Earth and Atmospheric Sciences, Department of


Date of this Version



Published in Quaternary Science Reviews 28:1–2 (January 2009), pp. 120–136; doi: 10.1016/j.quascirev.2008.08.021 Copyright © 2008 Elsevier Ltd. Used by permission.


Taphonomic issues pose fundamental challenges for Quaternary scientists to recover environmental signals from biological proxies and make accurate inferences of past environments. The problem of microfossil preservation, specifically diatom dissolution, remains an important, but often overlooked, source of error in both qualitative and quantitative reconstructions of key variables from fossil samples, especially those using relative abundance data. A first step to tackling this complex issue is establishing an objective method of assessing preservation (here, diatom dissolution) that can be applied by different analysts and incorporated into routine counting strategies. Here, we establish a methodology for assessment of diatom dissolution under standard light microscopy (LM) illustrated with morphological criteria for a range of major diatom valve shapes. Dissolution data can be applied to numerical models (transfer functions) from contemporary samples, and to fossil material to aid interpretation of stratigraphic profiles and taphonomic pathways of individual taxa. Using a surface sediment diatom-salinity training set from the Northern Great Plains (NGP) as an example, we explore a variety of approaches to include dissolution data in salinity inference models indirectly and directly. Results show that dissolution data can improve models, with apparent dissolutionadjusted error (RMSE) up to 15% lower than their unadjusted counterparts. Internal validation suggests improvements are more modest, with bootstrapped prediction errors (RMSEP) up to 10% lower. When tested on a short core from Devils Lake, North Dakota, which has a historical record of salinity, dissolution-adjusted models infer higher values compared to unadjusted models during peak salinity of the 1930s–1940s Dust Bowl but nonetheless significantly underestimate peak values. Site-specific factors at Devils Lake associated with effects of lake level change on taphonomy (preservation and re-working, implied by dissolution data) may override model improvements incorporating dissolution. Dissolution-adjusted salinity models are also applied to a 150-year sediment record from Spiritwood Lake, North Dakota, which suggests that this lake has a damped and lagged response to major regional climate forcing of salinity during the Dust Bowl. At this site, dissolution data also suggest different taphonomic behavior of taxa related to their seasonal patterns of growth and sedimentation. Thus, dissolution data can improve models, and aid interpretation of sedimentary profiles as records of limnological, ecological and environmental change, filtered by taphonomy.