Can a paleodrought record be used to reconstruct streamflow?: A case study for the Missouri River Basin

Abstract

Recent advances in paleoclimatology have revealed dramatic long‐term hydroclimatic variations that provide a context for limited historical records. A notable data set derived from a relatively dense network of paleoclimate proxy records in North America is the Living Blended Drought Atlas (LBDA): a gridded tree‐ring‐based reconstruction of summer Palmer Drought Severity Index. This index has been used to assess North American drought frequency, persistence, and spatial extent over the past two millennia. Here, we explore whether the LBDA can be used to reconstruct annual streamflow. Relative to streamflow reconstructions that use tree rings within the river basin of interest, the use of a gridded proxy poses a novel challenge. The gridded series have high spatial correlation, since they rely on tree rings over a common radius of influence. A novel algorithm for reconstructing streamflow using regularized canonical regression and inputs of local and global covariates is developed and applied over the Missouri River Basin, as a test case. Effectiveness in reconstruction is demonstrated with reconstructions showing periods where streamflow deficits may have been more severe than during recent droughts (e.g., the Civil War, Dust Bowl, and 1950s droughts). The maximum persistence of droughts and floods over the past 500 years far exceeds those observed in the instrumental record and periods of multidecadal variability in the 1500s and 1600s are detected. Challenges for an extension to a national streamflow reconstruction or applications using other gridded paleoclimate data sets such as adequate spatial coverage of streamflow and applicability of annual reconstructions are discussed.

The impact of the Three Gorges Dam on summer streamflow in the Yangtze River Basin

“…have previously been used to assess the performance of streamflow models (Chen, Hao, Devineni, & Lall, 2014;Devineni, Lall, Pederson, & Cook, 2013;Ho et al, 2016). The comparison between streamflow before and after dam operation provides an approach to assessing the impact of the dam on either exacerbating or mitigating extreme events.…”

“…A positive value indicates that the model has better predictive skill than using the climatology of the verification period.The RE and CE statistics are similar to the Nash-Sutcliffe efficiency score(McCuen, Knight, & Cutter, 2006;Nash & Sutcliffe, 1970) andhave previously been used to assess the performance of streamflow models(Chen, Hao, Devineni, & Lall, 2014;Devineni, Lall, Pederson, & Cook, 2013;Ho et al, 2016). Theoretically, the values of both RE and CE range from the negative infinity to a maximum of 1, with higher values indicating better prediction skill.…”

Streamflow Reconstruction in the Upper Missouri River Basin Using a Novel Bayesian Network Model

“…We conclude that the BSM model improves upon the PCR model in cross validation with the CE statistic. Cook et al (), Devineni et al (), Ho et al () and Rao et al () all found results that were similar in their reconstructions; in particular, it was a bit more difficult to achieve positive values with the CE statistic than with the RE statistic, and hence a slightly greater number of negative values were encountered in the validation with this statistic.…”

“…Barnett et al (2010) were able to explain 44% to 65% variance in their streamflow reconstructions for the Upper Green River Basin. Ho et al (2016) were able to explain an even greater share of the variance in the Missouri River Basin, where the adjusted R 2 values of their reconstructions ranged from 0.56 to 0.90. However, there is a significant overlap in the range of R 2 values in this study with those in our study.…”

“…While there are many more examples that can be cited, we cite just a few prominent ones for the sake of brevity. The RE statistic was also used by Cook and Jacoby (1983), Cook et al (2013), Devineni et al (2013), Bracken et al (2016), and Ho et al (2016) for validation. Cook and Jacoby (1983), for streamflow reconstructions in the Potomac River, found RE values not too far from the median values seen in Figure 8; typically between 0.2 and 0.5, with three notable exceptions.…”