James Madison University Terrestrial Paleoclimate Lab. First, we developed a new in-house soil standard from the uppermost Bt horizon of the Frederick Soil Series (a Typic Paleudult). Visible roots were hand-picked prior to grinding and homogenizing batches of bulk soil in a Shatterbox. We used the resulting powders to test an existing LOI methodology for precision and accuracy. Oven-dry (105-110°C for 2 hr) samples were roasted in sanitized porcelain crucibles in a Ney 525 muffle furnace. We systematically tested the effect of crucible position (Trial 1), sample mass (Trial 2), and crucible type (Trial 3) on measured LOI, while maintaining a constant temperature of 950°C. In Trial 1, we hypothesized that samples located closest to the furnace door would have lower LOI values than samples in the back of the furnace. We tested crucible position by arranging samples in a 3 x 3 grid and roasting samples for 90 minutes. Variance of LOI was high, driven by two samples closest to the door with lower LOI values. Trial 1 was repeated for 105 minutes, which equalized LOI across samples. In Trial 2, we hypothesized that sample mass would be inversely correlated with LOI due to incomplete roasting of large samples. We tested the effect of sample mass by roasting a batch of nine samples that varied from 0.09 to 24 g. We found that above a mass of 3 g, LOI was consistent and generally similar to expected values from Trial 1. However, LOI rapidly increased with decreasing mass below 3 g. This effect was reproduced with a replicate trial; thus we hypothesized that high LOI values in small samples (< 3 g) result from mass measurement uncertainty induced by static electricity buildup on the soil surface while using the analytical balance. We further tested for this effect in Trial 3, where we repeated Trial 2 with taller crucibles that have a smaller cross-sectional area. Here, we found the increase of LOI with smaller mass occurred at slightly lower threshold (~ 1 g), suggesting that static-related uncertainty is indeed related to sample surface area. In summary, LOI measurements in our lab should be performed on oven-dry samples > 3 g at 950 °C for 105 minutes.