Atmospheric rivers (ARs), concentrated filaments of atmospheric moisture, are responsible for significant rainfall events along western coastlines globally. In western North America, and particularly California, ARs are linked to major flood-producing storms. The state's topography, especially the Sierra Nevada, amplifies the effects of ARs, leading to dramatic variations in precipitation totals and influencing the balance between surplus and drought. Recent examples include the winter of 2022-2023, marked by numerous AR storms, which significantly replenished snowpack and reservoirs, ending widespread drought but also causing major flooding and landslides, prompting disaster declarations. Climate models predict increased extreme precipitation events in California, potentially driven by enhanced AR activity, leading to greater flood risks. However, current flood hazard projections suffer from uncertainties due to limited instrumental data. Lake sediments offer a valuable archive for reconstructing past precipitation patterns beyond historical and instrumental records, providing insights into long-term climate-flood relationships. A key challenge is identifying high-resolution sediment records containing clear signatures of extreme precipitation events. This study utilizes a sediment core from Leonard Lake, California, chosen for its location within a high-frequency AR landfall zone and its potential to record extreme precipitation due to its steep-sided basin and proximity to the coast. The study employs a combination of lithological, grain size, and geochemical data with robust age control to reconstruct AR activity, using Si/Al ratios and clay content as proxies calibrated against modern instrumental records of IVT and IWV. This novel approach is enhanced by time-uncertainty analysis, critical for accurate quantitative interpretation in paleosciences.

Existing research highlights the significance of atmospheric rivers in producing extreme precipitation events, particularly in California. Studies using instrumental data and climate models have pointed toward an increasing likelihood of extreme precipitation events in the future due to strengthening AR activity. However, the limited temporal extent of instrumental data poses a major constraint on accurately predicting flood hazards. Past research has utilized lake sediments as archives for reconstructing long-term climate variability, revealing centennial-to-millennial-scale pluvial events in California. However, these studies often lack the temporal resolution necessary to capture shorter-term AR events. The identification of reliable proxies for paleo-AR activity is crucial, and previous research suggests elevated siliciclastic elements (Si, Ti, Al) in runoff as a potential fingerprint of extreme precipitation events. This study builds upon these previous findings, using high-resolution sediment data to reconstruct AR activity and employing rigorous statistical methods to account for uncertainties in the data and age-depth modeling.

A high-resolution sediment core (446 cm) was collected from Leonard Lake, Mendocino County, California, a strategically chosen location based on its proximity to a high-frequency AR landfall zone. The core was analyzed for lithological characteristics (clay layers, fine laminations, homogenous deposits, and sand layers), grain size distribution, and geochemical composition (Si, Al, Ti, K, Fe, Br) using X-ray fluorescence (XRF). Seventeen samples were radiocarbon (¹⁴C) dated, and short-lived radionuclides (²¹⁰Pb, ¹³⁷Cs, ²²⁶Ra) were measured in the upper 60 cm using gamma spectrometry. Bayesian software (Plum) was used to construct an age-depth model, accounting for uncertainty in radiometric dating. The GeoChronR package in R was employed to propagate age model uncertainty throughout the analyses. Principal component analysis (PCA) was used to examine correlations between different elements and to verify the trends in correlations. The instrumental record of AR activity was represented using integrated vapor transport (IVT) and integrated water vapor (IWV), data obtained from the Gershunov database. A calibration-in-time approach was used to establish a relationship between the sediment proxies (Si/Al and clay percentage) and the instrumental AR metrics. A linear regression model, accounting for regression dilution, was then used to reconstruct paleo-IVT from the sediment data. Ensemble methods were used to propagate uncertainties from the age-depth model and regression analysis to obtain a robust estimate of paleo-AR activity.

The Leonard Lake sediment core revealed prominent light greenish-gray clay layers interspersed with fine laminations and homogenous zones. These lithological features, coupled with geochemical analysis, suggest a relationship between clay layer deposition and periods of extreme precipitation associated with ARs. Si/Al ratios were found to be strongly correlated with modern instrumental records of IVT (rmedian = 0.63, pmedian = 0.02) and IWV (rmedian = 0.48, pmedian = 0.09), validating the use of Si/Al as a proxy for paleo-AR intensity. The reconstructed IVT record indicates that the highest median IVT values since the onset of the Medieval Climate Anomaly occurred in the late 20th century. However, the largest IVT peaks were found around 860 CE and 70 CE, exceeding the range of variability observed in the instrumental record. The reconstruction reveals a notable increase in IVT during the Little Ice Age and periods of lower IVT during the Medieval Climate Anomaly (600-725 CE, 1250-1350 CE, and the 1500s), aligning with known historical drought periods. The reconstructed data show remarkable repeating patterns interpreted as hydrologically forced changes in the lake's depositional processes. The correspondence between fine laminations and increased Fe/Al supports the idea that sedimentary Fe was created in situ as a result of anaerobic decomposition. This sequence suggests a process whereby large storms raise lake levels, creating anoxic bottom water conditions that facilitate the preservation of fine laminations until relatively dry conditions lowered lake levels. Dry conditions are likely recorded by homogenous packages that indicate oxygenated bottom waters and bioturbation.

The findings demonstrate that California experienced periods of extreme precipitation associated with ARs that significantly exceeded the range observed in the instrumental record. The strong correlation between Si/Al ratios in the sediment core and modern IVT data validates the use of this proxy for reconstructing past AR activity. The reconstructed IVT record aligns well with known paleoclimatic periods, such as the Medieval Climate Anomaly and the Little Ice Age, and corroborates evidence from other regional climate reconstructions (tree ring data). The study highlights the importance of extreme precipitation events in shaping pluvial and drought cycles in California over millennia. The results challenge the notion that AR activity is directly linked to the Northern Hemisphere temperature gradient, as indicated by the decreasing IVT trend between 1300 and 1500 CE despite a strong latitudinal temperature gradient and an increasing trend in IVT from 1500 CE to the present day as the gradient weakened. The reconstructed data provide a crucial long-term perspective on AR activity in California, improving the accuracy of risk assessments for precipitation-related hazards and informing future planning strategies. Although the study acknowledges limitations concerning potential changes in the relationship between Si/Al and extreme precipitation through time and other factors that might affect siliciclastic transport and preservation, the consistent sedimentation rate and the relatively undisturbed environment of Leonard Lake support the reliability of the proxy record.

This study provides the first high-resolution reconstruction of atmospheric river activity in California over the past 3,200 years. Using Si/Al ratios in lake sediments as a proxy for AR intensity, the research demonstrates that extreme precipitation events have been a significant driver of both pluvial and drought periods in the region. The findings reveal that past AR activity exceeded the range observed in the instrumental era, highlighting the potential for future extreme events. Future research could explore the mechanisms driving long-term variations in AR activity, investigating potential linkages with broader climate patterns and refining the understanding of the relationship between sediment proxies and AR intensity. Further investigation into other similar lake systems across the region could enhance the spatial coverage of paleo-AR reconstructions.

The study acknowledges potential limitations inherent in all paleo-reconstructions. One limitation is the potential for non-stationarity—the possibility that the relationship between Si/Al and extreme precipitation may have changed over time. Other factors, including lake level changes and shifts in vegetation, may influence siliciclastic transport and preservation. While the authors argue that the processes affecting Si/Al and clay content have been relatively stable, this assumption cannot be fully verified. Further, the reconstruction does not capture individual storm events, limiting the detail of individual storm records. The study is focused on a single site (Leonard Lake); broader spatial coverage could provide a more comprehensive view of regional AR variability.