Threat by marine heatwaves to adaptive large marine ecosystems in an eddy-resolving model

The study investigates how climate change will alter marine heatwaves (MHWs) in Large Marine Ecosystems (LMEs), which support ~95% of global fish catches but lie in coastal zones where low-resolution climate models perform poorly. The research tests whether, even if marine organisms fully adapt to gradual mean sea surface temperature (SST) warming, changes in variability and extremes would still intensify MHW exposure. The authors highlight observed trends of longer, more frequent MHWs largely driven by mean SST increases, and note shortcomings of defining future MHWs relative to historical baselines (mean warming-inclusive thresholds), which embed a built-in increase. They propose using a high-resolution eddy-resolving Earth system model and a future threshold (90th percentile computed in the future climatology) to distinguish effects of mean warming from higher moments of SST variability, aiming to quantify risks to LMEs under a warming climate and to assess the necessity of high-resolution modeling to capture coastal processes.

Observations show MHWs have become longer, more frequent, and more extensive in recent decades, primarily due to mean SST warming. Many regional and global models project stronger and more frequent MHWs under warming scenarios (e.g., RCP8.5), often approaching near-permanent heatwave states when defined relative to historical baselines. Prior work commonly uses mean warming-inclusive thresholds, which conflate mean warming with changes in variability. Studies advocate moving thresholds to isolate changes in higher moments. Low-resolution (~1°) models struggle to resolve boundary currents, coastal upwelling, and mesoscale eddies that influence MHWs, leading to biases in coastal/LMEs. High-resolution regional/global models (~0.1° ocean) better reproduce observed MHW frequency, duration, and intensity. Ecological studies document impacts such as coral bleaching, seagrass decline, and fish distribution shifts linked to MHWs, underscoring socioeconomic vulnerability in LMEs.

Models and scenarios: The authors use a 250-year high-resolution CESM1.3 simulation (CESM-HR) spanning 1850–2100 with CMIP5 historical forcings to 2005 and RCP8.5 thereafter. CESM-HR employs 0.25° atmosphere/land and 0.1° ocean/sea-ice (eddy resolving). Comparisons are made to a low-resolution CESM (CESM-LR; ~1° atmosphere and ocean) and a CMIP5 20-model ensemble (low-resolution). For model-observation intercomparison in the historical period, all fields are interpolated to 1°. Future vs historical comparisons with CESM-HR use the native 0.25° grid. Observations: Daily NOAA OISST v2.1 (0.25°) and the GHRSST Multi-Product Ensemble (GMPE v2.0, 0.25°) provide observational benchmarks for 1982–2011. LMEs: Fifty-four LMEs within 70° N/S are analyzed, grouped by adjacent continents (North America, South America, Europe, Africa, Asia, Australia). LMEs are coastal regions typically ≥200,000 km², encompassing continental shelves, currents, and estuaries. MHW definition and metrics: For each grid cell, a daily threshold is computed as the 90th percentile within an 11-day moving window over a 30-year period, smoothed with a 31-day moving average. An MHW is ≥5 consecutive days above threshold; events separated by ≤2 days are merged. Metrics include annual frequency, total annual MHW days, and mean intensity (mean SST anomaly relative to climatology during events). Analysis is limited to 70° N/S to avoid direct ice/snow influence. Threshold strategies: Two thresholds are used for future (2071–2100) MHWs: (1) mean warming-inclusive threshold: the 90th percentile based on the historical period (1975–2004); (2) future threshold: the 90th percentile based on the future period (2071–2100). The latter removes mean warming to highlight changes due to variability and persistence. Pseudo scenario for attribution: A pseudo warming-only scenario adds the 30-year mean SST difference (future minus historical) to historical daily SST, isolating the role of mean warming in MHW changes. Analyses: The study evaluates historical MHW frequency and mean intensity against OISST/GMPE; computes zonal means and LME-group statistics; assesses future changes in annual MHW days and mean intensity under both thresholds globally and over LMEs; examines relationships with detrended SST variance; and contrasts behaviors in western boundary current (WBC) regions. Statistical significance (e.g., correlations, P<0.05) is reported where relevant.

- Model fidelity: CESM-HR reproduces observed spatial patterns and zonal means of MHW frequency and mean intensity better than CESM-LR and CMIP5. Biases in annual MHW frequency relative to OISST are −0.31, −0.55, and −0.60 events per year for CESM-HR, CESM-LR, and CMIP5, respectively (≈15%, 27%, 30%). CESM-HR also more closely matches observed MHW mean intensity distributions, though intensity in WBC regions is overestimated relative to OISST but comparable to GMPE maxima in North American LMEs. - Dominant role of mean warming under historical baseline: Using the mean warming-inclusive threshold (historical 90th percentile), projected increases by 2071–2100 (RCP8.5, CESM-HR) are very large: mean annual MHW days increase by 287.2 days between 70° N/S, implying near-permanent MHWs in many equatorial/subtropical regions; mean intensity increases by 1.2°C, with larger increases in the Northern Hemisphere. A pseudo warming-only scenario reproduces ≥94% of these changes, indicating mean warming dominates. - Residual increases after removing mean warming: Using the future threshold (future 90th percentile), global increases are modest but positive: mean annual MHW days increase by 2.8 days and mean intensity by 0.2°C (2071–2100 vs 1975–2004). The pronounced hemispheric dipole in intensity seen with the historical baseline disappears, yielding more uniform increases. - LME-specific results: Historically, LMEs (70° N/S) exhibit 27.4–40.6 annual MHW days (mean 33.2). Under the mean warming-inclusive threshold, LME annual MHW days soar to 351.4 by 2071–2100; mean intensity more than doubles from 1.2°C to 2.9°C. Under the future threshold, 98% of LMEs show increased annual MHW days (mean +2.8 days), and 93% show increased mean intensity (mean +0.2°C). Increases in days are linked to greater persistence (higher SST autocorrelation) despite some frequency decreases; intensity increases correlate with increased detrended SST variance. - Variability link: Strong correlations between MHW mean intensity and detrended SST standard deviation across LMEs in both historical (R=0.97) and future (R=0.96) periods. Enhanced variability is attributed to strengthened modes (e.g., ENSO, north tropical Atlantic variability, positive Indian Ocean Dipole) and nonlinear evaporation–SST relationships. - WBC behavior in high-resolution model: CESM-HR projects distinct, larger changes in major WBC regions (Kuroshio Extension, Gulf Stream, Zapiola Anticyclone, Agulhas Return Current, East Australian Current, South Pacific storm track), often with meridional dipoles (increases on poleward flank, decreases equatorward), linked to shifts in frontal positions and changes in SST variance. Such distinct WBC behavior is not evident in CESM-LR/CMIP5. - LME stress persistence: Under the future threshold, the correlation between future and historical LME mean intensities is 0.9 (P<0.05), compared with 0.6 using the mean warming-inclusive threshold, indicating LMEs stressed today will remain among the most stressed even after adjusting baselines. - Fisheries relevance: LMEs with the highest catches (top 15; category I) tend to experience larger increases in annual MHW days and intensity than lower-catch LMEs (category II), implying potentially greater socioeconomic impacts.

The findings show that while mean SST warming drives the majority of projected MHW increases when using historical baselines, even after removing the mean warming effect with a future threshold, most LMEs still face increased MHW exposure in both duration and intensity. This indicates that full biological adaptation to mean warming would not eliminate growing heat-stress risks. High-resolution, eddy-resolving modeling is crucial to capture coastal processes, WBC dynamics, and realistic MHW characteristics affecting LMEs, which low-resolution models underestimate or misrepresent. The strong linkage between MHW intensity and SST variance suggests that projected strengthening of climate variability modes (e.g., ENSO, tropical Atlantic, Indian Ocean Dipole) will further exacerbate MHW severity in many LMEs. The persistence of relative stress ranking across LMEs implies that regions currently vulnerable will remain so, necessitating targeted adaptation and management strategies, particularly in high-catch LMEs where impacts may be amplified. These results refine risk assessments by separating mean warming effects from variability-driven extremes and emphasize the need for improved resolution in climate projections relevant to coastal ecosystems and fisheries.

Using an eddy-resolving high-resolution Earth system model and a future-threshold definition of MHWs, the study demonstrates widespread increases in annual MHW days and mean intensity across most LMEs by late century, even under the optimistic assumption that organisms fully adapt to mean warming. High-resolution modeling significantly improves historical fidelity and reveals distinct WBC responses critical for coastal risk assessment. The work highlights substantial ecological, social, and economic risks to LMEs and calls for response strategies by communities and policymakers. Future research should develop multimodel ensembles of high-resolution simulations under multiple forcing scenarios to quantify uncertainties, explore mechanisms (e.g., variability mode changes, frontal shifts), and provide early warning for LME management and fisheries planning.

- Model ensemble size: Primary projections rely on a single high-resolution model (CESM-HR); while supported by CESM-LR/CMIP5, a multi-model high-resolution ensemble is not yet available. - Resolution and observations: Satellite SST (0.25°) may underestimate mesoscale variability in WBC/coastal regions compared to the 0.1° ocean model, complicating bias assessments. - Spatial scope: Analyses exclude high latitudes (>70° N/S) due to ice/snow threshold issues; results may not generalize to polar LMEs. - Threshold choices: Results depend on the MHW threshold methodology (90th percentile windows, event merging rules); alternative definitions could yield quantitative differences. - Forcing scenario: Projections use RCP8.5; other scenarios (e.g., SSP pathways) may produce different magnitudes of change. - Biological adaptation: The future-threshold framework assumes full adaptation to mean warming, which may not be achievable given rates of change and species-specific constraints, potentially underestimating ecological risks.