Uncertainty in applied tree-ring reconstructions: Klamath River basin streamflow

Thursday, 18 December 2014
Steven B Malevich and Connie A Woodhouse, University of Arizona, Tucson, AZ, United States
Tree-ring based reconstructions of hydroclimatic variability are useful for water resource management. One of the challenges in generating reconstructions is assessing uncertainties and how they impact use in resource management. This project focuses on recent reconstructions of streamflow for the Upper Klamath River basin. The reconstructions provide several centuries of historical context to drought and hydroclimate variability found in the instrumental record of the past century. The quality of tree-ring reconstructions greatly depends on the quality and the availability of tree-ring data and the instrument records used for reconstruction calibration. This project focuses on the quality and potential uncertainty of instrumental records. A number of hydroclimate instrumental records are available in the Klamath River basin. Unfortunately these records have deficiencies. Complex basin hydrology and a long history of water use have made it difficult to estimate the river’s natural flow. This project develops a methodology to combine uncertainty from the instrumental record with uncertainty from tree-ring reconstructions. The result is a Monte Carlo method which considers complex and potential non-linear interactions. Probability distributions are created for each observation in the instrumental record. An algorithm independently designs, trains, and tests a tree-ring reconstruction for potential outcomes of the instrumental record’s distributions. Preliminary results give probability distributions for past streamflow events. These can be used for commonly applied drought analysis. This includes exceedance probabilities relative to instrumental record drought for single and multiple-year moving averages.