Tree ring cellulose δ18O of Fokienia hodginsii in northern Laos: A promising proxy to reconstruct ENSO?

Authors


Abstract

[1] A tree ring cellulose δ18O chronology of Fokienia hodginsii from 1951 to 2002 is established by four cores from four different trees to explore its potential for reconstructing climatic variations in northern Laos. This chronology is the first tree ring δ18O chronology from Laos as well as the first for this species. We compare variations of tree ring cellulose δ18O and ring width between cores. The results suggest that the δ18O time series are more strongly correlated between different trees than ring width time series are. In addition, correlation analysis shows a significantly negative correlation between tree ring cellulose δ18O and the May–October Palmer Drought Severity Index (PDSI), while the tree ring width indices have a poor correlation with PDSI. The spatial correlation analysis between tree ring cellulose δ18O and global sea surface temperature and the time series correlation analysis between tree ring cellulose δ18O and the Multivariate El Niño-Southern Oscillation (ENSO) Index (MEI) during the investigated period (1951–2002) reveal that ENSO has an important effect on tree ring celluloseδ18O in Laos. Our results indicate that the tree ring cellulose δ18O of Fokienia hodginsii is a promising proxy to reconstruct PDSI in Southeast Asian countries, such as Laos, and to discuss their teleconnections with global atmospheric circulations such as ENSO.

1. Introduction

[2] Southeast Asia is governed by a monsoon climate with distinct rainy and dry seasons. Monsoon variations, especially monsoon-associated precipitation changes, have an important effect on the livelihood of the people [Lau and Yang, 1997] and civilization shifts during the past several hundred years [Buckley et al., 2010]. However, the instrumental climate record in this area is too short to understand the monsoonal variability in this area. Proxy-based paleoclimate records are therefore necessary for a better understanding of monsoon variations.

[3] Tree rings are one of the most important archives in paleoclimate research because of their high temporal resolution and accuracy of dating. Although tropical trees may not produce verifiably annual rings, many previous researchers [Buckley et al., 2007a, 2010; D'Arrigo et al., 1997; Pumijumnong and Eckstein, 2011; Sano et al., 2009] have found increasing success in using tropical trees (including evergreen and deciduous trees) with annual rings to reconstruct paleoclimates that are due to the strong annual cycles in rainfall in Southeast Asia. Tree ring width indices (TRWi) are often used in this region, but tree ring cellulose δ18O exhibits certain advantages over using tree ring widths in some cases. First, the physiological controls of tree ring cellulose δ18O are well understood [Flanagan et al., 1991; Roden et al., 2000; Sternberg et al., 1986; Yakir and DeNiro, 1990]. Second, tree ring cellulose δ18O is not as strongly affected by the physiological characteristics of individual trees as ring width is [Raffalli-Delerce et al., 2004], and it has a limited effect of age and a lower autocorrelation [McCarroll and Loader, 2004; Shi et al., 2011]. Third, tree ring cellulose δ18O can be employed as a chronometer and climatic indicator for trees lacking visible rings [Anchukaitis et al., 2008; Anchukaitis and Evans, 2010; Evans and Schrag, 2004; Poussart and Schrag, 2005; Poussart et al., 2004] and can supply an intra-annual resolution signal [Managave et al., 2010; Roden et al., 2009]. Fourth, even for trees not growing at their “ecological” limit, it is possible to extract a strong climate signal using stable isotope methods, in which ring widths (and density) do not routinely yield useful climatic information [Loader et al., 2008].

[4] Plant physiological models and tree ring isotopic studies [Barbour et al., 2004; Li et al., 2011b; Robertson et al., 2001; Roden et al., 2000] suggest that the relative humidity and oxygen isotope composition (δ18O) of the source water are the main factors for controlling variations of tree ring cellulose δ18O. These two factors are influenced by many local climatic parameters (e.g., temperature, precipitation, and duration of monsoon season). Thus, an approach that uses tree ring cellulose δ18O to reconstruct climatic information is both species and site sensitive. For example, the tree ring cellulose δ18O of teak trees from western and central India shows a significant positive correlation with the amount of rainfall, while the tree ring cellulose δ18O of teak from southern India shows a significant negative correlation with the amount of rainfall [Managave et al., 2011]. Poussart and Schrag [2005] analyzed oxygen isotopes of 11 trees in Thailand. The three samples collected in Lampang did not produce coherent δ18O records, and two other samples lacked seasonal isotopic signals. Therefore, we need to further study site and species sensitivities for tree ring isotopic ratios to understand the climatic implications of tree ring cellulose δ18O under local climate conditions.

[5] In this paper, we report measurements of the oxygen isotopic composition of tree ring cellulose in four cores from four trees of Fokienia hodginsii in northern Laos from 1951 to 2002. The isotopic data are then statistically analyzed to explore the potential for extracting climate signals. We develop a tree ring cellulose δ18O chronology and perform spatial and temporal correlation analyses with climate data. On the basis of the results, we discuss teleconnections between climate in northern Laos and global atmospheric circulations such as El Niño-Southern Oscillation (ENSO).

2. Materials and Methods

2.1. The δ18O Model

[6] Because there is no isotopic fractionation during the transportation of water from soil to leaf [White et al., 1985] and because the isotopic signature of xylem water is imprinted on the final cellulose product [Deniro and Epstein, 1979], many models of tree ring cellulose δ18O usually focus on two processes: (1) changes in the oxygen isotope ratio of water as it moves through the plant (effect of leaf water enrichment) and (2) changes in the isotope ratio of oxygen as it becomes incorporated into the cellulose molecule through biochemical effects [Sternberg, 2009].

[7] The isotopic composition of leaf water is modified from xylem water by transpiration, which leads to a consequent enrichment of the δ18O of the leaf water. This process, described by Craig and Gordon [1965], is

display math

where the subscripts l, x, and a refer to leaf water, xylem water, and water vapor in the air outside the leaf, respectively; ε*(εk) is the equilibrium (kinetic) isotopic fractionation factor; and ea/ei is the ratio of ambient vapor pressure to intercellular vapor pressure. Assuming that atmospheric water vapor is isotopically equilibrated with precipitation, δ18Oa can be equal to δ18Op − ε* and ea/ei can be equal to the relative humidity (h) [Dongmann et al., 1974]. The modified Craig-Gordon equation can be simplified to

display math

Equation (2) demonstrates that the oxygen isotope of leaf water was mainly controlled by relative humidity and the oxygen isotope of source water.

[8] Roden et al. [2000] summarized the process from the leaf water to the stem cellulose, as expressed by

display math

where δ18Oc represents the δ18O of tree ring cellulose, f0 is the exchanged proportion of carbohydrate oxygen with xylem water, εo is the net biological isotopic fractionation factor between the xylem water and the exchanged oxygen in the carbohydrate, and +27‰ is the equilibrated isotopic exchange ratio between the leaf water and the photosynthesized carbohydrate [Deniro and Epstein, 1979; Yakir and DeNiro, 1990]. There is a negative correlation [Dansgaard, 1964] between δ18Op and rainfall in tropical areas (amount effect). Therefore, we can see clearly how climate exerts an effect on tree ring cellulose oxygen isotope by equations (2) and (3).

[9] Because the sampling sites of this study were not located close to meteorological stations, we did not have sufficient environmental data to quantitatively interpret our data using such models. However, the models are conducive to understanding how the data we obtained are related to environmental variables.

2.2. Sampling Sites and Local Climate

[10] The tree ring samples used in this study were derived from 5 mm increment cores collected from old-growthFokienia hodginsii (>300 years). These trees grow in the Phu Leuy (PL) mountain area (20°17′N, 103°55′E, 1340 m) in northern Laos (Figure 1a). The age range of this tree species generally reaches several hundred years. Three cores from different directions per tree at breast height were collected in the summer of 2003. According to the 1961–1990 meteorological data (Figure 1b) from the Luang Prabang weather station (19.9°N, 101.2°E), which is located 140 km west of Phu Leuy at an elevation of 305 m, the annual mean temperature is 24.4°C and the mean annual precipitation is 1442.9 mm, 85% of which falls in the monsoon months from May to October.

Figure 1.

(a) Map of the study area, showing the locations of the Phu Leuy (PL) Fokienia site (black triangle) and four gridded PDSI spots (black circles) [Dai et al., 2004]. (b) Monthly mean temperature (black circles) and precipitation (gray bar) (1961–1990) at the Luang Prabang instrumental station.

2.3. Cross Dating

[11] The cores were air dried and polished with increasingly fine sandpaper until rings were clearly visible. The ring widths were measured using a sliding stage micrometer interfaced directly with a computer for measurement capture with a resolution of 0.01 mm. Cross dating was performed in the laboratory by matching ring width variations for all cores to determine the absolute year of each ring. Quality control was carried out using the COFECHA program software [Holmes, 1983]. We collected 112 cores from 38 trees in total. The 26 cores from the 15 trees could be used for cross dating. TRWi, which are commonly used for paleoclimate reconstruction [Buckley et al., 2007a; Cook et al., 2010], were calculated by ARSTAN [Cook, 1985].

2.4. Sample Preparation and α-Cellulose Extraction

[12] We selected the cores for isotope analysis based on the following two principles: (1) ring width of cores is wide enough to produce samples for isotope analysis and (2) cores have more than 300 rings, which are helpful for further paleoclimate reconstruction and reduce potential age-related effects [Treydte et al., 2006]. Four cores covering a period of 52 years (from 1951 to 2002), namely, PL104b, PL108a, PL126a, and PL136b, were selected to investigate the relationship between tree ring cellulose δ18O and the climate. Many of the rings were narrow, so we used whole annual rings, including earlywood and latewood, for the isotopic analyses.

[13] Most researchers separate annual rings before α-cellulose extraction. Each ring was treated independently, so the number of samples per batch was limited. The number of samples per batch of the Brendel [Brendel et al., 2000] and modified Brendel [Evans and Schrag, 2004] methods is 56 and 80, respectively. Given that chemically treating hundreds or thousands of samples is time consuming, researchers [Esper et al., 2010; Gagen et al., 2011b] have analyzed wood blocks of 5 or 10 years old together for a tree ring sample with a 1000 year length, in which interannual variations cannot be retained. Therefore, a method that can treat hundreds or thousands of rings simultaneously is necessary to save time. We employed “the plate method,” which is similar to the method of Li et al. [2011c]. In this method, the α-cellulose is extracted directly from the wood plate rather than from individual rings, and the grinding step is omitted. This method drastically reduces the time needed.

[14] The following is the detailed procedure. Before α-cellulose extraction, we cut cores into plates along surfaces perpendicular to the cellulose fiber directions with a thickness of 1 mm using a low-speed diamond wheel saw. We then selected one plate with the greatest width. In general, one core taken by a 5 mm increment borer can be cut into two or three plates (1 mm per plate), whileLi et al. [2011c] employed a thickness of 3.5–4.0 mm per plate, which requires more time for the chemical reaction and consumes more samples. Although it depends on the ring widths, one plate is usually enough to produce two samples for isotopic measurement per year. Thus, we can keep another plate to observe and check the wood anatomy after obtaining isotopic data. To maintain the shape of the wood plate during the α-cellulose extraction, the plate was packed by two Teflon punch sheets and was put into a glass tube for chemical reactions.

[15] The chemical protocol for extracting α-cellulose from the tree rings generally follows the Jayme and Wise method [Green, 1963; Loader et al., 1997]. The glass tube containing the wood plate was treated first with toluene and ethanol (1:1) in an ultrasonic water bath for 30 min, then with acetone in an ultrasonic water bath for 30 min in a fume hood to remove lipids, and then with an acidified NaClO2 solution in a water bath (70°C) for 60 min. The step of decomposing the lignin by an acidified NaClO2 solution had to be repeated a few times until the color of the plate turned white. A 17 wt % NaOH solution was poured into a glass tube containing the wood plate in a water bath (80°C) for 60 min to remove hemicellulose, and this step was repeated another two times. After this step, the wood plate was extremely fragile. The wood plate was gently washed with a diluted HCl solution and distilled water.

[16] The cellulose plate was then dried in an oven overnight. For a batch of samples, the above steps took 4 days. For the purpose of saving time, we extracted many cores simultaneously. The separation of the tree ring sample from the α-cellulose plate was performed under a binocular microscope. We wrapped 120–180μg of each α-cellulose sample in silver foil in duplicate for isotope measurements, although there were no duplicates of several samples because of the narrow ring width.

2.5. Stable Isotopic Analysis

[17] The oxygen isotope ratios (18O/16O) of the cellulose samples were determined by a continuous flow system with a pyrolysis-type elemental analyzer (TC/EA) and an isotope ratio mass spectrometer (Delta plus XL) at the Research Institute of Humanity and Nature, Kyoto, Japan. The celluloseδ18O values were calculated by comparison with the laboratory working standard International Atomic Energy Agency (IAEA) C3 cellulose, which was inserted frequently during the measuring process. The isotopic results for oxygen are presented in δ notation as per mil (‰) with respect to the international stable oxygen isotope standard, Vienna standard mean ocean water (VSMOW): δ18O = [(Rsample/Rstandard) − 1] × 1000‰, where Rsample and Rstandard are the 18O/16O ratios in the sample and standard, respectively. The value of the oxygen isotope ratios was obtained as an average of the duplicate analyses on an annual tree ring cellulose sample. The analytical uncertainties for repeated measurements of the IAEA cellulose and the tree ring samples were approximately 0.2‰ (n = 49) and 0.3‰ (= 168) (1σ), respectively.

2.6. Expressed Population Signal

[18] The expressed population signal (EPS) is used to evaluate whether the composite time series captures a representative site signal [McCarroll and Loader, 2004; Wigley et al., 1984]. EPS = (n*rmean)/[(n*rmean) + (1 − rmean)], where n is the number of trees at the site and rmean is the mean correlation coefficient among all the series. The EPS of tree ring width chronology and tree ring cellulose δ18O chronology is 0.9 and 0.803, respectively. Although an EPS > 0.85 is used to suggest that the composite record accurately represents the mean variance of the population and yields a signal relatively free of noise because of individual variations, an EPS < 0.85 does not necessarily indicate that the record is an inaccurate representation of the population signal [English et al., 2010]. For example, English et al. [2010] indicate that the δ18O of cactus in South America effectively records the ENSO-related precipitation, although the EPS of five cactuses ofδ18O is 0.66. Our results show that our tree ring cellulose δ18O chronology has a close relationship with climate (discussed in a later section).

2.7. Climate Analysis

[19] The Fokienia at PL were from 1300 m in the forested mountains, while the Luang Prabang weather station was below 400 m. Komonjinda [2003]analyzed elevational temperature-rainfall-relative humidity profiles in northern Thailand and illustrated how poorly low-elevation stations reflect mountain climates. Therefore, the CRU TS3.0 (http://badc.nerc.ac.uk/data/cru/) gridded temperature and precipitation data sets with 0.5° × 0.5° resolution are employed here. We produced an average from the four CRU grid points (19.75°N, 103.25°E; 19.75°N, 103.75°E; 20.25°N, 103.25°E; and 20.25°N, 103.75°E) containing the study site.

[20] The Palmer Drought Severity Index (PDSI) [Palmer, 1965] incorporates monthly precipitation, temperature, moisture supply, and moisture demand into a hydrological accounting system [Dai et al., 2004], whereby positive and negative values of the PDSI correspond to wet and dry conditions, respectively. To more clearly investigate the influence of hydroclimate on cellulose δ18O, we compared the chronology of the PL cellulose δ18O with the average value of four gridded global PDSI (Figure 1a) series produced by Dai et al. [2004].

[21] We also examined the spatial correlation between the PL cellulose δ18O chronology and the sea surface temperature (SST) from the NCDC v2 data set [Smith and Reynolds, 2004] to identify large-scale forcing effects on tree ring celluloseδ18O and the local climate. For this purpose, we utilized the Royal Netherlands Meteorological Institute (KNMI) Climate Explorer (http://www.knmi.nl/) [van Oldenborgh and Burgers, 2005]. The Multivariate ENSO Index (MEI) based on the six main observed variables (sea level pressure, zonal and meridional components of the surface wind, sea surface temperature, surface air temperature, and total cloudiness fraction of the sky) over the tropical Pacific was used to monitor ENSO [Wolter and Timlin, 2011]. High MEI values indicate El Niño events. We compared the PL cellulose δ18O chronology with the MEI from 1951 to 2002 to explore the influences of ENSO on local climate [Wolter and Timlin, 2011]. The δ18O data were annual, while the MEI was bimonthly. Thus, we employed the average value of the MEI from the previous November to the current October as the annual value of the current year.

3. Results

3.1. The Statistics of the Ring Width and δ18O Chronology

[22] The ring width and δ18O data are shown in Figure 2. The ring width time series show marker years in 1965, 1973, and 1986 (Figure 2a). Table 1 presents the correlation coefficients (r) among ring width time series that were chosen for isotope analysis. The mean correlation coefficient of these four cores is 0.227 (effective degrees of freedom (edf) = 39, four series). The tree ring width chronology is illustrated in Figure 2a by a bold line. The mean autocorrelation (1 year lag) of ring width chronology is 0.389.

Figure 2.

Time series of (a) 26 cores of ring width and (b) 4 cores of tree ring cellulose δ18O for Fokienia hodginsii growing in northern Laos. (Black bold line, tree ring width indices of STD chronology (Figure 2a) and the average δ18O of four cores (Figure 2b); black dashed line, trend of tree ring cellulose δ18O in last 50 years; gray shading, ±1 standard deviation of ring width chronology and tree ring cellulose δ18O.)

Table 1. Correlation Coefficients Among Time Series of Ring Widths and δ18Oa
rPL104bPL108aPL126b
  • a

    Effective degrees of freedom are 39 (Ring Widths) and 50 (δ18O).

  • b

    p < 0.05.

  • c

    p < 0.01.

Ring Widths
PL108a0.218  
PL126a0.2490.193 
PL136b0.1770.1940.331b
δ18O
PL108a0.491c  
PL126a0.564c0.449c 
PL136b0.546c0.385b0.598c

[23] The inter-tree variability forδ18O exhibited a range of 0.3‰–2‰, and the typical inter-tree variability forδ18O summarized by Leavitt [2010]fell within the range of 1‰–4‰. The mean inter-tree correlation coefficients ofδ18O is 0.506 (edf = 50, four series),

[24] A visual inspection shows several marker years in δ18O time series in Figure 2b (e.g., 1956–1957, 1964, 1971, 1980–1981, 1983, 1992, and 1998). An annually resolved mean tree ring cellulose δ18O record was developed in northern Laos over the period 1951–2002 (Figure 2b, bold line). The δ18O range is from 22.3‰ to 25.9‰, and the average is 24.14‰. The standard deviation is 0.78‰. An increasing trend (Figure 2b, dashed line) is observed in the mean tree ring cellulose δ18O series calculated by a linear regression as 0.15‰/decade. The mean autocorrelation (1 year lag) of δ18O chronology is 0.129.

Figure 3.

Correlation coefficients (r) among (a) temperature, precipitation, PDSI, and TRWi and (b) tree ring cellulose δ18O of Fokienia hodginsii. Horizontal dotted line indicates 99% confidence level, and the lowercase letters indicate (left to right) the previous October, November, and December.

3.2. Correlation Between Ring Width and δ18O Chronology and Climate

[25] Correlation analyses were carried out between TRWi-tree ring celluloseδ18O and climatic parameters (Figure 3). All the data are from the common period of 1951–2002. For TRWi, there is a positive relationship with precipitation from the previous October to the current March and a negative relationship with monsoon-associated precipitation (May–September). The correlation coefficient between TRWi and precipitation in the previous December is 0.35. The TRWi was negatively correlated with temperature from February to May and positively correlated with temperature in August and September (Figure 3a). The correlation coefficient between TRWi and temperature in February is −0.28. Buckley et al. [2007a] obtained similar correlations using Pinus merkusii in the neighborhood of Vientiane, the capital of Laos. TRWi shows a weak relationship with PDSI.

[26] Tree ring cellulose δ18O (Figure 3b) shows significant positive correlations with temperature from May to August (r = 0.64, p < 0.001, edf = 50) and negative correlations with precipitation in May (r = −0.34, p < 0.05, edf = 50). There is a significant negative relationship between tree ring cellulose δ18O and PDSI from May to October (r = −0.66, p < 0.001, edf = 49). Tree ring cellulose δ18O accounts for 41% and 44% of the actual temperature from May to August and monsoon season PDSI variance, respectively.

[27] Spatial correlations between δ18O chronology and monsoon season maximum temperature, cloud cover (Figure 4), and global SST (Figure 6) in the past 50 years were determined to assess the regional representative of δ18O chronology and explore its links with remote oceans. The δ18O chronology has a significant positive correlation with monsoon season maximum temperature in northern Laos, Thailand, and Vietnam, and has a significant negative correlation with monsoon season cloud cover in northern Laos and eastern Myanmar. There is a significant positive correlation between the δ18O chronology and the prior winter (December, January, February (DJF)) and premonsoon (March, April, May (MAM)) Niño 3.4 SST. The correlation coefficient between the δ18O chronology and premonsoon (MAM) Niño 3.4 SST is higher. Besides, there is a consistently high correlation (r = 0.773, p < 0.001, edf = 44) between tree ring cellulose δ18O and the MEI from 1951 to 2002 (Figure 5). The PL δ18O chronology can explain 60% of the variance of MEI.

Figure 4.

Spatial correlation between the tree ring cellulose δ18O and (a) May–October temperatures and (b) cloud cover during 1951–2002. Black triangle shows the study area. Correlations of >∣0.3∣ are significant at the p < 0.05 level, assuming 44 effective degrees of freedom. Correlations not significant at the 95% level have been masked out.

Figure 5.

Time series of tree ring cellulose δ18O chronology and MEI from 1951 to 2002. Blue line, multivariate ENSO index; red line, tree ring cellulose δ18O (edf = 44).

4. Discussion

4.1. Tree Ring Width Chronology

[28] The relationship between TRWi and climate (Figure 3a) suggests that wet conditions in the premonsoon period (March–May) are conducive for tree growth. Considering the negative relationship between ring width and precipitation in the monsoon season and the low mean daily sunshine duration in June, July, and August, dense cloud cover in summer may limit photosynthesis activity, resulting in narrower rings [Buckley et al., 2007b; Fritts, 1976]. Based on correlation analysis, narrow rings form when spring droughts and summer floods happen in 1 year. In addition, the TRWi are weakly correlated with PDSI. However, the TRWi of teak in northwestern Thailand [Buckley et al., 2007a] and Fokienia in northern Vietnam [Sano et al., 2009] showed a positive relationship with premonsoon PDSI. The effects of climate fluctuations on tree growth are site specific [Spiecker et al., 1996]. The sampled trees originate from a closed-canopy forest site, where frequent endogenous disturbance is common, resulting in reduced climate sensitivity. On the whole, tree ring width chronology in sampling site has weak statistical association with climatic parameters, but tree ring width chronology provides excellent age control, which is important for building up preciseδ18O time series and information on tree growth.

4.2. The δ18O Chronology and Its Climate Signal

[29] Compared with those of tree ring width, correlation coefficients of δ18O time series are higher (Table 1). The δ18O chronology has a higher signal-to-noise ratio than ring widths (rδ18O > rring width) based on the same sample number. The isotopic records of teak in the India subcontinent exhibit higher mutual coherence than the ring width records [Managave et al., 2011]. Stable isotopes usually display higher signal strengths in comparison with equivalent ring width and density series [Gagen et al., 2011a].

[30] The correlation analyses indicate that the temperature and PDSI of the monsoon season have an important effect on tree ring cellulose δ18O. In the monsoon season, high temperature usually indicates less rainfall and low relative humidity, which can make the tree ring cellulose δ18O heavier. It is easy to understand the significantly negative correlation between the tree ring cellulose δ18O and PDSI from equations (2) and (3). Low PDSI indicates a dry condition. Low soil moisture could result in the evaporation of soil water, so the source water that tree roots take up becomes heavier. On the other hand, low relative humidity also leads to enhanced evapotranspiration, which causes the oxygen isotopes of leaf water to be enriched [Roden et al., 2000]. The fact that tree ring cellulose δ18O has highly negative correlations with the monsoon season PDSI suggests that the trees in the study area grow mainly in the monsoon season. A previous dendrometer study on teak in northern Thailand [Buckley et al., 2001] showed that the circumference increased following rain events, particularly during the wet season. Poussart and Schrag [2005] estimated seasonal and annual growth rates by oxygen and carbon isotopes and found that approximately 60% of the annual growth of several evergreen species occurred during the rainy season (May to October). In addition, tree ring cellulose δ18O has a weak relationship with the climate of the previous winter, which indicates that the climate in the previous winter exerts little influence on the oxygen isotopes of earlywood. Hill et al. [1995] revealed similar results for oak. Although tree ring cellulose δ18O reflects monsoon season rainfall, rainfall is not the main factor controlling the variations of tree ring cellulose δ18O. Poussart and Schrag [2005] measured the tree ring cellulose δ18O of deciduous trees in Thailand and suggested that the local precipitation only explained approximately 10%–20% of the variance of tree ring cellulose δ18O.

[31] The tree ring cellulose δ18O model (equations (2) and (3)) can be used to explain the correlation between tree ring cellulose δ18O and monsoon season maximum temperature and cloud cover. Maximum temperature-cloud cover usually has a negative–positive correlation with relative humidity, which is a significant factor for controlling tree ring celluloseδ18O. The spatial analysis (Figure 4) indicates that the tree ring cellulose δ18O chronology in northern Laos can represent the regional hydroclimate in central Indo-China.

[32] The δ18O chronology has an increasing trend during the last 50 years, which indicates that the monsoon season in this area has become drier or the δ18O of source water has changed (equations (2) and (3)). The tree ring cellulose δ18O in the Asian monsoon area (southeast Tibetan Plateau [Shi et al., 2011], Nepal Himalaya [Sano et al., 2010], and North China [Li et al., 2011b]) reveals a similar increasing trend during the last 50 years. Based on these results, the increasing trend reflects regional signal. Singhrattna et al. [2005] indicate that the temperature and precipitation exhibit an increasing and decreasing trend, respectively, in the monsoon seasons of the last 50 years in central Thailand. In addition, the meteorological data from China also indicate that in the past 50 years, the monsoon region in China became more arid [Fu and Ma, 2008]. According to these climatic data, the increasing trend of our δ18O chronology mainly reflects variations of dryness-wetness status in the monsoon season in northern Laos.

4.3. Links With Remote Oceans

[33] ENSO, which is often linked with extreme weather events such as flooding and drought, has an important effect on climate and society in Asian monsoon areas [Diaz and Markgraf, 2000]. Many previous reports [Buckley et al., 2005, 2007a; Sano et al., 2009] have investigated the relationship between the climate in Southeast Asia and ENSO using TRWi. Moreover, TRWi has been used in the southwestern United States to reconstruct the history of ENSO [Li et al., 2011a; Mann et al., 2000]. However, tree ring cellulose δ18O usually displays higher signal strengths than TRWi. Based on our results, the δ18O chronology shows a close link with premonsoon (MAM) Niño 3.4 SST and has similar variations with MEI during the last 50 years. In addition, the δ18O chronology shows peak values in 1958, 1965, 1969, 1972, 1983, 1987, 1992, 1998, and 2002, when El Niño events happened in the previous winter and/or the current summer. Lower values of tree ring cellulose δ18O emerge in 1956, 1964, 1971, 1989, 1999, and 2000, when La Niña events occurred in the prior winter and/or the current summer. Our results appear to show links between tree ring cellulose δ18O in Laos and ENSO, and they indicate that the tree ring cellulose δ18O in Laos may be a promising proxy to record ENSO. Previous studies in Indonesia [Poussart et al., 2004] and Peru [Evans and Schrag, 2004] showed that intra-annual tree ring celluloseδ18O time series had the potential to record El Niño. Compared with previous studies [Evans and Schrag, 2004, Poussart et al., 2004], our δ18O chronology can reflect El Niño using fewer samples, which is an advantage of a combination of cross dating by ring width and the isotope data. Ring width chronology provides precise age control. On the other hand, tree ring cellulose δ18O captures higher climatic signals. Therefore, combining cross dating results by ring width and stable isotope data is a powerful tool for doing tropical dendroclimatology.

[34] We can infer that the following factors cause the link between tree ring cellulose δ18O in northern Laos and ENSO. First, El Niño results in a reduction of summer monsoon rainfall in Southeast Asia [Chen and Yoon, 2000; Singhrattna et al., 2005]. The decrease in rainfall leads to a dry condition in summer. The low relative humidity makes the δ18O of leaf and soil water more enriched. On the other hand, the oxygen isotope of precipitation becomes heavier because of the decrease of precipitation amount (amount effect) [Dansgaard, 1964]. Poussart and Schrag [2005] discussed the relationship between the minimum value of tree ring cellulose δ18O in a year and wet season rainfall and suggested that the extent of the rainfall amount effect (monsoon intensity) was at least partially recorded in the minima of tree ring cellulose δ18O. Second, El Niño events result in temperature rise in Laos, which will make lower relative humidity and lead to more enriched tree ring cellulose δ18O (equations (2) and (3)). This relationship is consistent with the observation that temperatures (all year except June, July, and August) in Laos will increase during El Niño events (http://www.knmi.nl/research/global_climate/enso/effects/#temperature).

[35] ENSO-monsoon relationships are characterized by a nonstationary process. For example, an inverse relationship between ENSO and the Indian summer monsoon has broken down since the 1980s [Kumar et al., 1999]. Recent research by Singhrattna et al. [2005] reveals an increase in the influence of ENSO over central Thailand since the 1980s. The correlation coefficient between tree ring cellulose δ18O in northern Laos and the MEI increased from 0.617 (p < 0.001, edf = 25, 1951–1979) to 0.857 (p < 0.001, edf = 21, 1980–2002). This result supports the idea of Singhrattna et al. [2005]that El Niño patterns centered in the eastern Pacific have tended to place the descending arm of the Walker circulation in the Thailand-Indonesian region since the 1980s, resulting in reduced rainfall over the region during warm-phase events [Singhrattna et al., 2005]. In addition, the positive relationship between the tree ring cellulose δ18O in northern Laos and the Indian Ocean SST (previous winter and premonsoon) is illustrated by Figure 6, which points to an effect on climate in northern Laos by the Indian monsoon because part of the rainfall in the monsoon season in Southeast Asia is transported by the Indian summer monsoon [Singhrattna et al., 2009].

Figure 6.

Spatial correlation plots for tree ring cellulose δ18O chronology with (a) prior winter and (b) premonsoon NCDC v2 sea surface temperatures (1951–2002). Black triangle shows the study area; black rectangle shows Niño region 3.4. Correlations of >∣0.3∣ are significant at the p < 0.05 level, assuming 44 effective degrees of freedom. Correlations not significant at the 95% level have been masked out.

5. Conclusions

[36] The developed tree ring cellulose δ18O chronology of Fokienia hodginsii in northern Laos from 1951 to 2002 was analyzed in this paper. A new method was employed for extracting tree ring α-cellulose directly from a wholewood plate for oxygen isotope analyses that enabled us to save time. The results obtained demonstrate that it is possible to extract a strong monsoon season climate signal using tree ring celluloseδ18O from Fokienia hodginsii in northern Laos. Our results indicate that tree ring cellulose δ18O has a significant negative correlation with the May–October regional PDSI and a significant positive correlation with the MEI. This study indicates that the tree ring cellulose δ18O of Fokienia hodginsii is a promising proxy to reconstruct PDSI in Southeast Asian countries such as Laos and global atmospheric circulations such as ENSO. A comparison between the tree ring cellulose δ18O chronology and ENSO shows that the link between ENSO and drought in northern Laos has been enhanced since the 1980s. Further research using a long-term tree ring celluloseδ18O chronology combined with a tree ring network will shed more light on the variability of climate in Southeast Asia and how it relates to global atmospheric circulations on longer time scales.

Acknowledgments

[37] This study was funded by an environmental research grant from the Sumitomo Foundation, Japan, an FS research grant from Research Institute of Humanity and Nature, Kyoto, Japan, and grant-in-aid for Japan Society for the Promotion of Sciences Fellows (23242047 and 23–10262). We deeply appreciate the helpful comments from three anonymous reviewers to improve the manuscript.

Ancillary