Comparison of techniques to reconstruct palaeoclimates of China

Investigating the past climate is helpful to distinguish humankind’s contribution on presentclimate change, and also helpful to understand the historic or geological evolution of global environmental conditions. A number of different types of records provide information about past climate changes, including petrology, sedimentology, glaciology, dendrochronology, palynology and others. Various different techniques can be used to make quantitative reconstructions from palaeodata. All of these methods have their shortcomings and advantages. In this thesis, I examine two reconstruction methods, which have been widely applied to make reconstructions of climate during pre-Quaternary times, specifically the coexistence approach and leaf traits analysis. The coexistence approach (CoA) assumes that the climate of a fossil assemblage can be defined from the overlap between the climate ranges of the individual taxa, where the climate range of each taxon is defined by the climate range under which it grows today. For taxa that are no longer extant, the range is defined as that of the nearest living relative (NLR). The method assumes that it is possible to define the modern climate tolerance accurately and also that taxa were physically present at the site. In Chapter 2, I test the impact of these two assumptions on CoA reconstructions of mean annual precipitation (MAP), mean annual temperature (MAT), mean temperature of the warmest month (MTWA) and mean temperature of the coldest month (MTCO) using modern pollen data from the Qinghai-Tibetan Plateau. I find that the data quality of NLRs and the exotic pollen seriously affect the reconstructed MAP, MAT, MTWA and MTCO. The uncertainties are also relatively large, especially for those three temperature parameters, even those two factors considered. Thus the complementary method should be explored. Methods based on the correlation between leaf physiognomy and climate variables provide an alternative approach to reconstructing past climate. The two most widely used methods are like leaf margin analysis (LMA) and the climate leaf analysis multivariate program (CLAMP). The disadvantages of these methods are identified in various studies, such as using the limited leaf traits and the problem of correlation within traits. Hence, we are exploring our leaf traits-climate model method. I first establish the modern relationships between leaf traits and climates (Chapter 3) based on a large data set of modern trait observations from China. I used logistic regression techniques to investigate the relationships between summer temperature (measured by the accumulated temperature sum above 0°C), plant-available moisture (measured by the ratio of actual to equilibrium transpiration) and seasonality (measured by the daily mean growing season temperature when temperatures are >5°C) and the frequencies of 25 leaf morphometric traits, collected from 98 sites sampling the range of climate and vegetation types found in China. Results show that these morphometric traits vary along climate gradients in a predictable and understandable way. Different traits combination can outline the specific climate space. Leaf traits responding to one or more climate variables indicate that traits could play multiple functions on the adaptation for the moisture and temperature. Many specific relationships between traits and bioclimate variables are conservative across all woody life forms. These findings lay a stronger foundation for using morphometric traits to reconstruct past climates. In Chapter 4, I apply the independent relationships between specific leaf traits and individual climate variables to build predictive models for estimating the length of the growing season (GDD0: growing degree days above a baseline of 0°C) and plant-available moisture (α: the ratio of actual to equilibrium evapotranspiration). I then apply these models to predict the paleoclimate of four fossil leaf floras: from the Fushun Basin (Eocene), from Shanwang Basin (Middle Miocene), from Xiaolongtan Basin (Late Miocene), and from Shengzhou (Pliocene) in China. These geological times are examples of climate intervals when CO₂ was higher than today, and as such provide opportunities to examine how the climate system has responded to enhanced greenhouse gas concentrations. Results show that our models have relatively small reconstruction biases under modern conditions. The reconstructed paleoclimates by the modified models are comparable with previous reconstructions, but our results show more constraints rather than large uncertainties like previous reconstructions. Paleoclimate changes for these four sites are consistent with the evolution of climates in this region and compatible with enhanced monsoon conditions during these high CO₂ intervals. In summary, my thesis makes three important contributions to the field of palaeoclimate reconstruction. Firstly, by quantifying the impact of extra-local pollen on climate reconstructions for the Tibetan Plateau, I demonstrate the unreliability of the coexistence approach when applied in open vegetation. This work suggests that the coexistence approach should only be used when pollen samples can be combined with e.g. macrofossil evidence that would demonstrate the local presence of individual species. Secondly, through applying generalised linear modelling (GLM) technique to establish the independent relationships between multiple climate variables and leaf morphometric traits, after removing the influence of interactions between these variables, I have shown why univariate correlations as used in standard methods such as leaf margin analysis (LMA) and the climate leaf analysis multivariate program (CLAMP) are noisy and unreliable. The GLM methodology provides a better way to use leaf traits to reconstruct climate. Finally, I have developed a new multi-model technique based on these independent trait-climate relationships to reconstruct palaeoclimates in China from fossil leaf floras, and demonstrated that this provides well-constrained estimates of temperature and moisture variables.