Vegetation and climate change in eastern continental Asia during the last 22 ka inferred from pollen data synthesis

Diedrich.Fritzsche [ at ]


The spatial and temporal climate and vegetation patterns and their driving forces across eastern continental Asia are very complex and ecosystems in the arid/semi-arid areas are particularly sensitive to climate change. To investigate quantitatively the former climate and past biogeographical patterns at a broad spatial scale, and to understand the relationship between climate and vegetation on a long temporal scale, I have established a taxonomically harmonized and temporally standardized fossil pollen dataset (274 pollen records) covering the last 22,000 years at 500-year resolution across the eastern part of continental Asia, and a corresponding modern pollen dataset of 2559 sites. I proceeded to use the relationships between modern pollen and plant distribution and climate to make quantitative reconstructions of past plant distributions and past climate. To determine the pollen-percentage thresholds of a taxon’s presence/absence and dominance/presence, logistic regression was employed based on modern pollen percentages and corresponding plant distributional data for 14 key arboreal pollen taxa (Abies, Betula, Castanea, Castanopsis, Cyclobalanopsis, Fagus, Juglans, Larix, Picea, Pinus, Pterocarya, Quercus, Tilia, Ulmus) from China. The obtained pollen threshold values are reasonable when compared with pervious local or regional pollen‒vegetation relationship studies. I used the fossil pollen dataset to investigate past tree distribution changes based on pollen threshold values with a high probability to represent plant presence and/or dominance. The 14 major arboreal taxa show marked changes in their spatial and temporal distributions and follow different expansion patterns. The thermophilous (Castanea, Castanopsis, Cyclobalanopsis, Fagus, Pterocarya) and eurythermal (Juglans, Quercus, Tilia, Ulmus) broadleaved tree taxa were restricted to the current tropical or subtropical areas of China during the Last Glacial Maximum (LGM), spread northward since ca. 14.5 ka BP, and flourished in the Holocene. Their expansion patterns imply that both temperature and precipitation are important driving forces of their distribution. Betula and conifer taxa (Abies, Picea, Pinus) retained a wider distribution during the LGM and flourished in the different stages of the Holocene among different regions, indicating that precipitation plays a more important role than temperature for their distribution. Since the late mid-Holocene the abundance but not the spatial extent of most trees has decreased: these decreases might be caused by enhanced human impact. For the quantitative climate reconstruction, I employed a series of numerical analyses to determine which climatic variable is the most reliable for a pollen-based reconstruction, which numerical method is the most reliable for my modern pollen dataset, and how large a spatial extent the calibration set should cover to be the most robust for a single pollen record. Modern pollen‒climate relationship investigations using canonical correspondence analysis (CCA), and Huisman-Olff-Fresco (HOF) models, indicate that mean annual precipitation (Pann) is the most important climatic determinant of pollen distribution. The model performances under cross-validation for the modern analogue technique (MAT) and weighted-averaging partial least squares (WA-PLS) also imply that Pann is the most promising climate variable for the reconstruction. Testing for spatial autocorrelation reveals that the better model performances of MAT are most probably caused by spatial autocorrelation, while the WA-PLS models are only slightly influenced. Hence, the WA-PLS models of my modern pollen dataset are deemed most suitable for reconstructing past climate. However, these WA-PLS models produced high root mean square errors of prediction (RMSEPs) because of the strong noise in the pollen data which is related to the long climatic gradients. It is necessary to compile a unique pollen‒climate calibration set of relatively narrow spatial extent around each fossil pollen record. Here, I used the statistical significance test and analogue quality to determine the most appropriate spatial extent to use for the calibration sets for three examples (pollen spectra from Lake Ganhai, Lake Qinghai and Lake Dahu), and concluded that a calibration set with ca. 1000 km around each fossil pollen record should be the most suitable for reconstructing climate. I created a series of calibration sets from the the modern pollen dataset with a 1000 km spatial extent around each available fossil pollen record (112 records covering the period between 10 and 2 cal ka BP) and established WA-PLS models and reconstructed Pann for these records. To synthesize the general evolution patterns of these Pann reconstructions, a Fuzzy C-means algorithm was employed to cluster those Pann records with good model statistics for WA-PLS and significant temporal changes (49 records) to make a reliable reconstruction. Results of the clustering show that Pann evolved differently between regions producing three patterns: general increase, general decrease, and unimodal with a mid Holocene (ca. 8–4 cal ka BP) Pann optimum. In eastern China, the spatial distribution of Pann during the Holocene is similar to that of the past 50 years, with a weak East Asian Summer Monsoon (EASM) in the early Holocene (10–8 cal ka BP), a strong EASM in the mid Holocene (8–4 cal ka BP), and an even weaker EASM in the late Holocene (4–2 cal ka BP) than in the early Holocene. The EASM evolution during the Holocene implies its intensity has teleconnections with the sea surface temperature in the central and eastern tropical Pacific Ocean and North Atlantic Ocean.

Item Type
Thesis (PhD)
Primary Division
Primary Topic
Publication Status
Eprint ID
Cite as
Cao, X. (2015): Vegetation and climate change in eastern continental Asia during the last 22 ka inferred from pollen data synthesis , PhD thesis, Mathematisch-Naturwissenschaftliche Fakultät / Institut für Erd- und Umweltwissenschaften.

Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email

Geographical region

Research Platforms


Edit Item Edit Item