Sources of local and regional variability in the MBT'/CBT paleotemperature proxy: Insights from a modern elevation transect across the Eastern Cordillera of Colombia

Distributions of brGDGTs (branched glycerol dialkyl glycerol tetraethers) in soils have been utilized in the literature as a proxy for temperature and pH via the MBT' and CBT indices (methylation of branched tetraethers and cyclization of branched tetraethers, respectively). However, there is substantial scatter in the empirical relationships between the global MBT'/CBT index and modern temperature, resulting in uncertainty as large as±5.0°C in reconstructed paleotemperature. In this study, we sought to determine the magnitude of several sources of calibration error using a new set of samples spanning a large gradient in elevation (ca. 3000m) and temperature (ca. 16°C) across an Andean transect in the Eastern Cordillera of Colombia. Using in situ temperature monitoring, this field study of an equatorial montane location provided an opportunity to investigate several potential confounding factors affecting the MBT'/CBT paleothermometer, including the mismatch between air temperature and actual soil temperature, and the effect of soil moisture on brGDGT production. In the well-constrained environment of our study, we determined the RMSE (root mean squared-error) for a local calibration of the MBT'/CBT proxy to be 3.0°C, in contrast to the 5.0°C value reported in the most recent global calibration. A comparison between in situ soil temperature measurements at our soil sampling locations and interpolated weather station temperature data indicates that errors in interpolation schemes are not likely to be a significant contributor to the observed scatter in the MBT'/CBT-temperature relationship. Rather, we conclude that the majority of the 3°C scatter in this calibration for the northern Andes arises from transient, site specific variation in factors such as soil moisture and nutrient availability, with several cases showing a difference of up to 7.8°C between closely spaced sites (<600m in distance and<50m in elevation apart). In addition, we compared our dataset with five published regional soil elevation transects and found that that the MBT' index correlated strongly with temperature at sites with high precipitation, such as Colombia (R² 0.67, 0.47), but correlated weakly with temperature at arid sites (R² 0.01, 0.04). Instead, the arid sites showed a strong relationship among MBT', CBT and pH that was absent from the higher precipitation transects. This bolsters previous findings that the relationship between brGDGT distributions and temperature may change at sites with low soil moisture. This heterogeneous response in brGDGT distributions to temperature and pH may be largely responsible for the significant scatter in the global calibration dataset. We therefore conclude that the RMSE of 3°C in our local calibration may represent a lower boundary on the error in MBT'/CBT-based temperature reconstruction in moist regions, but further work is needed to reliably identify suitable conditions for the application of this proxy to ancient samples.