A Northwestern Pacific Three-Dimensional Mesoscale Eddy Full-scope Dataset (NWP-3D-Eddy-Full-scope Dataset)

Distinct from existing eddy datasets, this dataset provides comprehensive 3D mesoscale eddy information for the Northwestern Pacific (99°E－170°E、20°S－66°N) spanning 1993–2023. Key variables include 3D position, type (subsurface-intensified or not), vertical thickness, vertical tilt displacement, 3D centroid trajectory, and horizontal speed of 3D centroid.The dataset is provided in both MAT and JSON formats, with a total volume of 174 GB.

It comprises two subsets: a 3D mesoscale eddy identification subset (daily files: eddy_daily_YYYYMMDD.json/eddy_daily_YYYYMMDD.mat) and a 3D mesoscale eddy tracking subset (eddy_3d_ID_tracks.json/eddy_3d_ID_tracks.mat). These files are compatible with standard data analysis tools, including MATLAB, Python, and other programming languages commonly used in oceanographic and atmospheric research.The 3D eddy identification subset, generated daily using Eulerian statistics, documents all detected 3D mesoscale eddy structures.

Each record includes the eddy identification (ID) number, date, polarity, type, vertical thickness, center and boundary locations, radius, intensity, 3D centroid coordinates, and vertical tilt displacement. The 3D eddy tracking subset, constructed using Lagrangian statistics and organized by individual eddy lifespan (exceeding 30 days), contains complete lifecycle information for each tracked eddy. This includes the ID number, date, polarity, type, vertical thickness, center and boundary locations, radius, intensity, 3D centroid trajectory, vertical tilt displacement, and horizontal propagation speed.

File specific fields are as follows:3D eddy identification dataset (eddy_3d_daily_YYYYMMDD.json/mat)(1)  ID: Unique eddy identifier, assigned daily in sequential order (i.e., the nth eddy identified on that day).(2)  date: Date of eddy detection.(3)  polarity: Direction of eddy rotation. 1 for cyclonic eddy, -1 for anticyclonic eddy.(4)  type: Type of the eddy. 1 denotes a subsurface-intensified eddy, while NaN indicate non-subsurface-intensified eddy.(5)  depth: Influence depth.

All depth layers in which the 3D eddy structure is located.(6)  thickness: Vertical extent of eddy’s influence. Difference between the deepest depth and the uppermost depth.(7)  center_lat: Latitude of the eddy center in each layer.(8)  center_lon: Longitude of the eddy center in each layer.(9)  shapes: Eddy boundary latitude and longitude of the eddy in each layer, stored in dot pairs, with the first row being the longitude and the second row being the latitude.(10) size: Radius of the eddy in each layer (unit: km).(11) intensity: Intensity of the eddy in each layer (unit: m/s).(12) tilt: Vertical tilt parameters of the 3D eddy, including the maximum tilt distance (unit: km), the depth of the maximum tilt distance (unit: m), and the tilt angle (unit: degree).(13) centroid: 3D center parameters of mass location of the 3D eddy​. including longitude, latitude, and depth.3D eddy tracking dataset (eddy_3d_ID_tracks.json/mat)(1)  ID: Eddy identification number, the same ID for the same eddy track.(2)  date: Eddy date sequence value, representing the entire number of days since a fixed pre-determined date (0 January 0000) and its fractional value, in pre-Islamic calendar form.(3)  polarity: Direction of eddy rotation. 1 for cyclonic eddy, -1 for anticyclonic eddy.(4)  type: Type of the eddy. 1 denotes a subsurface-intensified eddy, while NaN indicate non-subsurface-intensified eddy.(5)  depth: Influence depth.

All depth layers in which the 3D structure is located during the eddy life cycle.(6)  thickness: Vertical extent of eddy’s influence. The depth of the deepest structure minus the uppermost depth during its life cycle.(7)  centroid_lat: ​​Latitude for the spatial centroid of an 3D eddy throughout its lifetime.​(8)  centroid_lon: Longitude for the spatial centroid of an 3D eddy during its life cycle.(9)  centroid_dep: Depth for the spatial centroid of an 3D eddy during its life cycle.(10) size: Eddy radius for each layer during its life cycle (unit: km).(11) shapes: Eddy boundary longitude and latitude positions for each layer during its life cycle, stored in point pairs, with longitude in the first row and latitude in the second row.(12) center: Longitude and latitude of the eddy center for each layer during its life cycle, stored in the form of dot pairs, with the first column being the longitude and the second column being the latitude.(13) intensity: Eddy intensity for each layer during its life cycle (unit: m/s).(14) tilt: Vertical tilt of the eddy structure during its life cycle, the three values are, in order, the maximum tilt distance of the eddy (unit: km), the depth at which the maximum tilt distance of the eddy is located (unit: m), and the tilt angle (unit: degree).(15) speed: Instantaneous horizontal speed of the eddy during its life cycle (unit: m/s).To facilitate data access and analysis, the 3D mesoscale eddy identification dataset is packaged by 31 years to reduce the size of individual compressed files.

The tracking dataset is similarly organized into four categories based on the coefficient of variation (CV) of eddy thickness and the relative fluctuation (RF) of centroid depth, reducing the overall file count. CV is defined as the ratio of the standard deviation of thickness () to the mean thickness (), and RF is defined as the ratio of the max centroid depth range to the mean thickness. The classification thresholds are as follows: Type I, stable in both thickness and centroid position (CV < 0.5, RF < 1.0), further subdivided into more stable (CV < 0.5, RF < 0.6) and less stable (CV < 0.5, 0.6 ≤ RF < 1.0); Type II, compact thickness with significant centroid drift (CV < 0.5, RF ≥ 1.0); Type III, minimal centroid drift but substantial thickness deformation (CV ≥ 0.5, RF < 1.0); and Type IV, unstable in both thickness and centroid drift (CV ≥ 0.5, RF ≥ 1.0).The Northwestern Pacific Three-Dimensional Mesoscale Eddy Full-Scope Dataset enables novel insights into the complete vertical extent and temporal evolution of oceanic mesoscale eddies.

It supports investigations of vertical biogeochemical processes and facilitates the validation and refinement of high-resolution ocean and climate models. Furthermore, it serves as a benchmark dataset for machine learning applications in 3D eddy detection and dynamic prediction.