Reports: ND251690-ND2: Re-examination of environmental controls on bacterial branched tetraether lipids using North American soils
Yongsong Huang, Brown University
This ACS-PRF grant has made tremenduous impact on my scientific career. It bridged a major funding gap for me from NSF due to severe budget cuts in recent years. With this funding, I was able to complete the remaining work and publish key papers that allow me successful obtain two new NSF grants and an Brown University seeding grant starting from 2015, in addition to producing exciting new results from the proposed work. I feel very grateful to ACS-PRF for the timely and generous funding. This grant also helped support my graduate students (James Dillon) and visiting scientist Dr. Qingqiang Chen, as well as undergraduate students.
The main objective of the study is to use a large set of soils in the North America with well-defined environmental conditions to assess how the distribution of bacteria-derived branched glycerol dialkyl glycerol tetraethers (b-GDGT) respond to temperature, precipitation, soil pH and other potential impact factors. To achieve this goal, we analyze b-GDGTs in 103 strategically selected soil samples throughout North America (Fig.1), and integrate all available data to re-assess environmental impact on b-GDGT distribution. The patterns of methylation and cyclization (MBT and CBT indices) in the ubiquitous, bacteria-derived branched glycerol dialkyl glycerol tetraethers (b-GDGT) have been shown to vary systematically with temperature and soil pH (Weijers et al., 2007). There have been major efforts to calibrate b-GDGT distributions against environmental variables in globally distributed soils (including 135 North American soils) (Dirghangi et al., 2013; Peterse et al., 2012). However, there are two problems regarding soil b-GDGTs in the North America: 1) published data so far are geographically limited and insufficient for establishing robust regional calibrations along key environmental gradients; and 2) there have been no attempts to apply multiple linear regression (MLR) of fractional abundance, which has been applied successfully in African and Chinese soils, to North American soils. We combine our data with the previously published b-GDGT analyses from 135 North American soils and perform systematic statistical analyses to obtain the optimal results.
Combining our new results with the published 135 soil b-GDGT data allows much more comprehensive geographic coverage of soils from different environmental and climatic conditions and permits more robust statistical assessment of factors controlling b-GDGT distributions in North America. Our systematic division of samples using different ecological or environmental criteria (e.g., vegetation type, precipitation range, temperature range, pH range etc.) provides the opportunity to perform in-depth analysis of the regional differences in factors controlling b-GDGTs in soils. Our large data set also allows us to systematically compare effectiveness of empirical indices (MBT/CBT or MBT/CBT) with stepwise multiple linear regression in identifying the best linear regression coefficients based on b-GDGT distributions. Overall, we find that stepwise multiple linear regression approach yields stronger linear regression coefficients between b-GDGT distributions and environmental parameters (AAT, soils pH, and AAP) than MBT/CBT or MBT/CBT indices, although both approaches identify similar environmental factors as the predominant (or unimportant) controls on b-GDGT distributions in regional calibrations.
In almost all the cases, MLR yielded higher R2
values and smaller errors than the empirical MBT/CBT indices, supporting the
use of MLR over MBT/CBT indices for paleoclimate reconstructions. Notably, annual
average precipitation (AAP) played a major role, in many cases stronger than annual
average temperature (AAT), in controlling the b-GDGT distributions in North
America. Similar results (i.e., strong precipitation control on b-GDGT
distributions) are obtained from published data for Chinese soils using MLR.
The strongest correlation between AAP and b-GDGT is found in mid-latitude
grasslands (R2 = 0.78), or when soil pH values are greater
than 8 (R2 = 0.88). A temperature effect is most apparent
only in regions of high AAP (e.g., >1200 mm) (R2 = 0.75)
or with mixed forest vegetation (R2 = 0.70). The effect of
soil pH on b-GDGT distribution is strong in almost all regions. Our results suggest
that climate reconstructions using b-GDGTs involve a more complex set of
controls than previously realized, including vegetation and AAP effects in arid
regions.
An important new finding of our study is that, over the scale
of North American, the distributions of b-GDGTs are generally more strongly
correlated with precipitation (represented by AAP) than temperature
(represented by AAT). Specifically, when all samples are taken into
consideration, the R2 values of linear regression between AAP
and b-GDGT distributions are 0.705 (for MBT/CBT), 0.700 (for MBT/CBT), and
0.74 (for MLR), respectively. In contrast, the R2 values of
linear regression between AAT and b-GDGT distributions are 0.138 (for MBT/CBT),
0.137 (for MBT/CBT), and 0.132 (for MLR), respectively. Impact of pH on
b-GDGTs is strong in almost all environments.