Reconstruction of October Mean Temperature since 1796 in Wuying Based on Tree Ring Data

Abstract

The monthly mean temperature for October in the Fenglin National Natural Reserve of Wuying, in Heilongjiang province, was reconstructed for the period running from 1796 to 2004 using RES tree ring chronology. The explained variance of the reconstruction is 34.8%. In the past 209 years, there are 4 colder and 4 warmer periods according to the reconstructed series. A period of 3.33-year is found significant based on the power spectrum method. Abrupt changes are also detected in the reconstructed series with 30-year time scale based on the smoothing t -test, smoothing F -test and Le Page test methods. Significant abrupt changes in mean value are observed for around 1871 and 1900, and a significant abrupt change in standard deviation is observed for around 1851.

Keywords

Wuying ; tree ring ; temperature reconstruction

1. Introduction

Tree ring data have been applied widely in climate change research on historical periods because of their high veracity, continuity and high resolution, and because they are easy to obtain [ Fritts, 1991  ; Cook et al., 1991 ]. Recently, there has been large progress in international tree ring climatology, in which a number of studies reconstructed local and regional temperature change [ Briffa et al., 1995  ; Jacoby et al., 2000 , Davi et al., 2003  and Esper et al., 2002 ]. The research on tree ring climatology in China has focused on western arid and semi-arid area, and many temperature and precipitation series have been reconstructed [ Yuan and Li, 1999 , Liu et al., 2003  and Shao et al., 2005 ]. Futhermore, there are a few researches on tree ring in northeastern China. For example, maximum temperature series from January to April in Changbai Mountain were reconstructed using the tree ring chronology of Changbai larch, spruce-fir and korean pine [ Shao and Wu, 1997 ]. Minimum temperature series from February to September were reconstructed using multiple tree ring chronologies [ Zhu , 2006 ]. Xiaoxing’anling is located in the north of northeastern China. Cold has significant influence on agricultural production in the region. Therefore, it is very important to study climate change characteristics in Xiaoxing’anling. Most of the researches on climate change in northeastern China used to apply proxy data such as tree ring due to a deficit in instrumentation. However, few researches reconstructed the climate of Xiaoxing’anling by using tree ring data.

The object of the paper is to reconstruct monthly mean temperature series for October in Wuying since 1796 and to analyze the characteristics based on the tree ring chronology developed by the Fenglin National Natural Reserve of Wuying, in Heilongjiang province, and observed data recorded by Wuying meteorological station. The results could be used to validate climate models and predict climate change.

2. Data

2.1. Tree ring materials

The sampling site, Fenglin National Natural Reserve of Wuying, is located in the southern slope of Xiaoxing’anling, in Heilongjiang province. The plant cover is warm zone needle-leaves and broad-leaves forest, including red pine (Pinus koraiensis), spruce-fir, and larch. The dominant species are red pines. The forest stock of red pine accounts for 2/3 of the total storage. They kept their original state as they are not much influenced by human behavior. The canopy density is about 0.4. A sample collection was conducted in September 2006, and 21 Pinus koraiensis trees were selected, and 2 increment cores were taken from each tree, making a total of 42 cores. The sampling site was located in a high and cold area (48°14’N, 129°12’E, a.s.l. 360–460 m).

In laboratory, all samples were dried, mounted and sanded to smooth surface. Then cross-dating was done by using the Skeleton Plot method according to basic processing program of tree ring sample [ Stocks and Smiley, 1968 ]. The samples were measured within 0.01 mm. Quality control of crossdating and measurement was performed by the COFECHA program [ Holmes, 1983 ]. The cores which did not match the master series well were removed. The average correlation coefficient between trees was 0.566, the mean sensitivity was 0.197, and the autocorrelation coefficient was 0.819. These data showed that trees in this site are extensively controlled by the same climate factors. The lag effect is quite evident, and tree growth of the current year influences the next year growth.

Tree ring width chronologies were developed by the ARSTAN program [ Cook and Kairiukstis, 1990 ]. By comparing repeatedly original sample cores, we selected a spline function of 190-year step length to eliminate tree-age-related growing trends. The length of the established chronology is 282 years. To ensure the reliability of the reconstructed climate, a chronology length of 209 years from 1797 to 2005 was selected to reconstruct the climate. The number of sample core is 16 in 1797. The sub-sample signal strength is more than 0.85. Three kinds of chronologies were obtained, e.g., standard (STD), residual (RES), and autoregressive (ARS). Table 1 presents the statistics of the three chronologies at Fenglin, and Table 2 shows the statistics of common interval analysis (1862–2004) for standardized and residual series. Statistics from Table 1  and Table 2 showed that the generality of RES series is higher than STD series which means the numerous low frequency oscillations of STD series are non-synchronized, and the high frequency oscillation is synchronous. Thus, RES chronology is selected for the climate response analysis of tree growth.

Table 1. Statistics of three chronologies in Fenglin

Chronology type	Mean	Median	Mean sensitivity	Standard deviation	First order autocorrelation
STD	0.990	0.970	0.132	0.172	0.483
RES	0.992	0.985	0.148	0.134	−0.010
ARS	0.980	0.956	0.134	0.165	0.466

Table 2. Statistics of common interval analysis for standardized and residual series in Fenglin (1862–2004)

Statistical item	STD	RES
Mean correlation between all series	0.280	0.382
Mean correlation between trees	0.271	0.375
Mean correlation between samples of a tree	0.525	0.571
Signal/noise ratio	6.332	10.201
Variance of 1st PC (%)	31.51	40.40
Expressed population signal	0.864	0.911

2.2. Climatic data

Climatic data included dekad mean temperature, dekad total precipitation, monthly mean temperature, monthly mean minimum temperature, monthly mean maximum temperature, and monthly total precipitation from 1959 to 2004 at Wuying station (48°07’N, 129°15’E, a.s.l. 299.1 m). When data were absent in August 1961, data from Yichun and Wuyiling stations were taken by using regression interpolation. The consistency of climatic series is performed based on Mann-Kendall [ Mann, 1945 ] and Double-mass analysis method [ Kohler, 1949 ]. The results showed that temperature and precipitation records of Wuying station don’t have random abrupt changes and uneven distribution, and that it could represent the actual climate change. A first-order autocorrelation analysis of the monthly mean temperature and total precipitation showed that monthly climatic data don’t have significant lag effect year after year.

The annual total precipitation of Wuying varies from 470 to 900 mm, and precipitation from June to September occupies more than 70% of the total, and precipitation occurs mostly in July, about 150 mm. Annual mean temperature is about 0 °C. Monthly mean temperature is below 0 °C from January to March, November and December, and more than 10 °C from May to September (Fig. 1 ).

Figure 1. Climatic monthly total precipitation and monthly mean temperatures in Wuying (1959–2004 mean)

3. Tree growth response analysis to climatic elements

The relationships between tree growth and climatic factors were analyzed by using correlation function and response function [ Fritts, 1976 ]. The analysis period is 1959–2004. Climatic factors are monthly mean temperature and monthly total precipitation from September of last year to October of this year. The results showed that response of tree growth to climatic factors is relatively remarkable, and variations in monthly mean temperature and precipitation can explain 60.3% of the tree radial growth increment. Moreover, response of tree growth to temperature in the preceding October is more significant than other factors. The results of correlation analysis showed that there are significant positive correlations between RES chronology and monthly mean temperature, monthly mean maximum temperature, and monthly mean minimum temperature, among which correlation coefficients with monthly mean temperature are the highest. This has definite physiological significances. If soil temperature is low in the previous period of the growing season, leaf and root growth is limited, net photosynthesis is reduced, and tree nutrition transformation storage is weakened, so, narrow ring in the next year is easy to form [ Wu , 1990 ]. Pinus koraiensis belongs to the group of temperate deciduous arbors. Under the condition of natural carbon dioxide concentration and light saturation, minimum photosynthetic temperature, optimum photosynthetic temperature, and maximum photosynthetic temperature of temperate deciduous arbors are −3 to −1 °C, 15–25 °C, and 40–45 °C, respectively. Photosynthesis can’t occur when temperature is lower than the minimum photosynthetic temperature. The photosynthetic rate is constant when temperature is equal to the optimal photosynthetic temperature. The photosynthetic rate shows a linear increasing trend when temperature is between the minimum photosynthetic temperature and the lower limit of optimal photosynthetic temperature. The photosynthetic rate shows a linear decreasing trend when temperature is between the upper limit of optimal photosynthetic temperature and the maximum photosynthetic temperature. The research [ Wang , 1999 ] showed that a great diurnal temperature range is favorable for photosynthate accumulation, because during day time high temperature is beneficial to photosynthesis, and low temperature weakens the respiration consumption at night. The phenophase of natural red pine forest is affected significantly by temperature [ ECFC , 1999 ]. When the mean temperature increases up to 4.5 °C in the second or last ten days of April, sap begins to flow. If ground temperature is greater than 0 °C at the same time, root begins to respire. From May to September, the mean temperature of Wuying is more than 10 °C, and monthly mean maximum temperature is more than 18 °C, which is suitable for tree germination, leaf emergence, flowering, and fruition. In October, when the mean temperature is 2.0 °C, monthly mean maximum temperature is 9.2 °C, and monthly mean minimum temperature is −4.3 °C, the plant could still proceed the photosynthesis. Moreover, great diurnal temperature range is favorable to photosynthate accumulation.

In addition, the correlation between dekad temperature and precipitation in Wuyuing and Pinus koraiensis RES chronology showed that temperature variation in April has some influence on the radial growth of Pinus koraiensis, in accordance with the phonological characteristics of red pine forest [ ECFC , 1999 ] .

4. Temperature reconstruction and test

Table 3 showed that correlation coefficient is the highest between temperature of October in the current year (T_t ) and the next year tree chronology (RES_t+1 ). Further calculation elucidates that T_t is not significantly related with the third year tree chronology. Therefore, we reconstructed the mean temperature for October in Wuying by using the next year tree chronology. During the calibration years from 1960 to 2004, the ultimate transfer function is as follows.

$T_{t}=0.01RES_{t+1}-5.01$

( 1)

Table 3. Correlation coefficients between RES chronology and monthly mean temperature, monthly mean maximum temperature, monthly mean minimum temperature, and monthly total precipitation

Month	P-Sep	P-Oct	P-Nov	P-Dec	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct
Tmean	0.063	0.589	−0.025	−0.003	0.111	0.175	−0.044	0.175	0.090	−0.029	0.241	0.094	−0.100	−0.174
Tmax	0.055	0.485	−0.066	−0.046	0.058	0.153	−0.104	0.192	0.159	−0.009	0.240	0.135	−0.046	−0.135
Tmin	0.094	0.518	0.021	0.024	0.151	0.212	−0.013	0.142	−0.034	−0.005	0.172	−0.030	−0.136	−0.171
P	0.068	0.030	−0.069	0.044	0.054	0.123	0.006	0.042	−0.037	0.026	0.077	−0.100	−0.178	0.117

Notes: Coefficients at the statistical significant confidence level of 99% are in bold. P indicates the preceding year.

The correlation coefficient of the transfer function is 0.590. The explained variance is 34.8%. F -test value is 23.448, far more than the critical value at 0.0001 significant level. The mean temperature for October at Wuying from 1796 to 2004 was reconstructed using the transfer function. The correlation coefficients are more than 0.7 between Wuying station and some other stations including Yichun, Harbin, Mohe, Bei’an and Mudanjiang, which indicates that the reconstructed temperature at Wuying can represent big spatial scope.

Figure 2 a showed the comparison results between the reconstructed and observed series. The high frequency variation of the reconstructed series is in consistency with that of observed series. Stability, reliability, and accuracy of the reconstructed series were tested by using the “leave-one-out” [ Yin et al., 1996 ]. The test results are shown in Table 4 . The correlation coefficient (R) is 0.590 between the reconstructed and observed series, and the correlation coefficient of cross-validation (r) is 0.543. They both reach the significant level at 0.001, indicating that the transfer function is stable. Reduction Error (R_E ) is the statistics used to test climate reliability, here 0.292, showing that the reconstruction series is credible. S1 represents the years with the same sign of one-order difference between the reconstructed and observed series. S2 represents the years with the same sign of difference between the two series relative to the average of observed series. S 1 and S 2 are both more than the critical value at 99% confidence level, which shows that there is a consistency between the reconstructed and observed series. ts is the average product number, which is used to validate whether there is a significant difference between the amount of the same-sign years and the opposite-sign years of reconstructed and observed series. The value of ts is 2.698 which is more than the critical value at 99% confidence level, showing that the reconstructed series is reliable.

Figure 2. (a) Observed (solid line) and reconstructed (dashed line) October mean temperature series (1960–2004), (b) the reconstructed October mean temperature series of Wuying site during 1796–2004 (thick line denotes the 11-year running-average), and (c) the number of cores

Table 4. Statistical parameters of the reconstructed October mean temperature of Wuying

R	r	RE	ts	S1	S2
0.590	0.543	0.292	2.698	34(30, 32)	36(31, 33)

Note: Values in parentheses indicate the critical values at 95% and 99% confidence level.

5. Characteristic analysis and discussion of the reconstructed temperature series

Figure 2 b showed the reconstructed October mean temperature of Wuying. Temperature of 209 years can be divided into four colder and four warmer periods. The four colder periods are 1801–1847, 1861–1872, 1900–1937, and 1951–1978, and the four warmer periods are 1848–1860, 1873–1899, 1938–1950, and 1979–1999. The duration of the colder period before 1847 is the longest, and its occurrence is synchronous with the little ice age generally acknowledged [ Portter , 1986 ].

The main researches of tree ring in northeastern China were conducted on the reconstruction of minimum temperature from February to September [ Zhu , 2006 ], and maximum temperature from January to April [ Shao and Wu, 1997 ] in Changbai Mountain. The correlation coefficients of monthly mean temperature with monthly mean minimum and monthly mean maximum temperature are more than 0.7 during 1959–2004 at Wuying and Changbai stations, which showed that trend in monthly mean temperature is consistent with those in monthly mean minimum and maximum temperature. As can be seen from Table 5 , October mean temperature is comparable to Changbai temperature from February to September and from January to April in low-frequency variation. Therefore, we can use the two reconstructed temperature series of Changbai to compare it with the reconstructed series of Wuying. The four colder periods and the four warmer periods of Wuying are almost one-to-one in correspondence with the colder and warmer periods of Changbai from February to September. Changbai Mountain is in a low-temperature stage before 1830 and in high-temperature stages from late 1940s to early 1960s and from late 1970s to late 1990s, which are consistent with our results. Mutual verification of the reconstructed series also further illustrated that the reconstruction is credible.

Table 5. Correlation coefficients between temperatures at Wuying and Changbai stations

Series	Series	Correlation coefficients	Coefficients of 11-year running average
October mean temperature of Wuying	Mean temperature from Jan to Apr of Wuying	0.240	0.639
Mean temperature from Jan to Apr of Wuying	Mean temperature from Jan to Apr of Changbai	0.770	0.930
October mean temperature of Wuying	Mean temperature from Jan to Apr of Changbai	0.239	0.706
October mean temperature of Wuying	Mean temperature from Feb to Sep of Wuying	0.330	0.602
Mean temperature from Feb to Sep of Wuying	Mean temperature from Feb to Sep of Changbai	0.630	0.540
October mean temperature of Wuying	Mean temperature from Feb to Sep of Changbai	0.278	0.706

Periodicity of the reconstructed series was analyzed by the power spectrum method [ Huang, 2004 ]. The longest lag is 70 years which is the equivalent of 1/3 of the total length (209 years). The results showed there are periodic characteristics of 5.6-year, 4.67-year, 3.33-year, and 2.46-year, among which the period of 3.33-year is the most significant. It is noteworthy that a period of around 5-year is consistent with double oscillation periods of solar activity. The correlation coefficient is 0.15 between temperature series and sunspot number of 9-year moving-average from 1796 to 2000 (P<0.02), which showed that solar activity may have influence on the temperature of this area.

The abrupt changes in the reconstructed series were also detected by the methods of smoothing t-test, smoothing F-test [ Fu and Wang, 1992 ], and Le Page test [ Yanetani, 1992 ]. In general, 30-year mean is used as basic climatic condition of one area, consequently in this paper the abrupt changes of 30-year time scale were carried out in the reconstructed series. Test results were showed in Figure 3 . The significant abrupt changes in the mean series are observed for around 1871 and 1900 according to smoothing t -test and Le Page test. The abrupt trend in 1871 is from cold to warm, while it is from warm to cold in 1900. Abrupt change in mean value also probably occurs in 1834. The Abrupt changes in 1834 and 1990 are consistent with the jump of winter mean minimum temperature at Daxigou meteorological station in the river source region of Urumchi [ Yuan and Li, 1999 ]. Significant abrupt changes in standard deviation are found for around 1851, from little to great according to smoothing F -test and Le Page test. Abrupt changes in standard deviation probably occur in 1891, 1924 and 1956. Abrupt changes for around 1851 and 1891 were also detected in Fuping at Qinling [ Liu and Shao, 2003 ].

Figure 3. Analysis of abrupt changes of the reconstructed October mean temperature series (a) t -test, (b) F -test, and (c) Le Page test

6. Conclusions

(1) Local temperature in the preceding October is the main factor for the growth of Pinus koraiensis in the Fenglin National Natural Reserve of Wuying, in Heilongjiang province. Temperatures in late April also affect tree ting radial width to some extent. Response of tree growth to temperature has some physiological significance.

(2) October mean temperature of Wuying was reconstructed for the period 1796–2004 using residual tree ring chronology. In the past 209 years, there are 4 colder periods and 4 warmer periods according to the reconstructed series. The most significant period of 3.33-year is detected. Abrupt changes were also detected at 30-year time scale. Significant abrupt changes of mean value are observed for around 1871 and 1900, and significant abrupt changes of standard deviation are found for around 1851.

Acknowledgements

The project was supported by the Special Research Program for Public-welfare Forestry (No. 200804001), National Science and Technology Support Program (No. 2007BAC29B01) and the Natural Science Foundation of China (No. 40705032).

References

1. Briffa et al., 1995 K.R. Briffa, P.D. Jones, F.H. Schweingruber, et al.; Unusual twentieth-century summer warmth in a 1,000-year temperature record from Siberia; Nature, 376 (1995), pp. 156–159
2. Cook and Kairiukstis, 1990 E.R. Cook, L.A. Kairiukstis; Methods of Dendrochronology: Applications in Environmental Sciences, Kluwer Academic Publishers (1990), p. 394
3. Cook et al., 1991 E.R. Cook, T. Bird, M. Peterson, et al.; Climate change in Tasmania inferred from a 1,089-year tree-ring chronology of Huon Pine; Science, 253 (1991), pp. 1266–1268
4. Davi et al., 2003 N.K. Davi, G.C. Jacoby, G.C. Wiles; Boreal temperature variability inferred from maximum latewood density and tree-ring width data, Wrangell Mountain region, Alaska; Quaternary Research, 60 (3) (2003), pp. 252–262
5. ECFC (Editorial Committee of Forest in China), 1999 ECFC (Editorial Committee of Forest in China); Forest in China (in Chinese), China Forestry Press (1999), p. 1161
6. Esper et al., 2002 J. Esper, E.R. Cook, F.H. Schweingruber; Low-frequency signals in long tree-ring chronologies for reconstructing past temperature variability; Science, 295 (5563) (2002), pp. 2250–2253
7. Fritts, 1976 H.C. Fritts; Tree Rings and Climate, Academic Press Incorporation (1976), p. 567
8. Fritts, 1991 H.C. Fritts; Reconstruction Large-scale Climate Patterns from Tree-Ring Data, University of Arizona Press (1991), p. 286
9. Fu and Wang, 1992 C. Fu, Q. Wang; The definition and detection of the abrupt climatic change; Chinese Journal of Atmospheric Sciences (in Chinese), 16 (4) (1992), pp. 482–493
10. Holmes, 1983 R.L. Holmes; Computer-assisted quality control in tree-ring dating and measurement; Tree-Ring Bulletin, 43 (1983), pp. 69–78
11. Huang, 2004 J. Huang; Statistic Analysis and Forecast Methods in Meteorology (in Chinese), China Meteorological Press (2004), p. 298
12. Jacoby et al., 2000 G.C. Jacoby, N.V. Lovelius, O.I. Shumilov, et al.; Long-term temperature trends and tree growth in the Taymir region of northern Siberia; Quaternary Research, 53 (3) (2000), pp. 312–318
13. Kohler, 1949 M.A. Kohler; On the use of double-mass analysis for testing the consistency of meteorological records and for making required adjustment; Bull. Amer. Meteor. Soc., 30 (1949), pp. 188–189
14. Liu and Shao, 2003 H. Liu, X. Shao; Reconstruction of January to April mean temperature at Qinling MTS from 1789 to 1992 using tree ring chronologies; Journal of Applied Meteorological Science (in Chinese), 14 (2) (2003), pp. 188–196
15. Liu et al., 2003 Y. Liu, Q. Cai, P. Won-Kyu, et al.; Tree-ring precipitation records from Baiyinaobao, Inner Mongolia since A.D. 1838; Chinese Science Bulletin, 48 (11) (2003), pp. 1140–1145
16. Mann, 1945 H.B. Mann; Non-parametric test against trend; Econometrika, 13 (1945), pp. 245–259
17. Portter, 1986 S.C. Portter; Pattern and forcing of Northern Hemisphere glacier variations during the last millennium; Quaternary Research, 26 (1986), pp. 27–48
18. Shao and Wu, 1997 X. Shao, X. Wu; Reconstruction of climate change on Changbai Mountain Northeast China using tree-ring data; Quaternary Sciences (in Chinese), 1 (1997), pp. 76–85
19. Shao et al., 2005 X. Shao, L. Huang, H. Liu, et al.; Reconstruction of precipitation variation from tree rings in recent 1,000 years in Delingha, Qinghai; Science in China Series D: Earth Sciences, 48 (7) (2005), pp. 939–949
20. Stocks and Smiley, 1968 M.A. Stocks, T.L. Smiley; An Introduction to Tree Ring Dating, University of Chicago Press (1968), p. 73
21. Wang, 1999 Z. Wang; Plant Physiology (in Chinese), China Agriculture Press (1999), p. 492
22. Wu, 1990 X. Wu; Tree Ring and Climate Change (in Chinese), China Meteorological Press (1990), p. 369
23. Yanetani, 1992 T. Yanetani; Discontinuous changes of precipitation in Japan after 1900 detected by the Le Page test; J. Meteor. Soc. Japan, 70 (1) (1992), pp. 95–103
24. Yin et al., 1996 X. Yin, H. Liu, X. Shao; Flow analysis climate reconstruction program using tree ring chronologies; Collection of Researches on Climate Change and Numerical Modeling (in Chinese), Progromme 85-913-2 committee, 1, , China Meteorological Press (1996), pp. 116–126
25. Yuan and Li, 1999 Y. Yuan, J. Li; Reconstruction and analysis of 450 years’ winter temperature series in the Urumqi river source of Tianshan Mountains; Journal of Glaciology and Geocryology (in Chinese), 21 (1) (1999), pp. 64–70
26. Zhu, 2006 H. Zhu; A Temperature Reconstruction for Eastern Part of Northeast China Based on a Tree Ring Width Network (in Chinese)  ; School of Geographic Sciences and Remote Sensing, Beijing Normal University (2006), p. 90