SOMHW4D: a circumpolar dataset of four-dimensional marine heatwave events in the Southern Ocean

Amid ongoing climate change, marine heatwaves (MHWs) have received increasing attention due to their impacts on marine ecosystems and human societies. In the Southern Ocean, MHWs may exert profound effects on regional ecosystems by altering sea-ice conditions, primary productivity, and habitat conditions for species. Although publicly available tools have provided standardized approaches for MHW identification, existing studies have mainly focused on the sea surface or individual depth layers, making it difficult to characterize the three-dimensional structure of MHWs within the water column and their continuous spatiotemporal evolution. This limitation constrains a comprehensive assessment of the ecological impacts of MHWs. Here, we developed the Southern Ocean Marine Heatwave 4D dataset (SOMHW4D), a circumpolar four-dimensional MHW dataset for the Southern Ocean from 1993 to 2025 based on the GLORYS12V1 reanalysis. SOMHW4D identifies MHWs within a longitude–latitude–depth–time framework. By constructing daily three-dimensional connected objects and tracking them between consecutive time steps, the dataset represents MHWs as continuously evolving four-dimensional thermal anomaly structures. SOMHW4D provides three levels of data products: grid-level, three-dimensional object-level, and event-level data, including MHW occurrence and intensity information, daily three-dimensional thermal anomaly structures, and four-dimensional tracking results. Grid-level and three-dimensional object-level data are provided as NetCDF files, together with event-level summary tables. SOMHW4D provides a systematic data basis for characterizing the spatial distribution, vertical structure, and lifecycle evolution of MHWs in the Southern Ocean, and can support studies of MHW ecological impacts.

 

grid level:

SOMHW4D_grid_daily3d_snapshot_YYYY.nc (1993-2025)

This file stores daily three-dimensional marine heatwave occurrence masks for a given year, indicating whether MHW conditions occur at each longitude, latitude, depth, and date. Ice-shelf masking has been applied. The dimension order of mhw_mask is lat × lon × depth × time.

lon: Longitude coordinate in degrees east.

lat: Latitude coordinate in degrees north.

depth: Depth coordinate in meters, positive downward.

time: Time coordinate, expressed as days since January 1 of the corresponding year.

date_yyyymmdd: Calendar date in YYYYMMDD format for direct date identification.

mhw_mask: Daily 3D MHW occurrence mask. 0 indicates non-MHW conditions, and 1 indicates MHW conditions.

 

SOMHW4D_stats_1993_2025.nc

Annual MHW frequency at each grid cell and depth. Values range from 0 to 1, where larger values indicate more frequent MHW occurrence within the corresponding year. The dimension order is lat × lon × depth × time.

frequency: Annual MHW frequency. 

 

SOMHW4D_frequency_1993_2025.nc

SOMHW4D_stats_1993_2025.nc

This file stores grid-level statistics of annual marine heatwave frequency over 1993-2025, including the long-term trend, interannual variability, and statistical significance at each grid cell and depth. The dimension order of trend, std, and pvalue is lat × lon × depth.

trend: Linear trend in annual MHW frequency over 1993-2025, expressed in yr^-1.

std: Interannual standard deviation of annual MHW frequency over 1993-2025.

pvalue: P-value associated with the linear trend.

 

 

object level:

object label:

SOMHW4D_object_label_YYYY.nc (1993-2025)

This file stores daily three-dimensional MHW object ID label fields for a given year. The dimension order of object_id is lat × lon × depth × time.

date_yyyymmdd: Calendar date in YYYYMMDD format for direct date identification.

object_id: Daily 3D MHW object ID label field. 0 indicates no valid object, and positive values indicate valid daily object IDs.

Object IDs are unique within each day. Therefore, date_yyyymmdd + object_id should be used as the unique key for identifying a daily 3D object.

 

object metrics:

object_metrics.zip

SOMHW4D_object_daily_metrics_YYYY.csv (1993-2025)

This file stores daily metrics of individual 3D MHW objects. Each row represents one 3D MHW object identified on a specific date. The object_id is unique within each day, while event_id links daily objects belonging to the same tracked 4D MHW event.

date: Date of the daily 3D object.

year: Year of the daily object.

object_id: Daily 3D object ID; unique within each day.

event_id: 4D MHW event ID linking daily objects across time.

voxel_count: Number of 3D voxels included in the object.

layer_count: Number of depth layers occupied by the object.

daily_volume: Daily 3D volume of the object, in m3.

daily_min_depth: Shallowest depth occupied by the object, in m.

daily_max_depth: Deepest depth occupied by the object, in m.

daily_max_vertical_extent: Maximum vertical extent of the object, in m.

daily_max_thickness: Maximum vertical thickness within a single horizontal grid cell, in m.

daily_horizontal_footprint_area: Daily horizontal footprint area of the object, in km2.

daily_type: Vertical type of the object, such as surface-subsurface or subsurface.

footprint_cell_count: Number of horizontal grid cells within the object footprint.

ice_covered_cell_count: Number of ice-covered grid cells within the object footprint.

ice_fraction: Fraction of the object footprint covered by sea ice.

lon_min: Minimum longitude of the object footprint.

lon_max: Maximum longitude of the object footprint.

lat_min: Minimum latitude of the object footprint.

lat_max: Maximum latitude of the object footprint.

The combination of date and object_id uniquely identifies a daily 3D object.

event level:

event_level.zip:

SOMHW4D_event_summary_xxxx_xxxxk.csv (0000-2248)

This file stores summary metrics of tracked 4D MHW events. Each row represents one tracked event formed by linking daily 3D MHW objects through time.

event_id: Unique 4D MHW event ID.

start_date: Start date of the event.

end_date: End date of the event.

duration: Duration of the event, in days.

birth_reason: Diagnostic label describing how the event was initiated.

termination_reason: Diagnostic label describing how the event ended.

is_left_censored: Indicator of whether the event already existed at the beginning of the study period.

is_right_censored: Indicator of whether the event continued beyond the end of the study period.

max_volume: Maximum daily volume of the event during its lifetime, in m3.

mean_volume: Mean daily volume of the event during its lifetime, in m3.

max_object_count: Maximum number of daily 3D objects belonging to the event on a single day.

multi_object_days: Number of days during which the event consisted of multiple daily 3D objects.

 

Daily volume and object-count metrics are calculated from the object-level metrics associated with each event_id.

These files have not been filtered to retain only events with duration >= 5 days. Users may apply their own filtering based on the duration field.

This study was funded by the National Key Research and Development Program of China (grant number 2023YFE0104500), Shanghai Top-tier Talent Program of Eastern Talent Plan (grant number BJKJ2024059), and the Innovative Talent International Cooperation Training Project of Chinese Scholarship Council (grant number CXXM20230022).