Environmental and Climatic History During the Past Centuries in the Hengduan Mountains (Southwest China) Derived from Tree Rings

Home , Meili Snow Mountains, Tanggula Mountains

Environmental and Climatic History during the Past Centuries in the Hengduan Mountains (Southwest China) Derived From Tree Rings

Den Naturwissenschaftlichen Fakultäten Der Friedrich-Alexander-Universität Erlangen-Nürnberg zur Erlangung des Doktorgrades

vorgelegt von Ze-Xin Fan aus Yunnan / China

Als Dissertation genehmigt von den Naturwissenschaftlichen Fakultäten der Friedrich-Alexander-Unversität Erlangen-Nürnberg

Tag der Mündlichen Prüfung: 21. July 2009

Vorsitzender der Prof. Dr. Eberhard Bänsch Promotionskommision:

Erstberichterstatter: Prof. Dr. Achim Bräuning

Zweitberichterstatter: Prof. Dr. Dieter Eckstein

Declaration I

Declaration

I hereby declare that my dissertation entitled: “Environmental and Climatic History during the Past Centuries in the Hengduan Mountains (Southwest China) Derived from Tree Rings” is my own work under the supervising of Prof. Dr. Achim Bräuning. The author is responsible for all data analyses and interpretation of the results. The main content work has not previously been submitted for a degree at any tertiary institution. I also confirm that I have fully acknowledged by name all of those individuals and organizations that have contributed to the research for this dissertation.

Erlangen, 20.7.2009 ______

Ze-Xin Fan

Acknowledgements II

Acknowledgements

Initially I would like to express my gratitude to the Max-Planck-Gesellschaft, who provided me with a 3-years fellowship during my research work in Germany, and supported me for attending the international tree-ring conferences in Latvia and Poland. I would especially like to thank Mrs. Sabine Panglung for her perfect organization of my fellowship and for patient communications.

I would like to express my sincere gratitude to my supervisor Prof. Dr. Achim Bräuning. He helped me to set up the research plan, taught me how to approach scientific issues from various perspectives, and how to write a concise scientific paper. I would especially like to thank Prof. Bräuning and Dr. Aster Gebrekirstos for their warm friendliness and hospitality during my stay in Germany.

I am very indebted to Dipl.-Geogr. Iris Burchardt who helped me building the first tree-ring chronology, gave me many invaluable suggestions during my lab work, and improved the English language of this thesis. Thanks go to Dipl.-Geogr. Peter von Schnakenburg and Dipl.- Geogr. Iris Burchardt who translated English summary into German. I highly appreciate the staff from the Institute of Geography, University of Erlangen-Nürnberg, Mrs. Sabine Donner, Dr. Thomas Sokoliuk, Prof. Dr. Michael Richter, Dr. Jussi Grießinger, Dipl.-Geogr. Franziska Volland-Voigt, Dipl.-Geogr. Sebastian Feick, Dipl.-Geogr. Malanie Ehmisch-Feick and ….

I am also very grateful to my Chinese promoter Prof. Dr. Cao KunFang for continually stimulating me during my study and for supporting me financially during my field work. Thanks to Prof. Dr. Li QingJun for sharing some of his tree-ring samples. My gratitude also goes to M.Sc. Zhu ShiDan and Mr. Di ChengQiang for their energetic assistance in the field, and many warm hospitable Tibetans for their enthusiastic and friendliness in those remote villages.

Many thanks to Prof. Yang Bao for his invaluable suggestions on my research project and his enthusiasm and helps during my stay in Germany. I also thank Prof. Wang Lily and Prof. Ram R. Yadav for all their comments on my research program and thesis. I also thank Prof. Shao XueMei and Dr. Liang ErYuan who kindly provided their tree-ring data. I appreciate Prof. Dr. Dieter Eckstein who reviewed this thesis. Thanks to Prof. Dr. Rupert Bäumler, Prof. Dr. Werner Nezadal and Prof. Dr. Michael Richter who gave the examination.

Thanks also go to colleagues Dr. Zhang JiaoLin, Dr. Chen JunWen, Dr. Wei ZuoDong, Dr. Zhang ChuanLin, Dr. Hao GuangYou, Dr. Zhang YongJiang, M.Sc. Chen YaJun, M.Sc. Wang AiYing, M.Sc. Fu PeiLi, Dr. Tian QinHua, Dr. Li JinBao for their friendliness and help.

Finally, I would like to express my sincere respects to my parents and to my parents-in-law for their continuous support. Most of all, I want to express my special thanks to my wife Mrs. Ji ShiMei and my son Fan XiaoHui, for all their consideration, understanding, patience and love.

Table of contents III

Table of Contents

Declaration...... I

Acknowledgements...... II

Table of Contents ...... III

Abbreviations...... VI

List of Figures...... VII

List of Tables...... XII

1. Introduction and Research Aims...... 1

1.1 Climate Change ...... 1 1.2 Tree Rings as Climate Proxies ...... 2 1.3 Tree Growth and Climate...... 3 1.3.1 Tree-Ring Formation...... 3 1.3.2 Dendroclimatology...... 4 1.3.3 Climate-Growth Responses ...... 4 1.3.4 Tree-Ring Parameters ...... 5 1.4 Climatological Relevancy of the Tibetan Plateau...... 6 1.5 Dendroclimatology on the Tibetan Plateau...... 7 1.6 Dendroclimatology in the Hengduan Mountains...... 10 1.7 Objectives and Research Questions...... 11 1.8 Outlines of this Thesis ...... 12

2. Physical Setting...... 13

2.1 Location of the Study Area...... 13 2.2 Climate Conditions...... 15 2.3 Regional Climate Patterns...... 18 2.4 Vegetation ...... 20 2.5 Land Use and Forest Disturbances...... 24

3. Data and Methods ...... 26

3.1 Tree-Ring Network...... 26 3.2 Tree-Ring Measurement and Cross-Dating...... 26 3.3 Tree-Ring Standardization...... 29 3.4 Common Signal Extraction ...... 31 3.5 Climate Data ...... 33 3.6 Calibration and Reconstruction...... 33 3.6.1 Climate-Growth Response...... 33 3.6.2 Transfer Functions...... 34

Table of contents IV

4. Climate-Growth Responses of High-Elevation Conifers...... 37

4.1 Data Source and Methods...... 37 4.2 Internal Chronology Statistics ...... 38 4.3 Comparison of Chronologies...... 40 4.4 Climate-Growth Response...... 41 4.5 Effects of Winter Climate on Tree Growth...... 45 4.6 Site-Specific Climate-Growth Responses ...... 47 4.7 Species-Specific Climate-Growth Responses ...... 48

5. Temperature Reconstructions...... 49

5.1 Tree Ring Width-Based Annual Temperature Reconstruction ...... 49 5.1.1 Chronology Comparisons...... 50 5.1.2 Climate-Growth Relationship...... 52 5.1.3 Annual Temperature Reconstruction ...... 54 5.2 Tree-Ring-Density Based Summer Temperature Reconstruction...... 56 5.2.1 Chronology Statistics...... 59 5.2.2 Growth-Climate Relationships ...... 60 5.2.3 Warm Season Temperature Reconstruction...... 63 5.2.4 Spatial Correlation Analysis...... 65 5.2.5 Comparison with Regional Records...... 68

6. Drought Reconstruction ...... 69

6.1 Data Source and Methods...... 69 6.2 Chronology Statistics ...... 70 6.3 Correlation and Principal Component Analysis ...... 72 6.4 Climate-Growth Responses ...... 73 6.5 Drought Reconstruction...... 75 6.6 Spectral Analysis ...... 77 6.7 Spatial Correlations and Teleconnections...... 78

7. Larch Radial Growth and Insect Defoliation ...... 81

7.1 Impact of Insect Defoliation on Tree Growth...... 81 7.2 Detection of Larch Insect Defoliation Signals...... 82 7.3 Outbreaks History Inferred from Tree-Ring Data ...... 83

8. General Discussion and Recommendations ...... 88

8.1 Growth-Climate Relationships...... 88 8.2 Temperature History ...... 89 8.2.1 Annual Temperature Variability...... 89 8.2.2 Summer Temperature Variability ...... 90 8.3 Drought History...... 91

Table of contents V

8.4 Comparison of Different Climate Variables ...... 92 8.5 Recommendations for Further Research ...... 94 8.5.1 Extension of the Chronology Lengths ...... 94 8.5.2 Expansion of Sample Area...... 95 8.5.3 Tree-Ring Parameters ...... 96 8.5.4 The Problem of Larch Insect Defoliation...... 96

9. Summary-Zusammenfassung...... 98

10. References...... 107

Curriculum Vitae ...... 128

List of Publications...... 130

Abbreviations VI

Abbreviations

AC =Autocorrelation AGR =Average Growth Rate CE =Coefficient of Efficiency CRU =Climatic Research Unit EPS =Expressed Population Signal HCA =Hierarchical Cluster Analysis IPCC =Intergovernmental Panel of Climate Change MS =Mean Sensitivity MSL =Mean Segment Length MTM =Multi-Taper Method MXD =Maximum Latewood Density PCA =Principal Component Analysis RDA =Redundancy Analysis PDSI =Palmer Drought Severity Index RE =Reduction of Error SAM =Southwest Asian Monsoon SNR =Signal-to-Noise Ratio TP =Tibetan Plateau TRW =Total Ring Width

List of figures VII

List of Figures

Figure 1.1 Observed changes in global average surface temperature (red) and China temperature (blue) anomaly. Reference period is 1961–1990. Both are smoothed with a 5-year FFT-filter (Fast Fourier Transform) and shown in bolded lines. Data source: Global temperature data (Brohan et al., 2006; http://www.cru.uea.ac.uk), China temperature data (Wang and Gong, 2000; Wang et al., 2004)...... 2 Figure 1.2 Schematic overview on wood structure and tree-ring parameters of Picea brachytyla, produced by the Lignostation densitometry system (Rinntech, Germany)...... 3 Figure 1.3 Topographical view of the Tibetan Plateau and surrounding areas. Approximate present geographical extent of Southwest Asian and East Asian summer monsoons are indicated in dashed lines (redraw after Araguas-Araguas et al., 1998; Böhner and Lehmkuhl, 2005)...... 7

Figure 2.1 Topographical map of the Tibetan Plateau. Black square highlights the central Hengduan Mountains region...... 13 Figure 2.2 Photographic views of the three parallel rivers and ridges in the central Hengduan Mountains, south- western China. Photos: Bräuning A., 2004; Fan Z.X., 2006...... 14 Figure 2.3 Mean air temperature (°C, upper) and precipitation sums (mm, lower) distributions during winter (Nov-Feb) and summer (Jun-Sep) seasons over the Yunnan Province. Meteorological data for 119 stations were obtained from the National Meteorological Information Centre (NMIC) of China. Mean seasonal values were calculated over the period of 1961–2004 and interpolated with the Kriging method...... 15 Figure 2.4 Climate diagrams for the six meteorological stations in the central Hengduan Mountains...... 17 Figure 2.5 Linear trends of air temperatures (°C/10yr, left) and precipitation sums (mm/10yr, right) during winter (NDJF) and summer (JJAS) seasons over the Yunnan Province for the period 1961–2004. Stations with significant (95%) Mann-Kendall trend test are encircled...... 18 Figure 2.6 Temporal changes of regional monthly maximum (Tmax), mean (TEM), minimum (Tmin) temperatures in the summer (a, June-September) and winter (b, prior November-February) seasons; (c) temporal changes of regional diurnal temperature range (DTR). The regional temperature series were developed from two high-elevation meteorological stations of Shangri-la and Daocheng. Values are expressed as anomalies (reference period = 1959–2004) and bold lines were smoothed using a 10-year low-pass filter. R indicates the linear correlation coefficients. The asterisks indicate the 99% significant level and ns denote no significant...... 19 Figure 2.7 Repeat photo-pairs from the Mingyong Glacier site (a, b) and the Baima Snow Mountain site (c, d) in the central Hengduan Mountains. The original photographs in (a) and (c) were taken by Joseph Rock in 1923 (Rock, 1926); repeat photographs (b, d) were taken by Robert Moseley. Adapted and slightly modified from Baker and Moseley (2007)...... 20 Figure 2.8 Distribution patterns of selected conifer taxa in south-eastern Tibet, western Sichuan and northwestern Yunnan Province (from Franzel et al., 2003)...... 21 Figure 2.9 Photographic view of the vegetation types along the elevation gradient of the east slope of the Meili Snow Mountains (Photos: Bräuning, 2004). (a) Sub-arid vegetation in the Lancang River valley, juniper patches are protected by a Buddhist monastery, inlet shows a juniper stem decorated with religious symbols; (b) pine forest at about 2700~3000 m a.s.l.; (c) mixed oak forest at south-facing slopes and spruce forest in the valleys; (d) fir forest near the glacier at 3500 m a.s.l...... 22 Figure 2.10 View of two upper tree line sites at Baima pass and Daxueshan pass in the central Hengduan Mountains (Fhotos: Fan, 2006)...... 23 Figure 2.11 Examples of forest disturbances and land use history in the central Hengduan Mountains (Fhotos: Fan, 2005; 2006). (a) Commercial wood transport of Torreya yunnanensis from subalpine forests at east of the

List of figures VIII

Nujiang River; (b) land use penetrates into the alpine forest area at the Haba Snow Mountains; (c) vegetation after forest fire occurred in 1987 at the Daxueshan Mountains...... 24

Figure 3.1 Locations of the tree-ring sampling sites and meteorological stations in the central Hengduan Mountains, southwestern China...... 27 Figure 3.2 Classification dendrogram of tree ring-width site chronologies based on hierarchical clustering of Ward’s method. The symbols of tree species are identical as in Figure 3.1...... 32

Figure 4.1 Map of selected tree-ring sites and meteorological stations in the central Hengduan Mountains. Square highlights the area covered by sixteen gridded climate data points of CRUts2.1 (0.5 × 0.5, Mitchell and Jones, 2005)...... 37 Figure 4.2 Eight tree ring-width residual chronologies. The site codes are consistent with those of Table 4.1. Tree-ring series are smoothed with a 15–year cubic spline (thick lines)...... 40 Figure 4.3 Relative positions of the eight ring-width chronologies according to the three significant factors resulting from principal component analysis over the period 1850–1999. The explained variances of the factors are indicated in brackets. The site codes are consistent with those given in Table 4.1. Groups of similar chronologies are indicated by identical symbols...... 41 Figure 4.4 Correlation between the four fir ring-width residual chronologies and regional monthly mean

maximum (Tmax), mean (TEM), minimum temperature (Tmin), precipitation (PRE) and Palmer drought severity index (PDSI). The correlation coefficients were calculated from previous year July to current year October over the common period 1951–2002. The horizontal dashed lines indicate the 95% significance level...... 43 Figure 4.6 Biplot of the redundancy analyses (RDA) calculated from the eight residual chronologies and the monthly climate parameters for the period 1951–2002. Significant (p < 0.05) climate factors are indicated by vectors (black arrows); the longer the vector the more important the climate parameter. The correlation between the variables is illustrated by the cosine of the angle between two vectors. Vectors pointing in nearly the same direction indicate a high positive correlation, vectors pointing in opposite directions have a high negative correlation, and vectors crossing at right angles are related to a near zero correlation (Legendre and Legendre 1998). P = Precipitation, T = Temperature, Tn = Minimum temperature, (t-1) = year before ring formation, numbers represent months (e.g. 3 = March)...... 45 Figure 4.7 Comparisons of (a) PC#1 with winter (prior November to February) mean temperatures, (b) PC#2 with summer (June to August) mean temperatures and (c) PC#3 with spring (March to May) PDSI. Values are adjusted for their mean and standard deviations. Bolded lines are smoothed with a 10-year low-pass filter. R indicates the linear correlation coefficients between PCs and climatic variables over the period 1951–1999 and p denotes their significant levels...... 46

Figure 5.1.1 Locations of the sample sites and meteorological stations...... 49 Figure 5.1.2 The four residual chronologies from the central Hengduan Mountains. The sample depths through time are shown in the lower sections of each graph. Pink lines represent annual values; blue lines are 15-year cubic smoothing splines...... 51 Figure 5.1.3 Correlation (columns) and response (dot-lines) functions coefficients between radial growth and regional monthly mean temperature (left) and total monthly precipitation (right). Correlations are computed from previous year May to current year October over 1958–2003. Horizontal dashed lines denote the 95% significance level of the correlation. Asterisks denote significance (p < 0.05) of response function based on bootstrapping tests...... 52

List of figures IX

Figure 5.1.4 Correlation (columns) and response (dot-lines) functions coefficients between PC#1 of four spruce

chronologies and the regional monthly mean (TEM) (a), minimum (Tmin) (b), maximum temperature (Tmax) (c) and precipitation (PRE) (d) from previous year May to current year October over the common period 1958– 2003. The horizontal dotted and dashed lines donate the 99% and 95% significance level for the correlation function, respectively. Asterisks denote significance (p < 0.05) of response function based on bootstrapping tests...... 53 Figure 5.1.5 (a) Comparison of the actual (black line) and reconstructed (grey line) annual (previous October through current September) mean temperature for the common period 1959–2003. (b) Reconstructed annual temperature in the central Hengduan Mountain over the past 250 years. The thin line represents the annual value and the thick line was smoothed with an 11-year FFT-filter (Fast Fourier Transform) to emphasize long- term fluctuations. The horizontal grey line is the instrumental regional mean temperature for period 1959– 2003...... 55 Figure 5.2.1 Map showing the locations of wood densitometry sampling sites and meteorological stations. Bold letters show the locations of the tree-ring based temperature reconstructions from the south (STP) and eastern (ETP) Tibetan Plateau (Bräuning and Mantwill, 2004), the upper source region of Yangtze River (SRY; Liang et al., 2008) and west Sichuan Plateau (WS; Shao and Fan, 1999)...... 56 Figure 5.2.2 Photographic views of the two sample sites of Hongpo (a, HP_P) and Bitahai (b, BT_P) in the central Hengduan Mountains, south-western China. Photos: Fan, Z.X. (2006)...... 57 Figure 5.2.3 Comparison between the HP_P (blue) and BT_P (red) ring-width residual chronologies. (a) Expressed population signal (EPS) statistic (calculated over 30 years lagged by 15 years); (b) the two residual chronologies adjusted to the same mean and variance over the common period 1688–2005; (c) the sample depth through time; (d) the 15-year low-pass filtered components. The correlation coefficients were calculated for 1780–2005 when EPS exceeds the threshold of 0.85 for both chronologies...... 59 Figure 5.2.4 Comparison between the HP_P (blue) and BT_P (red) maximum latewood density residual chronologies. (a) Expressed population signal (EPS) statistic (calculated over 30 years lagged by 15 years); (b) the two residual chronologies adjusted to the same mean and variance over the common period 1724–2005; (c) the sample depths through time; (d) their 15-year low-pass filtered components. The correlation coefficients were calculated for 1820–2005 when EPS exceeds the threshold of 0.85 for both chronologies...... 60 Figure 5.2.5 Climate response of ring width (TRW) for the sites HP_P (white), BT_P (grey) and their regional chronology (RC, black) using (a) maximum temperatures (Tmax), (b) mean temperatures (TEM), (c) minimum temperatures (Tmin), and (d) precipitation sums (PRE). Correlations were calculated from previous year July to current year October over 1958–2004 common periods. Horizontal dashed lines denote the 95% significance levels. Numbers on the x-axis refer to seasonal means of prior November-February (-11/2), prior November-April (-11/4), prior October-September (-10/9), and prior November-October (-11/10) ...... 61 Figure 5.2.6 Climate response of maximum latewood density (MXD) for the sites HP_P (white), BT_P (grey) and their regional chronology (RC, black) using (a) maximum temperatures (Tmax), (b) mean temperatures (TEM), (c) minimum temperatures (Tmin) and (d) precipitation sums (PRE). Correlations were calculated from previous year July to current year October over 1958–2004 common periods. Horizontal dashed lines denote the 95% significance levels. Numbers on the x-axis refer to seasonal means of April-May (4/5), July- September (7/9), April-September (4/9), April-October (4/10), April-September but excluding June (4/9*), respectively...... 62 Figure 5.2.7 (a) Comparison between actual and estimated mean warm season (April to September) temperature for their common period 1958–2004; (b) warm season temperature reconstruction for the central Hengduan Mountains derived from maximum latewood density. Thin line represents annual values; the bold line was smoothed with a 15-year low pass filter...... 65 Figure 5.2.8 Spatial correlations of (a) instrumental and (b) reconstructed summer (April-September) temperatures with regional gridded April-September temperatures for the period 1958–2001. The analyses were performed using the KNMI climate explorer (Royal Netherlands Meteorological; http://climexp.knmi.nl),

List of figures X

the gridded climate dataset was developed by the Climatic Research Unit (Mitchell and Jones, 2005; CRUts 2.1)...... 66 Figure 5.2.9 Graphical comparison of various temperature reconstructions for the western China derived from tree-ring records. (a) warm season (April-September) mean temperature reconstruction in the central Hengduan Mountains (CHM; this study); (b) late-summer (August-September) reconstruction in the southern Tibetan Plateau (STP; Bräuning and Mantwill, 2004); (c) late-summer temperature reconstruction in the eastern Tibetan Plateau (ETP; Bräuning and Mantwill, 2004); (d) summer (June to August) minimum temperature reconstruction in the upper source region of Yangtze River (SYR; Liang et al., 2008); (e) winter half year temperature reconstruction from west Sichuan Plateau (WS; Shao and Fan, 1999). All series were adjusted for their long-term means over period 1750–1994, and smoothed with a 15-year low-pass filter to emphasize long-term fluctuations...... 67

Figure 6.1 Map of the sampling sites in the central Hengduan Mountains, south-western China...... 69 Figure 6.2 Photograpic views of the tree-ring sampling sites at Yanmen (YE_T and YM_A; a) and Tacheng (GK_P and YC_T; b) in the central Hengduan Mountains. Photos: Fan Z.X. (2006)...... 70 Figure 6.3 The four residual chronologies from Baima Snow Mountains, NW Yunnan. The expressed population signal statistic (EPS) through time is shown in the upper section of each graph. Thin lines represent annual values; bold lines were smoothed with an 11-year-FFT-filter (Fast Fourier Transform) to emphasize long-term growth variations. The site codes match with those of Table 6.1...... 71 Figure 6.4 Correlation coefficients between radial growth and monthly mean temperature (dot-lines) and total monthly precipitation (columns) at nearby meteorological stations of Deqin (gray) and Weixi (white). Correlations are computed from previous year July to current year September over 1957–2000 for Deqin and 1961–2000 for Weixi. Horizontal dashed lines denote the 95% levels of significance...... 73 Figure 6.5 Correlation (columns) and response function (dot-lines) coefficients between the first principal component (PC#1) of the residual chronologies and regional monthly a) total precipitation, b) mean temperature, c) relative humidity for 1961–2000, and d) PDSI data for 1951–2000. The coefficients were calculated from previous year July to current year September. Horizontal dashed and bold lines denote the 95% and 99% significance levels, respectively. The asterisks indicate the 95% significance level for response functions based on bootstrapping tests...... 74 Figure 6.6 Actual (grey) and reconstructed (dark) March to May PDSI during their common period 1951–2000. The estimation explains 42% of the actual PDSI variance in this common period...... 76 Figure 6.7 a) Average cross-chronology correlation (mean is 0.47) of the four residual chronologies, calculated for 30–year periods with 15–year overlaps. b) The reconstruction of March-May PDSI in the central Hengduan Mountain region over the past 350 years. The thin line represents the annual value and the thick line was smoothed with an 11–year FFT-filter (Fast Fourier Transform) to emphasize long-term fluctuations. The grey line indicates the ±2SD values...... 76 Figure 6.8 Multi-Taper method (MTM) power spectra of the reconstructed spring PDSI. The bold line indicates the null hypothesis; the dashed, dash-doted, and dotted lines indicate the 90%, 95% and 99% significance levels, respectively. Numbers associated to peaks indicate the periodicity of the signals...... 77 Figure 6.9 Spatial correlations of (a) actual and (b) reconstructed spring (March-May) PDSI with concurrent GPCC gridded precipitation data set for the period 1951–2000. The analyses were performed using the KNMI Climate Explorer...... 78 Figure 6.10 Spatial correlation of reconstructed PDSI with global sea surface temperature data set (ERSST.v2) during prior winter (a, Nov-Feb) and pre-monsoon (b, Mar-May) seasons for the period 1951–2000. The analyses were performed using the KNMI Climate Explorer (http://climexp.knmi.nl)...... 79

List of figures XI

Figure 7.1 Example of wood structure (upper graph) and corresponding maximum latewood density profile (lower graph) of Larix potaninii at Baima site. The density data was produced by the Lignostation densitometry system (Rinntech, Germany)...... 82 Figure 7.2 (a) Standardized maximum latewood density chronology of Larix potaninii at Baima site, dashed line indicates 2 standard deviations; (b) Sample depths (line) and percentage (bars) of larch trees shown reduction of MXD as identified by program OUTBREAK without any corrections; (c) Comparison of TRW chronologies between the host (BM_L, Larix potaninii) and non-host (BM_A, Abies georgei) species at the Baima site; (d) Sample depths (line) and percentage (bars) of larch trees with ring-width growth reduction as identified by program OUTBREAK (bars) controlled by the non-host (Abies georgei) master chronology. Shade area line indicates the 30% threshold...... 84 Figure 7.3 (a, c, e) Ring-width chronologies of larch (blue lines) and non-host species (orange lines) at three sites in the central Hengduan Mountains; (b, d, f) Sample depths (lines) and percentages (bars) and of larch trees with ring-width reduction as identified by program OUTBREAK controlled by non-host chronologies. Shaded area indicates the 30% threshold...... 85 Figure 7.4 Multi-taper method (MTM) power spectra of the ring-width standard chronologies at the Baima (a, b), Bitahai (c, d), Haba (e, f) and Yulong (g, h) sites. The blue line indicates the null hypothesis, the cyan, green and red lines indicate the 90%, 95% and 99% significance level, respectively...... 87

Figure 8.1 Comparison of annual temperature, summer temperature and spring PDSI reconstructions in the central Hengduan Mountains. The series are rescaled to z-scores referred from their long-term means, and smoothed with a 15-year low-pass filter...... 93 Figure 8.2 Photographic view of the mixed forest site at the northern edge of Gaoligong Mountains. Inlet shows a tree stem of Taiwania flousiana Gaussen. Photos: Fan Z.X. (2005)...... 95

List of tables XII

List of Tables

Table 2. 1 Locations and characteristics for meteorological stations from the central Hengduan Mountains. Bold names indicate climate station used for tree growth calibration. For locations, see also Figure 3.1 ...... 16 Table 2. 2 Correlation matrix of mean annual temperature (upper right) and total annual precipitation (lower left) among six meteorological stations during the 1961–2004 (N = 44). * Significant at p < 0.05, ** Significant at p < 0.01...... 17

Table 3. 1 Locations and characteristics of the tree-ring network in the central Hengduan Mountains, south- western China...... 28

Table 4. 1 Site locations and characteristics...... 38 Table 4. 2 Descriptive statistics of the eight total ring width chronologies...... 39 Table 4. 3 Correlation coefficients of the eight residual chronologies for the well replicated period 1850–1999. * Significant at p < 0.05; ** Significant at p < 0.01...... 39 Table 4. 4 Correlation coefficients between eight residual chronologies and seasonal climatic variables of monthly mean temperature (TEM), precipitation (PRE) and Palmer drought severity index (PDSI) for the period 1951– 2002 (n = 53)...... 44 Table 4. 5 Summary of redundancy analysis statistics ...... 44

Table 5.1. 1 Site characteristics and chronology statistics...... 50 Table 5.1. 2 Pearson correlations among the residual chronologies for the period 1850–2003. All correlation are significant at the p < 0.05 level...... 50 Table 5.1. 3 Eigenvalues of principal component analysis of the four spruce residual chronologies for the period 1750–2003...... 51 Table 5.1. 4 Statistics of the leave-one-out calibration results for the common period 1959–2003...... 54

Table 5.2. 1 Site information and tree-ring chronologies statistics...... 58 Table 5.2. 2 Statistics of calibration and leave one-out verification results for the common period 1958–2004. ... 64

Table 6. 1 Site locations and chronology statistics...... 72 Table 6. 2 Correlation matrix of the four ring-width chronologies for the common period 1655–2005. The site codes are identical with those in Table 6.1. All correlations are significant at p < 0.01...... 72 Table 6. 3 Statistics of calibration-verification test results for the common period 1951–2000...... 75

1. Introduction and research aims 1

1. Introduction and Research Aims

1.1 Climate Change

Climate change is currently an issue of scientific challenge and political debate. According to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2007), global surface temperature has been increasing 0.74 ±0.18 ºC during the 20th century and is projected to rise 1.8~4.0 ºC in the 21st century (Figure 1.1; Brohan et al., 2006). Climate warming is also evident from observations of widespread melting of snow and ice and the rising of the global average sea level. Although any human-induced changes in climate will be embedded in a background of natural variations, most of the observed increase in global average temperature since the mid-20th century is very likely due to the reported increase in anthropogenic (man-made) greenhouse gas concentrations (IPCC, 2007). Series of annual mean temperature anomalies for the period 1880–2003 for ten climatic regions of China were established (Wang and Gong, 2000; Wang et al., 2004). The temperature series for the whole of China was obtained by averaging over the ten regions considering the regional weights, which are proportional to the size of each (Figure 1.1). During the periods 1880–2003, the linear trend of mean temperature in China is 0.58 °C/100yr, which is a little less than the global mean. Positive anomalies over China during the 1920s and 1940s are noticeable (Wang et al., 2004). During the last two decades, the warming trend (0.3 ºC/10yr) over China is much stronger than the trend of global mean temperature (0.19 ºC/10yr) (Wang and Gong, 2000). Annual mean temperature of China in 1998 was 1.38 ºC higher than normal. 1998 was the warmest year since A.D. 1880 in China and is in agreement with the estimation of the global mean temperature (Wang et al., 2001). Knowledge of past climate is required in order to assess the range of natural variability of the climate system and to evaluate the role of anthropogenically induced changes over recent centuries. However, instrumental and historical records of past climate are usually only available for the most recent century or even shorter (about 50 years) in western China. The inference of meaningful long-term climatic information requires the rigorous interpretation of proxy data (Bradley and Jones, 1992; Bradley et al., 2003). Long proxy-based climatic series are important for evaluating the amplitudes of natural fluctuations of the climate system and to place recent warming in a long-term context (Crowley, T.J., 2000; Esper et al., 2002). Proxy data, such as time-series of the thickness and composition of lake sediments, glacial ice layers, growth increments in corals, speleothems and trees, and documentary evidence are the key to the understanding and quantification of past climatic variations (Jones and Mann, 2004; Moberg et al., 2005). Data from these archives are generally compared and correlated with instrumental measurements to quantify their climatic sensitivity and signal, and are subsequently used to document climate prior to the period of instrumental data.

1. Introduction and research aims 2

1.2 Tree Rings as Climate Proxies

Tree rings provide important proxy data for paleo-environmental studies and reconstructions of various climate elements (Bradley and Jones, 1992; Luckman, 1996). In particular, the last millennium is considered as a suitable time interval for the study of the natural background variability in relation to climate change detection and an appropriate period in relation to the longevity of several tree species (Mann et al., 1998; Mann and Bradley, 1999; IPCC, 2007). The major strengths of tree rings as climatic proxies are i) annual resolution, verifiable by cross- dating, ii) the existence of large-scale geographic patterns of synchronous inter-annual variability, iii) the increasing availability of extensive networks of tree-ring chronologies covering large parts of the northern hemisphere outside of the tropics, and iv) the possibility of using simple linear models of climate-growth relationships that can easily be verified and calibrated (Hughes, 2002). Their weaknesses include: i) an intrinsic sampling bias, given that tree-ring information is available only for terrestrial regions of the globe, ii) the fact that methods used to extract growth signals from tree-ring series retain only certain wavelengths of climate variability (IPCC, 2007), iii) the complexity of biological responses to climate forcing, and iv) the presence of non- climatic variability in the series attributable to intrinsic growth trends and other non-climatic disturbances (Fritts, 1976).

1. Introduction and research aims 3

1.3 Tree Growth and Climate

1.3.1 Tree-Ring Formation

Radial growth of trees is composed of annual increments of xylem in the stem. Besides the tropical belt (with exceptions of certain higher elevation and arid regions), where no periodic dormancy in plant growth exists, the disruption of the growing season by winter is consistent and strong enough to induce clearly discernible layers of increment. Changes in anatomical features of the wood cells formed at various times throughout the growing season generally allow annual rings to be defined. Most conifers form rather large cells with thin cell walls (earlywood) during the earlier growing season, whereas tend to form smaller cells with thicker walls (latewood) during the later growing season (Figure 1.2; Schweingruber, 1996). This difference between earlywood and latewood will create the annual tree rings (Fritts, 1976). Tree rings thus reflect the annual growth cycle of the trees which in turn is largely influenced by the environmental circumstances during the growing processes.

Figure 1.2 Schematic overview on wood structure and tree-ring parameters of Picea brachytyla, produced by the Lignostation densitometry system (Rinntech, Germany).

1. Introduction and research aims 4

1.3.2 Dendroclimatology

Dendrochronology is the science of dating annual growth layers (rings) in woody plants (Fritts, 1971). The common practice of dendrochronology is the synchronization of annual growth increments among the samples from one tree, then among several trees of one stand, and lastly among regional dataset of samples. This technique is called cross-dating and is the most characterizing feature of dendrochronology (Fritts, 1976). The pattern formed over time by the differently sized of tree rings each year, particularly in trees from the same site, can be matched between trees. The visual matching of tree-ring patterns along with some statistical analysis allows dendrochronologists to eliminate any dating errors and to produce series of ring-width measurements which are exactly dated. The “cross-dating” of ring width measurements enables the determination of precise calendar dates for each growth ring (Stokes and Smiley, 1996). Dendroclimatology studies the relationship between the variations of tree-ring characteristics and climate over a common period. This relationship can be used to estimate the values of climatic variables in periods prior to available instrumental climate data. Dendroclimatological examination is therefore an empirical test for the detection of climate variables bearing significant impact for tree-ring variability. Ideally, the strength and the sign of the variables can be determined. If a strong co-variation between climate and tree-rings is evident and relatively stable over time, a statistical equation for the relationship can be successfully established and verified (Fritts, 1976; Cook and Kairiukstis, 1990). However, climate is not the only factor influencing tree-ring growth, in many cases it is not even the most significant factor, but its signal can be strengthened in the processes of tree-ring data analysis by tree-ring standardization (Cook, 1985). Dendroclimatological studies maximize the climatic signal through careful selection of sampling site, large sample replication and statistical decomposition of tree-ring variability (i.e. standardization), which will be discussed in chapter 3.

1.3.3 Climate-Growth Responses

Climate-growth relationships are the basis for tree-ring based climate reconstructions (Briffa et al., 1998a; Tessier et al., 1997). Furthermore, tree-ring growth responses to climate change provide crucial information to assess future forest productivity, vegetation dynamics and tree species distributions (e.g. Peterson and Peterson, 2001; Saxe et al., 2001; Frenzel et al., 2003; Thuiller et al., 2005; Tardif et al., 2006). The physiological mechanisms by which climatic parameters are translated into radial growth variations are complex, because radial growth in any given year integrates the effects of climate conditions during and prior to growth, local site conditions and physiological differences of tree species (Fritts, 1976). In different biogeographic zones, changes in temperature, precipitation and radiation are known to limit tree growth in different degrees (Churkina and Running, 1998). At local scale, local factors such as aspect, elevation, wind exposure, substrate, animal and human influence all can influence annual ring formation. For example, in mountain regions, radial growth of trees

1. Introduction and research aims 5 from high elevations generally reflects temperature variations, whereas growth rates of trees from lower elevations generally mirror precipitation changes (La Marche, 1974; Schweingruber, 1996). Similarly, the temperature signal is maximized at the temperature limits of tree growth at the northern high latitudes (Briffa et al., 1998b). However, because of the co-variation of various climatic parameters and complex plant physiological reactions and processes, attempts to define growth responses in terms of a single controlling factor often fail (Fritts, 1976). For example, during the relatively short vegetation period during summer, even at cold sites, water availability becomes the dominating factor for tree growth, once temperatures are high enough to allow any growth at all (e.g. Anfodillo et al., 1998; Carrer et al., 1998).

1.3.4 Tree-Ring Parameters

In addition to the commonly used total ring width, variations in the width of earlywood/latewood, maximum latewood density or stable isotopic composition of tree rings can be measured and cross-dated. Earlywood and latewood contain distinct signals (McCarroll et al., 2003): earlywood cellulose produced from starch or similar stores synthesized during the previous year’s growth carries a signal from the previous year, while latewood carries a signal from the current year. Here, total ring width (TRW) and maximum latewood density (MXD) are used for the reconstruction of past environmental conditions. TRW is an expression of cell numbers and cell-enlargement, and may reflect a variety of environmental conditions during and/or prior to ring formation (Fritts, 1976). The use of networks of chronologies sampled from both northern latitudinal and upper tree-line environments has allowed the development of high quality regional temperature reconstructions from many regions in the Northern Hemisphere (D’Arrigo and Jacoby, 1993; Jacoby et al., 1996; Mann et al., 1998; Esper et al., 2002; 2003a; Briffa et al., 2004; Frank et al., 2007; Büntgen et al., 2008). Tree-ring density measurements reflect cell size and cell-wall thickness (Schweingruber et al., 1978). An increase of density is due to a decrease in radial cell diameter and/or to an increase in cell wall thickness (Heger et al., 1974; Yasue et al., 2000). Compared with ring width, measurement of MXD data is much more expensive and time-consuming. However, the advantage of MXD records is that the common signal between trees/sites is generally stronger than among ring-width series (e.g. Wilson and Luckman, 2003). MXD series from many conifer tree species growing in cool-moist regions generally correlate strongly with spring and late summer temperature variability (Schweingruber and Briffa, 1996). Based on tree-ring width and maximum latewood density data, high quality long-term climate variations have been successfully reconstructed from e.g. Fennoscandia (Briffa et al., 1988; Grudd, 2008), North America (Briffa et al., 1992; Wang L., 2001; Davi et al., 2003; D’Arrigo et al., 2004; Luckman and Wilson, 2005), Siberia (Briffa et al., 2001; Kirdyanov et al., 2008), European Alps (Frank

1. Introduction and research aims 6 and Esper, 2005; Büntgen et al., 2006), high Asia (Hughes, 2001; Bräuning and Mantwill, 2004) and the northern hemisphere (Briffa et al., 2004; Mann et al., 2008).

1.4 Climatological Relevancy of the Tibetan Plateau

The mountain regions of western China and the Tibetan Plateau (TP) are an area which is especially sensitive to climatic change (Zheng, 1996; Liu and Chen, 2000). The TP, has an average elevation above 4000 m a.s.l. and an extension of more than 2 million km2 (Figure 1.3), which acts as a heating surface during spring and summer. Thus, the TP is a climatically very important region due to its influence on large-scale atmospheric circulation patterns over Asia, including the Asian summer monsoon (Murakami, 1987; Ding, 1992; Webster et al., 1998). Analyses of temperature records from 78 climate stations above 2000 m a.s.l. from the TP and adjacent mountain areas have revealed positive seasonal temperature trends for summer (JJA) and winter (DJF) between 0.09 ºC and 0.32 ºC per decade for the period 1955–1996 (Liu and Chen, 2000). The TP represents a huge heating surface in subtropical latitudes, which plays a key role in driving the Asian summer monsoon circulation (Fu and Fletcher, 1985; Li and Yanai, 1996; Webster et al., 1998). As a consequence of global warming, the contrast in summer temperature between the Asian continent and the surrounding oceans is increasing. Consequently, global circulation models predict an increase in summer monsoon rainfall (Meehl and Washington, 1993; Hulme et al., 1994; Böhner and Lehmkuhl, 2005). Climate conditions (i.e. temperature, snow cover) over the TP also affect the intensity of the summer monsoon circulation and the rainfall distribution in the adjoining forelands (Barnett et al., 1988; Kripalani et al., 2003). Comprehensive knowledge about the past monsoon behaviour can provide deeper insight into possible future circulation patterns. The TP is affected by Indian and East Asian summer monsoons from the southwest and southeast, respectively, by westerly airflow in both summer and winter. Regional climate conditions over the TP are far from uniform but vary considerably according to topographic conditions and exposure to moisture bringing winds. Stable isotope of precipitation from different climatic regions over TP shows that the northern limit of the summer monsoon is north of the Yalongzangbo river into the middle of the Tibetan Plateau around 34 ~ 35 °N (Figure 1.3; Araguas-Araguas et al., 1998; Tian et al., 2007). The Tanggula Mountains seem to mark a climatic threshold: to the north of this range, precipitation from monsoon is largely replaced by that derived from westerly depressions, or by continental water recycling (Tian et al., 2001; Wang, 2006). Temperature reconstructions derived from ice cores (Yao et al., 2006a) and multiple archives (Yang et al., 2003) indicate that decadal-scale temperature fluctuations occurred asynchronously in different part of the vast TP during the last millennium. Understanding of spatial and temporal climate variations over the TP is thus vital to improve out knowledge of Asian monsoon dynamics. To relate these complex spatio-temporal differences of paleoclimate to atmospheric

1. Introduction and research aims 7 circulation patterns that are the driving factors of climate variability, a dense network of past climate information has to be established. However, meteorological stations in the TP were not installed before 1950. In addition, the spatial distribution of the stations is sparse. This limits our ability to examine current climate regimes in a long-term perspective. Therefore high-resolution proxy data like tree rings are needed to shed more light on the climate history of the TP.

Figure 1.3 Topographical view of the Tibetan Plateau and surrounding areas. Approximate present geographical extent of Southwest Asian and East Asian summer monsoons are indicated in dashed lines (redraw after Araguas-Araguas et al., 1998; Böhner and Lehmkuhl, 2005).

1.5 Dendroclimatology on the Tibetan Plateau

In recent years, considerable efforts have been made to reconstruct past climate change on the TP over the past millennia. Climate proxy data were derived from ice cores, tree rings, lake sediments, groundwater profile and glacial geomorphology over TP (Holmes et al., 2008). However, each proxy has its strengths and limitations with respect to dating accuracy, resolution,

1. Introduction and research aims 8 and accurancy of interpretation. During the past decades, a growing body of tree-ring based poleoclimatic reconstructions were derived over TP, especially in the eastern part of the TP. Growth-climate relationships in different climatic regions over the TP and across environmental gradients were evaluated (e.g. Bräuning, 1994; 2001a; Zhang et al., 2003; Shao et al., 2005; Liang et al., 2006a; 2008). In the semi-arid cold regions of the northeastern TP, radial growth of Sabina przewalskii and Picea crassifolia are mainly influenced by late spring to early summer (April-June) precipitation. This reaction pattern has been found in the Qilian Mountains (Gou et al., 2005; Tian et al., 2007), in the Dulan area (Zhang et al., 2003; Sheppard et al., 2004; Shao et al., 2005; Li et al., 2008a) and in the Anyemaqen Mountains (Gou et al., 2007a). However, the effects of spring or summer precipitation on radial growth of these species vanish at high elevation sites (Liu et al., 2006; Peng et al., 2008). The same holds true for north and east-facing slopes (Liang et al., 2006a) where higher amounts of moisture for tree growth are available than on south-facing slopes. Trees growing in the cold–moist environment near the upper treeline in eastern Tibet are sensitive to temperature variations during the winter as well as during the summer season (Shao and Fan, 1999; Bräuning, 2001a; Liang et al., 2008). In the last decades, the knowledge of past climate changes in the TP derived from dendroclimatic studies has drastically increased. A number of century- to millennial-long tree- ring chronologies have been constructed from the TP. In the Dulan area of north-eastern TP, Kang et al. (1997) developed an 1835 years ring-width chronology from living Qilian juniper (Sabina przewalskii). Zhang et al. (2003) found that annual radial growth of Qilian juniper mainly reflects variations of regional spring precipitation. Spring moisture availability inferred from Dulan tree-rings was extended back to B.C. 326. Annual precipitation was reconstructed over the past 2500 years from composites of living trees and archaeological juniper wood (Sheppard et al., 2004). Liu et al. (2006) reconstructed annual precipitation for the Dulan area since A.D. 850. Their reconstruction highlighted that low-frequency fluctuations of precipitation are associated with regional temperature changes on a decadal- to century-scale, and revealed a warm-wet and cold-dry climatic pattern over the north-eastern TP. Based on a network of 11 Qilian juniper ring-width chronologies, Shao et al. (2005) developed an annual precipitation reconstruction for the nearby region of Delingha and Wulan, which covered the past millennium and accounted for 66% of the instrumental climatic variance for the 1955–2002 period. Methodological improvements have also been made for cross-dating old juniper trees (c.a. 1000 years), which form many missing rings in the arid or sub-arid area of western China (Shao et al., 2003). A 3500-year tree-ring master chronology, so far the longest chronology in China, was constructed using well-replicated samples from living, dead trees and archaeological wood in the Dulan area (Shao et al., 2007). Dendrochronological studies were conducted in the Qilian Mountain region since the 1970s. Zhuo (1981) sampled Qilian juniper and Qinghai spruce in the Qilian Mountains. Zhang and Wu (1997) constructed a 700-year ring-width chronology. Their analyses with meteorological data indicated that radial growth of Qilian juniper correlates positively with

1. Introduction and research aims 9 spring precipitation and negatively with spring temperature. A dry/wet index was then reconstructed for the Qilian Mountains since A.D. 1300. Gou et al. (2001) concluded that spring precipitation is the main limiting factor for radial growth of Qilian junipers in the central Qilian Mountains, and compared their reconstruction of spring precipitation with glacier fluctuation records from the Qilian Mountains. Based on a regional chronology of Qilian juniper, 1319 years of runoff variability were reconstructed for the Heihe River (Kang et al., 2002). Liu et al. (2005) found that Qilian junipers growing near the upper tree line are mainly influenced by winter temperature variations. Temperature variability in the Qilian Mountains was reconstructed for the past millennium and its low-frequency variations were compared with northern hemisphere temperature reconstructions. During recent years, considerable progress has also been achieved in the south of the Qinghai plateau and the Anyemaqen Mountains region. Qin et al. (2003) developed a network of Sabina tibetica which covers the past 500 years, and reconstructed a spring dry/wet index. Based on four ring-width chronologies of Balfour spruce (Picea balfouriana), Liang et al. (2008) reconstructed summer minimum temperature since A.D. 1624 for the upper source region of the Yangtze River. Peng et al. (2008) analyzed growth-climate responses of Qilian juniper along altitude gradients, and revealed that radial growth of Qilian juniper is mainly limited by spring precipitation at low elevations, however, the importance of spring/summer precipitation on radial growth vanishes at high elevation sites. Gou et al. (2007b) reconstructed winter half-year (October-April) minimum temperature for the Anyemaqen Mountains region, which shows a strong warming trend (0.5 ºC/10yr) in the late 20th century that clearly exceeds the range of temperature variability over the previous 400 years. Meanwhile, they also reconstructed summer half-year maximum temperature using independent tree-ring data for the same region (Gou et al., 2008a), and revealed the asymmetric variability between maximum and minimum temperatures in northeastern TP (Gou et al., 2008b). In the southern and south-eastern Tibetan Plateau, Wu and Lin (1981) reconstructed temperature changes during the past 2000 years in Lhasa. Bräuning (1994; 2001b) developed tree-ring networks coving 28–34 ºN, 87–103 ºE, thus a wide range of different climatic conditions from the moist eastern and southern fringes of the TP to the semiarid part of southern central TP. Five dendroecological regions or growth provinces were defined in the south-eastern TP, according to multivariate analyses on tree-ring growth patterns and growth-climate interactions (Bräuning, 2000). Summer monsoon intensity was reconstructed during the past millennium based on ring-width, maximum latewood density and carbon isotope data. During the past millennium, intensification and weakening of summer monsoon were found to be associated with major climatic shifts during the so-called ‘Medieval Warm Period’ (AD 1150- 1400) and ‘Little Ice Age’ (AD 1400-1900) (Bräuning and Grießinger, 2006). Based on a network of 22 MXD chronologies of high elevation conifer sites, late summer (August- September) temperature over the past 400 years was reconstructed for the southeastern Tibetan Plateau (Bräuning and Mantwill, 2004). Glacier fluctuations during the ‘Little Ice Age’ were

1. Introduction and research aims 10 studied by determining the ages of trees growing on glacier deposits in eastern Tibet (Bräuning, 2006). Liu et al. (2003) discussed climatic signals of carbon isotopes in annual rings of Abies spectabilis, and reconstructed dry/wet histories over the past 350 years for the Nyingchi area of southeastern Tibet. Recently, annually resolved δ13C- and δ18O-isotope chronologies of tree-ring cellulose were constructed along a hydrological gradient from south-eastern Tibet to the central TP (Helle et al., 2002; Grießinger et al., 2008).

1.6 Dendroclimatology in the Hengduan Mountains

The north-south oriented Hengduan Mountains form the southern rim of the Tibetan Plateau, and are strongly exposed to the southwest Asian monsoon. The region is extremely rich in conifer species and is an important refuge area where conifers survived in the harsh conditions during the cold climate stages during the Pleistocene, when Tibet was free of forest (Frenzel et al., 2003). Presently, the region is covered by widespread forests dominated by various coniferous species and thus has a great potential for dendroclimatological studies (Wu et al., 1988; 1989; Bräuning, 1994). Despite the widespread forest cover, however, only a few dendroclimatological studies have been conducted in the Hengduan Mountain ranges (Wu et al., 1988; Shao and Fan, 1999; Bräuning, 2001a). Shao and Fan (1999) reconstructed winter minimum temperature for the west Sichuan Plateau since A.D. 1650, using four tree ring-width chronologies from Picea balfouriana. Wu et al. (2006) discussed the growth-climate relationships of five conifer species from West Sichuan Province, and reconstructed summer temperature for the past century using maximum latewood density data of Pinus densata (Wu et al., 2005). Tree-ring sampling sites from the work of Bräuning (1994, 2000) cover the north part of the Hengduan Mountain, but such studies are still scarce at the central and south part of the Hengduan Mountains. Based on tree-ring width data from the upper tree line and lower-altitude arid area, Wu et al. (1988) published a reconstruction of fluctuations of air temperature and annual precipitation during the last 400 years. Some of their sample sites are located within the present study area. Unfortunately, their original raw data are not available, although filtered version of regional chronologies were published. On the other hand, they did not provide information about the strength of the tree growth-climate signal or the reliability of the reconstructions. In addition, their sampling work was conducted during the end of the 1970s, when only very short meteorological records were available (~ 20 years), which did not allow estimations on the strength of the tree growth-climate relationship or the reliability of the reconstruction. Meanwhile, their data does not document environmental changes that occurred during the last two decades, a period of special interest.

1. Introduction and research aims 11

1.7 Objectives and Research Questions

The primary objective of this study was to develop high-resolution climate proxy data to extend the few existing climate records for the central Hengduan Mountains region, and to close a gap of dendroclimatological studies sites so far existing in south-western China. To meet the overall goal of reconstructing climate fluctuations, a series of specific steps were identified: a network of tree-ring chronologies from various conifer species was developed in the central Hengduan Mountains; total ring width (TRW) and maximum latewood density (MXD) were processed to build chronologies which contain regional climatic variability of different seasonal signals; climate data (temperature, precipitation and Palmer drought severity index) were assembled for the central Hengduan Mountains; regional climatic series were developed for better spatial representation and for the improvement of calibration models. Under the scope of reconstructing environmental and climatic history during the past centuries, the following questions will be discussed: 1) What are the differences of climate-growth responses of high elevation conifers, with respect of different tree species, habitats and parameters (TRW and MXD)? 2) To which extents are the past climate could be reconstructed from TRW and MXD data, as well as from different geophysical locations (i.e. altitudes)? What are the characteristics of the past temperature fluctuations and drought variability during the past few centuries? 3) What are the contexts of local climate reconstructions within a broad regional scale? How are the spatiotemporal synchronizations and discrepancies between the present reconstructions and other existing tree-ring-based climate reconstructions, as well as climate proxies like ice cores, historical documents and glacier fluctuation records from nearby regions? How is the local drought variability connected to large scale circulation changes? 4) To which extents are local tree-ring series influenced by other environmental factors than climate, i.e. insect infestations, forest fires, geomorphodynamics and others factors?

1. Introduction and research aims 12

1.8 Outlines of this Thesis

This dissertation is divided in ten chapters. Chapter 1 provides an overview of the rationale for this study, introducing tree rings as promising climate proxies in general, summarizing recent achievements by utilizing tree-ring data as climate proxies on the Tibetan Plateau, and presenting the aims of this study. Chapter 2 describes the characteristics of geographical setting, climate conditions and vegetation zonations of the central Hengduan Mountains. Chapter 3 presents a tree-ring network and the sampling strategies. The methods of tree-ring measurement, cross-dating and chronology development are described. Then, statistical methods to extract common ‘climate-like’ signals and to develop reliable climatic reconstructions are explained. Chapter 4 addresses the nature of climatic signals in the spruce and fir ring-width chronologies along an altitude gradient. It contributes to an assessment of the suitability of tree- ring data for climate reconstructions, as well as a better understanding of the impact of recent climatic change on subalpine tree growth. Chapter 5 presents an annual temperature reconstruction developed from four spruce ring- width chronologies near the upper timberline. Two well-replicated maximum latewood density chronologies were used to develop a summer temperature reconstruction. Existing proxy-based temperature reconstruction and glacier fluctuation records from surrounding regions were compared with the present findings. Chapter 6 presents a 350–years spring drought (Palmer drought severity index, PDSI) reconstruction developed from four low-elevation ring-width chronologies. Chapter 7 contributes the preliminary results of the larch insect outbreak history. Chapter 8 consists of a general discussion and conclusions about the findings of this thesis and recommendations for further research

Chapter 9 and 10 provide the summary and reference list, respectively.

2. Physical setting 13

2. Physical Setting

2.1 Location of the Study Area

The Hengduan Mountains (22~32.05° N, 97~103° E) are located in the south-eastern Tibetan Plateau (Figure 2.1). The Chinese name for the geographically complex region is “Hengduan Shan”, which means “traverse cutting mountains”. The mountain ranges run roughly north to south and define the south-eastern edge of the Tibetan Plateau. They converge within a corridor stretching only 90 km from east to west in the central portion of the Hengduan Mountains. The most geographically distinctive part is the “Three River Gorges County” on the frontier between Tibet, Sichuan, and Yunnan, with its tremendous bow-shaped geomorphologic structure. The parallel mountain ranges are separated by deep, narrowly incised major Asian rivers (Figure 2.2): the Jinsha Jiang (upper Yangtze River), Lancang Jiang (upper Mekong River) and Nu Jiang (upper Salween River). In the central portion of the Hengduan Mountains, they converge within a 90 km corridor. The resulting landscape patterns include extreme topographic gradients between deeply incised parallel gorges in about 1500 m a.s.l. and glaciated peaks over 6700 m a.s.l. within a horizontal distance of 20 km or even less.

Figure 2.1 Topographical map of the Tibetan Plateau. Black square highlights the central Hengduan Mountains region.

2. Physical setting 14

Figure 2.2 Photographic views of the three parallel rivers and ridges in the central Hengduan Mountains, south-western China. Photos: Bräuning A., 2004; Fan Z.X., 2006.

The height of the mountain ranges declines toward the south. Eighty percent of land surface areas are above 3000 m a.s.l. in the central portion of Hengduan Mountains. Some mountain peaks are higher than 5000 m, such as the Gaoligong Mountain (5128 m), Meili Snow Mountain (6740 m), Baima Snow Mountain (5476 m), Haba Snow Mountain (5396 m) and Yulong Snow Mountain (Jade Dragon, 5596 m).

2. Physical setting 15

2.2 Climate Conditions

Climatologically, the Hengduan Mountain region is a transition zone between the lowland subtropical climate of Southeast Asia and the highland climate of the Tibetan Plateau. The climate of this region is mainly dominated by the southwest Asian monsoon (SAM) rhythm (Xu et al., 2003), characterized by a seasonal change of the wind system. During winter (November to April), the southern branch of the westerlies influences the western part of this region, causing clear, cold and dry conditions. During summer (June to September), the SAM is the dominant climate factor. The deeply incised gorges of the parallel rivers form passageways for the monsoonal air masses, resulting in a steep moisture gradient from south-east to north-west in southern Tibet (Chang, 1981). Under the driving force of the heated Tibetan Plateau, monsoonal air masses originating from the Bay of Bengal flow north along the river gorges, causing heavy rainfall and cloudy weather (Wen, 1989).

Figure 2.3 Mean air temperature (°C, upper) and precipitation sums (mm, lower) distributions during winter (Nov-Feb) and summer (Jun-Sep) seasons over the Yunnan Province. Meteorological data for 119 stations were obtained from the National Meteorological Information Centre (NMIC) of China. Mean seasonal values were calculated over the period of 1961–2004 and interpolated with the Kriging method.

2. Physical setting 16

The central Hengduan Mountains is located in the northwestern part of Yunnan Province. Figure 2.3 presents the spatial distributions of averaged seasonal air temperature and precipitation over the Yunnan Province for 1961–2004. Temperatures generally decrease along the latitude gradient from south to north. The central part is charactered as high temperatures and low precipitation, which are the main reasons for the development of hot-dry vellays in this area. In the northwest part of Yunnan Province (the present study area), temperature gradients are more abruptly than in other regions. During winter season, the west part of Hengduan Mountains is probably influenced by westerly, which introduce a sharp gradient of precipitation depression from west to east. The meteorological stations in the Hengduan Mountains are sparse and generally located in the valley floors, and were not established until the 1950s. According to meteorological records from the central part of the Hengduan Mountains (Table 2.1; Figures 2.4 and 3.1), mean annual temperatures range from 4.3 to 14.6 ºC, temperatures in January (coldest month) range from -5.4 to 7.5 ºC, while temperatures in July (warmest) range from 12 to 21 ºC. The annual sum of precipitation ranges from 636 mm to 1736 mm, 43%–87% of which is related to the summer monsoon season from June to September. The frost-free period is between 120-200 days. In general, the orographic snowline lies between 4800 and 5200 meters. Due to the topographical complexity, this region displays a wide variety of microclimates. Air temperature decreases from south to north and from low to high elevations (Table 2.1; Figure 2.4). Temperature decrease accounts to ~0.57 ºC per 100 meters elevation increase. Precipitation decreases from west to east and from south to north and increases with increasing elevation (The Editing Committee of “The Baima National Nature Reserve”, 2003). Correlations of annual mean temperature among the stations are generally high, especially for the high- elevation stations (Table 2.2). However, correlations of precipitation among these stations are low and indicate the strong influence of local topography on rainfall distribution.

Station Latitude Elev. Period Ann. Jan. Jul. Total Jun-Sep Name /Longitude (m) Temp. Temp. Temp. Prep. Prec. (°C) (°C) (°C) (mm) (%) Daocheng 29.05/100.03 3729 1957–2004 4.3 –5.4 12.0 653 87 Deqin 28.48/98.92 3320 1957–2004 5.2 –2.5 12.3 644 59 Shangri-la 27.83/99.70 3276 1958–2004 5.8 –3.2 13.5 636 73 Lijiang 26.87/100.22 2393 1951–2004 12.6 6.0 18.1 970 81 Weixi 27.17/99.28 2325 1961–2004 11.5 3.8 18.5 968 54 Gongshan 27.75/98.67 1583 1961–2004 14.6 7.5 21.2 1736 43

2. Physical setting 17

Figure 2.4 Climate diagrams for the six meteorological stations in the central Hengduan Mountains.

Table 2. 2 Correlation matrix of mean annual temperature (upper right) and total annual precipitation (lower left) among six meteorological stations during the 1961–2004 (N = 44). * Significant at p < 0.05, ** Significant at p < 0.01.

Daocheng Deqin Shangri-la Lijiang Weixi Gongshan Daocheng 1 0.86** 0.96** 0.54** 0.85** 0.44** Deqin 0.37* 1 0.93** 0.45** 0.82** 0.44** Shangri-la 0.26 0.72** 1 0.47* 0.85** 0.46** Lijiang 0.56** 0.34* 0.25 1 0.67** 0.43** Weixi –0.03 0.53** 0.71** –0.06 1 0.60** Gongshan –0.13 0.37* 0.42** –0.25 0.49** 1

2. Physical setting 18

2.3 Regional Climate Patterns

Figure 2.5 presents the linear trends of mean air temperatures (°C/10yr) and precipitation (mm/10yr) over the Yunnan Province for the period 1961–2004. A non-parameter Mann-Kendall test was also performed for detecting trends. Over the past 44 years, winter temperatures were mostly increasing over the majority of the whole province. The picture of summer temperatures looks different, most positive trends are found for stations south of ~24 °N. The trends of precipitation are mostly not significant, although summer precipitation show negative trends for most stations, as well as winter precipitation for the area south of ~24 °N.

Figure 2.5 Linear trends of air temperatures (°C/10yr, left) and precipitation sums (mm/10yr, right) during winter (NDJF) and summer (JJAS) seasons over the Yunnan Province for the period 1961–2004. Stations with significant (95%) Mann-Kendall trend test are encircled.

In the central Hengduan Mountains region, the analyses of climate data revealed that during the past fifty years regional mean annual temperatures (TEM) have been increasing significantly by 0.03 °C yr–1, with an increase of 0.028 °C yr -1 and 0.033 °C yr -1 for summer and winter

2. Physical setting 19 season, respectively (Figure 2.6). The warming trend was nearly double the average global warming of 0.019 °C yr–1 during the past half century (Wang and Gong, 2000). However, this –1 warming trend was mainly caused by an increase of minimum temperatures (Tmin, 0.09 °C yr ) instead of maximum temperatures (Tmax). As a result, regional diurnal temperature ranges (DTR) have been decreasing since the early 1980s. These findings coincide with other instrumental climate records at regional scales (Wilson and Luckman, 2002) and high elevation sites (Diaz and Bradley, 1997). Increased cloud cover may lead to a decrease of daily maximum temperature due to greater reflection of incoming radiation from the upper surface of clouds, and an increase of night time minimum temperature because of enhanced downward radiation from clouds (Easterling et al., 1997; Dai et al., 1999). Meanwhile, a decreasing trend in summer precipitation is registered (Baker and Moseley, 2007). In combination, these opposing trends between summer temperatures and precipitation seem to be a major cause for the observed retreat of glaciers and for a rise of the upper tree line evidenced from repeat photographs in the central Hengduan Mountains (Figure 2.7; Backer et al., 2005; Baker and Moseley, 2007).

Figure 2.6 Temporal changes of regional monthly maximum (Tmax), mean (TEM), minimum (Tmin) temperatures in the summer (a, June-September) and winter (b, prior November-February) seasons; (c) temporal changes of regional diurnal temperature range (DTR). The regional temperature series were developed from two high-elevation meteorological stations of Shangri-la and Daocheng. Values are expressed as anomalies (reference period = 1959–2004) and bold lines were smoothed using a 10-year low-pass filter. R indicates the linear correlation coefficients. The asterisks indicate the 99% significant level and ns denote no significant.

2. Physical setting 20

Figure 2.7 Repeat photo-pairs from the Mingyong Glacier site (a, b) and the Baima Snow Mountain site (c, d) in the central Hengduan Mountains. The original photographs in (a) and (c) were taken by Joseph Rock in 1923 (Rock, 1926); repeat photographs (b, d) were taken by Robert Moseley. Adapted and slightly modified from Baker and Moseley (2007).

2.4 Vegetation

From a biogeographical point of view, the area is located in a transitional zone between the south-eastern Tibetan Plateau and the north-western edge of the Yunnan Plateau. During the Pleistocene, the Hengduan Mountains were not completely covered by ice, making them a refuge area for plant species and wildlife. Thus, the region is one of the hot spots of plant biodiversity in the world (Franzel et al., 2003; Xu and Wilkes, 2004). It is rich in plant diversity with a high degree of endemism, and is the worldwide center for the distribution of many types of flower, such as rhododendron and the primrose. For example, in NW Yunnan Province there are 84 rare or endangered plant species that belong to 49 families and 84 genera (Xu and Wilkes, 2004). After alpine mosaic, evergreen broadleaf forest has the second highest number of rare and endangered plant species in the region, especially in the western edges of the central Hengduan Mountains (Guo and Long, 1998). The most important vegetation types in the region are alpine mosaic and a variety of natural forest ecosystems. Presently, the region is covered by widespread forests dominated by various coniferous species (Frenzel et al., 2003). Figure 2.8 show the distribution pattern of major coniferous species within a large area of eatern Tibetan Plateau.

2. Physical setting 21

Figure 2.8 Distribution patterns of selected conifer taxa in south-eastern Tibet, western Sichuan and northwestern Yunnan Province (from Franzel et al., 2003).

On the other hand, the pronounced altitudinal climatic gradients lead to a differentiation of mountain forests into several altitudinal belts (Figure 2.9; Winkler, 1996). At elevations below 2700 meters, the river valleys are usually covered by drought-adapted scrublands and are presently free of forest. However, isolated groves of Juniper forests can be found under the protection of Buddhist monasteries. Warm-temperate coniferous forests dominated by different pine species occur on the upper montane level between 2700 m and 3300 m a.s.l. Mixed forests of broadleaved and sclerophyllous oak species (e.g. Quecus pannosa Hand.-Mazz.) can be found in fire protected areas. This vegetation belt is strongly influenced by human activity.

2. Physical setting 22

Figure 2.9 Photographic view of the vegetation types along the elevation gradient of the east slope of the Meili Snow Mountains (Photos: Bräuning, 2004). (a) Sub-arid vegetation in the Lancang River valley, juniper patches are protected by a Buddhist monastery, inlet shows a juniper stem decorated with religious symbols; (b) pine forest at about 2700~3000 m a.s.l.; (c) mixed oak forest at south-facing slopes and spruce forest in the valleys; (d) fir forest near the glacier at 3500 m a.s.l.

Cold-temperate coniferous forests cover the subalpine vegetation zone from 3100 m to up to the upper treeline. Distinct forest types are recognized according to the dominate conifer species: hemlock (Tsuga dumosa (D. Don) Eichler), spruce (Picea brachytyla var. complanata (Mast.) Cheng ex Rehd, Picea likiangensis Pritz), fir (Abies georgei Orr, Abies ernestii Rehd. and Abies forrestii C.C. Rogers) and larch (Larix potaninii var. macrocarpa Law) (Yang and Shen, 1984). Under more humid conditions, open spruce forests dominated by Picea likiangensis and Picea brachytyla are found. There are also mixed forests of spruce and fir. Forest stands dominated by

2. Physical setting 23

Abies georgei occur at elevations between 3500 and 4000 m. The upper timberline in the area is formed by fir and larch and is situated at ca. 4000 m a.s.l., whereas isolated trees reach even higher than 4200 m a.s.l (Figure 2.10). Larix potaninii is a deciduous, shade intolerant pioneer species. The larch forests usually occur at sunny, fire-influenced and well-drained slopes in the 3400~4000 m belt. Larch is associated with fir and spruce, or forms pure larch stands at the upper treeline. Glacial moraines and other slopes prone to disturbances (e.g. fire) may support dense stands of larch as an early successional forest type (The Editing Committee of “The Baima National Nature Reserve”, 2003). Besides elevation, forest zonation in the subalpine zone is affected by other factors, e.g. slope, aspect, soil chemistry and fire history. Southward facing slopes may support oak forests well into the subalpine zone. For example, Quercus anquifolioides grows at 3500 m or higher where it forms mixed stands with fir or larch in sunny, drier locations.

Figure 2.10 View of two upper tree line sites at Baima Snow Mountains and Daxueshan pass in the central Hengduan Mountains (Fhotos: Fan, 2006).

2. Physical setting 24

2.5 Land Use and Forest Disturbances

The land area encompassed by Three Parallel Rivers is one of the world’s least disturbed temperate ecological areas. Alpine forests supply for human beings as important ecosystem services, such as livestock grazing, collection of medicinal plants and fungi. Such practices appear to have been sustained for millennia. However, since 1980, government policies have promoted rapid economic growth in the region, stimulating the demand for the resources of the area, especially timber. Some alpine systems in the study area are now beginning to show signs of fragmentation and degradataion (Xu and Wilkes, 2004; Buntaine et al., 2007)

Figure 2.11 Examples of forest disturbances and land use history in the central Hengduan Mountains (Fhotos: Fan, 2005; 2006). (a) Commercial wood transport of Torreya yunnanensis from subalpine forests at east of the Nujiang River; (b) land use penetrates into the alpine forest area at the Haba Snow Mountains; (c) vegetation after forest fire occurred in 1987 at the Daxueshan Mountains.

2. Physical setting 25

In the study region, timber extraction was among the most important sources of villagers’s income and local government tax revenue before 1990s. Following the disastrous flooding of Yangtze River in 1998, natural forests have become increasingly valued by the government for watershed pretection (Zhao et al., 2001). A ban on logging in all natural forest has had significant adverse impacts on rural livelihood. However, since fuelwood cuting in state forests is now illegal, village forests are suffering from unsustainable demand and illegal logging is still practiced. Local land use has important impacts on the sustainability of alpine ecosystems. Figure 2.11 show some examples of human land use and other forest disturbances, such as logging and forest fires. Forest fire is an important cause of deforestation in the alpine area. The common practices of human-induced fire are offen associated with agricultural production, such as clearing swidden field and grazing land for livestocks. Other disturbances such as forest pests were also reported in the study area. For example, an outbreak of Dendrolimus angulata occurred during 1984–1987 has destroyed more than 20 000 ha of native fir (Abies georgei) forest in the Shangri- la Country (Luo, 1989).

3. Data and Methods 26

3. Data and Methods

3.1 Tree-Ring Network

Sampling was carried out between 1999 and 2007. During several field campaigns, a network of 22 well-replicated ring-width chronologies was established (Figure 3.1; Table 3.1). Tree species include Abies georgei, Abies forrestii, Abies ernestii, Picea brachytyla, Picea likiangensis, Tsuga dumosa, Pseudotsuga forrestii and Larix potaninii var. macrocarpa. Most sampling sites were located in nature reserves. The Gaoligong Reserve, the Bitahai National Park and the Baima Snow Mountain Reserve are national nature reserves which were established since the beginning of 1980s. The sampling sites were generally located along the warm- temperate and cold-temperate forest zones from 2800 to 4200 m a.s.l. Increment borers were used to extract cores from the bole at approximately 1.3 m above ground. Samples were taken mostly from living trees or occasionally from fallen trees with intact wood to extend chronology length. In most cases, two cores were extracted from opposite sides of the bole in a direction orthogonal to the slope to avoid slope-imposed growth distortions. Occasionally, only one core was retrieved due to local site conditions, slope conditions or internal rot. To minimize non-climatic influences on ring growth, only trees with no obvious injury (i.e. soil erosion, rock fall) or disease (i.e. fungi) were sampled. The tallest and largest diameter trees were the superior sampling targets to increase the length of the derived tree-ring chronologies.

3.2 Tree-Ring Measurement and Cross-Dating

The air-dried cores were mounted on sample holders with vertical alignment of the tracheids. The surfaces were then prepared with sharp razor blades, and surface contrast was enhanced with chalk. A good surface preparation helps to identify ring borders clearly under a microscope since micro-rings sometimes consist of only two or three rows of tracheids. Ring widths were measured with a LINTAB II measuring system coupled with the TSAP software (RINNTECH, Germany; Rinn, 2003) with a resolution of 0.01 mm. Wood density measurements were carried out with the Lignostation densitometry system (RINNTECH, Germany; Schinker et al., 2003). First, the cores were glued on wooden holders again with vertical alignment of the tracheids. Wood surface was then smoothed with a fly milling in consecutive runs with decreasing speed and milling depth from the previous surface. Thus, a high surface quality was obtained. As a next step, relative density was measured along the smoothed wood surface using a high-frequency dielectric scanner along pre-placed tracks. For each core, six parallel paths were scanned with a distance-in-between ≥ 0.1 mm. Maximum latewood density (MXD) values were extracted from each density profile. Then, an average sequence was calculated year-by-year from the six MXD series.

3. Data and Methods 27

Figure 3.1 Locations of the tree-ring sampling sites and meteorological stations in the central Hengduan Mountains, southwestern China.

Cross-dating was achieved using both visual comparison of growth patterns and statistical tests (Sign-test and t-test) in the software package TSAP (Stokes and Smiley, 1996; Rinn, 2003). Gleichläufigkeit (GLK) is a sign-test statistic, which measures the year-to-year agreement between the interval trends of two series based upon the sign of agreement, usually expressed as a percentage of cases of agreement (Schweingruber, 1988). Student’s t-test is a widely known test for significance of correlations. The combination of both GLK and t-value is realized in the cross-dating index (CDI), which is a powerful parameter in cross-dating. A criterion of GLK > 60%, t-value > 4 and CDI > ~30 was considered to be acceptable for cross-dating in this study. Based on the recommendation of statistical tests, problematic parts of the samples were visually reexamined after improving the wood surface quality and enhancing contrast and resolution etc., so that in many cases, formerly missing rings could be identified (i.e. wedging rings, very narrow rings). Cores with poor quality (e.g. fragmented, rotten and undated) were removed from the final chronologies.

3. Data and Methods 28

Table 3. 1 Locations and characteristics of the tree-ring network in the central Hengduan Mountains, south-western China.

Site Site Species Latitude Elev. No. No. Data Period Length No. Code /Longitude (m) Trees Cores Type (years) 1 DX_A ABGE 28.57/99.82 4200 17 26 TRW 1718–2004 287 2 BM_A ABGE 28.38/98.99 4100 24 49 TRW 1651–1999 349 3 HB_A ABGE 27.37/100.07 3400 14 18 TRW 1760–2007 248 4 YL_A ABFO 27.16/100.24 3250 45 83 TRW 1723–2005 283 5 BT_P PCBR 27.82/99.99 3580 40 50 TRW 1623–2007 385 MXD 6 HP1_P PCBR 28.25/98.91 3600 18 32 TRW 1688–2005 318 MXD 7 HP2_P PCBR 28.24/99.01 3500 32 46 TRW 1738–2005 268 MXD 8 YB_P PCBR 28.40/98.76 3380 15 15 TRW 1696–2003 308 9 YL_P PCLI 27.14/100.22 3200 22 32 TRW 1641–2006 366 10 GK_P PCLI 27.58/99.35 3240 28 38 TRW 1429–2005 577 11 YC_T TSDU 27.59/99.29 3150 32 49 TRW 1542–2005 464 12 YM_A ABER 28.04/99.02 3200 17 19 TRW 1489–2005 517 MXD 13 YE_T TSDU 28.04/98.98 3100 16 30 TRW 1393–2005 613 14 MY_P PTFO 28.45/98.77 2870 10 19 TRW 1770–2005 236 15 BM1_L LAPO 28.38/98.98 4200 34 71 TRW 1702–2003 302 MXD 16 BM2_L LAPO 28.37/99.01 4100 13 26 TRW 1656–1999 344 17 BT_L LAPO 27.82/99.99 3580 33 53 TRW 1755–2007 253 18 HB_L LAPO 27.35/100.07 3650 9 11 TRW 1689–2007 319 19 YL_L LAPO 27.17/100.25 3450 16 34 TRW 1459–2007 549 20 DL1_L LAPO 27.83/98.44 3300 19 38 TRW 1411–2005 595 21 DL2_L LAPO 27.89/98.41 3200 21 35 TRW 1486–2008 520 22 DL_T TSDU 27.88/98.40 3150 21 35 TRW 1530–2005 476

ABGE: Abies georgei; ABFO: Abies forrestii; ABER: Abies ernestii; PCBR: Picea brachytyla; PCLI: Picea likiangensis; TSDU: Tsuga dumosa; PTFO: Pseudotsuga forrestii; LAPO: Larix potaninii var. macrocarpa Elev.: Elevation; TRW: total ring width; MXD: maximum latewood density

3. Data and Methods 29

In general, increments of fir trees have low year-to-year variability (sensitivity). Therefore, marker rings are infrequent so that they have relative low GLK and CDI and comparable t-values statistics. Larch is highly sensitive and thus provides significant statistics (i.e. t-value, CDI), but one can encounter very narrow rings, light rings or missing rings frequently, which have to be identified by cross-dating. Picea brachytyla usually has good wood quality and almost no missing rings. Wedging or missing rings are more likely to occur in old trees of Picea likiangensis or Tsuga dumosa growing at relatively dry sites, when carbon stores from previous year are insufficient to start growth in some extremely dry years.

3.3 Tree-Ring Standardization

The process by which series of ring measurements are converted into a series of chronology indices which represent the magnitude of annual variations of “common” forcing on tree growth is called standardization (Cook and Kairiukstis, 1990). In theory, the total variation of tree-rings can be divided into components of variation due to tree aging, climate, pulses of local endogenous and stand-wide exogenous growth disturbances, and other random factors unique to each tree or radius within the tree. The total observed tree-ring variance may be expressed as a sum of all these factors, the so-called “Linear Aggregate Model” (Cook, 1985), which implies that if the variation in ring width due to all other factors (i.e. noise) but climate can be determined, captured and removed, then theoretically the remaining variance will be composed of a pure climate signal. Through the procedure of standardization, dendroclimotologists maximize the climate signal and minimize all other factors (i.e. noise) as described in the “Linear Aggregate Model”. Trees growing in open-canopy situations, such as the upper timer line, often show a decreasing trend in radial increment with tree aging. This age effect reflects the geometric constraint of adding an equal volume of wood to a stem of increasing radius, and can generally be removed by applying a linear or negative exponential function. In addition to the aging effect, trees growing in a more closed-canopy forest often show growth pulses (i.e. ‘suppression’ and ‘release’), frequently resulting from gap-phase stand development (White et al., 1999). Tree- specific endogenous disturbances (‘endogenous noise’) can be removed as far as possible by applying more flexible cubic spline functions (Cook, 1985). Endogenous disturbances can also be diminished by only using master chronologies with very high replication (number of trees). Exogenous disturbances represent the characteristic response of a tree to a stand wide disturbance. Such disturbances can be caused by fires, storms, diseases, insect infestations, logging, pollution, etc. The identification and removal of exogenous disturbance pulses, however, is not straight forward. Careful selection of sampling sites/trees (i.e. natural old-growth forests, no obvious injuries) can help to minimize the influence of non-climatic factors in the resulting chronologies. In addition, sampling many sites within a region may help to identify possible

3. Data and Methods 30 anomalous sites, and site-specific disturbances can be averaged out by using many site chronologies to calculate a final regional chronology. The standardization procedure was performed with the software ARSTAN (Dendrochronology Program Library; Cook, 1985). Before standardization, the variance of each series caused by ranging sampling replication was stabilized using a data-adaptive power transformation based on the local mean and standard deviation (Cook and Peters, 1997). In order to remove low-frequency trends due to tree aging and stand dynamics, a double detrending procedure was performed for ring-width series discussed in chapter 4, 6. This procedure involved first adjusting a negative exponential or a linear regression function to the raw data. Then, the resulting sequences were detrended with a cubic smoothing spline with a 50% frequency- response cut-off equal to 2/3 of the series length. A more flexible spline function (fixed 120-year spline with a 50% frequency-response cut-off at wavelengths of 60 years) was used in chapter 5 (section 5.1), which allows maximizing the common signals among individual tree-ring series and retains annual to multi-decadal scale climatic variability. A negative or zero slope linear regression function was utilized for detrending maximum latewood density (MXD) series, since their age-related trend is expressed by a linear decrease of MXD with tree age. The tree-ring indices were calculated by subtracting the predicted values from those of the raw data in the same year. Since radial growth in any given year integrates the effects of conditions of the plant in previous years (Fritts, 1976), tree-ring series usually contain some degree of autocorrelation. The tree-physiological lag effects include needle retention and the remobilization of stored food reserves for growth in the following spring. This is the so-called ‘physiological preconditioning’ (Fritts, 1976) or ‘biological memory’ (Büntgen et al., 2007). The auto-correlation of tree-ring series can be modeled by applying an autoregressive model, which has been implemented into the program ARSTAN, producing pre-whitened ‘residual’ indices (Cook, 1985). The autoregressive order can be objectively determined by the minimum Akaike Information Criterion (AIC) procedure (Akaike, 1974, Cook and Kairiukstis 1990). Decomposition of growth-related auto-correlation components is essential to reduce the degrees of freedom needed for significance tests, and for modeling high-frequency variability in tree-ring series. However, this procedure may also deliberately remove some extent of climate-related low-frequency signal embodied in the tree-ring series. To produce a master chronology, detrended indices from individual cores are averaged year by year. Averaging the standardized indices increases the (climatic) signal-to-noise ratio (SNR), since the climatically related variance is not lost by averaging, whereas non-climatic “noise” will be partially cancelled in the averaging process. In order to reduce the influence of outliers (or extreme values) in tree-ring indices, the biweight robust mean was calculated in addition to the arithmetic mean (Cook and Kairiukstis, 1990). The variance of the mean-value function, however, depends on the number of series averaged together and their inter-series correlation

3. Data and Methods 31

(Wigley et al., 1984). Variance stabilization, as described in Osborn et al. (1997), was applied to eliminate variance changes resulting from changing sample replication over time. Several descriptive statistics commonly used in dendrochronology were calculated to compare site chronologies. These included average growth rate (AGR, mm/year) and mean segment length (MSL, years). The standard deviation (SD) estimates the variability of measurements for the whole series, the mean sensitivity (MS) is an indicator of the relative changes in ring-width variance between consecutive years, while the first-order autocorrelation (AC1) assesses relationships with previous growth (Fritts, 1976).

3.4 Common Signal Extraction

The common signal is a statistical quantity representing the common variability present in all tree-ring series at a particular site (Briffa and Jones, 1990). The following parameters were calculated to identify the signal strength of the individual chronologies and their suitability for dendroclimatological studies. The mean inter-series correlation (Rbar or Rbt) describes the strength of the common signal in a chronology by averaging the internal correlations of all cores in a chronology. Signal strength can also be inferred from the percent variance explained by the first principal component (PC#1). Higher values of Rbt and PC#1 indicate greater similarity in the inter-annual growth patterns among the sampled trees.

Expressed population signal (EPS) is a function of Rbt and series replication, which can be calculated as:

N r bt EPS = , 1+ (N −1)r bt where N is the sample replication, r bt is the mean inter-series correlation (Wigley et al., 1984).

EPS is expected to measure chronology confidence and reliability, and a level of 0.85 of EPS is considered to indicate a satisfactory quality for climate reconstruction (Wigley et al., 1984; Briffa and Jones, 1990). In order to assess the temporal variability of the common signal in each chronology, Rbar and EPS are calculated in defined time-windows, where sample depth does not change significantly. In the present study, a 30-year moving window with 15-year overlaps was used. Hierarchical Cluster Analyses (HCA) was performed to detect similarity and relative distances of different site chronologies. The HCA proceeds in stepwise calculations, leading from n clusters of one object (tree site) to one cluster containing all objects (Jongman et al., 1987). The similarity is assessed by the average distance of all possible pairs of sites between two clusters, measured in squared Euclidian distance (Oberhuber et al., 1998; Bräuning, 1999a). The agglomeration the clusters is based on the Ward’s method. This method evaluates the cluster

3. Data and Methods 32 variance for each cluster, and maximize variance between clusters while minimize variance within the clusters. As shown in Figure 3.2, the HCA classification generates six groups of site chronologies. The different groups reflect the differences of tree species and their growing environments. Larch chronologies stand on three separate groups (G1, G2, G4), and the chronology of Pseudotsuga forrestii (G3, MY_P) stand out from all others. Most chronologies of fir and spruce from relative high elevations are fall within the same group (G5), which may contain regional temperature signals. On the other hand, the last group (G6) encompasses site chronologies from relative low elevations where trees may sensitive to moisture stress.

Figure 3.2 Classification dendrogram of tree ring-width site chronologies based on hierarchical clustering of Ward’s method. The symbols of tree species are identical as in Figure 3.1.

Chronologies with similar growth variations are expected to contain similar ‘climate-like’ signals. Principal component analysis (PCA) was used to extract common ‘climate-like’ signals among site chronologies (chapters 4, 5, 6). PCA involves the extraction of orthogonal (uncorrelated) principal components (or factors) from an original set of correlated variables by employing a variance maximizing (“varimax”) rotation of the original variable space (Richman 1986). PCA is employed to reduce the number of variables and to detect a possible structure in the relationships between variables. Thhe first few principal component encompasses most of the variability of the original variables. Therefore, PCA is applied as a data reduction or structure detection method and is widely used in dendroclimatology (Briffa, 1995). Because of the strong inter-correlation of the tree-ring data, the number of components retained for further analyses was determined according the criteria of eigenvalues > 1. The varimax rotation was used to aid the interpretation of the individual site loadings on each eigenvector (Richman, 1986).

3. Data and Methods 33

3.5 Climate Data

Instrumental data near the sampling sites were utilized to detect the most important climatic factors influencing tree growth. Monthly mean temperatures (maximum, mean and minimum), precipitation sums and relative humidity data were obtained from the National Meteorological Information Centre (NMIC) of China. Daily temperature range (DTR) was calculated by subtracting monthly maximum temperature by the minimum temperature. Data homogeneity was tested by applying the double-mass analysis techniques, a graphical technique (Kohler, 1949). Meteorological stations in this region are sparse and generally located in the valley floors. In order to emphasize the spatial representation, monthly mean temperatures (TEM) and precipitation (PRE), as well as monthly mean minimum (Tmin) and maximum temperatures

(Tmax), from a high resolution (0.5° × 0.5°) gridded dataset were also accessed from the Climatic Research Unit (http://www.cru.uea.ac.uk; CRUts2.1; Mitchell and Jones, 2005). In chapter 4, mean values were calculated from 16 grid-boxes which cover the region 27.25–28.75 °N and 98.75–100.25 °E over the 1951–2002 period. The CRU gridded dataset was also used to develop a field correlation with instrumental and reconstructed temperatures for the central Hengduan Mountains in chapter 5. The Palmer Drought Severity Index (PDSI; Palmer, 1965) estimates plant moisture availability by integrating precipitation, soil moisture demand and supply into a highly simplified hydrological accounting system. PDSI has been employed to investigate spatial and temporal variations of moisture conditions across Europe (Briffa et al., 1994; van der Schrier et al. 2006), North America (Cook et al., 1999) and north China and Mongolia (Li et al., 2006; 2008b). A global PDSI dataset on a 2.5º × 2.5 º grids was developed by Dai et al. (2004). The grid points closest to our sampling sites were extracted for the period 1951–2002. Monthly PDSI series were used to test the influence of soil moisture on tree growth (chapter 4) and to develop a drought reconstruction (chapter 6). Usually, regional averages of climate series are more representative for a wide area than those of single station data. A regional meteorological data series was developed by applying the techniques outlined in Jones and Hulme (1996). Monthly values for each station were standardized as z-scores relative to their respective means over the common period, and averaged to monthly z-scores for the regional average series. The monthly z-scores were then converted to ‘absolute’ values using the grand means and standard deviations of the original monthly series.

3.6 Calibration and Reconstruction

3.6.1 Climate-Growth Response

3. Data and Methods 34

Correlation and response function analyses were performed to quantify the tree-ring response to climatic variables (i.e. temperature, precipitation), by using the software DENDROCLIM2002 (Biondi and Waikul, 2004). The correlation/response analyses, involved empirical modeling to calculate coefficients between tree-ring chronologies and meteorological data over their common period, and for a biological year from the summer prior to growth (i.e. previous July) until the end of the current growing season (i.e. current October). Response functions deal with the problem of multicollinearity, which is commonly encountered in multi- variable sets of meteorological data. Stepwise multiple regressions were computed on uncorrelated principal components to assess climate-growth relationships. The significance of the calculated partial regression coefficients were estimated based on 1000 bootstrapped estimates obtained by random extraction with replacement from the initial data set (Guiot, 1991).

Redundancy analysis (RDA) is another effective method, but not frequently used, for detecting the climate-growth relationships (Trouet et al., 2001; Tardif et al., 2006; Friedrichs, 2008). RDA is the canonical extension of PCA and intends to display the main trends in variation of a multidimensional data set in a reduced space of a few linearly independent dimensions (Legendre and Legendre, 1998). In RDA, however, the canonical axes differ from the principal components (PCs) in that they are constrained to be linear combination of supplied environmental variables (ter Braak and Smilauer, 2002). RDA may be understood as a two-step process: (1) each tree-ring chronologies are regressed on the climate variables and the fitted values are computed; (2) a PCA is then carried out on the matrix of fitted values to obtain the eigenvalues and eigenvectors (Legendre and Legendre, 1998). The climate variables were selected using a forward selection on the basis of the goodness of fit and tested for significance using a Monte Carlo permutation test based on 999 random permutations. This procedure was repeated until a variable was tested non-significant at the 5% level. In chapter 4, the RDA method was employed to quantify climate influence on tree growth, using the program CANOCO (Version 4.5; ter Braak and Smilauer, 2002).

3.6.2 Transfer Functions

Since the climate-growth relationships are statistically determined, the ‘physiologically meaningful’ dominant climatic variables influencing radial growth can be targeted for climate reconstruction. Transfer functions are used in tree-ring based reconstructions, where tree-ring chronologies serve as the predictor (independent) variables and the ‘target’ climatic series serve as the predictand (dependent) variable. After the common ‘climate-like’ signal among tree-ring site chronologies was extracted by using principal component analyses, a linear regression function was employed to transfer principal component scores (mostly PC#1) into series of pact climate variability. The principle of uniformitarianism, a crucial keystone in any palaeoenvironmental study, is assumed to be true for tree-ring-based climate reconstructions. In dendroclimatology, this

3. Data and Methods 35 principle implies that climate-growth relationships are stable over time so that the nature of past climate can be reliably inferred from statistically derived tree-ring calibrations (Fritts, 1976). The reliability of a particular statistical model is usually verified for a time-period independent of calibration (Cook and Kairiukstis, 1990). Successful verification supports the validity of the uniformitarianism principle in this case, and therefore the validity of the reconstruction. Usually, the full calibration period is split into two or more sub-periods, the model is first calibrated on the earlier period and then verified with independent data from the later period, and vice versa. This is called the ‘split-sample verification’. However, short meteorological datasets may result in non-robust split-sample verification due to the artificial selection of the verification period. In this case, the ‘leave-one-out’ cross-validation method is superior (Michaelsen, 1987), and was used in this study to verify our reconstructions. The ‘leave-one-out’ cross-validation is similar to the ‘split-sample verification’, but models are repeatedly estimated using the data set after omitting one observation each time from calibration, and the estimated model is used to generate a predicted value for the predictand for the deleted observation. Finally, a time series of the predictions assembled from the deleted observations is compared to the observed predictand to compute validation statistics of model accuracy (Michaelsen, 1987). Verification is processed by visually comparison of the actual and estimated climate data, and by statistical measures including several tests of association (Fritts, 1976; Cook and Kairiukstis, 1990). The sign test (ST) is a non-parametric test of the similarity between series based upon account of the number of agreements and disagreements in sign of departure from their mean in the observed and reconstructed series (Fritts, 1976; Cook et al. 1994). If the number of agreements exceeds the number of disagreements significantly, the reconstruction passes. A significance table is detailed by Fritts (1976). The sign test only measures the agreement in the direction of change from one year to the next. In contrast, the product means test (Pmt) takes into account both the direction and magnitude of the changes when testing for agreement between series. The Pmt is calculated from the cross-products of the actual and estimated yearly departures from their respective mean values. If the mean positive cross-product is significantly greater than the mean negative cross-product based on the standard t-test, the reconstruction passes the verification test (Fritts, 1976). The reduction of error (RE) statistic tests whether a reconstruction provides a better estimate of climatic variability than simply using the mean of the meteorological data in the calibration period (Cook and Kairiukstis, 1990; Cook et al. 1994). It is calculated as

⎡ n ⎤ ˆ 2 ⎢ ∑(xi − xi ) ⎥ i=1 RE = 1.0 − ⎢ n ⎥ , ⎢ 2 ⎥ (x − xc ) ⎢∑ i ⎥ ⎣ i=1 ⎦ where xi and xˆi are the actual and estimated climate data in year i of the verification period, and

xc is the mean of the actual data in the calibration period. The value of RE can range from

3. Data and Methods 36 negative infinity to a maximum value of 1.0 which indicates perfect estimation. If the total difference between the estimates and the actual data is less than the total difference between the calibration mean and the actual data, the RE statistic will be positive. It should be noted that if there is a trend within the instrumental data (i.e. the mean of the calibration and verification periods are different), RE can be greater than the square of the Pearson’s correlation coefficient. In this case, the coefficient of efficiency (CE) should be used. It is calculated similarity to the RE but replaces xc by the mean of the actual data in the verification period ( xv ). The CE statistic is similar to RE except that its benchmark for determining model skill is the verification period and not the calibration period. Like RE, CE can range from negative infinity to a maximum value of 1.0, and any positive value indicates that there is merit to the reconstruction (Cook et al. 1994; Wilson et al. 2006).

4. Climate-growth responses 37

4. Climate-Growth Responses of High-Elevation Conifers

4.1 Data Source and Methods

In order to improve our understanding of tree rings as proxies for climate reconstructions and to estimate the ecological responses of subalpine trees to climate change, profound knowledge about the influence of climatic variables on tree growth is needed. In this chapter, climate- growth relationships of major forest types that are characteristic for different environmental conditions in the central Hengduan Mountains are investigated. Eight tree-ring chronologies of the major tree species Abies and Picea that are aligned along an elevation gradient from 3200 to 4200 m a.s.l. were selected to study the influence of altitude on tree growth (Figure 4.1). In total, 302 cores from 202 living fir (Abies georgei and A. forrestii) and spruce (Picea brachytyla and P. likiangensis) trees were collected (Table 4.1; Table 4.2). The sites are located in six Nature Reserves: Meili Snow Mountain (YB_P), Baima Snow Mountain (BM_A, HP_P), Daxueshan (DX_A), Bitahai National Park (BT_P), Haba Snow Mountain (HB_A) and Yulong Snow Mountain (YL_A, YL_P; note that _A stands for fir sites, _P indicates spruce sites).

4. Climate-growth responses 38

Due to the relatively high absolute elevation of the valley floors (around 2000 m a.s.l.), we will refer to high, middle and low elevation sites in the context of this study as locations situated above 3500 m, between 3300–3500 m and below 3300 m, respectively. Ring-width chronologies were developed according to the techniques described in chapter 3. To assess the similarity among the eight individual residual chronologies, correlation and principal component analyses (PCA) were applied over the common period 1850–1999. Growth-climate response analyses were examined by computing correlation between individual tree-ring residual chronologies and regional climate variables over the 1951–2002 period, using a 16-month window from July of the year prior to growth until the current-year October. Redundancy analysis (RDA) was applied to detect the most important growth influencing climate variables over the common interval 1951–2002, using the program CANOCO (Version 4.5; ter Braak and Smilauer, 2002).

4.2 Internal Chronology Statistics

Table 4.1 and Table 4.2 list the locations and descriptive statistics of the eight ring-width chronologies. The chronologies cover the last 200–350 years, with mean tree ages ranging from c.a. 150 years (HB_A) to 280 years (YL_P) (Table 4.1). Low mean growth rates are associated with increasing elevation and stand age. The chronologies generally display a low year-to-year variability (mean sensitivity, MS), which is typical for conifers growing in humid environments. However, chronologies from lower elevation sites (e.g. YL_A and YL_P) display a higher standard deviation (SD) and MS (Table 4.2; Figure 4.2). Trees from dry forest habitats often show higher inter-annual growth variability than trees from temperature-limited sites (Fritts, 1976; Bräuning, 2001a; Liang et al., 2006a). The trees from lower elevations probably suffer more from occasional moisture stress than those growing at higher elevation sites.

Table 4. 1 Site locations and characteristics

Site Species Latitude Longitude Elev. (m) MSL AGR (mm) DX_A A. georgei 28.57 99.82 4200 150 0.76 BM_A A. georgei 28.38 98.99 4100 237 0.57 HB_A A. georgei 27.37 100.07 3400 146 1.06 YL_A A. forrestii 27.16 100.24 3250 149 1.88 BT_P P. brachytyla 27.82 99.99 3580 234 0.81 HP_P P. brachytyla 28.25 98.91 3500 186 1.40 YB_P P. brachytyla 28.40 98.76 3380 218 1.25 YL_P P. likiangensis 27.14 100.22 3200 277 1.07

DX_A, Daxueshan; BM_A, Baima Snow Mountain; HB_A, Haba Snow Mountain; YL_A, YL_P, Yulong Snow Mountain; BT_P, Bitahai Nature Reserve; HP_P, Hongpo at Baima Snow Mountain; YB_P, Yubeng at Meili Snow Mountain; Elev., elevation; MSL, mean segment length; AGR, average growth rate.

4. Climate-growth responses 39

Table 4. 2 Descriptive statistics of the eight total ring width chronologies

+ + + + Site Cores(Trees) Period SD* MS* AC1* Rbt EPS PC1 SNR DX_A 26 (17) 1718–2004 0.25 0.14 0.73 0.54 0.93 45 14.1 BM_A 49 (24) 1651–1999 0.26 0.14 0.79 0.47 0.94 30 16.4 HB_A 18 (14) 1760–2007 0.26 0.16 0.70 0.37 0.74 29 2.7 YL_A 32 (19) 1723–2005 0.32 0.24 0.58 0.53 0.93 39 13.3 BT_P 50 (40) 1623–2007 0.22 0.14 0.65 0.48 0.96 37 23.1 HP_P 80 (51) 1688–2007 0.28 0.18 0.70 0.59 0.97 36 36.1 YB_P 15 (15) 1696–2003 0.24 0.15 0.71 0.39 0.80 29 4.1 YL_P 32 (22) 1641–2006 0.28 0.18 0.72 0.50 0.92 36 12.1

SD, standard deviation; MS, mean sensitivity; AC1, first-order autocorrelation; Rbt, mean inter-series correlation; EPS, expressed population signal; PC1, percent variance explained by the first principal component (PC1); SNR, signal-to-noise ratio. * Calculated for the standardized chronologies prior to autoregressive modeling. + Calculated for prewhitened chronologies over the common period 1900–1999.

Mean inter-series correlations (Rbt) range from 0.37 to 0.59, and expressed population signals (EPS) vary between 0.74 and 0.97. The first principal component (PC#1) explains more than 29% of the total variance in all individual series. As a consequence of different sample depth and Rbt, the signal-to-noise ratio (SNR) ranges from 2.7 (HB_A) to 36.1 (HP_P) (Table 4.2). The amount of variance explained by PC#1 and the high SNR values indicate that the developed chronologies contain strong common signals. The combination of relatively high values of Rbt and EPS confirms that the chronologies are suitable for growth-climate relationship studies (Wigley et al., 1984). The high first-order autocorrelations (AC1) reflect a high persistence of the ring-width chronologies, indicating a significant impact of previous year’s climate on current year’s ring width, probably caused by carry-over effects of carbohydrates used for earlywood formation (Fritts, 1976).

Table 4. 3 Correlation coefficients of the eight residual chronologies for the well replicated period 1850–1999. * Significant at p < 0.05; ** Significant at p < 0.01.

BM_A HB_A YL_A BT_P HP_P YB_P YL_P DX_A 0.53** 0.07 0.22** 0.18* -0.02 0.10 0.11 BM_A 0.21* 0.18* 0.23** 0.22** 0.22** 0.23** HB_A 0.38** 0.34** 0.28** 0.26** 0.38** YL_A 0.10 0.09 0.06 0.47** BT_P 0.40** 0.34** 0.26** HP_P 0.42** 0.13 YB_P 0.27**

4. Climate-growth responses 40

Figure 4.2 Eight tree ring-width residual chronologies. The site codes are consistent with those of Table 4.1. Tree-ring series are smoothed with a 15–year cubic spline (thick lines).

4.3 Comparison of Chronologies

Correlations between chronologies reflect differences caused by species and distance among sites (Table 4.3). The four fir chronologies display significant inter-site correlations except between DX_A and HB_A, with the highest correlation found between the two highest sites DX_A and BM_A (r = 0.53, p < 0.01). The four spruce chronologies are correlated significantly among each other except for HP_P and YL_P. Correlation coefficients are higher between neighboring sites of different species (e.g. between YL_A and YL_P) than between distant sites of the same species. PCA revealed that the first three rotated factors have eigenvalues > 1 and account for 34%, 17% and 15% of the total variance, respectively, or cumulatively 66% of the total variance (Figure 4.3). According to the factor loadings of the first three PCs, the site chronologies can be divided into three groups (Figure 4.3). This is consistent with the results of the correlation

4. Climate-growth responses 41 analyses. The loadings of PC#1 describe the environmental signals that are common between the Picea brachytyla chronologies (BT_L, HP_L, and YB_L). PC#2 represents the common variances of the two high-elevation sites (DX_A and BM_A) of Abies georgei. PC#3 encompasses variables of the chronologies at low elevation sites (YL_A, YL_P and parts of HB_A).

Figure 4.3 Relative positions of the eight ring-width chronologies according to the three significant factors resulting from principal component analysis over the period 1850–1999. The explained variances of the factors are indicated in brackets. The site codes are consistent with those given in Table 4.1. Groups of similar chronologies are indicated by identical symbols.

4.4 Climate-Growth Response

The climate-growth response patterns of the four fir chronologies are more or less similar, with a positive response to temperatures in the early winter (November-December) previous to growth and in the summer of the growing season (June-August; Figure 4.4; Table 4.4). However, climate-growth relationships change considerably along the altitudinal gradient in this study: the most pronounced positive effects of temperature on radial growth occur at treeline sites like DX_A and BM_A, whereas at the low elevation site YL_A, correlation with temperature is still significant, but negative. In contrast, fir growth rates at low elevation sites show positive correlations with spring (March-May) precipitation and with PDSI. These findings confirm that water supply during the spring season is crucial for fir growth at low elevations.

4. Climate-growth responses 42

Figure 4.4 Correlation between the four fir ring-width residual chronologies and regional monthly mean maximum (Tmax), mean (TEM), minimum temperature (Tmin), precipitation (PRE) and Palmer drought severity index (PDSI). The correlation coefficients were calculated from previous year’s July to current year’s October over the common period 1951–2002. The horizontal dashed lines denote the 95% significance level.

Radial growth of spruce trees at middle elevations is influenced by temperature variations throughout the growing season (Table 4.4, Figure 4.5 a-c). Compared to fir, spruce growth is limited by low winter temperatures within a wider seasonal window. Warm temperatures in the growing season have a positive effect on radial growth, but most of the correlations found are not statistically significant. Negative correlations occur between radial growth of spruce and

4. Climate-growth responses 43 temperatures in the previous late summer season from August to September, accompanied by a positive influence of precipitation in the same season. At the low elevation site YL_P, correlations between radial growth and temperatures are mostly not significant. However, precipitation in February and spring PDSI correlate significantly with tree growth (Figure 4.5 d).

Figure 4.5 Correlation between four spruce ring-width residual chronologies and regional monthly mean maximum (Tmax), mean (TEM), minimum temperature (Tmin), precipitation (PRE) and Palmer drought severity index (PDSI). The correlation coefficients were calculated from previous year July to current year October over the common period 1951–2002. The horizontal dashed lines indicate the 95% significance level.

4. Climate-growth responses 44

Table 4. 4 Correlation coefficients between eight residual chronologies and seasonal climatic variables of monthly mean temperature (TEM), precipitation (PRE) and Palmer drought severity index (PDSI) for the period 1951–2002 (n = 53).

Climatic DX_A BM_A HB_A YL_A BT_P HP_P YB_P YL_P variable TEM (p8-p9) 0.06 -0.13 -0.07 -0.21 -0.38** -0.11 -0.31* -0.23 TEM (p11-p12) 0.22 0.42** 0.30* 0.00 0.23 0.36** 0.46** 0.06 TEM (p11-2) 0.12 0.26 0.21 -0.05 0.31* 0.47** 0.49** -0.005 TEM (6-8) 0.48** 0.20 0.19 -0.01 0.00 0.07 0.08 0.06 PRE (3-5) -0.13 -0.02 0.03 0.42** -0.23 0.08 -0.22 0.10 PDSI (3-5) 0.09 -0.07 0.13 0.59** -0.04 0.30* 0.06 0.30*

The numbers in the parenthesis represent consecutive number of months, while p represents the year before ring formation. * Significant at p < 0.05; ** Significant at p < 0.01.

Redundancy analysis (RDA) is a direct extension of multiple regressions applied to multivariate data (Legendre and Legendre, 1998) and enables the analysis of the combined effects of climate factors on tree growth. According to the RDA results, all site chronologies shared a positive loading on axis 1(Figure 4.6), which explains 21% of the total variance (Table 4.5). November temperature during the year prior to ring formation is the most relevant climatic factor limiting radial growth of spruce and fir at high- and middle-elevation sites (Figure 4.6). High temperatures in September of the previous year have negative effects on radial growth at most high- and middle-elevation sites. At low-elevation sites (i.e. YL_A, YL_P), spring precipitation is the most important factor for growth, and temperature in May is negatively associated with radial growth. The RDA results are generally in agreements with those of correlation analyses, while highlight the most important climatic variables influencing tree-ring growth.

Table 4. 5 Summary of redundancy analysis statistics

Canonical Canonical Canonical Canonical axis 1 axis 2 axis 3 axis 4 Eigenvalues a 0.21 0.084 0.027 0.007 Species-environment correlation b 0.675 0.633 0.504 0.352 Cumulative percentage variance of 63.0 88.3 96.4 98.5 species-environment relation (%) c a Variance in a set of variables explained by a canonical axis. b Amount of the variation in species composition that may be “explained” by the environmental variables. c Amount of variance explained by the canonical axes as a fraction of the total explained variance.

4. Climate-growth responses 45

Figure 4.6 Biplot of the redundancy analyses (RDA) calculated from the eight residual chronologies and the monthly climate parameters for the period 1951–2002. Significant (p < 0.05) climate factors are indicated by vectors (black arrows); the longer the vector the more important the climate parameter. The correlation between the variables is illustrated by the cosine of the angle between two vectors. Vectors pointing in nearly the same direction indicate a high positive correlation, vectors pointing in opposite directions have a high negative correlation, and vectors crossing at right angles are related to a near zero correlation (Legendre and Legendre 1998). P = Precipitation, T = Temperature, Tn = Minimum temperature, (t-1) = year before ring formation, numbers represent months (e.g. 3 = March).

The first PC which represents growth variability at middle elevations correlated significantly with winter temperatures (R = 0.56; Figure 4.7a) during the period of 1951–1999. At inter- annual to decadal scales, there are generally agreements between PC#2 and summer temperatures (R = 0.40; Figure 4.7b), as well as between PC#3 and spring PDSI (R = 0.48; Figure 4.7c). In the study area, the climate during 1940s and 1950s were noticeable warm. Regional temperatures have been increasing significantly since 1960s, by 0.04 °C yr-1 and 0.022 °C yr-1 for winter (Prior November to February) and summer season (June to August), respectively (Figure 4.7a and b). This temperature maximum will soon be overtopped, if the warming trend observed for the last decades will continue. Under this warming trend, enhanced tree growth anomalies at high- to middle-elevation sites are evident during the past two decades.

4.5 Effects of Winter Climate on Tree Growth

The most consistent growth-climate response in the studied trees is associated with positive correlation with winter temperature at high- and middle-elevation sites (Table 4.4, Figures 4.4, 4.5 and 4.6). Bräuning (2001a) demonstrated that trees growing in the cold-moist environment near the upper treeline in eastern Tibet are sensitive to temperature variations, especially in the winter season. In the west Sichuan Plateau, northeast of our study area, winter minimum temperatures (last December to current February) were found to be the most crucial factor

4. Climate-growth responses 46 limiting radial growth of Balfour spruce (Picea balfouriana) (Shao and Fan, 1999) and some other coniferous species (Wu et al., 2006). Winter temperatures are also found to limit radial growth of Picea crassifolia in northeastern Tibet (Liang et al., 2006a), Juniperus przewalskii in the Qilian Mountains (Kang et al., 1997; Gou et al., 2007b), and Picea schrenkiana in the Tianshan Mountains (Yuan and Li, 1999). Thus, positive growth-responses to winter temperatures at high-elevations sites seem to be a widespread phenomenon in the mountainous areas of eastern Tibet.

Figure 4.7 Comparisons of (a) PC#1 with winter (prior November to February) mean temperatures, (b) PC#2 with summer (June to August) mean temperatures and (c) PC#3 with spring (March to May) PDSI. Values are adjusted for their mean and standard deviations. Bolded lines are smoothed with a 10- year low-pass filter. R indicates the linear correlation coefficients between PCs and climatic variables over the period 1951–1999 and p denotes their significant levels.

4. Climate-growth responses 47

However, similar reaction patterns are also documented from other humid mountain regions of the world. In the North Cascade Mountains, Peterson and Peterson (1994) found that growth of four investigated species correlated positively with previous November temperatures. Winter temperature has also been found to influence radial growth of various tree species of northeastern America (Pederson et al., 2004). The detrimental effect of low winter temperatures on tree growth, especially at treeline sites, supports the hypothesis that winter temperature is one of the dominant controls on the location of the timberline ecotone (Körner, 1998). Winter carbon shortage, as evidenced from insufficient non-structural carbohydrates (NSC) and sugar-to-starch ratio, has been suggested to influence the survival and growth of coniferous trees at the upper tree line on the eastern edge of the Tibetan Plateau (Li et al., 2008). The positive influence of winter temperature on radial growth has been attributed to the increase of stem carbohydrate storage in a warm late autumn, thus enhancing earlywood growth in the following spring (Kang et al., 1997; Gou et al., 2007b). Moreover, ring width of high elevation conifers is often reduced by low winter temperature as a consequence of bud damage, frost desiccation and reduced root activity due to low soil temperature (Körner, 1998). Defoliation and bud mortality deplete the pool of stored carbohydrates and reduce a tree’s potential for future growth and photosynthetic capture (Lazarus et al., 2004). In addition, after cold winters with delayed snow melt, the following vegetation period is shortened, which may lead to a reduced earlywood width in the following year (Ettl and Peterson, 1995; Peterson and Peterson, 2001; Bräuning, 2001a).

4.6 Site-Specific Climate-Growth Responses

The effects of steep environmental gradients on forest composition and patterns of stand development in the central Hengduan Mountains are well documented (i.e. Yang and Shen, 1984; Yu et al., 1989). The results of this study emphasize that these gradients cause markedly diverse patterns of tree growth responses to climatic variations. Conifers growing at high and middle elevations are mainly limited by temperature variations, whereas moisture availability in the spring season is more important at lower elevation sites (Table 4.4, Figure 4.4 and Figure 4.5). This is also clearly reflected by the reaction to PDSI. High-elevation fir sites (DX_A, BM_A) show negative correlations, whereas low-elevation fir sites (HB_A, YL_A) show positive correlations with PDSI (Figure 4.4). For spruce, this picture looks different, since absolute elevations of spruce sites are lower than for fir (Table 4.1). Nevertheless, the highest spruce site (BT_P) shows lower correlations to PDSI than all other sites. Precipitation in June has a negative influence on radial growth at higher elevations, especially at the upper treeline sites DX_A and BM_A. At high elevation sites, abundant rainfall is generally combined with enhanced cloudiness and reduced radiation input and lower temperatures. Monthly mean temperature (TEM) and monthly precipitation sums (PRE) in May and June show inverse correlation of -0.56 and -0.54 (p < 0.01), respectively. As a consequence of abundant rainfall during early summer (May-June), tree growth at high elevations is

4. Climate-growth responses 48 constrained by light availability and temperature (e.g. Graham et al., 2003; Bräuning and Mantwill, 2004).

4.7 Species-Specific Climate-Growth Responses

Fir and spruce trees show different climate response behavior, both seasonally and in magnitude (Figure 4.4 and Figure 4.5). Compared with spruce, fir responds to winter temperature variations within a narrower seasonal window, mostly confined to November of the year prior to growth. In the summer season (June-August), low temperatures limit radial growth of fir more severely than growth of spruce (compare Figure 4.4 and Figure 4.5). For spruce growing at middle elevations, radial growth is mainly influenced by temperature variations in the winter season, whereas growth response to summer temperature variations is only weak (Table 4.4; Figure 4.5). Radial growth of spruce trees correlates negatively (positively) with temperatures (precipitation) in the late summer of the previous year. Lagged effects of summer temperature and precipitation on growth in the following year are commonly observed in tree-ring studies of subalpine conifers (Villalba et al., 1994; Ettl and Peterson, 1995; Peterson and Peterson, 2001; Tardif and Stevenson, 2001). Warm and dry late summers can reduce the accumulation of carbohydrate reserves by limiting photosynthesis through drought stress, by increasing respiration rates, and by diverting energy reserves to current-year growth (Fritts, 1976). At low elevation sites, growth of A. forrestii is more severely restricted by spring moisture availability than growth of P. likiangensis, indicating that this spruce species is more tolerant of droughts.

5. Temperature reconstructions 49

5. Temperature Reconstructions

5.1 Tree Ring Width-Based Annual Temperature Reconstruction

According to the climate-growth analyses of various tree species of the central Hengduan Mountains, temperatures are generally the most important factor influencing tree growth near the upper timber line. In this chapter, four tree ring-width chronologies from Picea brachytyla were selected to reconstruct regional annual temperature variability during the past 250 years. In total, 112 increment cores from 83 trees were selected from four sites near the upper timber line (Figure 5.1.1; Table 5.1.1). The chronologies were developed according the techniques described in chapter 3. The raw ring-width series were standardized using a fixed 120-year spline (with a 50% frequency- response cut-off at wavelengths of 60 years). This detrending method allows maximizing the common signals among individual tree-ring series; however, it also removes information on century-scale climate variability. Therefore, the discussion on past climate variability will focus on multidecadal scale, and does not include long-term trends.

Figure 5.1.1 Locations of the sample sites and meteorological stations.

5. Temperature reconstructions 50

Table 5.1. 1 Site characteristics and chronology statistics.

Site Lat/ Lon Elev Trees Period AGRa MSb AC1b Rbarc EPSc (m) (cores) (mm) HP1_P 28.25/98.91 3600 18 (32) 1688-2005 1.35 0.16 0.73 0.32 0.85 HP2_P 28.24/99.01 3500 32 (46) 1738-2005 1.43 0.18 0.70 0.46 0.96 BT_P 27.82/99.99 3540 18 (19) 1634-2003 0.85 0.16 0.63 0.29 0.83 YB_P 28.40/98.76 3280 15 (15) 1696-2003 1.25 0.15 0.71 0.21 0.74

HP1_P, HP2_P: Hong Po at the Baima snow mountain (Picea brachytyla); BT_P: Bitahai Nature Reserve (Picea brachytyla); YB_P: Yu Beng at the Meili snow mountain (Picea brachytyla); Lat: latitude; Lon: longitude; Elev: elevation; AGR: average growth rate; MS: mean sensitivity; AC1: first-order autocorrelation; Rbar: mean inter-series correlation; EPS: expressed population signal. a Calculated for raw ring-width values. b Calculated for ARSTAN standard chronologies. c Calculated for ARSTAN residual chronologies for 30-year intervals with 15-year overlaps

5.1.1 Chronology Comparisons

The moving Rbar and EPS statistics of the site chronologies signal strength range from 0.21 to 0.46 and from 0.74 to 0.96, respectively. All chronologies meet the 0.85 EPS criterion after AD 1750, except for site YB_P whose EPS is above 0.80 at 1750 and reaches the 0.85 limit only after 1780. Common signals between the individual chronologies were assessed by applying correlation analysis and principal component analysis. Figure 5.1.2 shows that the five chronologies share a lot of common decadal variation, which is also reflected by the significant correlations between the chronologies (Table 5.1.2). The average correlation between all chronologies is 0.47 for the well-replicated period 1850–2003, with the correlation amount to 0.81 between the neighbouring sites HP1_P and HP2_P (Table 5.1.2). Only the eigenvalue of the first principal component (PC#1) was greater than one and PC#1 accounts for 54.8% of the total variance for the period 1750–2003 (Table 5.1.3). The factor loadings for PC#1 of all chronologies are positive and very similar in magnitude (0.89 for HP1_P, 0.80 for HP2_P, 0.61 for BT_P, 0.61 for YB_P, respectively).

Table 5.1. 2 Pearson correlations among the residual chronologies for the period 1850–2003. All correlation are significant at the p < 0.05 level.

HP1_P HP2_P BT_P YB_P HP1_P 1 0.81 0.44 0.48 HP2_P 1 0.35 0.37 BT_P 1 0.35 YB_P 1

5. Temperature reconstructions 51

Figure 5.1.2 The four residual chronologies from the central Hengduan Mountains. The sample depths through time are shown in the lower sections of each graph. Pink lines represent annual values; blue lines are 15-year cubic smoothing splines.

Table 5.1. 3 Eigenvalues of principal component analysis of the four spruce residual chronologies for the period 1750–2003.

Component Eigenvalue Variance (%) Cumulative variance (%) 1 2.19 54.8 54.8 2 0.87 21.8 76.6 3 0.70 17.4 94.0 4 0.24 6.0 100

5. Temperature reconstructions 52

5.1.2 Climate-Growth Relationship

Regional climate series were developed from two high-elevation stations of Shangeri-la and Deqin (see Figures 2.4 and 3.1). Correlation analysis with regional climate data indicates that the radial growth rates are mainly influenced by temperature conditions. Warm temperatures have a positive impact on tree growth throughout the year, especially in the winter season (Figure 5.1.3). Correlations between ring width and precipitation are generally low and rarely exceed the 95% significance level. Previous year’s rainfall during summer and autumn generally has a positive influence on tree growth in the following year (Figure 5.1.3 and Figure 5.1.4d).

Figure 5.1.3 Correlation (columns) and response (dot-lines) functions coefficients between radial growth and regional monthly mean temperature (left) and total monthly precipitation (right). Correlations are computed from previous year May to current year October over 1958–2003. Horizontal dashed lines denote the 95% significance level of the correlation. Asterisks denote significance (p < 0.05) of response function based on bootstrapping tests.

5. Temperature reconstructions 53

Figure 5.1.4 Correlation (columns) and response (dot- lines) functions coefficients between PC#1 of four spruce chronologies and the regional monthly mean (TEM) (a),

minimum (Tmin) (b), maximum

temperature (Tmax) (c) and precipitation (PRE) (d) from previous year May to current year October over the common period 1958–2003. The horizontal dotted and dashed lines donate the 99% and 95% significance level for the correlation function, respectively. Asterisks denote significance (p < 0.05) of response function based on bootstrapping tests.

Rainfall during the growing period (July and August) also stimulates growth. January precipitation shows a significant positive influence on ring width at sites HP1_P and HP2_P. Possibly, winter precipitation enhances soil moisture which enhances available moisture resources for tree growth during the early growing season. In general, correlation coefficients between individual ring-width chronologies and monthly climate data are relatively weak, but temperature has a higher influence on tree growth at these subalpine sites than precipitation. The correlation calculations between PC#1, which represents the common signal of the spruce chronologies, and regional climate data show a higher number of significant correlations than for the individual chronologies. PC#1 correlates positively with the mean and minimum temperatures from October prior to growth through the current growth year, and the maximum temperature in the winter season (previous November to April, r = 0.55, p < 0.01) (Figure 5.1.4). The highest correlation was found between PC#1 and the regional annual mean temperature from

5. Temperature reconstructions 54 previous October to current September (r = 0.659, p < 0.01), which was therefore reconstructed by using the PC#1 as predictor variable. Results of response function analyses generally confirm the climate-growth relationships derived from linear correlations for the individual chronologies as well as for PC#1 (Figures 5.1.3 and 5.1.4). However, some response function coefficients between ring width and temperature during winter months show a drastic decrease which points to some multicollinearity between temperature records of consecutive months. Except for March, the response for temperature on tree growth remains positive throughout the annual cycle from November of the year before growth until October of the growth year (Figure 5.1.4). Since our final temperature reconstruction includes temperature over the whole year, the multicollinearity between some temperature series is not problematic. Tree growth near the upper treeline in the study area is mainly influenced by temperature conditions, especially in the winter season (Figures 5.1.3 and 5.1.4). Ring width of high elevation conifers is often reduced by low winter temperatures as a consequence of bud damage, frost desiccation and reduced root activity due to low soil temperature (Körner, 1998). In addition, after cold winters with delayed snow melt, the following vegetation period is shortened; an early start of winter enhances the consumption of stored carbohydrates, which may lead to a reduced earlywood width in the following year (Bräuning, 2001a; Gou et al., 2007a).

5.1.3 Annual Temperature Reconstruction

A linear regression model (Y = 5.182 + 0.351*X) was developed to reconstruct the annual (previous October to current September) temperature history in the central Hengduan Mountains. As shown in Table 5.1.4, the model accounts for 42% of the actual regional temperature variance during the period 1959–2003. The leave-one-out cross-validation method was employed to evaluate the statistical fidelity of this model (Michaelsen, 1987). The model yielded significant verification statistics as measured by sign test and product mean test. The RE value is 0.37, indicating that there is some skill in the derived reconstruction (Fritts, 1976; Table 5.1.4). The temperature reconstruction derived from this model shows good agreement with the actual regional temperature (Figure 5.1.5a).

Table 5.1. 4 Statistics of the leave-one-out calibration results for the common period 1959–2003.

2 2 Period R R R adj F r Sign test Pmt RE 1959–2003 0.659 0.433 0.421 33.0** 0.61 32/13** 2.67* 0.37

R is correlation coefficient; r is the correlation coefficient between the recorded data and the leave-one- out-derived estimates. Pmt is the product mean test. Sign test is sign of paired observed and estimated departures from their mean on the basis of the number of agreement/disagreements; RE is the reduction of error, any positive value indicates there is some sense in the reconstruction (Fritts, 1976). * Significant at p < 0.05; ** Significant at p < 0.01.

5. Temperature reconstructions 55

The annual temperature (previous October to September) reconstructed over the last 250 years is shown in Figure 5.1.5b. Although the reconstruction was based on the residual tree-ring chronologies, considerable decadal-scale temperature variability was retained in the reconstruction. Cold episodes occurred around 1810–1820, 1860–1970 and in the 1960–1980s. Warm periods occurred in the 1780s, 1825–1860, 1930–1960, and from 1990 to present.

Figure 5.1.5 (a) Comparison of the actual (black line) and reconstructed (grey line) annual (previous October through current September) mean temperature for the common period 1959–2003. (b) Reconstructed annual temperature in the central Hengduan Mountain over the past 250 years. The thin line represents the annual value and the thick line was smoothed with an 11-year FFT-filter (Fast Fourier Transform) to emphasize long-term fluctuations. The horizontal grey line is the instrumental regional mean temperature for period 1959–2003.

5. Temperature reconstructions 56

5.2 Tree-Ring-Density Based Summer Temperature Reconstruction

Maximum latewood density (MXD) has been demonstrated to be a promising proxy of summer temperatures (e.g. Briffa et al., 2004; Büntgen et al., 2006). However, the application of such technique in palaeoclimatic studies is still scarce on the Tibetan Plateau (Bräuning and Mantwill, 2004; Wu et al., 2005). Here, a newly developed technique, namely ‘high-frequency densitometry’ (RinnTech, Germany), was employed to process tree-ring density data and to reconstruct summer temperature during the past centuries. Two sites of P. brachytyla near the upper timberline were selected (Figure 5.2.1). The Hongpo (HP_P) site (28.25 °N, 98.91 °E, 3500 m a.s.l.) was located on the west slope of the Baima Snow Mountains. The Bitahai (BT_P) site (27.82 °N, 99.99 °E, 3580 m a.s.l.), ~120 km south-east of the HP site, was located in the Bitahai Nature Reserve. Figure 5.2.2 shows the photographic views of the spruce forests at the two sampling sites.

Figure 5.2.1 Map showing the locations of wood densitometry sampling sites and meteorological stations. Bold letters show the locations of the tree-ring based temperature reconstructions from the south (STP) and eastern (ETP) Tibetan Plateau (Bräuning and Mantwill, 2004), the upper source region of Yangtze River (SRY; Liang et al., 2008) and west Sichuan Plateau (WS; Shao and Fan, 1999).

5. Temperature reconstructions 57

Figure 5.2.2 Photographic views of the two sample sites of Hongpo (a, HP_P) and Bitahai (b, BT_P) in the central Hengduan Mountains, south-western China. Photos: Fan, Z.X. (2006).

Total ring width (TRW) chronologies were first developed for all core samples (130 cores/91 trees) from two sites. As compared with in chapter 4, TRW raw data were detrended here to preserve more low-frequency climate signal. Meanwhile, ring-width chronologies were compared with some more climatic variables (i.e. maximum/minimum temperatures) and their seasonalities. Due to the limitation of wood quality for density procedure, 62 cores (58 trees) with high wood quality (i.e. without wood decomposition by fungal attack) were further processed for density measurement. A regional meteorological data series was developed from monthly records of two nearby high elevation stations Shangri-la (27.83 °N, 98.76 °E, 3276 m a.s.l.) and Daocheng (29.05 °N, 100.3 °E, 3729 m a.s.l.) (Figure 5.2.1), by applying the techniques outlined by Jones and Hulme (1996). The climate-tree growth relationships were investigated by using correlation analyses between tree-ring data and meteorological records for their common period 1958–2004. Simple correlations were calculated between the tree-ring residual chronologies and regional monthly climate variables (precipitation, mean temperature as well as maximum and minimum temperature) from July prior to growth to October of the current growth year. In addition, various seasonal means of climate variables and their correlations with tree-ring data were calculated.

5. Temperature reconstructions 58

Table 5.2. 1 Site information and tree-ring chronologies statistics.

Site Date type Location Elev. C/T Time span MSL SD+ MS+ AC1+ Rbar* EPS > (Lat./Lon.) (m) (A.D.) 0.85*

HP_P TRW 28.25/98.91 3 500 80/51 1688–2005 186 0.28 0.20 0.70 0.59 1780/21 MXD 31/29 1724–2005 179 0.14 0.11 0.51 0.41 1820/15

BT_P TRW 27.82/99.99 3 580 50/40 1623–2007 234 0.22 0.15 0.65 0.48 1690/17 MXD 31/29 1623–2006 235 0.15 0.11 0. 59 0.46 1750/17

RC TRW 130/91 1623–2007 204 0.25 0.18 0.68 0.46 1690/18 MXD 62/58 1623–2006 207 0.15 0.11 0.55 0.39 1750/20

HP_P, Hongpo at Baima Snow Mountain; BT_P, Bitahai Nature Reserve; RC, regional chronology; TRW, total ring width; MXD, maximum latewood density; Lat., latitude; Lon., longitude; Elev., elevation; C/T, cores/trees; MSL, mean segment length; SD, standard deviation; MS, mean sensitivity; AC1, first-order autocorrelation; Rbar, mean inter-series correlation; EPS > 0.85, year/no. of cores when expressed population signal exceeds the 0.85 threshold. + Calculated for ARSTAN standard chronologies. * Calculated for ARSTAN residual chronologies for 30-year intervals with 15-year overlaps.

5. Temperature reconstructions 59

Figure 5.2.3 Comparison between the HP_P (blue) and BT_P (red) ring-width residual chronologies. (a) Expressed population signal (EPS) statistic (calculated over 30 years lagged by 15 years); (b) the two residual chronologies adjusted to the same mean and variance over the common period 1688–2005; (c) the sample depth through time; (d) the 15-year low-pass filtered components. The correlation coefficients were calculated for 1780–2005 when EPS exceeds the threshold of 0.85 for both chronologies.

5.2.1 Chronology Statistics

The two spruce TRW chronologies ranged in length from 318 years at HP_P to 385 years at BT_P site. Based on the EPS statistics, the TRW chronologies met signal strength acceptances after A.D. 1780 for HP_P and A.D. 1690 for BT_P (Table 5.2.1; Figure 5.2.3). The MXD chronologies met the 0.85 EPS criterion for signal strength acceptance after A.D. 1820 (HP_P) and A.D. 1750 (BT_P), respectively (Figure 5.2.4). The chronologies generally showed a low mean sensitivity (MS) and high first-order autocorrelation (AC1), which is typical for spruce trees in humid environments. Compared with the TRW chronologies, the MXD chronologies of P. brachytyla usually displayed lower MS, standard deviations (SD) and AC1 (Table 5.2.1). The mean inter-series correlations (Rbar) ranged from 0.41 to 0.59, which indicated that the developed chronologies contain considerable common signals, and were thus suitable for climate change studies.

5. Temperature reconstructions 60

Figure 5.2.4 Comparison between the HP_P (blue) and BT_P (red) maximum latewood density residual chronologies. (a) Expressed population signal (EPS) statistic (calculated over 30 years lagged by 15 years); (b) the two residual chronologies adjusted to the same mean and variance over the common period 1724–2005; (c) the sample depths through time; (d) their 15-year low-pass filtered components. The correlation coefficients were calculated for 1820–2005 when EPS exceeds the threshold of 0.85 for both chronologies.

Although the distance between these two sites was relatively far (~120 km), significant correlations were found between the paired two site chronologies. TRW chronologies and their 15-year low-pass filter components correlated with r = 0.39 and r = 0.41 respectively over 1780- 2005 period (Figure 5.2.3). The respective correlations between the two MXD chronologies (1820–2005) were 0.44 and 0.61 (Figure 5.2.4). This suggested that spruce trees growing near the upper tree line are influenced by some common regional climate signal. In order to dampen possible site-specific effects on tree growth and to emphasize regional-scale climate signals, all tree-ring index series of TRW and MXD were averaged to form two regional chronologies (RC). The EPS statistics were above the threshold of 0.85 after A.D. 1690 for TRW and A.D. 1750 for MXD regional chronologies (Table 5.2.1).

5.2.2 Growth-Climate Relationships

Temperatures were generally the main factors influencing TRW and MXD for trees growing at humid high-elevation sites (Figure 5.2.5; Figure 5.2.6). The growth-climate responses were

5. Temperature reconstructions 61 very similar for the two site chronologies and generally the regional chronologies showed higher correlations with temperature than the individual site chronologies. For TRW, the positive influence of temperature was predominant, while the importance of precipitation was weak except for the positive impact of precipitation in January and August of the growth year (Figure 5.2.5). For the regional TRW residual chronology, temperatures during the winter season (prior November until current February) played a crucial role (r = 0.43; p < 0.01). Although the influence of temperature in the later growing season was also important, significant correlations only occurred for the current August.

Figure 5.2.5 Climate response of ring width (TRW) for the sites HP_P (white), BT_P (grey) and their regional chronology (RC, black) using (a) maximum temperatures (Tmax), (b) mean temperatures (TEM), (c) minimum temperatures (Tmin), and (d) precipitation sums (PRE). Correlations were calculated from previous year July to current year October over 1958–2004 common periods. Horizontal dashed lines denote the 95% significance levels. Numbers on the x-axis refer to seasonal means of prior November- February (-11/2), prior November-April (-11/4), prior October-September (-10/9), and prior November- October (-11/10), respectively.

5. Temperature reconstructions 62

Figure 5.2.6 Climate response of maximum latewood density (MXD) for the sites HP_P (white), BT_P (grey) and their regional chronology (RC, black) using (a) maximum temperatures (Tmax), (b) mean temperatures (TEM), (c) minimum temperatures (Tmin) and (d) precipitation sums (PRE). Correlations were calculated from previous year July to current year October over 1958–2004 common periods. Horizontal dashed lines denote the 95% significance levels. Numbers on the x-axis refer to seasonal means of April-May (4/5), July-September (7/9), April-September (4/9), April-October (4/10), April- September but excluding June (4/9*), respectively.

Spruce ring-width chronologies from the upper timberline are sensitive to temperature, especially in the winter season (Figure 5.2.5). This was also reported for other high elevation conifers from west Sichuan (Shao and Fan, 1999) and the eastern Tibetan Plateau (Bräuning, 2001a; Gou et al., 2007b; Liang et al., 2008). Winter temperatures have been found to constrain radial growth in different temperate species and ecosystems in eastern North American (Pederson et al., 2004). Warm conditions in the late autumn might increase carbohydrate storages in the stem, and thus enhance earlywood growth in the following spring (Gou et al., 2008b). Cold late-winter conditions, on the other hand, might cause bud damage, frost desiccation, and fine root mortality due to low soil temperature (Körner, 1998). The positive

5. Temperature reconstructions 63

correlations with Tmin in February indicate that low night-time temperatures in the late winter probably cause frost damages and thus have a negative impact on radial growth (Figure 5.2.5). MXD correlated with the temperature variations in the warm season of the current growth year (Figure 5.2.6). Correlations between MXD residual chronologies and monthly mean temperature were significant from April to September except for June. The correlation between the regional MXD residual chronology and warm season (April until September) mean temperature reached 0.64 (p < 0.01), whereas split-summer temperature (April through September, excluding June) produced higher correlations (r = 0.72, p < 0.01). The summer daily maximum temperature (Tmax) showed a stronger influence on MXD than the night-time minimum values (Tmin). Precipitation in the growing season has mostly negative influence on MXD although positive correlations occur in August. Since the common signals between different TRW site chronologies were low (Figure 5.2.3), more ring-width series are needed to develop a robust winter temperature reconstruction. Therefore, only warm season temperature was reconstructed using the MXD chronologies. Compared with TRW data, the MXD chronologies show a consistent response to a wider window of warm season temperature (Figure 5.2.6). Similar results were reported by various studies and thus MXD was used for reconstructing summer temperature variations in the boreal and temperate climate zones (D’Arrigo et al., 1992; Schweingruber et al. 1993; Davi et al., 2003) and in subtropical mountain regions (Hughes, 2001; Davi et al., 2002; Bräuning and Mantwill, 2004). Temperature conditions influence the number and size of latewood cells, and needle mass which determines the amount of photosynthates available for cell wall thickening during the late growing season (Hughes, 2001; Veganov et al., 2006). Warmer temperature during late summer can contribute to latewood cell wall thickening and thus lead to a denser latewood. In addition, latewood formation appears to benefit from photosynthate production and elevated hormone levels in the early growth season (April-May) (Conkey et al., 1986; Veganov et al., 2006). TRW at individual sites correlated with MXD chronologies over 1820–2005 period (r = 0.12 for HP_P, 0.50 for BT_P and 0.38 for RC, respectively), which indicates that particular years with narrow rings show corresponding lower values of latewood density. Annual rings formed under unfavourable environmental conditions consist of only several tracheids in a radial dimension with radial sizes smaller than usual and with thinner cell walls.

5.2.3 Warm Season Temperature Reconstruction

Based on the results of the growth-climate analyses, a transfer function was calculated based on a linear regression model, using the prewhitened mean summer temperature as the dependent variable and the regional MXD residual chronology as the independent variable. In order to reflect temperature variability for the whole warm season, April to September mean temperature was reconstructed for the study area. The reconstruction accounted for 40.9% of the actual summer temperature variance (Table 5.2.2), and agrees well with variations of the actual

5. Temperature reconstructions 64 temperature during their common period 1958–2004 (Figure 5.2.7a). The reconstructed summer temperature series covered the period A.D. 1750–2006 of reliable internal signal strength (Figure 5.2.7b), based on the sample depths (> 20 cores) and commonly acceptable EPS statistics (> 0.85; Wigley et al., 1984) of the regional MXD chronology (Table 5.2.1).

Table 5.2. 2 Statistics of calibration and leave one-out verification results for the common period 1958– 2004.

Calibration (Model: Y = 6.076 + 4.419X) 2 2 Period R R Radj SE 1958–2004 0.64 0.409 0.396 0.35

Standard Verification 1st Difference Verification Period R ST Pmt RE R ST Pmt RE 1958–2004 0.60** 33/14** 3.31** 0.36 0.80** 35/11** 4.78** 0.63

** Significant at p < 0.01. R, correlation coefficient; SE, standard error of the estimate; ST, sign test, which counts the number of agreements and disagreements between the reconstructed and the instrumental climatic data; Pmt, product mean test; RE, reduction of error. Any positive value of RE indicates that there is confidence in the reconstruction (Fritts, 1976).

The leave-one-out cross-validation method was employed to verify our reconstruction, since the meteorological data set available was too short to carry out a robust split-sample calibration (Michaelsen, 1987). Evaluative statistics include the Pearson’s correlation coefficient (R), reduction of error (RE), product mean test (Pmt) and sign test (ST) (Fritts et al., 1976). The cross-validation test yielded a positive RE (0.36), indicating predictive skill of the regression model (Table 5.2.2). Statistically significant sign test and product mean test between the recorded data and the leave-one-out-derived estimates were additional indications for the reconstruction’s validity. When 1st differences of the actual and leave-one-out estimated series were used, even better verification statistics were obtained, indicating that the developed model was successful in tracking the high-frequency temperature variability over the calibration period. Warm season temperature reconstruction, as derived from MXD, contains inter-annual to multi-decadal temperature variability. Remarkable periods of high summer temperature occurred during 1750s, 1820–1850s, 1880–1890s, 1930–1950s and 1990–present. On the opposite, the periods 1790–1810s, 1860–1870s, 1900–1920s and 1960–1985 were relatively cold (Figure 5.2.7b).

5. Temperature reconstructions 65

Figure 5.2.7 (a) Comparison between actual and estimated mean warm season (April to September) temperature for their common period 1958–2004; (b) warm season temperature reconstruction for the central Hengduan Mountains derived from maximum latewood density. Thin line represents annual values; the bold line was smoothed with a 15-year low pass filter.

5.2.4 Spatial Correlation Analysis

To demonstrate that the reconstruction and instrumental records reflect regional-scale temperature variability, spatial correlation analyses were taken between these data and the CRUts 2.1 dataset (Mitchell and Jones, 2005) of all grid cells available for a user-defined region. The analyses were achieved using the KNMI climate explorer (Royal Netherlands Meteorological Institute; http://climexp.knmi.nl). Instrumentally recorded warm season temperatures in our study area (two stations means), as well as reconstructed summer temperatures, correlated significantly with gridded surface temperatures on a regional scale (Figure 5.2.8). Spatial correlation fields were similar for the instrumental and reconstructed temperature variability, although correlations were lower for the latter. The highest correlation fields were confined to the north-south orientated Hengduan Mountain ranges. These results indicate that our reconstruction captures a great part of the regional temperature variability of the Hengduan Mountains and the south-eastern Tibetan Plateau.

5. Temperature reconstructions 66

Figure 5.2.8 Spatial correlations of (a) instrumental and (b) reconstructed summer (April-September) temperatures with regional gridded April-September temperatures for the period 1958–2001. The analyses were performed using the KNMI climate explorer (Royal Netherlands Meteorological; http://climexp.knmi.nl), the gridded climate dataset was developed by the Climatic Research Unit (Mitchell and Jones, 2005; CRUts 2.1).

5. Temperature reconstructions 67

Figure 5.2.9 Graphical comparison of various temperature reconstructions for the western China derived from tree-ring records. (a) warm season (April-September) mean temperature reconstruction in the central Hengduan Mountains (CHM; this study); (b) late-summer (August-September) reconstruction in the southern Tibetan Plateau (STP; Bräuning and Mantwill, 2004); (c) late-summer temperature reconstruction in the eastern Tibetan Plateau (ETP; Bräuning and Mantwill, 2004); (d) summer (June to August) minimum temperature reconstruction in the upper source region of Yangtze River (SYR; Liang et al., 2008); (e) winter half year temperature reconstruction from west Sichuan Plateau (WS; Shao and Fan, 1999). All series were adjusted for their long-term means over period 1750–1994, and smoothed with a 15-year low-pass filter to emphasize long-term fluctuations.

5. Temperature reconstructions 68 5.2.5 Comparison with Regional Records

Several tree-ring based temperature reconstructions in surrounding areas have recently been developed (Shao and Fan, 1999; Bräuning and Mantwill, 2004; Liang et al., 2008). Based on MXD data of multi-species from 22 temperature sensitive sites, four growth regions were outlined from northeast to southwest of the Tibetan Plateau (Bräuning and Mantwill, 2004). Late summer (August-September) temperature reconstructions from the eastern (ETP) and southern (STP) region of Tibet (Figure 5.2.1, Figure 5.2.9) of those four growth regions are compared with our reconstruction. Shao and Fan (1999) developed four Balfour spruce (Picea likiangensis var. balfouriana) TRW chronologies and reconstructed winter (prior December to February) minimum temperature in west Sichuan (WS in Figure 5.2.1, Figure 5.2.9). Their calibration model explained 58% of the actual temperature variances for 1960–1993. Using four TRW series of Balfour spruce, Liang et al. (2008) developed a summer (June-August) minimum temperature reconstruction (R2 = 0.27 for 1957–2002) for the upper source region of the Yangtze River on the Tibetan Plateau (SRY in Figure 5.2.1, Figure 5.2.9). Our reconstruction mirrors similar warm/cold intervals as temperature reconstructions in the nearby regions (Figure 5.2.9). Cold summers of 1810s, 1860s, 1900–1910s, and 1970s found in the present study are consistent with low summer temperatures in southern Tibet (Bräuning and Mantwill, 2004) and the source region of Yangtze river (Liang et al., 2008). Cold condition in the 1800s, 1900s and 1970s are also reported for the northeastern Tibetan Plateau (Gou et al., 2008b), the Eastern Himalayan Region (Bhattacharyya and Chaudhary, 2003) and Nepal (Cook et al., 2003). The warm 1820–1850s, 1880s, 1930–1950s are consistent between the compared records (Figure 5.2.9), and were also reported as warm periods by Wu et al. (1988) in the Hengduan Mountain region, and in Kashmir (Western Himalaya; Hughes, 2001). Some differences existing between the reconstructions (i.e. in the 1760s, 1800s) may reflect the local influence of different geographic features or difference in seasonality of the various temperature reconstructions.

6. Drought reconstruction 69

6. Drought Reconstruction

6.1 Data Source and Methods

In this chapter, four tree ring-width chronologies were selected from low-elevation sites to reconstruct regional drought variability during the past 350 years. The site chronologies come from three coniferous tree species, i.e. Picea likiangensis Pritz, Tsuga dumosa (D. Don) Eichler and Abies ernestii Rehd. In total, increment cores from 93 trees (136 cores) from four sites on the west facing slopes of the Baima Snow Mountains entered the analyses (Figure 6.1). Figure 6.2 shows the photographic views of the sampling sites. The raw ring-width series were standardized using the double detrending method described in chapter 3. Autoregressive modelling was used to remove persistence from each series, producing pre-whitened ‘residual’ indices.

Figure 6.1 Map of the sampling sites in the central Hengduan Mountains, south-western China.

6. Drought reconstruction 70

Figure 6.2 Photograpic views of the tree-ring sampling sites at Yanmen (YE_T and YM_A; a) and Tacheng (GK_P and YC_T; b) in the central Hengduan Mountains. Photos: Fan Z.X. (2006).

In order to assess the similarity among the four site chronologies, correlation and principal component analyses (PCA) were applied over the common period 1655–2005. The two grid points (26.25 ºN, 98.75 ºE and 28.75 ºN, 98.75 ºE) of the global net of Palmer drought severity index (PDSI; Dai et al., 2004) data next to our sampling sites were used for detecting the growth response to moisture conditions. A regional series of climate data was created from the two meteorological stations for their common period of 1961–2000, and from the two PDSI grid points for 1951–2000. Pearson correlation coefficients were calculated between the ring width chronologies and the monthly series of temperature and precipitation from both Weixi (1961– 2000) and Deqin station (1957–2000), for a 15 month period ranging from July of the summer prior to growth until September of the growth year. Bootstrapped correlation and response functions were performed to analyze the relationships between PC#1 and regional climatic variables (Biondi and Waikul, 2002). PC#1 was compared with the regional climate series (precipitation, temperature, relative humidity and PDSI) for a 15–month period from July of the summer prior growth to September of the growth year.

6.2 Chronology Statistics

Descriptive statistics of the four standard chronologies are shown in Table 6.1. All chronologies exceed 460 years and one 613 years old tree (Tsuga dumosa) was found at YE_T site. The average mean sensitivity (a measure of the inter-annual variability in tree ring series) at all sites was around 0.18, indicating that the ring-width chronologies show relative low inter-

6. Drought reconstruction 71 annual variability, which is characteristic for trees growing in humid environments. First-order autocorrelation ranged from 0.66 to 0.71, which documents that the chronologies contain low- frequency variance generated by climate and by tree physiological lag effects. The latter include needle retention and the remobilization of stored food reserves for growth in the following spring (Fritts, 1976). The Rbar and EPS statistics of the site chronologies signal strength ranged from 0.28 to 0.37 and from 0.77 to 0.87, respectively. Three site chronologies meet the 0.85 EPS criterion after AD 1655, except for site YE_T whose EPS was above 0.75 at 1655 and reached the 0.85 limit only after 1685 (Figure 6.3d). Nevertheless, we regarded the period 1655–2005 (351 years) as common period of acceptable chronology quality for further analyses.

Figure 6.3 The four residual chronologies from Baima Snow Mountains, NW Yunnan. The expressed population signal statistic (EPS) through time is shown in the upper section of each graph. Thin lines represent annual values; bold lines were smoothed with an 11-year-FFT-filter (Fast Fourier Transform) to emphasize long-term growth variations. The site codes match with those of Table 6.1.

6. Drought reconstruction 72

Table 6. 1 Site locations and chronology statistics.

Site Lat./Lon. Elev Trees Period MSL AGRa MSb AC1b Rbarc EPSc (degree) (m) (cores) (year) (mm)

GK_P 27.58/99.35 3240 28 (38) 1429–2005 285 1.30 0.19 0.69 0.31 0.85 YC_T 27.59/99.29 3150 32 (49) 1542–2005 222 1.34 0.19 0.71 0.28 0.85 YM_A 28.04/99.02 3200 17 (19) 1489–2005 273 1.09 0.18 0.66 0.37 0.87 YE_T 28.04/98.98 3100 16 (30) 1393–2005 297 1.08 0,17 0.69 0.28 0.77

GK_P: Picea likiangensis; YM_A: Abies ernestii; YC_T and YE_T: Tsuga dumosa Lat: latitude, Lon: longitude; Elev: Elevation; MSL: mean segment length (tree age); AGR: average growth rate; MS: mean sensitivity; AC1: first-order auto-correlation; Rbar: mean inter-series correlation; EPS: expressed population signal a calculated for raw ring-width values b calculated for ARSTAN standard chronologies c calculated for ARSTAN standard chronologies for 30–years intervals with 15–year overlaps

6.3 Correlation and Principal Component Analysis

Despite of the different species analyzed, the four chronologies correlated to each other significantly over the common period 1655–2005, with correlation coefficients ranging from 0.40 to 0.55 (p<0.01) (Table 6.2). Ring-width patterns were very similar among the four chronologies, especially concerning high and middle frequency growth variations (Figure 6.3). The average cross-chronology correlation (calculated for a 30–year moving window) was 0.47 (p < 0.001) over 1655–2005, and ranged from 0.20 to 0.66 (Figure 6.7a). The principal component analysis of the four residual chronologies showed that only the eigenvalue of PC#1 was greater than one and that PC#1 accounts for 60.5% of the total variance. The four chronologies showed common positive loadings on PC#1: 0.78 for GK_P, 0.79 for YC_T, 0.75 for YM_A and 0.79 for YE_T, respectively. Therefore, PC#1 reflects the common growth response to regional climatic variations, and the score of PC#1 can be used to evaluate the regional climate-growth relationships and to indicate regional climate variability. Other PCs were insignificant, and can probably be assigned to small-scale growth variations caused by site-specific growth reactions or local growth disturbances.

Table 6. 2 Correlation matrix of the four ring-width chronologies for the common period 1655–2005. The site codes are identical with those in Table 6.1. All correlations are significant at p < 0.01.

YC_T YM_A YE_T

GK_P 0.55 0.44 0.44 YC_T 0.40 0.49 YM_A 0.52

6. Drought reconstruction 73

6.4 Climate-Growth Responses

Correlation analyses indicated that tree growth was mainly affected by early spring precipitation, especially during January, March and May (Figure 6.4). The correlation coefficients between regional seasonal precipitation (March to May) and the residual ring-width index chronologies were significant (p < 0.05) except for site YE_T (r = 0.53 for GK_P, 0.47 for YC_T, 0.37 for YM_A, and 0.29 for YE_T). Compared with the other two species, the radial growth of Tsuga dumosa showed a weaker (positive) spring-precipitation response, but stronger (negative) response to temperatures (Figure 6.4c, d). When standard chronologies instead of residual chronologies are compared with climate parameters, we found the same general patterns but lower correlations (not shown). The correlation coefficients were mostly positive between the PC#1 of the four residual chronologies and regional monthly precipitation, relative humidity and PDSI (Figure 6.5a, c, d), but negative between PC#1 and regional monthly temperature (Figure 6.5b). Response function analyses indicated that the water availability in March and April was a crucial factor impacting radial growth. The highest correlation between PC#1 and PDSI occurred for the season March to May (r = 0.65, p < 0.01) which was therefore reconstructed by using PC#1 of the four residual chronologies as predictor variable.

Figure 6.4 Correlation coefficients between radial growth and monthly mean temperature (dot-lines) and total monthly precipitation (columns) at nearby meteorological stations of Deqin (gray) and Weixi (white). Correlations are computed from previous year July to current year September over 1957–2000 for Deqin and 1961–2000 for Weixi. Horizontal dashed lines denote the 95% levels of significance.

6. Drought reconstruction 74

Figure 6.5 Correlation (columns) and response function (dot-lines) coefficients between the first principal component (PC#1) of the residual chronologies and regional monthly a) total precipitation, b) mean temperature, c) relative humidity for 1961–2000, and d) PDSI data for 1951–2000. The coefficients were calculated from previous year July to current year September. Horizontal dashed and bold lines denote the 95% and 99% significance levels, respectively. The asterisks indicate the 95% significance level for response functions based on bootstrapping tests.

In southern Tibet, dry and warm conditions before the onset of the summer monsoon cause drought stress to the trees and are thus limiting growth (Bräuning, 1999b; Bräuning and Grießinger, 2006). Thus, tree growth benefits from former winter and current spring precipitation, which increase the soil moisture content during the early part of the growing season. Later on, after the onset of the summer monsoon season, enough moisture is available to satisfy the water

6. Drought reconstruction 75 demand of the trees. Most of the correlations between tree growth and temperature were negative, especially at the two Tsuga dumosa sites. High temperatures during the growing season enhance evapotranspiration and thus decrease soil moisture availability (LeBlanc and Terrell, 2001). This growth response is not surprising for trees growing on steep slopes in a subtropical climate. Our study sites are located at middle elevations, far away from the upper tree line which is situated around 4200 meters a.s.l. in the Baima Mountain region.

6.5 Drought Reconstruction

A linear regression model (Y = 0.818X – 0.0724) was developed to reconstruct the drought history for the central Hengduan Mountain region. During the common period of tree rings and PDSI data (1951–2000), the reconstruction accounted for 42% of the actual PDSI variance (Table 6.3). The spring PDSI estimates derived from this model were in high agreement with the yearly departures from the long-term mean in the observed data (Figure 6.6). Split-sample calibration-verification and leave-one-out cross verification methods (Michaelsen, 1987) were employed to evaluate the statistical fidelity of this model. The reduction error (RE) and the coefficient of efficiency (CE) were positive; indicating significant skill in the tree-ring estimates (Fritts, 1976). The results of the sign test and product mean test demonstrated the validity of the regression model (Table 6.3).

Table 6. 3 Statistics of calibration-verification test results for the common period 1951–2000.

Split-sample calibration-verification Calibration Verification

Period R R2 F Period Sign test Pmt RE CE

1951–1980 0.65 0.42 20.4** 1981–2000 14/6** 2.50** 0.318 0.298 1981–2000 0.68 0.47 15.9** 1951–1980 22/8* 2.51** 0.374 0.359 1951–2000 0.65 0.42 34.5**

Leave-one-out verification 1951–2000 0.61 38/12** 2.75** 0.372

*Significant at p < 0.05, ** significant at p < 0.01 R is correlation coefficient; Sign test is sign of paired observed and estimated departures from their mean on the basis of the number of agreements/disagreements; Pmt is product mean test; RE is the reduction of error, any positive value indicates that there is some sense in the reconstruction (Fritts, 1976); CE is the coefficient of efficiency.

6. Drought reconstruction 76

Figure 6.6 Actual (grey) and reconstructed (dark) March to May PDSI during their common period 1951–2000. The estimation explains 42% of the actual PDSI variance in this common period.

Figure 6.7 a) Average cross-chronology correlation (mean is 0.47) of the four residual chronologies, calculated for 30–year periods with 15–year overlaps. b) The reconstruction of March-May PDSI in the central Hengduan Mountain region over the past 350 years. The thin line represents the annual value and the thick line was smoothed with an 11–year FFT-filter (Fast Fourier Transform) to emphasize long-term fluctuations. The grey line indicates the ±2SD values.

The spring (March–May) PDSI reconstruction (Figure 6.7b) showed that wet periods prevailed in the 1690s, 1715–1730, 1750s, 1780s, 1825–1850,1900s, 1930–1960, and 1990– present. Extremely wet years (≥ 2 SD) occurred in AD 1703, 1824, 1835, 1940, and 1998. In

6. Drought reconstruction 77 contrast, the intervals AD 1700–1715, 1733–1745, 1790–1820, 1860–1890, 1910–1925, and 1960–1990 were relatively dry. Extremely dry years (≤ 2 SD) were more frequent than extremely wet years and were concentrated during AD 1730–1800 and 1870–1900, and after 1980, respectively. From the higher mean correlation between the four individual chronologies (Figure 6.7a), it can be concluded that these were time periods of strong climatic forcing of regional tree growth patterns. Particularly dry years occurred in AD 1670, 1706, 1735–1736, 1757, 1766, 1772, 1792, 1800, 1820, 1870, 1887, 1897, 1987, and 1999.

6.6 Spectral Analysis

Multi-taper method (MTM) spectral analysis (Mann and Lees, 1996) was used to evaluate frequency domains in local drought variability (Figure 6.8). A decadal broadband power was identified at 10 years (p < 0.05), which resembles other findings from Mongolia and northern China and suggests an influence of solar forcing (Pederson et al., 2001; Li et al., 2006). Significant high frequency peaks were found at 2.3 years and 5.2–5.5 years (p < 0.01), as well as at 2.1, 2.8, and 3.0 years (p < 0.05). These periods fall within the range of variability of the El Niño–Southern Oscillation (ENSO). The latter has been found to have strong, but variable influence on the strength of the South Asian summer monsoon system (e.g. Charles et al., 1997; Kane, 2006). Thus, like in other annually resolved climate archives from the Tibetan Plateau (e.g. in ice cores) (Yang et al., 2000), ENSO variability might be reflected in the long-term growth curves of moisture-sensitive trees.

Figure 6.8 Multi-Taper method (MTM) power spectra of the reconstructed spring PDSI. The bold line indicates the null hypothesis; the dashed, dash-doted, and dotted lines indicate the 90%, 95% and 99% significance levels, respectively. Numbers associated to peaks indicate the periodicity of the signals.

6. Drought reconstruction 78

Figure 6.9 Spatial correlations of (a) actual and (b) reconstructed spring (March-May) PDSI with concurrent GPCC gridded precipitation data set for the period 1951–2000. The analyses were performed using the KNMI Climate Explorer.

6.7 Spatial Correlations and Teleconnections

Actual and reconstructed spring (March-May) PDSI were correlated with regional monthly gridded precipitation data set developed by the Global Precipitation Climatology Centre (GPCC,

6. Drought reconstruction 79

Fuchs et al., 2007). As shown in Figure 6.9, regional spring drought variability (i.e. PDSI) highly correlates with gridded surface precipitation in the Hengduan Mountains and nearby regions. Similar spatial patterns are found for the correlations between reconstructed PDSI and surface precipitation data (Figure 6.9b). These results indicated that the present reconstruction represents drought variability on a regional scale. In addition, the correlations are reversed for the northern Tibetan Plateau, which may reflect their differences of rainfall sources.

Figure 6.10 Spatial correlation of reconstructed PDSI with global sea surface temperature data set (ERSST.v2) during prior winter (a, Nov-Feb) and pre-monsoon (b, Mar-May) seasons for the period 1951–2000. The analyses were performed using the KNMI Climate Explorer (http://climexp.knmi.nl).

In order to detect large-scale forcing effects on the response of local moisture availability, reconstructed spring PDSI were correlated with global sea surface temperatures from the NCDC data set (ERSST.v2) developed by Smith and Reynolds (2004). As shown in Figure 6.10, negative correlation correlations were found between the reconstructed spring PDSI and the equatorial Pacific SST. In other words, warmer conditions in the tropical ocean during the pre-

6. Drought reconstruction 80 monsoon season correspond to drought conditions in the study region. Buckley et al. (2007) suggest that anomalously warm SST in the tropical Pacific, specifically in the area of Nino 3.4, may be the major contributing factor to decadal scale drought over north-western Thailand. Although the correlations were rather weak, the results suggested that El Niño-like conditions may strengthen droughts over the region in some extent. However, reconstructed PDSI was also negatively correlated with SST in the Indian Ocean, especially in the Bay of Bengal and the area of Indonesia. This suggests that ENSO is not the only factor contributing to rainfall variability in the study region, and therefore growth variability of moisture-sensitive trees.

7. Larch radial growth and insect defoliation 81

7. Larch Radial Growth and Insect Defoliation

7.1 Impact of Insect Defoliation on Tree Growth

Although tree-ring data have been widely used as climate proxies, many studies have demonstrated the complexity of the relationship between tree growth and other environmental parameters, such as insect infestations, forest fires, geomorphodynamics (Schweingruber, 1996). Outbreaks of defoliating insects can affect forest carbon dynamics by reducing growth and increasing tree mortality over large areas (Kurz et al., 2008). The study of the long-term history on outbreak frequency, severity and extent over broad temporal and spatial scales is essential to understand the dynamics of ecosystems (Pickett and White, 1985; Parish and Antos, 2002; Zhang and Alfaro, 2003), and to evaluate the accuracy of tree-ring-based climatic reconstructions (Trotter et al., 2002; Caccianiga et al., 2008). Dendrochronological methods can be used to analyze growth reductions in annual rings and changes in morphology that are indicative of defoliation, and thus to reconstruct the history of insect outbreaks (e.g. Schweingruber, 1979; Swetnam and Lynch, 1993; Speer et al., 2001; Esper et al., 2007). A number of insect species have been documented to influence radial growth in trees, such as the eastern spruce budworm (Choristoneura fumiferana Clem.; Swetnam et al., 1985; Boulanger and Arseneault, 2004; Campbell et al., 2006; Fraver et al., 2007), forest tent caterpillar (Malacosoma disstria Hubner; Hogg et al., 2002; Huang et al., 2008), larch sawfly (Pristiphora erichsonii (Httg.); Girardin et al., 2001) and larch bud moth (Zeiraphera diniana; Weber, 1997; Rolland et al., 2001; Nola et al., 2006). In the European Alps, Zeiraphera diniana is a common defoliator feeding on sub-alpine larch forests (Baltensweiler et al., 1977). In years of high population densities, larvae may completely defoliate larch trees, but defoliation rarely results in tree mortality. Historical records of defoliation documented that outbreaks have recurred with a high degree of regularity every 8–9 years since the 1850s (Baltensweiler and Rubli, 1999; Dormont et al., 2006). As a result of insect-defoliation, changes in wood anatomy are evident in the years of damage, e.g. reduction in cell-wall thickness in the latewood causes so-called ‘light rings’ to be formed (Schweingruber, 1979). Periodic outbreaks of larch bud moth (Zeiraphera diniana) have been widely studied by means of host/non-host comparison (Nola et al., 2006) or examination of the wood anatomy in pointer years (Weber, 1997; Rolland et al., 2001) and latewood density variations (Esper et al., 2007) in the European Alps. Abrupt reduction of radial growth is found synchronously in multiple trees of Larix potaninii at various sampling sites in the central Hengduan Mountains. An examination of wood density of larch trees from Baima site (Figure 3.1, Figure 7.1) revealed that narrow rings occur in combination with a significant depression of maximum latewood density (MXD). For example, MXD was reduced abruptly in 1946 and 1983 whereas a relatively gradual but remarkable decrease occurred in 1926, 1964 and 1972 (Figure 7.1). The reduction of ring width and wood

7. Larch radial growth and insect defoliation 82 density continues for three or more years, and after then growth recovers gradually. Such changes in tree-ring growth and wood structure indicates possible insect-defoliation (outbreaks) occurring in those years, especially as in 1946 and 1983. However, such information is rarely documented in the region of the Hengduan Mountains. There are very few records about insect outbreaks in the study area, since the forestry service was not established until 1991. Dendrochronological techniques, i.e. the comparison of host and non-host tree-ring indices, may add to our knowledge about the history of insect defoliations, and to reconstruct the frequency/intensity of outbreaks of this larch defoliator.

7.2 Detection of Larch Insect Defoliation Signals

Larix trees from four sites were sampled at the north-south oriented mountain ranges, namely the Baima Mountain (BM), Bitahai Nature Reserve (BT), Haba Snow Mountain (HB) and Yulong Snow Mountain (YL), respectively. For each site, increment cores were collected from larch trees (host, Larix potaninii) and non-host trees available nearby the larch stands (Abies georgei at BM and HB site, Picea brachytyla at BT and P. likiangensis at YL site) (Figure 3.1, Table 3.1). Ring width was measured for all samples and maximum latewood density was processed for Larix potaninii at Baima site (see methods in chapter 3). To remove the age/size related trend and other low frequency trends not related to defoliation or climatic events, the individual tree- ring series were detrended with a flexible curve of a 60–yr cubic spline with 50% frequency- response (Cook and Peters, 1981). This means that almost all variance within the individual series at time scales longer than two decades were removed. A tree-ring index was obtained by

7. Larch radial growth and insect defoliation 83 calculating the ratio between the actual measurements and the fitted splines. The subsequent detrended series were then averaged to form a site chronology (Cook and Kairiukstis, 1990). To identify potential outbreaks, a host/non-host comparison was conducted using the program OUTBREAK (Swetnam and Lynch, 1993; Holmes and Swetnam, 1996; Speer et al., 2001). The program compares host and non-host chronologies to identify the occurrence of insect attacks. In this analysis, climatic variation contained in each host index series was removed (“corrected”) by subtracting the non-host index series. Before proceeding with subtraction, the variance in both series was adjusted for mean and standard deviation differences and each index value in the non-host chronology exceeding 1.0 was raised to a fractional power of 0.3. This adjustment is intended to suppress the effect of increased radial growth in the non- host chronology which would introduce apparent growth reductions that have been incorrectly diagnosed as potential outbreaks (Holmes and Swetnam, 1996). The program sequentially reads the corrected index values for each tree chronology, and identifies years and periods that meet or exceed some assigned parameters. Identification of insect-defoliation signature requires the following parameters simultaneously: (1) tree-ring (TRW/MXD) indices ≥ 1.28 standard deviations below the mean; (2) growth reduction of this magnitude lasting at least for 3 years; (3) index value for the first year of the inferred outbreak ≤ 60% of the previous year. The minimum length of outbreaks was fixed at three years to eliminate possible effects of individual tree responses to a yearly, specific environmental condition. It has also been reported that larch cannot survive more than three years of complete defoliation and more than eight years of moderate defoliation (Girardin et al., 2001). Because outbreak reconstructions often show only some percentage of trees recording an outbreak, a threshold with at least 30% of the infected trees was considered to be an outbreak. The study of growth periodicity was carried out by means of spectral analysis. Multi-taper method (MTM) was performed to evaluate significant dominant periods of the host and non-host standard chronologies (Mann and Lees, 1996).

7.3 Outbreaks History Inferred from Tree-Ring Data

The maximum latewood density index at Baima site decreased abruptly (≥ 2SD) during 1890–1894, 1926–1930, 1947–1949, and 1984–1987 (Figure 7.2a). These reductions occurred synchronously in most individual trees, during that more than 40% of trees are identified by the program OUTBREAK under the assigned criteria without any corrections. These outbreaks inferred from MXD data are consistent with four major outbreaks (> 50% trees infested) inferred from the TRW dataset but controlled by a master chronology of non- host species (BM_A; Abies georgei). The medium depression of latewood density during 1964– 1968 was not detected by host/non-host comparison. So it might not to be caused by insect defoliation but the result of cold environments during that period as recorded from meteorological data and temperature reconstructions (see Figure 5.1.5 and Figure 5.2.7). The

7. Larch radial growth and insect defoliation 84 results confirm that the host/non-host comparison is successful in extracting insect defoliation events on larch trees in this area. Other defoliation events can be inferred at Baima site, i.e.1728– 1733, 1746–1751, 1803–1812, 1840–1845 and 1856–1858.

Figure 7.2 (a) Standardized maximum latewood density chronology of Larix potaninii at Baima site, dashed line indicates 2 standard deviations; (b) Sample depths (line) and percentage (bars) of larch trees shown reduction of MXD as identified by program OUTBREAK without any corrections; (c) Comparison of TRW chronologies between the host (BM_L, Larix potaninii) and non-host (BM_A, Abies georgei) species at the Baima site; (d) Sample depths (line) and percentage (bars) of larch trees with ring- width growth reduction as identified by program OUTBREAK (bars) controlled by the non-host (Abies georgei) master chronology. Shade area line indicates the 30% threshold.

7. Larch radial growth and insect defoliation 85

Figure 7.3 (a, c, e) Ring-width chronologies of larch (blue lines) and non-host species (orange lines) at three sites in the central Hengduan Mountains; (b, d, f) Sample depths (lines) and percentages (bars) and of larch trees with ring-width reduction as identified by program OUTBREAK controlled by non-host chronologies. Shaded area indicates the 30% threshold.

7. Larch radial growth and insect defoliation 86

Compared with the non-host chronologies, most larch trees growing at the other three sites (BT_L, HB_L and YL_L) show periodic reduction in radial growth, which may be associated with outbreaks of a larch defoliator (Figure 7.3). Synchronously growth reductions can be found for low-elevation sites, such as during 1783–1790, 1804–1809, 1827–1834, 1846– 1850, 1873–1878, 1900–1909, 1939–1945, 1974–1977 and 1981–1986. However, these periods of outbreaks differ from those of high-elevation site at Baima (compare Figures 7.2 and 7.3) The Multi-taper method (MTM) spectral analyses for the four pairs of host and non-host standard chronologies are shown in Figure 7.4. The frequency domain of the larch chronology at Baima site (BM_L) is significant at 7–9 years, while the other three larch chronologies (BT_L, HB_L and YL_L) present peaks at 11–13 year periods that exceed the 99% confidence limit. However, no significant frequency domains within the same periodicity can be found from the paired non-host chronologies (BT_P, HB_A and YL_P). The spectral analyses reveal that non- host (fir, spruce) trees were at least not infested by the same insect species of larch defoliator. Evidences from tree-ring studies have proven that larch trees (Larix potaninii) growing in the study area have been infected by insect defoliator repeatedly. As compared with TRW data, MXD data from Baima site show more significantly depression during the past century (Figure 7.2). Comparison of host ring-width indices and the paired non-host master chronology detected synchronous growth reductions across different sampling sites. However, the frequencies and extents of inferred outbreaks at high-elevation site of Baima are far from consistent with the other three investigated sites. The harsh environmental conditions at high elevations may regulate the stability of insect populations. Spectral analyses indicate that outbreaks occurred regularly every 11-13 years at lower elevation sites (Figure 7.4). However, these conclusions are preliminary due to the lack of documental reports on larch insect defoliation history and the methodologically induced uncertainties.

7. Larch radial growth and insect defoliation 87

Figure 7.4 Multi-taper method (MTM) power spectra of the ring-width standard chronologies at the Baima (a, b), Bitahai (c, d), Haba (e, f) and Yulong (g, h) sites. The blue line indicates the null hypothesis, the cyan, green and red lines indicate the 90%, 95% and 99% significance level, respectively.

8. General discussion and recommendations 88

8. General Discussion and Recommendations

Climate change has been accelerating glacier retreats, treeline advances and ecosystem vulnerabilities in many mountain systems in the world (e.g. Barnett, et al., 2005; Aizen et al., 2007; Devi et al., 2008; Batllori and Gutiérrez, 2008). Glacier retreats were observed to be very fast during the past century in the central Hengduan Mountains (Figure 2.7; He et al., 2003; Baker and Moseley, 2007). Climate change, along with fire suppression policies, was supposed to accelerate advance of the upper treeline and loss of biodiversity in high-altitude alpine meadows (Figure 2.7; Bao et al., 2001; Baker et al., 2005). High climatic variability may result in potential impacts on the socioeconomic development in this ecologically complex region. To evaluate the influences of future climatic changes on vulnerable mountain ecosystems, a thorough understanding of past climate changes is needed in a context of long-term perspective. It has been demonstrated that tree rings are a highly suitable proxy for understanding the historical environmental and climate variability. High elevation conifers in the central Hengduan Mountains have shown to be an excellent source of paleoenvironmental records because they are highly sensitive to climatic and environmental variations. In addition, the forests at most of the subalpine area have been rarely disturbed by logging and other human induced disturbances, which allow enhancing the climatic signals embodied in the proxy record. The present study contributes tree-ring proxy data for the study of climatic and environmental variability across a wide range of spatial and temporal scales.

8.1 Growth-Climate Relationships

Eight tree-ring width chronologies were developed from high-elevation coniferous stands (fir and spruce) along an elevation gradient in the central Hengduan Mountains (chapter 4). A profound examination of growth-climate relationships was taken with respect to different species and environmental conditions (chapter 4). In order to emphasize the spatial representation and homogeneity of meteorological data which is necessary for a detailed growth-climate analysis, we developed a regional climate data series from a high resolution (0.5°× 0.5°) gridded dataset of CRUts2.1 (Climatic Research Unit; Mitchell and Jones, 2005). As indicated by correlation analyses and PCA, tree-ring chronologies from different altitudinal belts can be distinguished by different climate-growth relationships. Winter temperatures are found to be the most consistent climatic factor limiting radial growth of fir and spruce growing at high to middle-elevations. In general, the climate-growth relationships found at different altitudes in this region resemble the reaction patterns found in other humid mountain systems on the Tibetan Plateau. However, the magnitude and seasonal domain of growth responses to climate are species- and habitat specific. Radial growth of fir growing at high- elevation sites is enhanced by a normal or warm summer (June and July) temperatures during the current growing season. For spruce trees growing at middle elevations, radial growth is mainly

8. General discussion and recommendations 89 influenced by temperature variations within a wider winter season, while the influence of summer temperature on radial growth is low. At lower elevations, tree growth is limited by spring moisture availability. Specially, growth of A. forrestii is more severely restricted by spring moisture availability than growth of P. likiangensis, which indicates that this spruce species is more tolerant of droughts. In the study area, instrumental regional annual temperatures have been increasing significantly by 1.4 °C and 1.65 °C during the past fifty years for summer and winter season, respectively (Figure 2.6). This warming trend is unprecedented during the past 250 years, as far as can be derived from tree-ring based annual temperature reconstructions (chapter 5). Under the current warming trend, high- to middle-elevation conifers may benefit from the increasing temperatures, especially from rising winter temperatures in the central Hengduan Mountains. In contrast, low-elevation forest sites may increasingly suffer from drought stress during the premonsoon spring season. This study may contribute to our understanding of tree rings as proxies for climate reconstruction and to estimate the growth response of subalpine trees to further climatic changes in this ecologically complex area. Incorporation of species specific climate-growth reactions into simulation models may be helpful to clarify the biogeographic problems of the Pleistocene refuge areas of Chinese conifers (Frenzel et al., 2003) and to refine forecast models of the regional impact of climate change on the forests in the central Hengduan Mountains

8.2 Temperature History

8.2.1 Annual Temperature Variability

Based on four ring-width chronologies of Picea brachytyla near the upper timberline, annual temperatures were reconstructed over the past 250 years for the central Hengduan Mountains (chapter 5; section 5.1). Temperature conditions, especially in the winter season, mainly affect the radial growth of trees growing near the upper treeline. The first principal component, extracted from four site chronologies, was used to develop a calibration model, which accounts for 42% of the instrumental temperature variance during 1958–2003. The reconstruction revealed that cool episodes occurred during the 1810s, 1860s, and 1960–1980s. Warm intervals occurred in the 1780s, 1850s, 1940–1960, and during the last two decades. Evidence of warm periods in the 1780s and in the 1800s was reported from northeast Tibet (Gou et al., 2007b), the Western Himalaya (Hughes, 1992; 2001; Yadav et al., 2004) and Nepal (Cook et al., 2003). The warm period from 1830 to 1850 and the cool period in the 1860s have also been detected by Wu (1988) in our study area. The warm 1850s are consistent with the warm winters in east Tibetan Plateau (Bräuning, 2006) and warm summers in Kashmir, Western Himalaya (Hughes, 2001). Temperature was relative stable from 1880 to 1940, except for the slightly cool period in the 1910–1920s.

8. General discussion and recommendations 90

Annual temperature was relatively low from 1810 to 1820. The coldest year occurred in 1817. A markedly cold spring in 1817 was also reported for the western Himalaya (Hughes, 1992). This may be linked with two volcanic eruptions of Tambora (Indonesia) in A.D. 1809 and 1815. These eruptions probably influenced the atmospheric circulation patterns of the monsoonal currents, and have been linked with the strong depression in tree growth in Tibet and Nepal (Cook et al., 2003; Bräuning and Mantwill, 2004). Instrumental, historical and various tree-ring evidence show widespread cold conditions in 1816, especially in eastern North America and Western Europe and various ice core records invariably show a strong acidity signal associated with this year (Briffa et al., 1998c). Comparison with other records from nearby regions confirms the reliability of our reconstruction. Based on instrumental data and other proxies (ice cores, tree rings), Wang et al. (2004) showed that temperature anomalies during the period 1920–1950 are noticeable positive over China throughout the last century, especially in south-western China and on the Tibetan Plateau. Warm conditions around 1950, and the cool period around 1970 were also reported in West Sichuan (Shao and Fan, 1999) and Tibet (Briffa et al., 2001; Bräuning and Mantwill, 2004). During the past 20 years, an exceptional increase of annual temperatures can be observed. This warming trend is almost at the same level than during the warmest period observed, i.e, around 1950. This temperature maximum will soon be overtopped, if the warming trend observed for the last 20 years will continue.

8.2.2 Summer Temperature Variability

Based on two well-replicated MXD chronologies of Picea brachytyla, warm season (April to September) mean temperatures were reconstructed over the past 257 years for the central Hengduan Mountains (chapter 5, section 5.2). The calibration model accounts for 41% of the instrumental temperature variance during the period of 1958–2004. Spatial field-correlation analyses using instrumental and reconstructed series revealed similar patterns. This indicates that the reconstruction represents a high amount of the regional summer temperature variability over the Hengduan Mountains. Warm summers occurred during 1750s, 1820–50s, 1880–1890s, 1930–1950s and 1990–present; while summer seasons during 1790–1810s, 1860–1870s, 1900– 1920s and 1960–1985 were relatively cold. The reconstruction successfully captures recent abrupt climatic changes. Comparison with other tree ring-based temperature reconstructions shows high coherency in the timing of warm/cold episodes at the decadal scale across the south-eastern Tibetan Plateau, i.e. cold 1800s, 1860s, 1900–1910s, 1970s; and warm 1750s, 1825–1850s, 1880s, 1930–50s. The spatiotemporal discrepancies among different temperature reconstructions may reflect the influences of different geographic features; bands of climatic signals preserved and target season windows.

8. General discussion and recommendations 91

The 1930–50s is likely the warmest period in the last few centuries seen in the tree-ring records, and was notable throughout the 20th century over China, especially in the south-western part of China and on the Tibetan Plateau (Wang et al., 2004). The cold period in the early 20th century and high temperatures in the 1940s and 1950s are also indicated by δ18O variations in the Puruogangri and Malan ice cores in northern Tibet (Wang et al., 2003; Yao et al., 2006b). The pronounced negative summer temperature trends from 1970 to 1990 on the Tibetan Plateau are probably the consequence of enhanced cloudiness and rainfall at the upper treeline and thus of increasing monsoon intensity (Bräuning and Mantwill, 2004). The cooling in the 1960s and warming in the 1980s seems to be more pronounced and started earlier in the reconstructions of Shao and Fan (1999) and Liang et al. (2008). This may reflect asymmetric variation patterns of maximum and minimum temperatures during regional warming processes (Figure 2.6). A delayed warming of summer maximum temperature after the 1980s in comparison with rising winter minimum temperature was observed in tree-ring based temperature reconstructions and meteorological records on the north-eastern Tibetan Plateau (Gou et al., 2008b). It should be noted that MXD seems to be more strongly influenced by summer day-time maximum temperatures than by nighttime minimum values (Figure 5.2.6; Wilson and Luckman, 2003); whereas there are considerable winter minimum temperature signals in ring-width chronologies of Liang et al. (2008), although their reconstruction concentrated on summer season temperatures. Temperature variations during the last century are also inferred from glacial fluctuation records from nearby mountain peaks. Advance and retreat phases of monsoonal-temperate glaciers are largely controlled by changes in temperature, especially by summer temperature (Shi et al., 2002; Yang et al., 2008). The cold periods of the 1800–1810s and 1900–1920s are consistent with glacier advances or standstill recorded from Lhamcoka (31°49'N, 99°06'E, 6168 m a.s.l.) in the Chola Shan (Bräuning, 2006). Zheng et al. (1999) reported that the Mingyong Glacier of the Meili Snow Mountain (28°29' N, 98°47' E, 6740 m a.s.l.) has retreated by 2 km between 1932 and 1959, then advanced by 830–930 m during 1959–1971 and by 70 m between 1971 and 1982. The terminus of the glacier has retreated by ~80 m between 1998 and 2002 and nearly by 110 m from 2002 to 2004 (Baker and Moseley, 2007). The Baishui No.1 glacier of the Jade Dragon Snow Mountain (27°10'N, 100°13'E, 5600 m a.s.l.), southeast of our sampling sites, also advanced by 800 m between 1957 and 1982, and then retreated by 150 m between 1982 and 1997, and by 100 m between 1998 and 2002 (He et al., 2003).

8.3 Drought History

Four tree ring-width chronologies of three species (Picea likiangensis Pritz, Tsuga dumosa (D. Don) Eichler and Abies ernestii Rehd.) were developed from low elevation sites (chapter 6). Although the four chronologies are from different species, significant correlations exist among the chronologies (mean r = 0.47), and the first principal component accounts for 60.5% of total

8. General discussion and recommendations 92 variance over their common period 1655–2005. Correlation and response function analyses showed that premonsoon (March, April) precipitation and relative humidity and Palmer drought severity index (PDSI) have positive effects on radial growth, while high temperatures in March influence tree growth negatively. This indicates that tree growth is generally limited by spring moisture availability. The spring (March to May) drought reconstruction was verified with independent data, and accounts for 42% of the actual PDSI variance during the common period 1951–2000. Although our PDSI reconstruction was based on the residual tree-ring chronologies, considerable decadal scale moisture variability was retained in our reconstruction. Wet springs occurred during AD 1690s, 1715–1730, 1750s, 1780s, 1825–1850, 1900s, 1930–1960, and 1990–present. Dry springs occurred during AD 1700–1715, 1733–1745, 1790–1820, 1860–1890, 1910–1925, and 1960–1990. Spectral analysis revealed a very strong periodicity of 2–5 years within the reconstructed drought variability, which implies that the El Niño-Southern Oscillation may have some impact on the local moisture availability. The dry springs in the 1700s, 1730s and 1790s–1820s (Figure 6.7b) were also detected as dry periods in the same area by Wu et al. (1988). The dry periods in the 1800s are synchronous with dry conditions in Mongolia and in the western Himalaya (Pederson et al., 2001; Singh et al., 2006). Other intervals in the 18th century were relative wet, especially the period 1715–1730. The period 1825–1850 was a relatively prolonged wet period, and abrupt growth increases occurred in 1824–1825, 1835–1836, and in 1846. This period has also been reported as a prolonged phase with wet springs (AD.1820s to 1840s) in the western Himalayan region of India (Singh et al., 2006). The most severe drought period in the past 350 years occurred during 1910–1925. This severe drought has also been recorded in tree-rings from North China (Li et al., 2006; Liang et al., 2006b) and southern Tibet (Bräuning and Grießinger, 2006). Liang et al. (2006b) combined tree-ring records and historical records (meteorological, hydrological and documentary evidence) and reported that the 1920s drought was severe and sustained. It extended across north China, southern Ningxia, eastern Qinghai, Gansu and eastern Xinjiang in northwest China. Spring climate was relatively wet from 1930 to 1960, which was also reported by Wu et al. (1988). The period 1960–1990 was relatively dry, especially in 1983–1988, which coincides with reports about a regional drought disaster in 1986 by the county annals of Deqin (The Editing Committee of “The annals of Deqin County”, 1997). The last ca. 15 years were comparatively wet, but were still within the range of fluctuations over the last 350 years.

8.4 Comparison of Different Climate Variables

A comparison of different reconstruction series developed in this study is shown in Figure 8.1. Annual temperature has been increasing continuously during the past two decades. However, variations of summer temperatures are still fall within the long-term trends during the past 257 years. Similarily, analyses of instrumental climate data indicated that winter temperatures have

8. General discussion and recommendations 93 been increasing more significantly than summer temperatures during the past fifty years (Figure 2.6). The warm periods of 1930–50s and cold periods of 1960s and 1975–1985 are quite predominant in the study area (Figures 5.1.5, 5.2.7 and 8.1). Although these climatic shifts are generally consistent with other findings from neary regions (Shao and Fan, 1999; Bräuning and Mantwill, 2004; Liang et al., 2008), however, the warmest and coldest years seem to be delayed in the study area as compared with other regions on the northern Tibetan Plateau (Figure 5.2.9) and eastern China (Wang et al., 2004). At decadal scale, warm temperatures are corresponding with relative wet conditions in the study region, such as during A.D. 1825–1850, A.D. 1940s–1950s. In addition, the cold periods especially during 1960s and 1980s correspond to low PDSI values and therefore precipitations. The similar trends between temperature and drought variability are not surprising in the study regions, where monsoon climate is predominant. These results are in agreements with the findings from northeastern Tibetan Plateau (Liu et al., 2006), which may imply the combination of warm-wet and cold-dry climate pattern in the study region. On the other hand, due to the different seasonality of climate that can be reconstructed so far from temperature or moisture- sensitive trees, such comparison could make more sense as further climatic reconstructions can be developed from tree rings and other natural archives.

8. General discussion and recommendations 94

8.5 Recommendations for Further Research

The present study presents a new tree-ring network from the central Hengduan Mountains, where previously high-resolution climate proxies were missing. Tree growth responses to climate were evaluated along environmental gradients, and high-elevation ring width and density measurements were used to estimate inter-annual to multi-decadal temperature variations prior to the period of instrumental measurements. Furthermore, spring drought variability during the past 350 years was reconstructed from low-elevation site chronologies. However, the potential of tree rings as a climate proxy in the study region are not yet explored completely. Further efforts should be made to develop more comprehensive tree-ring networks, and to combine various tree- ring parameters like maximum latewood density and stable isotopes to shed more light on the spatial and temporal variability of the past climate changes in this environmentally sensitive region.

8.5.1 Extension of the Chronology Lengths

One objective of future research in the Hengduan Mountains region should be to extend tree- ring chronologies back in time. The development of tree-ring series covering the past six centuries to the past millennium would allow putting the recent climat changes in the context of periods during the so called “Little Ice Age” and the “Medieval Warm Period”. Compared to trees growing in the cold-dry environment of the interior Tibetan Plateau, tree species growing under humid conditions in the Hengduan Mountains are not so long-lived. Thus, the current developed chronologies generally cover only the past three to five centuries. However, old trees can be found at remote, well-preserved forest sites far away from human influences. Several species have been found to live longer than 550 years in the study area, such as Larix potaninii, Tsuga dumosa and Picea likiangensis. The oldest tree of Tsuga dumosa found so far is 613 years old (Table 3.1). Meanwhile, robust reconstructions only extend back for the past 250 to 350 years, due to limitations of internal signal strength in dendroclimatology. Further sampling in those old-growth forest sites would potentially contribute more robust reconstructions extending back to the past six centuries. In addition, further efforts should exend tree-ring chronologies by exploring wood samples from historical buildings and other sources of sub-fossil woods. Traditional detrending methods (i.e. curve-fitting) are not capable to preserve long-term climatic fluctuation at timescales beyond the life span of individual trees, as described by the “segment length curse” (Cook et al., 1995). Therefore, in the present study past climatic information was generally discussed at annual to decadal time scales. For the further construction of long tree-ring series, new approaches like “Regional Curve Standardization” (RCS; Esper et al., 2003b) should be applied to retain more low-frequency climate information at multi-decadal to centennial scales.

8. General discussion and recommendations 95

8.5.2 Expansion of Sample Area

Despite the progress achieved in the present study, there are important geographic gaps of tree ring coverage in the Hengduan Mountain ranges. Further sampling north- and southwards along the ranges between the Salween and Mekong River may contribute to a more comprehensive understanding of the climate-growth relationships of various tree species along the passage ways of the monsoonal air masses. Moreover, additional long-time chronologies could be constructed through explorations of old-growth, well-preserved sampling sites. Due to the steep topography of the river gorges, human activities are restricted far away from the upper timberlines, so there is still a high potential for dendroclimatological sampling near the southern peaks. For example, I found promising sampling sites at the northern edge of the Gaoligong Mountains in the western range of the Hengduan Mountains (Figure 8.2). Although other dendroclimatological studies from the northern (Bräuning, 1999b) and northeastern (Shao and Fan, 1999; Wu et al., 2005) area of the present study already exist, there are still spatial gaps between the intensive tree-ring sampling around Qamdo and the ‘Three River Gorges’ area. Expanding spatial coverage of tree ring sampling sites would contribute to develop spatially more representative climate proxies, which would improve climate reconstructions in the future.

Figure 8.2 Photographic view of the mixed forest site at the northern edge of Gaoligong Mountains. Inlet shows a tree stem of Taiwania flousiana Gaussen. Photos: Fan Z.X. (2005).

8. General discussion and recommendations 96

8.5.3 Tree-Ring Parameters

Tree-ring widths may reflect environmental influences ranging from temperature to precipitation variations, depending upon the local ecological conditions at each site and ecophysiological differences of each tree species. However, maximum latewood density and stable isotope compositions in annual rings normally contain more coherent climate signals, independent of site conditions or tree species. The potential of using MXD to reconstruct summer temperature variability is not completely explored. The procedure for MXD data is time-consuming and the newly developed less expensive system called high-frequency densitometry (Rinntech, Germany) is still in an experimental stage. Moreover, successful data processing depends considerably upon wood sample quality (i.e. cores without breaks or rotten parts), and higher resolution is needed for getting accurate data for extremely narrow rings. Further research should also apply this technique to other long-lived conifer species, i.e. Larix potaninii, Abies ernestii and Tsuga dumosa, to extend the present reconstructions back in time. Wood anatomy and cell structure for trees growing near the upper and lower timber line may help to document intra-annual disturbances of growth due to extreme climate conditions (i.e. severe frosts, seasonal droughts). Such details should be integrated with climate reconstructions derived from ring width and density data. Furthermore, tree ring carbon and oxygen isotopes for selected sampling sites should be processed to reveal more comprehensive climate information with respect to different climatic variables and seasonal domains.

8.5.4 The Problem of Larch Insect Defoliation

There are several problems with the method of host/non-host comparison. Although the paired host/non-host sites are very close, the similarity between host and non-host chronologies is influence by the defoliations on host trees. Moreover, the method of host/non-host comparison is based on the assumption that host and non-host species are both sensitive to the same climate conditions and response to similar climatic factors. Such assumptions may introduce bias due to the diversity of climatic responses of different tree species and the spatially diverse local environmental conditions (see chapter 4). The criteria for detecting outbreaks vary in different studies and are somewhat arbitrary. Thus, appropriate criteria should be developed by combing field observations and tree growth data in the future. An intensive investigation of local information on insect outbreaks should be assembled. Further work should try to increase sample replication of both host and non-host chronologies, and thus reduce the uncertainties of the host/non-host comparison. Wood anatomical features, i.e. light rings, should be investigated to improve the temporal resolution for the reconstruction of the history of insect outbreaks. Additional sampling sites from nearby mountains are needed to detect spatial fluctuations of insect populations during historical periods. Insect defoliation regulates radial growth and thus influences climatic signals preserved in tree ring width and density series of larch trees. Further utilization larch tree ring data as a climatic proxy will only

8. General discussion and recommendations 97 be successful if appropriate approaches are applied to correct insect-defoliation signals (Büntgen et al., 2005; Esper et al., 2007).

9. Summary - Zusammenfassung 98

9. Summary - Zusammenfassung

This thesis studies the historical climatic and environmental variability in the central Hengduan Mountains, southwestern China, by using dendroclimatological techniques. A tree- ring network of twenty-two total ring width (TRW) and five maximum latewood density (MXD) chronologies has been developed. The sample sites cover the Three River Gorges region (27–29 °N, 98–100 °E), ranging from the western slope of the Gaoligong Mountains to the eastern slope of the Yulong Snow Mountains. The chronologies generally cover the past three to five centuries. The studied tree species include Abies georgei, A. forrestii, A. ernestii, Picea brachytyla, P. likiangensis, Tsuga dumosa, Pseudotsuga forrestii and Larix potaninii var. macrocarpa.

Site TRW and MXD chronologies were developed according to standard dendroclimatological techniques. Correlation and cluster analyses were performed to assess the similarity among site chronologies. Principal component analysis was employed to determine the relative distance of chronologies, and to extract common ‘climate-like’ signals. Regional climatic series were developed from instrumental data series (i.e. temperature, precipitation). Correlation and response functions were performed to quantify the tree-ring response to climatic variables. Linear regression was utilized to transfer orthogonal principal component scores (PCs) into climatic variability prior to the instrumental periods. The climate reconstructions were compared with existing climate proxies from surrounding regions.

The analyses of meteorological data revealed that regional annual mean temperature have been increasing by 0.03 ºCyr–1 during the past fifty years, with an increase rate of 0.028 ºCyr–1 and 0.033 ºCyr–1 for summer and winter season, respectively. The warming trend was nearly double the global mean trend of 0.019 ºCyr–1. However, this warming trend was mainly caused by an increase of minimum temperatures (0.09 ºCyr–1) instead of maximum temperatures. The regional daily temperature range (DTR) has been decreasing since the early 1980s.

Improved understanding of tree growth responses to climate is needed to model and predict forest ecosystem responses to current and future climatic variability. The effects of inter-annual climate variations on radial growth of high-elevation (> 3200 m a.s.l.) conifers were evaluated. Eight TRW chronologies were developed from fir (Abies georgei and A. forrestii) and spruce (Picea brachytyla and P. likiangensis) stands along an elevation gradient from 3200 to 4200 m a.s.l. Correlation and principal component analyses for the eight TRW residual chronologies identified three groups of sites, representing different patterns of growth-climate relationships. Correlation and redundancy analyses with regional climate data revealed that radial growth of trees growing from high- to middle elevations is sensitive to low temperatures during winter season. In contrast, radial growth of fir growing at high-elevation sites is enhanced by normal or warm summer temperatures (June and July) during the current growing season. For spruce trees growing at middle elevations, the influence of summer temperature on radial growth is low. At

9. Summary - Zusammenfassung 99 low elevation sites, trees display low sensitivity to temperature variations. However, spring moisture availability becomes crucial for radial growth regardless of tree species. High- to middle-elevation conifers in the central Hengduan Mountains may benefit from the current climate warming, especially from rising winter temperatures.

Variations in ring width and wood density of Picea brachytyla were used to develop high- resolution climate proxy data to extend the existing instrumental temperature record in the central Hengduan Mountains. TRW chronologies from four Picea brachytyla stands near the upper treeline show similar growth patterns, and significant correlations are embodied among the four chronologies. A principal component analysis for the four spruce chronologies indicated that the first component accounts for 54.8% of the total variance over the period 1750–2003. Climate-growth response analysis revealed that radial growth is mainly controlled by temperature variations, especially in the winter season. Based on the first principal component of the four spruce TRW chronologies, annual mean temperatures (from previous October until September) were reconstructed for the past 250 years. The calibration model accounts for 42% of the actual temperature variance during the period 1959–2003 covered by both meteorological records and tree-ring data. The reconstruction shows that the central Hengduan Mountains experienced some cool episodes during the 1810s, 1860s, and during 1960–1980. Warm intervals occurred during the 1780s, 1850s, 1940–1960 and in the past two decades. The cool period of the 1810s may be linked to two volcanic eruptions of Tambora (Indonesia) in 1809 and 1815. These general patterns are mostly in accordance with other records from nearby regions. During the past 250 years, the strongest warming trends were found during the past two decades, which was confirmed by a significant increase of winter temperatures in the instrumental records.

Two well-replicated MXD chronologies of Picea brachytyla were developed. Growth- climate response analyses showed that MXD of trees growing in the sub-alpine zone are mainly influenced by summer temperature variability. Based on a MXD regional chronology, a warm season (April-September) temperatures were reconstructed for the period A.D. 1750–2006. The climate/tree-growth model accounts for 41% of the instrumental temperature variance during the period 1958–2004. Warm summers occurred during 1750s, 1820–50s, 1880–1890s, 1930–1950s and 1990–present; while the periods of 1790–1810s, 1860–1870s, 1900–1920s, and 1960–1985 were relatively cold. Spatial climate correlation analyses with gridded land surface data revealed that the warm season temperature reconstruction contains a strong regional temperature signal for the Hengduan Mountain ranges. The present reconstruction successfully captured recent abrupt climatic changes and generally agreed with other tree-ring-based temperature reconstructions from nearby regions on a decadal timescale. In addition, reconstructed summer temperature variations were consistent with recorded glacier fluctuations in the surrounding high mountain areas over the past century.

Four tree ring-width chronologies of three species (Picea likiangensis Pritz, Tsuga dumosa (D. Don) Eichler and Abies ernestii Rehd.) were developed from low elevation sites in the

9. Summary - Zusammenfassung 100 central Hengduan Mountains. Although the four chronologies come from different species, significant correlations exist among the chronologies (mean r = 0.47), and the first principal component accounts for 60.5% of total variance over their common period 1655–2005. Correlation and response function analyses showed that premonsoon (March, April) precipitation and relative humidity and Palmer drought severity index (PDSI) have positive effects on radial growth, while temperature in March influences tree growth negatively. This indicates that tree growth at low-elevation sites is generally limited by spring moisture availability. Based on the first principal component of four ring-width chronologies, spring (March-May) drought variability (PDSI) was reconstructed for the past 350 years. The spring (March to May) drought reconstruction was verified with independent data, and accounts for 42% of the actual PDSI variance during their common period 1951–2000. Wet springs occurred during AD 1690s, 1715– 1730, 1750s, 1780s, 1825–1850, 1900s, 1930–1960, and 1990–present. Dry springs occurred during AD 1700–1715, 1733–1745, 1790–1820, 1860–1890, 1910–1925, and 1960–1990. A strong 2–5 years periodicity was found in the reconstructed drought sequence, implying that the El Niño-Southern Oscillation (ENSO) has some influence on the local drought variability.

9. Summary - Zusammenfassung 101

Zusammenfassung

In dieser Arbeit wird mit Hilfe dendroklimatologischer Techniken die historische Klima- und Umweltvariabilität im zentralen Hengduan Gebirge im südwestlichen China untersucht. Es wurde ein Jahrringnetzwerk aus 22 Jahrringbreitenchronologien (total ring width–TRW) und fünf Chronologien der maximalen Spätholzdichte (maximum latewood density–MXD) erstellt. Die Untersuchungsstandorte decken die Drei-Schluchten-Region (27–29 °N, 98–100 °E) ab und reichen von der westlichen Abdachung des Gaoligong Gebirges bis zur östlichen Abdachung des Yulong Schnee-Gebirges reichen. Die Chronologien umfassen im allgemeinen die letzten drei bis 5 Jahrhunderte. Für diese Arbeit wurden die Baumarten Abies georgei, A. forrestii, A. ernestii, sowie Picea brachytyla, P. likiangensis, Tsuga dumosa, Pseudotsuga forrestii und Larix potaninii var. macrocarpa untersucht.

Die Standortchronologien der Jahrringbreite und der maximalen Spätholzdichte wurden gemäß dendroklimatologischer Standardtechniken erstellt. Zur Untersuchung der Ähnlichkeit zwischen den Standortchronologien wurden Korrelationen und Clusteranalysen durchgeführt. Eine Hauptkomponentenanalyse wurde durchgeführt, um die relative Distanz der Chronologien festzustellen und um „klimaähnliche“ Signale herauszufiltern. Regionale Klimareihen wurden aus Klimadaten von stationären Messungen entwickelt (z. B. Temperatur und Niederschlag). Um die Reaktion der Jahrringe auf die klimatischen Parameter zu quantifizieren, wurden Korrelationen und „response functions“ berechnet. Mit Hilfe einer linearen Regression wurden die voneinander unabhängigen Hauptkomponenten (PCs) in klimatische Variabilität vor den Zeiträumen der Instrumentenmessung transformiert. Die so erhaltenen Klimarekonstruktionen wurden mit vorhandenen Klima-Proxy-Daten benachbarter Standorte verglichen.

Die Analyse der meteorologischen Daten ergab, dass die regionale Jahresmitteltemperatur in den letzten 50 Jahren um 0.03ºC pro Jahr angestiegen ist, mit einer Rate von 0.028°C im Sommer und einer Rate von 0.033°C im Winter. Diese Erwärmung war annähernd doppelt so hoch wie der Trend des globalen Mittels von 0.019ºC pro Jahr. Die Erwärmung geht jedoch hauptsächlich auf einen Anstieg der Temperaturminima zurück (0.09 ºCyr–1) und nicht auf die Temperaturmaxima. Die regionale Tagestemperaturamplitude (daily temperature range - DTR) ist seit den frühen 1980er Jahren gesunken.

Ein verbessertes Verständnis der Wachstumsreaktionen der Bäume auf das Klima ist nötig, um die Reaktionen der Waldökosysteme auf aktuelle und zukünftige Klimaschwankungen zu modellieren und vorherzusagen. Die Auswirkungen von interannuellen Klimaschwankungen auf das radiale Wachstum von Koniferen in Hochlagenstandorten (> 3200 m ü.M.) wurde untersucht. Jeweils acht Jahrringbreitenchronologien wurden von Tannen- (Abies georgei und A. forrestii) und Fichtenbeständen (Picea brachytyla und P. likiangensis) entlang eines Höhengradienten von 3200 bis 4200 m ü.M erstellt. Durch Korrelationen und Hauptkomponentenanalyse der acht „residual“ Jahrringbreitenchronologien nach der Trendbereinigung durch das Programm

9. Summary - Zusammenfassung 102

ARSTAN wurden drei Gruppen von Standorten identifiziert, die unterschiedliche Muster von Klima-Wachstumsbeziehungen repräsentieren. Korrelationen und Redundanzanalysen mit regionalen Klimadaten ergaben, dass das radiale Wachstum von Bäumen, die in großen bis mittleren Höhenlagen wachsen, sensitiv auf niedrige Temperaturen im Winterhalbjahr reagiert. Im Gegensatz dazu wird das radiale Wachstum von Tannen in Hochlagenstandorten durch normale oder warme Sommertemperaturen (Juni oder Juli) der jeweiligen Vegetationsperiode verstärkt. Für Fichten in mittleren Höhenlagen ist der Einfluss der Sommertemperatur auf das radiale Wachstum gering. Auf den niedrig gelegenen Standorten zeigen die Bäume eine geringe Sensitivität gegenüber Temperaturvariationen. Jedoch ist die Wasserverfügbarkeit im Frühling, unabhängig von der Baumart, ausschlaggebend für das radiale Baumwachstum. Nadelbäume in mittlerer bis großer Höhenlage im zentralen Hengduan Gebirge könnten von der derzeitigen Klimaerwärmung profitieren, insbesondere von den steigenden Wintertemperaturen.

Zeitreihen der Jahrringbreite und der maximalen Spätholzdichte von Picea brachytyla wurden genutzt um hochaufgelöste Klimaproxydaten zu erstellen und um die existierenden Temperaturaufzeichnungen des zentralen Hengduan Gebirges zu verlängern. Die Jahrringbreitenchronologien von Picea brachytyla Beständen nahe der oberen Baumgrenze zeigen untereinander ähnliche Wachstumsmuster und es existieren signifikante Korrelationen innerhalb der vier Chronologien. Eine Hauptkomponentenanalyse zeigt, dass der erste Komponent 54.8% der absoluten Varianz innerhalb des Zeitraums von 1750–2003 erklärt. Die Analyse der Klima-Wachstumsbeziehungen erbrachte, dass das radiale Wachstum hauptsächlich von Temperaturschwankungen gesteuert wird, insbesondere im Winterhalbjahr des Vorjahres. Basierend auf der ersten Hauptkomponente der vier Jahrringbreitenchronologien wurde die Jahresdurchschnittstemperatur (vom Oktober des Vorjahres bis zum September) für die letzten 250 Jahre rekonstruiert. Das Kalibrationsmodell erklärt 42% der aktuellen Temperaturschwankungen während des Zeitraums von 1959–2003, der sowohl von meteorologischen Daten, als auch durch Jahrringdaten abgedeckt wird. Die Rekonstruktion zeigt, dass das zentrale Hengduan Gebirge einige kältere Episoden in den 1810er und den 1860er Jahren und während der Jahre 1960 bis 1980 durchlief. Warme Zeiträume traten in den 1780ern, den 1850ern, von 1940 bis 1960 und in den letzten beiden Jahrzehnten auf. Die kalte Periode der 1810er Jahre könnte mit den Vulkanausbrüchen des Tambora (Indonesien) in den Jahren 1809 und 1815 zusammenhängen. Diese allgemeinen Muster befinden sich überwiegend in Übereinstimmung mit Ergebnissen aus benachbarten Regionen. Während der letzten 250 Jahre wurde die stärkste Erwärmungstendenz während der letzten zwei Dekaden gefunden, was durch einen deutlichen Anstieg der Wintertemperatur bei den meteorologischen Messergebnissen bestätigt wurde.

Zwei gut belegte Chronologien der maximalen Spätholzdichte von Picea brachytyla wurden erstellt. Die Analyse der Klimawachstumsbeziehungen zeigte, dass die maximale Spätholzdichte von Bäumen der sub-alpinen Stufe hauptsächlich von der Variabilität der Sommertemperaturen

9. Summary - Zusammenfassung 103 beeinflusst wird. Basierend auf einer regionalen Chronologie der maximalen Spätholzdichte wurde eine Temperaturrekonstruktion der warmen Jahreszeit (April–September) für den Zeitraum von 1750 bis 2006 erstellt. Das Klima-/ Baumwachstumsmodell erklärt 41% der direkt gemessenen Temperaturvarianz während der Periode 1958 bis 2004. Warme Sommer treten während der 1750er Jahre und von 1820–1850, von 1880–1890, 1930–1950 und von 1990 bis Heute auf. Hingegen waren die Zeiträume von 1790–1810s, von 1860–1870, von 1900–1920s, und von 1960–1985 relativ kalt. Eine räumliche Klimakorrelation mit Rasterdaten der Landoberfläche erbrachte, das die Rekonstruktion der Sommertemperatur ein starkes regionales Temperatursignal für das Hengduan Gebirge enthält.

Die vorliegende Rekonstruktion erfasst erfolgreich aktuelle abrupte Klimaveränderungen und stimmt im Allgemeinen auf einer dekadischen Zeitskala mit anderen jahrringbasierten Temperaturrekonstruktionen benachbarter Regionen überein. Darüber hinaus stehen die rekonstruierten Sommertemperaturvariationen im Einklang mit gemessenen Gletscherschwankungen des vergangenen Jahrhunderts in den umgebenden Hochgebirgsregionen. Vier Jahrringbreiten-Chronologien von drei Baumarten (Picea likiangensis Pritz, Tsuga dumosa (D. Don) Eichler and Abies ernestii Rehd.) wurden von niedrig gelegenen Standorten im zentralen Hengduan Gebirge erstellt. Obwohl die vier Chronologien von verschiedenen Arten stammen, existieren deutliche Korrelationen zwischen den Chronologien (mittleres r = 0.47), und die erste Hauptkomponente hat einen Anteil von 65 % der Gesamtvarianz über ihren gemeinsamen Zeitraum von 1655–2005.

Die Korrelations- und „response function“-Analyse zeigt, dass die Niederschläge vor dem Monsun (im März, April), die relative Feuchte und der Palmer drought severity index (PDSI) positive Auswirkungen auf das radiale Wachstum haben, während die Märztemperatur das Baumwachstum negativ beeinflusst. Das weist darauf hin, dass das Baumwachstum an niedrig gelegenen Standorten allgemein von der Wasserverfügbarkeit im Frühjahr gesteuert wird. Basierend auf der ersten Hauptkomponente der vier Jahrringbreitenchronologien wurde die Dürrevariabilität (PDSI) im Frühjahr (März bis Mai) für die vergangenen 350 Jahre rekonstruiert. Die Rekonstruktion der Frühjahrstrockenheit (März bis Mai) wurde mit unabhängigen Daten verifiziert und erklärt 42% der aktuellen PDSI Varianz während des gemeinsamen Zeitraums von 1951 bis 2000. Nasse Frühjahre traten in den 1690er Jahren, von 1715–1730, in den 1750er und 1780er Jahren, von 1825–1850, um 1900, von 1930–1960 und von 1990 bis heute auf. Trockene Frühjahre gab es in den Jahren 1700–1715, 1733–1745, 1790–1820, 1860–1890, 1910–1925 und von 1960–1990. Eine starke, zwei bis fünfjährige Periodizität wurde in der rekonstruierten Abfolge der Trockenheit entdeckt, was andeuten könnte, dass El Niño-Southern Oscillation (ENSO) einen gewissen Einfluss auf die lokale Dürrehäufigkeit hat.

9. Summary - Zusammenfassung 104

中文摘要

本文利用树木年代气候学方法研究了中国西南横断山中部地区过去几个世纪以来气候和环境变化的历史。根据树轮气候学研究的要求，在研究区域内进行了系统的树木年轮采样。样点覆盖了北纬 27～29 度和东经 98～100 度的区域，至西向东分布于高黎贡山自然保护区与玉龙雪山保护区之间。研究树种包括：长苞冷杉（Abies georgei）,川西冷杉（Abies forrestii）,云南黄果冷杉（Abies ernestii）,油麦掉云杉（Picea brachytyla）, 丽江云杉（Picea likiangensis）,云南铁杉（Tsuga dumosa）,澜沧黄杉（Pseudotsuga forrestii）和大果红杉（Larix potaninii var. macrocarpa）。利用 LINTAB II 树轮测量仪测定年轮宽度，用德国 RINNTECH 公司生产的 LignoStation 树轮密度仪测量年轮密度。在精确交叉定年的基础上，总共建立了 22 个树轮宽度年表和 5 个最大晚材密度年表。

采用相关和聚类分析方法比较年表之间的相似程度，利用主成份分析提取年表间的公共气候信号。建立器测气象数据的区域平均序列，采用相关分析和响应函数分析建立树轮宽度（密度）指数和器测气象资料之间的数量关系。以主成份提取的主分量作为预报因子，利用线性回归模型重建过去气候变化的历史。重建的气候序列与周边地区已有的古气候代用资料进行对比。

对气象数据的分析显示，横断山中部地区在过去 50 年来气温有显著上升，其中年平均温度每年上升 0.03 oC，夏季（6 月至 9 月）和冬季（11 月至翌年 2 月）温度分别每年上升 0.028 oC 和 0.033 oC。近半个世纪以来区域气候变暖的幅度接近全球平均气温上升幅度（0.019 oC）的两倍。然而，气候变暖主要表现为夜间最低温度的上升，而白天最高气温的上升相对较少，至 80 年代初期以来平均昼夜温差呈下降趋势。

树轮-气候响应模式的研究是气候重建的基础，同时可为预测森林生态系统对未来气候变化的响应提供重要的理论依据。本文首先讨论了高海拔地区（3200 米以上）几个主要针叶树种径向生长与气候因子的相关关系。对不同海拔梯度上建立的 8 个年轮宽度年表进行主成份及气候相关分析显示，8 个树轮宽度年表可以划分为 3 种不同类型的树木生长与气候响应模式。在海拔 4000～4200 米的高山林线附近，长苞冷杉（A. georgei）的径向生长主要受冬季（前年 11 月）和夏季（6～7 月）温度影响；生长于中山海拔（3300～ 3500 米左右）的油麦掉云杉(P. brachytyla)则主要受冬季低温的限制，而夏季温度的影响较弱；生长于低海拔地区（3300 米以下）的丽江云杉(P. likiangensis)和川滇冷杉(A. forrestii)对温度的敏感性降低，其径向生长与春季（3～5 月）温度呈负相关而与春季降水呈正相关，表明低海拔地区树木生长主要受春季水分供给能力的限制。不同的树轮-气候响应模式反映了海拔梯度上水热组合的变化导致的限制树木生长主要气候因子的差异，

9. Summary - Zusammenfassung 105

与不同树种在垂直地带性植被上的分布相对应。在横断山地区气候变暖背景下，该区域中高海拔地区的树木生长可能受益于未来温度的持续上升，特别是冬季温度的上升。

利用高海拔样点的油麦掉云杉（P. brachytyla）的树轮宽度和密度数据，获取高分辨率气候代用资料。首先，利用 4 条油麦掉云杉年轮宽度序列，重建了研究区域过去 250 年来年平均温度变化的历史。该区 4 个树轮宽度序列之间的变化趋势比较一致，且年表之间具有显著的正相关关系。在 1750～2003 的公共区间，对 4 个云杉年表进行主成份分析，结果表明第一主分量解释方差量达 54.8%。与气候因子的相关分析和响应函数分析表明高海拔地区树木的径向生长主要受温度因子特别是冬季温度的限制。以第一主分量作为预报因子建立线性回归方程，重建方程的方差解释量达 43%。重建结果显示，至 1750 年以来，横断山中部地区存在 3 个明显的冷期，分别为 1810s, 1860s, 1960～80s；明显的暖期分别为 1780s, 1850s, 1940～60 和过去 20 年。与周边地区重建的温度变化序列相比，本文重建结果与周边地区重建的温度变化历史具有较好的一致性。过去 20 年来年平均气温呈显著上升趋势，其温度上升的幅度超出了过去 250 年内的任何时段。

利用两个高样本重复的油麦吊云杉最大晚材密度序列，重建了 1750 年以来该地区夏季（4～9 月）平均气温变化的历史。在公共区间 1958～2004，转换方程的解释方差量达 41%，且方程通过稳定性检验。重建结果显示，过去 250 年间横断山中部地区经历了明显的冷暖交替，其中 1750s, 1820～50s, 1880～90s, 1930～50s, 1990 至今为夏季温度较平均值高的时段，而 1790～1810s, 1860～1870s, 1900～1920s, 1960～1985 则为相对寒冷的时段。与地表网格气象数据的空间遥相关分析表明，重建的夏季温度序列具有较强的区域代表性。同时，在 10 年尺度上该重建的夏季气温序列与邻近地区已有树轮记录的温度变化历史具有较好的一致性。另外，重建的夏季冷暖变化与周边地区季风温冰川的进退历史记录相对应，这近一步证明了该重建结果的可靠性。

利用在横断山中部地区低海拔样点建立的四个树木年轮宽度序列，重建了过去 350 年来该地区 Palmer 干旱指数（Palmer drought severity index，PDSI）变化的历史。结果表明，尽管四个树轮宽度序列来自不同树种，分别为丽江云杉 (P. likiangensis), 云南铁杉（T. dumosa）和云南黄果冷杉（A. ernestii），但各年表之间存在显著的相关关系和相似的变化趋势。在 1655～2005 年的公共区间，四个年表之间平均相关系数达 0.47，通过主成份分析提取的第一主分量解释方差量达 65.5%。相关和响应函数分析表明，树木径向生长与春季(3～4 月)降水呈显著正相关，而与 3 月温度呈负相关。同时，春季相对湿度和 Palmer 干旱指数(PDSI)对树木生长有显著的影响，说明春季水分供给是限制低海拔地区树木径向生长的关键因子。采用第一主分量作为预报因子，重建了该地区 1655 年以来春季(3～5 月份)干旱指数(PDSI) 的变化历史，重建方程的方差解释量达 42%。重建结果显示，研究区域相对湿润的时段为：1690s, 1715～1730, 1750s, 1780s, 1825～1850, 1900s, 1930～1960, 1990 至今；相对干燥的时段为：1700～1715, 1733～1745, 1790～1820,

9. Summary - Zusammenfassung 106

1860～1890, 1910～1925, 1960～1990。功率谱分析显示重建的干旱指数序列具有明显的 2～5 年准周期，并推测厄尔尼诺南方涛动（ENSO）对该地区干湿变化有着一定的影响。

10. References 107

10. References

Aizen, V.B., Kuzmichenok, V.A., Surazakov, A.B., Aizen, E.M., 2007. Glacier changes in the Tien Shan as determined from topographic and remotely sensed data. Global and Planetary Change 56, 328–340.

Akaike, H., 1974. A new look at the statistical model identification. IEEE Transaction on Automatic Control AC19 (6), 716–723.

Anfodillo, T., Rento, S., Carraro, V., Furlanetto, L., Urbinati, C., Carrer, M., 1998. Tree water relations and climatic variations at the alpine timberline: seasonal changes of sap flux and xylem water potential in Larix decidua Miller, Picea abies (L.) Karst. and Pinus cembra L. Annals of Forest Science 155, 159–172.

Araguas-Araguas, L., Froehlich, K., Rozanski, K., 1998. Stable isotope composition of precipitation over southeast Asia. Journal of Geophysical Research 103, 28721–28742.

Baker, B.B., Bachelet, D., Daly, C., Ma, J., Moseley, R.K., Shi, X.Z., Sun, J.H., Shlisky, A., 2005. Effects of climate change and land management practices in the Hengduan Mountains of northwestern Yunnan, PRC: options for alpine conservation. In: Price, M., (Eds.) Global Change in Mountain Regions. Dumfrieshire, Sapiens Publishing, pp 272–273.

Baker, B.B., Moseley, R.K., 2007. Advancing treeline and retreating glaciers: implications for conservation in Yunnan, P.R. China. Arctic, Antarctic and Alpine Research 39(2), 200–209.

Baltensweiler, W., Benz, G., Bovey, P., Delucchi, V., 1977. Dynamics of larch bud moth populations. Annual Review of Entomology 22, 79–100.

Baltensweiler, W., Rubli, D., 1999. Dispersal - an important driving force of the cyclic population dynamics of the larch bud moth. Forest Snow and Landscape Research 74, 3–153.

Bao, W.K., Wu, N., He, S.C., Gema, J.C., Yang, P.F., Zhongyong, C.L., 2001. A study of alpine and subalpine rangeland communities in upstream Deqin: types and attributes. Journal of Mountain Science 9, 226–230. (in Chinese with English abstract).

Barnett, T.P., Dümenil, L., Schlese, U., Roeckner, E., 1988. The effect of Eurasian snow cover on global climate. Science 239, 504–507.

Barnett, T.P., Adam, J.C., Lettenmaier, D.P., 2005. Potential impacts of a warming climate on water availability in snow-dominated regions. Nature 438, 303–309. Doi: 10.1038/nature04141.

10. References 108

Batllori, E., Gutiérrez, E., 2008. Regional tree line dynamics in response to global change in the Pyrenees. Journal of Ecology 96, 1275–1288.

Bhattacharyya, A., Chaudhary, V., 2003. Late-summer temperature reconstruction of the Eastern Himalayan Region based on tree-ring data of Abies densa. Arctic, Antarctic and Alpine Research 35(2), 196–202.

Biondi, F., Waikul, K., 2004. DENDROCLIM2002: A C++ program for statistical calibration of climate signals in tree-ring chronologies. Computers and Geosciences 30(3), 303–311.

Böhner, J., Lehmkuhl, F., 2005. Environmental change modelling for Central and High Asia: Pleistocene, present and future scenarios. Boreas 34, 220–231.

Boulanger, Y., Arseneault, D., 2004. Spruce budworm outbreaks in eastern Quebec over the last 450 years. Canadian Journal of Forest Research 34, 1035–1043.

Bradley, R.S., Jones, P.D., 1992. Climate science A.D. 1500: Introduction. In: Bradley R.S., Jones, P.D., (Eds.) Climate science A.D. 1500. Routledge, London, pp. 1–16.

Bradley, R.S., Briffa, K.R., Cole, J., Hughes, M.K., Osborn, T.J., 2003. The climate of the last millennium. In: Alverson, K.D., Bradley, R.S., Pedersen, T.F., (Eds.) Paleoclimate, golbal change and the future. Springer-Verlag, Berlin, pp. 105–141.

Bräuning, A., 1994. Dendrochronology for the last 1400 years in eastern Tibet. GeoJournal 34, 75–95.

Bräuning, A., 1999a. Zur Dendroklimatologie Hochtibets während des letzten Jahrtausends. Dissertationes Botanicae, Band 312. J. Cramer Berlin, Stuttgart.

Bräuning, A., 1999b. Dendroclimatological potential of drought-sensitive tree stands in Southern Tibet for the reconstruction of the monsoonal activity. IAWA Journal 20 (3), 325–338.

Bräuning, A., 2000. Ecological division of forest regions of eastern Tibet by using of dendroecological analyses. Marburger Geographische Schriften 135, S111–127.

Bräuning, A., 2001a. Combined view of various tree ring parameters from different forest habitats in Tibet for the reconstruction of seasonal aspects of Asian Monsoon variability. The Palaeobotanist 50, 1–12.

Bräuning, A., 2001b. Climate history of the Tibetan Plateau during the last 1000 years derived from a network of Juniper chronologies. Dendrochronologia 19(1), 127–137.

10. References 109

Bräuning, A., Mantwill, B., 2004. Summer temperature and summer monsoon history on the Tibetan Plateau during the last 400 years recorded by tree rings. Geophysical Research Letters 31, L24205, Doi: 10.1029/2004GL020793.

Bräuning, A., 2006. Tree-ring evidence of ‘Little Ice Age’ glacier advances in southern Tibet. The Holocene 16(3), 369–380.

Bräuning, A., Grießinger, J., 2006. Late Holocene variations in monsoon intensity in the Tibetan-Himalayan region-evidence from tree rings. Journal of the Geological Society of India 68(3), 485–494.

Briffa, K., Jones, P.D., Schweingruber, F.H., 1988. Summer temperature patterns over Europe: a reconstruction from 1750 A.D. based on maximum latewood density indices of conifers. Quaternary Research 30, 36–52.

Briffa, K.R., Jones, P.D., 1990. Basic chronology statistics and assessment. In: Cook, E.R., Kairiukstis, L.A., (Eds.) Methods of dendrochronology: application in environmental science. Kluwer Academic Press, Netherlands, p 146.

Briffa, K.R., Jones, P.D., Schweingruber, F.H., 1992. Tree-ring density reconstructions of summer temperature patterns across Western North America since 1600. Journal of Climate 5, 733–754.

Briffa, K.R., Jones, P.D., Hulme, M., 1994. Summer moisture variability across Europe 1892– 1991: an analysis based on the Palmer drought severity index. International Journal of Climatology 14, 475–506. Doi: 10.1002/joc.3370140502.

Briffa, K.R., 1995. Interpreting high-resolution proxy climate data: the example of dendroclimatology. In: von Storch, H., Navarra, A., (Eds.) Analysis of climate variability: applications of statistical techniques. Springer-Verlag, Berlin. pp. 77–94.

Briffa, K.R., Schweingruber, F.H., Jones, P.D., Osborn, T.J., Shiyatov, S.G., Vaganov, E.A., 1998a. Reduced sensitivity of recent tree-growth to temperature at high northern latitudes. Nature 391, 678–682.

Briffa, K.R., Schweingruber, F.H., Jones, P.D., Osborn, T.J., Harris, I.C., Shiyatov, S.G., Vaganov, E.A., Grudd, H., 1998b. Trees tell of past climates: but are they speaking less clearly today? Philosophical Transactions of the Royal Society of London (Series B- Biological Sciences) 353, 65–73.

Briffa, K.R., Jones, P.D., Schweingruber, F.H., Osborn, T.J., 1998c. Influence of volcanic eruptions on Northern Hemisphere summer temperature over the past 600 years. Nature 393, 450–455.

10. References 110

Briffa, K.R., Osborn, T.J., Schweingruber, F.H., Harris, I.C., Jones, P.D., Shiyatov, S.G., and Vaganov, E.A., 2001. Low frequency temperature variations from northern tree ring density network. Journal of Geophysical Research 106, 2929–2941.

Briffa, K.R., Osborn, T.J., Schweingruber, F.H., 2004. Large-scale temperature inferences from tree rings: a review. Global and Planetary Change 40, 11–26.

Brohan, P., Kennedy, J.J., Harris, I., Tett, S.F.B., Jones, P.D., 2006. Uncertainty estimates in regional and global observed temperature change: A new data set from 1850. Journal of Geophysical Research 111, D12106, Doi: 10.1029/2005JD006548.

Buckley, B.M., Palakit, K., Duangsathaporn, D., Sanguantham, P., Prasomsin, P., 2007. Decadal scale droughts over northwestern Thailand over the past 448 years: link to the tropical Pacific and Indian Ocean sectors. Climate Dynamics 29, 63–71. Doi: 10.1007/s00382-007- 225-1.

Buntaine, M.T., Mullen, R.B., Lassoie, J.P., 2007. Human use and conservation planning in alpine areas of northwestern Yunnan, China. Environment, Development and Sustainability 9, 305–324.

Büntgen, U., Esper, J., Frank, D.C., Nicolussi, K., Schmidhalter, M., 2005. A 1052-year tree-ring proxy for Alpine summer temperatures. Climate Dynamics 25, 141–153. Doi: 10.1007/s00382-005-0028-1.

Büntgen, U., Frank, D.C., Nievergelt, D., Esper, J., 2006. Summer temperature variations in the European Alps, AD 755–2004. Journal of Climate 19, 5606–5623.

Büntgen, U., Frank, D.C., Kaczka, R.J., Verstege, A., Zwijacz-Kozica, T., Esper, J., 2007. Growth responses to climate in a multi-species tree-ring network in the Western Carpathian Tatra Mountain, Poland and Slovakia. Tree Physiology 27, 689–702.

Büntgen, U., Frank, D.C., Grudd, H., Esper, J., 2008. Long-term summer temperature variations in the Pyrenees. Climate Dynamics 31(6), 615–631.

Caccianiga, M., Payette, S., Fillion, L., 2008. Biotic disturbance in expanding subarctic forests along the eastern coast of Hudson Bay. New Physiologist 178(4), 823–834.

Campbell, R., Smith, D.J., Arsenault, A., 2006. Multicentury history of western spruce budworm outbreaks in interior Douglas-fir forests near Kamloops, British Columbia. Canadian Journal of Forest Research 36, 1758–1769.

Carrer, M., Anfodillo, T., Urbinati, C., Carraro, V., 1998. High-altitude forest sensitivity to global warming: results from long-term and short-term analyses in the eastern Italian Alps.

10. References 111

In: Beniston, M., Innes, J.L., (Eds.) The impacts of climate variability of forest. Springer- Verlag, Berlin, Heidelberg, pp. 171–189.

Chang, D.H.S., 1981. The vegetation zonation of the Tibetan Plateau. Mountain Research and Development 1(1), 29–48.

Charles, C.D., Hunter, D.E., Fairbanks, R.G., 1997. Interaction between the ENSO and the Asian monsoon in a coral record of tropical climate. Science 277, 925–928.

Churkina, G., Running, S.W., 1998. Contrasting climatic controls on the estimated productivity of global terrestrial biomes. Ecosystems 1(2), 206–215.

Conkey, L.E., 1986. Red spruce tree-ring widths and densities in eastern North America as indicators of past climate. Quaternary Research 26, 232–243.

Cook, E.R., Peters, K., 1981. The smoothing spline: A new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin 41, 45–53.

Cook, E.R., 1985. A time-series analysis approach to tree-ring standardization. PhD dissertation, The University of Arizona Press, Tucson.

Cook, E.R., Kairiukstis, A., 1990. Methods of dendrochronology: applications in the environmental sciences. Kluwer Academic Press, Dordrecht.

Cook, E.R., Briffa, K.R., Jone, P.D., 1994. Spatial regression methods in dendroclimatology: A review and comparison of two techniques. International Journal of Climatology 14, 379–402.

Cook, E.R., Briffa, K.R., Meko, D.M., Graybill, D.A., Funkhouser, G., 1995. The ‘segment length curse’ in long tree-ring chronology development for paleoclimatic studies. The Holocene 5(2), 229–237.

Cook, E.R., Peters, K., 1997. Calculating unbiased tree-ring indices for the study of climatic and environmental change. The Holocene 7, 361–370.

Cook, E.R., Meko, D.M., Stahle, D.W., Cleaveland, M.K., 1999. Drought reconstructions for the continental United States. Journal of Climate 12, 1145–1162.

Cook, E.R., Krusic, P.J., Jones, P.D., 2003. Dencroclimatic signals in long tree-ring chronologies from the Himalayas of Nepal. International Journal of Climatology 23, 707–732.

Crowley, T.J., 2000. Causes of climate change over the past 1000 years. Science 289, 270–277.

10. References 112

D’Arrigo, R.D., Jacoby, G.C., Free, R.M., 1992. Tree-ring width and maximum latewood density at the North American tree line: parameters of climatic change. Canadian Journal of Forest Research 22, 1290–1296.

D’Arrigo, R.D., Jacoby, G.C., 1993. Secular trends in high northern latitude temperature reconstructions based on tree rings. Climatic Change 25, 163–177.

D’Arrigo, R.D., Mashig, E., Frank, D.C., Jacoby, G.C., Wilson, R., 2004. Reconstructed warm season temperature for Nome, Seward, Peninsula, Alaska. Geophysical Research Letters 31, L09202, Doi: 10.1029/2004GL019756.

Dai, A., Trenberth, K.E., Karl, R., 1999. Effects of clouds, soil moisture, precipitation, and water vapour on diurnal temperature range. Journal of Climate 12(8), 2451–2473.

Dai, A., Trenberth, K.E., Qian, T., 2004. A global data set of Palmer Drought Severity Index for 1870–2002: relationship with soil moisture and effects of surface warming. Journal of Hydrometeorology 5, 1117–1130.

Davi, N.K., D’Arrigo, R.D., Jacoby, G.C., Buckley, B., Kobayashi, O., 2002. Warm-season annual to decadal temperature variability for Hokkaido, Japan, inferred from maximum latewood density (AD 1557–1990) and ring width data (AD 1523–1990). Climatic Change 52, 201–217.

Davi, N.K., Jacoby, G.C., Wiles, G.C., 2003. Boreal temperature variability inferred from maximum latewood density and tree-ring width data, Wrangell Mountain region, Alaska. Quaternary Research 60, 252–262.

Devi, N., Hagedor, F., Moiseev, P., Bugmann, H., Shiyatov, S., Mazepa, V., Rigling, A., 2008. Expanding forests and changing growth forms of Siberian larch at the Polar Urals treeline during the 20th century. Global Change Biology 14, 1581–1591.

Diaz, H.F., Bradley, R.S., 1997. Temperature variations during the last century at high elevation sites. Climatic Change 36, 253–379.

Ding, Y.H., 1992. Effects of the Qinghai-Xizang (Tibetan) Plateau on the circulation features over the plateau and its surrounding areas. Advances in Atmospheric Sciences 9(1), 112–130. (in Chinese with English abstract).

Dormont, L., Baltensweiler, W., Choquet, R., Roques, A., 2006. Larch- and pine-feeding host races of the larch bud moth (Zeiraohera diniana) have cyclic and synchronous population fluctuations. Oikos 115, 299–307.

10. References 113

Easterling, D.R., Horton, B., Jones, P.D., Peterson, T.C., Karl, T.R., Parker, D.E., Salinger, M.J., Razuvayev, V., Plummer, N., Jamerson, P., Folland, C., 1997. Maximum and minimum temperature trends for the global. Science 277, 364–367.

Esper, J., Cook, E.R., Schweingruber, F.H., 2002. Low-frequency signals in long tree-ring chronologies for reconstructing past temperature variability. Science 295, 2250–2253.

Esper, J., Shiyatov, S.G., Mazepa, V.S., Wilson, R.J.S., Graybill, D.A., Funkhouser, G., 2003a. Temperature-sensitive Tien Shan tree ring chronologies show multi-centennial growth trends. Climate Dynamics 21, 699–706.

Esper, J., Cook, E.R., Krusic, P.J., Peters, K., Schweingruber, F.H., 2003b. Tests of the RCS method for preserving low-frequency variability in long tree-ring chronologies. Tree-Ring Research 59, 81–98.

Esper, J., Büntgen, U., Frank, D.C., Nievergelt, D., Liebhold, A., 2007. 1200 years of regular outbreaks in apline insects. Proceedings of the Royal Society 274, 671–679.

Ettl, G.J., Peterson, D.L., 1995. Growth response of subalpine fir (Abies lasiocarpa) to climate in the Olympic Mountains, Washington, USA. Global Change Biology 1, 213–230.

Frank, D., Esper, J., 2005. Temperature reconstructions and comparisons with instrumental data from a tree-ring network for the European Alps. International Journal of Climatology 25(11), 1437–1454.

Frank, D., Esper, J., Cook, E.R., 2007. Adjustment for proxy number and coherence in a large- scale temperature reconstruction. Geophysical Research Letters 34 L16709, Doi: 10.1029/2007GL030571.

Fraver, S., Seymour, R. S., Speer, J.H., White, A.S., 2007. Dendrochronological reconstruction of spruce budworm outbreaks in northern Maine, USA. Canadian Journal of Forest Research 37, 523–529.

Frenzel, B., Bräuning, A., Adamczyk, S., 2003. On the problem of possible last-glacial forest- refuge-areas within the deep valleys of Eastern Tibet. Erdkunde 57, 182–198.

Friedrichs, D.A., 2008. Spatio-temporal patterns of tree-growth response to climate change. PhD dissertation, University of Bonn, Germany. pp. 1–98.

Fritts, H.C., 1971. Dendroclimatology and dendroecology. Quaternary Research 1, 419–449.

Fritts, H.C., 1976. Tree-rings and climate. Academic Press, London.

10. References 114

Fu, C., Fletcher, J.O., 1985. The relationship between Tibet tropical ocean thermal contrast and interannual variability of Indian monsoon rainfall. Journal of Climate and Applied Meteorology 24, 841–847.

Fuchs, T., Schneider, U. Rudolf, B., 2007. Global precipitation analysis products of the GPCC. Global Precipitation Climatology Centre (GPCC), DWD, Internet publication, 1–10.

Girardin, M.P., Tardif, J., Bergeron, Y. 2001. Radial growth analysis of Larix laricina from the Lake Duparquet area, Québec, in relation to climate and larch sawfly outbreaks. Ecoscience 8(1): 127–138.

Gou, X.H., Chen, F.H., Wang, Y.J., Shao, X.M., 2001. Spring precipitation reconstructed in the east of Qilian Mountain during the last 280 a by tree ring width. Journal of Glaciology and Geocryology 23(3), 292–262. (in Chinese with English abstract).

Gou, X.H., Chen, F.H., Yang, M.X., Li, J.B., Peng, J.F., Jin, L.Y., 2005. Climatic response of thick leaf spruce (Picea crassifolia) tree-ring width at different elevations over Qilian Mountains, northwestern China. Journal of Arid Environment 61, 513–524.

Gou, X.H., Chen, F.H., Cook, E.R., Jacoby, G.C., Yang, M.X., Li, J.B., 2007a. Streamflow variations of the Yellow River over the past 593 years in western China reconstructed from tree rings. Water Resources Research 43, W06434. Doi: 10.1029/2006WR005705.

Gou, X.H., Chen, F.H., Jacoby, G.C., Cook, E.R., Yang, M.X., Peng, J.F., Zhang, Y., 2007b. Rapid tree growth with respect to the last 400 years in response to climate warming, northeastern Tibetan Plateau. International Journal of Climatology 27, 1497–1503.

Gou, X.H., Peng, J.F., Chen, F.H., Yang, M.X., Levia, D.F., Li, J.B., 2008a. A dendrochronological analysis of maximum summer half-year temperature variations over the past 700 years on the northeastern Tibetan Plateau. Theoretical and Applied Climatology 93(3–4), 195–206.

Gou, X.H., Chen, F.H., Yang, M.X., Gordon, J., Fang, K.Y., Tian, Q.H., Zhang, Y., 2008b. Asymmetric variability between maximum and minimum temperatures in Northeastern Tibetan Plateau: Evidence from tree rings. Science in China (Series D) 51, 41–55.

Graham, E.A., Mulkey, S.S., Kitajima, K., Phillips, N.G., Wright, S.J., 2003. Cloud cover limits

net CO2 uptake and growth of a rainforest tree during tropical rainy seasons. Proceedings of the National Academy of Sciences 100, 572–576.

Grießinger, J., Bräuning, A., Thormas, A., Schleser, G.H., 2008. Stable oxygen isotopes in juniper trees from the Tibetan plateau as a proxy for monsoonal activity. TRACE: Tree Rings in Archaeology, Climatology and Ecology, Vol. 6, 92–95.

10. References 115

Grudd, H., 2008. Torneträsk tree-ring width and density AD 500–2004: a test of climatic sensitivity and a new 1500-years reconstruction of north Fennoscandian summers. Climate Dynamics 31, 843–857. Doi: 10.1007/s00382-007-0358-2.

Guiot, J., 1991. The bootstrapped response function. Tree-Ring Bulletin 51, 39–41.

Guo, H.J., Long, C.L., 1998. Yunnan’s Biodiversity. Yunnan Science and Technology Press, Kunming, China. (in Chinese).

He, Y.Q., Zhang, Z.L., Yao, T.D., Chen, T., Pang, H.X., Zhang, D., 2003. Modern changes of the climate and glaciers in China’s monsoonal temperature glacier region. Acta Geographica Sinica 58(4), 550–558. (in Chinese with English Abstract).

Heger, L., Parker, M.L., Kennedy, R.W., 1974. X-ray densitometry: a technique and an example of application. Wood Science 7, 140–148.

Helle, G., Schleser, G.H., Bräuning, A., 2002. Climate history of the Tibetan Plateau for the last 1500 years as inferred from stable isotopes in tree rings. Proceeding of the International Conference on the Study of Environmental Change Using Isotope Techniques, IAEA CN– 80–80, Vienna 22–27–04–2001, 301–311.

Hogg, E.H., Brandt, J.P., Kochtubajda, B., 2002. Growth and dieback of aspen forests in northwestern Alberat, Canada, in relation to climate and insects. Canadian Journal of Forest Research 32, 823–832.

Holmes, R.L., Swetnam, T.W., 1996. Program OUTBREAK user manual, Dendrochronoloy program library. Laboratory of Tree-Ring Research, University of Arizona, Tucson, Arizon, USA.

Holmes, J.A., Cook, E.R., Yang, B., 2009. Climate change over the past 2000 years in Western China. Quaternary International 194(1–2), 91–107.

Huang, J., Tardif, J., Denneler, B., Bergeron, Y., Berninger, F., 2008. Tree-ring evidence extends the historic northern range limit of severe defoliation by insects in the aspen stands of western Quebec, Canada. Canadian Journal of Forest Research 38, 2535–2544.

Hughes, M.K., 1992. Dendroclimatic evidence from the western Himalaya. In: Bradley, R.S., Jones, P.D., (Eds.) Climate since A.D. 1500. Routledge, London, pp. 415–431.

Hughes, M.K., 2001. An improved reconstruction of summer temperature at Srinagar, Kashmir since 1660 AD, based on tree-ring width and maximum latewood density of Abies pindrow [Royle] Spach. The Palaeobotanist 50, 13–19.

10. References 116

Hughes, M.K., 2002. Dendrochronology in climatology - the state of the art. Dendrochronologia 20, 95–116.

Hulme, M., Zhao, Z., Jiang, T., 1994. Recent and future climatic change in East Asia. International Journal of Climatology 14, 637–658.

IPCC, 2007. Summary for Policymakers. In: Solomon S., Qin, D., Manning M., Chen, Z., Marquis, M., Averyt K.B., Tignor, M., Miller H.L., (Eds.) Climate change 2007: The physical science basis. Contribution of working group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, New York, pp. 1–18.

Jacoby, G.C., D’Arrigo, R.D., Davaajamts, T., 1996. Mongolian tree rings and 20th-century warming. Science 273, 771–773.

Jones, P.D., Hulme, M., 1996. Calculating regional climatic time series for temperature and precipitation: methods and illustrations. International Journal of Climatology 16, 361–377.

Jones, P.D., Mann, M.E., 2004. Climate over past millennia. Reviews of Geophysics 42, 1–42.

Jongman, R.H.H., ter Braak, C.J.F., van Tongeren, O.F.R., 1987. Data analysis in community and landscaope ecology. Centre for Agricultural Publishing and Documentation, Wageningen, 299p.

Kane, R.P., 2006. Unstable ENSO relationship with Indian regional rainfall. International Journal of Climatology 26, 771–783.

Kang, X.C., Graumlich, L. J., Sheppard, P. R., 1997. A 1835–year tree-ring chronology and its preliminary analyses in Dulan region, Qinghai, Chinese Science Bulletin 42(13), 1122–1124. (in Chinese).

Kang, X.C., Cheng, G.D., Kang, E.S., Zhang, Q.H., 2002. Mountain pass run off reconstruction of Heihe River during past 1000 years using tree-ring. Science in China (Series D) 32(8), 675–685. (in Chinese).

Kirdyanov, A.V., Treydte, K.S., Nikolaev, A., Helle, G., Schleser, G.H., 2008. Climate signals in tree-ring width, density and δ13C from larches in Eastern Siberia (Russia). Chemical Geology 252, 31–41.

Kohler, M.A., 1949. On the use of double-mass analysis for testing the consistency of meteorological records and for making required adjustments. Bulletin of the American Meteorological Society 30, 188–189.

10. References 117

Körner, C., 1998. A re-assessment of high elevation treeline positions and their explanation. Oecologia 115, 445–459.

Kripalani, R.H., Ashwini, K., Sanade, S.S., 2003. Western Himalayan snow cover and Indian monsoon rainfall: a re-examination with INSAT and NCEP/NCAR data. Theoretical and Applied Climatology 74, 1–18.

Kurz, W.A., Dymond, C.C., Stinson, G., Rampley, G.J., Neilson, E.T., Carroll, A.L., Ebata, T., Safranyik, L., 2008. Mountain pine beetle and forest carbon feedback to climate change. Nature 452, 987–990.

La Marche, V.C., 1974. Paleoclimatic inferences from long tree-ring records. Science 183, 1043–1048.

Lazarus, B.E., Schaberg, P.G., DeHayes, D.H., Hawley, G.J., 2004. Severe red spruce winter injury in 2003 creates unusual ecological event in the northeastern United States. Canadian Journal of Forest Research 34, 1784–1788.

LeBlanc, D., Terrell, M., 2001. Dendroclimatic analyses using Thornthwaite–Mather–Type evapotranspiration models: a bridge between dendroecology and forest simulation models. Tree-ring Research 57(1), 55–66.

Legendre, P., Legendre, L., 1998. Numerical ecology. Elsevier Scientific, New York.

Li, J.B., Cheng, F.H., Cook, E.R., Guo, X.H., Zhang, Y.X., 2006. Drought reconstruction for north central China from tree rings: the value of the Palmer drought severity index. International Journal of Climatology 27, 903–909.

Li, J.B., Cook, E.R., D’Arrigo, R., Chen, F., Gou, X.H., Peng, J.F., Huang, J.G., 2008a. Common tree growth anomalies over the northeastern Tibetan Plateau during the last six centuries: implications for regional moisture change. Global Change Biology 14, 2096–2107. Doi: 10.1111/j.1365-2486.2008.01603.x.

Li, J.B., Cook, E.R., D’Arrigo, R., Chen, F.H., Gou, X.H., 2008b. Moisture variability across China and Mongolia: 1951–2005. Climate Dynamics Doi: 10.1007/s00382-008-0436-0.

Li, M.H., Xiao, W.F., Wang, S.G., Cheng, G.W., Cherubini, P., Cai, X.H., Liu, X.L., Wang, X.D., Zhu, W.Z., 2008. Mobile carbohydrates in Himalayan treeline trees I. Evidence for carbon gain limitation but not for growth limitation. Tree Physiology 28, 1287–1296.

Li. C., Yanai, M., 1996. The onset and interannual variability of the Asian summer monsoon in relation to land-sea thermal contrast. Journal of Climate 9, 358–375.

10. References 118

Liang, E.Y., Shao, X.M., Eckstein, D., Huang, L., Liu, X.H., 2006a. Topography- and species- dependent growth responses of Sabina prezewalskii and Picea crassifolia to climate on the northeast Tibetan Plateau. Forest Ecology and Management 236, 268–277.

Liang, E.Y., Liu, X.H., Yuan, Y.J., Qin, N.S., Fang, X.Q., Huang, L., Zhu, H.F., Wang, L.L., Shao, X.M. 2006b. The 1920s drought recorded by tree rings and historical documents in the semi-arid and arid area of Northern China. Climatic Change 79, 403–432.

Liang, E.Y., Shao, X.M., Qin, N.S., 2008. Tree-ring based summer temperature reconstruction for the source region of the Yangtze River on the Tibetan Plateau. Global and Planetary Change 61, 313–320.

Liu, L.S., Shao, X.M., Liang, E.Y., 2006. Climate signals from tree ring chronologies of the upper and lower treelines in the Dulan region of the northeastern Qinghai-Tibetan Plateau. Journal of Integrative Plant Biology 48(3), 278–285.

Liu, X.D., Chen, B.D., 2000. Climatic warming in the Tibetan Plateau during recent decades. International Journal of Climatology 20, 1729–1742.

Liu, X.H., Qin, D.H., Shao, X.M., Chen, T., Ren, J.W., 2003. Climatic significance of stable carbon isotope in tree rings of Abies spectabibis in southeastern Tibet. Chinese Science Bulletin 48(18), 2000–2004.

Liu, X.H., Qin, D.H., Shao, X.M., Chen, T., Ren, J.W., 2005. Temperature variations recovered from tree-rings in the middle Qilian Mountains over the last millennium. Science in China (Series D) 48(4), 521–529.

Liu, Y., An, Z.S., Ma, H.Z., Cai, Q.F., Liu, Z.Y., Kutzbach, J.K., Shi, J.F., Song, H.M., Sun, J.Y., Yi, L., Li, Q., Yang, Y.K., Wang, L., 2006. Precipitation variation in the northeastern Tibetan Plateau recorded by the tree rings since 850 AD and its relevance to the Northern Hemisphere temperature. Science in China (Series D) 49, 408–420.

Luckman, B.H., 1996. Dendrochronology in global change research. In: Dean, J.S., Meko, D.M., Swetnam, T.W., (Eds.) Tree rings, environment and humanity. Tucson, Radiocarbon. pp. 3– 24.

Luckman, B.H., Wilson, R.J.S., 2005. Summer temperature in the Canadian Rockies during the last millennium: a revised record. Climate Dynamics 24, 131–144.

Luo, Z.F., 1989. Study on the bionomic characteristics of Cosmolriche saxosimilis Lajonquiere. Yunnan Forestry and Technology 1, 63–73. (in Chinese).

10. References 119

Mann, M.E., Lees, J., 1996. Robust estimation of background noise and signal detection in climatic time series. Climate Change 33, 409–445.

Mann, M.E., Bradley, R.S., Hughes, M.K., 1998. Global-scale temperature patterns and climate forcing over the past six centuries. Nature 392, 779–787.

Mann, M.E., Bradley, R.S., 1999. Northern hemisphere temperatures during the past millennium: inferences, uncertainties, and limitations. Geophysical Research Letters 26, 759–762.

Mann, M.E., Zhang, Z.H., Hughes, M.K., Bradley, R.S., Miller, S.K., Rutherford, S., Ni, F.B., 2008. Proxy-based reconstructions of hemispheric and global surface temperature variations over the past two millennia. Proceedings of the National Academy of Sciences 105(36), 13252–13257. Doi: 10.1073/pnas.0805721105.

McCarroll, D., Jalkanen, R., Hicks, S., Tuovinen, M., Gagen, M., Pawellek, F., Eckstein, D., Schmitt, U., Autio, J., Heikkinen, O., 2003. Multiproxy dendroclimatology: a pilot study in northern Finland. The Holocene 13 (6), 831–841.

Meehl, G.A., Washington, W.M., 1993. South Asian summer monsoon variability in a model with doubled atmospheric carbon dioxide concentration. Science 260, 1101–1104.

Michaelsen, J., 1987. Cross-validation in statistical climate forecast models. Journal of Climate and Applied Meteorology. 26, 1589–1600.

Mitchell, T.D., Jones, P.D., 2005. An improved method of constructing a database of monthly climate observations and associated high resolution grids. International Journal of Climatology 25, 693–712.

Moberg, A., Sonechkin, D.M., Holmgren, K., Datsenko, N.M., Karlen, W., 2005. Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy data. Nature 433, 613–617.

Murakami, T., 1987. Effects of the Tibetan Plateau. In: Chang, C.P., Krishnamurti, T.N., (Eds.) Monsoon meteorology. Oxford Monographs on Geology and Geophysics 7, 235–270.

Nola, P., Morales, M., Motta, R., Villalba, R., 2006. The role of larch budmoth (Zeiraphera diniana Gn.) on forest succession in a larch (Larix decidua Mill.) and Swiss stone pine (Pinus cembra L.) stand in the Susa Valley (Piedmont, Italy). Trees-Structure and Function 20(3), 371–382.

Oberhuber, W., Stumböckm M., Kofler, W., 1998. Climate-tree-growth relationships of Scots pine stands (Pinus szlvestris L.). Trees-Structure and Function 13, 19–27.

10. References 120

Osborn, T.J., Briffa, K.R., Jones, P.D., 1997. Adjusting variance for sample size in tree-ring chronologies and other regional mean timeseries. Dendrochronologia 15, 89–99.

Palmer, W.C., 1965. Meteorological drought. Weather Bureau Research Paper 45. U.S. Department of Commerce, Washington, DC, p. 58.

Parish, R., Antos, J. A., 2002. Dynamics of an old-growth, fire-initiated, subalpine forest in southern interior British Columbia: tree-ring reconstruction of 2 year cycle spruce budworm outbreaks. Canadian Journal of Forest Research 32, 1947–1960.

Pederson, N., Cool, E.R., Jacoby, G.C., Peteet, D.M., Griffin, K.L., 2004. The influence of winter temperatures on the annual radial growth of six northern range margin tree species. Dendrochronologia 22, 7–29.

Pederson, N., Jacoby, G.C., D’Arrigo, R.D., Cook, E.R., Buckley, B.M., 2001. Hydrometeorological reconstructions for Northeastern Mongolia derived from tree rings: 1651–1995. Journal of Climate 14, 872–881.

Peng, J.F., Gou, X.H., Chen, F.H., Li, J.B., Liu, P.X., Zhang, Y., 2008. Altitudinal variability of climate-tree growth relationships along a consistent slope of Anyemaqen Mountains, northeastern Tibetan Plateau. Dendrochronologia 26(2), 87–96.

Peterson, D.W., Peterson, D.L., 1994. Effects of climate on radial growth of sub-alpine conifers in the North Cascade Mountains. Canadian Journal of Forest Research 24, 1921–1932.

Peterson, D.W., Peterson, D.L., 2001. Mountain hemlock growth responds to climatic variability at annual and decadal time scales. Ecology 82, 3330–3345.

Pickett, S.T.A., White, P.S., 1985. The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York.

Qin, N.S., Shao, X.M., Jin, L.Y., Wang, Q.C., Zhu, X.D., Wang, Z.Y., Li, J.B., 2003. Climate change over southern Qinghai Plateau in the past 500 years recorded in Sabina tibetica tree rings. Chinese Science Bulletin 48(22), 2484–2488.

Richman, M.B., 1986. Rotation of principal components. Journal of Climate 6, 293–355.

Rinn, F., 2003. TSAP-Win: Time series analysis and presentation for dendrochronology and related applications. Version 0.55, User reference. Heidelberg, Germany.

Rock, J.F., 1926. Through the great river trenches of Asia. National Geographic 50(2), 135–186.

Rolland, C., Baltensweiler, W., Petitcolas, V., 2001. The potential for using Larix decidua ring widths in reconstructions of larch budmoth (Zeiraphera diniana) outbreak history:

10. References 121

dendrochronological estimates compared with insects surveys. Trees-Structure and Function 15, 414–424.

Saxe, H., Cannell, M.G.R., Johnsen, B., Ryan, M.G., Vourlitis, G., 2001. Tree and forest functioning in response to global warming. New Phytologist 149, 369–399.

Schinker, M.G., Jamsem, N., Spiecker, H., 2003. High-frequency densitometry - a new method for rapid evaluation of wood density variations. IAWA Journal 24(3), 231–239.

Schweingruber, F.H., Fritts, H.C., Bräker, O.U., Drew, L.G., Schär, E., 1978. The X-ray technique as applied to dendroclimatology. Tree-Ring Bulletin 38, 61–91.

Schweingruber, F.H., 1979. Auswirkungen des Lärchenwicklerbefalls auf die Jahrringstruktur der Lärche. Schweizerische Zeitschrift für Forestwesen 130, 1071–1093.

Schweingruber, F.H., 1988. Tree rings - basics and applications in dendrochronology. Reidel, Dordrecht, Boston, London, p. 276.

Schweingruber, F.H., Briffa, K.R., Nogler, P., 1993. A tree-ring densitometric transect from Alaska to Labrador: comparison of ring-width and maximum-latewood-density chronologies in the conifer belt of northern North America. International Journal of Biometeorology 37, 151–169.

Schweingruber, F.H., 1996. Tree rings and environment dendroecology. Haupt, Bern, p. 609.

Schweingruber, F.H., Briffa, K.R., 1996. Tree-ring density networks of climate reconstruction. In: Jones, P.D., Bradley, R.S., Jouzel, J., (Eds.) Climatic variations and forcing mechanisms of the last 2000 years. NATO ASI Series Vol. 41. Springer-Verlag. Berlin, pp. 43–46.

Shao, X.M., Fan, J.M., 1999. Past climate on west Sichuan plateau as reconstructed from ring- widths of dragon spruce. Quaternary Sciences 1, 81–89. (in Chinese with English abstract).

Shao, X.M., Fang, X.Q., Liu, H.Y., Huang, L., 2003. Dating the 1000–year-old Qilian juniper in Mountains along the eastern margin of the Qaidam Basin. Acta Geographical Sinica 58, 90– 100. (in Chinese with English abstract).

Shao, X.M., Huang, L., Liu, H.B., Liang, E.L., Feng, X.Q., Wang, L.L., 2005. Reconstruction of precipitation variation from tree rings in recent 1000 years in Delingha, Qinghai. Science in China (Series D) 48(7), 939–949.

Shao, X.M., Wang, S.Z., Xu, Y., Zhu, H.F., Xu, X.G., Xiao, Y.M., 2007. A 3500–year master tree-ring dating chronology from the northeastern part of Qaidam Basin. Quaternary Sciences 27(4), 477–485. (in Chinese with English abstract)

10. References 122

Sheppard, P.R., Tarasov, P.E., Graumlich, L.J., Heussner, K.-U., Wagner, M., Österle, H., Thompson, L.G., 2004. Annual precipitation since 515 BC reconstructed from living and fossil juniper growth of northeastern Qinghai Province, China. Climate Dynamics 23, 869– 881.

Shi, Y.F., 2002. Characteristics of late Quaternary monsoonal glaciations on the Tibetan Plateau and in East Asia. Quaternary International 97–98, 79–81.

Singh, J., Park, W.-K., Radav, R.R., 2006. Tree-ring-based hydrological records for western Himalaya, India, since A.D. 1560. Climate Dynamics 26, 295–303.

Smith, T.M., Reynolds, R.W., 2004. Improved extended reconstruction of SST (1854–1997). Journal of Climate 17, 2466–2477.

Speer, J.H., Swetnam, T.W., Wickman, B.E., Youngblood, A., 2001. Changes in pandora moth outbreak dynamics during the past 622 years. Ecology 82(3), 679–697.

Stokes, M.A., Smiley T.L., 1996. An introduction to tree-ring dating. The University of Arizona Press, Tucson.

Swetnam, T. W., Thompson, M. A., Sutherland, E.K., 1985. Using dendrochronology to measure radial growth of defoliated trees. U.S. Forest Service, Agriculture Handbook, No. 639.

Swetnam, T. W., Lynch, A. M., 1993. Multicentury, regional-scale patterns of western spruce budworm outbreaks. Ecological Monographs 63(4), 399–424.

Tardif, J.C, Stevenson, D., 2001. Radial growth-climate association of Thuja occidentalis L. at the northwestern limit of its distribution, Manitoba, Canada. Dendrochronologia 19(2), 179– 187.

Tardif, J.C., Conciatori, F., Nantel, P., Gagnon, D., 2006. Radial growth and climate responses of white oak (Quercus alba) and northern red oak (Quercus rubra) at the northern distribution limit of white oak in Quebec, Canada. Journal of Biogeography 33, 1657–1669. ter Braak, C.J.F., Smilauer, P., 2002. CANOCO reference manual and CanoDraw for user’s guide Windows: software for canonical community ordination, Version 4.5. Microcomputer Power, Ithaca, NY.

Tessier, L., Guibal, F., Schweingruber, F.H., 1997. Research strategies in dendroecology and dendroclimatology in mountain environments. Climatic Change 36, 499–517.

The Editing Committee of “The annals of Deqin county”, 1997. The annals of Deqin county. Yunnan Nationality Press, Kunming, p. 68. (in Chinese).

10. References 123

The Editing Committee of “The Baima national nature reserve”. 2003. The Baima national nature reserve. Yunnan Nationality Press, Kunming, pp. 218–230. (in Chinese)

Thuiller, W., Lavorel, S., Arujo. M.B., Sykes, M.T., Prentice, I.C., 2005. Climate change threats to plant diversity in Europe. Proceedings of the National Academy of Sciences 102, 8245– 8250.

Tian, L., Masson-Delmotte, V., Stievenard, M., Yao, T., Jouzel, J., 2001. Tibetan Plateau summer monsoon northward extent revealed by measurements of water stable isotopes. Journal of Geophysical Research 106(D22), 28081–28088.

Tian, L., Yao, T., MaClune, K., White, J.W.C., Schilla, A., Vaughn, B., Vachon, R., Ichiyanagi, K., 2007. Stable isotopic variations in west China: A consideration of moisture sources. Journal of Geophysical Research 112(D10112), doi: 10.1029/2006JD007717.

Tian, Q.H., Gou, X.H., Zhang, Y., Peng, J.F., Wang, J.S., Chen, T., 2007. Tree-ring based drought reconstruction (A.D. 1855–2001) for the Qilian Mountains, Northwestern China. Tree-Ring Research 63(1), 27–36.

Trotter, R.T., Cobb, N.S., Whitham, T.G., 2002. Herbivory, plant resistance, and climate in the tree ring record: interactions distort climatic reconstruction. Proceedings of the National Academy of Sciences 99, 10197–10202.

Trouet, V., Haneca, K., Coppin, P., Beeckman, H., 2001. Tree ring analysis of Brachystegia spiciformis and Isoberlinia tomentosa: evaluation of the ENSO-signal in the miombo- woodland of eastern Africa. IAWA Journal 22: 385–399.

Vaganov, E.A., Hughes, M.K., Shashkin, A.V., 2006. Growth dynamics of conifer tree rings- images of past and future environments. Ecological Studies 183, Springer-Verlag, Berlin, Heidelberg. van der Schrier, G., Briffa, K.R., Jones, P.D., Osborn, T.J., 2006. Summer moisture variability across Europe. Journal of Climate 19, 2818–2834. Doi: 10.1175/JCLI3734.1.

Villalba, R., Veblen, T.T., Ogden, J., 1994. Climatic influences on the growth of subalpine trees in the Colorada Front Range. Ecology 75, 1450–1462.

Wang, B., 2006. The Asian Monsoon. Spring/Praxis Publishing Ltd, Chichester, UK. pp 787.

Wang, L.L., Payette, S., Bégin, Y., 2001. 1300–year tree-ring width and density series based on living, dead and subfossil black spruce at tree-line in Subarctic Québec, Canada. The Holocene 11(3), 333–341.

10. References 124

Wang, N.L., Yao, T.D., Pu, J.C., Zhang, Y.L., Sun, W.Z., Wang, Y.Q., 2003. Variations in air temperature during the last 100 years revealed by δ18O in the Malan ice core from the Tibetan Plateau. Chinese Science Bulletin 48, 2134–2138.

Wang, S.W., Gong, D.Y., 2000. Enhancement of the warming trend in China. Geophysical Research Letters 27(16), 2581–2584.

Wang, S.W., Gong, D.Y., Zhu, J.H., 2001. Twentieth-century climatic warming in China in the contest of the Holocene. The Holocene 11(3), 313–321.

Wang, S.W., Zhu, J.H., Cai, J.N., 2004. Interdecadal variability of temperature and precipitation in China since 1880. Advances in Atmospheric Sciences 21(3), 307–313.

Weber, U. M., 1997. Dendroecological reconstruction and interpretation of larch budmoth (Zeiraphera diniana) outbreaks in two central Alpine valleys of Switzerland from 1470– 1990. Trees-Structure and Function 11(5), 277–290.

Webster, P. J., Magana, V.O., Palmer, T.N, Shukla, J., Tomas, R. A., Yanai, M., Yasunari, T., 1998. Monsoons: Processes, predictability, and the prospects for prediction. Journal of Geophysical Research 103(C7), 14451–14510.

Wen, C.J., 1989. Influence of the relief on conditions of water and heat in the Hengduan Mountains region. Mountain Research 7(1), 66–73. (in Chinese with English abstract).

White, P.S., Harrod, J., Romme, W.H., Betancourt, J., 1999. Disturbance and temporal dynamics. In: Szaro, R.C., Johnson, N.C., Sexton, W.T., Malk, A.J., (Eds.) Ecological stewardship: A common reference for ecosystem management. Elsevier Science, Oxford. Volume 2, 281– 312.

Wigley, T., Briffa, K.R., Jones, P.D., 1984. On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Applied Meteorology and Climatology 23, 201–213.

Wilson, R.J.S., Luckman, B.H., 2002. Tree-ring reconstruction of maximum and minimum temperatures and the diurnal temperature range in British Columbia, Canada. Dendrochronologia 20(3), 257–268.

Wilson, R.J.S., Luckman, B.H., 2003. Dendroclimatic reconstruction of maximum summer temperatures from upper treeline sites in Interior British Columbia, Canada. The Holocene 13 (6), 851–861.

10. References 125

Wilson, R., Tudhope, A., Brohan, P., Briffa, K., Osborn, T., Tett, S., 2006. Two-hundred-fifty years of reconstructed and modeled tropical temperatures. Journal of Geophysical Research 111, C10007, Doi: 10.1029/2005JC0003188.

Winkler, D., 1996. Forests, forest economy and deforestation in the Tibetan Prefectures of West Sichuan. Commonwealth Forest Review 75(4), 295–301.

Wu, P., Wang, L.L., Shao, X.M., 2005. Reconstruction of summer temperature from maximum latewood density of Pinus densata in west Sichuan. Acta Geographica Sinica 60(6), 998–1 006. (in Chinese with English abstract).

Wu, P., Wang, L.L., Huang, L., 2006. A preliminary study on the tree-ring sensitivity to climage change of five endemic conifer species in China. Geographical Research 25 (1), 43–52. (in Chinese with English abstract).

Wu, X.D., Lin, Z.Y., 1981. Some characteristics of the climatic changes during the historical time of Qinghai-Xizang Plateau. Acta Meteorologica Sinica 39 (1), 90–97. (in Chinese).

Wu, X.D., Li, Z.Y., Sun, L., 1988. A preliminary study on the climatic change of the Hengduan Mountains area since 1600 A.D. Advances in Atmospheric Sciences 5(4), 437–443.

Wu, X.D., Sun, L., Zhan, X.Z., 1989. A preliminary study on reconstructing past climate in the middle Xizang Plateau by using tree-ring data. Acta Geographica Sinica 44(3), 334–342. (in Chinese with English abstract).

Xu, X.D., Miao, Q.J., Wang, J., Zhang, X.J., 2003. The water vapour transport model at the regional boundary during the Meiyu Period. Advances in Atmospheric Sciences 20, 333–342.

Xu, J.C., Wilkes, A., 2004. Biodiversity impact analysis in northwest Yunnan, southwest China. Biodiversity and Conservation 13, 959–983.

Yadav, R.R., Park, W.-K., Singh, J., Dubey, B., 2004. Do the western Himalayas defy global warming? Geophysical Research Letters 31 L17201, Doi: 10.1029/2004GL020201.

Yang, B., Bräuning, A., Shi, Y.F., 2003. Late Holocene temperature fluctuations on the Tibetan Plateau. Quaternary Science Reviews 22, 2335–2344.

Yang, B., Bräuning, A., Dong, Z., Zhang, Z., Jiao, K., 2008. Late Holocene monsoonal temperature glacier fluctuations on the Tibetan Plateau. Global and Planetary Change 60, 126–140.

Yang, M.X., Yao, T.D., He, Y.Q., Thompson, L.G., 2000. ENSO events recorded in the Guliya ice core. Climatic Change 47, 401–409.

10. References 126

Yang, Q.Y., Shen, K.D., 1984. On vertical zonation of the Northwestern Yunnan. Acta Geographica Sinica 39(2), 141–147. (in Chinese with English abstract).

Yao, T.D., Qin, D.H., Xu, B.Q., Yang, M.X., Duan, K.Q., Wang, N.L., Wang, Y.Q., Hou, S.G., 2006a. Temperature change over the past millennium recorded in ice cores from the Tibetan Plateau. Advances in Climate Change Research 2(3), 99–103, (in Chinese with English abstract).

Yao, T.D., Guo, X.J., Thompson, L., Duan, K.Q., Wang, N.L., Pu, J.C., Xu, B.Q., Yang, X.X., Sun, W.Z., 2006b. δ18O record and temperature change over the past 100 years in ice cores on the Tibetan Plateau. Science in China (Series D) 49, 1–9.

Yasue, K., Funada, R., Kobayashi, O., Ohtani, J., 2000. The effects of tracheid dimensions on variations in maximum density of Picea glehnii and relationships to climatic factors. Trees- Structure and Function 14, 223–229.

Yu, Y.D., Liu, L.H., Zhang, J.H., 1989. Vegetation regionalization of the Hengduan Mountains region. Mountain Research 7(1), 47–55. (in Chinese with English abstract).

Yuan, Y.J., Li, J.F., 1999. Reconstruction and analysis of 450 years winter temperature series in Urumqi River Source of Tianshan Mountains. Journal of Glaciology and Geocryology 21(1), 64–70. (in Chinese with English abstract).

Zhang, Q.B., Alfaro, R.I., 2003. Spatial synchrony of the two-year cycle budworm outbreaks in central British Columbia, Canada. Oikos, 102, 146–154.

Zhang, Q.B., Cheng, G.D., Yao, T.D., Kang, X.C., Huang, J.G., 2003. A 2,326–year tree-ring records of climate variability on the northeastern Qinghai-Tibetan Plateau. Geophysical Research Letters 30, L141739, Doi: 10.1029/2003GL017425.

Zhang, Z.H., Wu, X.D., 1997. Climatic reconstruction in the Qilian Mountains during past 700 years using tree-ring data. Chinese Science Bulletin 42(8), 849–851. (in Chinese).

Zhao, J.C., Xu, J.C., Qi, K., 2001. Community Survey Report on the Natural Forest Protection and Upland Conservation Programs in Yunnan. Yunnan Science and Technology Publisher, Kunming, China. (in Chinese).

Zheng, B.X., Zhao, X.T., Li, T.S., Wang, C.Y., 1999. Features and fluctuation of the Melang Glacier in the Mainri Mountain. Journal of Glaciology and Geocryology 21(2), 145–150, (in Chinese with English abstract).

Zheng, D., 1996. The system of physico-geographical region of the Qinghai-Xizang (Tibet) Plateau. Science in China (Series D) 39, 410–417. (in Chinese with English abstract).

10. References 127

Zhuo, Z.D., 1981. The tendency on the change of climate and glacier in Qilianshan region based on the measurement of tree-rings of Sabina przewalskii. Acta Phytoecologia et Geobotanica Sinca 5(1), 12–26. (in Chinese).

Curriculum vitae 128

Curriculum Vitae

Personal data Name Ze-Xin Fan Date of Birth 05 June 1980 Place of Birth Yunnan, PR China

Education 1995/9–1998/7 High School in Zhaotong Prefecture No.1 Middle School

1998/9–2002/7 Bachelor in Biology School of Life Science, Southwest University (Chongqing, China)

2002/9–2005/7 Master in Ecology Xishuangbanna Tropical Botanical Garden, the Chinese Academy of Sciences (Kunming, China) Advisor: Prof. Dr. Kun-Fang Cao Master Thesis: Axial and radial variation of xylem anatomy of six angiosperms and three coniferous tree species in Yunnan, China

2005/11– Ph.D. candidate in Geography Institute of Geography, Friedrich-Alexander-University Erlangen-Nuremberg (Erlangen, Germany) Advisor: Prof. Dr. Achim Bräuning Ph.D. Thesis: Environmental and climatic history during the past centuries in the Hengduan Mountains (southwest China) derived from tree rings

Work 2005/7–2005/10 Project employer in the Xishuangbanna Tropical Botanical Garden, the Chinese Academy of Sciences

Fellowship 2005/11–2008/11 Ph.D. fellowship from the Max-Planck-Society

Curriculum vitae 129

Publications 130

List of Publications

(Peer reviewed articles)

1) Ze-Xin Fan, Achim Bräuning, Kun-Fang Cao, Shi-Dan Zhu. 2009. Growth-climate responses of high-elevation conifers in the central Hengduan Mountains in southern China. Forest Ecology and Management 258(3): 306-313. Doi: 10.1016/j.foreco.2009.04.017.

2) Ze-Xin Fan, Achim Bräuning, Kun-Fang, Cao. 2008. Annual temperature reconstruction from the central Hengduan Mountains, as deduced from tree rings. Dendrochronologia 26: 97–107.

3) Ze-Xin Fan, Achim Bräuning, Bao Yang, Kun-Fang Cao. 2009. Tree ring density-based summer temperature reconstruction for the central Hengduan Mountains in southern China. Global and Planetary Change 65: 1–11. Doi: 10.1016/j.gloplacha.2008.10.001.

4) Ze-Xin Fan, Achim Bräuning, Kun-Fang Cao. 2008. Tree-ring based drought reconstruction in the central Hengduan Mountains (China) since A.D. 1655. International Journal of Climatology 28: 1879–1887. Doi: 10.1002/joc.1689.

5) Ze-Xin Fan, Kun-Fang Cao, Peter Becker. 2009. Axial and radial variations in xylem anatomy of angiosperm and conifer trees in Yunnan, China. IAWA Journal 30(1): 1–13.

6) Ze-Xin Fan, Kun-Fang Cao, Shou-Qing Zou. 2005. Axial and radial changes in xylem anatomical characteristics in six evergreen broadleaved tree species in Ailao Mountain, Yunnan. Acta Phytoecologia Sinica 29(6): 968–975. (in Chinese with English abstract)

7) Ze-Xin Fan, Kun-Fang Cao. 2005. Hypothesis on the limitation of tree height growth. Chinese Bulletin of Botany 22(5): 632–640. (in Chinese with English abstract)

8) Z.-Q. CAI, M. SLOT and Z.-X. FAN. 2004. Leaf development and photosynthetic properties of three tropical tree species with delayed greening. Photosynthetica 43(1): 91–98.

Publications 131

(Conference series)

1) Ze-Xin Fan, Achim Bräuning. 2008. Growth-climate relationships of high-elevation conifers in the central Hengduan Mountains, China. TRACE – Tree Rings in Archaeology, Climatology and Ecology, Vol. 7, 50–56.

2) Ze-Xin Fan, Achim Bräuning. 2007. Tree-ring based drought reconstruction in the central Hengduan Mountains region (China) since A.D. 1655. TRACE – Tree Rings in Archaeology, Climatology and Ecology, Vol. 6, 57–62.

3) Ze-Xin Fan, Kun-Fang Cao, Peter Becker. 2006. Axial and radial changes in xylem anatomy of angiosperm and coniferous trees in Yunnan, China. 7th International Dendrochronology Conference, Beijing, China. 11–17th June, 2006.

4) Ze-Xin Fan, Kun-Fang Cao, Becker Peter. 2005. Axial and radial changes in xylem anatomy and theoretical hydraulic conductance of angiosperm and coniferous trees in Yunnan, China. Frontiers in Tropical Biology and Conservation - Annual Meeting of the Association for Tropical Biology and Conservation, Uberlandia, Brazil. 24–28th July, 2005.