Sea-ice loss in Eurasian Arctic coast intensifies heavy Meiyu-Baiu rainfall associated with Indian Ocean warming

In early summer, the East Asian summer monsoon advances northward into the subtropics and encounters cooler mid-to-high latitude air, forming a quasi-stationary rain belt from June into July (Meiyu/Changma/Baiu). Extreme Meiyu-Baiu events (e.g., 1998, 2020) cause severe flooding, large economic losses, and casualties across densely populated regions from central-eastern China to Japan, underscoring the need for improved understanding and prediction. Traditionally, heavy Meiyu-Baiu rainfall is linked to prior-winter El Niño and subsequent Indian Ocean (IO) warming, which induces an Indo–Northwest Pacific anticyclone, enhances low-level southwesterlies, moisture supply, and ascent over the Meiyu-Baiu region. However, reliance on low-latitude SST anomalies alone results in limited prediction skill in some years, notably failing to capture the record July 2020 rainfall, though June 2020 was predicted. Rapid Arctic sea-ice loss and amplified Arctic warming have known impacts on mid- to high-latitude circulation, suggesting Arctic sea ice as a potential predictor for East Asian summer rainfall. Observations from 2020 hint that early-summer sea-ice loss along the Siberian coast contributed to heavy Meiyu-Baiu rainfall via East Siberian warming and high-pressure anomalies. Yet, observational relationships are non-stationary and limited by short records. The study hypothesizes that Arctic sea ice significantly influences East Asian summer rainfall only in conjunction with IO SST conditions. Preliminary modeling indicated improved simulation of 2020 rainfall when observed Arctic sea-ice concentration anomalies were prescribed alongside observed IO SST anomalies.

Prior work has established a strong link between preceding-winter El Niño events, subsequent IO warming in spring-summer, and enhanced Meiyu-Baiu rainfall via the Indo-Pacific Ocean Capacitor mechanism and an anomalous Northwest Pacific anticyclone that strengthens southwesterlies and moisture transport (e.g., Xie 2009, 2016). Nonetheless, predictive skill based solely on low-latitude SSTs can be inadequate for certain years, such as July 2020. Concurrently, studies implicate Arctic amplification and sea-ice loss in modulating mid-high latitude circulation and potentially East Asian precipitation. Observational analyses suggest that early-summer sea-ice reductions along the Siberian coast and associated East Siberian high-pressure anomalies may have contributed to the 2020 extreme Meiyu-Baiu event. However, reported Arctic–East Asia precipitation links vary across studies and periods due to short reliable observational records, non-stationary relationships, and confounding influences (ENSO, teleconnections). This study builds on these insights by positing a synergistic effect: Arctic sea-ice anomalies modulate Meiyu-Baiu rainfall primarily when IO warming is present.

- Models and data: Utilized a 100-member CESM2 Large Ensemble (CESM2LE) of historical simulations (1950–2014) with CMIP6 forcing. Ensemble means at each grid point were removed to extract unforced variability. Observations/reanalyses include GPCP v2.3 precipitation, ERA5 precipitation/reanalysis, and HadISST SST and sea-ice concentration (SIC). Anomalies are relative to 1981–2010 climatology; area weighting used for averaging. - Case definitions: Define May IO SST anomalies; years outside ±1 standard deviation are warm IO (WIO) or cold IO (CIO). Define May Kara Sea (KS) sea ice extent (SIE) anomalies; outside ±1 standard deviation are low SIE (LSI) or high SIE (HSI). Combine to form WIO-LSI, WIO-noLSI, LSI-noWIO, CIO-HSI, CIO-noHSI cases. CESM2LE suggests May KS SIE is generally uncorrelated with May IO SST and preceding-winter Niño3.4 on interannual timescales, implying independence. - Analyses: Composite precipitation anomalies over June–July; probability distributions of monthly accumulated precipitation anomalies for June and July; daily evolution of rain belt; scatter of July precipitation vs May IO SST across cases. Decompose July precipitation into large-scale and convective components; compute CAPE changes; analyze vertical profiles of specific humidity, temperature, and lapse rate (∂T/∂p). Diagnose 1000–700 hPa vertically integrated moisture flux/convergence and 700–300 hPa mean horizontal temperature advection, separating dynamic (wind-driven), thermodynamic (humidity/temperature-gradient), and nonlinear components. Examine geopotential height differences (Z200/Z500/Z850) to identify circulation patterns. - Atmospheric-only experiments: Forced CAM6 (atmospheric component of CESM2; 1.25°×0.9°, 32 levels) with monthly SIC and SST anomalies composited from CESM2LE WIO-LSI and WIO-noLSI cases relative to a 1995–2005 climatological control to isolate atmospheric responses to Arctic SIC/SST differences under WIO. - Statistical prediction: Multiple regression model for July precipitation anomalies using standardized May IO SST and an exponential function of standardized May KS SIE to capture stronger impacts of sea-ice loss than gain. Coefficients (b1, b2) estimated from all CESM2LE years (1950–2014; 6500 years). Applied to observations (1950–2020) with variance rescaling to correct magnitude bias. Compared two models: IOSST-only and IOSST&KSSIE (with sea ice). Evaluated skill via correlation maps, prediction bias time series, July 2020 case study, and RMSE over 1996–2020. Statistical significance via two-tailed t-tests at 5% level.

- Synergistic effect: In years with IO warming (WIO), concurrent May Kara Sea sea-ice loss (LSI) substantially enhances Meiyu-Baiu rainfall. WIO-LSI composites show markedly stronger June–July precipitation over the Meiyu-Baiu region than WIO-noLSI; LSI alone without WIO does not produce comparable anomalies. - Risk amplification: Relative to WIO-noLSI, the distribution of accumulated precipitation shifts rightward in both June and July under WIO-LSI, with heavier right tails. Probability of rainfall comparable to or exceeding 2020 increases from ~7% to ~13% in June and from ~5% to ~11% in July (approximately doubled risk). In July, the median precipitation anomaly is ~45 mm in WIO-LSI, about 50% greater than in WIO-noLSI; June median is ~45% higher but contributes only ~27% of the June–July mean change. - Season timing: WIO-LSI prolongs the Meiyu-Baiu season into July without substantially altering onset; positive anomalies persist through July versus waning after mid-July in WIO-noLSI. - Process and components: Increase is dominated by convective precipitation. Over the Meiyu-Baiu region, convective precipitation increases by nearly 90% in WIO-LSI versus WIO-noLSI, while large-scale precipitation increases by <10%. - Extreme days: Although WIO-LSI accounts for ~4% of July simulation days, it contributes ~10% of extremely heavy rainy days. Nearly 13% of July days in WIO-LSI experience extremely heavy rainfall versus just over 8% in WIO-noLSI (1.6× increase in likelihood). - Thermodynamics and dynamics: CAPE increases more in WIO-LSI, with maxima south of Japan and along the middle–lower Yangtze. Western Meiyu-Baiu region exhibits mid–upper tropospheric cooling and modest near-surface warming, steepening the lower-tropospheric lapse rate and enhancing instability; humidity increases near-surface (west) and throughout the column (east). - Moisture and temperature advection: Lower-level (1000–700 hPa) moisture convergence increases notably, especially in the east, largely due to dynamic (wind) effects that account for >70% of the increase. Mid–upper (700–300 hPa) cooling corresponds to enhanced cold advection driven mainly by meridional wind anomalies. - Circulation pattern: Differences feature a quasi-barotropic dipole over East Asia (high over East Siberia, low to its south near the Meiyu-Baiu region), producing a cyclonic anomaly that aids moisture convergence and cooling aloft. CAM6 forced with WIO-LSI minus WIO-noLSI SIC/SST anomalies reproduces a similar dipole, implicating Eurasian Arctic sea-ice loss and coastal SST warming as drivers. - Asymmetry: High SIE can partly mitigate deficits under cold IO (CIO), but impacts are weaker, indicating an asymmetric combined effect favoring enhancement under WIO. - Prediction improvement: A two-parameter IOSST&KSSIE model improves July precipitation predictions versus IOSST-only. Regional bias reductions are substantial in years with low spring sea ice (e.g., 1997, 2007, 2011, 2012, 2015). For 1996–2020, RMSE decreases from 1.09 to 0.91. For July 2020, mean bias over the Meiyu-Baiu region is reduced by 39%, with about 50% reductions around Wuhan and Shanghai.

The findings demonstrate that late-spring Kara Sea sea-ice loss can strongly modulate and intensify IO-warming–induced Meiyu-Baiu rainfall by dynamically and thermodynamically enhancing convective activity, particularly in July. Physically, additional ice-free ocean along the Eurasian Arctic coast absorbs more solar radiation in late spring and warms the lower atmosphere, fostering a positive height anomaly over North Asia. This strengthens a ridge and favors atmospheric blocking over East Siberia, while inducing a cyclonic anomaly over and south of the Meiyu-Baiu region. The resulting wind anomalies drive stronger lower-tropospheric moisture convergence and enhanced mid–upper tropospheric cold advection, which steepens lapse rates and elevates CAPE, increasing convective precipitation and the likelihood of extreme daily rainfall. The circulation pathway appears distinct from previously emphasized zonally oriented teleconnections (e.g., circumglobal or Eurasian wave trains), with the primary influence confined to East Asia. Importantly, when IO warming is present, incorporating Arctic sea-ice information substantially improves seasonal predictions of July Meiyu-Baiu rainfall and extreme-event risk, offering actionable lead time of weeks to months for preparedness.

This study establishes that Eurasian Arctic (Kara Sea) sea-ice loss in May significantly intensifies Indian Ocean warming–induced Meiyu-Baiu rainfall, prolongs the rainy season into July, elevates convective activity, and substantially increases the probability of extreme rainfall events comparable to 2020. Mechanistically, sea-ice loss and associated Arctic coastal SST warming generate a quasi-barotropic dipole over East Asia that enhances lower-level moisture convergence and mid–upper tropospheric cooling, boosting instability and convection. A simple statistical model combining May IO SST and an exponential function of May KS SIE materially improves seasonal prediction skill and reduces bias, particularly in extreme years like 2020. Future work should assess robustness across models and higher-resolution simulations, improve dynamic prediction of early-summer Arctic sea ice to extend lead times, and develop more sophisticated statistical or machine learning frameworks to capture nonlinearities and additional predictors.

- Model dependence: Quantitative results may depend on CESM2LE and CAM6 configurations; coarse resolution likely underrepresents extreme daily rainfall intensity and variability. - Simplified prediction model: The two-parameter regression explains a limited fraction of July variance; despite bias reduction, predictability remains modest, indicating that late-spring Arctic sea ice is not a sole determining factor. - Observational constraints: Reliable observational records are relatively short; concurrence of strong WIO and low KS SIE is rare, limiting empirical sample sizes. - Sea-ice predictability: Poor predictive skill for early-summer Arctic SIE constrains practical lead time for forecasts. - Mechanistic isolation: Atmospheric-only experiments lack air–sea–ice feedbacks and show damped responses; some regional displacement of circulation anomalies remains.