Monthly paleostreamflow reconstruction from annual tree-ring chronologies

Paleoclimate reconstructions are increasingly used to characterize annual climate variability prior to the instrumental record, to improve estimates of climate extremes, and to provide a baseline for climate change projections. To date, paleoclimate records have seers limited engineering use to estimate hydrologic risks because water systems models and managers usually require streamflow input at the monthly scale. This study explores the hypothesis that monthly streamflows can be adequately modeled by statistically decomposing annual flow reconstructions. To test this hypothesis, a multiple linear regression model for monthly streamflow reconstruction is presented that expands the set of predictors to include annual streamflow reconstructions, reconstructions of global circulation, and potential differences among regional tree-ring chronologies related to tree species and geographic location. This approach is used to reconstruct 600 years of monthly streamflows at two sites on the Bear and Logan rivers in northern Utah. Nash-Sutcliffe Efficiencies remain above zero (0.26-0.60) for all months except April and Pearson's correlation coefficients (R) are 0.94 and 0.88 for the Bear and Logan rivers, respectively, confirming that the model can adequately reproduce monthly flows during the reference period (10/1942 to 9/2015). Incorporating a flexible transition between the previous and concurrent annual reconstructed flows was the most important factor for model skill. Expanding the model to include global climate indices and regional tree-ring chronologies produced smaller, but still significant improvements in model fit. The model presented here is the only approach currently available to reconstruct monthly streamflows directly from tree-ring chronologies and climate reconstructions, rather than using resampling of the observed record. With reasonable estimates of monthly flow that extend back in time many centuries, water managers can challenge systems models with a larger range of natural variability in drought and pluvial events and better evaluate extreme events with recurrence intervals longer than the observed record. Establishing this natural baseline is critical when estimating future hydrologic risks under conditions of a non-stationary climate. (C) 2017 Elsevier B.V. All rights reserved.