Field test of an autocorrelation technique for determining grain size using a digital camera

An extensive field test using Rubin's (2004) autocorrelation technique shows that median and mean grain size can be determined with suitable accuracy using a digital camera and associated autocorrelation when compared to traditional methods such as mechanical sieving and settling-tube analysis. The field test included 205 sediment samples and > 1200 digital images from a variety of beaches on the west coast of the United States, with grain sizes ranging from sand to granules. To test the accuracy of the digital-image grain-size algorithm, we compared results with manual point counts of a large image data set in the Santa Barbara littoral cell. Grain sizes calculated using the autocorrelation algorithm were highly correlated with the point counts of the same images (r2=0.93; n=79) and had an error of only 1%. Although grain sizes calculated from digital images give an accurate result for grains in the image, natural lateral and vertical variability in grain size can cause differences between grain size measured in digital images of the bed surface and grain size measured by sieving a grab sample that includes subsurface sediment. Lateral spatial variability was tested by analyzing the results of up to 100 images taken in a series of 1 m2 sample areas. Comparisons of calculated grain sizes and grain sizes measured from grab samples show small differences between surface sediment and grab samples on high- energy dissipative beaches with well-sorted sediment such as in the Pacific Northwest (r2 > 0.92; n=115). In contrast, on less dissipative, more poorly sorted beaches such as Ocean Beach in San Francisco, differences between surface and subsurface grain size are greater (r2 > 0.70; n=67; within 3% accuracy). In all field tests the autocorrelation method was able to predict the mean and median grain size with ~96% accuracy, which is more than adequate for the majority of sedimentological applications. When properly automated for large numbers of samples, the autocorrelation technique is roughly 2 orders of magnitude faster than traditional grain- size analysis, saving time and money, without sacrificing accuracy in measuring average grain size; however, the technique may not be as accurate in measuring the fine and coarse tails of a size distribution.