Introduction
The variation in strength, timing, and duration of the Asian summer monsoon
(ASM) system affects the life and economy of many millions of people living in
south and east Asia (;
). In remote areas, such as the Tibetan Plateau (TP),
reliable climate records are short and scattered. Nevertheless, a recent
weakening trend of the ASM precipitation amount was reported in several
studies (;
; ). The decline in air humidity was
explained by a reduction in the thermal gradient between the surface
temperatures of the Indian Ocean and the TP due to global warming
. Different locations and
climate archives reveal contemporaneous strengthened monsoonal precipitation
(; ;
). This discrepancy may be explained by the high
variability of the monsoon circulation itself and also by the limited
number of available palaeoclimate studies and resulting climate modelling
uncertainties. Thus, for a better understanding of the circulation system as
a whole, but also for the verification of climate change scenarios, a keen
demand for reliable climate reconstructions exists for the TP. With
increasing numbers of palaeoclimatic records, forecast and climate projection
precision increases and can be helpful for facilitating targeted decision-making
regarding water and resource management.
Location of the study site Lhamcoka (green pentagon) and other proxy
archives mentioned in the text. Green triangles: tree-ring
δ18O chronologies; yellow triangle: tree-ring width
chronology; green asterisk: ice cores; green circle: lake sediments. Red
rectangles indicate climate stations.
The northward movement of the Intertropical Convergence Zone (ITCZ) in the
Northern Hemisphere in boreal summer is amplified over the Asian continent by
the thermal contrast between the Indian Ocean and the TP
. Convective rainfalls
during the summer monsoon season between June and September are strongly
altered by the complex topography of the Himalayas and western Chinese
mountain systems (e.g. ;
; ). Extreme
climatic events that may have devastating effects and the long-term trends of
ASM intensity are therefore in the focus of numerous climate reconstruction
efforts (e.g. ; ; ). Most of these studies use tree-ring width
as a proxy for palaeoclimate reconstructions. Nonetheless, several studies
demonstrated that δ18O of wood cellulose is a strong indicator of
hydroclimatic conditions (; ;
; ). Even if tree stands might have
been influenced by external disturbances (e.g. competition, insect attacks,
or geomorphological processes) they still reflect variations of the local
hydroclimate accurately . Recently published tree-ring
δ18O chronologies from the TP show a common strong response to
regional moisture changes. successfully
reconstructed August precipitation over the past 800 years. They demonstrated
reduced precipitation during the Medieval Warm Period (MWP), stronger
rainfalls during the Little Ice Age (LIA), decreasing precipitation rates
since the 1810s, and slightly wetter conditions since the 1990s. In addition,
shorter δ18O chronologies from the central Himalayas showed
consistent negative correlations to summer precipitation (,
2011, 2013). The detected recent reduction of monsoonal precipitation has
been interpreted as a reaction to increased sea surface temperatures (SSTs)
over the tropical Pacific and Indian Ocean
. Strong responses to regional
cloud cover changes were found for tree-ring δ18O chronologies from
the south-eastern TP (; ;
). The local moisture reduction starting in the middle of
the 19th century is less pronounced than for south-west Tibet, and is
associated with complex El Niño–Southern Oscillation (ENSO)
teleconnections . Existing tree-ring δ18O
chronologies on the north-eastern part of the TP respond to local
precipitation and relative humidity (;
). Except for a relatively
short summer moisture-sensitive time series
, no long-term δ18O
chronologies or reliable reconstructions have been conducted for the eastern
TP so far. It still remains unclear to what extent the MWP, LIA, and the
modern humidity decrease are reflected in tree-ring δ18O on the
eastern TP, where the influence of the ASM, the Indian Summer monsoon and the
westerlies overlap.
We present a new, well-replicated 800-year δ18O chronology,
representing a unique archive for studying the past hydroclimate in eastern
Tibet. We applied response and transfer functions and obtained a reliable
reconstruction of summer relative humidity (July + August). We compared the
long-term trend of our chronology to other moisture-sensitive proxy archives
from several sites over the TP and discuss climatic control mechanisms on the
relative humidity.
Lhamcoka tree-ring δ18O isotope chronology. (a)
Individual δ18O time series of five individuals. The coarse
resolution between 1707 and 1864 results from
shifted block pooling. (b) Running EPS (calculated for 25-year
intervals, lagged by 10 years) and number of trees used for the
reconstruction (solid line). Dashed line represents the theoretical EPS
threshold of 0.85. (c) Tree-ring δ18O chronology spanning
the period AD 1193–1996. Green solid line represents a 50-year smoothing
spline. The red dashed line marks the turning point towards heavier isotope
ratios after ∼ 1870.
Material and methods
Study site – Lhamcoka
Lhamcoka is located on the eastern TP (see Fig. green
pentagon). During a field campaign in 1996, 16 living Juniperus tibetica trees were cored twice in order to enhance the chance of detecting
missing rings. The samples were collected from a steep, south-east exposed
slope at an elevation of 4350 ma.s.l. (31∘49′ N,
99∘06′ E). The oldest tree is 801 years old, resulting in
an overall chronology time span of AD 1193–1996. The average single core
length is 633 years, with single segment lengths of 801 yr, 697 yr, 668 yr, 528 yr,
and 469 yr. The
chronology is not biased by an age trend as it was supposed for different
high-altitude mountain ecosystems (;
).
We applied a spline-based trend analysis and revealed non-systematic trends during the first 100
years after germination (graph not shown here). Therefore, a “juvenile” effect is not likely to
affect our chronology, justifying the retention of the oldest parts of each single core. Juniper
forms the upper timberline in the region due to its cold temperature tolerance .
The species' annual tree-ring growth is limited by temperature and spring precipitation (February–April)
(see Lhamcoka E site description in ). Therefore, the early wood formation is
negatively affected by spring conditions, leading to growth reduction of the annual growth rings.
Due to the steep slope angle of more than 30∘ and well-drained substrate properties at the study
site, ground water influence can be excluded. Therefore, we assume the trees δ18O source
water properties are mainly controlled by the oxygen isotope configuration of summer precipitation,
although it is known that snow-derived meltwater input affects the source water properties of trees
. According to dry and cold winter monsoon conditions
(see climate diagram in Fig. ), a high and persistent snow cover
at our study site is not likely. Hence, 13 % of potential solid precipitation falling
between October and April will probably not strongly influence the source water properties at our study site.
Lhamcoka is influenced by the Indian summer monsoon system with typical
maxima of temperature and precipitation during the summer months (see climate
diagram in Fig. ). The nearby climate station Derge
(3201 m a.s.l., 50 km from the sampling site) records 78 % (541 mm) of
annual precipitation between June and September, which is in accordance with
common monsoonal climate properties . The Derge climate
record (data provided by the China Meteorological Administration) revealed
increasing temperatures of about 0.6 ∘C during the period
1956–1996, whereas the amount and interannual variability of precipitation
remained constant within these 41 years.
Five trees were chosen for isotope analysis to adequately capture inter-tree
variability of δ18O . The trees were selected
for the (i) old age of the cores, to maximize the length of the derived
reconstruction, (ii) avoidance of growth asymmetries due to slope processes,
(iii) sufficient amounts of material (samples with wider rings were
favoured), and (iv) high inter-correlation among the tree-ring width series
of the respective cores.
Sample preparation
We used the tree-ring width master chronology of in
order to date each annual ring precisely. The dated tree-rings were cut with
a razor blade under a microscope. δ18O values were measured from
each tree individually in annual resolution. During periods of the chronology
with extremely narrow rings, we used shifted block pooling to obtain
sufficient material . Pooling was applied between the
years 1707 and 1864 (see chronology parts with missing expressed population
signal (EPS) in Fig. ). To obtain pure α-cellulose,
we followed the chemical treatment presented in . The
α-cellulose was homogenized with an ultrasonic unit and the
freeze-dried material was loaded into silver capsules . The
ratio of 18O / 16O was determined in a continuous flow mass
spectrometer (Delta V Advantage; Thermo Fisher Scientific Inc.). The standard
deviation for the repeated measurement of an internal standard was better
than 0.25‰.
Statistical analyses
We used standard dendrochronology techniques of chronology building, model
building, and verification for reliable climate reconstruction
. All analysis were conducted with the open source
statistical software R (http://cran.r-project.org/). The stable isotope
chronology was calculated within the “dplR” package developed by
and the dendroclimatological correlation and response
analyses were conducted by the “bootRes” package . The
pooling method we executed required a running mean calculation. Thus, the
presented chronology has a quasi-annual resolution, smoothed with a 5-year
running mean filter. To evaluate the isotope chronology reliability, the EPS
(introduced by ) and the
Gleichläufigkeit (GLK) were computed. The EPS expresses the
variance fraction of a chronology in comparison with a theoretically infinite
tree population, whereas the GLK specifies the proportion of
agreements/disagreements of interannual growth tendencies among the trees of
the study site. The EPS is interrupted within our δ18O chronology
at parts where we applied shifted block pooling.
Results
Chronology characteristics
The Lhamcoka δ18O chronology is defined by a mean of 21.27‰
and global minima/maxima of 18.24‰/24.83‰. The values are
similar to results from nearby studies
(; ;
). Moreover, the trees within the chronology are
characterized by a common signal that is expressed by an EPS of 0.88 and a
highly significant GLK of 0.57 (p<0.01). Thus, our selected trees are
likely to be affected by a common force, a prerequisite for compiling a reliable
mean δ18O chronology. The chronology can be
sub-divided into two parts (see Fig. ). The younger
section (AD 1868–1996) shows a pronounced trend of about 2‰ towards
heavy isotope ratios. Within this segment, the year with the heaviest ratio
was detected in AD 1943 (24.8‰). Before the late 1870s, the isotope
δ18O values oscillate around the chronology mean. A phase of
considerable low δ18O values is obvious from AD 1200 to 1300. Within
this section, the lightest isotope ratio was detected in AD 1272
(18.2‰). The signal strength (EPS) occasionally drops below the
commonly accepted threshold of 0.85 during several periods. One reason might
be the imprecise cutting of very narrow rings (ring width
< 0.2 mm). A mix of several rings produces a signal that cannot be
related with certainty to a specific year, a well-known problem when using
very old trees (; ). Nevertheless,
we have confidence in the Lhamcoka chronology due to an EPS above the
threshold during the period AD 1300–1700.
Climatic response of tree-ring stable oxygen isotopes
We conducted linear correlation analyses between the δ18O values
and monthly climate data as well as calculated seasonal means of climate
elements. The available climate record of station Derge covers 41 years
(AD 1956–1996) and correlations were calculated for temperature (mean),
precipitation, relative humidity, sunshine hours (duration of global
radiation > 120 W m-2), and vapour pressure (see Fig. ).
Response of tree-ring δ18O to monthly/seasonal
temperature, precipitation, relative humidity, sunshine hours, and vapour
pressure over the period AD 1956–1996. Gray and black bars indicate
correlations significant at p<0.05 and p<0.01,
respectively; p indicates months/seasons of the previous year.
Summer moisture conditions explain most of the variance of the δ18O
chronology during the calibration period (AD 1956–1996). The stable oxygen
isotopes are highly significantly (p<0.01) correlated with
precipitation, relative humidity, sunshine hours, and vapour pressure during
July and August. Highest (negative) correlations were obtained with relative
humidity during July (r=-0.73) and July/August (r=-0.71). Thus, if
relative humidity is high, transpiration is lowered and the depletion of
light 16O due to leaf water fractionation is reduced.
Additionally, weak and non-significant relationships were found with the mean
temperature in all months/seasons. Thus, concepts of integrated
temperature–moisture indexes, e.g. the vapour pressure difference (VPD:
), are unlikely to explain more of the variance
in our data. However, we calculated the VPD as the difference between water
vapour saturation pressure (E) and vapour pressure (e) and correlated the
VPD time series against the δ18O during the calibration period.
From this we obtained significant but slightly weaker relationships with VPD
(r=0.68, p<0.01) since relative humidity and VPD are both
influenced by temperature. Moreover, sunshine hours are positively related to
the δ18O variation. This association of high sunshine hours, less
cloudiness, decreased relative humidity, and thus increased δ18O
values was corroborated by findings for south-east Tibet .
Very weak correlations were found with climate conditions during the previous
year. Therefore, plant physiological carryover effects as well as stagnating
soil water can be regarded as insignificant factors for tree-ring δ18O
variations. The explained variance of the linear regression model between annual stable oxygen isotope values and relative humidity is 53 %. Hence,
the δ18O value mainly depends on relative humidity, which is in
accordance with the findings of
. Although highest
correlations were obtained with single months (July: r=-0.73 (p<0.01)), the reconstruction was established for the summer season (mean
relative humidity of July and August). In terms of using the wood cellulose
of a single year, the humidity reconstruction of the major growing season is
more robust than for single months.
Reconstruction of relative humidity
We employed a linear model for the reconstruction of relative humidity over
the past 800 years. The linear relationship was achieved for the
δ18O values and instrumental records of relative humidity at
climate station Derge between AD 1956 and 1996. The model was validated
according to the standard methods presented in , and
. We applied the leave-one-out validation procedure due to
the short time period of available climate data. The model statistics are
summarized in Table .
Verification statistics according to the linear transfer model of
δ18O and relative humidity within the calibration period AD 1956–1996.
Sign test (ST)
0.73 (p<0.1)
Product-moment correlation (PMC)
0.67 (p<0.01)
Product means test (PMT)
3.3 (p<0.01)
Reduction of error (RE)
0.45
Coefficient of efficiency (CE)
0.45
The validation tests indicated that (1) the number of agreements between the
reconstructed climate series and the meteorological record is determined
according to the sign orientation significantly different from a pure chance
driven binomial test (ST); (2) the cross-correlation between the
reconstruction and the measurement is highly significant (PMC, PMT), and (3)
the reconstruction is reliable due to a positive RE and CE, indicating the
reconstruction is better than the calibration period mean .
Thus, our linear model is suitable for climate reconstruction purposes. The
model related to the reconstruction of summer relative humidity is described
as RHJA=-2.3⋅δ18O+125.3 (RHJA
expressed in percentage). A negative relationship between tree-ring stable
oxygen isotopes and relative humidity was documented properly in several
studies around the globe and among different species
(;
;
;
). However, due to varying
environmental settings (e.g. climate, soil) and different biological leaf
properties
,
the slopes of the regression function differ significantly among study sites
and species. Hence, δ18O inferred model parameters from a
neighbouring summer relative humidity reconstruction (June–August) using
Abies trees differ from our regression model
.
Summer (July + August) relative humidity reconstruction
AD 1193–1996 for the eastern TP. Solid black and red lines represent
50-year and 150-year smoothing splines, respectively. Red dashed line
emphasizes the turning point towards drier conditions (∼1870s). The
horizontal gray line illustrates mean relative summer humidity (RH = 72.4 %). Vertical dashed lines mark relatively dry periods. The
Medieval Warm Period (MWP) and Little Ice Age (LIA) are emphasized in yellow
and blue.
Our reconstruction reveals several phases of high and low summer humidity
(see Fig. ). Negative deviations from the mean value
(72.4 %; sd = 4.9 %) occurred during AD 1300–1345, AD 1475–1525,
AD 1630–1670, and AD 1866–1996 (periods are emphasized with dashed vertical lines
in Fig. ). The most pronounced relative humidity
depression started in the late 1870s (dashed red line in Figure
) and lasts until approximately the 1950s. The period is
characterized by the driest summer in AD 1943 (RH = 68.4 %). The
remarkable moisture reduction since the end of the LIA has been validated for
the southern and south-eastern part of the TP (;
; ). After
approximately the 1950s, a clear trend towards even drier conditions is
attenuated (trend slope = 0.01, p=0.63). This finding is in accordance
with results from the central and south-eastern TP
(; ; ) and might
be caused by uneven warming trends of the northern and equatorial Indian
Ocean sea surface temperatures
. More humid periods were
detected during AD 1193–1300, AD 1345–1390, AD 1455–1475, and AD 1740–1750, with the
highest relative humidity in AD 1272 (RH = 83.5 %). Thus, the MWP is
characterized by the highest humidity values within the past 800 years.
Similar conditions were observed for inner Asia and the
northern TP (;
) but were not corroborated for the central
TP . The moderate oscillation of our humidity
reconstruction during the LIA contrasts results of increasing and decreasing
moisture trends at different parts of the TP (;
; ). We identified extreme
interannual humidity variations by calculating the third standard deviation
of the first differences. Years with humidity variations above 10 % were
detected in AD 1960/1961, AD 1946/1947, AD 1941/1942, AD 1706/1707, AD 1253/1254, AD 1238/1239,
AD 1233/1234, AD 1230/1231, and AD 1225/1226.
Spatial correlation of July–August relative humidity (ERA interim
data, AD 1979–2013) at the (a) 500 hPa and (b) 300 hPa
pressure level. Colour code represents the Pearson correlation coefficient.
White lines delineate the 95 % significance level. Proxy location is shown
by the light green dot.
Discussion
Lhamcoka is located at the assumed boundary zone of air masses from the
Indian Ocean, the South and North Pacific, and Central Asia .
Thus, our study site is likely influenced by the monsoon circulation (Indian
and Southeast Asian monsoon) as well as by the westerlies
. In particular, the long-term spatiotemporal
modulation of the monsoon circulation systems has been intensively studied
(e.g. ;
;
) and may significantly
control the moisture availability at our study site. The precondition for the
formation of the monsoon is the land–sea surface temperature gradient
between the Asian land mass and the surrounding oceans
. However, the monsoon circulation system
shows variations at interannual and intraseasonal timescales
. In particular, the ENSO
impact on the monsoon circulation has been studied extensively (e.g.
;
;
). We tested the
influence of ENSO on our humidity reconstruction and achieved no significant
relationships, implying an ENSO decoupled climate variability at our proxy
site (see interactive discussion of this paper
). On an intraseasonal timescale,
the Madden–Julian Oscillation (MJO) modulates the monsoonal precipitation
, where the 30- to 90-day zonal
propagation of cloud clusters causes breaks in and strengthening of the
monsoonal precipitation . More
recently, the monsoon circulation system has been affected by greenhouse gas
and aerosol emissions
(;
). Both induce a positive anomaly of
monsoonal precipitation due to the strengthening of the thermal gradient in
the upper troposphere.
However, in this study, we primarily focus on the controls of relative
humidity at our study site, rather than targeting large-scale atmospheric
circulation influences immediately. Therefore, we conducted correlation
analysis of the July–August relative humidity at the grid cell of our study
site with the July–August relative humidity in the area of
0–45∘ N, 40–120∘ E (ERA Interim data:
http://apps.ecmwf.int/datasets/data). Beforehand, we examined the
accordance of our summer relative reconstruction and the ERA interim data
(mean relative humidity July–August). The significant relationship (r=0.77, p<0.01) suggests that the ERA interim data are likely to
represent our relative humidity reconstruction. As shown in
Fig. a, significant correlations at the 500 hPa
pressure level are found with almost the entire TP. This suggests a regional
signal, reflecting the strong connection of moisture variability at our study
site with moisture variability over the whole TP. However, significant
negative relationships were found with the south-west and south-east Asian
regions. These correlations are even more evident on the 300 hPa level
(Fig. b) and show a remarkable spatial pattern.
Interestingly, the negative correlation in south-west Asia contains the
region where defined
an index for the westerly wave activity (west central Asia:
35–40∘ N, 60–70∘ E). The significance of this finding is
corroborated by strong correlations of the mean summer relative humidity in
200 hPa of the west central Asian region and our proxy record (r=-0.58,
p<0.05). Several studies highlight the general influence of the
ASM as the major driver for Tibetan moisture variability
(; ;
). However, the results
of ,
,
, and our findings indicate that
the mid-latitude westerlies influence should be taken into consideration in
future studies.
For an analysis of the regional representativeness of our data set, we
compared the Lhamcoka δ18O chronology with six moisture-sensitive
proxies from the TP (see Fig. and locations
in Fig. ), including normalized tree-ring (TR)
δ18O records (Ranwu TR: ; Reting TR:
), tree-ring width data (Dulan TR:
), accumulation records (Dasuopu and Dunde ice cores:
) and lake sediments (Qinghai sediment:
). We found significant positive correlations between
our time series and the Ranwu (r=0.55, p<0.01), Reting (r=0.23, p<0.01), Dunde (r=0.16, p<0.01) and
Qinghai (r=0.22, p<0.1) data sets. Only the tree-ring width
series of Dulan is negatively correlated with the δ18O values of
Lhamcoka (r=-0.16, p<0.01). The snow accumulation rate of
Dasuopu ice core has no relationship with our δ18O chronology (r=-0.04, p=0.3). In the case of weak correlations (|r| < 0.2)
and due to the degrees of freedom (DF >100), significance levels
alone might be misleading and indicate only a statistical and not a causal
relationship. However, strong relationships between the tree-ring
δ18O chronologies of Lhamcoka and Ranwu, and partially Reting, are
reasonable since moisture reconstructions from these sites rely on the same
proxy (δ18O of tree-ring cellulose) and the trees grown under
similar climate conditions. Relationships with the more northern sites
(Dunde, Dulan, Qinghai) are difficult to verify according to a clearly
detectable westerly influence at these sites. We adapted the colour scheme of
Fig. and highlighted the MWP (yellow polygon), LIA (blue
polygon), and the remarkable humidity decline since the late 1870s (dashed
red line) in Fig. 7. The MWP is characterized by more humid conditions on the
eastern TP (Lhamcoka), a drier phase on the central plateau (Reting) and
moderate humidity conditions on the northern plateau (Dulan). During the LIA,
a remarkable moisture increase occurred at the central and southern plateau
(Reting, Dasuopu). Although humidity was high according to these archives,
the ASM was weak during that time (;
).
Multiproxy comparison of tree-ring data (TR), ice core, and lake
sediment data. TR: Lhamcoka, this study; Ranwu, ; Reting,
; Dulan, . Ice: Dasuopu and
Dunde, . Sediment: Qinghai, .
Locations of the several proxies are shown in Fig. 1. Z-scores were derived
from raw proxy data and not from reconstructions. High positive Z-scores
indicating dry conditions for TR and sediment records, whereas high Z-scores
of ice accumulations represent humid
conditions.
Thus, the findings for Reting and Dasuopu revealed moisture conditions during
cold phases and even drier circumstances during warm periods which might be
contrary to findings of and
. The sudden moisture decrease since the late 1870s
affects the eastern (Lhamcoka), southern (Dasuopu), and central (Reting)
parts of the TP. Reasons for the sudden moisture decline were discussed in
detail by . They address the effect of the moisture decrease on
the reduction in the thermal gradient induced by uneven land-ocean temperature rise caused by
aerosol and greenhouse gas loads. In fact, under rising northern hemispheric air temperatures
, the air moisture load over
sea is increased and the meridional moisture transport is contemporaneously hampered due to the black
aerosol induced solar dimming effect .
In addition,
attributed the moisture decline to the
weakening of the easterly trade wind system along the equatorial Pacific
since the middle of 19th century. Moreover, decreasing varve thicknesses
imply a weakening Asian summer monsoon over the past 160 years
. The aforementioned analysis revealed a link to warm
phases of ENSO and an anomalous regional Hadley circulation. However, their
explanation approach remains incomplete due to dynamic issues associated with
rising temperatures and a weakening South Asian summer monsoon. Therefore, a
terminal explanation is not given yet and should be discussed in future
studies.
In comparison to tree-ring sites located further south (e.g.
; ; ), the distinct
humidity decline is more pronounced on the central and eastern TP. From that
observation, , concluded that there is a weakening of the
monsoon over the last 100–200 years due to uneven SST variation (equatorial
vs. northern Indian Ocean regions). To test this hypothesis, we calculated
the averaged SST anomalies of the equatorial and northern Indian Ocean
(2.5∘N–2.5∘ S, 52.5–112.5∘ E;
22.5—27.5∘ N, 52.5–112.5∘ E). As shown in Fig. 7, a
slight SST increase in both regions beginning in approximately the 1950s is
obvious. Besides, the gradient constantly decreases, but has restrengthened
since approximately the 1970s. This finding contrasts with a generally
weakening monsoon circulation over the past 100–200 years deduced from a
thermal gradient reduction. Therefore, the various moisture variations of the
southern and central/eastern TP during the last 100–200 years might show influences of varying local air mass characteristics.
Sea surface temperature anomalies in different regions of the Indian
Ocean: Bay of Bengal–North Indian ocean (SST BB:
equatorial and northern Indian Ocean (2.5∘N–2.5∘ S, 52.5–112.5∘ E;
22.5—27.5∘ N, 52.5–112.5∘ E) and equatorial Indian Ocean (SST EIO:
2.5∘N–2.5∘ S, 52.5–112.5∘ E) . Difference
between the two time series is marked with a blue line.