Extracting paleoclimate signals from sediment laminae: An automated 2-D image processing method

摘要

Lake sediments commonly contain laminations and the occurrence and quantitative attributes of these microstrata contain signals of their depositional environment, limnological conditions, and past climate. However, the identification and measurement of laminae and their attributes remains a largely semi-manual process that is tedious, labor intensive, but subject to human error. Here, we present a method to automatically measure and accurately extract lamina properties from sediment core images. This method is comprised of four major components: (1) image enhancement that includes noise reduction and contrast enhancement to improve signal-to-background ratio and resolution of laminae; (2) identification of 1-D laminae for a user-chosen area in an image; (3) laminae connectivity analyses on the 1-D laminae to obtain a lamina stratigraphy; and (4) extraction and retrieval of the primary and derived lamination stratigraphic data. Sediment core images from Lake Hitchcock and Lake Bosumtwi were used for algorithm development and testing. Our experiments show a complete match between laminae produced by the software and manual process for images from Lake Hitchcock. Quantitative comparisons reveal an insignificant discrepancy in the number of laminae identified automatically by the software and manually by researchers, and in over 90% of the cases the position mismatch of individual laminae is less than one pixel between the software and the manual method for the experimental images from Lake Bosumtwi.