Main content area

Reconstruction of annual runoff since CE 1557 using tree-ring chronologies in the upper Lancang-Mekong River basin

Yang, Bing, Chen, Xiaohong, He, Yanhu, Wang, Jiawen, Lai, Chengguang
Journal of hydrology 2019 v.569 pp. 771-781
El Nino, databases, growth rings, regression analysis, runoff, standard deviation, watersheds
In this study, with the use of a multiple linear regression approach, the tree-ring chronologies of eight sampling sites in the upper Lancang-Mekong River Basin were developed to provide a 449-year (CE 1557–2005) reconstruction of the annual runoff, thus placing recent changes in annual runoff into a long-term context. These eight tree-ring chronologies have recently been archived in publicly available databases through the International Tree-Ring Data Bank. Reconstruction results showed a good correlation coefficient of 0.662 (n = 39, p-value < 0.01) between the reconstructed and the observed annual runoff. The adjusted coefficient (R2) for the degrees of freedom is 42.3%, which meets the precision requirements of reconstruction. The reconstructed runoff displays a trend toward more moist conditions: there were 37 extremely wet years and 23 extremely dry years, exceeding the mean ±1 standard deviation, during the past 449 years. Empirical mode decomposition (EMD) was used to fully analyze and understand the multi-scale variation of the reconstructed runoff. Six intrinsic mode function (IMF) components with different scales were obtained and the sum of all components can be reverted to the original variable sequence. The first and second IMF mainly reflect the change characteristics of the interannual scale of reconstructed sequence. Both are likely controlled by the Quasi-Biennial Oscillation (QBO) and the El Niño-Southern Oscillation (ENSO), respectively. The third IMF showed a 10–13 year scale fluctuation, which is very similar to the solar activity of an 11-year cycle. The fourth and fifth IMF mainly represents multidecadal and centennial oscillations, and they have shown coherent variations with predecessors’ reconstructed May-September precipitations in the same time domain. The lowest frequency component (the residue IMF) represents the trend term of the original signal.