A TEST OF SEMI-AUTOMATED VARVE-COUNTING SOFTWARE USING DRILL CORE PHOTOGRAPHS, PLEISTOCENE LAKE TIGHT; TEAYS VALLEY, WEST VIRGINIA

Count-My-Varves is open-source, image recognition software for semi-automated counting of lacustrine varves. The program was developed and first used with digital line scans of drill cores of Holocene lake deposits and was suggested to be equally suitable for use with outcrop photos, and by extension, drill core photos. We assessed the potential of this program for speeding the varve counting process and providing accurate count results using photos of drill core.

Our evaluation used digital photos of archival drill core from deposits of Early Pleistocene (Calabrian) Lake Tight, an ice-dammed lake that occupied the Teays River Valley in West Virginia. Two series of photos were utilized, one from 20 years ago, taken shortly after initial collection and processing of the drill core, and the other from 2024 of the archived core. Samples of the archival core selected for study were sprayed with matte finish polyurethane and photographed using a polarizing filter to enhance contrast and reduce glare. Software counts were performed and compared to two separate manual counts of easily discerned varves conducted by different operators. One manual count was completed Sanderson & Robertson (2005) shortly after the drill core was collected and the other completed in 2024 for the archived core.

Reliable software counts were only achievable following strict procedures. The accuracy of the software counts was heavily influenced by the average varve thickness value that must be input at the start of the semi-automated counting process. This limitation was accommodated using short interval samples of similar varve thickness. Designed for processing of large line scan data sets, the software has toggles for splitting or combining varves that do not fall within 50% of the average thickness. Acceptable results were achieved only by disabling this feature. Additionally, the user-selected area where the count will be performed is critical to obtaining accurate counts and must be a very clean (undisturbed) and generally a very narrow section of the sample. Following this approach and using short interval samples of similar varve thickness, the software can routinely yield counts that match manual counts. Although, due to these constraints it provides little advantage over manually counting varves in outcrop or drill core photographs.