PP53B-2346
Examination of Pedogenic Weathering Indices in the Coterminous USA Using a Large Nationwide Soil Database
Friday, 18 December 2015
Poster Hall (Moscone South)
Nick Bader, Whitman College, Walla Walla, WA, United States
Abstract:
Geologists routinely characterize soil and paleosol samples by their bulk elemental composition via X-ray diffraction or mass spectrometry. As a result, pedogenic weathering indices based on these bulk analyses, such as the chemical index of alteration, have been widely employed in paleosol studies in order to reconstruct paleoclimate variables such as mean annual temperature and mean annual precipitation. These pedogenic weathering indices provide valuable climate estimates for terrestrial environments where climate proxies are sparse. However, because soil properties are a function of a large number of environmental variables, pedogenic weathering indices tend to have only modest predictive power for individual climate variables. Past researchers have used various datasets to develop these climate functions, notably the Marbut (1935) dataset. In this study, we used the Smith at al. (2013) dataset to examine the effectiveness of these weathering indices in estimating current climate variables and soil properties. The Smith et al. dataset includes elemental analysis (via mass spectrometry) and mineralogical analysis (via X-ray diffraction) for A and C horizons for 4857 soils in the coterminous United States. The Smith et al. data was collected for baseline estimation of environmental contaminants, and lacks B-horizon characterization. The data was joined to climate variables calculated from the PRISM dataset, and to additional soil variables from the United States Department of Agriculture's SSURGO database. The data was divided into a training subset and a testing subset using generalized random tesselation stratified sampling. Correlations between pedogenic indices and climate variables are generally modest, indicating that much of the variability in climate data is not explained by these variables. Simple indices such as the chemical index of alteration tend to perform better than experimental pedogenic weathering indices that account for differences between the A and C horizon. There are large differences in model performance between the eastern and western United States, suggesting that models developed for the East may perform poorly in the West, and vice versa.