Moisture dipole over the Tibetan Plateau during the past five and a half centuries

Abstract

The South Asian Monsoon and mid-latitude Westerlies are two important controls on Tibetan Plateau (TP) fresh water resources. Understanding their interaction requires long-term information on spatial patterns in moisture variability on the TP. Here we develop a network of 23 moisture-sensitive tree-ring chronologies from major juniper forests in a north-south transect on the eastern TP. Over the past five and a half centuries, we find that these chronologies cluster into two groups, North and South, of ∼33° N. Southern and northern regional chronology subsets are positively and significantly correlated with May-June Palmer Drought Severity Indices (PDSI). The meridional moisture stress gradient reconstructed from these data suggests substantial stochastic variation, yet persistent moisture stress differences are observed between 1463-1502 CE and 1693-1734 CE. Identification of these patterns provides clues linking them with forced or intrinsic tropical-extratropical interactions and thus facilitates studies of interannual-decadal dipole variations in hydroclimate over the TP.

Figure 1. Location of the 23 sites for tree-ring network studies on the Tibetan plateau.

The six grids in red represent the northern region PDSI data grids and the five grids in green represent the southern region PDSI data grids used to develop the regional reconstruction targets. The site numbers are in the order from north to south and the site information is listed in Supplementary Table 1.

Figure 2. Spatial structure of the 23 site chronologies in the common period 1,753–2,000 CE.

The size of circles represents value of loadings for sites in the first rotated EOF (a) and the second rotated EOF (b; the filled circle represents value of opposite sign). Cluster analysis shows linkage relations of the northern sites (coloured in pink) and southern sites (coloured in green; c). The tree-ring site numbers are arranged in the order of latitude from north to south and are consistent with those in Fig. 1 and Supplementary Table 1.

Figure 3. Relationships between regional tree-ring chronologies and monthly PDSI.

Response coefficients were calculated for months from October of prior growth year to September of growth year in 1953–2005 for the north (represented by pink bars) and south (represented by green bars) regions of eastern Tibetan Plateau. Response coefficients significant at P<0.05 (as tested by 1,000 bootstrap replications) are denoted by stars.

Figure 4. Spatiotemporal patterns in May–June PDSI on the TP for the past five and a half centuries.

Reconstruction of May–June PDSI (curve in black) from network of tree-ring chronologies (the number of component site chronologies is shown) spans 1,442–2,005 CE for the northern TP (a) and 1,451–2,006 CE for the southern TP (b). Moisture gradient index (MGI, southern reconstruction minus northern reconstruction) between the two regions is calculated (curve in pink) (c), and the calculation also includes the number of positive MGI in a 31-year window (the value is positioned in the 16th year of the window) sliding over the past five and a half centuries (curve in pink) with a background (in grey) of the same calculation but derived from 1,000 first-order autoregressive simulations of the reconstructions of both regions (d). The time series of moisture reconstructions in a and b are smoothed by a 20-year Fast Fourier Transform filter. The grey shading in curves a, b and c denotes 95% confidence interval for prediction of the mean values. The upper and lower borders in d indicate 99% range of the 1,000 simulated values. The green bars indicate two climate states of prolonged south–north moisture dipole.