Yongsong Huang, PhD, Brown University
In the past year, we analyzed a total of 88 soils for this study. The soils were mostly collected, dried and stored by the United States Department of Agriculture National Resources Conservation Service (USDA-NRCS) National Soil Survey Center located in Lincoln, Nebraska, between 2007 and 2010. The sampled depths all begin at the surface and extend to depths that range from 3 to 36 cm deep. Soil type/temperature regimes included mesic, isomesic, thermic, frigid, cryic, and gelic. We collected an additional five samples from near Toolik Lake, Alaska. These samples were taken from locations with different soil and vegetation characteristics including a dry heath stand (well drained), moist acidic tundra and non-acidic moist tundra (poorly drained; organic and mineral layers both collected).
The spatially heterogeneous arctic landscape allowed these samples to be collected within a kilometer of each other and they were therefore subjected to the same climatic regime, but different local conditions. Environmental variables were collected and archived in a database by the USDA-NRCS National Soil Survey Center. We extracted and sorted variables from the database through use of matrix programming code. We measured pH values of samples collected from Toolik lake using a 1:2 by mass soil : water mixture and an Accumet AP 115 pH meter. We obtained surface temperature, accumulated precipitation, and evapotranspiration values in .nc format from the NOAA Monthly/Seasonal Composites (http://www.esrl.noaa.gov/psd/cgi-bin/data/composites/printpage.pl) and converted the data into ArcGIS maps for further analysis. The temperature is measured in degrees K, and annual accumulated precipitation is measured in kg m-2. These data represent the mean values from 1980-2009 at 0.3 degree latitude x 0.3 degree longitude resolution across the U.S. Data were extracted for specific sample locations in ArcGIS. We compared sample point temperature values, when possible, against local weather station data (the closest climate station data to our site in terms of latitude, longitude, and elevation).
We analyzed samples using an HPLC-APCI-MS with an Agilent 1200 series instrument coupled to a single quadrupole mass spectrometer to determine the distributions of branched GDGTs in the samples, after purifying the lipid fractions containing these compounds, using the published procedures.
We hypothesized in our original proposal that a wider range of environmental variables play a part in controlling the distribution of GDGTs than has been found previously (TOC). Our results show that changes in the branched GDGT distributions show significant correlation with changes in environmental pH (pH= -1.316CBT + 7.3; r2 = 0.2547; or pH= -1.34 CBT + 7.131; r2 = 0.4451 if excluding samples with pH>8; pH= -1.665CBT + 7.94; r2 = 0.6084 if only samples from Montana and South Dakota are included), as well as precipitation amounts (Precipitation = 3.012 MBT + 0.7793; r2 = 0.6312). However, the relationship of branched GDGTs to temperature is relatively weak (east coast soils: r2 = 0.269; Montana and South Dakota soils: r2 = 0.0604; Midwest soils: r2 = 0.177). Our results suggest that temperature reconstructions involve a more complex set of controls than previously realized including possibly vegetation effects and certainly water throughflow / soil moisture effects on the GDGT-producing bacteria’s optimal growing season. It is likely that the distribution of branched GDGTs is controlled by precipitation / soil moisture in two ways. The first is by shifting the optimal growing season, causing the distributions of GDGTs that are preserved to represent the mean annual temperature of that season rather than the whole year in sites where this effect occurs. The second is through an overabundance of GDGT III caused by water saturation in soils. This will impose a significant challenge to reconstructing temperatures from soils with differing precipitation and soil moisture regimes; however, it is also possible that future studies can take advantage of this relationship in order to uncover paleo-precipitation information as well as paleo-temperature. We also found that there exists a pH threshold around pH 8, above which GDGT abundances do not continue to linearly redistribute and therefore pH reconstructions are not accurate.
We are continuing in the second year of the project to extract more soil samples, targeting areas that were not studied in the first year. We are also working on using multivariate linear regression method to determine if the fractional abundances of various GDGTs (rather than MBT, CBT) can be better formulated to environmental parameters.