Terrestrial vegetation is regarded as a sensitive indicator of climate change (Pauli
The widespread coniferous forests in Central Asia, comprising mostly local tree species (
Therefore, the aims of this study are to: (1) develop a tree-ring width-based regional chronology for the Alatau Mountains; (2) reconstruct historical NDVI series for the study area and explore its variations, and (3) assess climatic signals inhering in the newly developed NDVI reconstruction.
The Alatau Mountains form a boundary between China and Kazakhstan and are part of the Tian Shan Mountains of Central Asia. The Alatau Mountains extend more than 450 km from east to west, and the maximum elevation is approximately 4464 m a.s.l. The Alatau Mountains experience a temperate continental climate. The northern slopes of the Alatau Mountains have more annual precipitation (900–1000 mm) than the southern slopes and a lower snowline (3600 m a.s.l.). Approximately two-thirds of the mountains’ glaciers are distributed in the northern areas (He
A long-lived Schrenk spruce (
Following normal dendrochronological techniques (Speer, 2010), the sampled tree-ring cores were dried naturally and mounted on a wooden plank with grooves. Then, each core was sanded with abrasive paper and marked with needles under a microscope. Every ring on the sanded cores was measured under a binocular microscope using a Velmex Measuring System at a resolution 0.001 mm. This measuring system, the standard of North America’s Dendrological Research Community, is specifically designed for the researcher to perform non-contact measurement analysis. The system is composed of the following three main components: a UniSlide® dovetail rapid advance screw motion assembly, a linear encoder with one micron (0.001 mm) resolution, and a Velmex VRO™ Encoder Readout. Two programs, COFECHA (Grissino-Mayer, 2001) and ARSTAN (Cook and Krusic, 2005), were utilized for cross-dating quality control and to develop chronologies. To eliminate age trends that were affected by factors other than climate, the tree-ring width series were detrended using a negative exponential method,
For further analyses, we selected the precipitation, mean temperature, mean maximum temperature and mean minimum temperature per month of the Climatic Research Unit Time-Series (CRU TS 4.00; New
We used 13-year reciprocal filters to decompose the developed chronologies of the tree rings into high- and low-pass components to evaluate the patterns of variation in various frequency ranges (Yuan
Fig. 2a shows that the highest temperature periods in the study area were in summertime (from June to August), with peaks in July. Most of the annual precipitation fell in May, June, and July (120.1 mm). The two peak values of precipitation in May (42.2 mm) and November (30.1 mm), which account for 12.6% and 8.9%, respectively, of the total annual precipitation. Wintertime has less precipitation than summer. August is typically the wettest month of the year (PDSI=–0.46) for the study area (Fig. 2b). The climate data for 1901–present reveal remarkable rising tendencies in the study area for the following: annual total precipitation (
Fig. 2c indicates that the monthly NDVI values from January to March are relatively low (not exceeding 0.140) and without obvious fluctuations, but have an evident increase starting at April (0.248), and peaking in July (0.536), when the study area reaches the heathiest vegetation period of the year. Thereafter, the monthly NDVI values maintain a long-term decrease until December. An increasing trend of annual mean NDVI since 1982 is less significant (
After cross-dating quality control, two cores from the AMA site and four cores from the KBZ site were rejected because of the low correlations between the subseries and the master series. Ultimately, 50 cores from 28 spruces at the AMA site, 81 cores from 42 spruces at the KBZ site, and 45 cores from 23 spruces at the BSK site were used to develop ring-width chronologies. The oldest tree at the KBZ site was nearly 195 years old. In order to carry out a further analysis, we utilized the standard (STD) chronologies that contained the common variations of respective series of tree samples and retained a low-through high-frequency common variance. This variance is hypothesized as being dependent on climate (Cook, 1985). Three chronologies, the depths of their samples, EPS, and Rbar are demonstrated in Fig. 3. General statistics of these chronologies for a common period of analysis (from 1900 to 2010) are listed in Table 1. The reliable lengths of AMA, KBZ, and BSK chronologies were 107 (1910–2016), 167 (1850–2016), and 137 (1880–2016) years, respectively.
Statistical characteristics of chronologies over the common period of 1900–2010Statistic AMA KBZ BSK ARC Standard deviation (SD) 0.198 0.224 0.414 0.208 Mean sensitivity (MS) 0.180 0.186 0.379 0.180 First-order autocorrelation (AC1) 0.396 0.496 0.577 0.458 Interseries correlation (trees) 0.417 0.484 0.738 0.418 Interseries correlation (all series) 0.431 0.488 0.753 0.422 Mean within-tree correlation 0.786 0.805 0.961 0.808 Signal-to-noise ratio (SNR) 9.858 62.012 18.293 62.025 Expressed population signal (EPS) 0.908 0.984 0.948 0.984 The first principal component (PC#1) 0.492 0.518 0.799 0.461 First year EPS >0.85 1910 1850 1880 1850
In addition to the reciprocal filters, we also used Pearson correlation coefficients to analyze the three sets of data, which included original unfiltered data, high-pass filtered data, and low-pass filtered data. As shown in Table 2, the correlations for the three chronologies employing the low-pass filter are relatively low, but the confidence levels for the correlation coefficients were all over 99.9% in the case of the original, high-frequency, and low-frequency domains over the common period of 1910–2016. In the common period, the years of 1924 (value: 1.452), 1970 (1.421), 1993 (1.319), 1973 (1.307), 1994 (1.292), 1950 (1.282), 1951 (1.275), 1923 (1.27), 1948 (1.264), and 1990 (1.257) are regarded as 10 highest-value years of AMA chronology, and the 10 lowest-value years appear in 2008 (0.511), 1945 (0.607), 1913 (0.614), 1917 (0.625), 1914 (0.632), 1911 (0.645), 2000 (0.675), 1927 (0.694), 1995 (0.705), and 2014 (0.708). The 10 highest-value years of KBZ chronology are 1993 (1.499), 1994 (1.448), 1968 (1.419), 1962 (1.395), 1973 (1.392), 1969 (1.369), 1924 (1.349), 2005 (1.341), 1989 (1.336), and 1970 (1.304), and the 10 lowest-value years appear in 1945 (0.457), 1927 (0.496), 1917 (0.509), 1944 (0.554), 1943 (0.596), 2000 (0.602), 1928 (0.626), 1938 (0.648), 1946 (0.66), and 2015 (0.686). The 10 highest-value years of BSK chronology are 1929 (1.678), 1935 (1.573), 1970 (1.556), 1952 (1.517), 1964 (1.502), 1960 (1.473), 1962 (1.473), 1993 (1.467), 2010 (1.456), and 1954 (1.44), and the 10 lowest-value years appear in 2008 (0.001), 1985 (0.137), 2000 (0.211), 1997 (0.226), 1927 (0.249), 1991 (0.249), 1917 (0.286), 1977 (0.304), 1974 (0.322), and 2014 (0.334). After comparing these extreme-value years, two highest-value years (1970 and 1993) and three lowest-value years (1917, 1927 and 2000) were observed in the three chronologies. The results indicated a good coherence of extreme-values among these chronologies. As a result, we combined all the ring-width data from the AMA, KBZ, and BSK sites in order to establish a regional chronology (ARC) that was longer and more replicated.
Correlation coefficients for three chronologies over the common period of 1910–2016. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Significant at p<0.001. Note: The table shows correlation coefficients. Results for the original, high-, and low-pass filtered chronologies are shown.Original ( High-frequency ( Low frequency ( AMA KBZ BSK AMA KBZ BSK AMA KBZ BSK 1 0.783 0.660 1 0.842 0.682 1 0.641 0.683 / 1 0.624 / 1 0.759 / 1 0.388 / / 1 / / 1 / / 1
The ARC chronology and the depths of the samples are shown in Fig. 3d, and the general descriptive statistics are listed in Table 1. Values of standard deviation (SD) and mean sensitivity (MS) for the chronologies obtained using the ARC chronology are slightly lower than those of the chronologies for single sampling sites because of the combination of samples. The first-order autocorrelation (AC1) assesses relationships between tree growth in a current year and previous growth. These values ranged from 0.396 to 0.577, and revealed that chronologies possessed low-frequency variance, which was affected by the lag effects of climate and tree physiology. The interseries correlations of the three chronologies obtained from sites of single sampling and the regional chronology all exceed 0.4. These values indicated a good coherence in the tree-growth. The ARC chronology with higher signal-to-noise ratios (SNR) and EPS demonstrated more climatic signals in the regional chronology. The reliable length of the ARC chronology was 167 (1850–2016) years based on the initial year of EPS>0.85.
A strong biological lag effect was indicated by the high values of AC1 from 0.396 to 0.577 for the chronologies from single and composite sites (Table 1). Therefore, the climatic data per month (1901–2015) from July in the previous year to October in the current year were applied to assess the influence of climatic factors on radial growth of spruces in our study area. The results of the correlation analysis revealed that the relationship between the radial growth of spruces and precipitation was positive in general, and that significant correlations appeared in July of the previous year and from March to July of the current year (Fig. 4). There was a negative correlation between the tree-ring width of spruces and the temperature. Generally, these chronologies had very negative correlations with the temperature at the end of the growing season of the previous year and in the middle of the growing season of the current year. Furthermore, almost all of monthly the positive monthly correlation coefficients between chronologies and PDSI from July of the previous year to October of the current exceeded the 0.05 significance level. The above results demonstrated that moisture was the main climatic limitation on development of spruces’ tree rings in the study area. More precipitation at the end of the growing season of the previous year and the fast growing season of the current year may enhance the potential for accumulating water reserves in the soil, thus resulting in larger leaves (Liu
The optimal temperature for evergreen conifers to conduct photosynthesis ranges from 10°C to 25°C. If the temperature falls below –3°C to –5°C or raises above 35°C to 42°C, photosynthesis by conifers may cease (Wang, 2000). Fig. 2a reveals that the mean temperature ranged from 4.3°C to 18.3°C in April to October, and from –5.4°C to –4.5°C in March to November, respectively. Therefore, April–October was taken as the growing season for spruces in our study area. The monthly NDVI data (1982–2006) from April to October were utilized to assess the connection between vegetation coverage and tree growth (Fig. 5). Significant correlations were found from July to October based on a 0.05 significance level and peaked in September. By testing various combinations of months, we found that the maximal correlation coefficient was between the ARC chronology and the July–October NDVI (
We computed the correlation coefficients between spruces’ growth in a radial pattern within our study area and various assemblages of months for the NDVI for the period 1982–2006, so that the most suitable season for reconstruction could be selected. To do this, we reconstructed the NDVI from July to October by using of the regional chronology. We also used a linear regression model to portray the relationship between the ARC chronology and the NDVI. The model is as follows:
(n=25,
where NJuly.–Oct. is the July–October NDVI for the study area and ARC refers to the regional chronology. The model accounts for 33.3% of the NDVI variance during the calibration period 1982–2006 for the ARC chronology. Fig. 6a shows that the reconstructed NDVI tracks the observations well. The results of the Leave-one-out and Bootstrap tests (100 iterations in the recomputation process) revealed that values of
Verification results from the Leave-one-out and Bootstrap tests for the NDVI reconstruction.Statistic Calibration Leave-one-out Mean (Range) Bootstrap (100 Iterations) Mean (Range) 0.58 0.58 (0.52–0.63) 0.57 (0.34–0.77) Squared multiple correlation (R2) 0.33 0.33 (0.27–0.40) 0.34 (0.12–0.60) Adjusted squared multiple correlation ( 0.30 0.30 (0.24–0.37) 0.31 (0.08–0.58) Standard error (SE) 0.017 0.017 (0.015–0.018) 0.16 (0.11–0.19) 11.45 11.06 (8.10–14.56) 13.21 (3.17–32.39) Durbin-Watson (D/W) 1.82 1.84 (1.60–2.05) 1.22 (0.49–1.95)
Statistics of split-sample calibration-verification tests for the NDVI reconstruction.Statistic Calibration (1982–1994) Verification (1995–2006) Calibration (1994–2006) Verification (1982–1993) Full calibration (1982–2006) 0.67 0.41 0.44 0.68 0.58 0.43 0.17 0.19 0.46 0.33 0.38 / 0.12 / 0.30 / 0.10 / 0.35 / / 0.09 / 0.34 / / 3.01 / 3.49 /
The above results, which indicates significant skill in the tree-ring estimates, all demonstrate that the model (1) is stable and reliable. Therefore, we used model 1 to reconstruct the NDVI around the Alatau Mountains for July to October during the period of 1850–2016; an average value of 0.418 and a standard deviation of
The values for extreme years of the NDVI reconstruction are now listed. The 10 highest value years were 1897 (value: 0.447), 1993 (0.442), 1924 (0.441), 1970 (0.440), 1962 (0.440), 1994 (0.440), 1973 (0.438), 1952 (0.437), 1852 (0.437), and 1922 (0.435), and the 10 lowest-value years were 1945 (0.389), 1927 (0.389), 1917 (0.390), 1879 (0.390), 1866 (0.391), 2008 (0.393), 1944 (0.396), 2000 (0.396), 1938 (0.398), and 1867 (0.399). It can be seen that the difference between the highest-value (1897) and lowest-value (1945) years is 0.058.
The interdecadal variability was conspicuous. When a 21-year moving average was applied to the reconstruction (Fig. 6b), three periods of dense and sparse vegetation coverage could be differentiated. The periods of dense vegetation coverage (above the mean value of the reconstruction) were 1860–1870 (average value: 0.419), 1891–1907 (0.421), and 1950–1974 (0.423). The sparse vegetation coverage periods (below the mean) were 1871–1890 (0.417), 1908–1949 (0.416), and 1975–2006 (0.416).
The lowest-value years of 1917 and 1938 in the newly reconstructed NDVI series coincide with the historical documents that record a lack of rainfall in summer and autumn leading to a great drought occurring in the Ili region (Wen
Spatial correlation was carried out to evaluate the regional significance of the ring-width based NDVI reconstruction. The NDVI series has a correlations (>0.3) with the gridded PDSI data for July–October for 1901–2015 in a large area between approximately 40°–48°N and 68°–82°E. The highest correlations (>0.4) appear in southern Kazakhstan (Fig. 7). The newly reconstructed NDVI series for the Alatau Mountains (NAM) were compared with four climatic reconstructions based on tree-ring data: (1) August–January standardized precipitation evapotranspiration index for the southeastern Kazakhstan (SKZ, 1785–2014; Chen
Correlation coefficients between the NDVI and climatic reconstructions over the common period of 1850–2005. Significant at p < 0.01 Significant at p < 0.01 Significant at p < 0.05 Significant at p < 0.01 Significant at p < 0.01 Significant at p < 0.01 Significant at p < 0.05 Significant at p < 0.01 Significant at p < 0.01 Significant at p < 0.01 Significant at p < 0.05Original( High-frequency ( Low-frequency ( SKZ PNL PIL PAK SKZ PNL PIL PAK SKZ PNL PIL PAK 0.507 0.460 0.371 0.197 0.580 0.508 0.332 0.189 0.498 0.326 0.452 0.190
To further analyze the large-scale climate anomalies associated with the newly developed NDVI series, two typical years (1993 and 2008) were investigated. As shown in Fig. 9a, above normal rainfall was evident around the Alatau Mountains, which accounts for the highest NDVI in 1993. An anomalous westerly wind was pronounced over Eurasia between 40°–60°N, which brought air rich in water vapor from the Atlantic Ocean, and the Mediterranean and Caspian Seas. As a result, a rainy belt was observed from Western Europe to the Alatau Mountains. In contrast, below normal rainfall was evident around the Alatau Mountains in 2008 (Fig. 9b), which is well represented by the associated lowest NDVI since the 1950s. Unlike the effects of the remarkable zonal water vapor transport in 1993, precipitation in the Alatau Mountains is normally controlled by dry air from high latitudes (He
A 167-year, regional, ring-width chronology was developed from 93 living and healthy spruces on the northern slopes of the Alatau Mountains. The results of correlation analyses revealed that more precipitation and lower temperatures in the growth season of the previous and current years might help to form a wider tree ring in the target coniferous species. Because of the coherence between ring width and vegetation coverage under the same water limitations, a July–October NDVI reconstruction for the Alatau Mountains was developed using the regional chronology. The reconstruction matched the observed data well, and its lowest-value years precisely captured drought events mentioned in historical documents. A spatial analysis and correlations with other tree-ring based climatic reconstructions indicated that the newly reconstructed NDVI series was affected by large-scale climatic oscillations.
The preliminary study in this paper provides new knowledge about the variations in vegetation coverage for the Alatau Mountains and will be helpful in improving the tree-ring dataset for Central Asia. However, this study is limited because it is local research based on only three sampling areas. Thus, further studies should be carried out at a larger spatial scale to verify the results presented in this paper.