A grain-size distribution report lands on your desk with a column of percentile values, a curve, and a textural class name like "moderately sorted fine sand." If you trained in structural or geotechnical work, that line of descriptors can feel like a formality rather than a tool. It is not. The percentile values encode the engineering behavior of a soil or sediment directly: permeability, compactability, susceptibility to piping, suitability for filtration media, and regulatory classification under ASTM D422 or the equivalent ISO 17892-4 procedure all flow from the same three numbers. This guide walks through D10, D50, and D90 with the precision they deserve, then explains the Folk and Ward (1957) statistical framework so the textural classification at the bottom of the report becomes a working reference, not boilerplate.

What the grain-size curve is actually recording

A grain-size distribution curve plots cumulative percent finer on the vertical axis against grain diameter (in millimetres, or phi units) on the horizontal axis. The curve is constructed from two combined procedures: mechanical sieving for the coarser fraction and hydrometer or laser diffraction analysis for the finer fraction below 0.075 mm. The point where these two datasets join is called the split point, and a discontinuity there often signals a laboratory error rather than a real feature of the sediment. When you receive a report, check the curve for a kink at that junction before reading any derived statistics. The shape of the full curve tells you more than any single number: a near-vertical segment means a narrowly graded material; a gently sloping, near-horizontal curve means the sample contains a wide spread of grain sizes. Engineering behavior changes significantly between the two, even when the median diameter is identical.

D10, D50, and D90: reading the percentiles correctly

D10 is the grain diameter at which 10 percent of the sample by mass is finer. D50 is the median diameter. D90 is the diameter below which 90 percent of the sample falls. These are read directly off the cumulative curve by drawing a horizontal line at the relevant percentage and dropping a vertical to the diameter axis. In practice, laboratories interpolate between sieve sizes using a log-linear function, so the reported values are not raw measurements but fitted estimates. D10 drives hydraulic conductivity calculations: Hazen's approximation sets k (cm/s) at roughly 0.01 times D10 squared when D10 is between 0.1 mm and 3 mm and the coefficient of uniformity is below 5. D50 governs the median transport threshold in sediment mobility assessments and is the primary descriptor in most soil classification schemes. D90 appears in filter design, where the U.S. Bureau of Reclamation's filter criteria and the equivalent criteria used in Australian dam engineering both use the ratio of the filter D15 to the base material D85 or D90. Treating these three numbers as independent facts rather than related points on a single curve is the most common interpretive mistake.

Coefficient of uniformity and coefficient of curvature

Two derived ratios appear on most reports alongside the percentile values. The coefficient of uniformity, Cu, equals D60 divided by D10. A value below 4 indicates a uniform or poorly graded material; above 6 for gravels and above 4 for sands, ASTM D2487 classifies the material as well-graded, which carries implications for compaction and bearing capacity. The coefficient of curvature, Cc, equals the square of D30 divided by the product of D10 and D60. A Cc between 1 and 3, combined with a qualifying Cu, puts gravels into the GW class and sands into the SW class under the Unified Soil Classification System. Both Cc and Cu can appear acceptable in a bimodal sample that is actually gap-graded, meaning a middle fraction is missing entirely. The curve will show a flat shelf at some intermediate diameter. This condition increases internal erosion risk in embankments and should prompt additional scrutiny regardless of what the derived ratios suggest.

The Folk and Ward framework: why phi units matter

Robert Folk and William Ward published their graphical statistical measures in 1957 as a way to characterize sediment texture using a small number of percentile readings taken directly from a cumulative curve plotted in phi units. Phi is defined as the negative base-2 logarithm of diameter in millimetres. Fine sediments have positive phi values; gravels have negative ones. Converting to phi is necessary because grain-size distributions are log-normal, and arithmetic statistics applied to millimetre values produce meaningless results. The Folk-Ward measures use six percentile values: phi5, phi16, phi25, phi50, phi75, phi84, and phi95. The graphic mean is calculated as the average of phi16, phi50, and phi84. Inclusive graphic standard deviation, which Folk and Ward used as the sorting index, combines these six values into a single formula. Skewness and kurtosis are also defined graphically. Most laboratory software computes all four statistics automatically, but understanding what each measures prevents misreading the output.

Sorting classes and what they tell you about depositional history

Inclusive graphic standard deviation produces six sorting classes. Values below 0.35 phi indicate very well sorted sediment, typical of aeolian dune sands or beach foreshore deposits. Between 0.35 and 0.50 phi is well sorted; 0.50 to 0.71 phi is moderately well sorted; 0.71 to 1.00 phi is moderately sorted; 1.00 to 2.00 phi is poorly sorted; and above 2.00 phi is very poorly sorted. Alluvial fan deposits in active tectonic settings routinely return values above 2.00 phi. River channel sands in low-gradient systems typically fall between 0.50 and 1.00 phi. Sorting is not merely a descriptive label: poorly sorted materials have lower porosity and permeability than their median grain size alone would predict, because finer particles fill the voids between coarser grains. This directly affects remediation timeframes in contaminated-site work and drainage layer performance in civil infrastructure. When a report from a coastal reclamation project returns a sorting value of 1.3 phi for a supposed beach-compatible fill, that material will behave differently from the native shore sediment even if the D50 values match.

Skewness and kurtosis: the overlooked statistics

Folk-Ward graphic skewness measures the asymmetry of the distribution. Positive skewness means an excess of fine material relative to the median, common in fluvial overbank deposits where the energy wanes and fines settle out. Negative skewness indicates an excess of coarse material, typical of lag gravels or residual soils where fines have been winnowed. The threshold values are: strongly fine-skewed above 0.30, fine-skewed from 0.10 to 0.30, near-symmetrical from minus 0.10 to 0.10, coarse-skewed from minus 0.30 to minus 0.10, and strongly coarse-skewed below minus 0.30. Kurtosis compares the spread of the tails to the central portion of the distribution. A leptokurtic result, above 1.00, means the central fraction is very well sorted but the tails are wide, often produced by mixing of two separate sediment populations. Platykurtic distributions, below 0.90, suggest uniform sorting across the full range. Environmental consultants working on sediment provenance studies or contaminant pathway analysis use skewness and kurtosis to distinguish reworked from in-situ deposits, a distinction that affects where monitoring wells are placed and how fate-and-transport models are calibrated.

Applying the report: three practical checkpoints

Before signing off on a grain-size report, three checks catch most interpretive errors. First, confirm that the textural class name on the report is consistent with the Folk-Ward statistics computed in the body of the document. Automated naming routines occasionally misfire when samples sit near class boundaries, and the narrative description in a report may not be updated after laboratory reanalysis. Second, verify that D10, D50, and D90 are internally consistent with the shape of the curve provided. If D90 is less than twice D50 but the curve looks broadly graded, one of the two is wrong. Third, for any sample used in filter or drainage design, compare the full grading envelope, not just the percentile points, against the design specification band. A sample can meet D15 and D85 criteria while having an interior bulge or gap that falls outside the specification window. These three steps take less than ten minutes and routinely surface errors before they propagate into design calculations.

Grain-size reports are produced quickly and read quickly, which is where interpretation errors enter a project. The Folk and Ward statistics date from 1957 but remain the standard because they extract the maximum information from the minimum number of data points on a cumulative curve. Understanding what each statistic is geometrically measuring, rather than accepting the class label at face value, is the difference between using the report and simply filing it.