3-Hydroxy fatty acids (3-OH-FAs), derived from Gram-negative bacterial outer membranes, have received recent attention for their potential as new terrestrial pH and temperature proxies for palaeoclimate studies. Initial studies from altitudinal transects of contemporary soils – correlating bacterial 3-OH-FA compositions to air temperature and pH – have shown promising results. But the geographical extent of recent calibrations is limited. In this study, we analyse 3-OH-FA lipid distributions in 186 globally distributed soil samples to study the environmental factors controlling the relative distribution of the 3-OH-FA isomers. Our sample-set covers a wide range of temperatures (−0.4 to 27 °C) and pH (3.6–9.2). For the global compilation we find that the ratio of anteiso to normal 3-OH-FAs of the C15 or C17 homologues (RAN15 or RAN17) shows a strong linear relationship with mean annual air temperature (MAAT) (R2 = 0.48, p < 0.001 and R2 = 0.41, p < 0.001, respectively). Additionally, the negative logarithm of the ratio of the summed iso and anteiso to the total amount of normal 3-OH-FAs (RIAN) is also strongly anticorrelated with the soil pH (R2 = 0.66, p < 0.001). However, we find that for our 3-OH-FA based proxies there are significant differences in slope and intercept of the linear corrections at regional scales. Thus local or regional calibrations are likely preferable (at this stage of 3-OH-FA proxy development) for application to specific palaeoclimate archives. We also explore the relationship of 3-OH-FA isomer fractional abundances to environmental parameters using machine learning tool (a Gaussian Process (GP) emulator). This confirms the first order relationships to environmental parameters highlighted by the empirical equations and also derives several alternative GP emulator models for reconstructing MAAT and pH which give higher R2 values (0.66 for MAAT; 0.63 for pH) and lower RSME values (3.5 °C for MAAT; 0.76 for pH) compared to simple linear regressions at the global scale. We compare our 3-OH-FA based indices with bacterial branched glycerol dialkyl glycerol tetraethers (brGDGTs) based indices from the same soil samples. At a global scale RAN15 and RAN17 show negative correlations with the MBT′5ME-MAAT (MBT′5ME, methylation index of 5-methyl branched tetraethers) (r = −0.59, p < 0.001 and r = −0.42, p < 0.001, respectively), whilst RIAN shows strong linear correlations with the cyclisation ratio of branched tetraethers (CBT) (r = 0.77, p < 0.001). Similar to 3-OH-FA based temperature proxies, GDGT based temperature proxy MBT′5ME also showed different regional calibrations. Our new field-based correlations demonstrate the broad physiological response of Gram-negative bacterial cell membranes to external environmental changes on a global scale. We suggest that 3-OH-FA based proxies have widespread potential for palaeoenvironmental studies to estimate past MAAT and soil pH, but that regional/local and context specific calibrations may need to be applied.

Original link：10.1016/j.gca.2021.03.010