Effect of grain packing tightness on strain estimation from the Fry method

How tightly should a sample be packed for strain estimation by the Fry method? We address this issue using synthetic simulations of 900 images such that each image contains 200 randomly distributed grains, but differs from other images with respect to the packing tightness and the degree of sorting. Each image is coaxially distorted by various known strain ratios and the strain estimates from distorted images are obtained by the Fry method. The statistical errors in the strain estimates are found to grow larger with the decrease in the packing tightness irrespective of the level of distortion. We demonstrate that a progressive decrease in the packing tightness results in an increasingly clustered nature of grain center distribution and hence the larger errors. These results, obtained from the synthetic images, are corroborated by two natural examples of sandstone, one loosely packed and the other tightly packed. Based on the results of tests on synthetic and natural examples, we recommend that the Fry method should be used only on those samples that have ({>}30%) packing density.