Year-round marine heat extremes severely impact marine life during the first year at 1.5 °C warming
Data files
Jun 09, 2026 version files 301.88 KB
-
Database_1.csv
227.03 KB
-
Database_2.csv
8.24 KB
-
Database_3.csv
7.64 KB
-
Database_4.csv
7.42 KB
-
Database_5.csv
168 B
-
RCode.zip
25.36 KB
-
README.md
15.13 KB
-
Source.zip
10.90 KB
Abstract
Periods of anomalous ocean heat are becoming more intense and persistent, with consequences for biodiversity, food systems, and coastal resilience. Yet ocean impact assessments often emphasize summer, creating a seasonal bias that can underrepresent the year-round consequences of marine heat extremes. In 2023–2024, ocean surface warming exceeded 1.0°C above pre-industrial levels, coinciding with year-round marine heat extremes across much of the ocean. Here, we provide a rapid global assessment of marine impacts during this exceptional year. We show that summer impacts were frequent and severe, but 29% of non-tropical impacts occurred in cooler seasons. Events involving compound hazards occurred across seasons, though their timing varied by hazard type. Summer assessments alone are therefore insufficient to capture when, where, and how ocean heat threatens marine ecosystems. Rapid, year-round assessments are needed to understand how anomalous heat affects marine ecosystems and dependent societies, while informing preparedness plans for inevitable future extremes.
https://doi.org/10.5061/dryad.zgmsbccph
Description of the data and file structure
This repository is associated with the manuscript Emerging Ocean Impacts in the First Year at 1.5ºC Warming and contains the databases, supplementary information, extended data, R codes, and source files related to the paper. The datasets document global ocean impacts observed during the first year in which global mean surface air temperatures reached 1.5°C above pre-industrial levels and global mean sea surface temperatures exceeded 1°C. The data were compiled through an extensive literature search conducted in 17 languages, incorporating peer-reviewed studies and grey literature sources (e.g., governmental reports, news media) collected between June 2023 and December 2024.
This repository includes three CSV files (Database 1, Database 2, Database 3, Database 4, and Database 5) and three ZIP files (Rcode.zip, Figure_Source.zip, Event_source.zip).
Database_1.csv: Ocean Impact database
Database containing all extracted events/impacts occurring in the ocean systems from June 2023 to June 2024 and its related metadata. The dataset includes 201 recorded events occurring during the 2023–2024 global ocean heat anomaly and were classified into seven primary categories (e.g., low-oxygen extreme events, disease outbreaks, mass coral bleaching events).
Dataset Structure:
Event_ID: refers to the published source (identified by number) from which the data were extracted. In instances where multiple events were reported in the source (column Q, T, W, etc), an identifying letter was assigned to each individual event (a-m).
Extractor_ID: identifies the individual responsible for data collection (anonymised).
Date_Event: Denotes the extracted or approximated date of the event. The standardisation of approximations to the 15th of the month occurred in instances where no further specification could be identified in the sources.
Seasonal_Day: refers to the date of occurrence of each event based on the season and hemisphere, accounting for opposing seasonal cycles
Hemisphere: refers to the latitudinal geographical hemispheres
Region: classifies each event according to broad latitudinal biogeographic zones: Equatorial, austral, and boreal
Marine_Region: identifies the marine region the event was reported in
Location: most detailed location that could be extracted from sources
Latitude: latitudinal coordinate
Longitude: longitudinal coordinate
Impact: describes the main type of the reported event; a) low-oxygen extreme events, b) disease outbreaks, c) habitat loss, d) harmful algal blooms, e) mass coral bleaching events, e) population outbreaks, f) other extreme heat-related impacts – events primarily driven by heat extremes that did not fall under more specific categories such as mass coral bleaching events, and h) other for events that were indirectly linked to heat extremes or associated with extreme weather events.
Impact Outcome: describes the type of the impact outcome; 1) mass mortality, defined by rapid, large-scale die-offs that clearly exceeded basal mortality levels, 2) detrimental ecological impact, ecological degradation that did not result in mass mortality, though it may have included instances of mortality, or 3) no detrimental impact, events that did not result in ecological deterioration.
Lowest_Reported_Taxonomy: reports the lowest taxonomic level possible of the organisms mentioned in the source of the event.
Event_Description: a concise event/impact summary.
Source_Type: indicates the type of source from which the data were obtained, categorizing them by origin (e.g., peer-reviewed articles, governmental reports, social media, newspapers, non-profit reports, or preprints/university reports).
Source_Language: indicates the language of the source.
Source_Link: Link to the original website the event description was obtained from.
Additional_Source_Type_1: indicates the type of source from which additional information was obtained, categorizing them by origin (e.g., peer-reviewed articles, governmental reports, social media, newspapers, non-profit reports, or preprints/university reports). Additional_Source_Type_2 and Additional_Source_Type_3 follow the same definition.
Additional_Source_Language_1: indicates the language of the additional information source. Additional_Source_Language_2 and Additional_Source_Language_3 follow the same definition.
Additonal_Source_Link_1: website from which the additional information from the event description was obtained. Additonal_Source_Link_2 and Additonal_Source_Link_3 follow the same definition.
Ecosystem_Species__Type: This category describes the primary target system of impacts. It includes coral reefs (impacts on tropical shallow-water coral reef ecosystems), fisheries (affecting fish stocks, fishing practices, or target species), other organisms (non-charismatic or non-target species), charismatic animals (iconic or public-interest species like turtles, whales etc), direct human impact (effects on human health), vegetated habitats (impacts on seagrasses, mangroves, salt marshes, and other rooted coastal or marine plant systems), and unclassified. (Note: this is referred to as "Impact Classification" in the manuscript)
Higher_Source_Type: Most reliable source type used. We ranked source types by perceived reliability—peer-reviewed literature, preprints or university reports, governmental reports, non-governmental reports, newspapers, and social media—and assigned the highest-ranked available source as the representative for that event.
Post.search.peer.review.source.link: DOI for peer-reviewed papers either fully or partially corroborating the impact event that was published after December 2024.
Post.search.peer.review.sources>citation: Full reference for peer-reviewed papers either fully or partially corroborating the impact event that was published after December 2024.
Notes: This column states how the peer-reviewed paper identified post search either fully or partially corroborates the impact event.
Post.search.peer.review.sources2: DOI for second peer-reviewed papers either fully or partially corroborating the impact event that was published after December 2024.
Post.search.peer.review.sources>citation2: Full reference for second peer-reviewed papers either fully or partially corroborating the impact event that was published after December 2024.
Notes2: This column states how the second peer-reviewed paper identified post search either fully or partially corroborates the impact event.
Physical_impact: Whether impact events were classified as being linked to periods of anomalously high SSTs (High SSTs), Extreme weather event (storm), or Declining SIE (Sea-ice extent) .
Heating_days_within_57day_window: The number of heating days, defined as days when SST exceeded the seasonally varying 90th percentile calculated for the 1982–2022 climatology detected within the 57 day event window (see Methods).
Maximal_intensity_during_57 days_C: The maximum SST anomaly relative to this 90th percentile detected within the 57 day event window (see Methods).
3x3_method: Reports the coordinates (format: XX.XX N/S, XX.XX E/W) of the cell used when 3 × 3 grid relaxation was used (see Methods).
Database_2.csv : Ocean Impact IPCC
This dataset categorizes impacts based on their alignment with one or more impact/risk categories adapted from the IPCC framework for the Ocean, Food Security, and Food Production Systems sectors (IPCC 2022). The data are coded as binary values, where 1 = Yes (the IPCC category applies to the event) and 0 = No (the IPCC category does not apply to the event). For full descriptions of the IPCC categories used, refer to Extended Data Table 3.
Dataset structure:
Event_ID: refers to the published source (identified by number) from which the data were extracted. In instances where multiple events were reported in the source (column Q, T, W, etc), an identifying letter was assigned to each individual event (a-m).
Row 1, columns B–O: IPCC Physical climate change drivers, according to the IPCC framework for the Ocean, Food Security, and Food Production Systems sectors (IPCC 2022). Note climate drivers are repeated in multiple columns due to more than one Nature of risk (row 2) being assigned to each climate driver.
Row 2, columns B–O: IPCC Nature of risk (adapted), according to the IPCC framework for the Ocean, Food Security, and Food Production Systems sectors (IPCC 2022).
Database_3.csv: Avian Influenza Database
This database follows the same structure and classification system as Database_1 but is specifically focused on recorded impacts of avian influenza outbreaks.
Database_4.csv: Search Keywords
This database contains all the keywords used during the literature search to identify potential impacts. The literature search included both peer-reviewed studies and grey literature, including governmental and non-governmental reports, news media, and social media posts; thus the search is not reproducible. Searches were carried out using Google Scholar, Web of Science, and the Google search engine.
Dataset Structure:
Category: indicates the broader category each keyword belongs to; 1) Stressors (Environmental pressures and drivers of change). 2) Responses (Consequences, ecosystem impacts, and human/social effects)
Keywords: the specific terms used in the literature search.
Database_5.csv: Search Languages
This database lists the 17 different languages in which the literature search was conducted.
RCode.zip: Figure codes
RCode.zip: Figure code
This archive contains the R scripts used to recreate manuscript figures. After unzipping, the folder is organised by figure number (e.g., Figure_2, Figure_3, Figure_4, Supplementary_figures). Within each folder, separate R scripts are provided for each figure panel where applicable.
Script names correspond directly to the figure panels they generate (e.g., Figure_2A.R generates Figure 2A). After downloading and unzipping, the folder structure is as follows:
- Figure_2/: contains scripts for Figure 2 panels A and B.
- Figure_3/: contains scripts for Figure 3 panels A and B.
- Figure_4/: contains one script used to generate Figure 4.
- Supplementary figures/: contain scripts used to generate Supplementary Figure S3 and Supplementary Figures S5–S6.
Each script name corresponds to the figure or figure panel it generates. The scripts require the source data provided in Figure_Source.zip. To rerun the analyses, users should download and unzip both RCode.zip and Figure_Source.zip and set the working directory to a parent folder containing the source files.
RCode.zip: Figure descriptions
Figure 2. Global distribution of ocean ecosystem impacts and their seasonal timing, 2023–2024. (A) Distribution of ocean ecosystem impacts during the record-breaking period when mean ocean surface temperatures exceeded 1.0 °C above pre-industrial levels and air surface temperatures met or exceeded 1.5 °C, from June 2023 to June 2024. (B) Cumulative event count by seasonal day of year, standardized across hemispheres (day 1 = winter solstice in each hemisphere). The y-axis shows the running total of events, where the cumulative event count starts at 1 and increases by exactly 1 per event, reaching 201 at the final point (n = 201).
Figure 3. Latitudinal and temporal distribution of reported ocean ecosystem impact events. (A) Relative frequency of events by month within each region—boreal (>15° N), equatorial (15° N–15° S), and austral (>15° S)—expressed as the proportion of all events in the dataset (n = 201) that occurred in that region and month (June 2023–June 2024). Bars are stacked by impact type; overlaid lines show daily-mean SST anomalies from NOAA OISST (0.25°)123, referenced to a seasonally varying 1982–2022 climatology. (B) Relative frequency by latitude outside the equatorial zone (|lat| > 15°), expressed as the proportion of all events occurring in each 10° absolute-latitude bin. A negative relationship (±95% CI) between absolute latitude and impact frequency is evident, where frequencies are higher near the equator and decline toward higher latitudes (α = 0.63; β = 0.05; Pα < 0.001; Pβ < 0.001; R² = 0.98). The dashed line marks |lat| = 15°.
Figure 4. Distribution of ocean ecosystem impact events in 2023–2024 by IPCC risk categories. Rose graph showing the distribution and frequency of impact events across risk categories, categorized according to the IPCC impact classification framework (Table 3.5 in ref 116), adapted to specifically address risks to marine ecosystems (Table S1). Bar length represents the number of impact events recorded, with numbers at the end of each bar indicating n per category. Bold numbers around the outer radius (1–14) correspond to risk categories listed in the table below. A color gradient is applied, with darker, warmer colors representing higher impact counts.
Figure S3. Latitudinal distribution of the relative frequency of impacts across all categories, with impact frequency represented as bold lines. The equatorial region (15°S to 15°N) is represented by dashed lines, with the solid line at 0° indicating the equator. Latitude values left of the dashed lines represent impacts that occurred in the southern hemisphere, and those to the right of the dashed line represent the northern hemisphere.
Figure S5. Distribution of reports (n = 201) according to region, impact outcome and source type. Colors represent different source types. Reporting sources differed among regions (Fisher’s exact test; p = 0.0088). In contrast, we detected no significant association between impact outcome and source type (Fisher’s exact test; p = 0.1981).
Figure S6. Distribution of reports (n = 201) according to impact targets (or systems) and impact classification. Colors represent different source types. Reporting sources differed among impact targets (Fisher’s exact test; p = 0.0004) and impact classification (Fisher’s exact test; p = 0.0004).
The R version used for each analysis is specified within the corresponding R script. Required package installation commands, library loading commands, and package dependencies are also included directly within the scripts. Additional methodological details and package information are described in the corresponding manuscript/study methods section.
Source.zip: Figure source data
This zip file contains all the necessary source data for recreating all figures. The folder is organized per Figure number, and within it you can find the underlying data for each panel.
Literature search and inclusion criteria
Twenty extractors conducted searches across 17 languages to compile records of biological and ecological impacts associated with extreme ocean conditions between June 2023 and June 2024. To account for publication delays common in peer-reviewed literature, we included both primary research articles and grey literature sources, including governmental and non-governmental reports, news media, and social media posts. Searches were conducted between June 2023 and May 2025 using Google Scholar, Web of Science, Google and China National Knowledge Infrastructure (CKNI) to maximize geographic and linguistic coverage. Because of the large number of search terms used across multiple languages and the lack of repeatable search results for some of these search engines and platforms95, it was not possible to generate a single, reproducible search string. To ensure traceability of the search outcomes, we archived all retrieved reports and search terms used during the searches.
Sources were selected based on the following criteria: (1) impacts linked to periods of anomalously high SSTs or other anomalous ocean conditions during 2023–2024 (e.g., severe storms, cyclones); (2) impacts affecting marine species and/or ecosystems; and (3) impacts occurring between 1 June 2023 and 30 June 2024, corresponding to a period when global average SSTs exceeded 1.0 °C above pre-industrial levels. This benchmark reflects the established relationship between atmospheric and ocean surface warming, with 1.5 °C surface air warming corresponding to approximately 1.0 °C in ocean surface warming (Fig. S1).
A minimum of three extractors independently inspected each event and its supporting evidence to verify inclusion in the database and minimize extractor bias. Extractors first assessed the plausibility of each report by reviewing key elements, including geographic location and environmental context (i.e., confirmation that the reported ecosystem type and/or species were consistent with that location), to ensure events were accurately attributed and to minimize the risk of misclassification or spurious records. In addition, the nature of each ecological response was required to have a historical evidence base strongly supporting its association with an anomalous event (e.g., a well-established link between MHWs and coral bleaching/mortality). Event types without such supporting evidence were excluded from the database (e.g., n = 9 cases of H5N1). All grey literature reports included in the dataset were supported by at least two independent sources. Social media reports that could not be corroborated by at least one additional, independent non-social media source were excluded due to the common occurrence of misinformation in unverified social media content96. The resulting database comprised 187 impact events.
We further cross-referenced the 187 events in our database with the independently compiled database of Smith and colleagues12,97 that documented 100 impact events during extreme marine heatwaves of 2023 and 2024. Among these, 33 were captured in our database, 51 lacked documented biological or ecological impacts and therefore did not meet our inclusion criteria, and two occurred outside our study window (1 June 2023 to 30 June 2024). The remaining 14 events were incorporated into our dataset (Fig. S7), yielding a final database of 201 impact events.
Data extraction and classification
Most impact events in the final database were assigned to one of six primary impact categories: (a) low-oxygen extreme events, (b) disease outbreaks, (c) habitat loss, (d) harmful algal blooms, (e) mass coral bleaching events, and (f) population outbreaks (Table S2). Those reported as linked to unusually warm SSTs that did not fit these predefined categories were assigned to “other extreme heat-related impacts” to capture multi-species and/or context-specific impacts. Events involving indirect consequences of heatwaves (such as prey depletion leading to starvation of macrofauna) and those not associated with high SSTs but driven primarily by other anomalous conditions (e.g., severe storms or cyclones) were classified as “other” (Table S2). Events were further classified by severity: 1) Mass mortality—rapid, large-scale die-offs exceeding basal mortality levels26; 2) Detrimental ecological impact—ecological impacts without mass mortality, though some mortality may have occurred; and 3) No observed detrimental impact—events with no observable ecological deterioration.
For each event, we extracted the geographic location, the affected community or species (identified to the lowest possible taxonomic level), and the occurrence date. When sources did not provide a specific date but reported the month, we used the 15th of the reported month (i.e., ± two weeks). If specific locations were unavailable, we assigned coordinates to the center of the reported geographical area. Events were recorded as separate entries if they were ecologically distinct (i.e., involved different ecosystems) or occurred at different locations or occurred at least one month apart. To avoid pseudo-replication, when multiple reports of the same impact type occurred within a 50 km radius and one-month timeframe, they were consolidated into a single event. This approach follows common practice across marine disciplines and allows spatial overlap to be represented without inflating event counts, while still capturing differences in marine heatwave exposure and other impact patterns across space and time. For example, even for large coral reef systems (e.g., GBR), events separated by ≥50 km were treated as spatially independent, reflecting both standard reporting conventions in that field and the heterogeneous nature of impacts across large reef systems. The only exceptions were phenomena explicitly reported as single, spatially contiguous events, which standard reporting conventions treat as singular occurrences, for instance, the Ulva prolifera bloom in the Yellow Sea.
The occurrence date of each event was assigned a "seasonal day" to account for hemispheric differences in seasonal timing. The winter solstice was designated as seasonal day 1, corresponding to December 22nd in the Northern Hemisphere and June 21st in the Southern Hemisphere. This date marks the start of astronomical winter and represents the shortest day of the year in each respective hemisphere. From this reference point, the sequence progressed through spring, summer, and autumn, with the final day (December 21/June 20) assigned a value of 366 to accommodate the leap year in 2024.
Impacts were categorised based on their alignment with one or more impact/risk categories, adapted from the IPCC framework for the Ocean, Food Security, and Food Production Systems sectors (Ch. 3; Table 3.5 in ref 100). The IPCC identifies six key climate change drivers affecting marine ecosystems, each associated to specific risks/impacts. Events were classified according to 14 selected IPCC risks/ impacts, which were adapted to reflect the entries in our database. Events were linked to multiple impact/risk categories, where appropriate (Fig. 4). Full descriptions of the 14 adapted IPCC risks/impacts are provided in Table S1.
Detecting periods of anomalously high SSTs associated with ecological impact events
To test whether ecological impact events were spatially and temporally associated with periods of anomalously high SSTs, we analyzed a 57-day event window surrounding the timing of each impact (42 days preceding and 14 days following the reported date). This interval captures both the timeframe over which sub-lethal and lethal responses of marine ecosystems typically unfold during or shortly after MHW exposure56,101-105, and the uncertainty in reported event timing (± 14 days). For each window, we quantified (i) the number of heating days, defined as days when SST exceeded the seasonally varying 90th percentile calculated for the 1982–2022 climatology, and (ii) the maximum SST anomaly relative to this 90th percentile. All metrics were calculated using daily mean SSTs from NOAA’s OISST reanalysis product (0.25° resolution). Six impact events returned zero heating days for the grid cell at the reported location. In two cases, reports identified the impacts as heat-stress related, and in three cases, the events occurred on coastlines or in geomorphologically complex coastal settings where grid cell positioning can obscure open-water signals. For these events, we applied a 3 × 3 grid relaxation, calculated SST metrics for each surrounding cell, and retained the cell with the longest detected marine heatwave duration. For events at latitudes ≥70° (n = 4), where OISST reliability decreases due to sea-ice treatment, SST metrics are not reported.
Data analyses
To assess the relationship between latitude and the proportion of total impacts, we applied a Non-linear Least Squares (NLS) exponential decay model using the nls package in R. We evaluated model assumptions, including independence, homoscedasticity, and normality of residuals, through diagnostic plots of residuals versus fitted values, quantile-quantile (Q-Q) plots, and residual histograms. We selected the NLS model because it avoided unrealistic negative predictions at high latitudes, handled continuous response variables, and yielded a near-perfect fit (R² = 0.98). Unlike likelihood-based pseudo-R² in beta regression models, the NLS model provides a variance-based R², improving interpretability. We also assessed model adequacy by comparing the NLS model to generalized linear models with binomial and beta distributions, as well as linear and log-linear models, using both AIC and R² metrics. To ensure parameter robustness, we conducted bootstrap resampling with 1000 iterations, deriving confidence intervals for α and β from the bootstrap distribution. Impacts were grouped into 10-degree latitude bins for this analysis. We tested for differences in heating days and maximum SST anomalies between impacts classified as detrimental ecological impacts and those resulting in mass mortality using two-sided Welch t-tests in R. These comparisons were performed across the full database (n = 197) and within the mass coral bleaching event category (n = 109), as these groups had sufficient sample sizes and, in the latter case, represented a single ecosystem linked to a common physical driver. All data processing, analysis, and figures (except Fig. 1) were performed in R version 4.4.0. Figure 2 was generated using the tmap, sf, rnaturalearth, mapproj, and mapdata packages.
To assess potential bias in how different information sources report on ecological impacts, we evaluated associations between source type and multiple categorical variables, including region, impact outcome, impact targets (or systems), and impact classification (Figs. S5 and S6). For each event, we ranked source types by perceived reliability—peer-reviewed literature, preprints or university reports, governmental reports, non-governmental reports, newspapers, and social media—and assigned the highest-ranked available source as the representative for that event. For each variable pair (e.g., source type × region, source type × impact outcome), we constructed contingency tables and tested for overall dependence using Fisher’s exact tests with Monte Carlo simulation (9,999 permutations). When significant associations were detected, we performed post hoc analyses to identify specific differences between groups. For this, we conducted pairwise Fisher’s exact tests comparing individual source types within each factor level. All pairwise p-values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) procedure. These analyses were conducted in R version 4.4.0.