Response of sea surface temperature to atmospheric rivers

The study addresses how atmospheric rivers (ARs)—long, moisture-laden filaments responsible for most meridional water vapor transport—modify sea surface temperature (SST) through air–sea interactions, and how these modifications affect AR predictability beyond 10–20 days. ARs contribute substantially to precipitation and flooding, especially along the U.S. West Coast, yet prediction skill remains limited. Physical expectations suggest ARs increase downward heat fluxes (warming SST) while strong winds enhance vertical mixing (cooling SST). Prior work indicated entrainment-driven cooling and synoptic-scale air–sea feedbacks can improve AR prediction. However, the individual contributions of atmospheric and oceanic processes to SST tendencies during ARs, and their spatial dependence across the North Pacific, had not been quantified. This study uses an ocean state estimate and an SST budget framework to diagnose the processes governing SST tendency responses to ARs and to map their spatial and seasonal variability.

Previous studies identified climate modes (e.g., ENSO, MJO) as predictors of AR activity but with limited predictability horizons (~10–20 days). Shinoda et al. first composited ARs and highlighted the role of ocean entrainment in SST cooling downstream of AR cores. Sun et al. showed that using interactive ocean models (vs. fixed-SST) improves AR-related forecast skill for integrated water vapor and vapor transport, especially during strong cooling periods. Other works link ENSO to shifts in storm tracks affecting AR frequency and location. Studies have also shown mesoscale ocean features and fronts can influence AR intensity and heavy precipitation by modulating surface fluxes and moisture transport. The sensitivity of marine stratocumulus to large-scale conditions (including ENSO) is well documented and relevant to cloud-mediated shortwave changes during ARs.

- Data and period: Analyzed water years 1993–2017 (Oct–Mar) over the North Pacific (10°N–60°N, 120°E–120°W). Ocean variables and surface fluxes from ECCOv4r4 (1° MITgcm state estimate, 1992-01-01 to 2018-01-01), which conserves and closes the heat budget; atmospheric fields (IWV, IVT, winds, clouds) from ERA-Interim (79 km, 60 levels). Sensitivity checks with ERA5 yielded similar IWV/IVT. All variables were converted to anomalies by removing a 15-day-smoothed daily climatology. Data were regridded to a common 2°×2° grid by cosine-latitude-weighted averaging; SST budget closure uses only ECCOv4 fields. - AR detection: Built AR “objects” following a modified Guan & Waliser (2015) algorithm, with criteria including an IVT threshold (max of 100 kg m−1 s−1 or local 85th percentile within a 15-day window), minimum length >2000 km, orientation constraints relative to mean IVT, aspect ratio >2, no equatorial straddling, and mean poleward IVT >50 kg m−1 s−1. A grid-day is an AR day if covered by an AR object. - Mixed-layer diagnostics: Mixed layer depth (MLD) defined as shallowest depth where potential density exceeds surface by 0.03 kg m−3, using ECCOv4-consistent equation of state. Mixed-layer averages computed from surface to depth h over free surface η. - SST budget framework: Used mixed-layer potential temperature (θ̄) as proxy for SST and decomposed its local tendency θ̄_loc = ∂θ̄/∂t into contributions from shortwave, longwave, sensible, latent, dilution (PmE), advection (mean, eddy, and entrainment effects), vertical mixing (diffusion and entrainment), detrainment, and horizontal diffusion. Grouped terms into surface forcing θ̄_sfc and ocean dynamics θ̄_ocn such that θ̄_loc = θ̄_sfc + θ̄_ocn. Diagnostic terms derived consistently from ECCOv4 MITgcm outputs, with vertical mixing represented via GGL TKE scheme; advection includes Eulerian and bolus (eddy parameterization) velocities. Composites and statistics: For maps, composited anomalous tendencies on AR days; significance tested against daily climatology. For temporal analysis, computed means and variability across valid years (requiring ≥5 AR days in the selected months). - EOF analysis: Computed EOFs of AR-day frequency per water year to relate spatial shifts in AR activity to ENSO (NINO3.4).

- Ocean modulation of SST tendencies: Ocean dynamics frequently oppose surface-forced warming during ARs; in AR-active regions they can offset over 100% of the anomalous SST warming that surface fluxes would induce, effectively canceling or reversing SST tendency. - Dominant ocean processes: Ageostrophic (Ekman) advection and vertical mixing (diffusion and entrainment) are the leading oceanic contributors to SST tendency responses. Vertical mixing cooling strengthens with increased wind-driven turbulence and can occur even when surface forcing anomalies are weak. - Nonlinear regime behavior: Joint distributions show two regimes. For θ_sfc > ~0.2×10^−6 K s^−1, the ocean tends to oppose surface warming (θ_ocn becomes more negative). For θ_sfc below that threshold, the ocean amplifies surface-driven cooling (more negative θ_ocn). Linear fit between total tendencies indicates θ_sfc alone cannot explain θ_ocn (e.g., weak anomaly correlation R^2 ~0.07 after removing the mean slope relationship). - Spatial patterns (Oct–Mar): North of ~30°N, net local SST tendency often shows warming driven by enhanced longwave, sensible, and latent heat fluxes; cooling arises from reduced shortwave due to increased clouds. The oceanic response generally opposes the surface forcing via advective and mixing cooling. South of ~30°N, patterns tend toward cooling from shortwave reductions except in a notable exception region. - Strong marine warming region: Over [15°N–25°N, 120°W–135°W], ARs produce pronounced warming (>0.3 K day^−1) due to net decreases in total cloud cover that increase shortwave radiation, despite increased mid/high clouds; stratocumulus reduction (weakened inversion, AR-associated ascent/low SLP) leads to MLD shoaling and detrainment warming of the mixed layer. - Coastal California: Anomalous southerly winds during ARs reduce upwelling-favorable conditions, yielding net warming via anomalous Ekman (ageostrophic) advection; this is partially offset by increased vertical mixing cooling from enhanced turbulent kinetic energy penetration. - Variability hotspots: Interannual variability of the SST tendency response is strongest along the Kuroshio Extension and in the strong marine warming region. In the eddy-rich Kuroshio Extension, variability is dominated by advection and vertical mixing; mesoscale processes likely play a key role. - Seasonality: Surface-forced warming and oceanic cooling largely cancel in Oct–Nov and Feb–Mar. Vertical mixing cooling is strongest in fall (thin MLD, strong vertical temperature gradient at MLD base) and weak in Dec–Jan; modest mixing cooling reappears in Feb–Mar as AR forcing interacts with seasonal MLD evolution. - AR climatology and drivers: AR frequency peaks near 35°N. Leading AR-frequency EOFs represent meridional and zonal shifts and both correlate positively with NINO3.4 (EOF1 at 95% confidence; EOF2 at 80%), indicating ENSO influence on AR activity patterns.

The results clarify how AR-induced surface fluxes interact with ocean dynamics to shape SST tendencies across the North Pacific. Ocean processes frequently counteract atmospheric warming during ARs, leading to substantial cancellation in local SST tendencies that vary by region and season. Key regional behaviors include Ekman-advection-driven warming along coastal California and cloud-mediated shortwave-driven warming in the East Pacific stratocumulus region, contrasting with widespread mixing/advection-driven cooling elsewhere. These findings address the research question by quantifying the specific atmospheric and oceanic contributions to SST tendencies during ARs and mapping their spatial and temporal variability, thus explaining when and where the ocean amplifies or opposes atmospheric forcing. The implications for prediction are significant: models without interactive oceans will likely lose skill after AR passages, especially during AR families where successive events interact with pre-existing SST anomalies. Given projected increases in AR frequency with global warming, resolving SST changes due to surface fluxes, vertical mixing, and Ekman advection becomes critical for improved AR forecasting and climate projections.

This study decomposes the SST tendency response to atmospheric rivers across the North Pacific using a closed mixed-layer heat budget from a 25-year ocean state estimate. It demonstrates that ocean dynamics, particularly ageostrophic (Ekman) advection and vertical mixing, can negate or reverse surface-flux-driven warming during ARs, with strong spatial and seasonal dependence. Notable regional responses include Ekman-advection-driven warming along coastal California and cloud-regime-driven shortwave warming in the East Pacific stratocumulus region. These process-level insights highlight the necessity of interactive ocean components for skillful AR prediction and suggest that improved representation of air–sea fluxes, vertical mixing, and Ekman processes under AR conditions can enhance forecast performance. Future work should address freshwater-forcing impacts on stratification, explicitly resolve mesoscale and frontal-scale ocean–atmosphere interactions, incorporate two-way coupling to capture feedbacks, and further investigate ENSO-related modulation of AR activity and cloud regimes.

- Model resolution and dynamics: ECCOv4’s 1° ocean state estimate is not eddy-resolving; mesoscale eddy effects are parameterized via bolus velocity, limiting detailed interpretation in eddy-rich regions (e.g., Kuroshio Extension). - One-way coupling: The ocean simulation is forced with reanalysis fluxes and does not feed back to the atmosphere; potential two-way air–sea feedbacks during ARs are not represented. - Cloud and radiation complexity: The shortwave response hinges on cloud-regime changes (e.g., stratocumulus reduction), which are inferred from reanalysis composites and may involve processes not fully resolved or coupled. - AR detection choices: A slightly more selective AR identification than GW15 was used; results could be sensitive to detection thresholds and criteria. - Scope constraints: Physical mechanisms connecting ENSO to zonal shifts in AR frequency (EOF2) are not resolved and left for future work; detailed eddy–AR interactions were not pursued due to resolution limits.