2.2.1. Estimates of Pre-Development Central Valley Runoff from Tree-Ring Data

Annually resolved variations in hydroclimate before the start of instrumented weather records can be inferred from tree-ring records. For some tree species and climate regimes, tree growth is limited by drought stress, such that tree-ring chronologies, or standardized indices of ring width closely track the occurrence of wet and dry years [42–44]. A drought atlas from 835 tree-ring chronologies in North America, which covers two millennia, underscores the shortcomings of a relatively short instrumented record for characterizing extremes of hydroclimate [45]. An expanded network of 1285 chronologies identifies unmatched severe, widespread, persistent Southwest droughts in the medieval period [46], and independent tree-ring evidence from exposed stumps in lakes and rivers suggests that two such droughts in the Sierra Nevada may have lasted more than two centuries [47]. Paleo-simulations of Mono Lake from tree-ring data independently corroborate the timing

and magnitude of Stine's drought-induced low stands and suggest centennial-average precipitation and river runoff in the central Sierra Nevada as low as 75% of the 20th century values during the medieval period [48].

Most relevant to our characterization of pre-development San Francisco Estuary hydrology are quantitative tree-ring reconstructions of annual discharge or runoff for the Sacramento and San Joaquin Rivers. Streamflow reconstructions from tree-rings are generally done by linear regression, in which a time series of unimpaired runoff is calibrated with time series from a network of indices of annual tree-ring width. Regression approaches, which can vary greatly from one study to another, are reviewed elsewhere [49,50]. Reconstructions for many basins in the western United States are available at https://www.treeflow.info/ (accessed on 9 July 2021). The first such reconstruction, which estimated flow in the Sacramento River at Bend Bridge (see Figure 1), utilized a network of 17 tree-ring chronologies that dated back to 1560. This reconstruction indicated that the wettest (1854–1916) and driest (1917–1950) periods overlapped with the historical period for which gaged flows are available in Earle [51].

The accuracy of Sacramento River runoff reconstructions over the past 500 years was improved by a network of blue oak (*Quercus douglasii*) chronologies whose collection began in the mid-1990s [52]. These blue oak chronologies, along with new collections of western Juniper (*Juniperus occidentalis*), were utilized with other tree-ring chronologies to reconstruct Sacramento River runoff back to 869; this work showed that the instrumented flow record was deficient in representing long duration (e.g., decadal and longer) droughts and wet periods [53]. A more recent effort reconstructed annual runoff for the Sacramento and San Joaquin Rivers and their major tributaries for the interval 900–2012 [20]. In contrast to Earle [51], these reconstructions indicated that, while the instrumented record does not reflect the extreme single-year Central Valley droughts, it does include multi-year droughts of similar magnitude to the most extreme droughts of the long-term record. The reconstructions further indicated exceptionally long multi-decadal swings between wet and dry conditions in the medieval period. More recent work applying the Sacramento River reconstruction [20] underscores the spatial extent of medieval drought: multi-basin coverage of hydrologic drought during the 1100s in the Sierra Nevada as well as the Colorado Rockies [54].
