Reconstructing the flow of the Sacramento River since 1560.

Abstract

Tree-ring width time series have been shown to be useful in conjunction with regression techniques for extending instrumental records of climate variables such as precipitation, drought severity and annual streamflow. This study uses tree-ring data derived from 17 sites in northern California and eastern Oregon to develop reconstructions of annual Sacramento River streamflow for the period 1560-1980. Reconstruction methods using principal components analysis (PCA) and autoregressive (AR) modeling are tested and compared. Reconstructions based on both PCA-transformed and untransformed tree-ring data may be superior to reconstructions using only one of these two forms of data. Also, reconstructions using AR modeling of the tree-ring data are found to more accurately reproduce the time series structure of the instrumental streamflow record, but to explain less variance in the data set, than reconstructions which do not use AR modeling. The reconstructed streamflow series shows that the historical period includes the wettest (1854-1916) and driest (1928-1937) periods of the last 400 years, but that many other periods of sustained drought or high flows have also occurred. This reconstruction correlates well with a previous reconstruction of precipitation in California, and shows varying levels of agreement with tree-ring based reconstructions of climate done elsewhere in the western U.S.