Gregory T Pederson Connie A Woodhouse 2017 tabular digital data Reston, VA U.S. Geological Survey https://doi.org/10.5066/F79885ZT With increasing concerns about the impact of warming temperatures on water resources, more attention is being paid the relationship between runoff and precipitation, or runoff efficiency. Temperature is a key influence on Colorado River runoff efficiency, and warming temperatures are projected to reduce runoff efficiency. Here, we investigate the nature of runoff efficiency in the upper Colorado River (UCRB) basin over the past 400 years, with a specific focus on major droughts and pluvials, and to contextualize the instrumental period. We first verify the feasibility of reconstructing runoff efficiency from tree-ring data. The reconstruction is then used to evaluate variability in runoff efficiency over periods of high and low flow, and its correspondence to a reconstruction of late runoff season UCRB temperature variability. Results indicate that runoff efficiency has played a consistent role in modulating the relationship between precipitation and streamflow over past centuries, and that temperature has likely been the key control. While negative runoff efficiency is most common during dry periods, and positive runoff efficiency during wet years, there are some instances of positive runoff efficiency moderating the impact of precipitation deficits on streamflow. Compared to past centuries, the 20th century has experienced twice as many high flow years with negative runoff efficiency, likely due to warm temperatures. These results suggest warming temperatures will continue to impact runoff efficiency in wet or dry years, and that future flows will be less than anticipated from precipitation due to warming temperatures. Hydrologic and climataic reconstructions for the Upper Colorado River Basin. 12200901 20160901 ground condition Complete None planned -112.25 -104.125 43.75 33.0 Upper Colorado River Basin ISO 19115 Topic Category climatologyMeteorologyAtmosphere environment inlandWaters None Colorado River Tree-Rings Runoff Efficiency Temperature Reconstruction Precipitation Reconstruction USGS Thesaurus Hydrology Climatology USGS Metadata Identifier USGS:59b70011e4b08b1644ddf9a6 Geographic Names Information System Lees Ferry (historical) None. Please see 'Distribution Info' for details. None. Users are advised to read the data set's metadata thoroughly to understand appropriate use and data limitations. Gregory T Pederson U.S. Geological Survey Research Ecologist Mailing and Physical

2327 University Way, Suite 2

Bozeman MT 59715 USA 406-994-7390 406-994-6556 gpederson@usgs.gov University of Arizona USGS Biocomplexity Thesaurus vegetation None Pinus flexilis Pinus albicaulis Pinus edulis Pinus ponderosa Picea engelmannii Integrated Taxonomic Information System (ITIS) 2017 ONLINE_REFERENCE Washington, D.C. Integrated Taxonomic Information System (ITIS) http://itis.gov Gregory T Pederson U.S. Geological Survey Research Ecologist Mailing and Physical

2327 University Way, Suite 2

Bozeman MT 59715 USA 406-994-7390 406-994-6556 gpederson@usgs.gov keys and key characters Kingdom Plantae Subkingdom Viridiplantae Infrakingdom Streptophyta Superdivision Embryophyta Division Tracheophyta Subdivision Spermatophytina Class Pinopsida Subclass Pinidae Order Pinales Family Pinaceae Genus Pinus Species Pinus flexilis Species Pinus albicaulis Species Pinus edulis Species Pinus ponderosa Genus Picea Species Picea engelmannii No formal attribute accuracy tests were conducted. No formal logical accuracy tests were conducted. Data set is considered complete for the information presented, as described in the abstract. Users are advised to read the rest of the metadata record carefully for additional details. A formal accuracy assessment of the horizontal positional information in the data set has not been conducted. A formal accuracy assessment of the vertical positional information in the data set has either not been conducted, or is not applicable. The UCRB and surrounding areas contain a network of existing tree-ring chronologies from trees with growth that is moisture-limited and highly sensitively to hydroclimatic variability [e.g., Schulman 1956, Woodhouse and Lukas 2006]. These chronologies have been used to generate highly skillful reconstructions of streamflow [e.g., Stockton and Jacoby 1976, Hidalgo et al. 2000, Woodhouse et al. 2006, Meko et al. 2007]. Cool season moisture in the form of snowpack in the UCRB has also been reconstructed with a high degree of skill [Woodhouse 2003, Pederson et al. 2011]. The approach applied here, evaluating the difference between reconstructed water year streamflow and cool season precipitation to develop a reconstruction of runoff efficiency, relies on the ability of these trees to replicate both cool season precipitation and water year streamflow variability in the instrumental records. For streamflow, we used the reconstruction of the Colorado at Lees Ferry from Woodhouse et al. [2006] that covers years 1490-1997. The Lees B version of the streamflow reconstructions was selected because it is characterized by low-order autocorrelation similar to that in the naturalized flow record for the Colorado River at Lees Ferry (approximately r1 =0.25). The Lees B reconstruction model explains 84% of the variance in the gage record. Since no reconstruction of cool season precipitation exists, we generated a reconstruction using a process similar to that used for the flow reconstruction, but removing chronologies as potential predictors if they had been predictors in the Lees Ferry streamflow reconstruction model, and using the residual version of the chronologies to replicate the lack of autocorrelation in the October-April precipitation time series. The precipitation data used for calibrating the reconstruction are the PRISM [Daly et al. 2008] 4km resolution grid points for the upper Colorado River basin above Lees Ferry (Text S1, Figure S1). All other tree-ring chronologies available that ended in 1999 or later, and correlated significantly (p ≤ 0.05) with UCRB total October-April precipitation over the common interval, 1906-1999, for the full period and two halves of this period, were retained (25 chronologies). Stepwise regression yielded a model with four predictors, explaining 70% of the variance and spanned 1569-1999 (Figures 1 and S2). Model residuals met the assumptions of the multiple linear regression approach. Cross-validation of the model was performed using both a leave-one-out and a k-folds leave-10-out method replicated 100 times. Additional details on the tree-ring predictors and cool season precipitation reconstruction are provided in the supplemental information (Text S1, Table S1, Figure S2, and Data S1.). The correlation between the Lees Ferry streamflow and UCRB cool season precipitation reconstructions over the years 1906-1997 is r = 0.825 (p ≤ 0.05), while correlation between the two instrumental data records for the same period is r = 0.820 (p ≤ 0.05). Because runoff season temperature is a critical control on runoff efficiency (Woodhouse et al. 2016), we were interested in comparing reconstructions of runoff efficiency and temperature for the runoff season. As a close association is expected, a comparison could serve as an independent validation of the reconstruction-based runoff efficiency, though we acknowledge that temperature many not always be the dominant control on runoff efficiency. While temperature reconstructions for the general region [Salzer and Kipfmueller 2005], or as part of a gridded network do exist [e.g., Wahl and Smerdon 2012], we have elected not to use these for our comparison as they are not specifically for the UCRB (i.e. they are either too localized, or the gridded networks also integrate tree-ring information located far away from the basin), or are seasonal or annual measures of temperature that may not accurately reflect the important runoff season temperatures. To address these potential issues, we develop a new reconstruction of May-July average temperature (henceforth, ‘late runoff season’) for the UCRB using existing tree-ring data. Importantly, though the new temperature reconstruction is specific to the UCRB since it uses relatively local chronologies, the record still lacks information on the key spring months of March and April. So we interpret the reconstruction to be representative of the important late runoff season temperatures while noting the lack of sensitivity to early spring conditions. The independent reconstruction of UCRB May-July average temperature was produced as follows: Potential predictor chronologies for use in the reconstruction were selected by acquiring all publicly (ITRDB 2016, and PAGES2k Consortium 2017) available ring-width and maximum density records that fell within 100km of the UCRB (n = 23, Figure 1 and Data S1) that exhibited a positive and significant (p ≤ 0.1) correlation with observed May-July average temperature. The temperature data used for screening and calibrating the reconstruction are the PRISM [Daly et al. 2008] 4km resolution grid points for the upper Colorado River basin above Lees Ferry. The PRISM temperature data were compared against climate division data [nClimDiv dataset, version 2, see Vose et al. 2014] for the UCRB (Divisions: Colorado #2, Wyoming #3, and Utah #6 & 7) to ensure integrity of 20th century trends and variance (Figure S1). The final May-July average temperature reconstruction was produced by nesting successively longer, but statistically weaker, best-fit multiple linear regression models through time (see Figue S3 and Text S2 for more details). As with the precipitation reconstruction, models were cross-validated using both a leave-one-out and a k-folds leave-10-out method that was replicated 100 times. Reconstruction skill ranges from R2adj of 0.516 to 0.217, with cross-validation statistics from the leave-one-out and leave-10-out method equivalent showing the root mean squared validation error (RMSEv) ranging from 0.445 to 0.537 (Figure S3, Table S1). The nested reconstruction spans the years 1457-1993. Supporting data, statistics, and figures are provided in the supplemental section (Text S2, Figure 1, S1, S3, and Data S1). Text S1. UCRB October-April precipitation reconstruction details The chronologies in the initial pool for the reconstruction model included all available tree-ring chronologies from ponderosa pine (Pinus ponderosa), Douglas fir (Pseudotsuga menziesii), pinyon pine (Pinus edulis) from Colorado, southwestern Wyoming, and northeastern Utah (from International Tree Ring Databank and author Woodhouse’s collections), for the common years 1600-1999. The standardization and chronology development approach used follows Woodhouse et al. (2006): a negative exponential/straight line fit or a cubic spline two thirds the length of the series with a robust weighted mean to combine all series into a single site chronology (Cook et al., 1990). For this reconstruction, chronologies were prewhitened to remove low order autocorrelation. Chronologies were screened for significant (p < 0.05) correlations with UCRB total October-April precipitation over the common interval, 1906-1999, for the full period and two halves of this period. Calibration data used were from PRISM (Daley et al. 2008) for the Upper Colorado River basin. The gridded PRISM precipitation data have been found to be homogeneous across the Colorado River basin over the period tested (1916-2006) by Guentchev et al. (2010), in their comparison of three gridded datasets. A scatter plot of October-April total precipitation averaged for Colorado Division 2, Utah Divisions 6 and 7, and Wyoming Division 3 and PRISM data for the UCRB shows very close agreement among the two datasets (Figure S1). A total of 25 chronologies were retained for the regression model pool. Stepwise multiple linear regression was used to generate the reconstruction model, again following Woodhouse et al. 2006. Four chronologies entered into the regression equation explaining 70% of the variance (R2adj=0.691). The details of the predictor chronologies are in Data S1. The residuals met the assumptions of the multiple linear regression approach, which is supported by the Durbin-Watson statistic d = 2.183 and the lack of serial correlation (r = -0.100). The model was cross-validated using the Leave-One-Out and K-folds method in the “cvTools” package (Alfons 2012 ) (Figures S2, bottom panel, Table S1). In the K-folds cross-validation method, a random set of 10-sequential-observations were withheld from model fitting for use in prediction testing, with this function then replicated 100 times. Resulting cross-validation statistics provide a mean, median, and full distribution of prediction errors that are displayed in units of Root Mean Squared Error of the validation data (RMSEv) (Data S1). The observed and reconstructed October-May total precipitation time series are shown in Figure S2 (top and middle panels). Text S2. UCRB May-July temperature reconstruction details To produce a UCRB runoff season (May-July) average temperature reconstruction from proxies with local temperature sensitivity, we first obtained all maximum density and total ring-width tree-ring chronologies within 100 km from the basin that exhibit significant (p ≤ 0.10) and positive correlation (r > 0) to temperature (Figure 1). The tree-ring chronologies used here, are also a subset of the PAGES2k Consortium (2017) dataset of temperature sensitive proxies that were originally produced by Pederson and the PAGES North American 2k group for use in North American and Global temperature reconstructions. After obtaining the UCRB temperature sensitive tree-ring chronologies, we utilized chronology construction methods from Melvin and Briffa [2008] intended to retain low-frequency temperature variability most similar to the observational record, and more representative of the potential magnitude of UCRB temperature variability. In order to select a “best” reconstruction and bracket uncertainty in low-frequency temperature variability, we initially produced two independent sets of reconstructions using the variance stabilized “standard” and “signal-free” [see Melvin and Briffa 2008] versions of the tree-ring chronologies – methods described below. Of the two records, the best UCRB runoff season (May-July) average temperature reconstruction arose from the signal-free chronologies, and was selected due to the reconstructions superior calibration and cross-validation statistics (Table S1). For clarity, only the results for this reconstruction are presented. Chronologies in both the standard and signal-free version of the reconstructions were uniformly detrended in program “RCSsigFree45_v2b.app” [Cook et al. 2014] using an age-dependent spline required to retain a zero or negative slope with initial spine stiffness set to 50-years [Melvin et al. 2007]. This individual-series detrending method should preserve low-frequency climate signals, and positive monotonic growth trends (if present) in a similar fashion to using a negative exponential curve, but with some advantages over the juvenile growth years. Regional Curve Standardization was not used on this dataset since chronologies did not meet the assumptions and sample depth requirements for this method [Cook et al. 1990, Briffa and Melvin 2011]. Chronologies were constructed by calculating a bi-weight robust mean from the power-transformed residuals of the individually detrended ring-width series, which reduced potential trend distortion from end effects that can result from the ratio method of chronology calculation. To correct for artificial increases in variance due to decreasing sample depth through time, variance stabilization was performed using the Rbar method with correction of trend in variance provided by an age-dependent spline. All chronologies were truncated before entering into the reconstruction framework when sample size fell below 5 series. The final UCRB runoff season (May-July) average temperature reconstruction (Figure S3) was produced by nesting successively longer, but statistically weaker, best-fit multiple linear regression models through time. See Cook et al. [2004] and Pederson et al. [2011] for other examples of the nested reconstruction procedure. Regression models were constructed using a best subsets extensive search criteria within the “leaps” package [Lumley and Miller, 2009], with additional standard dendrochronological analyses and time series functions performed package “dplR” [Bunn 2008, Bunn 2010, and Bunn et al. 2015] in the analytical program R (R core team 2013). Selection of the final suite of “best” nested models was based on the Mallow's CP statistic, adjusted R2 values, and results from cross-validation. Models were cross-validated using the Leave-One-Out and K-folds method in the “cvTools” package (Alfons 2012 ), as described for the precipitation reconstruction above. Metadata for the predictor chronologies are in DataS1. The successively longer (but weaker) reconstructions were mean and variance scaled to match the mean and variance of the observed temperature record, then nested forward and backward in time to generate a single long “standard” and “signal-free” version of the reconstruction spanning 1457-1993 CE. This approach corrected for artificial variance reduction from regression that arises due to the reduced quality of the longer nested models. Reconstruction calibration and validation statistics are provided as a benchmark of model quality through time in Figure S3 and Table S1, and with the reconstruction data (Data S1). 20170804 The proxy record for runoff efficiency is based on the difference between the reconstructions of streamflow and cool season precipitation, and thus it is important to assess not only the skill of the individual reconstructions, but the relationships between these two variables. The two reconstructions were standardized and cool season precipitation was subtracted from water year flow to generate a difference series representing runoff efficiency. 20170804 Multi-century reconstructions of temperature, precipitation, and runoff efficiency for the Upper Colorado River Basin.csv Comma Separate Value (CSV) file containing data. Producer defined Year Water Year Producer defined 1220 2016 Obs.Oct-Apr.total.PPT(mm).PRISM Observed Prior October-April total precipitation for the Upper Colorado River Basin from PRISM data Producer defined 82.109949 310.351186 mm NA No measurement was recorded. Producer defined Obs.Oct-Apr.total.PPT(mm).nClimDiv Observed Prior October-April total precipitation for the Upper Colorado River Basin from climate division data Producer defined 82.7405 303.5935 mm NA No measurement was recorded. Producer defined Obs.Lees.WaterYr.flow(BCM) Observed Total Water Year Flows for the Upper Colorado River Basin as measured at Lee's Ferry Producer defined 6.682399353 29.82703333 Billion Cubic Meters NA No measurement was recorded. Producer defined Obs.May-Jul.Tave.(oC).nClimDiv Observed May-July average temperature for the Upper Colorado Basin from climate division Data Producer defined 13.73 18.16 Degrees Celcius NA No measurement was recorded. Producer defined Obs.MayJuly.Tave.(oC)_PRISM Observed May-July average temperature for the Upper Colorado Basin from PRISM Data Producer defined 13.4457 17.7052 Degrees Celcius NA No measurement was recorded. Producer defined Recon.LeesB.Flow.(MAF) Observed Total Water Year Flows for the Upper Colorado River Basin as measured at Lee's Ferry. Lee's B Reconstruction from Woodhouse et al. 2006 Producer defined 2106.06662 31358.7942 Million Acre Feet NA No measurement was recorded. Producer defined Recon.LeesB.Flow.(std) Observed Total Water Year Flows for the Upper Colorado River Basin as measured at Lee's Ferry. Lee's B Reconstruction from Woodhouse et al. 2006 Producer defined -2.80812243 2.35659336 Standard Deviation NA No measurement was recorded. Producer defined Recon.Oct-Apr.PPT.(std) Stardardized (z-scores, subtract mean and divide by standard deviation) Versions of the Oct-April total Precipitation Reconstruction Producer defined -3.0037375 3.02447534 standard deviations NA No measurement was recorded. Producer defined Recon.Runooff.Efficiency.(Flow-O-APPT.std) Reconstructed Runoff Efficiency derived from subtracting standardized Oct-Apr precipitation reconstruction from the Lees B streamflow reconstruction Producer defined -1.77924939 1.94097695 standard deviations NA No measurement was recorded. Producer defined Oct-Apr.PPT(Recon) Oct-April total precipitation (mm) for the Upper Colorado River Basin calculated from 4km PRISM data for all grid point located within the UCRB above the Lee's ferry gage Producer defined 89.95 332.08 millimeters NA No measurement was recorded. Producer defined L90CI_PPT Lower 90% Confidence Interval (annual values) for the Oct-April total precipitation reconstruction. Producer defined 70.68 317.52 millimeters NA No measurement was recorded. Producer defined U90CI_PPT Upper 90% Confidence Interval (annual values) for the Oct-April total precipitation reconstruction Producer defined 109.21 346.63 millimeters NA No measurement was recorded. Producer defined PPTrecon.20yrSpl Oct-April total precipitation reconstruction smoothed with a 20-year cubic smoothing spline Producer defined 185.63 237.58 millimeters NA No measurement was recorded. Producer defined L90CI.20yrSpl_PPT Lower 90% Confidence Interval smoothed with a 20-year cubic smoothing spline for the Oct-April total precipitation reconstruction. Producer defined 175.34 229.88 millimeters NA No measurement was recorded. Producer defined U90CI.20yrSpl_PPT Upper 90% Confidence Interval smoothed with a 20-year cubic smoothing spline for the Oct-April total precipitation reconstruction. Producer defined 195.92 245.41 millimeters NA No measurement was recorded. Producer defined R2_PPT R2 value or coefficent of variation of reconstruction Producer defined 0.7042 0.7042 NA No measurement was recorded. Producer defined R2adj_PPT Adjusted R2 value or the adjusted coefficent of variation of the reconstruction adjusted for degree's of freedom. Producer defined 0.6909 0.6909 NA No measurement was recorded. Producer defined LOOCV_PPT Leave-One-Out Cross Validation CV statistic for the reconstructoin shown in units of Root Mean Squared Error of Validation (RMSEv) Producer defined 24.35 24.35 Root Mean Squared Error NA No measurement was recorded. Producer defined k.10.Min_PPT Minimum CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 23.9 23.9 NA No measurement was recorded. Producer defined k.10.1Quar_PPT Lower Quartile CV (RMSEv) value from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 24.24 24.24 NA No measurement was recorded. Producer defined k.10.Median_PPT Median CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 24.45 24.45 NA No measurement was recorded. Producer defined k.10.Mean_PPT Mean CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 24.44 24.44 NA No measurement was recorded. Producer defined k.10.3Quar_PPT Upper Quartile CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 24.59 24.59 NA No measurement was recorded. Producer defined k.10.Max_PPT Maximum CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 25.13 25.13 NA No measurement was recorded. Producer defined MCP_PPTcrn Tree-ring chronologies in the Oct-April total precipitation reconstruction Producer defined 0.12 1.885 NA No measurement was recorded. Producer defined RIF_PPTcrn Tree-ring chronologies in the Oct-April total precipitation reconstruction Producer defined 0.095 1.916 NA No measurement was recorded. Producer defined GOU_PPTcrn Tree-ring chronologies in the Oct-April total precipitation reconstruction Producer defined 0.0 1.747 NA No measurement was recorded. Producer defined EFU_PPTcrn Tree-ring chronologies in the Oct-April total precipitation reconstruction Producer defined 0.0 2.423 NA No measurement was recorded. Producer defined Recon.May-Jul.Tave.(oC) Nested May-July average temperature (oC) reconstruction for the Upper Colorado River Basin (Annual Values). Variance of each nested reconstruciton was scaled to match that of the observational record, allowing for direct comparision to the observational period and removing aritifical variance reduction due to regression and the nesting of models with reduced skill. Producer defined 12.43409271 16.97402343 Degrees Celcius NA No measurement was recorded. Producer defined recon.20yrSpl_Tave May-July average temperature reconstruction smoothed with a 20-year cubic smoothing spline Producer defined 13.62671869 16.17207765 Degrees Celcius NA No measurement was recorded. Producer defined recon.50yrSpl_Tave May-July average temperature reconstruction smoothed with a 50-year cubic smoothing spline Producer defined 13.72479557 16.16197329 Degrees Celcius NA No measurement was recorded. Producer defined L90CI_Tave Lower 90% Confidence Interval (annual values) for the May-July average temperature reconstruction. Producer defined 11.69316444 16.57479711 Degrees Celcius NA No measurement was recorded. Producer defined L90CI.20yrSpl_Tave Lower 90% Confidence Interval smoothed with a 20-year cubic smoothing spline for the May-July average temperature reconstruction. Producer defined 12.97503566 15.85238545 Degrees Celcius NA No measurement was recorded. Producer defined L90CI.50yrSpl_Tave Lower 90% Confidence Interval smoothed with a 50-year cubic smoothing spline for the May-July average temperature reconstruction. Producer defined 13.10619484 15.87914275 Degrees Celcius NA No measurement was recorded. Producer defined U90CI_Tave Upper 90% Confidence Interval (annual values) for the May-July average temperature reconstruction. Producer defined 13.17502098 17.37324974 Degrees Celcius NA No measurement was recorded. Producer defined U90CI.20yrSpl_Tave Upper 90% Confidence Interval smoothed with a 20-year cubic smoothing spline for the May-July average temperature reconstruction. Producer defined 14.27840172 16.49176984 Degrees Celcius NA No measurement was recorded. Producer defined U90CI.50yrSpl_Tave Upper 90% Confidence Interval smoothed with a 50-year cubic smoothing spline for the May-July average temperature reconstruction. Producer defined 14.3433963 16.44480382 Degrees Celcius NA No measurement was recorded. Producer defined R2_Tave R2 value or coefficent of variation of each nested reconstruction Producer defined 0.282 0.5669 NA No measurement was recorded. Producer defined R2adj_Tave Adjusted R2 value or the adjusted coefficent of variation of each nested reconstruction adjusted for differing degree's of freedom arising from the changing number of predictor chronologies in each regression. Producer defined 0.2651 0.516 NA No measurement was recorded. Producer defined LOOCV_Tave Leave-One-Out Cross Validation CV statistic for the reconstructoin shown in units of Root Mean Squared Error of Validation (RMSEv) Producer defined 0.5815 0.6406 NA No measurement was recorded. Producer defined k.10.Min_Tave Minimum CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.5648 0.6343 NA No measurement was recorded. Producer defined k.10.1Quar_Tave Lower Quartile CV (RMSEv) value from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.5768 0.6401 NA No measurement was recorded. Producer defined k.10.Median_Tave Median CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.5832 0.6434 NA No measurement was recorded. Producer defined k.10.Mean_Tave Mean CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.584 0.6434 NA No measurement was recorded. Producer defined k.10.3Quar_Tave Upper Quartile CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.5893 0.6484 NA No measurement was recorded. Producer defined k.10.Max_Tave Maximum CV value (RMSEv) from the K-Folds leave 10 (sequential) values out cross-validation replicated 100 times with a different random 10 sequential observations witheld. Producer defined 0.6221 0.6593 NA No measurement was recorded. Producer defined az553x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.771 1.119 NA No measurement was recorded. Producer defined co517x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.802 1.108 NA No measurement was recorded. Producer defined co545_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.503 1.564 NA No measurement was recorded. Producer defined co552_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.606 1.336 NA No measurement was recorded. Producer defined co552x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.794 1.134 NA No measurement was recorded. Producer defined co553x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.845 1.134 NA No measurement was recorded. Producer defined co554_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.571 1.477 NA No measurement was recorded. Producer defined co554x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.772 1.169 NA No measurement was recorded. Producer defined co574x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.821 1.092 NA No measurement was recorded. Producer defined co586_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.594 1.312 NA No measurement was recorded. Producer defined co586x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.741 1.184 NA No measurement was recorded. Producer defined co633_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.638 1.572 NA No measurement was recorded. Producer defined FRE_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.49 1.434 NA No measurement was recorded. Producer defined nm529x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.768 1.084 NA No measurement was recorded. Producer defined ut510x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.889 1.096 NA No measurement was recorded. Producer defined ut512_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.633 1.489 NA No measurement was recorded. Producer defined ut512x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.867 1.157 NA No measurement was recorded. Producer defined wy022_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.644 1.419 NA No measurement was recorded. Producer defined wy022x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.734 1.118 NA No measurement was recorded. Producer defined wy025_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.451 1.349 NA No measurement was recorded. Producer defined wy025x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.609 1.116 NA No measurement was recorded. Producer defined wy030_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.368 1.476 NA No measurement was recorded. Producer defined wy030x_Tave.crn A positively correlated May-July temperature sensitive tree-ring chronology used in the temperature reconstruction. Signal-Free chronologies (detrended with an age-dependent spline with variance stabilization) used in the nested reconstruction framework. All chronologies were truncated when sample depths fell below 5 ring-width series. Producer defined 0.748 1.156 NA No measurement was recorded. Producer defined U.S. Geological Survey - ScienceBase mailing address

Denver Federal Center, Building 810, Mail Stop 302

Denver CO 80225 USA 1-888-275-8747 sciencebase@usgs.gov Unless otherwise stated, all data, metadata and related materials are considered to satisfy the quality standards relative to the purpose for which the data were collected. Although these data and associated metadata have been reviewed for accuracy and completeness and approved for release by the U.S. Geological Survey (USGS), no warranty expressed or implied is made regarding the display or utility of the data on any other system or for general or scientific purposes, nor shall the act of distribution constitute any such warranty. Digital Data https://doi.org/10.5066/F79885ZT None 20200826 Gregory T Pederson U.S. Geological Survey Research Ecologist Mailing and Physical

2327 University Way, Suite 2

Bozeman MT 59715 USA 406-994-7390 406-994-6556 gpederson@usgs.gov FGDC Biological Data Profile of the CDGSM FGDC-STD-001.1-1999 Record created using USGS Metadata Wizard tool. (https://github.com/usgs/fort-pymdwizard)