The chores of sample preparation were divided equally between labs at the University of Missouri and University of California, Santa Cruz. Because of subtle differences in procedure, several samples were split and prepared in both labs to test for reproducibility (Table 1). The first step in sample preparation was to remove pore water by freeze drying. Dried samples (typically 10-20 g) were stored in a desiccator to prevent moisture from being absorbed, and the dry weights were recorded. The samples then were transferred to 600-mL beakers and immersed in hydrogen peroxide to digest organic matter. After at least 24 hr of digestion and periodic stirring, 250 mL of sodium hexametaphosphate (Calgon) solution (4 g per 1000 mL deionized water) were added to each beaker to assist disaggregation and prevent clay flocculation. After sitting in Calgon solution for at least 12 hr, the beakers were immersed in an ultrasonic bath for 5-10 min to enhance disaggregation further. Suspensions were washed through a 63-µm screen to separate sand-sized grains from silt and clay. Each sand portion was collected, dried in an oven, and weighed. Each fraction <63 µm was collected in a large evaporating dish and transferred to a 500-mL Nalgene bottle. A centrifuge was used to reduce water volumes by roughly one-half (at 8000 rpm for 20 min), and sediment was washed out of the centrifuge tubes with Calgon solution. The concentrated suspensions were stored in 125-mL Nalgene bottles until analysis.
The SediGraph 5000ET grain-size analyzer measures the attenuation of X-rays by particles that are suspended in a solution (Jones et al., 1988). Comparisons among the results of SediGraph analysis and results using other instruments and techniques have been described by Stein (1985), Singer et al. (1988), and Camerlenghi et al. (1995). The SediGraph determines the concentration of particles remaining at decreasing depths within a suspension as a function of time. The principle of Stoke's Law of Settling is used to convert vertical profiles of suspension density to weight percentages of grain size. Before analysis, sample bottles were shaken vigorously for several minutes to resuspend and disaggregate the sediment particles. Approximately 60 mL of suspension were poured into the SediGraph chamber. In some cases, the concentrations had to be adjusted to fall within an acceptable range of kilocounts/s. Rigorous characterization of size fractions less than 0.5 µm requires settling times that are prohibitively long; in addition, absolute size data in the submicron range from the SediGraph are of questionable reliability (Singer et al., 1988). In most cases, measurements to 0.5 µm took ~10 min and allowed us to characterize the size distribution to between 30% and 40% cumulative mass finer.
Data output from SediGraph software includes a table and cumulative curve of mass percentages finer over a range of sizes from 63 to 0.5 µm, plus values of the median diameter (d50) and cumulative mass finer than 4 µm (clay fraction). Weight percentages of sand were calculated by dividing the weight of the sediment coarser than 63 µm (from sieving) by the total dry weight that was measured after freeze drying. We regarded the total dry weight minus the sand weight as equal to the weight of silt + clay, even though this difference includes an error equal to the weight of digested organic matter. The weight percentages of silt and clay fractions were calculated by multiplying the weight of silt + clay by the cumulative percents >4 µm and <4 µm, respectively.
Digital data were downloaded to a graphics application and replotted to extrapolate the cumulative curve (by linear extension) beyond the 75th percentile and to digitize the diameter values (in micrometers) at the 25th and 75th percentiles (d25 and d75). These quartile measures were used to calculate three statistics for the silt + clay fraction: mean (Me), geometric sorting coefficient (So), and geometric skewness (Skg), following Krumbein (1936). The relevant equations are
A symmetrical distribution of particle sizes results in a geometric skewness equal to unity. Sorting coefficients less than 2.5 are indicative of well-sorted samples; coefficients greater than 4.5 are indicative of poor sorting. We emphasize here that these statistics pertain only to the size fraction analyzed by SediGraph (<63 µm), not the total grain-size distribution.