Rain triggers seasonal stratification in a temperate shelf sea

Northwest Europe lies beneath the North Atlantic Storm Track and experiences frequent extratropical cyclones that bring strong winds and heavy rain. In temperate shelf seas, the seasonal transition from winter mixing to summer stratification sets up the spring phytoplankton bloom, a key driver of marine productivity with ecological and socioeconomic importance. The canonical view attributes stratification onset primarily to increased solar heating overcoming wind- and tide-driven mixing, with rainfall typically considered negligible as a buoyancy source. However, winter and early spring observational gaps have limited understanding of preconditioning processes. This study asks whether rainfall associated with spring storms can trigger the onset of seasonal stratification in a temperate shelf sea, examines the mechanisms and timing using high-resolution ocean glider observations in the Celtic Sea (March 2015), and evaluates how often rainfall contributes to stratification onset over multiple decades and how large-scale climate modes modulate that timing.

Prior work emphasizes solar heating overcoming tidal and wind mixing as the dominant mechanism for seasonal stratification onset on temperate shelves, with tidal cycles modulating timing. This supports classical stability–bloom frameworks and links timing to ecosystem phenology. Rainfall has been shown to promote near-surface buoyancy and stratification in subtropical and monsoonal oceans, and episodically in shelf environments, but has generally been discounted as a primary control for temperate stratification onset due to its perceived small buoyancy input relative to heat fluxes. The lack of high temporal and spatial observations in winter has hampered testing of rainfall’s role. Autonomous ocean gliders can resolve near-surface processes during adverse weather, offering the capability to capture storm-driven buoyancy and mixing and to revisit assumptions about storm impacts, which are often framed as delaying stratification via wind-driven mixing.

Observations: An autonomous underwater glider (Unit 419 “Fortyniner”) was deployed 22 March–2 April 2015 along repeat transects between the Central Celtic Sea site (49°24.00′N, 8°36.00′W) and the shelf break site CS2 (48°34.26′N, 9°30.58′W). Sensors included a pumped Sea-Bird CTD (pressure, temperature, conductivity) and a Wet Labs triplet (chlorophyll-a fluorescence, optical backscatter, CDOM). Data were processed with a MATLAB-based Glider Toolbox using a flight model (Merckelbach et al.) and thermal inertia corrections (Lueck & Picklo; Garau et al.). Only one calibration offset round was possible due to sparse CTDs. The glider covered ~315 km and completed one full CCS–CS2 transect. Meteorology and air–sea fluxes: Meteorological observations (10 m wind speed, air density, mean sea level pressure, relative humidity) came from the UK Met Office ODAS buoy at CCS. ERA-Interim reanalysis provided precipitation, evaporation, winds, pressure, and SST for multi-decadal analyses. Surface heat fluxes for 2015 were computed following Wihsgott et al. Stratification metric and budget: Stratification strength was quantified by potential energy anomaly (φ, J m⁻³), the mechanical energy per unit depth required to homogenize the water column: φ = −∫₀ʰ (ρ−ρ̄) g z dz, where ρ(z) is density over depth h and ρ̄ is depth-mean density. The temporal evolution was decomposed as dφ/dt = dφ_heat/dt + dφ_rain/evap/dt + dφ_wind/dt + dφ_tides/dt. Thermal buoyancy: dφ_heat/dt = α g Q_net / (2 C_p). Freshwater buoyancy from precipitation: combined as dφ_heat/dt + dφ_rain/evap/dt = α g Q_net / (2 C_p) + α₀ P Δρ, where P is precipitation rate and Δρ the seawater–freshwater density difference. Mixing terms: wind mixing dφ_wind/dt = c₁ k₁ ρ_a u₁³ / h and tidal mixing dφ_tides/dt = (3/2) c₂ k₂ ρ u₂² / h with mixing efficiencies c₁=0.023, c₂=0.004; bottom drag k₁=0.0025; tidal drag k₂=C₀ y_s² with C₀=0.0012 and slippage factor y_s=0.02. Time series of φ were computed with and without rain/evaporation to isolate freshwater buoyancy impacts. Rain-driven salinity change: The surface mixed layer salinity change from precipitation was estimated as ΔS ≈ S₀ (1 − Z_SML / (Z_SML + P)), where S₀ is initial mixed-layer salinity, Z_SML the mixed-layer depth, and P the rain depth, and compared to observed near-surface salinity changes during storm events. Modeling: A 3D hydrodynamic hindcast used the AMM7 configuration of NEMO v3.6 (nominal 7 km resolution; 51 hybrid z–s vertical levels) covering the NW European shelf and adjacent NE Atlantic (20°W–13°E, 40°N–65°N). Bathymetry derived from NOOS; initial/boundary conditions from ORCA025 GO5.0; atmospheric forcing from ERA-Interim. After a 1981 spin-up, the model ran continuously 1982–2015. Model validation shows small SST biases (<0.05 °C surface; <−0.01 °C domain mean). Modeled stratification onset dates were validated against glider observations with a ~3 h timing offset. Multi-decadal composites examined precipitation, wind speed, sea level pressure, and SST relative to modeled onset dates, and related interannual onset variability to AMV and NAO indices. Analyses: Correlation of observed vertical density gradient with temperature and salinity gradients assessed whether stratification was thermo- or halo-dominated. Chlorophyll fluorescence and backscatter were used to infer phytoplankton biomass response to stratification onset. Regional representativeness was checked against moorings and spatial consistency up to ~120 km from the shelf break.

• March 2015 case study: After a rain event on 25–26 March, near-surface potential density decreased by ~0.006 kg m⁻³ over 12 h and sustained stratification (φ>0) developed. Night-time changes in ∂σθ/∂z correlated strongly with ∂S/∂z (r=0.9) and weakly with ∂T/∂z (r=0.1; p<0.001), indicating haline control; temperature showed a cold cap that would have reduced buoyancy in the absence of freshwater input.
• Rain quantification: 13.1 mm rainfall (25–26 March) coincided with observed surface freshening ΔS=0.0124. Using an average Z_SML=44 m, the estimated ΔS from rain dilution was 0.0104, accounting for most of the observed freshening. A second 2.8 mm event on 27 March explained over half of additional freshening. Sustained winds (~10.3 m s⁻¹) likely contributed wind-driven freshwater advection.
• Timing impact: Including rain-induced freshwater buoyancy advanced sustained stratification onset by about a week relative to considering thermal inputs alone. Without rain, model calculations suggest weak stratification on 26 March would have been eroded by winds, with sustained onset delayed until 1 April (7 days later). Precipitation-induced sensible heat contributed up to ~0.35 W m⁻² (<1% of daytime heat flux), not a controlling factor.
• Biological response: Near-surface chlorophyll fluorescence increased immediately after rain-induced stratification, peaking on 28 March; backscatter increases supported a biomass rise rather than photoacclimation. Elevated chlorophyll persisted into April, consistent with process cruise data.
• Regional representativeness: Timing and conditions matched observations up to ~120 km from the shelf break; mooring data confirmed sustained stratification while the glider transited off-shelf.
• Multi-decadal analysis (1982–2015): Rainfall and low sea level pressure occurred on or shortly before modeled onset dates in 30 of 34 years (88%), indicating rainfall commonly contributes to triggering onset. Winds typically peaked after rainfall, favoring wind-driven freshwater transport that can reinforce halo-stratification. SST generally increased 2–3 days after onset during post-storm quiescence.
• Climate modulation: Stratification onset dates were nearly twice as variable during positive AMV relative to negative AMV phases (interquartile ranges ~20 vs ~13 days). Relationships with NAO were more ambiguous, but NAO likely also modulates variability. Case years highlight dynamics: 2012 exhibited early stratification and subsequent erosion by prolonged winds; 1997 showed unusually early onset (26 Feb) under calm post-storm conditions without subsequent erosion.

The study demonstrates that rainfall associated with spring storms can provide sufficient freshwater buoyancy to trigger seasonal stratification in a temperate shelf sea, even when thermal stratification alone would be insufficient or transient under concurrent wind and tidal mixing. This directly addresses the research question by identifying halo-stratification as an initial, often overlooked, mechanism for onset. The findings reframe the role of storms: beyond delaying stratification via wind-driven mixing, storms can also accelerate onset when rain-to-wind ratios favor stabilization. The decoupling between the timing of density stratification and subsequent thermal strengthening explains observed lags in SST increases after onset. Interannual variability in onset timing is further governed by large-scale atmospheric modes. Positive AMV phases (and often concurrent negative NAO) align with less frequent but potentially more intense storms and increased atmospheric blocking, allowing more variable onset driven by episodic rain events followed by quiescent periods. Negative AMV phases, with straighter jets and more persistent windy conditions, tend to delay or stabilize onset timing via frequent mixing events. These dynamics have significant ecological implications, as onset timing influences the spring bloom, trophic interactions, and fish recruitment. Globally, regions subject to atmospheric rivers and storm-related heavy precipitation may experience similar rain-triggered stratification processes, suggesting broad relevance for shelf sea physics under current and future climates.

High-resolution glider observations, combined with an energetics framework and a multi-decadal shelf-sea model, show that rainfall commonly triggers the onset of seasonal stratification on the NW European shelf. In the Celtic Sea case, rain advanced onset by about a week relative to thermal forcing alone, initiated halo-stratification, and supported an early phytoplankton biomass increase. Over 1982–2015, rainfall coincided with onset in 88% of years, and AMV phase modulated onset variability, with nearly double the interquartile range during positive AMV. These results challenge the prevailing view that storms predominantly delay stratification and emphasize the importance of storm rain-to-wind ratios and post-storm quiescence. Rainfall must be included as an initial trigger mechanism in shelf-sea stratification frameworks and ecosystem models. Future research should: (1) extend analyses across full AMV cycles and other climate modes; (2) quantify the coupled physical–biogeochemical impacts of rain-triggered onset on bloom phenology and trophic dynamics; (3) assess sensitivities to precipitation biases in reanalyses and projections; and (4) evaluate regional and global applicability, particularly in areas influenced by atmospheric rivers under warming climates.

• Observational scope: The primary case study relies on one glider deployment over ~10 days and a single shelf transect, with limited independent CTD calibrations due to data scarcity.
• Forcing uncertainties: ERA-Interim precipitation can have regional biases; discrepancies between observed and modeled precipitation may affect timing and magnitude of freshwater buoyancy estimates.
• Model resolution and physics: The AMM7 model’s ~7 km resolution and parameterized mixing may not fully resolve submesoscale processes or fine-scale rain–freshwater lenses; stratification onset timing showed a small (~3 h) offset versus observations.
• Temporal coverage: The 34-year hindcast does not encompass a full AMV cycle (60–80 years), limiting inference about multi-cycle robustness of climate–stratification relationships; NAO relationships were ambiguous.
• Attribution complexity: Co-occurring wind-driven advection and tidal variability complicate isolation of rain effects; estimates of wind- and tide-driven mixing efficiencies introduce parameter uncertainty.