VIRTUAL SANDSTONE PETROGRAPHY: COMPARISON OF DIGITAL AND MANUAL POINT-COUNTING METHODS

When the COVID-19 pandemic forced the revision of the in-progress sedimentary petrography course at the University of Mississippi, the final project had to be redesigned using existing resources. The decision was made to assess the accuracy of free point-counting software in comparison to traditional petrographic microscope-based point counting. While such comparisons have been made previously, we focused on accuracy of the relative percentages of constituents (i.e., grains, matrix, cement, pore space) counted in thin-section samples. We used four thin-section samples from the Mississippian Hartselle Sandstone of Mississippi and Alabama that were made for a previous project. Point-count results determined with a petrographic microscope were published earlier this year. Students were given digital photomicrographs of samples under both plane-polarized light and cross-polarized light and performed point counts using Rock.AR2 and JmicroVision software. A Pearson’s chi-squared test was performed in R for the data collected from each thin-section sample to test the null hypothesis that point-counting method is independent of constituent counts. Acceptance of the null hypothesis would indicate that point counting yields the same results regardless of the method used. Rejection of the null hypothesis would indicate that point-count results are dependent on the method used. For each of the four samples, the null hypothesis was rejected (all p values < 0.05). While this seems to show that neither of the software packages can provide the accuracy of traditional point counting, we cannot rule out the possibility of errors in point counting using still images that do not allow for the same range of mineral identification methods possible with the petrographic microscope. We also note that a single photomicrograph does not necessarily provide a representative sample of the constituent population. The magnification and size of the grains in photomicrographs also influences whether a single grain is counted multiple times when randomized points are targeted.