Tree ring-based annual streamflow reconstruction for the Heihe River in arid northwestern China from ad 575 and its implications for water resource management

Abstract

We developed two tree ring-width chronologies (Qilian juniper, Sabina przewalskii Kom.) for the inland Heihe River Basin in arid northwest China using a large number of tree-ring samples (217 samples/92 trees) with accurate information about pith offsets based on Regional Curve Standardization (RCS) and standard dendrochronological (STD) methodologies. Two 1422-year reconstructions of annual (August–July) streamflow for the upstream region of the Heihe River are presented. The STD and RCS reconstructions account for 53.4% and 57.2% of the actual streamflow variance during the period 1958–2006, respectively. Both reconstructions display considerable low frequency (multidecadal to multicentury) fluctuations, although the RCS based reconstruction is superior to the STD based reconstruction for retention of low-frequency trends. Low-flow years in ad 818–852, 1112–1196, 1453–1495 and 1680–1710, and high-flow periods in ad 868–1000, 1056–1094, 1228–1271, 1327–1440, 1510–1583 and 1877–2006 are detected in both reconstructions. Both the STD and the RCS reconstructions testify to the fact that the 20th century witnessed intensified pluvial conditions in the upstream region of the Heihe River in the context of the last 1500 years. The streamflow reconstructions are anticipated to be useful to water resource planning and management for the Heihe River Basin in arid northwestern China.

Keywords

arid northwest China dendrohydrology Heihe River streamflow reconstruction tree rings water resources management

Introduction

Water shortage due to drought is a very serious environmental, economic and social problem in semi-arid and arid regions of the world. The efficient utilization and management of water is a high priority in such areas, especially when problems are compounded by increasing populations and scarce water supplies. Today’s water resource planning is based primarily on modern hydrological records. Unfortunately, these records are typically less than 100 years in length and capture only a limited portion of the spectrum of natural hydro-climatic variability. This limits projections of future water resource variability and long-term planning and management of water resources since anthropogenic effects will strongly modulate natural variability (Bradley 2008; Huang, 1985; Jonathan et al., 2009; Kang et al., 2002; Li, 2000). We therefore have to rely on paleoclimatic records in order to place recent hydrological conditions in a long-term hydroclimatic context. Tree rings can provide annually resolved, exactly dated proxy climate information, and therefore have been widely used to extend existing instrumental streamflow records (Akkemik et al., 2008; Case and MacDonald, 2003; Cleaveland and Stahle, 1989; Li, 2000; Meko and Graybill, 1995; Meko and Woodhouse, 2005; Meko et al., 2001, 2007; Saito et al., 2008; Smith and Stockton, 1981; Stockton and Meko, 1975; Woodhouse, 2001; Woodhouse and Lukas, 2006; Woodhouse and Meko, 2002; Woodhouse et al., 2006). Furthermore, streamflow reconstructions from tree rings have been applied to water resource planning and management (Jacobs et al., 2005; Meko and Woodhouse, 2011; Meko et al., 2001; Woodhouse and Lukas, 2006; Young, 1995;). In China, a number of dendrohydrological reconstructions are established (Gou et al., 2007, 2010; Li, 2000; Liu et al., 2010; Qin et al., 2010; Yuan et al., 2005). These reconstructions, however, have mainly been carried out for the Xinjiang region and the Anemaqing Mountains of arid northwestern China and most reconstructions span several hundred of years.

In this study, we developed annual streamflow reconstructions dating back to ad 575 for the upper Heihe River Basin in arid northwest China based on a well-replicated data set of 217 samples/92 trees. In terms of chronology length and sample depth, this is a notable extension to our previous work (Qin et al., 2010) in which a 1009-year reconstruction of Yingluoxia streamﬂow gauge was established by using 111 samples/54 trees. Furthermore, these updated reconstructions have been created from a more robust calibration and verification process. Two versions of tree-ring chronologies were produced, using the traditional standard dendrochronological method (STD), and the Regional Curve Standardization (RCS) approach. The aim of the study was to gain an improved understanding of the natural variability of the Heihe River streamflow over the past 1500 years. An additional benefit of the research is the potential implication for water resource management, whereby the probabilistic analysis of streamflow in typical low-flow and high-flow time periods could be explored.

Study area

The Heihe River Basin is an inland river basin in the arid zone of northwestern China (Figure 1). The Qilian Mountains provide the main water resource area of the Heihe River, and have an elevation varying between 2000 and 5500 m. In the Heihe River basin, agriculture and ecosystem place heavy demand on water resources. The river’s flow in the upstream regions provides a water supply for agricultural production and ecosystem stabilization in the middle and lower reaches of the river basin. Vegetation and soil patterns show a distinct vertical zonation (for detailed information, see Qin et al., 2010).

Figure 1.

Map of the tree-ring sites near Qilian County (black triangles: Zhamashike, Youhulou and Binggou tree ring sites)

The sampling area is located in Qilian County, in the upstream region of the Heihe River in the central Qilian Mountains, northwestern China (Figure 1). The studied Qilian juniper (Sabina przewalskii Kom.), a dominant tree species, grows on south-facing slopes. The closest meteorological station is Qilian station (38°11′N, 100°15′E, 2787 m a.s.l.). A detailed description including climate diagrams of temperature and precipitation at this station is given elsewhere (Qin et al., 2010). In the study area, Yingluoxia (38°49′N, 100°10′E, 1694 m a.s.l.), a hydrological gauge station is located between the upstream and middle-stream areas, and records streamflow variations at the outlet of the mountainous areas of the upper reaches of the Heihe River. The mean annual streamflow is 15.7×108 m³/yr for the period 1957–2007, and the annual streamflow at Yingluoxia is fed predominantly by precipitation with only a small contribution (<10%) coming from ice or snowmelt. The seasonal distribution of precipitation and streamflow are consistent, confirming precipitation to be the main water source for the streamflow (Qin et al., 2010).

Materials and methods

Tree ring data

The Qilian juniper derived tree-ring width (TRW) data come from three locations; BG (17 trees), YHL (46 trees) and ZMSK (54 trees), all in the upper Heihe River region in the sub-alpine zone between 3100 and 3800 m in Qilian county (Figure 1). Ring width was measured with a LINTAB II measuring system with a resolution of 0.01 mm, and all cores were cross-dated using visual growth pattern matching and statistical tests in the software package TSAP (Rinn, 2003; Stokes and Smiley, 1968). Cores that correlated weakly with the master series, either due to age-related trends or different local habitats, were removed. Finally, 217 increment cores with accurate information of pith offsets from 92 trees were included for analysis. Because of a common climatic signal at the three sites (Qin et al., 2010), the composite or regional chronology was developed by using all increment cores collected at the three sites.

Age-related biological trend analysis

The segment length of the individual detrended tree-ring series determines the maximum timescale of retrievable climatic fluctuations (Cook et al., 1995). As a result, traditional tree-ring detrending processes removes low-frequency climate signals on timescales that are equal to the mean length of the samples. Cook et al. (1995) called this problem the ‘segment length curse’, as it limits the reconstruction capability of low- frequency climate trends from tree-ring data. A potential solution to this issue is the Regional Curve Standardization (RCS) method, which computes the expected value of the tree-ring data as a function of biological age, then uses the resulting growth curve to standardize the individual tree-ring series. This technique is theoretically appealing, and widely employed in dendroclimatic reconstructions for preserving low-frequency signals, especially multicentennial climate signals in long tree-ring chronologies (Briffa and Melvin, 2008; Briffa et al., 1992, 2001; D’Arrigo et al., 2006; Esper et al., 2002, 2003; Helama et al., 2005; Melvin, 2004; Melvin and Briffa, 2008; Nicault et al., 2010). It is crucial to the RCS method that all possible biological-growth populations are classified first (Esper et al., 2003). In this paper, the TRW series were arranged by biological (cambial) age according to the accurate pith-offsets of individual TRW records. It was noted that the entire TRW data could be classified into two different groups based on biological-growth curves. One population can be fitted with a Hugershoff function (108 samples; Figure 2a), and another can be approximated by a negative exponential function (109 samples; Figure 2b). This differs somewhat from those results reported by Xu and Shao (2006). In that study, it was suggested that only the Hugershoff function represented the biological trend of Qilian junipers growing in the Qaidam Basin close to our study area. That study was based on a small data set of about 70 tree-ring series. In this work, the Hugershoff group reaches a peak of maximum growth at 24–58 years, while the negative group peaks at 8–26 years (Figure 2c), indicating that the two groups are controlled by different local environments. The Regional Curves (RCs) were truncated at 900 biological years with a sample depth of more than 12 trees (Figure 2). For comparison, traditional STD methods were also employed. Since the mean segment length (MSL) of our data set is 658 years, while the MSL of the data used in the study of Esper et al. (2003) was 131 years, this study can be viewed as a case study investigating the application of RCS methodology to tree ring samples with a longer MSL.

Figure 2.

Two categories of biological-growth populations and sample depths

Chronology development

Both RCS and STD TRW chronologies were produced using the ARSTAN program (Cook, 1985). Following the above analyses, the Hugershoff and the negative exponential functions were applied to remove biological growth trends inherent in the raw ring-width series. Before standardization, an adaptive power transformation was applied to eliminate any heteroscedastic behavior in the original tree-ring series (Cook and Peters, 1997). This growth trend is removed from the individual tree-ring curves by subtraction rather than by division for the RCS and STD chronologies to prevent the resulting index curves from being flawed in the case of low growth rates (Esper et al., 2003). All detrended series were averaged to a mean chronology by computing the biweight robust mean in order to reduce the influence of outliers (Cook and Kairiukstis, 1990). Variance stabilization was applied to adjust for changes in variance associated with the declining sample size over time (Osborn et al., 1997). Thus two subchronologies RCS and STD, were built for this study region, following from the two detrending approaches. The arithmetic mean and the two sub chronologies are shown in Figure 3.

Figure 3.

The final STD chronology (left bottom panel) and its corresponding subchronologies (left upper and middle panels), and the final RCS chronology (right bottom panel) and its corresponding subchronologies (right upper and middle panels). Subchronologies are derived from the Hugershoff function group and the negative exponential function group.

Since the sample depth declines in the early portion of the tree-ring chronology, the theoretical or true population signal inherent in the chronologies weakens. The Expressed Population Signal (EPS; Wigley et al., 1984) was calculated to determine the most reliable periods of the STD and RCS chronologies by using a 30-year moving window with 15-year overlaps. A cutoff value of 0.85 of EPS is generally considered to be a reasonable threshold for acceptance of chronology quality (Wigley et al., 1984).

Hydro-meteorological data and dendroclimatic modeling

Monthly temperature and precipitation data (1957–2007) were obtained from the Qilian station of the National Meteorological Information Centre of China, and monthly streamflow (1957–2007) data were obtained from the Yingluoxia hydrological gauge station. The nearest Palmer Drought Severity Index (PDSI) gridpoint data (1957–2005, 2.5°×2.5°, the center is 38.75°N, 98.75°E) were derived from Climate & Global Dynamics (CGD) in Global/Land PDSI data-server v3 format (Dai et al., 2004).

Correlations were computed between the chronology and monthly averages of temperature, precipitation, PDSI and streamflow for the period from the previous May to the current October. We also carried out a response function analysis in order to preclude potential multicolinearity between the climate data series (Fritts, 1976). To test the validity of our climate growth model, we employed split-sample calibration and verification tests (Meko and Graybill, 1995) using the Pearson correlation coefficient (r), reduction of error (RE), sign test and coefficient of efficiency (CE) statistics. Both RE and CE are measures of shared variance between the actual and modeled series. Owing to the fact that the time period common to both the climate and the tree ring data is rather short, we also performed a leave-one-out verification (Michaelsen, 1987) for the derived climate–growth relationship.

Results and discussion

Chronology characteristics

The STD and RCS chronologies cover the span of 1575 years from AD 432 to 2006. The values for mean sensitivity (MS), inter-series correlation (R), EPS and effective chronology signal are all large (Table 1), indicating that both chronologies contain a strong common signal related to the climatic forcing of tree growth. It is notable that the first-order autocorrelation coefficient (AC1) is larger for the RCS chronology than for the STD chronology, indicating that more low-frequency variance is preserved in the RCS chronology.

Table 1.

Characteristics of the STD and RCS chronologies and the statistics calculated for the common period of analysis

Chronology characters^a							Common period of analysis (1200–1800)^b
		CL	ML	MS	STD	AC1	Rbar	Rbar1	Rbar2	Reff	EPS	EPS>0.85
	1	432–2008	621	0.157	0.265	0.42	0.398	0.519	0.367	0.657	0.933	575/3
RCS	2	581–2008	696	0.153	0.186	0.39	0.292	0.442	0.249	0.563	0.915	635/3
	3	575–2008	658	0.146	0.191	0.53	–	–	–	–	–	–
	1	432–2008	621	0.157	0.181	0.331	0.398	0.518	0.367	0.657	0.933	575/3
STD	2	581–2008	696	0.156	0.174	0.327	0.296	0.431	0.255	0.59	0.922	635/3
	3	575–2008	658	0.147	0.159	0.33	–	–	–	–	–	–

CL: chronology length; ML: mean length; MS: mean sensitivity; STD: standard deviation; AC1: first-order autocorrelation.

Rbar: mean inter-series correlation; Rbar1: within-trees Rbar; Rbar2: between-trees Rbar; Reff: effective chronology signal, the max value is 1.0, which corresponds to the best effective in chronology signal; EPS: expressed population signal; EPS>0.85, year and number of trees with signal strength up to 0.85, calculated for ARSTAN standard chronologies for 30-year intervals with 15-year overlaps.

Based on a threshold of 0.85 for EPS, the most confident period of the STD and RCS chronologies was reached at AD 575 with a sample depth of six samples. In comparison, the STD chronology has a weak signal strength at around AD 650 (EPS = 0.74) and 725 (EPS = 0.80) for the negative exponential curve group, and during the period AD 635–665 (EPS = 0.72–0.84) for the hugershoff curve group. The RCS chronology has a weak signal strength at around AD 650 (EPS = 0.75) and 725 (EPS = 0.79) for the negative exponential curve group and during AD 635–665 (EPS = 0.75–0.84) for the hugershoff curve group (Figure 3). We advise that the climatic interpretation for these parts of the chronologies be treated with caution.

Visual inspection indicates that the RCS and STD chronologies have a generally similar trend especially for the period ad 800–1700 (Figure 4a). To investigate the difference between them, a difference series between the two chronologies was calculated (Figure 4b). It is obvious that the maximum difference between the two chronologies occurs in the earliest 200 years and the latest 200 years (Figure 4b). The regular distribution of the young trees entering the chronology period (Figure 4c) and the number of tree-ring cores >200 (Esper et al., 2003) justify the application of the RCS in this case.

Figure 4.

Comparison between the STD /RCS chronologies and their sample depths. (a) The STD and RCS chronologies, (b) the difference between the STD/RCS chronologies, (c) sample depth

Hydroclimate–ring width relationship

The climate–tree growth relationships were analyzed using the software DENDROCLIM 2002 (Biondi and Waikul, 2004). For both chronologies, the correlation between monthly means of temperature, precipitation, streamflow, PDSI and tree growth were examined for the period from May of the year before growth to October of the growth year over the common period 1957–2006. Since the STD and RCS chronologies show similar correlation and response patterns to hydroclimate data, here we show only the results of the correlation and response function analysis between the RCS chronology and hydroclimatic factors (Figure 5). TRW-temperature correlation analysis shows that the correlation coefficients are positive for most months, with significant positive correlations occurring between the previous November and January of the current year. Significant positive correlations are found between monthly precipitation for July and August of the previous year and May and June of the current year, while a significant negative correlation is found for September of the current year. Correlation with monthly PDSI appears weak. The results of the response function analyses confirm these relationships (Figure 5), although the response values calculated between ring width and climate variables are lower for some months than those calculated for simple correlation coefficients. The highest correlation coefficients were found for the monthly streamflow variable. Ring width is positively correlated with streamflow for most months in the study period, except September and October of the current year. The correlations are significant at the 0.05 level for data from the previous August to the current July. The response function analyses show that correlations with streamflow in June and July of the current year are significant at the 0.05 level. These findings are consistent with earlier dendroclimatological studies carried out in the Qilian Mountains in Qinghai Province (Kang et al., 2002; Qin et al., 2010). It is noteworthy that streamflow has no direct influence on the growth of trees on the mountain slopes. However, streamflow and tree growth are both influenced by common climatic forcings. It is therefore reasonable to assume that tree growth in the study region is a good indicator of streamflow variation. We selected streamflow as the reconstructed variable in this study.

Figure 5.

Correlation (solid line) and response function (dashed dot line) between the RCS chronology and monthly mean values of temperature (a), precipitation (b), streamflow (c), and PDSI (d). The horizontal dashed line indicates the 95% confidence level for the correlation function. Response functions significant at the 0.05 level are marked with an asterisk

Reconstruction and validation

We computed the correlation coefficients between ring width and various seasonal assemblages of streamflow for the period 1958–2006 to determine the appropriate season for reconstruction. In agreement with previous research (Qin et al., 2010), we found that the time span from the previous August to the current July (P8–C7) is the best time span for streamflow reconstruction for the Yingluoxia gauge in the upper Heihe River. The correlation coefficients are 0.73 and 0.76 for the STD and RCS chronologies, respectively. A linear regression model (Y = 17.12X – 3.38, the STD chronology; Y = 13.41X – 2.25, the RCS chronology) was developed to reconstruct the annual (previous August to current July) streamflow history for the upper Heihe River (Figure 6). A comparison between the observed and the reconstructed annual streamflow reveals that low- and high-frequency variations in streamflow are adequately captured by the RCS and STD reconstructions (Figure 6). The models account for 53.4% and 57.2% of the actual streamflow variance during the period 1958–2006 for the STD and RCS chronologies, respectively (Table 2). The leave-one-out verification for the full period of available streamflow data resulted in RE values of 0.51 (0.55) for the STD (RCS) reconstructions, indicating that the reconstructions are robust (Fritts, 1976). The CE statistic is similar to RE except that its benchmark for determining model skill is the verification period rather than the calibration period. The CE values are greater than 0.49 for the different verification periods. All these statistics suggest that the reconstruction is reliable, and can be used to reconstruct past streamflow variations.

Figure 6.

Comparison of gauge flow records and the STD/RCS reconstructions at the Yingluoxia hydrological station during the period 1957–2006

Table 2.

Statistics calculated for the regression model using different calibration and verification periods for the STD/RCS reconstructions^a

Calibration	Verification	R_c	R_c²	R_adj²	R_v	R_v²	RE	CE
1958–1988	1989–2006	0.7/0.71	0.5/0.51	0.48/0.49	0.77/0.86	0.6/0.74	0.62/0.71	0.49/0.55
1976–2006	1958–1975	0.72/0.74	0.51/0.54	0.49/0.53	0.71/0.75	0.5/0.56	0.6/0.64	0.56/0.66
1958–2006	–	0.73/0.76	0.53/0.57	0.52/0.56	0.72/0.75^L	0.53/0.57 ^L	0.51/0.55^L	–

R_c: correlation coefficient for the calibration period; R_c²: explained variance in the regression model for the calibration period; R_v: correlation coefficient for the verification period; R_v²: explained variance of the regression model for the verification period; RE: reduction in the calculated error statistic for the verification period; CE: coefficient of efficiency for the verification period.

L indicates the leave-one-out verification.

The statistical characteristics of the reconstructions from the STD and RCS chronologies are shown in Table 3. The high discharge years, normal years, and low discharge years generally trisect each other over the past 1432 years. The mean streamflow during high-flow years, normal years, and low-flow years for the STD reconstruction (RCS reconstruction) are 16.69×10⁸ m³/yr (14.05×10⁸ m³/yr), 13.69×10⁸ m³/yr (11.10×10⁸ m³/yr), and 10.42×10⁸ m³/yr (8.30×10⁸ m³/yr), respectively. The long-term mean of the streamflow reconstruction for the STD reconstruction (the RCS reconstruction) is 13.65×108 m³/yr (12.08×10⁸ m³/yr), which is 2.08×10⁸ m³/yr (3.65×10⁸ m³/yr) lower than the mean annual streamflow 15.73×10⁸ m³/yr for the current period 1958–2006.

Table 3.

Characteristics of the STD and RCS streamflow reconstructions during the period ad 575–2006^a

Parameter	Max	Min	Mean	HF	PHF	N	PLF	LF
STD	23.85	3.83	13.65	16.07%	15.92%	36.83%	14.14%	17.04%
RCS	19.48	4.16	12.08	17.04%	13.47%	37.28%	14.51%	17.71%

Max, Min, and Mean are the maximum, minimum and mean values (10⁸ m³/yr) of the reconstructed streamflow series, respectively; HF, PHF, N, PLF, and LF denote percentage of years with the P values P>20%, 10% < P ≤ 20%, −10% ≤ P ≤ 10%, −20% ≤ P < −10%, and P > −20%, respectively. The parameter P, percentage of streamflow anomaly, is expressed as the ratio of annual streamflow anomaly (difference between annual streamflow and long-term average) to long-term average. Positive P values are indicative of high annual streamflow events whereas negative P values indicate low rates of river discharge.

The STD and RCS reconstructions show similar multidecadal to multicentury variation trends during the last 1500 years (Figure 7). Since water resource planning and management time horizons are 50 years in the Heihe River Basin, the annual streamflow series was smoothed with a 50-year low-pass (FFT) filter to highlight multidecadal hydrological changes. The filtered curve reveals that several long-lasting severe low streamflow epochs occurred in the past. Although the mean value differs between the STD and RCS reconstructions, the variation pattern is similar (Figure 7a, b). For both reconstructions, long-term below-average streamflow periods were identified at AD 818–852, 1112–1196, 1453–1495, and 1680–1710. Among these, the three lowest and extended streamflow intervals are centered on AD 810, 1130, and 1480. Above-average streamflow periods occurred at AD 868–1000, 1056–1094, 1228–1271, 1327–1440, 1510–1583, and 1877–2006. The most marked difference between the RCS and STD reconstructions is that the time span for the period of intensified high-flow beginning ad 1720, although present in both reconstructions, lasts 280 years in the former, and only 120 years in the latter. There are a growing number of papers that show RCS chronologies to exhibit long positive trends (Yadav, 2011; Yadav et al., 2011; Zhu et al., 2011). Melvin (2004) and Briffa and Melvin (2008) identified the ‘trend-in-signal’ and ‘contemporaneous-growth-rate’ biases, both of which possibly impart a longer scale positive trend to the RCS chronologies. Taking the Dulan sample set (a set of long-lived Qilian junipers in western China) as an example, Yang et al. (2011a, 2012a, 2012b) investigated the possible reason in detail by correlating deviations of individual tree-ring width records from their regional mean age-dependent curve (RC). They found that the intra-record correlations kept their positivity for various age shifts, and attributed it to the influence of fast- versus slow-grown trees. It was concluded that just the unification of these tree-ring series in the RC form created a spurious positive trend in the Dulan chronology. Compared with the reports of Yadav (2011), Yadav et al. (2011) and Zhu et al. (2011), the long positive trend in our RCS curve is not strong during the past centuries, as indicated by the comparison with the STD chronology, suggesting that possible bias issues in our RCS chronology are likely to be minor. At any rate, acquisition of large sample replication and the use of the signal-free approach (Melvin, 2004), multiple sub-RCS curves (Briffa and Melvin, 2008) and the eigen analysis (Yang et al., 2011a, 2011b, 2011c) could be important in future studies that look to mitigate the trend distortion effect and to combat the RCS sample depth issues.

Figure 7.

The STD (a) and RCS (b) streamflow reconstructions and comparison with a temperature reconstruction for China (c) (Yang et al., 2002). The thin black line shows annually resolved data, and the bold line is smoothed with a 50-year FFT filter. The horizontal thin line represents the long-term mean and the horizontal bold line represents the mean value for 1958–2006. The gray hatched bars show the main high flow intervals. The gray shaded areas show the obvious low flow intervals

The pluvial condition in the 20th century is comparable to the high-flow conditions evident around AD 868 and 1000 in both the STD and RCS reconstructions. The period AD 580–870 is a notable low-flow period in the RCS reconstruction, whereas it is generally a low-flow period with alternating high- and low-flow fluctuations in the STD reconstruction. Since the signal strength EPS in both the STD and RCS chronologies dropped below 0.85 in AD 635–665, the hydrological variability for this period should be treated with a degree of caution. Cook et al. (2010) developed the Monsoon Asia Drought Atlas (MADA) for the past 700 years using tree rings from 327 sites in monsoonal Asia, but the tree-ring data presented in this study are not included in the MADA data. Here we made a comparison between our streamflow reconstructions and the gridded PDSI reconstructions from the MADA. It was found that our STD and RCS reconstructions were consistent with the closest gridded PDSI reconstruction (38.75°N, 101.25°E) for the period 1300-2005 (figure not shown). Their correlation coefficients were 0.4 and 0.29 (p<0.001, n=706), and were 0.48 and 0.18 in low-frequency domain (>10 years), respectively. The low-flow phases centered at AD 1340, 1480, 1720, 1820, 1928 and 1975 agreed well with the MADA PDSI reconstruction. The most significant difference is the rapid rising trend in the recent decade in streamflow data which is not present in the PDSI reconstruction by Cook et al. (2010). The gridded moisture reconstruction is an indicator of Asian Monsoon drought for summer season. This correspondence indicates a close relationship between low- and high-flow variations and Asian monsoon drought and pluvials over the past seven centuries.

Focusing on the high-flow conditions evident in the 20th century, although the variability of the 20th century high-flow interval did not exceed the natural variability range (i.e. around 890) in either reconstruction, the persistence of such high streamflow conditions is unique in the context of the past 1500 years. Intensified pluvial conditions in the 20th century were also reported in northeastern Qinghai Province (Liu et al., 2006; Sheppard et al., 2004), the Tian Shan area in northwest China (Li et al., 2006), Tibet (Bräuning and Mantwill, 2004), Mongolia (Davi et al., 2009; Pederson et al., 2001), the western Himalaya (Singh and Yadav, 2005; Singh et al., 2009; Yadav et al., 2011), northern Pakistan (Treydte et al., 2006), monsoonal Asia in general (Anderson et al., 2002; Zhang et al., 2008), the western United States (Cook et al., 2004) and Germany (Wilson et al., 2005). It has been argued that the intensification of the hydrological cycle in the 20th century in different parts of the world could be attributable to regional temperature increases (Esper et al., 2002, 2007; Mann and Jones, 2003; Moberg et al., 2005; Yang et al., 2002, 2007). When viewed alongside a temperature reconstruction for China (Yang et al., 2002; Figure 7c), it can be seen that high-flow conditions generally correspond to warm periods and low-flow conditions generally coincide with cold periods (an exception to this occurs around AD 800). This conclusion is consistent with our previous research, which concluded that dry intervals generally coincide with cold periods in the Qilian Mountains and vice versa (Qin et al., 2011; Yang et al., 2010).

Power spectral analysis was used to detect multi-timescale variability (Figure 8). The dominant periodicities in our streamflow reconstructions are around 2–2.3, 2.7, 36–40, 47–51, 68–74, 136, 159, 191, 239, 318, 478, and 956 years, which are significant at the 0.05 level. It should be noted that although the 956-year period is significant, it should not be over interpreted as the whole series length is less than 1500 years and extreme low-frequency spectra are subject to edge effects. All these periods indicate that the hydrological variability of the Heihe River may have been influenced by forcing factors such as solar activity, tropical ocean–atmosphere coupling systems and other climate dynamic systems, as have been indicated in our previous research (Qin et al., 2010). Similar periodicities were detected in regional precipitation reconstructions carried out for arid Northwest China (Wang et al., 2004) and Tian Shan (Li et al., 2006), as well as in the Asian summer monsoon record (Zhang et al., 2008) and in a reconstruction of the water levels of Lake Erie in the Gulf of Alaska (Wiles et al., 2009). It should be noted that the spectral density of the RCS reconstruction is higher in long periodicities and lower in short periodicities. This suggests that the RCS approach is superior to the STD methodology in the retention of low frequency signal.

Figure 8.

Comparisons between power spectral analysis results for the STD/RCS streamflow reconstructions

Implications for water resource management

Temporal extension of modern gauge flow records is important to dendrohydrological studies, but the application of streamflow reconstructions to water resource planning and management is arguably more important. Studies of Lees Ferry in Colorado River (Young, 1995), Sacramento River in California (Meko et al., 2001), the lower Colorado River Basin (Jacobs et al., 2005), the Colorado headwaters region and the South Platte basin (Woodhouse and Lukas, 2006) have proven useful in providing key objective evidence for decision making and water resource management. Our streamflow reconstruction is potentially valuable to the Heihe River Basin, where rapid population increase and the expansion of the irrigation area in the middle reaches make efficient water management essential. To optimize utilization of water resources, unified water resource allocation projects in the Heihe River Basin were implemented by the Ministry of Water Resources in December, 1997. However, the water allocation project was formulated on the basis of probabilities calculated from limited streamflow observations (1957–1997) from the Yingluoxia gauging station. The mean value from the modern gauge flow record is 15.8×10⁸ m³/yr. According to the regulations, for example, when the runoff at Yingluoxia station reaches 15.8×10⁸ m³/yr, 9.50×10⁸ m³/yr water should be released into the downstream area of the Heihe River Basin. The gauge flow record is a limited segment of a long-term hydroclimatic history, and represents part of a period of intensified pluvial conditions that have been experienced over the past 1500 years. From the analysis above, low-flow and high-flow conditions lasting from several decades to centuries occurred several times in the past. It is clear that the short instrumental flow record does not capture the full spectrum of natural variability which is important to the water resource planning horizon. The severe low flow epochs that are seen to have occurred in the past might occur again in the future. Therefore it is necessary to provide water resource planners and policy-makers with information on patterns of streamflow variation between typical high and low flow periods, and to guide them in devising robust water resource planning and operation strategies.

Based on our reconstruction, discussions related to water resource planning were motivated by the analysis of different streamflow patterns (expressed as cumulative frequency distribution in river discharge) over time intervals typical for high and low flow conditions combined with the current water allocation program (97’s water allocation program) in the Heihe River Basin. We selected AD 931–980, 1381–1430 as a typical high-flow time horizon (50 years in length), and AD 791–840, 1131–1180, 1451–1500 as a typical low-flow time horizon during the past 1500 years. The multiyear mean value (15.8×10⁸ m³/yr) is a reference point for 97’s water allocation program. According to instrumental records (1958–2006) in P8-C7, the new reference point is 15.73×10⁸ m³/yr (which is very close to the reference point 15.8×10⁸ m³/yr), which reaches up to a cumulative frequency of 50% (regarded as a guaranteed rate for inflow). The streamflow patterns of cumulative frequency (probability) are different at various time intervals (Figure 9). If the reference point remains unchanged for high-flow periods, the value of 15.73×10⁸ m³/yr corresponds to a cumulative frequency of 38% (90%) in AD 931–980, and 54% (99%) in 1381–1430 in the STD (RCS) reconstructions, respectively. This implies that the probability that annual streamflow would not meet this reference point (the value of 15.73×10⁸ m³/yr) for the Heihe River Basin was 40% higher during the period AD 931–980 (90%) than for the instrumental period (50%) according to the RCS reconstruction. According to the STD streamflow reconstruction, it indicates that the probability of annual streamflow reaching this reference point, 15.73×10⁸ m³/yr, for the Heihe River Basin was 12% greater during AD 931–980. It is likely that the probability that annual streamflow would fail to the value 15.73×10⁸ m³/yr would be 4% (49%) higher in 1381–1430 for the STD (RCS) reconstruction. Such situations worsen in the case of low-flow periods. Taking AD 1451–1500 as an example, the value 15.73×10⁸ m³/yr corresponds to a cumulative frequency of 97% for this dry period of the reconstruction, implying that the probability of annual streamflow failing to reach this value of 15.73×10⁸ m³/yr would be 47% higher for this period for the STD reconstruction than for the instrumental period (50%). From the perspective of guaranteed river flow rate, it is an obviously impossible to create a water resource allocation program during low-flow periods, but such low-flow conditions might occur in the future. If we consider the full reconstruction period, AD 575–2006, we can expect to derive a more accurate accumulative frequency function since this data set is closer to the theoretical population of the streamflow data. One can see that the reference value 15.73×10⁸m³/yr corresponds to a 77% (95%) probability for the STD (RCS) reconstruction. It means that only 23% (5%) guaranteed rate of river discharge compared with 50% guatanteed rate of river flow for the instrumental period. Therefore, the current water allocation program is unreasonable from a long-term perspective.

Figure 9.

Cumulative frequency distribution for reconstructed annual flows. The upper blue curve represents the RCS-, and the lower red curve the STD-, reconstruction (colour figure available online). Cumulative frequency is plotted for different subperiods of typical high and low flow periods. The mean value 15.73×10⁸ m³/yr of instrumental flow (1958–2006) is indicated for reference

From the view of guaranteed rate of river flow, a 50% guaranteed rate provides an alternative reference point associated with the recent 50-year arithmetic mean value 15.73×10⁸m³/yr in the current water allocation program (Figure 10). Based on this new reference point, the 50% probability corresponds to 16.12×10⁸ m³/yr (13.69×10⁸ m³/yr) in AD 931–980, and 15.43×10⁸ m³/yr (12.22×10⁸ m³/yr) in 1381–1480 for the STD (RCS). This means that if similar streamflow conditions to those that occurred in the period ad 931–980 occur in the future, we should adjust the water resource allocation program on the basis of a river discharge of 16.12×10⁸ m³/yr (13.69×10⁸ m³/yr) as indicated in the STD (RCS) reconstruction. Similar discussions can be had regarding the low-flow periods and the full reconstruction period. Therefore a 50% guaranteed rate of river flow is also a reasonable reference for water resource planning in the Heihe River. Compared against the reference point of the 50-year mean value, a 50% guaranteed rate is a more reasonable reference for creating a water resource allocation program, especially in situations where river discharge is too large or too small for some periods.

Figure 10.

Cumulative frequency distribution of reconstructed annual flows. The upper blue curve represents the RCS-, and the lower red curve the STD-, streamflow reconstruction (colour figure available online). Cumulative frequency is plotted for different subperiods of typical high and low flow periods. The 50% guaranteed rate of instrumental flow (1958–2006) is indicated for reference

To sum up, there are different streamflow variation patterns (cumulative distributions) in different high and low flow periods. The 15.73×10⁸ m³/yr, a 50-year mean value, should be adjusted with time. The 50% guaranteed rate of river flow is also a robust and reasonable choice of reference point for water resource management. Here we have proposed alternatives of water division programs (Table 4) based on typical high and low flow time intervals during the last 1500 years.

Table 4.

Different water allocation programs under various high and low flow conditions

	Guaranteed rate for river flow	10	25	50	75	90	Multi-year average
High flow period	Instrument record (1958–2006)	19.47	17.16	15.73	13.86	12.37	15.73
	STD (1958–2006)	17.94	16.74	15.80	14.64	13.01	15.73
	RCS (1958–2006)	17.67	16.88	16.26	14.76	13.00	15.73
	STD (931–980)	18.76	17.42	16.12	13.28	11.89	15.34
	RCS (931–980)	14.15	13.49	13.62	10.77	9.78	11.98
	STD (1381–1430)	19.19	17.52	15.43	14.68	13.32	16.26
	RCS (1381–1430)	16.27	14.64	12.22	12.56	11.84	13.78
Low flow period	STD (791–840)	16.70	14.97	12.06	10.43	9.06	12.45
	RCS (791–840)	11.83	10.50	9.24	8.08	6.67	9.15
	STD (1131–1180)	15.35	13.87	12.26	9.75	8.50	11.77
	RCS (1131–1180)	12.40	11.41	9.74	8.16	6.82	9.72
	STD (1451–1500)	14.82	12.47	11.69	10.05	8.20	11.48
	RCS (1451–1500)	11.08	9.60	8.76	7.61	6.39	8.70
The full period	STD (575–2006)	17.06	15.62	13.75	11.80	9.95	13.64
	RCS (575–2006)	14.55	12.81	11.07	9.24	7.90	11.11
	97s water allocation program	19.00	17.10	–	14.20	12.90	15.80

Conclusions

An updated tree ring-width chronology was produced using 217 tree ring increment cores with accurate information of pith offsets and extended back to AD 432 in time. The STD and RCS reconstructions show similar multidecadal to multicentury variation trends for the last 1500 years. This reconstruction is by far the longest and most high-quality streamflow series in China. The persistence of the high streamflow conditions in the 20th century is unique in the context of the past 1500 years.

Two categories of tree growth patterns which can be described by a Hugershoff and negative exponential function, are detected in Qilian juniper species in the Qilian Mountains of northwestern China. It is more reasonable to remove the biological trend based on different biological-growth populations. The RCS detrended chronology is superior to the STD for its retention of low-frequency trends, although these are not very significant in this study. The long mean segment length (658 year) of our data set determines the large extent to which more low-frequency variance is preserved in the STD chronology.

Tree ring-based water resource planning and management is a new area of research. We have pointed out the necessity and urgency of making robust water resource planning and operation strategies in advance because different high and low streamflow conditions in the past can be expected to occur again in the future. We suggest that 50% guarantee rate of river flow is a reasonable reference for water resource planning in the Heihe River Basin.

Footnotes

Acknowledgements

The authors thank two anonymous reviewers for their thoughtful and constructive comments. This paper also benefited from revisions by the editor Mary Gagen.

The study was jointly funded by the National Basic Research Program of China (973 Program) (No. 2010CB950104), the Chinese Academy of Sciences (CAS) 100 Talents Project (No. 29082762), the National Science Foundation of China (Grant Nos. 41002050, 41071130), and the Chinese Academy of Sciences Visiting Professorship for Senior International Scientists (Grant No. 2009S1-38). Bao Yang gratefully acknowledges the support of the K.C. Wong Education Foundation, Hong Kong.

References

Akkemik

D’Arrigo

Cherubini

. (2008) Tree-ring reconstructions of precipitation and streamflow for north-western Turkey. International Journal of Climatology 28(2): 173–183.

Anderson

Overpeck

Gupta

(2002) Increase in the Asian SW Monsoon during the past four centuries. Science 297: 596–599.

Biondi

Waikul

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

Bradley

(2008) Holocene perspectives on future climate change. In: Battarbee

Binney

(eds) Natural Climate Variability and Global Warming: A Holocene Perspective. Chichester: Wiley-Blackwell, pp. 254–268.

Bräuning

Mantwill

(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:10.1029/2004GL02079310.1029/2004GL020793.

Briffa

Osborn

Schweingruber

. (2001) Low-frequency temperature variations from a northern tree ring density network. Journal of Geophysical Research 106(D3): 2929–2941.

Briffa

Melvin

(2008) A closer look at Regional Curve Standardization of tree-ring records: Justification of the need, a warning of some pitfalls, and suggested improvements in its application. In: Hughes

Diaz

Swetnam

(eds) Dendroclimatology: Progress and Prospects? Springer, pp. 113–146.

Briffa

Jones

Bartholin

. (1992) Fennoscandian summers from AD 500: Temperature changes on short and long timescales. Climate Dynamics 7: 111–119.

Case

MacDonald

(2003) Tree-ring reconstructions of streamflow for three Canadian Prairie Rivers. Journal of American Water Resources Association 39(3): 703–716.

10.

Cleaveland

Stahle

(1989) Tree-ring analysis of surplus and deficit runoff in the White River, Arkansas. Water Resources Research 25(6): 1391–1401.

11.

Cook

(1985) A time-series analysis approach to tree-ring standardization. PhD Thesis. The University of Arizona Press.

12.

Cook

Kairiukstis

(1990) Methods of Dendrochronology: Applications in the Environmental Sciences. Dordrecht: Kluwer Academic Publishers.

13.

Cook

Peters

(1997) Calculating unbiased tree-ring indices for the study of climatic and environmental change. The Holocene 7(3): 361–370.

14.

Cook

Anchukaitis

Buckley

. (2010) Asain monsoon failure and megadrought during the last millenniun. Science 328: 486–489.

15.

Cook

Briffa

Meko

. (1995) The ‘segment length curse’ in long tree-ring chronology development for palaeoclimatic studies. The Holocene 5: 229–237.

16.

Cook

Woodhouse

Eakin

. (2004) Long-term aridity changes in the western United States. Science 306: 1015–1018, doi: 10.1126/science.110258610.1126/science.1102586.

17.

Dai

Trenberth

Qian

(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.

18.

D’Arrigo

Wilson

Jacoby

(2006) On the long-term context for late twentieth century warming. Journal of Geophysical Research 111: D03103, doi:10.1029/2005JD00635210.1029/2005JD006352.

19.

Davi

Jacoby

D’Arrigo

. (2009) A tree-ring based drought index reconstruction for far Western Mongolia: 1565–2004. International Journal of Climatology 29(10): 1508–1514.

20.

Esper

Cook

Krusic

. (2003) Tests of the RCS method for preserving low frequency variability in long tree-ring chronologies. Tree-Ring Research 59(2): 81–98.

21.

Esper

Cook

Schweingruber

(2002) Low frequency signals in long tree-ring chronologies for reconstructing past temperature variability. Science 295: 2250–2253.

22.

Esper

Frank

Büntgen

. (2007) Long-term drought severity variations in Morocco. Geophysical Research Letters 34: doi: 10.1029/2007GL03084410.1029/2007GL030844.

23.

Fritts

(1976) Tree Rings and Climate. London: Academic Press.

24.

Gou

Chen

Cook

. (2007) Streamflow variations of the Yellow River over the past 593 years in western China reconstructed from tree rings. Water Resource Research 43: W06434, doi:10.1029/2006WR00570510.1029/2006WR005705.

25.

Gou

Deng

Chen

. (2010) Tree ring based streamflow reconstruction for the upper Yellow River over the past 1234 years. Chinese Science Bulletin 55: doi: 10.1007/s11434-010-4215-z10.1007/s11434-010-4215-z.

26.

Helama

Timonen

Lindholm

. (2005) Extracting long-period climate fluctuations from tree-ring chronologies over timescales of centuries to millennia. International Journal of Climatology 25: 1767–1779.

27.

Huang

(1985) Hydrology. Beijing: Higher Education Press.

28.

Jacobs

Meko

Nijssen

. (2005) Enhancing water supply reliability through improved predictive capacity and response. Oral presentation in session, Drought analysis, prediction, and impact mitigation, American Geophysical Union 2005 fall meeting, 7 December, San Francisco.

29.

Jonathan

Cook

Yang

(2009) Climate change over the past 2000 years in Western China. Quaternary International 194: 91–107.

30.

Kang

Cheng

Kang

. (2002) Mountainous run-off reconstruction of Heihe River during past 1000 years using tree-ring. Science in China (Series D) 32(8): 675–685.

31.

Gou

Cook

. (2006) Tree-ring based drought reconstruction for the central Tien Shan area, Northwest China. Geophysical Research Letters 33: L07715. doi: 10.1029/2006GL02580310.1029/2006GL025803.

32.

(2000) Dendrohydrology Research and Application. Beijing: Science Press.

33.

Liu

. (2006) Tree-ring precipitation and the relationship to north hemisphere mean temperature in Dulan, Qinghai since A.D. 850. Science in China (Series D) 36(5): 461–471.

34.

Liu

Sun

Song

. (2010) Tree-ring hydrologic reconstructions for the Heihe River watershed, western China since 1430 AD. Water Research 44: 2781–2792, doi:10.1016/j.watres.2010.02.01310.1016/j.watres.2010.02.013.

35.

Mann

Jones

(2003) Global surface temperatures over the past two millennia. Geophysical Research Letters 30(15): doi: 10.1029/2003GL01781410.1029/2003GL017814.

36.

Meko

Graybill

(1995) Tree-ring reconstruction of Upper Gila River discharge. Water Resource Bulletin 31: 605–616.

37.

Meko

Woodhouse

(2005) Tree-ring footprint of joint hydrologic drought in Sacramento and Upper Colorado River basins, western USA. Journal of Hydrology 308(1–4): 196–213.

38.

Meko

Woodhouse

(2011) Application of streamflow reconstruction to water resources management. Developments in Paleoenvironmental Research 11(4): 231–261.

39.

Meko

Therrell

Baisan

. (2001) Sacramento River flow reconstructed to AD 869 from tree rings. Journal of the American Water Resources Association 37(4): 1029–1039.

40.

Meko

Woodhouse

Baisan

. (2007) Medieval drought in the upper Colorado River Basin. Geophysical Research Letters 34(10): L10705. doi:10.1029/2007GL02998810.1029/2007GL029988.

41.

Melvin

(2004) Historical growth rates and changing climatic sensitivity of boreal conifers. Ph.D.Thesis. Climate Research Unit, School of Environmental Science, University of East Anglia.

42.

Melvin

Briffa

(2008) A ‘signal-free’ approach to dendroclimatic standardization. Dendrochronologia 26: 71–86.

43.

Michaelsen

(1987) Cross-validation in statistical climate forecast models. Journal of Applied Meteorology 26(11): 1589–1600.

44.

Moberg

Sonechkin

Holmgren

. (2005) Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy data. Nature 433(7026): 613–617.

45.

Nicault

Guiot

Edouard

. (2010) Preserving long-term fluctuations in standardization of tree-ring series by the adaptative regional growth curve (ARGC). Dendrochronologia 28: 1–12.

46.

Osborn

Briffa

Jones

(1997) Adjusting variance for sample-size in tree-ring chronologies and other regional mean time series. Dendrochronologia 15: 89–99.

47.

Pederson

Jacoby

D’Arrigo

. (2001) Hydrometeorological reconstructions for northeastern Mongolia derived from tree rings: AD 1651–1995. Journal of Climate 14: 872–881.

48.

Qin

Yang

Bräuning

. (2011) Regional extreme climate events on the northeastern Tibetan Plateau since AD 1450 inferred from tree rings. Global and Planetary Change 75: 143–154.

49.

Qin

Yang

Burchardt

. (2010) Intensified pluvial conditions during the twentieth century in the inland Heihe River Basin in arid northwestern China over the past millennium. Global and Planetary Change 72: 192–200.

50.

Rinn

(2003) TSAPWin: Time Series Analysis and Presentation for Dendrochronology and Related Applications, Version 0.55 User Reference. Heidelberg.

51.

Saito

Biondi

Salas

. (2008) A watershed modeling approach to streamflow reconstruction from tree-ring records. Environmental Research Letters 3(2): 1–6. doi:10.1088/1748-9326/3/2/02400610.1088/1748-9326/3/2/024006.

52.

Sheppard

Tarasov

Graumlich

. (2004) Annual precipitation since 515 BC reconstructed from living and fossil juniper growth of Northeast Qinghai Province, China. Climate Dynamics 23: 869–881.

53.

Singh

Yadav

(2005) Spring precipitation variations over the western Himalaya, India, since A.D. 1731 as deduced from tree rings. Journal of Geophysical Research 110: D01110. doi:10.1029/2004JD00485510.1029/2004JD004855.

54.

Singh

Yadav

Wilmking

(2009) A 694 year tree-ring based rainfall reconstruction from Himachal Pradesh, India. Climate Dynamics 33(7–8): 1149–1158.

55.

Smith

Stockton

(1981) Reconstructed streamflow for the Salt and Verde rivers from tree-ring data. Water Resources Bulletin 17(6): 939–947.

56.

Stockton

Meko

(1975) A long-term history of drought occurrence in western United States as inferred from tree rings. Weatherwise 28(6): 245–249.

57.

Stokes

Smiley

(1968) An Introduction to Tree-ring Dating. Chicago: University of Chicago Press, pp. 21–24.

58.

Treydte

Schleser

Helle

. (2006) Millennium-long precipitation record from tree-ring oxygen isotopes in northern Pakistan. Nature 440: 1179–1182.

59.

Wang

Yang

Bräuning

. (2004) Precipitation variations in arid and semiarid zones of northern China during the past 500 years. Chinese Science Bulletin 49(9): 883–887.

60.

Wigley

Briffa

Jones

(1984) On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Climate and Applied Meteorology 23(2): 201–213.

61.

Wiles

Krawiec

D’Arrigo

(2009) A 265-year reconstruction of Lake Erie water levels based on North Pacific tree rings. Geophysical Research Letters 36: L05705. doi:10.1029/2009GL03716410.1029/2009GL037164.

62.

Wilson

RJS

Luckman

Esper

(2005) A 500-year dendroclimatic reconstruction of spring–summer precipitation from the lower Bavarian forest region. Germany. International Journal of Climatology 25: 611–630.

63.

Woodhouse

(2001) A tree-ring reconstruction of streamflow for the Colorado Front Range. Journal of the American Water Resources Association 37(3): 561–569.

64.

Woodhouse

Lukas

(2006) Multi-century tree-ring reconstructions of Colorado streamflow for water resource planning. Climatic Change 78(2–4): 293–315.

65.

Woodhouse

Meko

(2002) Introduction to tree-ring-based streamflow reconstructions. Southwest Hydrology 1(2): 14–15.

66.

Woodhouse

Gray

Meko

(2006) Updated streamflow reconstructions for the Upper Colorado River basin. Water Resources Research 42: doi:10.1029/2005WR00445510.1029/2005WR004455.

67.

Shao

(2006) Standardization of Qilian Juniper Ring-width Series in the Eastern Margin of the Qaidam Basin. Acta Geographica Sinica 61(9): 919–928.

68.

Yadav

(2011) Tree ring evidence of a 20th century precipitation surge in the monsoon shadow zone of the western Himalaya, India. Journal of Geophysical Research 116: D02112. doi:10.1029/2010JD01464710.1029/2010JD014647.

69.

Yadav

Braeuning

Singh

(2011) Tree ring inferred summer temperature variations over the last millennium in western Himalaya, India. Climate Dynamics 36: 1545–1554. DOI 10.1007/s00382-009-0719-010.1007/s00382-009-0719-0.

70.

Yang

Bräuning

Johnson

Shi

(2002) General characteristics of temperature variation in China during the last two millennia. Geophysical Research Letters 29(9): 1324. doi:10.1029/2001GL01448510.1029/2001GL014485.

71.

Yang

Bräuning

Yao

. (2007) Correlation between the oxygen isotope record from Dasuopu ice core and the Asian Southwest Monsoon during the last millennium. Quaternary Science Reviews 26: 1810–1817. doi:10.1016/j.quascirev.2007.03.00310.1016/j.quascirev.2007.03.003.

72.

Yang

Qin

Huang

. (2010) Spatial and temporal patterns of variations in tree growth over the northeastern Tibetan Plateau during the period AD 1450–2001. The Holocene 20(8): 1235–1245. doi: 10.1177/095968361037199710.1177/0959683610371997.

73.

Yang

Sonechkin

Datsenko

. (2011a) The eigen analysis of tree-ring records: Part 1, a limited representative of regional curve. Theoretical and Applied Climatology 106: 489–497. DOI 10.1007/s00704-011-0451-710.1007/s00704-011-0451-7.

74.

Yang

Sonechkin

Datsenko

. (2012a) The eigen analysis of tree-ring records: Part 2, Posing the eigen problem. Theoretical and Applied Climatology DOI: 10.1007/s00704-011-0468-y10.1007/s00704-011-0468-y.

75.

Yang

Sonechkin

Datsenko

. (2012b) The eigen analysis of tree-ring records: Part 3, taking heteroscedasticity and sampling effects into consideration. Theoretical and Applied Climatology 10.1007/s00704-011-0498-510.1007/s00704-011-0498-5.

76.

Young

(1995) Coping with a severe sustained drought on the Colorado River: Introduction and overview. Water Resource Bulletin 316: 779–788.

77.

Yuan

. (2005) Reconstruction and analysis of the 355 a runoff of the Manas River on the north slopes of Tianshan Mountains. Journal of Glaciology and Geocryology 27(3): 411–417.

78.

Zhang

Cheng

Edwards

. (2008) A test of climate, sun, and culture relationships from an 1810-year Chinese cave record. Science 322: 940–942.

79.

Zhu

Shao

Yin

. (2011) August temperature variability in the southeastern Tibetan Plateau since AD 1385 inferred from tree rings. Palaeogeography, Palaeoclimatology, Palaeoecology 305(1–4): 84–92.