Reconstructions of river discharge and other hydrologic variables often exploit large available networks of tree-ring chronologies from multiple species and hydrologic settings. A common early step in such studies is screening to reduce the predictor data set and focus on chronologies with a strong hydrologic signal. A stepwise regression approach to screening is proposed and illustrated for reconstruction of April 1 snow-water equivalent (SWE) at three snow courses in the northern Sierra Nevada and Lake Tahoe region from a multi-species tree-ring network. SWE is regressed separately on each chronology lagged t-2 to t+2 years from the year of SWE. A chronology is accepted based on specified criteria for temporal stability of signal and skill of the lagged model in predicting SWE outside the calibration space. A cross-validation stepwise cutoff rule is applied to guard against over-fitting the lagged model. Illustration for a network of 23 chronologies of five snow-adapted species (Juniperus occidentalis, Pinus jeffreyi, Pinus ponderosa, Abies magnifica, and Tsuga mertensiana) underscores the critical importance of lags in the tree-ring response to SWE. For Abies and Tsuga, in particular, chronologies passing screening are characterized by lagged models with a positive coefficient on the year following the hydrologic anomaly (deep snowpack this year, wide ring the following year) and no dependence or a negative coefficient on the current year (deep snowpack, narrow current ring). The SWE signal is strongest for one particular Juniperus chronology whose regression explains 39% of the SWE variance at two snow courses. The strength of SWE signal varies greatly over sites within species. More than half of the Juniperus and Pinus chronologies were rejected by screening because of either weak or temporally unstable signal. A repeat of the regression screening using water-year precipitation (Global Historical Climate Network) instead of SWE suggests that different subsets of chronologies are optimal depending on the target hydologic variable. Few chronologies have a significant signal for the residual of SWE regressed on water-year precipitation. This suggests little snow-specific information on the moisture signal in annual ring width. Sub-annual ring measurements and quantitative wood anatomy are suggested as possible ways to help discriminate the rain and snow signals in tree rings from the region.

Year-to-year variability of precipitation and temperature has significant consequences for water management decision making. “Whiplash” is a term which describes this variability at its most severe, referring to events at various timescales in which the hydroclimate switches between extremes. Tree-rings in semi-arid environments like the Truckee-Carson River Basin (California/Nevada watersheds with headwaters in the Sierra Nevada) can provide proxy records of hydroclimate as their annual growth is tied directly to limitations in water-year rainfall and temperature, but traditional metrics of reporting explained variance do not distinguish a reconstruction’s sensitivity to whiplash events. In this study, a pool of total ring width indices from five snow-adapted conifer species (Abies magnifica, Juniperus occidentalis, Pinus ponderosa, Pinus jeffreyi, Tsuga mertensiana) were used to develop a series of standardized reconstructions of water-year PRISM precipitation (P12) using stepwise linear regression. A nonparametric analysis approach was then used to determine positive and negative whiplash events in reconstructed and instrumental precipitation records. Hypergeometric distribution of the resulting timeseries datasets illustrates relationships between reconstructions and recorded whiplash events and allows for determination of patterns in tree-ring growth response. The results of this study suggest that ring-width indices from the assessed conifer species in the snow-belt of the Sierra Nevada are often able to record consecutive years of opposing extreme precipitation and report such events through derived models. Negative WL events are tracked more consistently across species in site-specific reconstructions of P12 than positive ones. It appears that residual effects of a preceding year’s drought or pluvial do not necessarily suppress records of WL, though sensitivity to precursor conditions in tracking of WL events may differ across species, and the absolute WL events captured in a reconstruction vary.