top of page

DA-1

Data & Preprocessing

This section describes the meteorological datasets used to construct the quadrant-based adaptive observation error framework and the preprocessing procedures applied in the simulations. We first introduce the primary datasets: the ERA5 reanalysis data used as the background field and the microwave satellite sea surface temperature (MW SST) observations used as the observation field. We then present the mathematical formulations, including the Haversine distance and spherical bearing calculations, employed to transform the conventional latitude–longitude grid into a typhoon-relative local coordinate system. Finally, we describe the preprocessing framework developed to statistically isolate and estimate the quadrant-dependent representation error (F), separating it from the instrument-related observation error following the theoretical framework of Janjić et al. (2018).

1. SST(Sea Surface Temperature) Data

Temporal

& Spatial

Filtering

Spatial

Regridding

Unit

Standardization

​

​

​

  • Pre-processing

    • Flow-chart​​

​

​

Pink Poppy Flowers

To ensure analytical consistency across SST datasets

The spatiotemporal resolutions of the three SST datasets were interpolated to match those of OISST,

and all data were converted to a consistent unit of Kelvin (K).

2. TC(Tropical Cyclone) Track

​

  • Pre-processing

    • Flow-chart​​​​​​

JASO

Filtering

*JASO: July, Agust, September, October

Spatial

Filtering

Temporal

Interpolation

12z

Extraction

Step 1. Temporal & Spatial Filtering

To ensure the seasonal and spatial consistency of the analysis.

Step 2. Temporal Interpolation & 12z Data Sampling

To accurately match the typhoon track information with daily gridded Sea Surface Temperature (SST) datasets (reanalysis and satellite observations), all track data were temporally aligned to 12:00 UTC (12z).

Pink Poppy Flowers
Pink Poppy Flowers

From the typhoon best-track dataset, summer-season data over the Northwest Pacific—

where tropical cyclones frequently occur—were extracted and resampled to a daily temporal resolution.

3. Validation dataset

  • Analysis Resolution

    • time analysis range: 2024 daily

    • spatial analysis region: 110°E - 180°E, 0°N - 45°N

    • TC influence radius: ~500km from the TC center

  • data

    • The best tracks of TCs in the western North Pacific - Japan Meteorological Agency (JMA)​​

    • SST dataset (MWSST, OISST, ERA5 SST)

      • same preprocessed data compared with test dataset ​​​​​

The validation dataset consists of data from 2024

and was preprocessed in the same method as the test dataset (2014–2023).

4. TC-Relative Mapping

1. Calculation of TC Motion Vectors (u, v) and Heading Angle​​​​​​​

  • Motion Vector Calculation: Based on the latitude and longitude coordinates from the Best Track, the zonal (u) and meridional (v) velocity components (m/s) were calculated using the central difference method. The endpoints were adjusted using forward and backward differences to maintain temporal continuity

  • Heading Angle Definition: The storm-motion direction was derived from the zonal (u) and meridional (v) velocity components. It represents the translation direction of the typhoon, expressed as the clockwise angle (0–360°) from true north (0°).

​

2. Typhoon-Centric Relative Coordinate Transformation

To standardize the physical structure of the typhoon, the fixed lat-lon grid was transformed into a relative coordinate system centered on the storm.

  • Spherical Distance and Azimuth: The Haversine formula was used to calculate the great-circle distance (km) between the TC center and each SST grid point. Additionally, the azimuth from the center to each grid point was determined using spherical trigonometry.​

       *The azimuth is defined as the clockwise angle from north between the storm center and each grid point.

 

​

​

​

​

​

​

​

​

​

  • Relative Angle Rotation: The Relative Angle was derived by subtracting the TC's heading from the grid point's azimuth. This 'Heading-up' transformation ensures that the progress direction is always oriented forward (0°), regardless of the storm's actual track.

         * Relative angle = grid azimuth − storm heading angle

​

​

​

​

​

​

​

​

3. Quadrant Assignment

Following the experimental design of He (2024), the grid area was partitioned into four quadrants based on the calculated relative angles:

​

​

​

​

​

​

​

​​

Pink Poppy Flowers
Pink Poppy Flowers
Pink Poppy Flowers
Pink Poppy Flowers
Pink Poppy Flowers
Pink Poppy Flowers
Go

The storm motion vector and relative angle were computed, and the data were remapped onto a grid centered on the storm. Quadrants were then assigned based on the direction of storm propagation.

bottom of page