1. Introduction
Atmospheric rivers (ARs) play an important role in California’s water cycle. ARs deliver a large fraction of the annual total precipitation to California, often in the form of a few extreme events per year (Dettinger et al. 2011; Guan et al. 2010; Huning et al. 2017, 2019). The storms replenish reservoirs and nourish wetland habitat (Florsheim and Dettinger 2015; Guan et al. 2010; Ralph et al. 2019). Although most of the precipitation brought by ARs is beneficial, they are often also associated with floods (Dettinger et al. 2011).
Several studies have looked at the relationship between ARs and floods in the western United States. About 85% of floods on the Pacific coast, including California, Oregon, and Washington, from 1949 to 2015 resulted from AR storms (Corringham et al. 2019; Konrad and Dettinger 2017). Ralph et al. (2006) found that all seven of the floods along Northern California’s Russian River between 1997 and 2006 occurred during AR events. Similarly, Neiman et al. (2011) found that out of 48 peak daily river flow events in western Washington from 1998 to 2009, all except two fit into their definition of an AR. Although flooding is most commonly associated with strong ARs, flooding can occur with ARs of low intensity when they last for several consecutive days and when soil moisture is already high (Konrad and Dettinger 2017; Ralph et al. 2013). The characteristic flood risk of a region from an AR is influenced by the landscape features, including antecedent soil moisture, runoff potential, and orographic precipitation features (Albano et al. 2020). In addition, many studies have suggested that as the atmosphere warms under climate change, precipitation from ARs may increase and may cause more flood events (Espinoza et al. 2018; Gershunov et al. 2019; Huang et al. 2020a; Huang and Swain 2022; Massoud et al. 2019; Michaelis et al. 2022; Rhoades et al. 2020, 2021). A better understanding of the characteristics of the ARs and the local attributes where they are occurring can help to better predict potential flood events (Albano et al. 2020; Ralph et al. 2006).
Motivated by the impacts associated with ARs and projections of more intense ARs under climate warming, this paper examines AR occurrences alongside documented floods and levee failures in the Sacramento–San Joaquin delta (“the Delta”) from 1980 to 2021. The region has more than 60 agricultural islands that are protected by levees and has experienced hundreds of levee failures since the levees were first built in the mid-1800s. Florsheim and Dettinger (2015) found that since 1951, about 80% of 128 levee breaks in California’s Central Valley occurred during AR storms, and the authors concluded that “despite construction of levees and other flood control structures, climate and floods continue to cause unintentional levee breaks” (p. 138). The aim of this study is to improve knowledge about the characteristics of flood-producing ARs in the built environment by examining 1) the spatial and temporal characteristics of ARs that cause levee failures in the Delta and 2) the changes in historic patterns of intensity and frequency of ARs that may impact delta levees differently moving into the future.
2. Study area
We focus our study on levee failures within the Delta boundary, a 2800 km2 area on the western edge of the Central Valley at the confluence of the Sacramento and San Joaquin Rivers, in the California Water Code sec. 12220 (see Fig. 1). The Delta is composed of over 60 low-lying islands protected by about 1800 km of levees. There are eleven small towns within the Delta and several urban centers just outside its border.
The Delta. (a) Map of the Delta (dark gray) and the watersheds that feed into it. The Sacramento and San Joaquin Rivers and their associated watersheds are shown in blue and green, respectively, and the cities of San Francisco, Sacramento, and Los Angeles are shown for geographic context. The MERRA-2 coastal grid cells where landfalling ARs are identified are plotted with red boxes, and the approximate dividing line between the Sacramento and San Joaquin watersheds (38.5°N) is marked with a red dotted line. (b) Inset map of the Delta. Map of the state–federal flood control project levees; private levees with benefits to the state interests that are eligible for the state assistance, referred to as nonproject; and those that are neither project nor eligible for the state assistance, referred to as unattributed levees within the Delta. The USGS gauges used for streamflow analysis are plotted as points. Source for levee data: the California Department of Water Resources (DWR 2018, 2022).
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
a. History of the Delta
Geomorphic change and variability have long been the characteristics of the Delta (Malamud-Roam et al. 2006). Coastlines migrated inland from sea level rise thousands of years ago; however, the past 2000 or so years have been more stable (Waters et al. 2016). Humans have further transformed this expansive freshwater tidal wetland by building levees, clearing riparian forests, and grazing (Whipple et al. 2012). Hydraulic mining debris raised the depth of the rivers and channels in the nineteenth century. By the late 1800s, most of the Delta was leveed and farmed (Whipple et al. 2012). In more recent decades, upstream dams have amplified changes to sedimentation transport, which has increased land subsidence (Waters et al. 2016). The region is characterized by a Mediterranean climate with hot, dry summers and cool, wet winters. Precipitation in the Delta is a variable both temporally and spatially. Most precipitation falls during the winter months of October–April, and much of it occurs during winter storms, particularly ARs, which predominantly flow from the southwest to the northeast orthogonal to the mountain topography facing the Delta. In addition to precipitation that falls locally, the Delta receives freshwater from the Sacramento River watershed (70 000 km2) and the San Joaquin River watershed (40 000 km2), from Mount Shasta to Fresno [Delta Stewardship Council (DSC) 2013; United Research Services (URS) 2007] (Fig. 1).
Before European settlement in the nineteenth century, the Delta was a vast area of brackish and freshwater tidal marshes connected by narrow, branching tidal channels. The area was home to thousands of indigenous peoples who lived, hunted, fished, and harvested in the marshlands. Indigenous villages occupied the higher ground, protected by natural and artificially constructed levees and sand mounds. For thousands of years, it appears that groups of people moved around seasonally for harvesting and in response to floods (Whipple et al. 2012).
European contact began in the late 1700s, and hunting, ranching, and agriculture began in earnest in the early 1800s. After the California gold rush, conversion of the Delta’s marshland to farmland commenced. The Arkansas Swamp Lands Acts, passed in 1850 by the U.S. government, gave states the right to sell swamp and overflow lands within their borders. In 1861, the California Board of Swamp Land Commission was created by the state to encourage the reclamation of marshland for agriculture. Over the past 200 years, levees were built to prevent seasonal flooding of the fertile land for agricultural use. A series of floods throughout the mid-to-late 1800s caused significant damage and led to farmers building higher and wider levees; nonetheless, many islands were devastated by floods for the decades to come, well into the twentieth century.
For the first 100 years of farming in the Delta, flood losses were the burden of the farmers and residents (Thompson 2006). Floods occur when a levee fails, is overtopped, or seeps, which most often occurs during times of inclement weather due to high river flow or wind-driven waves. Reservoir releases, changes in the water export facilities, and construction or removal of other barriers, such as fish barriers, can alter the river flow. On occasion, levee failures have occurred on “sunny days” for nonweather-related reasons, including burrowing rodents, erosion, or penetration by pipes or other equipment (DSC 2013). When a levee fails, the island floods like a bowl if it is below sea level. This flooding can place pressure on the inside of the levee, leading to more seepage or breaching.
b. The Delta today
Today, the Delta is central for the state’s water supply, and highways, railways, shipping channels, natural gas pipelines, and power transmission lines all run across the Delta to service the urban areas outside. As a result, the state’s economic interests are now at stake when a flood occurs. The extensive system of levees, as well as the multiple dams along the Sacramento and San Joaquin Rivers, are therefore part of a complex flood management system. The Delta’s levees are designated as “project,” “nonproject,” or “unattributed” based on funding sources and maintenance responsibilities [DSC 2015; California Department of Water Resources (DWR) 1995, 2012]. Project levees, which account for about 620 km of the 1800 km of levees, are designated by law [CWC section 9110(e)] as levees for which the state has given assurances to the federal government that they will be prioritized for maintenance to meet federal 100-yr flood standards, as defined by the Federal Emergency Management Agency. The majority of levees are nonproject and maintained by local reclamation districts, which are funded by property assessments. The landowners of an island control its reclamation district. Islands vary in size, and some contain only one or a few landowners; therefore, capacity for levee maintenance is highly heterogeneous. Many of the nonproject levees have been acknowledged to have special benefits to state interests. Levees that are neither state–federal flood control levees nor have been recognized as beneficial to the state are referred to as unattributed levees.
After a major levee failure in 1972 and a series of devastating storms in the early 1980s, many of the nonproject levees were improved with financial assistance from the state. The Delta Levees Maintenance Subventions (Subventions) Program began in 1973 as a cost-share agreement between the state and local reclamation districts on the Delta’s agricultural islands to fund levee maintenance (DWR 2023a). After the devastating floods of 1986, the goal of the program became to increase the height of levees to 1 ft above the 100-yr flood stage (URS 2011). Most of the levees were subsequently upgraded further to meet the state’s hazard mitigation plan (HMP) standards. A little over a decade later, the U.S. Army Corps of Engineers Public Law 84–99 (PL 84–99) became the new goal, which involved improving the Delta’s levees to 6 in. over the basic HMP and therefore ensuring federal funding for repair if a levee was damaged (U.S. Army Corps of Engineers 2023). Overall, California has invested over $700 million in Delta levees since 1973.
However, there are estimates that it will cost a further $1.5 billion to sufficiently improve all of the levees to PL 84–99 standards (DSC 2015). In 2014, the state attempted to address this deficiency by launching the Delta Levee Investment Strategy (DLIS), whereby levees would be categorized by the level of priority for the state interests (very high priority, high priority, or other priority) in order to allocate the state funding for levee improvements. After years of discussion at over 50 council meetings, 70 public workshops, and hundreds of public comments, in 2023, the DLIS was approved by the DSC for implementation beginning in 2024 (DSC 2023). Despite the significant local, state, and federal interest in levee safety, though, it is unclear whether these projects will be enough to protect the Delta from the intensifying ARs that are expected to occur in a future climate (Corringham et al. 2022; Huang et al. 2020b; Rhoades et al. 2020).
3. Data and methods
In this study, we first look at the characteristics of the levee events. We then combine information on levee events with AR occurrence information. We assess trends in how the frequency of different types of levee events is changing over time and explore the differences in AR climatology between the two primary watersheds affecting the Delta.
a. AR detection and categorization
For this study, a global AR dataset through December 2021 was used. The dataset is based on the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), 0.5° × 0.625° 6-hourly global reanalysis (Gelaro et al. 2017) and identifies ARs based on requirements of intensity, direction, and geometry using the Guan and Waliser detection algorithm, which was originally introduced in Guan and Waliser (2015) and refined and validated in Guan et al. (2018). This dataset has been evaluated and used in many studies focused on ARs in many different regions around the world (Eiras-Barca et al. 2018; Francis et al. 2022; Huning et al. 2019; Liang et al. 2023; Massoud et al. 2019; Ralph et al. 2019).
The global AR dataset was narrowed by choosing only the ARs that made landfall on the western coast of North America between 36.5° and 41.5°N, which encompasses the Sacramento and San Joaquin River watersheds that feed into the Delta [U.S. Geological Survey (USGS) hydrologic unit code footprints 1802 and 1804, respectively]. For this study, the boundary between the San Joaquin and Sacramento watersheds was assigned to be 38.5° north latitude, where the Sacramento watershed is to the north and the San Joaquin watershed is to the south (Fig. 1). We define an “AR day” as any day with at least one 6-h period meeting the criteria to qualify as an AR, and we measure AR intensity using the total integrated water vapor transport (TIVT) (Ralph et al. 2017; Guan et al. 2018). TIVT is conceptually analogous to streamflow for terrestrial rivers (Zhu and Newell 1994) and is found to be a robust AR metric relatively insensitive to the input data resolution (Guan et al. 2018).
b. Levee events
We define a “levee event” as either (i) a levee failure leading to a flood or (ii) a near-failure that was averted by flood fighting. Several reports have documented levee events in the Delta. California Department of Water Resources (DWR)’s 2009 Delta Risk Management Study (DRMS) compiled a list of levee failures in the Delta–Suisun region (i.e., including the tidal estuary to the west of the Delta), collectively, to determine the sustainability of the levees for the next 100 years (URS 2009). According to DRMS, between 1900 and 2009, there were 158 floods caused by levee failures. In a dissertation, Hopf (2011) noted that DRMS included low-lying levees in Suisun Marsh. We filtered the records of levee failures from URS (2009) to exclude levees outside of the boundaries of the Delta. We used Hopf (2011) to confirm the list of levee failures within the Delta, the dates of failures, and locations. Further details about levee events were obtained by searching local daily newspapers using LexisNexus (Nexis Uni) with the search terms: “Flood! AND levee! AND delta AND (Sacramento OR California OR San Joaquin).” Many papers have not been digitized before 2000, so we searched microfilm by dates of known levee failures and averted failures when flood fighting occurred, which resulted in 345 articles from 12 different newspapers. Additionally, we conducted 25 semistructured interviews with landowners, levee engineers, and government agency staff. Interviewees were identified through participant observation of over 20 public meetings and workshops concerning flood management and analysis of hundreds of public comments, newspaper articles, op-eds, and blogs that mentioned the levees and flood risk [University of California Santa Cruz (UCSC) Institutional Review Board (IRB) Protocol No. 2635]. Interviews were recorded and transcribed verbatim.
c. Precipitation and streamflow
Precipitation data were retrieved from the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center (CPC), which reports total daily precipitation at 0.25° × 0.25° resolution, using the R package rnoaa (Chamberlain and Hocking 2023). Watershed-level precipitation was calculated as a spatially weighted average of all cells intersecting with the Sacramento and San Joaquin watersheds, respectively. Streamflow data were retrieved from the USGS National Water Information System (NWIS) using the R package dataRetrieval (De Cicco et al. 2023). Daily streamflow values in thousands of cubic feet per second (kcfs) were retrieved for two gauges, one on the Sacramento River (USGS 11447650) and one on the San Joaquin River (USGS 11303500). The Sacramento River gauge is in the northeastern corner of the Delta and the San Joaquin River gauge is in the southeastern corner (Fig. 1); therefore, these two gauges represent the bulk of surface flows into the Delta on any given day. Daily precipitation and streamflow data are used together to generate the results described in section 4b.
d. Data aggregation and statistical analysis
The datasets were aggregated into three separate products to perform the analysis presented in the results. The first product, discussed in section 4b, measures the empirical probability of being preceded by an AR day. We estimated the empirical probability that a levee event was preceded by an AR day of a given intensity at time intervals of 1–30 days. These probabilities were compared against the empirical probability of any day in the wet season (not just days with levee events) being preceded by an AR day. If we assume that AR days are randomly distributed throughout the wet season, then the probability of an AR day of a given intensity occurring within a specified time interval can be computed with the binomial distribution. We resampled the historical record 1000 times to generate the 90% bootstrapped confidence intervals.
The second product, described in section 4c, is a time series of levee failures and averted failures by water year (WY),1 divided (“normalized”) by the number of AR days in that WY. This correction accounts for the contribution of interannual hydroclimatic variability to the levee failure risk; for example, because the 1980s and 1990s were notably wet periods in California and much of the 2010s was spent in severe drought (California Office of Environmental Health Hazard Assessment 2024), an unmodified time series would be biased toward a negative trend. If we see a decreasing trend in the normalized data, though, we can infer that the relationship between the two is changing independent of changes in AR days. We tested for monotonic trends in both levee failures and averted failures using the Mann–Kendall test and for stepwise changepoints using the Pettitt test with a significance threshold of p = 0.05.
The third and final product, analyzed in section 4d, is a time series of annual maximum TIVT by the Delta watershed. To understand differences in frequency and magnitude of the most intense AR events in both watersheds, we fit nonstationary generalized extreme value (GEV) distributions to the annual maxima of the Sacramento and San Joaquin ARs by water year. The GEV is one of the most widely used distributions for extreme value analyses of hydroclimatic data. More recently, nonstationary formulations of the GEV distribution have been used to estimate changes to hydroclimatic extremes in a future climate (Cheng and AghaKouchak 2014; Vu and Mishra 2019). In this study, the location parameter of the distribution (a measure of centrality) was allowed to vary linearly with time in order to capture changes in the magnitude of annual maxima events over the historical record. Therefore, there are four distribution parameters to be estimated: ξ0, the location parameter at time t = 0 (in our case, the year 1980); ξ1, the annual rate of change of the location parameter; α, the scale parameter (a measure of dispersion); and κ, the shape parameter (a measure of distribution shape). Distributions were fit using the R package climextRemes (Paciorek et al. 2018), which supports estimation of nonstationary parameters. A Kolmogorov–Smirnov goodness-of-fit test was unable to reject the null hypotheses that the data followed the respective fitted distributions (p = 0.70 for the Sacramento watershed; p = 0.47 for the San Joaquin watershed).
4. Results
a. Characteristics of Delta levee events, 1980–2021
Overall, there were 57 recorded levee events across the 40 complete WYs in the period from 1980 to 2021 (see Table S1 in the online supplemental material). Levee events were first categorized by the outcome of the event. In a few instances, levees completely failed, and the islands were permanently flooded (“permanent failure”; 5/57 events). In many cases, the levees were repaired, and the flood waters were drained (“temporary failure”; 33/57 events), if, for instance, a cost–benefit analysis showed that the islands’ benefit to agriculture and water quality outweighed the cost of pumping water out (McCullough 1982). For 19 of the 57 events (“averted failure”), a levee failure was averted through emergency management or flood fighting. A flood fight might take place when flood conditions arise and would include actions involving organized placement of sandbags and rocks to reduce overtopping, divert water from structures, and control boils (Burnett 2012; Pappalardo 2014).
Most (50/57) of the levee events occurred in the 7-month wet season from October through April, which overlaps with periods of high AR activity in central California. The majority of failures (41/57) are concentrated in January and February. Levee failures have also occurred from poor maintenance, high winds, tidal action, burrowing rodents, and collisions with equipment or boats. Attribution was not determined for the dry season levee failures in this study.
Of the over 60 islands in the Delta, 32 experienced at least one levee event from 1980 to 2021, and 13 experienced multiple events (see Fig. S1). In all but one case, permanent levee failures were allowed on islands with a history of repeat flooding, which likely influenced the cost–benefit decision to let those islands stay flooded. In the next subsection, we will tie these levee events to AR occurrences in the Delta to understand the patterns of climatological behavior that lead to adverse outcomes.
b. Coupling timing of ARs and levee events, 1980–2021
There were 1376 AR days in the study period, an average of 32.8 days yr−1. To better understand the relationship between ARs and levee events at different intensities, we calculated the 50th and 75th percentiles of the TIVT distribution across all grid cells, which were found to be 318 and 515 × 106 kg s−1, respectively. The 632 AR days were below the 50th percentile, 358 were in the 50th–75th percentile, and 386 were above the 75th percentile (Table 1). Most of the AR days (1077; 78.3%) occurred in California’s wet season from October through April, which is also when the majority of levee events occurred in this historical record.
Distribution of AR days.
To compare levee events against AR occurrence, we removed the seven levee events that occurred outside of the wet season, as well as one wet-season levee event with a specific, nonweather-related cause of failure listed. This left 49 levee events that are potentially AR-driven or AR-influenced. Figure 2a shows the summary of levee events (left axis) and AR days (right axis) from WY 1981–2021. The annual number of AR days varies from a low of 15 in WY 1991 to a high of 50 in WY 2017. The 1980s saw multiple abnormally wet water years, with higher-than-average annual numbers of AR days (35.2 days yr−1 vs 32.8 days yr−1), and experienced the most levee events of any decade in the historical record.
Levee events and AR days. (a) Number of levee events (bars, left axis) and AR days (line, right axis) by WY. The bars indicating the number levee events are colored by levee failure outcome. (b) Relationship between the number of AR days and the number of levee events, where every point represents one WY. WY 1980 is excluded because it starts in October 1979, beyond the range of our analysis, but levee events in the early months of 1980 are included for reference.
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
We explore the relationship between AR days and levee events further in Fig. 2b. We find a clear threshold at the mean of 32.8 AR days yr−1, as indicated by the dashed line. During the 21 water years with <32.8 AR days, there were no recorded failures or averted failures. Of the water years with >32.8 AR days, 10/20 saw at least one levee event. A chi-squared test rejected the null hypothesis of independence between AR days and levee events (p = 0.000 77). We infer that while a wetter year does not guarantee a levee event, AR-driven levee events seem to be due more to an accumulation of hydrologic impacts rather than individual high-intensity events.
We next examine the larger climatological context around individual levee events. Of the 49 events that are potentially AR-induced, 24 (49.0%) occurred on an AR day. However, 40 (81.6%) occurred within 3 days after an AR day and 46 (93.9%) occurred within 2 weeks of an AR day. The time lag between AR occurrence and levee events could be due to one of several factors. Operational decisions at the upstream dams could create a significant delay between AR-induced precipitation and peak streamflow in the Delta. Also, if an upstream levee fails where an island is subsided, the flood waters will fill the space of the island and downstream flow and pressure might be reduced.
Figure 3 shows the probability of a levee event being preceded by an AR day of a given intensity, with the 2-week interval marked by a dotted line. As stated above, 93.9% of levee events were preceded by an AR day of any magnitude within the previous 2 weeks. Further analysis reveals that 91.8% of events were preceded by an AR day in the upper 50% of the TIVT distribution (>318 × 106 kg s−1) and 87.8% were preceded by an AR day in the upper 25% of the TIVT distribution (>515 × 106 kg s−1). We compare these numbers to the probability of being preceded by an AR day on any arbitrary day within the wet season, assuming a random distribution of AR days. Days when levee events have occurred are approximately 10% more likely to have had an AR day of any magnitude, 30% more likely to have had an AR day in the upper half of the TIVT distribution, and 50% more likely to have had an AR day in the upper quarter of the TIVT distribution sometime in the last 2 weeks, relative to the baseline wet-season rate. Across almost all intensities and all time intervals, the probability of being preceded by an AR day is significantly higher on days with levee events than it is on random days, far exceeding the bounds of the 90% confidence interval of the random distribution. Therefore, Fig. 3 supports the hypothesis that antecedent climatological and hydrological conditions have a significant impact on the probability of levee failure or near failure.
Probability of being preceded by an AR day. The X axis defines the time window and the Y axis reports the probability of an AR day occurring within that time window. For each panel, the dashed line indicates the probability of any random wet-season day being preceding by an AR day and the solid line indicates the probability of a day with a recorded levee event being preceded by an AR day. Across all intensities and time windows, days with levee events are significantly more likely to be preceded by an AR day than the average. The vertical dotted line marks a before-failure time window of 2 weeks, and the dark gray shaded areas represent bootstrapped 90% confidence intervals around the random distribution.
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
We further explore the importance of antecedent conditions by focusing on a few impactful events and considering precipitation and streamflow as intermediate variables between AR occurrence and levee events. We examine the 5 days with three or more levee events, shown in Fig. 4. These are 27 January 1983 (four levee events), 19 February 1986 (four levee events), 1 January 1997 (four levee events), 3 January 1997 (three levee events), and 1 February 1998 (six levee events). The 4/5 of these days are AR days, and on average, 4.8 of the previous 7 days were AR days as well. Each panel of Fig. 4 shows the watershed-average precipitation in both the Sacramento and San Joaquin watersheds, as well as the differential between average streamflow for that day of the year versus observed streamflow before, during, and after the day of failure. The time period shown ranges from a week prior to the failure to 3 days after.
Precipitation and streamflow before and after multiple-failure days. The watershed-average precipitation in the Sacramento (blue) and San Joaquin (green) watersheds, as well as the observed (solid line) vs average (dotted line) streamflow in the Sacramento River (blue) and the San Joaquin River (green) are shown. The vertical (gray) line represents one of the 5 days in the historical record with three or more levee failures. The dates are (a) 27 Jan 1983, (b) 19 Feb 1986, (c) 1 Jan 1997, (d) 3 Jan 1997, and (e) 1 Feb 1998.
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
Figure 4 shows that many of the days before and during those with multiple levee events have nonzero precipitation and that streamflow tends to be elevated above average levels in both the Sacramento and San Joaquin Rivers on days with multiple levee events. Both 27 January 1983 (Fig. 4a) and 1 January 1997 (Fig. 4c) coincide with the days of peak precipitation and 19 February 1986 (Fig. 4b) and 3 January 1997 (Fig. 4d) coincide with the days of peak streamflow on the Sacramento River. The two events in early January of 1997 are distinct events in our analysis. Clearly, they are connected by the atmospheric and hydrological conditions, but those events were driven by peak precipitation and peak streamflow, respectively. The peak streamflow value on the San Joaquin River tends to lag the peak on the Sacramento River by a day or two. Interestingly, 1 February 1998 (Fig. 4e) coincides with neither peak precipitation nor peak streamflow on either river, and the levee events on this day seem to be occurring before the precipitation event begins. Further investigation reveals that 5/6 of the levee events on this day were averted failures and 4/6 were on low-lying levees in the Suisun Marsh. The year 1998 was a notably wet water year with multiple periods of heavy rainfall throughout the winter in Northern California, so this could be an example of levee damage due to sustained pressure on the levee walls rather than a directly AR-driven failure.
Our analysis has revealed a strong relationship between AR days and levee events, both at the water year scale and at the scale of individual events. At the water year scale, we identified that weather-related levee events have never occurred during water years with less than 32.8 AR days. At the event scale, the probability of experiencing an AR day of any magnitude in the 2 weeks before a levee failure day was found to be significantly higher than it is before a randomly selected day in the wet season from October through April. We also did a deep dive into the 5 days with three or more recorded levee events, including both failures and averted failures. High levels of precipitation and streamflow were present in all cases, and ARs were the clear drivers in all but one case. Based on these findings, we posit that AR conditions in the days–weeks preceding levee failures are just as important, if not more so, than the AR conditions on the day of the failure event.
c. Observed changes in levee failure outcomes over time
Up until now, we have considered all types of levee events together. An important question to ask is whether the frequency of levee events is changing over time, and whether the number of failures versus averted failures is shifting. Because of the significant investment in the Delta that has occurred over the last few decades, we would expect to see an increase in flood resilience unless there was a marked increase in the frequency and intensity of levee events over this period. Absent other climatic changes, there would be a decrease in the number of both temporary and permanent failures and potentially a corresponding increase in the number of averted failures. The results of the trend analysis are shown in Fig. 5.
Changes in the levee event frequency. Levee failures and averted levee failures, normalized by the number of AR days in each WY. Points that represent WYs with <32.8 AR days are shown in gray.
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
We ran two sets of experiments for both failures and averted failures: one using the whole historical record and one focusing only on water years with >32.8 AR days in the wet season (points shown in black in Fig. 5). Recall from earlier that water years with <32.8 AR days are highly unlikely to lead to levee events; removing these points is another way to remove variability associated with the hydroclimate. Each experiment involved a Mann–Kendall test for monotonic trends and a Pettitt test for stepwise changes, for a total of eight hypothesis tests. Of the eight tests we ran, only the Mann–Kendall test on the full 40 WY historical record was statistically significant at the p = 0.05 threshold. The test found a significant (p = 0.018) negative trend in the number of levee failures. When water years with <32.8 AR days were removed, though, the trend remained negative but the test lost statistical significance (p = 0.087). The Mann–Kendall test for the averted failures found no trend in the full historical record and a weakly positive (p = 0.42) trend in the dataset with water years with <32.8 AR days removed. The Pettitt test was unable to determine a stepwise changepoint with statistical significance in any of our analyses.
While the interannual variability in ARs, precipitation, and flooding remains a dominant driver of levee failure risk, there is some evidence that the frequency of temporary and permanent failures has decreased since 1980. There is less evidence of a corresponding increase in averted failures. The differential between the two could be due to any of a number of factors, including improved levee quality due to increased investment or better organized and proactive local emergency preparation with support from the state for materials on the ready.
d. Differences between Sacramento and San Joaquin ARs
To see if there is a connection between AR occurrence locations and levee events, AR days were labeled and separated by the watershed in which they occurred. Across the study period, 1055 AR days occurred in the San Joaquin River watershed and 769 occurred in the Sacramento River watershed (Table 1). Note that some days may be classified as AR days in both watersheds, so the sum of these two numbers exceeds the overall total of 1376 AR days in our study. Per watershed, the Sacramento received the most AR days (20) in 1983, whereas the San Joaquin received the most (27) in 2017.
We fit nonstationary GEV distributions to the annual maximum TIVT in each watershed to analyze trends in extreme AR behavior over time. The annual rate of change of the location parameter, ξ1, was found to be −2.1 ± 3.1 × 106 kg s−1 yr−1 in the Sacramento watershed and 1.8 ± 3.7 × 106 kg s−1 yr−1 in the San Joaquin watershed. In both cases, the mean is less than one standard deviation from zero, so the change in the location parameter is not statistically significant (p = 0.56 for the Sacramento watershed; p = 0.64 for the San Joaquin watershed). However, these small differences add up to noticeable distributional shifts over the length of the historical record. For example, the TIVT with a 10% annual chance of occurrence in 1980 is 1291 × 106 kg s−1 in the Sacramento watershed and 1265 × 106 kg s−1 in the San Joaquin watershed. By 2021, the value has dropped to 1203 × 106 kg s−1 in the Sacramento watershed and increased to 1337 × 106 kg s−1 in the San Joaquin watershed. These findings are in line with the trends projected in Huang et al. (2020b) and Huang and Swain (2022), whose studies project a large increase in the upper tail of precipitation and surface runoff projections in the San Joaquin River watershed relative to that in the Sacramento River watershed in the future. We also note that the bootstrapped 50% and 90% confidence intervals expand in both watersheds in more recent years, particularly in the San Joaquin watershed (Fig. 6b). This is consistent with prior research about California’s hydroclimate; the annual precipitation variance started increasing notably around 1975 (He and Gautam 2016), and the state is projected to see even more “whiplash” between extreme wet and extreme dry years in the future (Swain et al. 2018).
Changes in annual maximum daily TIVT. Nonstationary GEV distributions fit to annual maxima by WY in the (a) Sacramento and (b) San Joaquin watersheds. TIVT values from all AR days are shown as small gray dots and annual maxima are shown as larger black dots. The line represents the bootstrapped median of the GEV distribution, the darker shaded area represents the bootstrapped 50% confidence interval, and the lighter shaded area represents the bootstrapped 90% confidence interval.
Citation: Weather, Climate, and Society 17, 1; 10.1175/WCAS-D-23-0136.1
5. Discussion
We have explored many of the factors that influence levee failure risk. In this study, we find that rather than being driven by individual high-intensity events, failures tend to occur after an accumulation of AR days and precipitation over a period of days–weeks. We demonstrated the influence of antecedent conditions and watershed memory on levee failure risk at both the seasonal and subseasonal scales. Of the 49 wet-season levee events that could not be attributed to a non-AR cause (i.e., burrowing rodents, strong tides and wind, poor levee maintenance, or a ship collision into the levee), 87.8% of them were preceded by an AR day with TIVT in the upper 25th percentile (>515 × 106 kg s−1) within the 2 weeks before the levee failure. Many of the years with the highest numbers of levee events also had the highest numbers of AR days in the wet season from October to April. These findings add further support to the concept that even though floods are most often associated with strong AR events, soil moisture from consecutive low intensity storms can contribute to flood conditions (Konrad and Dettinger 2017; Ralph et al. 2013).
The majority of levee events in our database occurred in the 1980s and 1990s. The number of levee events decreased substantially after the 2000s, which aligns with both investments in significant levee and emergency management improvements and an extended period of dry years starting in the 2010s. Funding for the Subventions Program increased over the four decades of this study (DWR 2023a). Furthermore, the Special Projects Program has been providing additional funds for some levee improvements since the late 1980s. In addition, there have been improvements in coordination between local, state, and federal agencies and residents for emergency preparation (DWR 2023b). The state launched the California Data Exchange Center in 1986 to coordinate and monitor reservoir functionality and inspections (DWR 2024). Local emergency managers and others across the state have effectively used that system to stay informed about real-time reservoir inflow and outflow, river gauge data, and weather forecast data so that they can increase the frequency of levee inspections and minimize levee failures.
The Delta levees are unique in that they receive substantial funding for maintenance and improvements through a cost-share program between private landowners and the state; however, there are some levees that are not maintained as well as others. Hopf (2011) concluded that levees that receive funding through the Subventions Program have experienced far fewer failures than before the program began in 1973, which aligns with our findings of nonproject levees in the Delta.
In this paper, we were motivated to explore potential reasons why the number of levee events has declined since 1980. Analysis of existing research pointed to a decline in levee failures due to improvements in levee integrity and a slight increase or stabilization in averted failures due to increased expertise and frequency of emergency preparation and response. We were interested in gaining a better understanding of the hydroclimatic variability due to exceptionally wet years in the 1980s–90s and drought in the 2010s. After controlling for hydroclimatic variability, there was a decreasing trend in the number of levee failures per year and no trend in the number of averted failures. We point to 2017, which had the highest number of AR days in the record (50) and only three levee events. Indeed, this could be used as evidence that emergency preparation (i.e., barges in the river loaded with rocks to take to a levee in distress before a failure occurs) should be seen as an effective element of a levee maintenance plan that starts well before an AR event.
While ARs are a key driver of the Delta floods and levee failures, there are multiple factors affecting the Delta flood risk that were not addressed in this paper, including site-specific factors such as widespread land subsidence due to excessive groundwater withdrawal and higher dynamic wave loading on the levees due to changing wind patterns, as well as hydroclimatic changes such as changing snowpack dynamics and shifts in El Niño–Southern Oscillation. Future work could explore the relative contribution of each of these factors to overall risk, which would help to inform future flood mitigation strategies.
Despite the substantial investments made in the Delta levees over the previous decades, it is not clear that these improvements will be sufficient to create resilience in a future climate with significantly different AR characteristics. Projections of ARs in warmer years consistently illustrate that strong ARs will increase in frequency and intensity and deliver increasingly large volumes of water, as the climate warms (Gershunov et al. 2019; Huang et al. 2020b; Huang and Swain 2022; Rhoades et al. 2020; Wuebbles et al. 2017). A warmer climate where there is less snow and more rain will increase winter river flow (Gershunov et al. 2019).
Furthermore, our work highlights that not all areas of the Delta will be impacted equally. We showed a weakly decreasing trend in annual maximum TIVT magnitude in the Sacramento watershed and a weakly increasing trend in the San Joaquin watershed. Changes in runoff in the Sacramento and San Joaquin Rivers resulting from shifts in the AR intensity could be a challenge to the current flood protection system (Gershunov et al. 2019; Rhoades et al. 2020). Additional analyses of the meteorological and land surface conditions, such as midlevel and ground temperatures, wind speed, soil saturation, and existing snowpack, would contribute to our understanding of hazardous impacts from ARs in the Delta.
6. Conclusions
In this paper, we presented the findings from a dataset of 57 levee events in the Sacramento–San Joaquin Delta from 1980 to 2021, including both real failures (permanent and temporary) and averted failures. We then examined the relationship between levee events and ARs. At the annual scale, all levee events in the historical record occurred during water years with an above-average number of AR days (greater than the annual mean of 32.8). At the daily scale, days with levee events were 10% more likely to be preceded by at least 1 AR day of any intensity and 50% more likely to be preceded by the most intense AR days in the 2 weeks prior to the failure event. A case study of the 5 days with three or more recorded levee events showed that the precipitation and streamflow conditions were often higher than average both in the days prior to the event and on the day of the event.
The results from this study show that while the AR intensity is one indicator of the likelihood of a levee failure, consecutive ARs of lower intensity can also be followed by levee failures. Levee improvements and better emergency preparation in recent years have likely contributed to the lower number of levee failures in the later years of this study; however, it is expected that the increase in the AR intensity will continue as the climate warms. A more in-depth understanding of the ARs that are potentially hazardous can help residents and flood managers in the Delta better prepare for floods.
Acknowledgments.
This work was partially funded by the Delta Stewardship Council Delta Science Program (California Sea Grant R/SF-86 1167). Development of the AR detection algorithm and databases was supported by NASA and the California Department of Water Resources. We thank the three reviewers who greatly improved the quality of this work.
Data availability statement.
The AR data used in this study are available from https://ucla.box.com/ARcatalog. Also, see https://dataverse.ucla.edu/dataverse/ar. Levee data supporting this study are provided as supplemental information accompanying this paper.
Footnotes
A WY is the 12-month period from 1 October to 30 September. It is designated by the calendar year in which it ends. As such, the first and last years of our study are incomplete WYs because our AR data began in January 1980 and ended in December 2021.
REFERENCES
Albano, C. M., M. D. Dettinger, and A. A. Harpold, 2020: Patterns and drivers of atmospheric river precipitation and hydrologic impacts across the western United States. J. Hydrometeor., 21, 143–159, https://doi.org/10.1175/JHM-D-19-0119.1.
Burnett, R., 2012: Flood fighting methods. Department of Water Resources Division of Flood Management, 42 pp., https://tehamacountywater.org/wp-content/uploads/2020/08/flood-fight-methods.pdf.
California Office of Environmental Health Hazard Assessment, 2024: 2018 Report: Indiciators of climate change in California: Precipitation. Accessed 14 April 2024, https://oehha.ca.gov/epic/changes-climate/precipitation.
Chamberlain, S., and D. Hocking, 2023: rnoaa: ‘NOAA’ Weather data from R, version 1.4.0. R package, accessed 11 April 2024, https://docs.ropensci.org/rnoaa/.
Cheng, L., and A. AghaKouchak, 2014: Nonstationary precipitation intensity-duration-frequency curves for infrastructure design in a changing climate. Sci. Rep., 4, 7093, https://doi.org/10.1038/srep07093.
Corringham, T. W., F. M. Ralph, A. Gershunov, D. R. Cayan, and C. A. Talbot, 2019: Atmospheric rivers drive flood damages in the western United States. Sci. Adv., 5, eaax4631, https://doi.org/10.1126/sciadv.aax4631.
Corringham, T. W., J. McCarthy, T. Shulgina, A. Gershunov, D. R. Cayan, and F. M. Ralph, 2022: Climate change contributions to future atmospheric river flood damages in the western United States. Sci. Rep., 12, 13747, https://doi.org/10.1038/s41598-022-15474-2.
De Cicco, L. A., Hirsch, R. M., Lorenz, D., Watkins, W. D., and Johnson, M., 2023: dataRetrieval: R packages for discovering and retrieving water data available from Federal hydrologic web services, version 2.7.13. R package, accessed 11 April 2024, https://doi.org/10.5066/P9X4L3GE.
Dettinger, M. D., F. M. Ralph, T. Das, P. J. Neiman, and D. R. Cayan, 2011: Atmospheric rivers, floods and the water resources of California. Water, 3, 445–478, https://doi.org/10.3390/w3020445.
DSC, 2013: The Delta Plan: Ensuring a Reliable Water Supply for California, a Healthy Delta Ecosystem, and a Place of Enduring Value. Delta Stewardship Council, 38 pp., https://deltacouncil.ca.gov/delta-plan.
DSC, 2015: State investments in delta levees. Delta Stewardship Council, 43 pp., https://cawaterlibrary.net/wp-content/uploads/2017/04/DSC_Levee_Investment_Strategy_Issue_Paper.pdf.
DSC, 2023: Delta levees investment strategy. Delta Stewardship Council, accessed 1 October 2023, https://deltacouncil.ca.gov/dlis/.
DWR, 1995: Sacramento-San Joaquin delta atlas. California Department of Water Resources, 134 pp., https://www.waterboards.ca.gov/waterrights/water_issues/programs/bay_delta/deltaflow/docs/exhibits/nmfs/spprt_docs/nmfs_exh4_dwr_1995.pdf.
DWR, 2012: Delta region integrated flood management key considerations and statewide implications. California Department of Water Resources, 80 pp., https://cawaterlibrary.net/wp-content/uploads/2017/05/Levee-Memo-DWR.pdf.
DWR, 2018: Legal Delta boundary. Accessed 8 October 2024, https://gis.data.cnra.ca.gov/datasets/57b02f8a5e77465f902376dbd9522585_0/about.
DWR, 2022: Delta levees centerlines 2017. Accessed 8 October 2024, https://gis.data.cnra.ca.gov/datasets/7995cfe94bcc4e15afe92f40efb66cc6_0/about.
DWR, 2023a: Delta levees maintenance subventions: Program history. Accessed 28 October 2023, https://water.ca.gov/Work-With-Us/Grants-And-Loans/Delta-Levees-Maintenance-Subventions.
DWR, 2023b: Delta levees special flood control projects: Program history. Accessed 28 October 2023, https://water.ca.gov/Work-With-Us/Grants-And-Loans/Delta-Levees-Special-Flood-Control-Projects.
DWR, 2024: Welcome to the California data exchange center. https://cdec.water.ca.gov/.
Eiras-Barca, J., N. Lorenzo, J. Taboada, A. Robles, and G. Miguez-Macho, 2018: On the relationship between atmospheric rivers, weather types and floods in Galicia (NW Spain). Nat. Hazards Earth Syst. Sci., 18, 1633–1645, https://doi.org/10.5194/nhess-18-1633-2018.
Espinoza, V., D. E. Waliser, B. Guan, D. A. Lavers, and F. M. Ralph, 2018: Global analysis of climate change projection effects on atmospheric rivers. Geophys. Res. Lett., 45, 4299–4308, https://doi.org/10.1029/2017GL076968.
Florsheim, J. L., and M. D. Dettinger, 2015: Promoting atmospheric-river and snowmelt-fueled biogeomorphic processes by restoring river-floodplain connectivity in California’s Central Valley. Geomorphic Approaches to Integrated Floodplain Management of Lowland Fluvial Systems in North America and Europe, Springer, 119–141.
Francis, D., R. Fonseca, N. Nelli, D. Bozkurt, G. Picard, and B. Guan, 2022: Atmospheric rivers drive exceptional Saharan dust transport towards Europe. Atmos. Res., 266, 105959, https://doi.org/10.1016/j.atmosres.2021.105959.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Gershunov, A., and Coauthors, 2019: Precipitation regime change in western North America: The role of Atmospheric Rivers. Sci. Rep., 9, 9944, https://doi.org/10.1038/s41598-019-46169-w.
Guan, B., and D. E. Waliser, 2015: Detection of atmospheric rivers: Evaluation and application of an algorithm for global studies. J. Geophys. Res. Atmos., 120, 12 514–12 535, https://doi.org/10.1002/2015JD024257.
Guan, B., N. P. Molotch, D. E. Waliser, E. J. Fetzer, and P. J. Neiman, 2010: Extreme snowfall events linked to atmospheric rivers and surface air temperature via satellite measurements. Geophys. Res. Lett., 37, L20401, https://doi.org/10.1029/2010GL044696.
Guan, B., D. E. Waliser, and F. M. Ralph, 2018: An intercomparison between reanalysis and dropsonde observations of the total water vapor transport in individual atmospheric rivers. J. Hydrometeor., 19, 321–337, https://doi.org/10.1175/JHM-D-17-0114.1.
He, M., and M. Gautam, 2016: Variability and trends in precipitation, temperature and drought indices in the state of California. Hydrology, 3, 14, https://doi.org/10.3390/hydrology3020014.
Hopf, F., 2011: Levee failures in the Sacramento-San Joaquin River Delta: Characteristics and perspectives. Ph.D. dissertation, Texas A&M University, 410 pp., https://oaktrust.library.tamu.edu/server/api/core/bitstreams/e5a6d0ac-0f31-46fb-87e0-5f28ee7befa5/content.
Huang, X., and D. L. Swain, 2022: Climate change is increasing the risk of a California megaflood. Sci. Adv., 8, eabq0995, https://doi.org/10.1126/sciadv.abq0995.
Huang, X., S. Stevenson, and A. D. Hall, 2020a: Future warming and intensification of precipitation extremes: A “double whammy” leading to increasing flood risk in California. Geophys. Res. Lett., 47, e2020GL088679, https://doi.org/10.1029/2020GL088679.
Huang, X., D. L. Swain, and A. D. Hall, 2020b: Future precipitation increase from very high resolution ensemble downscaling of extreme atmospheric river storms in California. Sci. Adv., 6, eaba1323, https://doi.org/10.1126/sciadv.aba1323.
Huning, L. S., S. A. Margulis, B. Guan, D. E. Waliser, and P. J. Neiman, 2017: Implications of detection methods on characterizing atmospheric river contribution to seasonal snowfall across Sierra Nevada, USA. Geophys. Res. Lett., 44, 10 445–10 453, https://doi.org/10.1002/2017GL075201.
Huning, L. S., B. Guan, D. E. Waliser, and D. P. Lettenmaier, 2019: Sensitivity of seasonal snowfall attribution to atmospheric rivers and their reanalysis-based detection. Geophys. Res. Lett., 46, 794–803, https://doi.org/10.1029/2018GL080783.
Konrad, C. P., and M. D. Dettinger, 2017: Flood runoff in relation to water vapor transport by atmospheric rivers over the western United States, 1949–2015. Geophys. Res. Lett., 44, 11 456–11 462, https://doi.org/10.1002/2017GL075399.
Liang, J., Y. Yong, and M. K. Hawcroft, 2023: Long-term trends in atmospheric rivers over East Asia. Climate Dyn., 60, 643–666, https://doi.org/10.1007/s00382-022-06339-5.
Malamud-Roam, F. P., B. Lynn Ingram, M. Hughes, and J. L. Florsheim, 2006: Holocene paleoclimate records from a large California estuarine system and its watershed region: Linking watershed climate and bay conditions. Quat. Sci. Rev., 25, 1570–1598, https://doi.org/10.1016/j.quascirev.2005.11.012.
Massoud, E. C., V. Espinoza, B. Guan, and D. E. Waliser, 2019: Global climate model ensemble approaches for future projections of atmospheric rivers. Earth’s Future, 7, 1136–1151, https://doi.org/10.1029/2019EF001249.
McCullough, C. A., 1982: 1980 Floods in the Sacramento-San Joaquin Delta. Storms, Floods, and Debris Flows in Southern California and Arizona, 1978 and 1980: Proceedings of a Symposium, National Academy Press, 207–214.
Michaelis, A. C., A. Gershunov, A. Weyant, M. A. Fish, T. Shulgina, and F. M. Ralph, 2022: Atmospheric river precipitation enhanced by climate change: A case study of the storm that contributed to California’s Oroville dam crisis. Earth’s Future, 10, e2021EF002537, https://doi.org/10.1029/2021EF002537.
Neiman, P. J., L. J. Schick, F. M. Ralph, M. R. Hughes, and G. A. Wick, 2011: Flooding in western Washington: The connection to atmospheric rivers. J. Hydrometeor., 12, 1337–1358, https://doi.org/10.1175/2011JHM1358.1.
Paciorek, C. J., D. A. Stone, and M. F. Wehner, 2018: Quantifying statistical uncertainty in the attribution of human influence on severe weather. Wea. Climate Extremes, 20, 69–80, https://doi.org/10.1016/j.wace.2018.01.002.
Pappalardo, E. A., 2014: The importance of levee performance in the reduction and evaluation of risk in the Sacramento-San Joaquin Delta. M.S. thesis, Dept. of Civil and Environmental Engineering, University of California, 24 pp.
Ralph, F. M., P. J. Neiman, G. A. Wick, S. I. Gutman, M. D. Dettinger, D. R. Cayan, and A. B. White, 2006: Flooding on California’s Russian River: Role of atmospheric rivers. Geophys. Res. Lett., 33, L13801, https://doi.org/10.1029/2006GL026689.
Ralph, F. M., T. Coleman, P. J. Neiman, R. J. Zamora, and M. D. Dettinger, 2013: Observed impacts of duration and seasonality of atmospheric-river landfalls on soil moisture and runoff in coastal Northern California. J. Hydrometeor., 14, 443–459, https://doi.org/10.1175/JHM-D-12-076.1.
Ralph, F. M., and Coauthors, 2017: Dropsonde observations of total integrated water vapor transport within North Pacific atmospheric rivers. J. Hydrometeor., 18, 2577–2596, https://doi.org/10.1175/JHM-D-17-0036.1.
Ralph, F. M., and Coauthors, 2019: ARTMIP-early start comparison of atmospheric river detection tools: How many atmospheric rivers hit northern California’s Russian River watershed? Climate Dyn., 52, 4973–4994, https://doi.org/10.1007/s00382-018-4427-5.
Rhoades, A. M., and Coauthors, 2020: The shifting scales of western U.S. landfalling atmospheric rivers under climate change. Geophys. Res. Lett., 47, e2020GL089096, https://doi.org/10.1029/2020GL089096.
Rhoades, A. M., M. D. Risser, D. A. Stone, M. F. Wehner, and A. D. Jones, 2021: Implications of warming on western United States landfalling atmospheric rivers and their flood damages. Wea. Climate Extremes, 32, 100326, https://doi.org/10.1016/j.wace.2021.100326.
Swain, D. L., B. Langenbrunner, J. D. Neelin, and A. Hall, 2018: Increasing precipitation volatility in twenty-first-century California. Nat. Climate Change, 8, 427–433, https://doi.org/10.1038/s41558-018-0140-y.
Thompson, J., 2006: Early reclamation and abandonment of the central Sacramento-San Joaquin delta. Sacramento Hist. J., VI, 41–72.
URS, 2007: Status and trends of Delta-Suisun services. Public Review DRAFT, 58 pp., https://cawaterlibrary.net/document/status-and-trends-of-delta-suisun-services/.
URS, 2009: Delta Risk Management Strategy. California Department of Water Resources from documents developed by URS Corporation/Jack R. Benjamin and Associates, Inc., accessed 28 October 2023, https://www.waterboards.ca.gov/waterrights/water_issues/programs/bay_delta/california_waterfix/exhibits/docs/SJRECWA/sjrecwa_3.pdf.
URS, 2011: Phase 2 Risk Reduction Report Section 4 Final. URS Corporation/Jack R. Benjamin and Associates, Inc., prepared for California Department of Water Resources.
U.S. Army Corps of Engineers, 2023: Levee rehabilitation & inspection program (Public Law 84-99). Accessed 1 October 2023, https://www.nwp.usace.army.mil/Missions/Flood-Risk-Management/Levees/PL84-99/.
Vu, T. M., and A. K. Mishra, 2019: Nonstationary frequency analysis of the recent extreme precipitation events in the United States. J. Hydrol., 575, 999–1010, https://doi.org/10.1016/j.jhydrol.2019.05.090.
Waters, C. N., and Coauthors, 2016: The Anthropocene is functionally and stratigraphically distinct from the Holocene. Science, 351, aad2622, https://doi.org/10.1126/science.aad2622.
Whipple, A., R. Grossinger, D. Rankin, B. Stanford, and R. Askevold, 2012: Sacramento-San Joaquin delta historical ecology investigation: Exploring pattern and process. SFEI-ASC’s Historical Ecology Program Rep. 672, 438 pp., https://www.sfei.org/documents/sacramento-san-joaquin-delta-historical-ecology-investigation-exploring-pattern-and-proces.
Wuebbles, D. J., D. W. Fahey, K. A. Hibbard, D. J. Dokken, B. C. Stewart, and T. K. Maycock, 2017: Climate Science Special Report: Fourth National Climate Assessment. Vol. I. U.S. Global Change Research Program, 470 pp., https://doi.org/10.7930/J0J964J6.
Zhu, Y., and R. E. Newell, 1994: Atmospheric rivers and bombs. Geophys. Res. Lett., 21, 1999–2002, https://doi.org/10.1029/94GL01710.
