Improving paleoclimate predictions from paleosol geochemistry: A case study from the Miocene of Argentina

Abstract

Paleoclimate records from past climate transitions for which there is no historical precedent have become a key focus for researchers looking to improve our understanding of the climate system. Records derived from terrestrial archives have become increasingly important as observations and modeling show that both heterogeneity and amplification of climate are more apparent on land relative to the oceans. Paleosol archives have served as the backbone of most terrestrial paleoclimate records in deep time, based on “climofunctions” which have evolved through time in terms of their approaches and complexity. Despite advancements, paleosol geochemistry-based paleoclimate proxies/models are still limited by gaps in training datasets, poorly constrained uncertainties, lack of accessibility for model inputs/code, and few response variables in paleoclimate outputs. Here we address these issues by assembling larger modern soil datasets compiled from international sources, apply new statistical methods via partial least squares (PLS) with secondary corrections and bootstrapping analyses, provide accessible user packages in a variety of modeling languages, and develop new response variables for climate predictions.The new Paleosol Geochemistry Paleoclimate Model (PGPM) produces estimates of mean annual temperature (MAT), mean annual range of temperature (MART), mean annual precipitation (MAP), and growing season precipitation (GSP) which have lower root mean square error and a wider range of applicability than any other available paleosol climofunctions. In addition, we test the PGPM by applying it to a north-south transect of Late Miocene localities from Argentina (Rio Iruya, La Viña, Palo Pintado, Cacheuta, Cerro Azul) for which estimates of MAT and MAP are published and qualitative descriptions of MART and GSP are available. We find that estimates of MAT range from 8 to 13℃, which are higher than previous paleosol-based estimates and more in line with expected modern values. We also find that MART values are comparable to modern (~24℃), and in line with modeled and paleobotany-based estimates. For MAP, estimates range from 350 to 1000 mmyr-1, which compare favorably to other paleosol-based estimates. GSP estimates range from 200 to 550 mmyr-1, which agree with modeled seasonality for the region and are novel from a proxy perspective.