A SOIL AND PALEOSOL INFORMATICS APPROACH TO DEVELOPING PALEOCLIMATE PROXIES

Paleopedological studies rely heavily on the use of contemporary soil characterization and whole-soil geochemical data. These soil science resources serve the needs of paleopedologists who reconstruct ancient climate and soil systems using proxies and pedotransfer functions that relate modern physical/chemical soil characterization data, whole-soil geochemical data, and climate parameters. The (paleo)pedologist currently faces a “data overload” problem due to a number of global and continental-based soil geochemical databases becoming widely available. We used a data analysis approach to construct the Baylor Paleosol Informatics Cloud (BU-PIC). The BU-PIC uniquely combines: (1) NRCS pedon data, (2) PRISM-based climate parameters, and (3) NLCD land-cover attributes. The BU-PIC will eventually incorporate (4) published paleosol data. This aggregation of data will allow paleopedologists to upload geochemical data and test and refine soil-derived paleoclimate proxies and paleopedotransfer functions. Although BU-PIC development is in the initial stages of data cleaning, initial analysis shows promising results. For example, the CALMAG molecular geochemical ratio of Al₂O₃/(Al₂O₃+CaO+MgO) * 100 from modern B horizons that formed in loess explains 82% of the variance in mean annual precipitation (MAP). A preliminary analysis of the sand-silt mineralogy, clay mineralogy, and exchangeable species suggests that CALMAG_loess values are driven by hydrolysis of weatherable minerals (e.g., plagioclase, biotite, and glass) and the alteration of clay minerals (e.g., smectite to kaolinite). Lower CALMAG_loess values are likely driven by the leaching of exchangeable Ca and Mg. We used chemical analyses of 21 previously published pedons with known MAP values to test the predictive strength of the CALMAG_loess regression equation. The average difference between observed and predicted MAP values was 10%, suggesting that CALMAG_loess holds great potential for predicting paleo-rainfall. Because parent material composition also controls elemental distribution, the proxy is only effective when oxide values from unknown MAP samples fall within the range of those data used to construct the regression.