Volume 49, Issue 19 e2022GL099582
Research Letter
Free Access

The Unprecedented Character of California's 20th Century Enhanced Hydroclimatic Variability in a 600-Year Context

Diana Zamora-Reyes

Corresponding Author

Diana Zamora-Reyes

Department of Hydrology and Atmospheric Sciences, University of Arizona, Tucson, AZ, USA

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Correspondence to:

D. Zamora-Reyes,

[email protected]

Contribution: Conceptualization, Methodology, Formal analysis, ​Investigation, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
Ellie Broadman

Ellie Broadman

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Contribution: Formal analysis, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
Erica Bigio

Erica Bigio

Department of Natural Resources and Environmental Science, University of Nevada, Reno, NV, USA

Contribution: Resources, Data curation, Writing - review & editing

Search for more papers by this author
Bryan Black

Bryan Black

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Contribution: Conceptualization, Methodology, ​Investigation, Writing - review & editing

Search for more papers by this author
David Meko

David Meko

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Contribution: Resources, Data curation, Writing - review & editing

Search for more papers by this author
Connie A. Woodhouse

Connie A. Woodhouse

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Contribution: Resources, Data curation, Writing - review & editing

Search for more papers by this author
Valerie Trouet

Valerie Trouet

Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ, USA

Contribution: Conceptualization, Methodology, ​Investigation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Funding acquisition

Search for more papers by this author
First published: 01 October 2022
Citations: 2

Abstract

Recently, year-to-year swings in California winter precipitation extremes have resulted in drought, wildfires, and floods causing billions of dollars in damage. These recent precipitation swings represent an increasing trend in variability of California's hydroclimate over the past decades. Here, we put this trend in a longer-term context using tree-ring-based precipitation, streamflow, and snow water equivalent reconstructions. We show that the statewide rise in hydroclimate variability in the 20th century is driven by an increasing trend in the magnitude of wet extremes. A prior period of strong variability in the 16th century, in contrast, is related to an increasing trend in the magnitude of dry extremes. Our results are consistent with climate model simulations that suggest an increasingly volatile future for California's hydroclimate and highlight the importance of collaboration between scientists and water resource managers to incorporate this increased variability into their decision-making and planning, acknowledging higher risks for compound events.

Key Points

  • California's hydroclimate shows enhanced variability caused by increasingly wet extremes during the 20th century

  • Tree-ring based hydroclimate reconstructions suggest that this statewide configuration has not been seen in the past 600 years

  • The 16th century shows enhanced hydroclimate variability driven by an increase in the magnitude of dry extremes, rather than wet extremes

Plain Language Summary

California's lack, as well as surplus, of rain has been in the news recently. Due to its Mediterranean climate, the state receives most of its rainfall from November to March and the amount varies from year to year. However, this variability has been increasing over the past decades in both rainfall and river flow. This study aims to place this 20th century increasing variability in a longer-term perspective using tree-ring records. We show that this statewide positive trend in variability of California's rainfall, river flow, and snow during the 20th century is caused by an increase in the magnitude of very wet years, which is a pattern that has not been seen in the past 600 years. We also show that the 16th century is another period with statewide increased variability that is caused by dry years getting drier. These results enhance our understanding of the 20th century rise in variability, which is expected to continue in the future, and highlights the need for scientists to work with local, state, and federal agencies to address this climatic variability in their decision-making process.

1 Introduction

Large year-to-year precipitation variability is a unique feature of California's Mediterranean climate, where winters (November–March) are wet, summers (May–August) are dry, and total annual rainfall ranges from 50% to 200% of the climatological average (Dettinger et al., 2011). Over the past decade, some of California's wet seasons have been unusually dry (e.g., 2012–2015 and 2020–2021) and, combined with unusually warm dry season conditions, have triggered an unprecedented and on-going multi-year drought (Belmecheri et al., 2016; Griffin & Anchukaitis, 2014). Meanwhile, other winter seasons (e.g., 2016–2017 and 2018–2019) have been unusually wet, with atmospheric river storms repeatedly steered toward California (CADWR, 20182020; Wang et al., 2017) by the Pacific jet stream in the upper atmosphere (Belmecheri et al., 2017; Wahl et al., 2019) causing precipitation networks across the state to record more than 150% of the climatological annual average (CADWR, 2020). These wet winters can alleviate the state's strained water supply (Dettinger et al., 2011), but have also led to catastrophic flooding (Hamilton et al., 2017; Michaelis et al., 2022; White et al., 2019). For example, in 2019 the city of Guerneville in Sonoma County was only accessible by boat as the Russian River covered roads in meters of water (Prodis Sulek, 2019). Another consequence of the abundant precipitation is fine fuel buildup, including grasses in California's chaparral ecosystems (Swain, 2021). When combined with katabatic winds, such as the Santa Ana winds (Kolden & Abatzoglou, 2018), such fuels have led to catastrophic fires (Thomas fire-2017, Camp fire-2019, August Complex-2020, and Dixie fire-2021) that have cost billions in damages and burned more than 2.5 million acres (Abatzoglou et al., 2018; CALFIRE, 2022; Swain, 2021; Williams et al., 2019).

Previous research on California's variable hydroclimate has been primarily focused on future changes in precipitation and streamflow (Berg & Hall, 2015; Gershunov et al., 2019; Mallakpour et al., 2018; Swain et al., 2018; Yoon et al., 2015) and few studies have analyzed observation-based variability (Black et al., 2018; Granger, 1979; Marston & Ellis, 2018; Pagano & Garen, 2005; Zamora-Reyes et al., 2022) or longer-term proxy-based (Black et al., 2014; Haston & Michaelsen, 1997) changes to California's hydroclimatic variability. Instrumental records, as well as future projections, generally report no significant trends in either mean annual or aggregated seasonal precipitation and streamflow (Berg & Hall, 2015; Black et al., 2018; Granger, 1979; Mallakpour et al., 2018; Pagano & Garen, 2005; Pratt & Mooney, 2013; Swain et al., 2018; Yoon et al., 2015; Zamora-Reyes et al., 2022). However, an increasing trend in the variability of aggregated seasonal precipitation and streamflow has been demonstrated from the mid-20th century onwards (Zamora-Reyes et al., 2022). Based on a network of instrumental station data across California, Zamora-Reyes et al. (2022) found that this recent increasing trend can be primarily attributed to an increase in the magnitude of extreme precipitation events during wet winter months, which propagates to winter, spring, and summer streamflow. That study, however, was based on less than 80 years of data and was restricted in its capacity to: (a) place the current hydroclimatic trend in a longer-term, pre-industrial context; and (b) understand whether and how the current spatial patterns of hydroclimatic extremes differ from those of the past. Most recent tree-ring-based California hydroclimate reconstructions have concentrated on megadroughts over the past millennia (E. K. Wise, 2016; Woodhouse et al., 2020) or on placing current dry periods in a long-term context (Biondi & Meko, 2019; Griffin & Anchukaitis, 2014; Touchan et al., 2021; Williams et al., 2020). Here, we use previously published tree-ring based precipitation, streamflow, and snow water equivalent (SWE) reconstructions across California to:
  1. Place the increased hydroclimate variability of the 20th and 21st century in California into the context of the past 600 years;

  2. Investigate past periods of enhanced hydroclimatic variability;

  3. Discern whether these past periods of enhanced variability are driven by wet or dry extremes, or both; and

  4. Determine the spatial extent of recent and past periods of enhanced hydroclimate variability across the Western United States (US).

2 Data and Methods

We compiled California precipitation, streamflow, and SWE tree-ring reconstructions and their associated instrumental data from two websites: TreeFlow (http://www.treeflow.info/California) and the National Oceanic and Atmospheric Administration Paleoclimate Database (https://www.ncei.noaa.gov/access/paleo-search/study/32352). We used six precipitation reconstructions, 13 streamflow reconstructions, and one SWE reconstruction that span most of California's latitudinal gradient (Figure S1; Table S1 in Supporting Information S1). Precipitation and streamflow reconstructions targeted aggregated water year (WY; 1 October to 30 September) totals, whereas the SWE reconstruction targeted the May 1st amount. The reconstructions vary in length (Table S1 in Supporting Information S1), but we focus on the period 1410–2010, which is covered by 75% of the records. Most of these reconstruction studies were funded by the California Department of Water Resources (CADWR) to help inform water managers about long-term conditions in major watersheds across the southwestern US, including the Sacramento, San Joaquin, and Colorado Rivers. These hydroclimatic reconstructions use the largest number of tree-ring chronologies in the area and explain between 54% and 78% of WY streamflow and precipitation variance, while the SWE reconstruction explains 55% of the SWE variance (Table S1 in Supporting Information S1).

We recognize that tree-ring reconstructions are imperfect recorders of WY hydrologic variables, especially during extremely wet years (Dannenberg et al., 2019), but 19 of our analyzed reconstructions include blue oak chronologies, whose records capture more than 80% of variance in instrumental California precipitation (Meko et al., 2011; Stahle et al., 2013). Moreover, tree-ring sites in regions of western North America located south of 45°N latitude typically capture both low and high winter precipitation extremes (E. K. Wise and Dannenberg, 2019).

We further acknowledge that the California hydroclimate reconstructions we used are based on a large, shared pool of tree-ring chronologies. To minimize overlap of chronologies contributing to reconstructions, we divided California into three regions: northern California (nCA), southern California (sCA), and the southern Sierra Nevada, and compiled regional reconstructions for each (Figure S1 and Table S1 in Supporting Information S1). We thus balanced the number of reliable reconstructions by placing them into geographically coherent hydroclimatic regions that covary (Abatzoglou et al., 2009) and have similar seasonal contributions to precipitation and streamflow (Zamora-Reyes et al., 2022). Within nCA and sCA, precipitation and streamflow reconstructions depend on some of the same tree-ring chronologies, and are strongly correlated with each other by nature (Meko et al., 20142018). Instrumental precipitation and streamflow data for these regions show similarly strong correlations (Belmecheri et al., 2016; Zamora-Reyes et al., 2022). However, each region is nearly independent from the other regions with a low number of shared chronologies among the three regions: 75%–91% of the chronologies in each region are unique to that region (Table S2 in Supporting Information S1).

We standardized all reconstructions to address their skewed distributions by following a procedure similar to the Standardized Precipitation Index (SPI), where we first fit a gamma distribution to all values and then transform each record to a normal distribution (Malevich et al., 2013; Mckee et al., 1993). Following this procedure, all the standardized series showed a normal distribution, as tested using Kolmogorov-Smirnov goodness-of-fit tests (p < 0.05). We also developed regional precipitation and streamflow time series in nCA and sCA by averaging all individual standardized reconstructions per variable and per region.

We analyzed long-term changes to the variability of the hydroclimatic reconstruction distributions for each individual (Figure S2 and S3 in Supporting Information S1) and regional (Figures 1-3) time series using 31-year running windows for three calculated metrics: (a) standard deviation, (b) 10th percentile, and (c) 90th percentile (Zamora-Reyes et al., 2022). Standard deviation is an indicator of variability, whereas the 10th and 90th percentile represent dry and wet extremes, respectively. We opted to use 31-year running windows for these metrics to follow the methodology used in previously published studies focused on long-term climate variability (Li et al., 2013; Liu et al., 2017; Trouet et al., 2018; Zamora-Reyes et al., 2022). To evaluate significant trends in the time series of these metrics, we applied two-tailed non-parametric Mann–Kendall trend (MKT) tests at a 95% significance level to 50-year running periods (lagged by one year) in each of the resulting time series. As such, for our reconstructions (1410–2010 CE), these metrics (standard deviation, 10th percentile, and 90th percentile) were plotted on the central year of the 31-year running periods (1426–1996), and 50-year trends were calculated for each year of these metric time series (1426–1475, 1427–1476, etc.). We then used the Sen's slope estimator to calculate the magnitude of these trends in each series. We also performed this analysis on an overlapping subset of instrumental and reconstructed records from 1940 to 2003 (Figures S4 and S5 in Supporting Information S1). The analysis of 50-year trends accommodates both the limited length of instrumental time series and our focus on decadal-scale variability trends.

Details are in the caption following the image

Time series and histograms for an analysis of running standard deviation, calculated using 31-year windows, over the period 1410 to 2010, for reconstructed hydroclimate data in three regions: northern California (a and d), the southern Sierra Nevada (b and e), and southern California (c and f). For northern California (a and d) and southern California (c and f), analyses were performed on streamflow (dark gray lines) and precipitation (light gray lines) data; for the southern Sierra Nevada (b and e), analyses were performed on snow water equivalent data (black lines). Panels a–c show the calculated 31-year running standard deviation time series. The x-axis in (a–c) (time) refers to the mid-year of each 31-year window. Panels (d–f) show normalized histograms of all 50-year trends from 1410 to 2010. In (d–f), vertical lines indicate the average 90% confidence level for each region, derived from resampled data. Periods of reconstructed increased hydroclimatic variability discussed in the text are highlighted with shading (a–c) or vertical lines (d–f) in magenta (1539–1588) and green (1938–1987). Note: magenta and green vertical lines in f overlap.

Details are in the caption following the image

As in Figure 1, but for 31-year running windows of 90th percentile values.

Details are in the caption following the image

As in Figure 1, but for 31-year running windows of 10th percentile values.

We identified two periods when significant (p < 0.05) 50-year trends of the same sign occurred at the same time in all three regions (hereafter, “synchronous periods”). To further investigate the significance of trends within these synchronous periods, we first determined 50-year periods with the steepest average slope for hydroclimate variability over the three regions (Table S3 in Supporting Information S1). Then, we built an empirical distribution of slope values for each region, variable, and metric. To do so, we resampled each series randomly without replacement using the same number of observations and recalculated the Sen's slope. We repeated this procedure 10,000 times and tabulated values. Synchronous periods when regional MKT had Sen's slope values ≤10th or ≥90th percentile from the empirical distribution were considered to have passed an additional test of significance (Burn & Hag Elnur, 2002).

3 Results

Standard deviations started to increase in the 1950s in sCA and the southern Sierra Nevada (Figures 1b and 1c) and in the early 1900s in nCA (Figure 1a). Across all regions, the 50-year period with the strongest trend, defined as the strongest average slope, is 1938–1987 (green area in Figures 1a–1c; Table S3 in Supporting Information S1), representing running standard deviations over the period 1923–2002. This recent trend was significantly stronger (p < 0.1) in all three regions than trends derived from resampled data (Figures 1d–1f). Precipitation and streamflow standard deviation reached a 600-year maximum in the late 20th century in nCA and sCA (Figures 1a and 1c). In the southern Sierra Nevada, SWE standard deviation peaked in 1973, but this peak was not unprecedented (Figure 1b).

The twentieth century increase in variability is primarily related to significant increases in the amplitude of wet extremes, as represented by the 90th percentile time series (Figure 2; Table S4 in Supporting Information S1), rather than increases in the amplitude of dry extremes, as represented by the 10th percentile time series (Figure 3; Table S5 in Supporting Information S1). The increase in wet extremes occurs across all regions and variables, and is unprecedented over the past 600 years for precipitation and streamflow in both nCA and sCA (Figures 2a and 2c). This is consistent with strong increasing trends in running standard deviation (Figures 1a and 1c). nCA streamflow is the only regional metric that shows a significant (p < 0.05) decreasing trend in the amplitude of dry extremes (Table S5 in Supporting Information S1) during the 20th century. In contrast, we find that the 90th percentile slope for 1938–1987 is significant (p < 0.1) in all three regions and variables compared to resampled data (Figures 2d–2f). We also found similar trends for each of the individual reconstructions (Figure S2 and S3 in Supporting Information S1). Moreover, the subset (1940–2003) of individual instrumental and reconstructed records (Figure S4 and S5 in Supporting Information S1) shows trends similar in direction and magnitude to those in the regional reconstructions (Table S6 in Supporting Information S1).

Another significant (p < 0.05) trend in California's hydroclimatic variability occurs in the 16th century and is synchronous across all three regions (Figure 1). The strongest 50-year trend across all regions and metrics occurred in 1539–1588 (magenta in Figures 1a–1c), representing variability over the period 1524–1603. This 16th century trend rivals the late 20th century trends in nCA and sCA and is stronger than the 20th century trend in southern Sierra Nevada SWE (Figures 1d–1f; Table S3 in Supporting Information S1). Unlike the 20th century trend, however, the 16th century trend stems primarily from a statewide, steep increase in the magnitude of dry extremes (10th percentile; Figure 3; Table S5 in Supporting Information S1), as opposed to the increase in wet extremes during the 20th century (90th percentile; Figure 2; Table S4 in Supporting Information S1). Only southern Sierra Nevada SWE shows increasing wet and dry extremes during the 16th century (Figures 2b and 3b). A comparison with resampled data, however, shows that the 16th century dry extremes for the southern Sierra Nevada are much more unusual than the wet extremes (Figures 3e vs. Figure 2e). There are two additional periods with increased hydroclimatic variability in at least one of the regions (1690–1739 and 1743–1792), but neither is synchronous statewide (Figure 1) or shows a distinct (wet or dry) driver (Figures 2 and 3). To demonstrate the robustness of the synchronous periods among these three regions, we combined the nCA and sCA hydroclimate (streamflow and precipitation) records into statewide records and performed standard deviation, 90th percentile, and 10th percentile trend analyses (Figure S6 in Supporting Information S1). These analyses confirm that the 16th century and 20th century contain the strongest statewide increases in variability over the past 600 years that are driven by increased magnitude in dry and wet extremes, respectively.

To assess the spatial extent of past (1539–1588) and recent (1938–1987) periods of significant (p < 0.1) increased hydroclimatic variability, we performed 31-year running metrics and 50-year trends using the North American Drought Atlas (NADA; Cook et al., 2010; 0–2005 CE), a 0.5° by 0.5° gridded network of tree-ring based reconstructions of June-August self-calibrating Palmer Drought Severity Index (sc-PDSI, or PDSI calculated with empirical constants replaced to reflect local climate (Wells et al., 2004)). We also performed this analysis for 1938–1987 using June-August sc-PDSI gridded data from the Climate Research Unit (CRU; 0.5° by 0.5°; 1901–2020; Barichivich et al., 2021) Mapping our three variability metrics for the NADA and CRU sc-PDSI datasets across the Western US reveals that significant (p < 0.05) standard deviation trends for the 16th and 20th century are generally higher in California than in other regions (Figures 4a, 4d and 4g). Spatial patterns in the 10th and 90th percentiles confirm that the peak in 20th century variability is primarily driven by increased magnitude of wet extremes (Figures 4b, 4e and 4h), whereas the 16th century peak stems primarily from an increase in dry extremes (Figures 4c, 4f and 4i). The increase in magnitude of the late 20th century sc-PDSI-inferred wet extremes is not limited to California, but is widespread throughout the Western US (Figures 4e and 4h). Meanwhile, the 16th century increase in magnitude of dry extremes is limited to south-central California, Nevada, Utah, and Arizona (Figure 4c).

Details are in the caption following the image

Sen's slope (multiplied by 10) calculated for 31-year running standard deviation (a, d, g), 90th percentile (b, e, h), and 10th percentile (c, f, i) of June-August self-calibrating Palmer Drought Severity Index (sc-PDSI) values for gridded data (0.5° × 0.5° resolution) in the western US. (a–f), values are derived from the North American Drought Atlas (Cook et al., 2010), reconstructed using tree-ring data, for the intervals 1539–1588 (a–c) and 1938–1987 (d)-(f). (g–i), instrumental data are from the Climate Research Unit sc-PDSI data set (Barichivich et al., 2021).

4 Discussion

The character of California's 20th century enhanced hydroclimate variability (Figure 1) is unprecedented over the past 600 years, driven by significant (p < 0.05) increases in the magnitude of wet extremes (Figure 2) and these increases extend into much of the American West (Figure 4). Moreover, precipitation variability, streamflow variability, and the magnitude of wet extremes peak in the late 20th century in nCA and sCA (Figures 1 and 2). The recent increase in variability and wet extremes is strongest in sCA and the southern Sierra Nevada (Figures 1b and 2b), consistent with the results of an instrumental-record based study (Zamora-Reyes et al., 2022). This finding supports a largely independent 400-year tree-ring based reconstruction study (Haston & Michaelsen, 1997) that suggests a wet-extreme driven 20th century increase in precipitation variability in south and central California, with five of the top 10 wettest 5-year periods occurring in the 20th century. We recognize that tree-ring data can underestimate wet extremes and, therefore, variability, but such underestimation will be consistent throughout the reconstruction. Even if a tree-ring record underestimates wet extremes overall, this will not affect relative (temporal) changes in standard deviation (Figure 1) or 90th percentiles (Figure 2).

The extended temporal context of our study demonstrates that increased variability started earlier in nCA (early 20th century) compared to other regions (Figure 1a vs. Figures 1b and 1c). This finding contextualizes an instrumental analysis showing the 20th century rise in hydroclimatic variability over nCA (and sCA) occurred after the 1950s (Zamora-Reyes et al., 2022). That study attributed enhanced hydroclimatic variability to increasing wet winter precipitation extremes that were imprinted in winter, spring, and summer streamflow. Unlike wet extremes, we find no significant (p < 0.05) trend in dry extremes (Figure 3), except for a decreasing trend in streamflow over nCA (Table S5 in Supporting Information S1). This decreasing trend is consistent with that of instrumental spring and summer streamflow (Zamora-Reyes et al., 2022), and with WY low flows in nCA over this period and into the mid-21st century (Mallakpour et al., 2018).

The synchronous increase in 20th century hydroclimatic variability throughout California, driven by wet extremes, is unique in the context of the past 600 years (Figures 1 and 2). An additional peak in hydroclimatic variability occurred in the 16th century that was comparable in statewide synchrony and amplitude (Figure 1), but was driven by an increase in the magnitude of dry extremes (Figures 3 and 4c). This 16th century peak in hydroclimatic variability occurred during the Little Ice Age (ca. 1500–1850 CE), when temperatures were cooler than today (Trouet et al., 2013) and much of the Western US experienced decreased fire activity (Trouet et al., 2010). In 1580 CE, the Western US experienced extreme drought (Hughes & Brown, 1992; Meko et al., 2001), which likely contributed to the increased magnitude of dry extremes in the late 16th century. The 16th century was further characterized by enhanced El Niño Southern Oscillation variability (Li et al., 2013; Liu et al., 2017) that, through its influence on California hydroclimate (Schonher & Nicholson, 1989; Yoon et al., 2015), might also have contributed to the 16th century variability peak.

Our work highlights the importance of understanding synchronous changes to California's hydroclimatic system during the pre-industrial era, versus that of modern and future hydroclimate. The current enhancement of the hydrologic cycle is driven by increasing wet extremes, often related to winter storms associated with atmospheric river moisture (Dettinger, 2016; Gonzales et al., 2019; Guan et al., 2016). Atmospheric river-fed storms are projected to increase in frequency and magnitude in the 21st century at the expense of low to moderate frontal storms (Gershunov et al., 2019; Swain et al., 2018), further increasing precipitation variability (Swain et al., 2018). As temperature continues to rise, less precipitation will be stored as snowpack (Siirila-Woodburn et al., 2021), more will be delivered through extreme precipitation events (Payne et al., 2020), and the relationship between precipitation and streamflow throughout the year will become more complex (Cho et al., 2021; Davenport et al., 2020; Huang et al., 2020; Rhoades et al., 2021). The transformation of California's water cycle appears imminent, as the state already experiences events with high rain to snow ratios and rain-on-snow events, resulting in nonlinear watershed responses (Davenport et al., 2020; Henn et al., 2020). For instance, the precipitation that almost destroyed Oroville Dam in February 2017 was related to an atmospheric river family that was enhanced by more than 10% because of anthropogenic warming (Michaelis et al., 2022).

5 Conclusion

The recent increased magnitude of wet extremes coincides with decadal-scale drought conditions in California (Williams, Cook, et al., 2022). California's rainy season has become shorter and sharper in the 20th century (Dong et al., 2019; Luković et al., 2021), a trend projected to persist into the end of the 21st century (Swain et al., 2018). As the climate continues to warm, drought will intensify (Williams, Cook, et al., 2022), and California's hydroclimatic variability will be further enhanced both by increases in wet extremes and by changes on the other end of the distribution (Swain et al., 2018; Yoon et al., 2015), leaving the state more vulnerable to wildfires (Abatzoglou & Williams, 2016; Goss et al., 2020), floods (Williams, Livneh, et al., 2022), and compound extreme events (Zscheischler et al., 2018). Our results highlight the importance of collaboration between scientists and public officials to produce research that can be used in decision-making (Siirila-Woodburn et al., 2021), especially as California moves toward climate-related mitigation and adaptation strategies concerning their infrastructure.

Acknowledgments

This work was supported by the National Science Foundation (AGS-1349942) and California Department of Water Resources (Agreement 4600011071).

    Data Availability Statement

    The tree-ring reconstruction data used for this study are publicly available and may be obtained from TreeFlow (https://www.treeflow.info/california) and the National Oceanic and Atmospheric Administration (NOAA) Paleoclimate Database (https://www.ncei.noaa.gov/access/paleo-search/study/32352). Gridded reconstruction data can be obtained from NOAA's Paleoclimate Database (https://www.ncei.noaa.gov/access/paleo-search/study/19119) and instrumental gridded data from the Climate Research Unit can be obtained through the World Meteorological Organization's Climate Explorer (http://climexp.knmi.nl/select.cgi?scpdsi).